21002 lines
648 KiB
JSON
21002 lines
648 KiB
JSON
|
[
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a",
|
||
|
|
"gt_answers": "A",
|
||
|
|
"image_path": "./data/IAM/a05/a05-125/a05-125-00-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "political",
|
||
|
|
"gt_answers": "POLITICAL",
|
||
|
|
"image_path": "./data/IAM/a05/a05-125/a05-125-00-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "c o w e n c h",
|
||
|
|
"gt_answers": "CORRESPONDENT",
|
||
|
|
"image_path": "./data/IAM/a05/a05-125/a05-125-00-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word writer",
|
||
|
|
"gt_answers": "writes",
|
||
|
|
"image_path": "./data/IAM/a05/a05-125/a05-125-00-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "be the change",
|
||
|
|
"gt_answers": "Banthe",
|
||
|
|
"image_path": "./data/IAM/a05/a05-125/a05-125-00-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "bomb",
|
||
|
|
"image_path": "./data/IAM/a05/a05-125/a05-125-01-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's name",
|
||
|
|
"gt_answers": "demonstrators",
|
||
|
|
"image_path": "./data/IAM/a05/a05-125/a05-125-01-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "throne of glass",
|
||
|
|
"gt_answers": "thronged",
|
||
|
|
"image_path": "./data/IAM/a05/a05-125/a05-125-01-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "i love you",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/a05/a05-125/a05-125-01-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the name george wilson",
|
||
|
|
"gt_answers": "Central",
|
||
|
|
"image_path": "./data/IAM/a05/a05-125/a05-125-01-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "lolly",
|
||
|
|
"gt_answers": "Lobby",
|
||
|
|
"image_path": "./data/IAM/a05/a05-125/a05-125-02-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the letter o",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/a05/a05-125/a05-125-02-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a signature",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/a05/a05-125/a05-125-02-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "h c",
|
||
|
|
"gt_answers": "House",
|
||
|
|
"image_path": "./data/IAM/a05/a05-125/a05-125-02-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/a05/a05-125/a05-125-02-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word person",
|
||
|
|
"gt_answers": "Commons",
|
||
|
|
"image_path": "./data/IAM/a05/a05-125/a05-125-02-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "ash",
|
||
|
|
"gt_answers": "last",
|
||
|
|
"image_path": "./data/IAM/a05/a05-125/a05-125-02-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "wwii",
|
||
|
|
"gt_answers": "night",
|
||
|
|
"image_path": "./data/IAM/a05/a05-125/a05-125-03-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "and",
|
||
|
|
"image_path": "./data/IAM/a05/a05-125/a05-125-03-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "f w",
|
||
|
|
"gt_answers": "formed",
|
||
|
|
"image_path": "./data/IAM/a05/a05-125/a05-125-03-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "a",
|
||
|
|
"image_path": "./data/IAM/a05/a05-125/a05-125-03-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "quieue",
|
||
|
|
"gt_answers": "queue",
|
||
|
|
"image_path": "./data/IAM/a05/a05-125/a05-125-03-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "stretching",
|
||
|
|
"image_path": "./data/IAM/a05/a05-125/a05-125-03-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word person",
|
||
|
|
"gt_answers": "for",
|
||
|
|
"image_path": "./data/IAM/a05/a05-125/a05-125-04-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "more",
|
||
|
|
"image_path": "./data/IAM/a05/a05-125/a05-125-04-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "than",
|
||
|
|
"image_path": "./data/IAM/a05/a05-125/a05-125-04-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the number 2000",
|
||
|
|
"gt_answers": "200",
|
||
|
|
"image_path": "./data/IAM/a05/a05-125/a05-125-04-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "yards",
|
||
|
|
"image_path": "./data/IAM/a05/a05-125/a05-125-04-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "outside",
|
||
|
|
"image_path": "./data/IAM/a05/a05-125/a05-125-04-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "wavy lines",
|
||
|
|
"gt_answers": "in",
|
||
|
|
"image_path": "./data/IAM/a05/a05-125/a05-125-05-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "ola",
|
||
|
|
"gt_answers": "Old",
|
||
|
|
"image_path": "./data/IAM/a05/a05-125/a05-125-05-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "pala co",
|
||
|
|
"gt_answers": "Palace",
|
||
|
|
"image_path": "./data/IAM/a05/a05-125/a05-125-05-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "ray",
|
||
|
|
"gt_answers": "Yard",
|
||
|
|
"image_path": "./data/IAM/a05/a05-125/a05-125-05-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the name of person",
|
||
|
|
"gt_answers": "About",
|
||
|
|
"image_path": "./data/IAM/a05/a05-125/a05-125-05-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "2000",
|
||
|
|
"image_path": "./data/IAM/a05/a05-125/a05-125-05-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word ali",
|
||
|
|
"gt_answers": "lobbied",
|
||
|
|
"image_path": "./data/IAM/a05/a05-125/a05-125-06-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "m p s",
|
||
|
|
"gt_answers": "MPs",
|
||
|
|
"image_path": "./data/IAM/a05/a05-125/a05-125-06-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "and",
|
||
|
|
"image_path": "./data/IAM/a05/a05-125/a05-125-06-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "harangued",
|
||
|
|
"image_path": "./data/IAM/a05/a05-125/a05-125-06-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "them",
|
||
|
|
"image_path": "./data/IAM/a05/a05-125/a05-125-06-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "on",
|
||
|
|
"image_path": "./data/IAM/a05/a05-125/a05-125-07-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "disarmament",
|
||
|
|
"image_path": "./data/IAM/a05/a05-125/a05-125-07-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "this",
|
||
|
|
"gt_answers": "This",
|
||
|
|
"image_path": "./data/IAM/k04/k04-115/k04-115-00-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "move",
|
||
|
|
"gt_answers": "move",
|
||
|
|
"image_path": "./data/IAM/k04/k04-115/k04-115-00-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "to day",
|
||
|
|
"gt_answers": "today",
|
||
|
|
"image_path": "./data/IAM/k04/k04-115/k04-115-00-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "you",
|
||
|
|
"gt_answers": "was",
|
||
|
|
"image_path": "./data/IAM/k04/k04-115/k04-115-00-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a snake",
|
||
|
|
"gt_answers": "a",
|
||
|
|
"image_path": "./data/IAM/k04/k04-115/k04-115-00-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "i love you",
|
||
|
|
"gt_answers": "reminder",
|
||
|
|
"image_path": "./data/IAM/k04/k04-115/k04-115-00-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the letter s",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/k04/k04-115/k04-115-00-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word that",
|
||
|
|
"gt_answers": "that",
|
||
|
|
"image_path": "./data/IAM/k04/k04-115/k04-115-00-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "fact",
|
||
|
|
"gt_answers": "fact",
|
||
|
|
"image_path": "./data/IAM/k04/k04-115/k04-115-00-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "william wilson",
|
||
|
|
"gt_answers": "Whilst",
|
||
|
|
"image_path": "./data/IAM/k04/k04-115/k04-115-01-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "she",
|
||
|
|
"gt_answers": "she",
|
||
|
|
"image_path": "./data/IAM/k04/k04-115/k04-115-01-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "automaticly",
|
||
|
|
"gt_answers": "automatically",
|
||
|
|
"image_path": "./data/IAM/k04/k04-115/k04-115-01-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "patted",
|
||
|
|
"gt_answers": "patted",
|
||
|
|
"image_path": "./data/IAM/k04/k04-115/k04-115-01-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/k04/k04-115/k04-115-01-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "hud",
|
||
|
|
"gt_answers": "hand",
|
||
|
|
"image_path": "./data/IAM/k04/k04-115/k04-115-01-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "f",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/k04/k04-115/k04-115-01-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/k04/k04-115/k04-115-02-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "authenticated",
|
||
|
|
"gt_answers": "agitated",
|
||
|
|
"image_path": "./data/IAM/k04/k04-115/k04-115-02-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "lucia",
|
||
|
|
"gt_answers": "Lucia",
|
||
|
|
"image_path": "./data/IAM/k04/k04-115/k04-115-02-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "he",
|
||
|
|
"gt_answers": "her",
|
||
|
|
"image_path": "./data/IAM/k04/k04-115/k04-115-02-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "brain",
|
||
|
|
"gt_answers": "brain",
|
||
|
|
"image_path": "./data/IAM/k04/k04-115/k04-115-02-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "was",
|
||
|
|
"image_path": "./data/IAM/k04/k04-115/k04-115-02-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "i love you",
|
||
|
|
"gt_answers": "working",
|
||
|
|
"image_path": "./data/IAM/k04/k04-115/k04-115-02-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "happx",
|
||
|
|
"gt_answers": "rapidly",
|
||
|
|
"image_path": "./data/IAM/k04/k04-115/k04-115-03-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the name of a person",
|
||
|
|
"gt_answers": "and",
|
||
|
|
"image_path": "./data/IAM/k04/k04-115/k04-115-03-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "she",
|
||
|
|
"gt_answers": "she",
|
||
|
|
"image_path": "./data/IAM/k04/k04-115/k04-115-03-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "decided",
|
||
|
|
"gt_answers": "decided",
|
||
|
|
"image_path": "./data/IAM/k04/k04-115/k04-115-03-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "that is all",
|
||
|
|
"gt_answers": "that",
|
||
|
|
"image_path": "./data/IAM/k04/k04-115/k04-115-03-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "she",
|
||
|
|
"gt_answers": "she",
|
||
|
|
"image_path": "./data/IAM/k04/k04-115/k04-115-03-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "must",
|
||
|
|
"image_path": "./data/IAM/k04/k04-115/k04-115-03-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the number 6",
|
||
|
|
"gt_answers": "be",
|
||
|
|
"image_path": "./data/IAM/k04/k04-115/k04-115-03-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the pampered",
|
||
|
|
"gt_answers": "prepared",
|
||
|
|
"image_path": "./data/IAM/k04/k04-115/k04-115-04-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "x and o",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/k04/k04-115/k04-115-04-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word counter",
|
||
|
|
"gt_answers": "counter",
|
||
|
|
"image_path": "./data/IAM/k04/k04-115/k04-115-04-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's name",
|
||
|
|
"gt_answers": "Orsinis",
|
||
|
|
"image_path": "./data/IAM/k04/k04-115/k04-115-04-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "dewars",
|
||
|
|
"gt_answers": "demands",
|
||
|
|
"image_path": "./data/IAM/k04/k04-115/k04-115-04-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "difficult",
|
||
|
|
"gt_answers": "difficult",
|
||
|
|
"image_path": "./data/IAM/k04/k04-115/k04-115-05-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "thought",
|
||
|
|
"gt_answers": "though",
|
||
|
|
"image_path": "./data/IAM/k04/k04-115/k04-115-05-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "hat",
|
||
|
|
"gt_answers": "that",
|
||
|
|
"image_path": "./data/IAM/k04/k04-115/k04-115-05-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "thomas",
|
||
|
|
"gt_answers": "would",
|
||
|
|
"image_path": "./data/IAM/k04/k04-115/k04-115-05-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "be",
|
||
|
|
"gt_answers": "be",
|
||
|
|
"image_path": "./data/IAM/k04/k04-115/k04-115-05-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "she",
|
||
|
|
"gt_answers": "She",
|
||
|
|
"image_path": "./data/IAM/k04/k04-115/k04-115-05-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "because",
|
||
|
|
"gt_answers": "became",
|
||
|
|
"image_path": "./data/IAM/k04/k04-115/k04-115-05-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word awake",
|
||
|
|
"gt_answers": "aware",
|
||
|
|
"image_path": "./data/IAM/k04/k04-115/k04-115-06-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "thoth",
|
||
|
|
"gt_answers": "that",
|
||
|
|
"image_path": "./data/IAM/k04/k04-115/k04-115-06-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word he",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/k04/k04-115/k04-115-06-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "peace",
|
||
|
|
"gt_answers": "pace",
|
||
|
|
"image_path": "./data/IAM/k04/k04-115/k04-115-06-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "you",
|
||
|
|
"gt_answers": "was",
|
||
|
|
"image_path": "./data/IAM/k04/k04-115/k04-115-06-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "slackening",
|
||
|
|
"gt_answers": "slackening",
|
||
|
|
"image_path": "./data/IAM/k04/k04-115/k04-115-06-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "how",
|
||
|
|
"gt_answers": "now",
|
||
|
|
"image_path": "./data/IAM/k04/k04-115/k04-115-07-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "he",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/k04/k04-115/k04-115-07-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "cooil",
|
||
|
|
"gt_answers": "coach",
|
||
|
|
"image_path": "./data/IAM/k04/k04-115/k04-115-07-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "stopped",
|
||
|
|
"gt_answers": "stopped",
|
||
|
|
"image_path": "./data/IAM/k04/k04-115/k04-115-07-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a",
|
||
|
|
"gt_answers": "A",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-00-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "small",
|
||
|
|
"gt_answers": "small",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-00-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word person",
|
||
|
|
"gt_answers": "man",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-00-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "in",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-00-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "bib",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-00-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "over all",
|
||
|
|
"gt_answers": "overalls",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-00-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word and",
|
||
|
|
"gt_answers": "and",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-00-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a snake",
|
||
|
|
"gt_answers": "a",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-00-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "batted",
|
||
|
|
"gt_answers": "battered",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-00-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "derby",
|
||
|
|
"gt_answers": "Derby",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-01-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "h",
|
||
|
|
"gt_answers": "hat",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-01-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "fused",
|
||
|
|
"gt_answers": "fussed",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-01-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "ground",
|
||
|
|
"gt_answers": "around",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-01-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "h",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-01-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "slove",
|
||
|
|
"gt_answers": "stove",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-01-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "2 ain 2",
|
||
|
|
"gt_answers": "2Aint",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-01-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "exacly",
|
||
|
|
"gt_answers": "exactly",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-02-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-02-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "r e",
|
||
|
|
"gt_answers": "Ritz",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-02-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "hotel",
|
||
|
|
"gt_answers": "Hotel",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-02-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "bf",
|
||
|
|
"gt_answers": "but",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-02-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word we",
|
||
|
|
"gt_answers": "we",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-02-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "e",
|
||
|
|
"gt_answers": "ve",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-02-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "gol",
|
||
|
|
"gt_answers": "got",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-02-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "our",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-02-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the name of the person",
|
||
|
|
"gt_answers": "little",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-02-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "comforts of home",
|
||
|
|
"gt_answers": "comforts",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-03-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word sad",
|
||
|
|
"gt_answers": "said",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-03-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "dowel",
|
||
|
|
"gt_answers": "Dowd",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-03-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "l",
|
||
|
|
"gt_answers": "He",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-03-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "introduced",
|
||
|
|
"gt_answers": "introduced",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-03-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "h",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-03-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "menu",
|
||
|
|
"gt_answers": "men",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-03-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "who",
|
||
|
|
"gt_answers": "who",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-04-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a",
|
||
|
|
"gt_answers": "d",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-04-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "come",
|
||
|
|
"gt_answers": "come",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-04-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the name of the person",
|
||
|
|
"gt_answers": "in",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-04-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the letter s",
|
||
|
|
"gt_answers": "as",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-04-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "roper",
|
||
|
|
"gt_answers": "Roper",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-04-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "smith",
|
||
|
|
"gt_answers": "Smith",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-04-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "shorty",
|
||
|
|
"gt_answers": "Shorty",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-04-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "mike",
|
||
|
|
"gt_answers": "Mick",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-04-10.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "go lightly",
|
||
|
|
"gt_answers": "Golightly",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-05-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "sweden",
|
||
|
|
"gt_answers": "Swede",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-05-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "andel",
|
||
|
|
"gt_answers": "and",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-05-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-05-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "good busker",
|
||
|
|
"gt_answers": "Sodbuster",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-05-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "this is person",
|
||
|
|
"gt_answers": "This",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-05-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "las",
|
||
|
|
"gt_answers": "last",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-05-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "was",
|
||
|
|
"gt_answers": "was",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-05-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-06-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "little",
|
||
|
|
"gt_answers": "little",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-06-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "hombre",
|
||
|
|
"gt_answers": "hombre",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-06-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "in",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-06-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "big",
|
||
|
|
"gt_answers": "big",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-06-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "foralls",
|
||
|
|
"gt_answers": "overalls",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-06-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "very",
|
||
|
|
"gt_answers": "Very",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-06-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "soon",
|
||
|
|
"gt_answers": "soon",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-06-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "he",
|
||
|
|
"gt_answers": "he",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-06-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "had",
|
||
|
|
"gt_answers": "had",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-07-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "h",
|
||
|
|
"gt_answers": "tin",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-07-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "mugs",
|
||
|
|
"gt_answers": "mugs",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-07-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word filled",
|
||
|
|
"gt_answers": "filled",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-07-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "with love",
|
||
|
|
"gt_answers": "with",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-07-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "hk",
|
||
|
|
"gt_answers": "hot",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-07-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "black",
|
||
|
|
"gt_answers": "black",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-07-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "coffee",
|
||
|
|
"gt_answers": "coffee",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-07-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "sweetened",
|
||
|
|
"gt_answers": "sweetened",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-08-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "th",
|
||
|
|
"gt_answers": "with",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-08-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "molases",
|
||
|
|
"gt_answers": "molasses",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-08-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "for",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-08-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the name of person",
|
||
|
|
"gt_answers": "all",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-08-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word and",
|
||
|
|
"gt_answers": "and",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-08-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the letter o",
|
||
|
|
"gt_answers": "for",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-08-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "us",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-08-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "there",
|
||
|
|
"gt_answers": "there",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-09-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word here",
|
||
|
|
"gt_answers": "were",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-09-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "hwo",
|
||
|
|
"gt_answers": "two",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-09-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "peaks",
|
||
|
|
"gt_answers": "plates",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-09-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the letter l",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-09-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "bean",
|
||
|
|
"gt_answers": "beans",
|
||
|
|
"image_path": "./data/IAM/n06/n06-163/n06-163-09-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "b f",
|
||
|
|
"gt_answers": "But",
|
||
|
|
"image_path": "./data/IAM/b02/b02-105/b02-105-00-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "they",
|
||
|
|
"gt_answers": "they",
|
||
|
|
"image_path": "./data/IAM/b02/b02-105/b02-105-00-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "we were",
|
||
|
|
"gt_answers": "were",
|
||
|
|
"image_path": "./data/IAM/b02/b02-105/b02-105-00-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the number 50",
|
||
|
|
"gt_answers": "so",
|
||
|
|
"image_path": "./data/IAM/b02/b02-105/b02-105-00-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "convict",
|
||
|
|
"gt_answers": "convinced",
|
||
|
|
"image_path": "./data/IAM/b02/b02-105/b02-105-00-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "and",
|
||
|
|
"gt_answers": "and",
|
||
|
|
"image_path": "./data/IAM/b02/b02-105/b02-105-00-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "hery",
|
||
|
|
"gt_answers": "they",
|
||
|
|
"image_path": "./data/IAM/b02/b02-105/b02-105-00-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word have",
|
||
|
|
"gt_answers": "have",
|
||
|
|
"image_path": "./data/IAM/b02/b02-105/b02-105-01-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word had",
|
||
|
|
"gt_answers": "had",
|
||
|
|
"image_path": "./data/IAM/b02/b02-105/b02-105-01-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "l",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/b02/b02-105/b02-105-01-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "false",
|
||
|
|
"gt_answers": "take",
|
||
|
|
"image_path": "./data/IAM/b02/b02-105/b02-105-01-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/b02/b02-105/b02-105-01-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "consequences",
|
||
|
|
"image_path": "./data/IAM/b02/b02-105/b02-105-01-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "mean",
|
||
|
|
"gt_answers": "Mean",
|
||
|
|
"image_path": "./data/IAM/b02/b02-105/b02-105-01-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "w h c",
|
||
|
|
"gt_answers": "while",
|
||
|
|
"image_path": "./data/IAM/b02/b02-105/b02-105-02-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a worm",
|
||
|
|
"gt_answers": "in",
|
||
|
|
"image_path": "./data/IAM/b02/b02-105/b02-105-02-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "germany",
|
||
|
|
"gt_answers": "Germany",
|
||
|
|
"image_path": "./data/IAM/b02/b02-105/b02-105-02-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "self",
|
||
|
|
"gt_answers": "itself",
|
||
|
|
"image_path": "./data/IAM/b02/b02-105/b02-105-02-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/b02/b02-105/b02-105-02-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "course",
|
||
|
|
"gt_answers": "course",
|
||
|
|
"image_path": "./data/IAM/b02/b02-105/b02-105-02-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "x",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/b02/b02-105/b02-105-02-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the name person",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/b02/b02-105/b02-105-03-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "can paly",
|
||
|
|
"gt_answers": "campaign",
|
||
|
|
"image_path": "./data/IAM/b02/b02-105/b02-105-03-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "has",
|
||
|
|
"gt_answers": "has",
|
||
|
|
"image_path": "./data/IAM/b02/b02-105/b02-105-03-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word destroyed",
|
||
|
|
"gt_answers": "dismayed",
|
||
|
|
"image_path": "./data/IAM/b02/b02-105/b02-105-03-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "a",
|
||
|
|
"image_path": "./data/IAM/b02/b02-105/b02-105-03-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "jocel",
|
||
|
|
"gt_answers": "good",
|
||
|
|
"image_path": "./data/IAM/b02/b02-105/b02-105-03-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "mary",
|
||
|
|
"gt_answers": "many",
|
||
|
|
"image_path": "./data/IAM/b02/b02-105/b02-105-04-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "people",
|
||
|
|
"gt_answers": "people",
|
||
|
|
"image_path": "./data/IAM/b02/b02-105/b02-105-04-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "fey",
|
||
|
|
"gt_answers": "they",
|
||
|
|
"image_path": "./data/IAM/b02/b02-105/b02-105-04-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "too",
|
||
|
|
"image_path": "./data/IAM/b02/b02-105/b02-105-04-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the name will",
|
||
|
|
"gt_answers": "will",
|
||
|
|
"image_path": "./data/IAM/b02/b02-105/b02-105-04-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "bc",
|
||
|
|
"gt_answers": "be",
|
||
|
|
"image_path": "./data/IAM/b02/b02-105/b02-105-04-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "glad",
|
||
|
|
"gt_answers": "glad",
|
||
|
|
"image_path": "./data/IAM/b02/b02-105/b02-105-04-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "who",
|
||
|
|
"gt_answers": "when",
|
||
|
|
"image_path": "./data/IAM/b02/b02-105/b02-105-05-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/b02/b02-105/b02-105-05-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "polly",
|
||
|
|
"gt_answers": "polling",
|
||
|
|
"image_path": "./data/IAM/b02/b02-105/b02-105-05-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the name of the person",
|
||
|
|
"gt_answers": "stations",
|
||
|
|
"image_path": "./data/IAM/b02/b02-105/b02-105-05-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "close",
|
||
|
|
"gt_answers": "close",
|
||
|
|
"image_path": "./data/IAM/b02/b02-105/b02-105-05-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "The",
|
||
|
|
"image_path": "./data/IAM/h01/h01-018/h01-018-00-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "numbers",
|
||
|
|
"gt_answers": "number",
|
||
|
|
"image_path": "./data/IAM/h01/h01-018/h01-018-00-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "f",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/h01/h01-018/h01-018-00-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word structurally",
|
||
|
|
"gt_answers": "structurally",
|
||
|
|
"image_path": "./data/IAM/h01/h01-018/h01-018-00-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "sepake",
|
||
|
|
"gt_answers": "separate",
|
||
|
|
"image_path": "./data/IAM/h01/h01-018/h01-018-00-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "schwellings",
|
||
|
|
"gt_answers": "dwellings",
|
||
|
|
"image_path": "./data/IAM/h01/h01-018/h01-018-00-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word la",
|
||
|
|
"gt_answers": "that",
|
||
|
|
"image_path": "./data/IAM/h01/h01-018/h01-018-00-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "is",
|
||
|
|
"image_path": "./data/IAM/h01/h01-018/h01-018-00-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "houses",
|
||
|
|
"gt_answers": "houses",
|
||
|
|
"image_path": "./data/IAM/h01/h01-018/h01-018-01-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "or",
|
||
|
|
"image_path": "./data/IAM/h01/h01-018/h01-018-01-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "fl h",
|
||
|
|
"gt_answers": "flats",
|
||
|
|
"image_path": "./data/IAM/h01/h01-018/h01-018-01-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the letter g",
|
||
|
|
"gt_answers": "or",
|
||
|
|
"image_path": "./data/IAM/h01/h01-018/h01-018-01-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "mother",
|
||
|
|
"gt_answers": "other",
|
||
|
|
"image_path": "./data/IAM/h01/h01-018/h01-018-01-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "quakes",
|
||
|
|
"gt_answers": "quarters",
|
||
|
|
"image_path": "./data/IAM/h01/h01-018/h01-018-01-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "build",
|
||
|
|
"gt_answers": "built",
|
||
|
|
"image_path": "./data/IAM/h01/h01-018/h01-018-01-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "or",
|
||
|
|
"image_path": "./data/IAM/h01/h01-018/h01-018-01-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "adapted",
|
||
|
|
"gt_answers": "adapted",
|
||
|
|
"image_path": "./data/IAM/h01/h01-018/h01-018-01-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word ber",
|
||
|
|
"gt_answers": "for",
|
||
|
|
"image_path": "./data/IAM/h01/h01-018/h01-018-01-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "separe",
|
||
|
|
"gt_answers": "separate",
|
||
|
|
"image_path": "./data/IAM/h01/h01-018/h01-018-02-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "occupation",
|
||
|
|
"gt_answers": "occupation",
|
||
|
|
"image_path": "./data/IAM/h01/h01-018/h01-018-02-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "hd",
|
||
|
|
"gt_answers": "and",
|
||
|
|
"image_path": "./data/IAM/h01/h01-018/h01-018-02-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "f oming",
|
||
|
|
"gt_answers": "forming",
|
||
|
|
"image_path": "./data/IAM/h01/h01-018/h01-018-02-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a",
|
||
|
|
"gt_answers": "a",
|
||
|
|
"image_path": "./data/IAM/h01/h01-018/h01-018-02-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "private",
|
||
|
|
"gt_answers": "private",
|
||
|
|
"image_path": "./data/IAM/h01/h01-018/h01-018-02-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "bd",
|
||
|
|
"gt_answers": "and",
|
||
|
|
"image_path": "./data/IAM/h01/h01-018/h01-018-02-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "shuchbally",
|
||
|
|
"gt_answers": "structurally",
|
||
|
|
"image_path": "./data/IAM/h01/h01-018/h01-018-02-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "separate",
|
||
|
|
"gt_answers": "separate",
|
||
|
|
"image_path": "./data/IAM/h01/h01-018/h01-018-03-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "h",
|
||
|
|
"gt_answers": "unit",
|
||
|
|
"image_path": "./data/IAM/h01/h01-018/h01-018-03-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "was",
|
||
|
|
"gt_answers": "was",
|
||
|
|
"image_path": "./data/IAM/h01/h01-018/h01-018-03-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "obtained",
|
||
|
|
"gt_answers": "obtained",
|
||
|
|
"image_path": "./data/IAM/h01/h01-018/h01-018-03-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "ad",
|
||
|
|
"gt_answers": "as",
|
||
|
|
"image_path": "./data/IAM/h01/h01-018/h01-018-03-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "previously",
|
||
|
|
"gt_answers": "previously",
|
||
|
|
"image_path": "./data/IAM/h01/h01-018/h01-018-03-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "together",
|
||
|
|
"gt_answers": "together",
|
||
|
|
"image_path": "./data/IAM/h01/h01-018/h01-018-03-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "william",
|
||
|
|
"gt_answers": "with",
|
||
|
|
"image_path": "./data/IAM/h01/h01-018/h01-018-03-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "he",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/h01/h01-018/h01-018-03-10.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word humber",
|
||
|
|
"gt_answers": "number",
|
||
|
|
"image_path": "./data/IAM/h01/h01-018/h01-018-04-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/h01/h01-018/h01-018-04-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "house holds",
|
||
|
|
"gt_answers": "households",
|
||
|
|
"image_path": "./data/IAM/h01/h01-018/h01-018-04-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "with",
|
||
|
|
"gt_answers": "with",
|
||
|
|
"image_path": "./data/IAM/h01/h01-018/h01-018-04-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "sol",
|
||
|
|
"gt_answers": "sole",
|
||
|
|
"image_path": "./data/IAM/h01/h01-018/h01-018-04-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "occupation",
|
||
|
|
"gt_answers": "occupation",
|
||
|
|
"image_path": "./data/IAM/h01/h01-018/h01-018-04-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "or",
|
||
|
|
"image_path": "./data/IAM/h01/h01-018/h01-018-04-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "sharing",
|
||
|
|
"gt_answers": "sharing",
|
||
|
|
"image_path": "./data/IAM/h01/h01-018/h01-018-04-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "suech",
|
||
|
|
"gt_answers": "such",
|
||
|
|
"image_path": "./data/IAM/h01/h01-018/h01-018-05-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "wellings",
|
||
|
|
"gt_answers": "dwellings",
|
||
|
|
"image_path": "./data/IAM/h01/h01-018/h01-018-05-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word and",
|
||
|
|
"gt_answers": "and",
|
||
|
|
"image_path": "./data/IAM/h01/h01-018/h01-018-05-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/h01/h01-018/h01-018-05-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "humber",
|
||
|
|
"gt_answers": "number",
|
||
|
|
"image_path": "./data/IAM/h01/h01-018/h01-018-05-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/h01/h01-018/h01-018-05-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "livis",
|
||
|
|
"gt_answers": "living",
|
||
|
|
"image_path": "./data/IAM/h01/h01-018/h01-018-05-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word rooms",
|
||
|
|
"gt_answers": "rooms",
|
||
|
|
"image_path": "./data/IAM/h01/h01-018/h01-018-05-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "occ +",
|
||
|
|
"gt_answers": "occu",
|
||
|
|
"image_path": "./data/IAM/h01/h01-018/h01-018-05-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word inspired",
|
||
|
|
"gt_answers": "pied",
|
||
|
|
"image_path": "./data/IAM/h01/h01-018/h01-018-06-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "b",
|
||
|
|
"gt_answers": "by",
|
||
|
|
"image_path": "./data/IAM/h01/h01-018/h01-018-06-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "teach",
|
||
|
|
"gt_answers": "each",
|
||
|
|
"image_path": "./data/IAM/h01/h01-018/h01-018-06-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "house hold",
|
||
|
|
"gt_answers": "household",
|
||
|
|
"image_path": "./data/IAM/h01/h01-018/h01-018-06-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "bf",
|
||
|
|
"gt_answers": "But",
|
||
|
|
"image_path": "./data/IAM/h07/h07-069a/h07-069a-00-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "whereas",
|
||
|
|
"gt_answers": "whereas",
|
||
|
|
"image_path": "./data/IAM/h07/h07-069a/h07-069a-00-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "he",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/h07/h07-069a/h07-069a-00-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "i love you",
|
||
|
|
"gt_answers": "national",
|
||
|
|
"image_path": "./data/IAM/h07/h07-069a/h07-069a-00-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "annual",
|
||
|
|
"gt_answers": "annual",
|
||
|
|
"image_path": "./data/IAM/h07/h07-069a/h07-069a-00-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "average",
|
||
|
|
"gt_answers": "average",
|
||
|
|
"image_path": "./data/IAM/h07/h07-069a/h07-069a-00-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "will",
|
||
|
|
"gt_answers": "will",
|
||
|
|
"image_path": "./data/IAM/h07/h07-069a/h07-069a-00-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "selfie",
|
||
|
|
"gt_answers": "settle",
|
||
|
|
"image_path": "./data/IAM/h07/h07-069a/h07-069a-00-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "down",
|
||
|
|
"gt_answers": "down",
|
||
|
|
"image_path": "./data/IAM/h07/h07-069a/h07-069a-01-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the letter h",
|
||
|
|
"gt_answers": "at",
|
||
|
|
"image_path": "./data/IAM/h07/h07-069a/h07-069a-01-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "som ethy",
|
||
|
|
"gt_answers": "something",
|
||
|
|
"image_path": "./data/IAM/h07/h07-069a/h07-069a-01-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "like",
|
||
|
|
"gt_answers": "like",
|
||
|
|
"image_path": "./data/IAM/h07/h07-069a/h07-069a-01-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a 4",
|
||
|
|
"gt_answers": "14",
|
||
|
|
"image_path": "./data/IAM/h07/h07-069a/h07-069a-01-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "per",
|
||
|
|
"image_path": "./data/IAM/h07/h07-069a/h07-069a-01-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "clf",
|
||
|
|
"gt_answers": "cent",
|
||
|
|
"image_path": "./data/IAM/h07/h07-069a/h07-069a-01-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "alfer",
|
||
|
|
"gt_answers": "after",
|
||
|
|
"image_path": "./data/IAM/h07/h07-069a/h07-069a-01-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word flu",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/h07/h07-069a/h07-069a-01-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "sauge",
|
||
|
|
"gt_answers": "bulge",
|
||
|
|
"image_path": "./data/IAM/h07/h07-069a/h07-069a-02-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "lgs",
|
||
|
|
"gt_answers": "has",
|
||
|
|
"image_path": "./data/IAM/h07/h07-069a/h07-069a-02-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "passed",
|
||
|
|
"gt_answers": "passed",
|
||
|
|
"image_path": "./data/IAM/h07/h07-069a/h07-069a-02-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "h",
|
||
|
|
"gt_answers": "in",
|
||
|
|
"image_path": "./data/IAM/h07/h07-069a/h07-069a-02-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "he",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/h07/h07-069a/h07-069a-02-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "l a",
|
||
|
|
"gt_answers": "new",
|
||
|
|
"image_path": "./data/IAM/h07/h07-069a/h07-069a-02-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "focus",
|
||
|
|
"gt_answers": "towns",
|
||
|
|
"image_path": "./data/IAM/h07/h07-069a/h07-069a-02-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/h07/h07-069a/h07-069a-02-10.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "peace tape",
|
||
|
|
"gt_answers": "percentage",
|
||
|
|
"image_path": "./data/IAM/h07/h07-069a/h07-069a-02-11.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "will",
|
||
|
|
"gt_answers": "will",
|
||
|
|
"image_path": "./data/IAM/h07/h07-069a/h07-069a-03-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "seville",
|
||
|
|
"gt_answers": "generally",
|
||
|
|
"image_path": "./data/IAM/h07/h07-069a/h07-069a-03-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "go",
|
||
|
|
"image_path": "./data/IAM/h07/h07-069a/h07-069a-03-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "o",
|
||
|
|
"gt_answers": "on",
|
||
|
|
"image_path": "./data/IAM/h07/h07-069a/h07-069a-03-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "lisus",
|
||
|
|
"gt_answers": "rising",
|
||
|
|
"image_path": "./data/IAM/h07/h07-069a/h07-069a-03-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "h",
|
||
|
|
"gt_answers": "in",
|
||
|
|
"image_path": "./data/IAM/h07/h07-069a/h07-069a-03-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "soul",
|
||
|
|
"gt_answers": "some",
|
||
|
|
"image_path": "./data/IAM/h07/h07-069a/h07-069a-03-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "awo",
|
||
|
|
"gt_answers": "towns",
|
||
|
|
"image_path": "./data/IAM/h07/h07-069a/h07-069a-03-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "i love you",
|
||
|
|
"gt_answers": "into",
|
||
|
|
"image_path": "./data/IAM/h07/h07-069a/h07-069a-03-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "flue",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/h07/h07-069a/h07-069a-03-10.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "middle",
|
||
|
|
"gt_answers": "middle",
|
||
|
|
"image_path": "./data/IAM/h07/h07-069a/h07-069a-04-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "seven lies",
|
||
|
|
"gt_answers": "seventies",
|
||
|
|
"image_path": "./data/IAM/h07/h07-069a/h07-069a-04-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "speaking",
|
||
|
|
"gt_answers": "reaching",
|
||
|
|
"image_path": "./data/IAM/h07/h07-069a/h07-069a-04-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "levels",
|
||
|
|
"gt_answers": "levels",
|
||
|
|
"image_path": "./data/IAM/h07/h07-069a/h07-069a-04-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/h07/h07-069a/h07-069a-04-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "pelope",
|
||
|
|
"gt_answers": "perhaps",
|
||
|
|
"image_path": "./data/IAM/h07/h07-069a/h07-069a-04-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "s 3",
|
||
|
|
"gt_answers": "23",
|
||
|
|
"image_path": "./data/IAM/h07/h07-069a/h07-069a-04-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "pe",
|
||
|
|
"gt_answers": "per",
|
||
|
|
"image_path": "./data/IAM/h07/h07-069a/h07-069a-05-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "ced",
|
||
|
|
"gt_answers": "cent",
|
||
|
|
"image_path": "./data/IAM/h07/h07-069a/h07-069a-05-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "h",
|
||
|
|
"gt_answers": "in",
|
||
|
|
"image_path": "./data/IAM/h07/h07-069a/h07-069a-05-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "some",
|
||
|
|
"gt_answers": "some",
|
||
|
|
"image_path": "./data/IAM/h07/h07-069a/h07-069a-05-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "focus",
|
||
|
|
"gt_answers": "towns",
|
||
|
|
"image_path": "./data/IAM/h07/h07-069a/h07-069a-05-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "sefe",
|
||
|
|
"gt_answers": "before",
|
||
|
|
"image_path": "./data/IAM/h07/h07-069a/h07-069a-05-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "",
|
||
|
|
"gt_answers": "it",
|
||
|
|
"image_path": "./data/IAM/h07/h07-069a/h07-069a-05-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "lepus",
|
||
|
|
"gt_answers": "begins",
|
||
|
|
"image_path": "./data/IAM/h07/h07-069a/h07-069a-05-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/h07/h07-069a/h07-069a-05-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "declive",
|
||
|
|
"gt_answers": "decline",
|
||
|
|
"image_path": "./data/IAM/h07/h07-069a/h07-069a-05-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "her",
|
||
|
|
"gt_answers": "Her",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-00-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the husband's secret",
|
||
|
|
"gt_answers": "husbands",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-00-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "gentleness",
|
||
|
|
"gt_answers": "gentleness",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-00-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word and",
|
||
|
|
"gt_answers": "and",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-00-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word equability",
|
||
|
|
"gt_answers": "amiability",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-00-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "his",
|
||
|
|
"gt_answers": "his",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-00-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "self fish",
|
||
|
|
"gt_answers": "unselfish",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-00-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "love",
|
||
|
|
"gt_answers": "love",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-01-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "fore",
|
||
|
|
"gt_answers": "for",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-01-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "her",
|
||
|
|
"gt_answers": "her",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-01-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "she",
|
||
|
|
"gt_answers": "she",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-01-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "regared",
|
||
|
|
"gt_answers": "regarded",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-01-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a s",
|
||
|
|
"gt_answers": "as",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-01-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "signs",
|
||
|
|
"gt_answers": "signs",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-01-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "off",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-01-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "weakness",
|
||
|
|
"gt_answers": "weakness",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-01-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "he",
|
||
|
|
"gt_answers": "He",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-01-11.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "wols",
|
||
|
|
"gt_answers": "was",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-01-12.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "oc",
|
||
|
|
"gt_answers": "a",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-01-13.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "futile",
|
||
|
|
"gt_answers": "futile",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-02-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "creature",
|
||
|
|
"gt_answers": "creature",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-02-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "who",
|
||
|
|
"gt_answers": "who",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-02-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "had",
|
||
|
|
"gt_answers": "had",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-02-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "not",
|
||
|
|
"gt_answers": "not",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-02-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word cren",
|
||
|
|
"gt_answers": "even",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-02-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word proved",
|
||
|
|
"gt_answers": "proved",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-02-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "capable",
|
||
|
|
"gt_answers": "capable",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-02-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "f",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-02-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "ring",
|
||
|
|
"gt_answers": "giving",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-02-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "her",
|
||
|
|
"gt_answers": "her",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-03-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "jo",
|
||
|
|
"gt_answers": "a",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-03-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "child",
|
||
|
|
"gt_answers": "child",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-03-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "whatever",
|
||
|
|
"gt_answers": "Whatever",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-03-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "passion",
|
||
|
|
"gt_answers": "passion",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-03-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "she",
|
||
|
|
"gt_answers": "she",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-03-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "had",
|
||
|
|
"gt_answers": "had",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-03-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "experienced",
|
||
|
|
"gt_answers": "experienced",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-03-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "h",
|
||
|
|
"gt_answers": "in",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-03-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the c",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-03-10.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "first",
|
||
|
|
"gt_answers": "first",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-04-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word months",
|
||
|
|
"gt_answers": "months",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-04-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "f",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-04-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "marriage",
|
||
|
|
"gt_answers": "marriage",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-04-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "had",
|
||
|
|
"gt_answers": "had",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-04-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "ben",
|
||
|
|
"gt_answers": "been",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-04-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word hosted",
|
||
|
|
"gt_answers": "ousted",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-04-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "x y",
|
||
|
|
"gt_answers": "by",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-04-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "content type",
|
||
|
|
"gt_answers": "contempt",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-04-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "she",
|
||
|
|
"gt_answers": "She",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-05-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "hd",
|
||
|
|
"gt_answers": "had",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-05-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "never",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-05-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "loved",
|
||
|
|
"gt_answers": "loved",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-05-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "her",
|
||
|
|
"gt_answers": "him",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-05-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "her",
|
||
|
|
"gt_answers": "Her",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-05-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "thoughts",
|
||
|
|
"gt_answers": "thoughts",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-05-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "rushed",
|
||
|
|
"gt_answers": "rushed",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-05-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "x",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-05-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "osin",
|
||
|
|
"gt_answers": "Orsini",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-05-10.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "since",
|
||
|
|
"gt_answers": "Since",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-06-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-06-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "meeting",
|
||
|
|
"gt_answers": "meeting",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-06-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "h",
|
||
|
|
"gt_answers": "in",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-06-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the h",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-06-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "dorian",
|
||
|
|
"gt_answers": "Doria",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-06-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "palace",
|
||
|
|
"gt_answers": "Palace",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-06-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "no",
|
||
|
|
"gt_answers": "no",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-06-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "word",
|
||
|
|
"gt_answers": "word",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-06-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "had",
|
||
|
|
"gt_answers": "had",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-06-10.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "cow",
|
||
|
|
"gt_answers": "come",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-07-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "frou",
|
||
|
|
"gt_answers": "from",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-07-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word hi",
|
||
|
|
"gt_answers": "him",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-07-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "and",
|
||
|
|
"gt_answers": "and",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-07-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "'this is'",
|
||
|
|
"gt_answers": "this",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-07-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "silence",
|
||
|
|
"gt_answers": "silence",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-07-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "oppressed",
|
||
|
|
"gt_answers": "oppressed",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-07-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "her",
|
||
|
|
"gt_answers": "her",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-07-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "spirits",
|
||
|
|
"gt_answers": "spirits",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-07-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word with",
|
||
|
|
"gt_answers": "with",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-08-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a b",
|
||
|
|
"gt_answers": "a",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-08-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "weight",
|
||
|
|
"gt_answers": "weight",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-08-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "f",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-08-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "miss giving",
|
||
|
|
"gt_answers": "misgiving",
|
||
|
|
"image_path": "./data/IAM/k04/k04-068/k04-068-08-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word ink",
|
||
|
|
"gt_answers": "I",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-00-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "m",
|
||
|
|
"gt_answers": "m",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-00-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word person",
|
||
|
|
"gt_answers": "afraid",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-00-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "we",
|
||
|
|
"gt_answers": "we",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-00-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "can't fly",
|
||
|
|
"gt_answers": "cant",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-00-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "take",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-00-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "yours",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-00-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "into",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-00-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "account",
|
||
|
|
"gt_answers": "account",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-00-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "at",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-01-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "all",
|
||
|
|
"gt_answers": "all",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-01-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "It",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-01-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the name person",
|
||
|
|
"gt_answers": "s",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-01-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "a",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-01-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "c",
|
||
|
|
"gt_answers": "rule",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-01-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-01-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "rule of law",
|
||
|
|
"gt_answers": "III",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-01-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "i love you",
|
||
|
|
"gt_answers": "You",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-01-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "see",
|
||
|
|
"gt_answers": "see",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-01-10.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "you",
|
||
|
|
"gt_answers": "you",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-01-12.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word night",
|
||
|
|
"gt_answers": "might",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-02-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "php",
|
||
|
|
"gt_answers": "stop",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-02-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word row",
|
||
|
|
"gt_answers": "work",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-02-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-02-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "have",
|
||
|
|
"gt_answers": "have",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-02-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "well",
|
||
|
|
"gt_answers": "well",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-02-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "for",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-02-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "all shook up",
|
||
|
|
"gt_answers": "all",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-02-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word lord",
|
||
|
|
"gt_answers": "sorts",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-02-10.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a signature",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-02-11.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "reasons",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-03-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "or",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-03-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "y",
|
||
|
|
"gt_answers": "you",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-03-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "might",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-03-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "leave",
|
||
|
|
"gt_answers": "leave",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-03-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "him",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-03-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "But",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-03-10.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "how",
|
||
|
|
"gt_answers": "how",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-03-11.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "butterfly",
|
||
|
|
"gt_answers": "utterly",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-04-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "extraordinary",
|
||
|
|
"gt_answers": "extraordinary",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-04-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the name person",
|
||
|
|
"gt_answers": "said",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-04-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "Serena",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-04-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "arrogy",
|
||
|
|
"gt_answers": "angrily",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-04-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "you",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-05-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "must",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-05-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "be",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-05-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "living",
|
||
|
|
"gt_answers": "living",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-05-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "in",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-05-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "he",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-05-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the name of the person",
|
||
|
|
"gt_answers": "nineteenth",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-05-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "edward cullen",
|
||
|
|
"gt_answers": "century",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-05-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "Oh",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-06-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "but",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-06-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "it",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-06-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "s",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-06-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a snake",
|
||
|
|
"gt_answers": "a",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-06-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "very",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-06-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "george washington",
|
||
|
|
"gt_answers": "general",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-06-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "rule",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-06-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "Mrs",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-06-11.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the name of the person",
|
||
|
|
"gt_answers": "Buttery",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-06-12.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "you",
|
||
|
|
"gt_answers": "you",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-07-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "ll",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-07-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "find",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-07-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "that",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-07-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "no",
|
||
|
|
"gt_answers": "no",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-07-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word independence",
|
||
|
|
"gt_answers": "insurance",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-07-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a signature of a person",
|
||
|
|
"gt_answers": "companies",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-07-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "or",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-08-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word belling",
|
||
|
|
"gt_answers": "building",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-08-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "societe",
|
||
|
|
"gt_answers": "societies",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-08-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a signature",
|
||
|
|
"gt_answers": "for",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-08-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "thomas jefferson",
|
||
|
|
"gt_answers": "that",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-08-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the name of the person",
|
||
|
|
"gt_answers": "matter",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-08-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a signature of a person",
|
||
|
|
"gt_answers": "will",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-08-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "elove",
|
||
|
|
"gt_answers": "allow",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-09-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "for",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-09-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "he",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-09-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "wife's signature",
|
||
|
|
"gt_answers": "wifes",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-09-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "edwin morgan",
|
||
|
|
"gt_answers": "earnings",
|
||
|
|
"image_path": "./data/IAM/r03/r03-084/r03-084-09-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "no",
|
||
|
|
"gt_answers": "No",
|
||
|
|
"image_path": "./data/IAM/r06/r06-035/r06-035-00-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "he",
|
||
|
|
"gt_answers": "he",
|
||
|
|
"image_path": "./data/IAM/r06/r06-035/r06-035-00-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "t hud",
|
||
|
|
"gt_answers": "hadnt",
|
||
|
|
"image_path": "./data/IAM/r06/r06-035/r06-035-00-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "read",
|
||
|
|
"gt_answers": "read",
|
||
|
|
"image_path": "./data/IAM/r06/r06-035/r06-035-00-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "t",
|
||
|
|
"gt_answers": "it",
|
||
|
|
"image_path": "./data/IAM/r06/r06-035/r06-035-00-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "mad ave",
|
||
|
|
"gt_answers": "Madame",
|
||
|
|
"image_path": "./data/IAM/r06/r06-035/r06-035-00-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "id",
|
||
|
|
"gt_answers": "did",
|
||
|
|
"image_path": "./data/IAM/r06/r06-035/r06-035-00-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "all",
|
||
|
|
"gt_answers": "all",
|
||
|
|
"image_path": "./data/IAM/r06/r06-035/r06-035-00-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word hue",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/r06/r06-035/r06-035-00-10.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "reading",
|
||
|
|
"gt_answers": "reading",
|
||
|
|
"image_path": "./data/IAM/r06/r06-035/r06-035-01-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "nawrence",
|
||
|
|
"gt_answers": "Lawrence",
|
||
|
|
"image_path": "./data/IAM/r06/r06-035/r06-035-01-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "died",
|
||
|
|
"gt_answers": "died",
|
||
|
|
"image_path": "./data/IAM/r06/r06-035/r06-035-01-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "h",
|
||
|
|
"gt_answers": "in",
|
||
|
|
"image_path": "./data/IAM/r06/r06-035/r06-035-01-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "huss",
|
||
|
|
"gt_answers": "this",
|
||
|
|
"image_path": "./data/IAM/r06/r06-035/r06-035-01-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "every",
|
||
|
|
"gt_answers": "very",
|
||
|
|
"image_path": "./data/IAM/r06/r06-035/r06-035-01-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "place",
|
||
|
|
"gt_answers": "place",
|
||
|
|
"image_path": "./data/IAM/r06/r06-035/r06-035-01-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "he",
|
||
|
|
"gt_answers": "He",
|
||
|
|
"image_path": "./data/IAM/r06/r06-035/r06-035-02-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word used",
|
||
|
|
"gt_answers": "used",
|
||
|
|
"image_path": "./data/IAM/r06/r06-035/r06-035-02-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "o",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/r06/r06-035/r06-035-02-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "coe",
|
||
|
|
"gt_answers": "come",
|
||
|
|
"image_path": "./data/IAM/r06/r06-035/r06-035-02-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "o",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/r06/r06-035/r06-035-02-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "pierre",
|
||
|
|
"gt_answers": "Pierres",
|
||
|
|
"image_path": "./data/IAM/r06/r06-035/r06-035-02-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "bar",
|
||
|
|
"gt_answers": "Bar",
|
||
|
|
"image_path": "./data/IAM/r06/r06-035/r06-035-02-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "again",
|
||
|
|
"gt_answers": "again",
|
||
|
|
"image_path": "./data/IAM/r06/r06-035/r06-035-02-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "audi",
|
||
|
|
"gt_answers": "and",
|
||
|
|
"image_path": "./data/IAM/r06/r06-035/r06-035-02-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "again",
|
||
|
|
"gt_answers": "again",
|
||
|
|
"image_path": "./data/IAM/r06/r06-035/r06-035-03-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "n",
|
||
|
|
"gt_answers": "No",
|
||
|
|
"image_path": "./data/IAM/r06/r06-035/r06-035-03-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word love",
|
||
|
|
"gt_answers": "he",
|
||
|
|
"image_path": "./data/IAM/r06/r06-035/r06-035-03-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "couldn't",
|
||
|
|
"gt_answers": "couldnt",
|
||
|
|
"image_path": "./data/IAM/r06/r06-035/r06-035-03-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "really",
|
||
|
|
"gt_answers": "really",
|
||
|
|
"image_path": "./data/IAM/r06/r06-035/r06-035-03-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "remember",
|
||
|
|
"gt_answers": "remember",
|
||
|
|
"image_path": "./data/IAM/r06/r06-035/r06-035-03-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word him",
|
||
|
|
"gt_answers": "him",
|
||
|
|
"image_path": "./data/IAM/r06/r06-035/r06-035-04-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "me",
|
||
|
|
"gt_answers": "he",
|
||
|
|
"image_path": "./data/IAM/r06/r06-035/r06-035-04-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "was",
|
||
|
|
"gt_answers": "was",
|
||
|
|
"image_path": "./data/IAM/r06/r06-035/r06-035-04-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "one",
|
||
|
|
"gt_answers": "one",
|
||
|
|
"image_path": "./data/IAM/r06/r06-035/r06-035-04-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "f",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/r06/r06-035/r06-035-04-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "he",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/r06/r06-035/r06-035-04-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "crowd",
|
||
|
|
"gt_answers": "crowd",
|
||
|
|
"image_path": "./data/IAM/r06/r06-035/r06-035-04-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "The",
|
||
|
|
"image_path": "./data/IAM/r06/r06-035/r06-035-04-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "sun",
|
||
|
|
"gt_answers": "sun",
|
||
|
|
"image_path": "./data/IAM/r06/r06-035/r06-035-05-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "cae",
|
||
|
|
"gt_answers": "came",
|
||
|
|
"image_path": "./data/IAM/r06/r06-035/r06-035-05-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "out",
|
||
|
|
"gt_answers": "out",
|
||
|
|
"image_path": "./data/IAM/r06/r06-035/r06-035-05-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "mart",
|
||
|
|
"gt_answers": "Mart",
|
||
|
|
"image_path": "./data/IAM/r06/r06-035/r06-035-05-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "wet",
|
||
|
|
"gt_answers": "went",
|
||
|
|
"image_path": "./data/IAM/r06/r06-035/r06-035-05-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "shopping",
|
||
|
|
"gt_answers": "shopping",
|
||
|
|
"image_path": "./data/IAM/r06/r06-035/r06-035-05-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "I",
|
||
|
|
"image_path": "./data/IAM/r06/r06-035/r06-035-05-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "sat",
|
||
|
|
"gt_answers": "sat",
|
||
|
|
"image_path": "./data/IAM/r06/r06-035/r06-035-06-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "in",
|
||
|
|
"image_path": "./data/IAM/r06/r06-035/r06-035-06-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "he",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/r06/r06-035/r06-035-06-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "square",
|
||
|
|
"gt_answers": "square",
|
||
|
|
"image_path": "./data/IAM/r06/r06-035/r06-035-06-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "reading",
|
||
|
|
"gt_answers": "reading",
|
||
|
|
"image_path": "./data/IAM/r06/r06-035/r06-035-06-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "fe",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/r06/r06-035/r06-035-06-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "patriot",
|
||
|
|
"gt_answers": "Patriote",
|
||
|
|
"image_path": "./data/IAM/r06/r06-035/r06-035-06-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "what am i",
|
||
|
|
"gt_answers": "What",
|
||
|
|
"image_path": "./data/IAM/p06/p06-096/p06-096-00-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "he",
|
||
|
|
"gt_answers": "his",
|
||
|
|
"image_path": "./data/IAM/p06/p06-096/p06-096-00-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "ary",
|
||
|
|
"gt_answers": "story",
|
||
|
|
"image_path": "./data/IAM/p06/p06-096/p06-096-00-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "will",
|
||
|
|
"gt_answers": "will",
|
||
|
|
"image_path": "./data/IAM/p06/p06-096/p06-096-00-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "be",
|
||
|
|
"gt_answers": "be",
|
||
|
|
"image_path": "./data/IAM/p06/p06-096/p06-096-00-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "z",
|
||
|
|
"gt_answers": "I",
|
||
|
|
"image_path": "./data/IAM/p06/p06-096/p06-096-00-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "recor",
|
||
|
|
"gt_answers": "reckon",
|
||
|
|
"image_path": "./data/IAM/p06/p06-096/p06-096-00-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "he",
|
||
|
|
"gt_answers": "he",
|
||
|
|
"image_path": "./data/IAM/p06/p06-096/p06-096-01-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "u",
|
||
|
|
"gt_answers": "ll",
|
||
|
|
"image_path": "./data/IAM/p06/p06-096/p06-096-01-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "corrie",
|
||
|
|
"gt_answers": "confine",
|
||
|
|
"image_path": "./data/IAM/p06/p06-096/p06-096-01-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word self",
|
||
|
|
"gt_answers": "himself",
|
||
|
|
"image_path": "./data/IAM/p06/p06-096/p06-096-01-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "h",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/p06/p06-096/p06-096-01-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "he",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/p06/p06-096/p06-096-01-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "nervous",
|
||
|
|
"image_path": "./data/IAM/p06/p06-096/p06-096-01-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "ory",
|
||
|
|
"gt_answers": "origin",
|
||
|
|
"image_path": "./data/IAM/p06/p06-096/p06-096-02-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/p06/p06-096/p06-096-02-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "her",
|
||
|
|
"gt_answers": "her",
|
||
|
|
"image_path": "./data/IAM/p06/p06-096/p06-096-02-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word peter",
|
||
|
|
"gt_answers": "recent",
|
||
|
|
"image_path": "./data/IAM/p06/p06-096/p06-096-02-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "ille",
|
||
|
|
"gt_answers": "illness",
|
||
|
|
"image_path": "./data/IAM/p06/p06-096/p06-096-02-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a signature",
|
||
|
|
"gt_answers": "His",
|
||
|
|
"image_path": "./data/IAM/p06/p06-096/p06-096-02-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "noles",
|
||
|
|
"gt_answers": "notes",
|
||
|
|
"image_path": "./data/IAM/p06/p06-096/p06-096-02-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "will",
|
||
|
|
"gt_answers": "will",
|
||
|
|
"image_path": "./data/IAM/p06/p06-096/p06-096-03-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "be",
|
||
|
|
"gt_answers": "be",
|
||
|
|
"image_path": "./data/IAM/p06/p06-096/p06-096-03-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word truthful",
|
||
|
|
"gt_answers": "truthful",
|
||
|
|
"image_path": "./data/IAM/p06/p06-096/p06-096-03-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "brf",
|
||
|
|
"gt_answers": "but",
|
||
|
|
"image_path": "./data/IAM/p06/p06-096/p06-096-03-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word wilbur",
|
||
|
|
"gt_answers": "what",
|
||
|
|
"image_path": "./data/IAM/p06/p06-096/p06-096-03-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "he",
|
||
|
|
"gt_answers": "he",
|
||
|
|
"image_path": "./data/IAM/p06/p06-096/p06-096-03-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "leave",
|
||
|
|
"gt_answers": "leaves",
|
||
|
|
"image_path": "./data/IAM/p06/p06-096/p06-096-03-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word au",
|
||
|
|
"gt_answers": "out",
|
||
|
|
"image_path": "./data/IAM/p06/p06-096/p06-096-03-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "will",
|
||
|
|
"gt_answers": "will",
|
||
|
|
"image_path": "./data/IAM/p06/p06-096/p06-096-04-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "mafer",
|
||
|
|
"gt_answers": "matter",
|
||
|
|
"image_path": "./data/IAM/p06/p06-096/p06-096-04-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a m",
|
||
|
|
"gt_answers": "most",
|
||
|
|
"image_path": "./data/IAM/p06/p06-096/p06-096-04-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "shayne",
|
||
|
|
"gt_answers": "Strange",
|
||
|
|
"image_path": "./data/IAM/p06/p06-096/p06-096-04-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "how",
|
||
|
|
"gt_answers": "how",
|
||
|
|
"image_path": "./data/IAM/p06/p06-096/p06-096-04-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word ju",
|
||
|
|
"gt_answers": "just",
|
||
|
|
"image_path": "./data/IAM/p06/p06-096/p06-096-04-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "hen",
|
||
|
|
"gt_answers": "then",
|
||
|
|
"image_path": "./data/IAM/p06/p06-096/p06-096-05-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "philip",
|
||
|
|
"gt_answers": "Philip",
|
||
|
|
"image_path": "./data/IAM/p06/p06-096/p06-096-05-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "hw",
|
||
|
|
"gt_answers": "was",
|
||
|
|
"image_path": "./data/IAM/p06/p06-096/p06-096-05-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "d",
|
||
|
|
"gt_answers": "so",
|
||
|
|
"image_path": "./data/IAM/p06/p06-096/p06-096-05-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "edward elaine",
|
||
|
|
"gt_answers": "certain",
|
||
|
|
"image_path": "./data/IAM/p06/p06-096/p06-096-05-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "hat",
|
||
|
|
"gt_answers": "that",
|
||
|
|
"image_path": "./data/IAM/p06/p06-096/p06-096-05-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "nicola",
|
||
|
|
"gt_answers": "Nicholas",
|
||
|
|
"image_path": "./data/IAM/p06/p06-096/p06-096-05-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "world",
|
||
|
|
"gt_answers": "would",
|
||
|
|
"image_path": "./data/IAM/p06/p06-096/p06-096-06-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "love",
|
||
|
|
"gt_answers": "never",
|
||
|
|
"image_path": "./data/IAM/p06/p06-096/p06-096-06-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "beray",
|
||
|
|
"gt_answers": "betray",
|
||
|
|
"image_path": "./data/IAM/p06/p06-096/p06-096-06-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "lin",
|
||
|
|
"gt_answers": "him",
|
||
|
|
"image_path": "./data/IAM/p06/p06-096/p06-096-06-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "he",
|
||
|
|
"gt_answers": "He",
|
||
|
|
"image_path": "./data/IAM/p06/p06-096/p06-096-06-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "loved",
|
||
|
|
"gt_answers": "loved",
|
||
|
|
"image_path": "./data/IAM/p06/p06-096/p06-096-06-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "sandra",
|
||
|
|
"gt_answers": "Sandra",
|
||
|
|
"image_path": "./data/IAM/p06/p06-096/p06-096-07-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "f o o",
|
||
|
|
"gt_answers": "too",
|
||
|
|
"image_path": "./data/IAM/p06/p06-096/p06-096-07-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "deeply",
|
||
|
|
"gt_answers": "deeply",
|
||
|
|
"image_path": "./data/IAM/p06/p06-096/p06-096-07-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "o",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/p06/p06-096/p06-096-07-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "i love you",
|
||
|
|
"gt_answers": "ruin",
|
||
|
|
"image_path": "./data/IAM/p06/p06-096/p06-096-07-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "she",
|
||
|
|
"gt_answers": "her",
|
||
|
|
"image_path": "./data/IAM/p06/p06-096/p06-096-07-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word faith",
|
||
|
|
"gt_answers": "future",
|
||
|
|
"image_path": "./data/IAM/p06/p06-096/p06-096-08-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "happiness",
|
||
|
|
"gt_answers": "happiness",
|
||
|
|
"image_path": "./data/IAM/p06/p06-096/p06-096-08-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "had",
|
||
|
|
"gt_answers": "Had",
|
||
|
|
"image_path": "./data/IAM/p06/p06-096/p06-096-08-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "love",
|
||
|
|
"gt_answers": "ever",
|
||
|
|
"image_path": "./data/IAM/p06/p06-096/p06-096-08-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "circum",
|
||
|
|
"image_path": "./data/IAM/p06/p06-096/p06-096-08-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "annie",
|
||
|
|
"gt_answers": "stances",
|
||
|
|
"image_path": "./data/IAM/p06/p06-096/p06-096-09-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word compelled",
|
||
|
|
"gt_answers": "conspired",
|
||
|
|
"image_path": "./data/IAM/p06/p06-096/p06-096-09-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "so",
|
||
|
|
"image_path": "./data/IAM/p06/p06-096/p06-096-09-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "ludlow",
|
||
|
|
"gt_answers": "cunningly",
|
||
|
|
"image_path": "./data/IAM/p06/p06-096/p06-096-09-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "phillip's",
|
||
|
|
"gt_answers": "Philips",
|
||
|
|
"image_path": "./data/IAM/p06/p06-096/p06-096-10-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word king",
|
||
|
|
"gt_answers": "spirits",
|
||
|
|
"image_path": "./data/IAM/p06/p06-096/p06-096-10-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "doreel",
|
||
|
|
"gt_answers": "soared",
|
||
|
|
"image_path": "./data/IAM/p06/p06-096/p06-096-10-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word the",
|
||
|
|
"gt_answers": "The",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-00-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "main",
|
||
|
|
"gt_answers": "main",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-00-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "tasks",
|
||
|
|
"gt_answers": "tasks",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-00-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "oc",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-00-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-00-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "rollerston",
|
||
|
|
"gt_answers": "Rolleston",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-00-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "committee",
|
||
|
|
"gt_answers": "Committee",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-00-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "who q",
|
||
|
|
"gt_answers": "whose",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-00-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "advice in black ink",
|
||
|
|
"gt_answers": "advice",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-00-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "sw",
|
||
|
|
"gt_answers": "we",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-01-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "wee",
|
||
|
|
"gt_answers": "were",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-01-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "invited",
|
||
|
|
"gt_answers": "invited",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-01-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "x",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-01-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "review",
|
||
|
|
"gt_answers": "review",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-01-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "wee",
|
||
|
|
"gt_answers": "were",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-01-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "ao",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-01-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "usc",
|
||
|
|
"gt_answers": "advise",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-01-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "i love you",
|
||
|
|
"gt_answers": "on",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-01-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "earbuds",
|
||
|
|
"gt_answers": "a",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-01-12.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-01-14.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "circumstances",
|
||
|
|
"gt_answers": "circumstances",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-02-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "if",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-02-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "ya",
|
||
|
|
"gt_answers": "any",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-02-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a, b, c",
|
||
|
|
"gt_answers": "in",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-02-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "which",
|
||
|
|
"gt_answers": "which",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-02-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the h",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-02-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "supply",
|
||
|
|
"gt_answers": "supply",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-02-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-02-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "morphine",
|
||
|
|
"gt_answers": "morphine",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-02-10.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word land",
|
||
|
|
"gt_answers": "and",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-03-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "heron",
|
||
|
|
"gt_answers": "heroin",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-03-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word and",
|
||
|
|
"gt_answers": "and",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-03-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "preparations",
|
||
|
|
"gt_answers": "preparations",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-03-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "containing",
|
||
|
|
"gt_answers": "containing",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-03-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "these are the times",
|
||
|
|
"gt_answers": "these",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-03-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "substances",
|
||
|
|
"gt_answers": "substances",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-03-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a and o",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-04-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "pet sons",
|
||
|
|
"gt_answers": "persons",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-04-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "addicted",
|
||
|
|
"gt_answers": "addicted",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-04-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "x",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-04-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "those",
|
||
|
|
"gt_answers": "those",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-04-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "drugs",
|
||
|
|
"gt_answers": "drugs",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-04-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "might",
|
||
|
|
"gt_answers": "might",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-04-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "b",
|
||
|
|
"gt_answers": "be",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-04-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "regared",
|
||
|
|
"gt_answers": "regarded",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-04-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "as",
|
||
|
|
"gt_answers": "as",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-05-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "medcally",
|
||
|
|
"gt_answers": "medically",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-05-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "accessible",
|
||
|
|
"gt_answers": "advisable",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-05-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "b",
|
||
|
|
"gt_answers": "b",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-05-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "i love you",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-05-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "precessions",
|
||
|
|
"gt_answers": "precautions",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-05-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "which",
|
||
|
|
"gt_answers": "which",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-05-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "medical",
|
||
|
|
"gt_answers": "medical",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-05-10.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word procedure",
|
||
|
|
"gt_answers": "practitioners",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-06-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word administrative",
|
||
|
|
"gt_answers": "administering",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-06-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the number seven",
|
||
|
|
"gt_answers": "or",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-06-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "prescribing",
|
||
|
|
"gt_answers": "prescribing",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-06-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "morphine",
|
||
|
|
"gt_answers": "morphine",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-06-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the number seven",
|
||
|
|
"gt_answers": "or",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-06-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "heron",
|
||
|
|
"gt_answers": "heroin",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-07-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "should",
|
||
|
|
"gt_answers": "should",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-07-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "adopt",
|
||
|
|
"gt_answers": "adopt",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-07-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a and o",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-07-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "avoid",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-07-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "abuse",
|
||
|
|
"gt_answers": "abuse",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-07-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "and",
|
||
|
|
"gt_answers": "and",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-07-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "ya",
|
||
|
|
"gt_answers": "any",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-07-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "administrative assistant",
|
||
|
|
"gt_answers": "administrative",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-08-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "measures",
|
||
|
|
"gt_answers": "measures",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-08-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "that's all",
|
||
|
|
"gt_answers": "that",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-08-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "semed",
|
||
|
|
"gt_answers": "seemed",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-08-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "experience",
|
||
|
|
"gt_answers": "expedient",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-08-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the number 6",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-08-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "secure",
|
||
|
|
"gt_answers": "secure",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-09-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "obsurance",
|
||
|
|
"gt_answers": "observance",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-09-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "o",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-09-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "hose",
|
||
|
|
"gt_answers": "those",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-09-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the precincts",
|
||
|
|
"gt_answers": "precautions",
|
||
|
|
"image_path": "./data/IAM/h02/h02-035/h02-035-09-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "no",
|
||
|
|
"gt_answers": "Nor",
|
||
|
|
"image_path": "./data/IAM/g06/g06-042k/g06-042k-00-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "iz",
|
||
|
|
"gt_answers": "is",
|
||
|
|
"image_path": "./data/IAM/g06/g06-042k/g06-042k-00-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "this is person",
|
||
|
|
"gt_answers": "this",
|
||
|
|
"image_path": "./data/IAM/g06/g06-042k/g06-042k-00-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "oh",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/g06/g06-042k/g06-042k-00-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "be",
|
||
|
|
"gt_answers": "be",
|
||
|
|
"image_path": "./data/IAM/g06/g06-042k/g06-042k-00-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "wonder",
|
||
|
|
"gt_answers": "wondered",
|
||
|
|
"image_path": "./data/IAM/g06/g06-042k/g06-042k-00-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "t",
|
||
|
|
"gt_answers": "at",
|
||
|
|
"image_path": "./data/IAM/g06/g06-042k/g06-042k-00-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "far",
|
||
|
|
"gt_answers": "for",
|
||
|
|
"image_path": "./data/IAM/g06/g06-042k/g06-042k-00-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "even",
|
||
|
|
"image_path": "./data/IAM/g06/g06-042k/g06-042k-00-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "to day",
|
||
|
|
"gt_answers": "today",
|
||
|
|
"image_path": "./data/IAM/g06/g06-042k/g06-042k-00-10.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "n",
|
||
|
|
"gt_answers": "in",
|
||
|
|
"image_path": "./data/IAM/g06/g06-042k/g06-042k-00-12.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/g06/g06-042k/g06-042k-01-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "1970s",
|
||
|
|
"gt_answers": "1960s",
|
||
|
|
"image_path": "./data/IAM/g06/g06-042k/g06-042k-01-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "no",
|
||
|
|
"gt_answers": "no",
|
||
|
|
"image_path": "./data/IAM/g06/g06-042k/g06-042k-01-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "cher",
|
||
|
|
"gt_answers": "cure",
|
||
|
|
"image_path": "./data/IAM/g06/g06-042k/g06-042k-01-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "has",
|
||
|
|
"gt_answers": "has",
|
||
|
|
"image_path": "./data/IAM/g06/g06-042k/g06-042k-01-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "been",
|
||
|
|
"gt_answers": "been",
|
||
|
|
"image_path": "./data/IAM/g06/g06-042k/g06-042k-01-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "fmd",
|
||
|
|
"gt_answers": "found",
|
||
|
|
"image_path": "./data/IAM/g06/g06-042k/g06-042k-01-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "far",
|
||
|
|
"gt_answers": "for",
|
||
|
|
"image_path": "./data/IAM/g06/g06-042k/g06-042k-01-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/g06/g06-042k/g06-042k-01-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "6tic",
|
||
|
|
"image_path": "./data/IAM/g06/g06-042k/g06-042k-02-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "don louis",
|
||
|
|
"gt_answers": "douloureux",
|
||
|
|
"image_path": "./data/IAM/g06/g06-042k/g06-042k-02-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "ass",
|
||
|
|
"gt_answers": "As",
|
||
|
|
"image_path": "./data/IAM/g06/g06-042k/g06-042k-02-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "early",
|
||
|
|
"gt_answers": "early",
|
||
|
|
"image_path": "./data/IAM/g06/g06-042k/g06-042k-02-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the number 20",
|
||
|
|
"gt_answers": "as",
|
||
|
|
"image_path": "./data/IAM/g06/g06-042k/g06-042k-02-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "o o o",
|
||
|
|
"gt_answers": "1830",
|
||
|
|
"image_path": "./data/IAM/g06/g06-042k/g06-042k-02-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "when",
|
||
|
|
"gt_answers": "when",
|
||
|
|
"image_path": "./data/IAM/g06/g06-042k/g06-042k-02-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "angeles",
|
||
|
|
"gt_answers": "Anglesey",
|
||
|
|
"image_path": "./data/IAM/g06/g06-042k/g06-042k-02-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "believed",
|
||
|
|
"gt_answers": "believed",
|
||
|
|
"image_path": "./data/IAM/g06/g06-042k/g06-042k-03-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "himself",
|
||
|
|
"gt_answers": "himself",
|
||
|
|
"image_path": "./data/IAM/g06/g06-042k/g06-042k-03-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "oh",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/g06/g06-042k/g06-042k-03-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "be",
|
||
|
|
"gt_answers": "be",
|
||
|
|
"image_path": "./data/IAM/g06/g06-042k/g06-042k-03-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "no",
|
||
|
|
"gt_answers": "on",
|
||
|
|
"image_path": "./data/IAM/g06/g06-042k/g06-042k-03-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/g06/g06-042k/g06-042k-03-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "point",
|
||
|
|
"gt_answers": "point",
|
||
|
|
"image_path": "./data/IAM/g06/g06-042k/g06-042k-03-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/g06/g06-042k/g06-042k-03-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "death",
|
||
|
|
"gt_answers": "death",
|
||
|
|
"image_path": "./data/IAM/g06/g06-042k/g06-042k-03-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/g06/g06-042k/g06-042k-04-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "new",
|
||
|
|
"gt_answers": "new",
|
||
|
|
"image_path": "./data/IAM/g06/g06-042k/g06-042k-04-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "gersman",
|
||
|
|
"gt_answers": "German",
|
||
|
|
"image_path": "./data/IAM/g06/g06-042k/g06-042k-04-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "creative",
|
||
|
|
"gt_answers": "curative",
|
||
|
|
"image_path": "./data/IAM/g06/g06-042k/g06-042k-04-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "method",
|
||
|
|
"gt_answers": "method",
|
||
|
|
"image_path": "./data/IAM/g06/g06-042k/g06-042k-04-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "know",
|
||
|
|
"gt_answers": "known",
|
||
|
|
"image_path": "./data/IAM/g06/g06-042k/g06-042k-04-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "as",
|
||
|
|
"gt_answers": "as",
|
||
|
|
"image_path": "./data/IAM/g06/g06-042k/g06-042k-04-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "homeopathy",
|
||
|
|
"gt_answers": "homoeopathy",
|
||
|
|
"image_path": "./data/IAM/g06/g06-042k/g06-042k-05-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "lord",
|
||
|
|
"gt_answers": "had",
|
||
|
|
"image_path": "./data/IAM/g06/g06-042k/g06-042k-05-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "ben",
|
||
|
|
"gt_answers": "been",
|
||
|
|
"image_path": "./data/IAM/g06/g06-042k/g06-042k-05-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "brought",
|
||
|
|
"gt_answers": "brought",
|
||
|
|
"image_path": "./data/IAM/g06/g06-042k/g06-042k-05-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "t",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/g06/g06-042k/g06-042k-05-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "hs",
|
||
|
|
"gt_answers": "his",
|
||
|
|
"image_path": "./data/IAM/g06/g06-042k/g06-042k-05-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "notice",
|
||
|
|
"gt_answers": "notice",
|
||
|
|
"image_path": "./data/IAM/g06/g06-042k/g06-042k-05-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "when",
|
||
|
|
"gt_answers": "When",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-00-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "mr",
|
||
|
|
"gt_answers": "Mr",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-00-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word brown",
|
||
|
|
"gt_answers": "Brown",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-00-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "sat",
|
||
|
|
"gt_answers": "sat",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-00-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "clown",
|
||
|
|
"gt_answers": "down",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-00-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "labour",
|
||
|
|
"gt_answers": "Labour",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-00-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "m p",
|
||
|
|
"gt_answers": "M Ps",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-00-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "cheered",
|
||
|
|
"gt_answers": "cheered",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-00-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "mr",
|
||
|
|
"gt_answers": "for",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-00-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the letter a",
|
||
|
|
"gt_answers": "a",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-01-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "full",
|
||
|
|
"gt_answers": "full",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-01-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "minute",
|
||
|
|
"gt_answers": "minute",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-01-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "amd",
|
||
|
|
"gt_answers": "and",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-01-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "even",
|
||
|
|
"gt_answers": "even",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-01-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "hs",
|
||
|
|
"gt_answers": "his",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-01-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "brittest",
|
||
|
|
"gt_answers": "bitterest",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-01-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "opposites",
|
||
|
|
"gt_answers": "opponents",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-01-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "om",
|
||
|
|
"gt_answers": "on",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-02-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "defense center",
|
||
|
|
"gt_answers": "defence",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-02-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "jones",
|
||
|
|
"gt_answers": "joined",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-02-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "n",
|
||
|
|
"gt_answers": "in",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-02-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "m",
|
||
|
|
"gt_answers": "Mr",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-02-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "powell",
|
||
|
|
"gt_answers": "Powell",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-02-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "devod",
|
||
|
|
"gt_answers": "devoted",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-02-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "half",
|
||
|
|
"gt_answers": "half",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-02-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "hms",
|
||
|
|
"gt_answers": "his",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-02-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "speech",
|
||
|
|
"gt_answers": "speech",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-02-10.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-03-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "giving",
|
||
|
|
"gt_answers": "giving",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-03-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "dobelus",
|
||
|
|
"gt_answers": "details",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-03-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word person",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-03-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "plans",
|
||
|
|
"gt_answers": "plans",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-03-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "love",
|
||
|
|
"gt_answers": "for",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-03-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "improving",
|
||
|
|
"gt_answers": "improving",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-03-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-03-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "hospital",
|
||
|
|
"gt_answers": "hospital",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-03-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "service",
|
||
|
|
"gt_answers": "service",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-03-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "om",
|
||
|
|
"gt_answers": "on",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-04-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "wich",
|
||
|
|
"gt_answers": "which",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-04-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "ineed",
|
||
|
|
"gt_answers": "indeed",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-04-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-04-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "government",
|
||
|
|
"gt_answers": "Government",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-04-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word is",
|
||
|
|
"gt_answers": "is",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-04-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "making",
|
||
|
|
"gt_answers": "making",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-04-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "progress",
|
||
|
|
"gt_answers": "progress",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-04-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "his",
|
||
|
|
"gt_answers": "His",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-05-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "basic",
|
||
|
|
"gt_answers": "basic",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-05-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "defense",
|
||
|
|
"gt_answers": "defence",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-05-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "f",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-05-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-05-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "health",
|
||
|
|
"gt_answers": "Health",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-05-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "service",
|
||
|
|
"gt_answers": "Service",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-05-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "cubs",
|
||
|
|
"gt_answers": "cuts",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-05-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "wcs",
|
||
|
|
"gt_answers": "was",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-05-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "fah",
|
||
|
|
"gt_answers": "that",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-05-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "even",
|
||
|
|
"gt_answers": "even",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-06-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "afe",
|
||
|
|
"gt_answers": "after",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-06-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-06-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "proposed",
|
||
|
|
"gt_answers": "proposed",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-06-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "changes",
|
||
|
|
"gt_answers": "changes",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-06-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "he",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-06-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "netf",
|
||
|
|
"gt_answers": "net",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-06-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "costa",
|
||
|
|
"gt_answers": "cost",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-06-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-06-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-06-10.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "sera",
|
||
|
|
"gt_answers": "service",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-07-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "c",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-07-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "h",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-07-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the exchanger",
|
||
|
|
"gt_answers": "Exchequer",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-07-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "will",
|
||
|
|
"gt_answers": "will",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-07-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "have faith",
|
||
|
|
"gt_answers": "have",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-07-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "increased",
|
||
|
|
"gt_answers": "increased",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-07-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "over",
|
||
|
|
"gt_answers": "over",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-07-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "hope",
|
||
|
|
"gt_answers": "three",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-08-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "years",
|
||
|
|
"gt_answers": "years",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-08-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "by",
|
||
|
|
"gt_answers": "by",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-08-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "twenty",
|
||
|
|
"gt_answers": "20",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-08-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "per",
|
||
|
|
"gt_answers": "per",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-08-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "cent",
|
||
|
|
"gt_answers": "cent",
|
||
|
|
"image_path": "./data/IAM/a01/a01-128/a01-128-08-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "The",
|
||
|
|
"image_path": "./data/IAM/d01/d01-123/d01-123-00-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word person",
|
||
|
|
"gt_answers": "verso",
|
||
|
|
"image_path": "./data/IAM/d01/d01-123/d01-123-00-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "f",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/d01/d01-123/d01-123-00-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "ahe",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/d01/d01-123/d01-123-00-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "roll",
|
||
|
|
"gt_answers": "roll",
|
||
|
|
"image_path": "./data/IAM/d01/d01-123/d01-123-00-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word moss",
|
||
|
|
"gt_answers": "was",
|
||
|
|
"image_path": "./data/IAM/d01/d01-123/d01-123-00-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "blank",
|
||
|
|
"gt_answers": "blank",
|
||
|
|
"image_path": "./data/IAM/d01/d01-123/d01-123-00-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word angel",
|
||
|
|
"gt_answers": "and",
|
||
|
|
"image_path": "./data/IAM/d01/d01-123/d01-123-00-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "one",
|
||
|
|
"gt_answers": "one",
|
||
|
|
"image_path": "./data/IAM/d01/d01-123/d01-123-00-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word camelot",
|
||
|
|
"gt_answers": "cannot",
|
||
|
|
"image_path": "./data/IAM/d01/d01-123/d01-123-01-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "hunk",
|
||
|
|
"gt_answers": "think",
|
||
|
|
"image_path": "./data/IAM/d01/d01-123/d01-123-01-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "f",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/d01/d01-123/d01-123-01-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "am",
|
||
|
|
"gt_answers": "an",
|
||
|
|
"image_path": "./data/IAM/d01/d01-123/d01-123-01-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "author",
|
||
|
|
"gt_answers": "author",
|
||
|
|
"image_path": "./data/IAM/d01/d01-123/d01-123-01-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "mark",
|
||
|
|
"gt_answers": "Mark",
|
||
|
|
"image_path": "./data/IAM/d01/d01-123/d01-123-01-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "for",
|
||
|
|
"gt_answers": "for",
|
||
|
|
"image_path": "./data/IAM/d01/d01-123/d01-123-01-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "ex example",
|
||
|
|
"gt_answers": "example",
|
||
|
|
"image_path": "./data/IAM/d01/d01-123/d01-123-01-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "slinging",
|
||
|
|
"gt_answers": "sending",
|
||
|
|
"image_path": "./data/IAM/d01/d01-123/d01-123-02-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "i love you",
|
||
|
|
"gt_answers": "out",
|
||
|
|
"image_path": "./data/IAM/d01/d01-123/d01-123-02-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "his",
|
||
|
|
"gt_answers": "his",
|
||
|
|
"image_path": "./data/IAM/d01/d01-123/d01-123-02-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "gospel",
|
||
|
|
"gt_answers": "Gospel",
|
||
|
|
"image_path": "./data/IAM/d01/d01-123/d01-123-02-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "locking",
|
||
|
|
"gt_answers": "lacking",
|
||
|
|
"image_path": "./data/IAM/d01/d01-123/d01-123-02-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/d01/d01-123/d01-123-02-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "endling",
|
||
|
|
"gt_answers": "ending",
|
||
|
|
"image_path": "./data/IAM/d01/d01-123/d01-123-02-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word nile",
|
||
|
|
"gt_answers": "while",
|
||
|
|
"image_path": "./data/IAM/d01/d01-123/d01-123-03-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "onl",
|
||
|
|
"gt_answers": "one",
|
||
|
|
"image_path": "./data/IAM/d01/d01-123/d01-123-03-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "no hole",
|
||
|
|
"gt_answers": "whole",
|
||
|
|
"image_path": "./data/IAM/d01/d01-123/d01-123-03-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "id",
|
||
|
|
"gt_answers": "side",
|
||
|
|
"image_path": "./data/IAM/d01/d01-123/d01-123-03-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "f",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/d01/d01-123/d01-123-03-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "hs",
|
||
|
|
"gt_answers": "his",
|
||
|
|
"image_path": "./data/IAM/d01/d01-123/d01-123-03-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "roll",
|
||
|
|
"gt_answers": "roll",
|
||
|
|
"image_path": "./data/IAM/d01/d01-123/d01-123-03-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "no c",
|
||
|
|
"gt_answers": "was",
|
||
|
|
"image_path": "./data/IAM/d01/d01-123/d01-123-03-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the name of the person",
|
||
|
|
"gt_answers": "unused",
|
||
|
|
"image_path": "./data/IAM/d01/d01-123/d01-123-03-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "The",
|
||
|
|
"image_path": "./data/IAM/d01/d01-123/d01-123-04-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "coelex",
|
||
|
|
"gt_answers": "codex",
|
||
|
|
"image_path": "./data/IAM/d01/d01-123/d01-123-04-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "form",
|
||
|
|
"gt_answers": "form",
|
||
|
|
"image_path": "./data/IAM/d01/d01-123/d01-123-04-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "aaa",
|
||
|
|
"gt_answers": "was",
|
||
|
|
"image_path": "./data/IAM/d01/d01-123/d01-123-04-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "tom",
|
||
|
|
"gt_answers": "not",
|
||
|
|
"image_path": "./data/IAM/d01/d01-123/d01-123-04-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the number 40",
|
||
|
|
"gt_answers": "so",
|
||
|
|
"image_path": "./data/IAM/d01/d01-123/d01-123-04-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "accomodating",
|
||
|
|
"gt_answers": "accommodating",
|
||
|
|
"image_path": "./data/IAM/d01/d01-123/d01-123-04-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "evan",
|
||
|
|
"gt_answers": "Even",
|
||
|
|
"image_path": "./data/IAM/d01/d01-123/d01-123-04-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "m",
|
||
|
|
"gt_answers": "in",
|
||
|
|
"image_path": "./data/IAM/d01/d01-123/d01-123-05-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/d01/d01-123/d01-123-05-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "casl",
|
||
|
|
"gt_answers": "case",
|
||
|
|
"image_path": "./data/IAM/d01/d01-123/d01-123-05-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "f",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/d01/d01-123/d01-123-05-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "he",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/d01/d01-123/d01-123-05-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "single sheet",
|
||
|
|
"gt_answers": "singlesheet",
|
||
|
|
"image_path": "./data/IAM/d01/d01-123/d01-123-05-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "gure",
|
||
|
|
"gt_answers": "quire",
|
||
|
|
"image_path": "./data/IAM/d01/d01-123/d01-123-05-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "am",
|
||
|
|
"gt_answers": "an",
|
||
|
|
"image_path": "./data/IAM/d01/d01-123/d01-123-05-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "exha",
|
||
|
|
"gt_answers": "extra",
|
||
|
|
"image_path": "./data/IAM/d01/d01-123/d01-123-05-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "sheet",
|
||
|
|
"gt_answers": "sheet",
|
||
|
|
"image_path": "./data/IAM/d01/d01-123/d01-123-06-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "glud",
|
||
|
|
"gt_answers": "glued",
|
||
|
|
"image_path": "./data/IAM/d01/d01-123/d01-123-06-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "on",
|
||
|
|
"gt_answers": "on",
|
||
|
|
"image_path": "./data/IAM/d01/d01-123/d01-123-06-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "might",
|
||
|
|
"gt_answers": "might",
|
||
|
|
"image_path": "./data/IAM/d01/d01-123/d01-123-06-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "have",
|
||
|
|
"gt_answers": "have",
|
||
|
|
"image_path": "./data/IAM/d01/d01-123/d01-123-06-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "ag",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/d01/d01-123/d01-123-06-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "be",
|
||
|
|
"gt_answers": "be",
|
||
|
|
"image_path": "./data/IAM/d01/d01-123/d01-123-06-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "gummil",
|
||
|
|
"gt_answers": "gummed",
|
||
|
|
"image_path": "./data/IAM/d01/d01-123/d01-123-06-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "om",
|
||
|
|
"gt_answers": "on",
|
||
|
|
"image_path": "./data/IAM/d01/d01-123/d01-123-06-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "over",
|
||
|
|
"gt_answers": "over",
|
||
|
|
"image_path": "./data/IAM/d01/d01-123/d01-123-07-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/d01/d01-123/d01-123-07-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "binding",
|
||
|
|
"gt_answers": "binding",
|
||
|
|
"image_path": "./data/IAM/d01/d01-123/d01-123-07-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "c",
|
||
|
|
"gt_answers": "or",
|
||
|
|
"image_path": "./data/IAM/d01/d01-123/d01-123-07-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/d01/d01-123/d01-123-07-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "whole",
|
||
|
|
"gt_answers": "whole",
|
||
|
|
"image_path": "./data/IAM/d01/d01-123/d01-123-07-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "codex",
|
||
|
|
"gt_answers": "codex",
|
||
|
|
"image_path": "./data/IAM/d01/d01-123/d01-123-07-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "nebuchadnezzar",
|
||
|
|
"gt_answers": "rebound",
|
||
|
|
"image_path": "./data/IAM/d01/d01-123/d01-123-07-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "lester",
|
||
|
|
"gt_answers": "Lester",
|
||
|
|
"image_path": "./data/IAM/m02/m02-072/m02-072-00-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "was",
|
||
|
|
"gt_answers": "was",
|
||
|
|
"image_path": "./data/IAM/m02/m02-072/m02-072-00-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "not",
|
||
|
|
"gt_answers": "not",
|
||
|
|
"image_path": "./data/IAM/m02/m02-072/m02-072-00-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "home",
|
||
|
|
"gt_answers": "home",
|
||
|
|
"image_path": "./data/IAM/m02/m02-072/m02-072-00-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "yet",
|
||
|
|
"gt_answers": "yet",
|
||
|
|
"image_path": "./data/IAM/m02/m02-072/m02-072-00-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "from",
|
||
|
|
"gt_answers": "from",
|
||
|
|
"image_path": "./data/IAM/m02/m02-072/m02-072-00-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "hs",
|
||
|
|
"gt_answers": "his",
|
||
|
|
"image_path": "./data/IAM/m02/m02-072/m02-072-00-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "formable",
|
||
|
|
"gt_answers": "formidable",
|
||
|
|
"image_path": "./data/IAM/m02/m02-072/m02-072-00-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "dickies",
|
||
|
|
"gt_answers": "duties",
|
||
|
|
"image_path": "./data/IAM/m02/m02-072/m02-072-01-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "and",
|
||
|
|
"gt_answers": "and",
|
||
|
|
"image_path": "./data/IAM/m02/m02-072/m02-072-01-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "his",
|
||
|
|
"gt_answers": "his",
|
||
|
|
"image_path": "./data/IAM/m02/m02-072/m02-072-01-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "wife",
|
||
|
|
"gt_answers": "wife",
|
||
|
|
"image_path": "./data/IAM/m02/m02-072/m02-072-01-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "m",
|
||
|
|
"gt_answers": "in",
|
||
|
|
"image_path": "./data/IAM/m02/m02-072/m02-072-01-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/m02/m02-072/m02-072-01-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "maner",
|
||
|
|
"gt_answers": "manner",
|
||
|
|
"image_path": "./data/IAM/m02/m02-072/m02-072-01-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "f",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/m02/m02-072/m02-072-01-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "middle brow",
|
||
|
|
"gt_answers": "middlebrow",
|
||
|
|
"image_path": "./data/IAM/m02/m02-072/m02-072-01-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "wives",
|
||
|
|
"gt_answers": "wives",
|
||
|
|
"image_path": "./data/IAM/m02/m02-072/m02-072-02-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "romantic",
|
||
|
|
"gt_answers": "romanced",
|
||
|
|
"image_path": "./data/IAM/m02/m02-072/m02-072-02-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "about me",
|
||
|
|
"gt_answers": "about",
|
||
|
|
"image_path": "./data/IAM/m02/m02-072/m02-072-02-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "chimie",
|
||
|
|
"gt_answers": "him",
|
||
|
|
"image_path": "./data/IAM/m02/m02-072/m02-072-02-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the name of person",
|
||
|
|
"gt_answers": "in",
|
||
|
|
"image_path": "./data/IAM/m02/m02-072/m02-072-02-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "his",
|
||
|
|
"gt_answers": "his",
|
||
|
|
"image_path": "./data/IAM/m02/m02-072/m02-072-02-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "absence",
|
||
|
|
"gt_answers": "absence",
|
||
|
|
"image_path": "./data/IAM/m02/m02-072/m02-072-02-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "lester",
|
||
|
|
"gt_answers": "Lester",
|
||
|
|
"image_path": "./data/IAM/m02/m02-072/m02-072-02-10.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "was",
|
||
|
|
"gt_answers": "was",
|
||
|
|
"image_path": "./data/IAM/m02/m02-072/m02-072-02-11.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a",
|
||
|
|
"gt_answers": "a",
|
||
|
|
"image_path": "./data/IAM/m02/m02-072/m02-072-02-12.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "country",
|
||
|
|
"gt_answers": "country",
|
||
|
|
"image_path": "./data/IAM/m02/m02-072/m02-072-03-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "lawyer",
|
||
|
|
"gt_answers": "lawyer",
|
||
|
|
"image_path": "./data/IAM/m02/m02-072/m02-072-03-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "and",
|
||
|
|
"gt_answers": "and",
|
||
|
|
"image_path": "./data/IAM/m02/m02-072/m02-072-03-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "we",
|
||
|
|
"gt_answers": "we",
|
||
|
|
"image_path": "./data/IAM/m02/m02-072/m02-072-03-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "wee",
|
||
|
|
"gt_answers": "were",
|
||
|
|
"image_path": "./data/IAM/m02/m02-072/m02-072-03-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "ver",
|
||
|
|
"gt_answers": "very",
|
||
|
|
"image_path": "./data/IAM/m02/m02-072/m02-072-03-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "small",
|
||
|
|
"gt_answers": "small",
|
||
|
|
"image_path": "./data/IAM/m02/m02-072/m02-072-03-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "m i r e b e s",
|
||
|
|
"gt_answers": "4microdebs",
|
||
|
|
"image_path": "./data/IAM/m02/m02-072/m02-072-03-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "wen h",
|
||
|
|
"gt_answers": "werent",
|
||
|
|
"image_path": "./data/IAM/m02/m02-072/m02-072-03-10.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word we",
|
||
|
|
"gt_answers": "we",
|
||
|
|
"image_path": "./data/IAM/m02/m02-072/m02-072-03-11.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "sally",
|
||
|
|
"gt_answers": "Sally",
|
||
|
|
"image_path": "./data/IAM/m02/m02-072/m02-072-04-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "here",
|
||
|
|
"gt_answers": "here",
|
||
|
|
"image_path": "./data/IAM/m02/m02-072/m02-072-04-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "she",
|
||
|
|
"gt_answers": "she",
|
||
|
|
"image_path": "./data/IAM/m02/m02-072/m02-072-04-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "had",
|
||
|
|
"gt_answers": "had",
|
||
|
|
"image_path": "./data/IAM/m02/m02-072/m02-072-04-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "dropped",
|
||
|
|
"gt_answers": "dropped",
|
||
|
|
"image_path": "./data/IAM/m02/m02-072/m02-072-04-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "hef",
|
||
|
|
"gt_answers": "her",
|
||
|
|
"image_path": "./data/IAM/m02/m02-072/m02-072-04-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "voice",
|
||
|
|
"gt_answers": "voice",
|
||
|
|
"image_path": "./data/IAM/m02/m02-072/m02-072-04-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "in",
|
||
|
|
"image_path": "./data/IAM/m02/m02-072/m02-072-04-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "homage",
|
||
|
|
"gt_answers": "homage",
|
||
|
|
"image_path": "./data/IAM/m02/m02-072/m02-072-04-10.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word o",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/m02/m02-072/m02-072-04-11.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/m02/m02-072/m02-072-04-12.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "economic",
|
||
|
|
"gt_answers": "economic",
|
||
|
|
"image_path": "./data/IAM/m02/m02-072/m02-072-05-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "system",
|
||
|
|
"gt_answers": "system",
|
||
|
|
"image_path": "./data/IAM/m02/m02-072/m02-072-05-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word that",
|
||
|
|
"gt_answers": "that",
|
||
|
|
"image_path": "./data/IAM/m02/m02-072/m02-072-05-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "w s",
|
||
|
|
"gt_answers": "was",
|
||
|
|
"image_path": "./data/IAM/m02/m02-072/m02-072-05-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "reversed",
|
||
|
|
"gt_answers": "reverenced",
|
||
|
|
"image_path": "./data/IAM/m02/m02-072/m02-072-05-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "when",
|
||
|
|
"gt_answers": "when",
|
||
|
|
"image_path": "./data/IAM/m02/m02-072/m02-072-05-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "he",
|
||
|
|
"gt_answers": "he",
|
||
|
|
"image_path": "./data/IAM/m02/m02-072/m02-072-05-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "thought",
|
||
|
|
"gt_answers": "thought",
|
||
|
|
"image_path": "./data/IAM/m02/m02-072/m02-072-05-10.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "we",
|
||
|
|
"gt_answers": "we",
|
||
|
|
"image_path": "./data/IAM/m02/m02-072/m02-072-06-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "night",
|
||
|
|
"gt_answers": "might",
|
||
|
|
"image_path": "./data/IAM/m02/m02-072/m02-072-06-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "get",
|
||
|
|
"gt_answers": "get",
|
||
|
|
"image_path": "./data/IAM/m02/m02-072/m02-072-06-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "g o",
|
||
|
|
"gt_answers": "GO",
|
||
|
|
"image_path": "./data/IAM/m02/m02-072/m02-072-06-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a ring",
|
||
|
|
"gt_answers": "a",
|
||
|
|
"image_path": "./data/IAM/m02/m02-072/m02-072-06-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the number 8",
|
||
|
|
"gt_answers": "bit",
|
||
|
|
"image_path": "./data/IAM/m02/m02-072/m02-072-06-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "better",
|
||
|
|
"gt_answers": "better",
|
||
|
|
"image_path": "./data/IAM/m02/m02-072/m02-072-06-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "f",
|
||
|
|
"gt_answers": "if",
|
||
|
|
"image_path": "./data/IAM/m02/m02-072/m02-072-06-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "he",
|
||
|
|
"gt_answers": "he",
|
||
|
|
"image_path": "./data/IAM/m02/m02-072/m02-072-06-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word entrada",
|
||
|
|
"gt_answers": "entered",
|
||
|
|
"image_path": "./data/IAM/m02/m02-072/m02-072-06-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "politics",
|
||
|
|
"gt_answers": "politics",
|
||
|
|
"image_path": "./data/IAM/m02/m02-072/m02-072-06-10.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "apart",
|
||
|
|
"gt_answers": "Apart",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-00-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "four",
|
||
|
|
"gt_answers": "from",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-00-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "there",
|
||
|
|
"gt_answers": "their",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-00-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "formal",
|
||
|
|
"gt_answers": "formal",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-00-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "admirably",
|
||
|
|
"gt_answers": "Admiralty",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-00-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "house",
|
||
|
|
"gt_answers": "House",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-00-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "talks",
|
||
|
|
"gt_answers": "talks",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-00-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "followed",
|
||
|
|
"gt_answers": "followed",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-00-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "by",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-00-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word lucy",
|
||
|
|
"gt_answers": "lunch",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-00-10.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "givern",
|
||
|
|
"gt_answers": "given",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-01-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "by person",
|
||
|
|
"gt_answers": "by",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-01-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "lady",
|
||
|
|
"gt_answers": "Lady",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-01-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "doroty",
|
||
|
|
"gt_answers": "Dorothy",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-01-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the name of the person",
|
||
|
|
"gt_answers": "Macmillan",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-01-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "wilfred",
|
||
|
|
"gt_answers": "with",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-01-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "mrs",
|
||
|
|
"gt_answers": "Mrs",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-01-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "kennedy",
|
||
|
|
"gt_answers": "Kennedy",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-01-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "ancl",
|
||
|
|
"gt_answers": "and",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-01-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "other",
|
||
|
|
"gt_answers": "other",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-01-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "guests",
|
||
|
|
"gt_answers": "guests",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-02-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "preset",
|
||
|
|
"gt_answers": "present",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-02-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "mr",
|
||
|
|
"gt_answers": "Mr",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-02-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "kennely",
|
||
|
|
"gt_answers": "Kennedy",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-02-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "and",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-02-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "mr",
|
||
|
|
"gt_answers": "Mr",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-02-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "machiavelli",
|
||
|
|
"gt_answers": "Macmillan",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-02-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "wet",
|
||
|
|
"gt_answers": "met",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-02-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "three",
|
||
|
|
"gt_answers": "three",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-02-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "uore",
|
||
|
|
"gt_answers": "more",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-02-10.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "hures",
|
||
|
|
"gt_answers": "times",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-03-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "yesterday",
|
||
|
|
"gt_answers": "yesterday",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-03-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "l",
|
||
|
|
"gt_answers": "In",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-03-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "paris",
|
||
|
|
"gt_answers": "PARIS",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-03-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "mr person",
|
||
|
|
"gt_answers": "Mr",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-03-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "deau",
|
||
|
|
"gt_answers": "Dean",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-03-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "ruk",
|
||
|
|
"gt_answers": "Rusk",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-03-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "us s",
|
||
|
|
"gt_answers": "US",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-03-10.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "secretary",
|
||
|
|
"gt_answers": "Secretary",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-03-11.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "f",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-03-12.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "state",
|
||
|
|
"gt_answers": "State",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-04-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word gage",
|
||
|
|
"gt_answers": "gave",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-04-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a",
|
||
|
|
"gt_answers": "a",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-04-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "go minute",
|
||
|
|
"gt_answers": "90minute",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-04-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "breifing",
|
||
|
|
"gt_answers": "briefing",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-04-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "ol",
|
||
|
|
"gt_answers": "on",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-04-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-04-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "vienna",
|
||
|
|
"gt_answers": "Vienna",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-04-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "talks",
|
||
|
|
"gt_answers": "talks",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-04-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "t",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-04-10.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "he",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-04-11.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "15-nation nation",
|
||
|
|
"gt_answers": "15nation",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-05-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "c",
|
||
|
|
"gt_answers": "Nato",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-05-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word council",
|
||
|
|
"gt_answers": "council",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-05-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "souce",
|
||
|
|
"gt_answers": "Some",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-05-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "f",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-05-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "history",
|
||
|
|
"gt_answers": "his",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-05-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "listeners",
|
||
|
|
"gt_answers": "listeners",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-05-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "salo",
|
||
|
|
"gt_answers": "said",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-05-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "he",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-05-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "was",
|
||
|
|
"gt_answers": "was",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-05-10.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word father",
|
||
|
|
"gt_answers": "rather",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-06-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "pessimistic",
|
||
|
|
"gt_answers": "pessimistic",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-06-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "and",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-06-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "falk",
|
||
|
|
"gt_answers": "talked",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-06-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "of",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-06-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a",
|
||
|
|
"gt_answers": "a",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-06-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "berlin",
|
||
|
|
"gt_answers": "Berlin",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-06-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "crisis",
|
||
|
|
"gt_answers": "crisis",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-06-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "later",
|
||
|
|
"gt_answers": "later",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-06-10.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "this",
|
||
|
|
"gt_answers": "this",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-06-11.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "yes",
|
||
|
|
"gt_answers": "year",
|
||
|
|
"image_path": "./data/IAM/a04/a04-023/a04-023-06-12.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "n",
|
||
|
|
"gt_answers": "In",
|
||
|
|
"image_path": "./data/IAM/j04/j04-039/j04-039-00-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/j04/j04-039/j04-039-00-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "preliminary",
|
||
|
|
"gt_answers": "preliminary",
|
||
|
|
"image_path": "./data/IAM/j04/j04-039/j04-039-00-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "experiments",
|
||
|
|
"gt_answers": "experiments",
|
||
|
|
"image_path": "./data/IAM/j04/j04-039/j04-039-00-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/j04/j04-039/j04-039-00-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "tracer",
|
||
|
|
"image_path": "./data/IAM/j04/j04-039/j04-039-00-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "was",
|
||
|
|
"image_path": "./data/IAM/j04/j04-039/j04-039-00-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "added",
|
||
|
|
"gt_answers": "added",
|
||
|
|
"image_path": "./data/IAM/j04/j04-039/j04-039-00-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/j04/j04-039/j04-039-00-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "a",
|
||
|
|
"image_path": "./data/IAM/j04/j04-039/j04-039-00-10.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "liter",
|
||
|
|
"image_path": "./data/IAM/j04/j04-039/j04-039-01-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "f",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/j04/j04-039/j04-039-01-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "effulent",
|
||
|
|
"gt_answers": "effluent",
|
||
|
|
"image_path": "./data/IAM/j04/j04-039/j04-039-01-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word and",
|
||
|
|
"gt_answers": "and",
|
||
|
|
"image_path": "./data/IAM/j04/j04-039/j04-039-01-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "immeditely",
|
||
|
|
"gt_answers": "immediately",
|
||
|
|
"image_path": "./data/IAM/j04/j04-039/j04-039-01-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word predicted",
|
||
|
|
"gt_answers": "coprecipitated",
|
||
|
|
"image_path": "./data/IAM/j04/j04-039/j04-039-01-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word with",
|
||
|
|
"gt_answers": "with",
|
||
|
|
"image_path": "./data/IAM/j04/j04-039/j04-039-01-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "astrium",
|
||
|
|
"gt_answers": "strontium",
|
||
|
|
"image_path": "./data/IAM/j04/j04-039/j04-039-02-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "sulp hate",
|
||
|
|
"gt_answers": "sulphate",
|
||
|
|
"image_path": "./data/IAM/j04/j04-039/j04-039-02-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "from",
|
||
|
|
"gt_answers": "from",
|
||
|
|
"image_path": "./data/IAM/j04/j04-039/j04-039-02-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "hot",
|
||
|
|
"gt_answers": "hot",
|
||
|
|
"image_path": "./data/IAM/j04/j04-039/j04-039-02-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "solution",
|
||
|
|
"gt_answers": "solution",
|
||
|
|
"image_path": "./data/IAM/j04/j04-039/j04-039-02-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "chemical",
|
||
|
|
"gt_answers": "Chemical",
|
||
|
|
"image_path": "./data/IAM/j04/j04-039/j04-039-02-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "exchange",
|
||
|
|
"gt_answers": "exchange",
|
||
|
|
"image_path": "./data/IAM/j04/j04-039/j04-039-02-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "wa",
|
||
|
|
"gt_answers": "was",
|
||
|
|
"image_path": "./data/IAM/j04/j04-039/j04-039-03-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word complete",
|
||
|
|
"gt_answers": "complete",
|
||
|
|
"image_path": "./data/IAM/j04/j04-039/j04-039-03-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "under",
|
||
|
|
"image_path": "./data/IAM/j04/j04-039/j04-039-03-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word these",
|
||
|
|
"gt_answers": "these",
|
||
|
|
"image_path": "./data/IAM/j04/j04-039/j04-039-03-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "conditions",
|
||
|
|
"gt_answers": "conditions",
|
||
|
|
"image_path": "./data/IAM/j04/j04-039/j04-039-03-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "t",
|
||
|
|
"gt_answers": "but",
|
||
|
|
"image_path": "./data/IAM/j04/j04-039/j04-039-03-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "when",
|
||
|
|
"gt_answers": "when",
|
||
|
|
"image_path": "./data/IAM/j04/j04-039/j04-039-03-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/j04/j04-039/j04-039-03-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "tracer",
|
||
|
|
"image_path": "./data/IAM/j04/j04-039/j04-039-03-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word wo",
|
||
|
|
"gt_answers": "was",
|
||
|
|
"image_path": "./data/IAM/j04/j04-039/j04-039-04-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "adolescence",
|
||
|
|
"gt_answers": "added",
|
||
|
|
"image_path": "./data/IAM/j04/j04-039/j04-039-04-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/j04/j04-039/j04-039-04-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "allline",
|
||
|
|
"gt_answers": "alkaline",
|
||
|
|
"image_path": "./data/IAM/j04/j04-039/j04-039-04-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the efficient",
|
||
|
|
"gt_answers": "effluent",
|
||
|
|
"image_path": "./data/IAM/j04/j04-039/j04-039-04-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "and",
|
||
|
|
"gt_answers": "and",
|
||
|
|
"image_path": "./data/IAM/j04/j04-039/j04-039-04-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "allowed",
|
||
|
|
"gt_answers": "allowed",
|
||
|
|
"image_path": "./data/IAM/j04/j04-039/j04-039-04-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "o",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/j04/j04-039/j04-039-04-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word stan",
|
||
|
|
"gt_answers": "stand",
|
||
|
|
"image_path": "./data/IAM/j04/j04-039/j04-039-05-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "for",
|
||
|
|
"gt_answers": "for",
|
||
|
|
"image_path": "./data/IAM/j04/j04-039/j04-039-05-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "several",
|
||
|
|
"gt_answers": "several",
|
||
|
|
"image_path": "./data/IAM/j04/j04-039/j04-039-05-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "days",
|
||
|
|
"gt_answers": "days",
|
||
|
|
"image_path": "./data/IAM/j04/j04-039/j04-039-05-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "before",
|
||
|
|
"gt_answers": "before",
|
||
|
|
"image_path": "./data/IAM/j04/j04-039/j04-039-05-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/j04/j04-039/j04-039-05-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word addition",
|
||
|
|
"gt_answers": "addition",
|
||
|
|
"image_path": "./data/IAM/j04/j04-039/j04-039-05-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "f",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/j04/j04-039/j04-039-05-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "lead",
|
||
|
|
"gt_answers": "lead",
|
||
|
|
"image_path": "./data/IAM/j04/j04-039/j04-039-05-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "carrier",
|
||
|
|
"gt_answers": "carrier",
|
||
|
|
"image_path": "./data/IAM/j04/j04-039/j04-039-06-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/j04/j04-039/j04-039-06-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "are covery",
|
||
|
|
"gt_answers": "recovery",
|
||
|
|
"image_path": "./data/IAM/j04/j04-039/j04-039-06-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the letter f",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/j04/j04-039/j04-039-06-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "lead 210",
|
||
|
|
"gt_answers": "lead210",
|
||
|
|
"image_path": "./data/IAM/j04/j04-039/j04-039-06-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "was",
|
||
|
|
"gt_answers": "was",
|
||
|
|
"image_path": "./data/IAM/j04/j04-039/j04-039-06-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "as",
|
||
|
|
"image_path": "./data/IAM/j04/j04-039/j04-039-06-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "much",
|
||
|
|
"gt_answers": "much",
|
||
|
|
"image_path": "./data/IAM/j04/j04-039/j04-039-06-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "as",
|
||
|
|
"image_path": "./data/IAM/j04/j04-039/j04-039-06-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "five",
|
||
|
|
"gt_answers": "five",
|
||
|
|
"image_path": "./data/IAM/j04/j04-039/j04-039-06-10.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word paper",
|
||
|
|
"gt_answers": "per",
|
||
|
|
"image_path": "./data/IAM/j04/j04-039/j04-039-07-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "cent",
|
||
|
|
"gt_answers": "cent",
|
||
|
|
"image_path": "./data/IAM/j04/j04-039/j04-039-07-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "low",
|
||
|
|
"gt_answers": "low",
|
||
|
|
"image_path": "./data/IAM/j04/j04-039/j04-039-07-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "when",
|
||
|
|
"gt_answers": "when",
|
||
|
|
"image_path": "./data/IAM/j04/j04-039/j04-039-07-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "corrected",
|
||
|
|
"gt_answers": "corrected",
|
||
|
|
"image_path": "./data/IAM/j04/j04-039/j04-039-07-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "for",
|
||
|
|
"gt_answers": "for",
|
||
|
|
"image_path": "./data/IAM/j04/j04-039/j04-039-07-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "gravimetric",
|
||
|
|
"gt_answers": "gravimetric",
|
||
|
|
"image_path": "./data/IAM/j04/j04-039/j04-039-07-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "recovery",
|
||
|
|
"gt_answers": "recovery",
|
||
|
|
"image_path": "./data/IAM/j04/j04-039/j04-039-07-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "f",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/j04/j04-039/j04-039-08-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "he",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/j04/j04-039/j04-039-08-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "carrier",
|
||
|
|
"gt_answers": "carrier",
|
||
|
|
"image_path": "./data/IAM/j04/j04-039/j04-039-08-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "mr person",
|
||
|
|
"gt_answers": "Mr",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094a/c03-094a-00-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "william",
|
||
|
|
"gt_answers": "William",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094a/c03-094a-00-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "lucas",
|
||
|
|
"gt_answers": "Lucas",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094a/c03-094a-00-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "morns",
|
||
|
|
"gt_answers": "Morris",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094a/c03-094a-00-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "tis",
|
||
|
|
"gt_answers": "is",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094a/c03-094a-00-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "always",
|
||
|
|
"gt_answers": "always",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094a/c03-094a-00-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "insen si",
|
||
|
|
"gt_answers": "insensi",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094a/c03-094a-00-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "hevely",
|
||
|
|
"gt_answers": "tively",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094a/c03-094a-01-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "pushing",
|
||
|
|
"gt_answers": "pushing",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094a/c03-094a-01-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "miss",
|
||
|
|
"gt_answers": "Miss",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094a/c03-094a-01-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "shela",
|
||
|
|
"gt_answers": "Sheila",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094a/c03-094a-01-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "allen",
|
||
|
|
"gt_answers": "Allen",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094a/c03-094a-01-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "his",
|
||
|
|
"gt_answers": "his",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094a/c03-094a-01-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "wife",
|
||
|
|
"gt_answers": "wife",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094a/c03-094a-01-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "always be yourself",
|
||
|
|
"gt_answers": "always",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094a/c03-094a-02-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "pally",
|
||
|
|
"gt_answers": "palely",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094a/c03-094a-02-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "appalling",
|
||
|
|
"gt_answers": "appealing",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094a/c03-094a-02-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "mr person",
|
||
|
|
"gt_answers": "Mr",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094a/c03-094a-02-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "James",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094a/c03-094a-02-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "maxwell",
|
||
|
|
"gt_answers": "Maxwell",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094a/c03-094a-02-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094a/c03-094a-03-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "editon",
|
||
|
|
"gt_answers": "editor",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094a/c03-094a-03-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word always",
|
||
|
|
"gt_answers": "always",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094a/c03-094a-03-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "comically",
|
||
|
|
"gt_answers": "comically",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094a/c03-094a-03-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "abash",
|
||
|
|
"gt_answers": "abashed",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094a/c03-094a-03-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "by",
|
||
|
|
"gt_answers": "by",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094a/c03-094a-03-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094a/c03-094a-03-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "events",
|
||
|
|
"gt_answers": "events",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094a/c03-094a-04-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word and",
|
||
|
|
"gt_answers": "and",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094a/c03-094a-04-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "mr",
|
||
|
|
"gt_answers": "Mr",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094a/c03-094a-04-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "abby",
|
||
|
|
"gt_answers": "Aubrey",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094a/c03-094a-04-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "richards",
|
||
|
|
"gt_answers": "Richards",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094a/c03-094a-04-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094a/c03-094a-04-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "father in law",
|
||
|
|
"gt_answers": "fatherin",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094a/c03-094a-04-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "law",
|
||
|
|
"gt_answers": "law",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094a/c03-094a-05-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "always love",
|
||
|
|
"gt_answers": "always",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094a/c03-094a-05-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "comically",
|
||
|
|
"gt_answers": "comically",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094a/c03-094a-05-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "grotesque",
|
||
|
|
"gt_answers": "grotesque",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094a/c03-094a-05-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the g",
|
||
|
|
"gt_answers": "they",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094a/c03-094a-05-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "we were",
|
||
|
|
"gt_answers": "were",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094a/c03-094a-05-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "not",
|
||
|
|
"gt_answers": "not",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094a/c03-094a-06-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "asked",
|
||
|
|
"gt_answers": "asked",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094a/c03-094a-06-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "h",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094a/c03-094a-06-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "module lake",
|
||
|
|
"gt_answers": "modulate",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094a/c03-094a-06-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "from",
|
||
|
|
"gt_answers": "from",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094a/c03-094a-06-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the iris",
|
||
|
|
"gt_answers": "their",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094a/c03-094a-06-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "set",
|
||
|
|
"gt_answers": "set",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094a/c03-094a-06-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "moods",
|
||
|
|
"gt_answers": "moods",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094a/c03-094a-06-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "but",
|
||
|
|
"gt_answers": "but",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094a/c03-094a-07-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "played",
|
||
|
|
"gt_answers": "played",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094a/c03-094a-07-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word with",
|
||
|
|
"gt_answers": "with",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094a/c03-094a-07-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "proper",
|
||
|
|
"gt_answers": "proper",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094a/c03-094a-07-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "efficiency",
|
||
|
|
"gt_answers": "efficiency",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094a/c03-094a-07-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "and i",
|
||
|
|
"gt_answers": "and",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094a/c03-094a-07-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "m",
|
||
|
|
"gt_answers": "in",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094a/c03-094a-07-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094a/c03-094a-08-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "case",
|
||
|
|
"gt_answers": "case",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094a/c03-094a-08-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the letter o",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094a/c03-094a-08-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "Mr",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094a/c03-094a-08-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "richards",
|
||
|
|
"gt_answers": "Richards",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094a/c03-094a-08-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "with love",
|
||
|
|
"gt_answers": "with",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094a/c03-094a-08-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "lauish",
|
||
|
|
"gt_answers": "lavish",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094a/c03-094a-08-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "andy",
|
||
|
|
"gt_answers": "and",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094a/c03-094a-08-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "suitably",
|
||
|
|
"gt_answers": "suitably",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094a/c03-094a-09-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "gaudi",
|
||
|
|
"gt_answers": "gaudy",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094a/c03-094a-09-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "colour",
|
||
|
|
"gt_answers": "colour",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094a/c03-094a-09-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "The",
|
||
|
|
"image_path": "./data/IAM/a03/a03-080/a03-080-00-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word person",
|
||
|
|
"gt_answers": "Foreign",
|
||
|
|
"image_path": "./data/IAM/a03/a03-080/a03-080-00-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the name of the person",
|
||
|
|
"gt_answers": "Ministry",
|
||
|
|
"image_path": "./data/IAM/a03/a03-080/a03-080-00-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "i love you",
|
||
|
|
"gt_answers": "spokesman",
|
||
|
|
"image_path": "./data/IAM/a03/a03-080/a03-080-00-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "added",
|
||
|
|
"image_path": "./data/IAM/a03/a03-080/a03-080-00-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "to day",
|
||
|
|
"gt_answers": "today",
|
||
|
|
"image_path": "./data/IAM/a03/a03-080/a03-080-01-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "howells",
|
||
|
|
"gt_answers": "however",
|
||
|
|
"image_path": "./data/IAM/a03/a03-080/a03-080-01-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "that",
|
||
|
|
"image_path": "./data/IAM/a03/a03-080/a03-080-01-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/a03/a03-080/a03-080-01-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "Brussels",
|
||
|
|
"image_path": "./data/IAM/a03/a03-080/a03-080-01-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "tuesday",
|
||
|
|
"gt_answers": "Treaty",
|
||
|
|
"image_path": "./data/IAM/a03/a03-080/a03-080-01-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "was",
|
||
|
|
"image_path": "./data/IAM/a03/a03-080/a03-080-02-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "irrelevant",
|
||
|
|
"image_path": "./data/IAM/a03/a03-080/a03-080-02-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word person",
|
||
|
|
"gt_answers": "because",
|
||
|
|
"image_path": "./data/IAM/a03/a03-080/a03-080-02-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "i love you",
|
||
|
|
"gt_answers": "research",
|
||
|
|
"image_path": "./data/IAM/a03/a03-080/a03-080-02-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "c",
|
||
|
|
"gt_answers": "on",
|
||
|
|
"image_path": "./data/IAM/a03/a03-080/a03-080-02-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/a03/a03-080/a03-080-02-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "various",
|
||
|
|
"image_path": "./data/IAM/a03/a03-080/a03-080-03-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "october",
|
||
|
|
"gt_answers": "rocket",
|
||
|
|
"image_path": "./data/IAM/a03/a03-080/a03-080-03-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word person",
|
||
|
|
"gt_answers": "stages",
|
||
|
|
"image_path": "./data/IAM/a03/a03-080/a03-080-03-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "jah",
|
||
|
|
"gt_answers": "had",
|
||
|
|
"image_path": "./data/IAM/a03/a03-080/a03-080-03-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "awkward",
|
||
|
|
"gt_answers": "already",
|
||
|
|
"image_path": "./data/IAM/a03/a03-080/a03-080-03-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "been",
|
||
|
|
"image_path": "./data/IAM/a03/a03-080/a03-080-03-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "concluded",
|
||
|
|
"image_path": "./data/IAM/a03/a03-080/a03-080-04-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "and",
|
||
|
|
"image_path": "./data/IAM/a03/a03-080/a03-080-04-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word person",
|
||
|
|
"gt_answers": "only",
|
||
|
|
"image_path": "./data/IAM/a03/a03-080/a03-080-04-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word cooper",
|
||
|
|
"gt_answers": "cooperation",
|
||
|
|
"image_path": "./data/IAM/a03/a03-080/a03-080-04-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a snake",
|
||
|
|
"gt_answers": "on",
|
||
|
|
"image_path": "./data/IAM/a03/a03-080/a03-080-04-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/a03/a03-080/a03-080-04-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person and person",
|
||
|
|
"gt_answers": "type",
|
||
|
|
"image_path": "./data/IAM/a03/a03-080/a03-080-04-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/a03/a03-080/a03-080-05-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "capule",
|
||
|
|
"gt_answers": "capsule",
|
||
|
|
"image_path": "./data/IAM/a03/a03-080/a03-080-05-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "was",
|
||
|
|
"image_path": "./data/IAM/a03/a03-080/a03-080-05-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "i love you",
|
||
|
|
"gt_answers": "involved",
|
||
|
|
"image_path": "./data/IAM/a03/a03-080/a03-080-05-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the letter a",
|
||
|
|
"gt_answers": "A",
|
||
|
|
"image_path": "./data/IAM/a03/a03-080/a03-080-05-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "lesser",
|
||
|
|
"image_path": "./data/IAM/a03/a03-080/a03-080-05-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "ollace",
|
||
|
|
"gt_answers": "obstacle",
|
||
|
|
"image_path": "./data/IAM/a03/a03-080/a03-080-05-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the letter a",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/a03/a03-080/a03-080-06-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "German",
|
||
|
|
"image_path": "./data/IAM/a03/a03-080/a03-080-06-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "you have a problem",
|
||
|
|
"gt_answers": "participation",
|
||
|
|
"image_path": "./data/IAM/a03/a03-080/a03-080-06-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "is",
|
||
|
|
"image_path": "./data/IAM/a03/a03-080/a03-080-06-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/a03/a03-080/a03-080-06-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "absence",
|
||
|
|
"image_path": "./data/IAM/a03/a03-080/a03-080-06-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/a03/a03-080/a03-080-06-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "any",
|
||
|
|
"image_path": "./data/IAM/a03/a03-080/a03-080-07-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "Ministry",
|
||
|
|
"image_path": "./data/IAM/a03/a03-080/a03-080-07-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "c",
|
||
|
|
"gt_answers": "or",
|
||
|
|
"image_path": "./data/IAM/a03/a03-080/a03-080-07-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word person",
|
||
|
|
"gt_answers": "Minister",
|
||
|
|
"image_path": "./data/IAM/a03/a03-080/a03-080-07-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the name person",
|
||
|
|
"gt_answers": "directly",
|
||
|
|
"image_path": "./data/IAM/a03/a03-080/a03-080-07-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "i am responsible",
|
||
|
|
"gt_answers": "responsible",
|
||
|
|
"image_path": "./data/IAM/a03/a03-080/a03-080-07-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "f",
|
||
|
|
"gt_answers": "for",
|
||
|
|
"image_path": "./data/IAM/a03/a03-080/a03-080-08-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "it",
|
||
|
|
"image_path": "./data/IAM/a03/a03-080/a03-080-08-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "and",
|
||
|
|
"image_path": "./data/IAM/a03/a03-080/a03-080-08-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "i love you",
|
||
|
|
"gt_answers": "therefore",
|
||
|
|
"image_path": "./data/IAM/a03/a03-080/a03-080-08-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the letter g",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/a03/a03-080/a03-080-08-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "any",
|
||
|
|
"image_path": "./data/IAM/a03/a03-080/a03-080-08-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word bucket",
|
||
|
|
"gt_answers": "budget",
|
||
|
|
"image_path": "./data/IAM/a03/a03-080/a03-080-08-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "under",
|
||
|
|
"image_path": "./data/IAM/a03/a03-080/a03-080-09-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word person",
|
||
|
|
"gt_answers": "which",
|
||
|
|
"image_path": "./data/IAM/a03/a03-080/a03-080-09-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "franz s",
|
||
|
|
"gt_answers": "funds",
|
||
|
|
"image_path": "./data/IAM/a03/a03-080/a03-080-09-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "scull",
|
||
|
|
"gt_answers": "could",
|
||
|
|
"image_path": "./data/IAM/a03/a03-080/a03-080-09-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "l",
|
||
|
|
"gt_answers": "be",
|
||
|
|
"image_path": "./data/IAM/a03/a03-080/a03-080-09-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word person",
|
||
|
|
"gt_answers": "appropriated",
|
||
|
|
"image_path": "./data/IAM/a03/a03-080/a03-080-09-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "mush",
|
||
|
|
"gt_answers": "Much",
|
||
|
|
"image_path": "./data/IAM/r03/r03-115/r03-115-00-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "more",
|
||
|
|
"gt_answers": "more",
|
||
|
|
"image_path": "./data/IAM/r03/r03-115/r03-115-00-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "spaceous",
|
||
|
|
"gt_answers": "spacious",
|
||
|
|
"image_path": "./data/IAM/r03/r03-115/r03-115-00-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "than",
|
||
|
|
"gt_answers": "than",
|
||
|
|
"image_path": "./data/IAM/r03/r03-115/r03-115-00-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "that",
|
||
|
|
"gt_answers": "that",
|
||
|
|
"image_path": "./data/IAM/r03/r03-115/r03-115-00-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "poky",
|
||
|
|
"gt_answers": "poky",
|
||
|
|
"image_path": "./data/IAM/r03/r03-115/r03-115-00-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "little",
|
||
|
|
"gt_answers": "little",
|
||
|
|
"image_path": "./data/IAM/r03/r03-115/r03-115-00-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "cottage",
|
||
|
|
"gt_answers": "cottage",
|
||
|
|
"image_path": "./data/IAM/r03/r03-115/r03-115-00-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "which",
|
||
|
|
"gt_answers": "which",
|
||
|
|
"image_path": "./data/IAM/r03/r03-115/r03-115-01-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "wash it",
|
||
|
|
"gt_answers": "wasnt",
|
||
|
|
"image_path": "./data/IAM/r03/r03-115/r03-115-01-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the number 6",
|
||
|
|
"gt_answers": "a",
|
||
|
|
"image_path": "./data/IAM/r03/r03-115/r03-115-01-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "bt",
|
||
|
|
"gt_answers": "bit",
|
||
|
|
"image_path": "./data/IAM/r03/r03-115/r03-115-01-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "practical",
|
||
|
|
"gt_answers": "practical",
|
||
|
|
"image_path": "./data/IAM/r03/r03-115/r03-115-01-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "realy",
|
||
|
|
"gt_answers": "really",
|
||
|
|
"image_path": "./data/IAM/r03/r03-115/r03-115-01-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/r03/r03-115/r03-115-01-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "reception",
|
||
|
|
"gt_answers": "reception",
|
||
|
|
"image_path": "./data/IAM/r03/r03-115/r03-115-02-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "room",
|
||
|
|
"gt_answers": "room",
|
||
|
|
"image_path": "./data/IAM/r03/r03-115/r03-115-02-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "wa",
|
||
|
|
"gt_answers": "was",
|
||
|
|
"image_path": "./data/IAM/r03/r03-115/r03-115-02-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "ooo",
|
||
|
|
"gt_answers": "too",
|
||
|
|
"image_path": "./data/IAM/r03/r03-115/r03-115-02-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "small",
|
||
|
|
"gt_answers": "small",
|
||
|
|
"image_path": "./data/IAM/r03/r03-115/r03-115-02-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "when",
|
||
|
|
"gt_answers": "when",
|
||
|
|
"image_path": "./data/IAM/r03/r03-115/r03-115-02-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "oddied",
|
||
|
|
"gt_answers": "divided",
|
||
|
|
"image_path": "./data/IAM/r03/r03-115/r03-115-02-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "sand",
|
||
|
|
"gt_answers": "and",
|
||
|
|
"image_path": "./data/IAM/r03/r03-115/r03-115-03-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "toad",
|
||
|
|
"gt_answers": "too",
|
||
|
|
"image_path": "./data/IAM/r03/r03-115/r03-115-03-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "big",
|
||
|
|
"gt_answers": "big",
|
||
|
|
"image_path": "./data/IAM/r03/r03-115/r03-115-03-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "when",
|
||
|
|
"gt_answers": "when",
|
||
|
|
"image_path": "./data/IAM/r03/r03-115/r03-115-03-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "not",
|
||
|
|
"gt_answers": "not",
|
||
|
|
"image_path": "./data/IAM/r03/r03-115/r03-115-03-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "and",
|
||
|
|
"gt_answers": "And",
|
||
|
|
"image_path": "./data/IAM/m01/m01-038/m01-038-00-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "hulk",
|
||
|
|
"gt_answers": "these",
|
||
|
|
"image_path": "./data/IAM/m01/m01-038/m01-038-00-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "lad",
|
||
|
|
"gt_answers": "had",
|
||
|
|
"image_path": "./data/IAM/m01/m01-038/m01-038-00-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "been",
|
||
|
|
"image_path": "./data/IAM/m01/m01-038/m01-038-00-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word wis",
|
||
|
|
"gt_answers": "his",
|
||
|
|
"image_path": "./data/IAM/m01/m01-038/m01-038-00-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word fear",
|
||
|
|
"gt_answers": "great",
|
||
|
|
"image_path": "./data/IAM/m01/m01-038/m01-038-00-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "guillaume",
|
||
|
|
"gt_answers": "qualities",
|
||
|
|
"image_path": "./data/IAM/m01/m01-038/m01-038-00-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "h",
|
||
|
|
"gt_answers": "so",
|
||
|
|
"image_path": "./data/IAM/m01/m01-038/m01-038-01-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a hacklive",
|
||
|
|
"gt_answers": "attractive",
|
||
|
|
"image_path": "./data/IAM/m01/m01-038/m01-038-01-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "ho",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/m01/m01-038/m01-038-01-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "her",
|
||
|
|
"gt_answers": "her",
|
||
|
|
"image_path": "./data/IAM/m01/m01-038/m01-038-01-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "The",
|
||
|
|
"image_path": "./data/IAM/m01/m01-038/m01-038-01-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word bar",
|
||
|
|
"gt_answers": "Base",
|
||
|
|
"image_path": "./data/IAM/m01/m01-038/m01-038-01-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "shon",
|
||
|
|
"gt_answers": "Station",
|
||
|
|
"image_path": "./data/IAM/m01/m01-038/m01-038-01-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "lad",
|
||
|
|
"gt_answers": "had",
|
||
|
|
"image_path": "./data/IAM/m01/m01-038/m01-038-01-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "joe",
|
||
|
|
"gt_answers": "gone",
|
||
|
|
"image_path": "./data/IAM/m01/m01-038/m01-038-02-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "be free",
|
||
|
|
"gt_answers": "before",
|
||
|
|
"image_path": "./data/IAM/m01/m01-038/m01-038-02-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "discharging",
|
||
|
|
"gt_answers": "discharging",
|
||
|
|
"image_path": "./data/IAM/m01/m01-038/m01-038-02-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "h",
|
||
|
|
"gt_answers": "its",
|
||
|
|
"image_path": "./data/IAM/m01/m01-038/m01-038-02-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "lemon",
|
||
|
|
"gt_answers": "narrow",
|
||
|
|
"image_path": "./data/IAM/m01/m01-038/m01-038-02-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "please",
|
||
|
|
"gt_answers": "plane",
|
||
|
|
"image_path": "./data/IAM/m01/m01-038/m01-038-02-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "f",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/m01/m01-038/m01-038-02-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "4ementalating",
|
||
|
|
"image_path": "./data/IAM/m01/m01-038/m01-038-03-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "energy",
|
||
|
|
"gt_answers": "energy",
|
||
|
|
"image_path": "./data/IAM/m01/m01-038/m01-038-03-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "along",
|
||
|
|
"gt_answers": "along",
|
||
|
|
"image_path": "./data/IAM/m01/m01-038/m01-038-03-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "he",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/m01/m01-038/m01-038-03-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "lewis carroll",
|
||
|
|
"gt_answers": "length",
|
||
|
|
"image_path": "./data/IAM/m01/m01-038/m01-038-03-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a cross",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/m01/m01-038/m01-038-03-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "me",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/m01/m01-038/m01-038-04-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "early's",
|
||
|
|
"gt_answers": "Earths",
|
||
|
|
"image_path": "./data/IAM/m01/m01-038/m01-038-04-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "imaginary",
|
||
|
|
"gt_answers": "imaginary",
|
||
|
|
"image_path": "./data/IAM/m01/m01-038/m01-038-04-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "louis hinds",
|
||
|
|
"gt_answers": "longitudes",
|
||
|
|
"image_path": "./data/IAM/m01/m01-038/m01-038-04-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word moving",
|
||
|
|
"gt_answers": "moving",
|
||
|
|
"image_path": "./data/IAM/m01/m01-038/m01-038-04-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "lashward",
|
||
|
|
"gt_answers": "eastward",
|
||
|
|
"image_path": "./data/IAM/m01/m01-038/m01-038-05-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "like",
|
||
|
|
"gt_answers": "like",
|
||
|
|
"image_path": "./data/IAM/m01/m01-038/m01-038-05-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "a",
|
||
|
|
"image_path": "./data/IAM/m01/m01-038/m01-038-05-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "knife - edged",
|
||
|
|
"gt_answers": "knifeedged",
|
||
|
|
"image_path": "./data/IAM/m01/m01-038/m01-038-05-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "l h",
|
||
|
|
"gt_answers": "twilight",
|
||
|
|
"image_path": "./data/IAM/m01/m01-038/m01-038-05-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a worm",
|
||
|
|
"gt_answers": "in",
|
||
|
|
"image_path": "./data/IAM/m01/m01-038/m01-038-05-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "redeemed",
|
||
|
|
"gt_answers": "reversed",
|
||
|
|
"image_path": "./data/IAM/m01/m01-038/m01-038-06-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "margaret",
|
||
|
|
"gt_answers": "progression",
|
||
|
|
"image_path": "./data/IAM/m01/m01-038/m01-038-06-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "It",
|
||
|
|
"image_path": "./data/IAM/m01/m01-038/m01-038-06-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "helen",
|
||
|
|
"gt_answers": "had",
|
||
|
|
"image_path": "./data/IAM/m01/m01-038/m01-038-06-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "harry truman",
|
||
|
|
"gt_answers": "brought",
|
||
|
|
"image_path": "./data/IAM/m01/m01-038/m01-038-06-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "ne",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/m01/m01-038/m01-038-06-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "greek",
|
||
|
|
"gt_answers": "First",
|
||
|
|
"image_path": "./data/IAM/m01/m01-038/m01-038-07-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "shye",
|
||
|
|
"gt_answers": "Stage",
|
||
|
|
"image_path": "./data/IAM/m01/m01-038/m01-038-07-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "t",
|
||
|
|
"gt_answers": "in",
|
||
|
|
"image_path": "./data/IAM/m01/m01-038/m01-038-07-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "he",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/m01/m01-038/m01-038-07-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the bar",
|
||
|
|
"gt_answers": "Thetan",
|
||
|
|
"image_path": "./data/IAM/m01/m01-038/m01-038-07-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "paul henry",
|
||
|
|
"gt_answers": "pattern",
|
||
|
|
"image_path": "./data/IAM/m01/m01-038/m01-038-07-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "f",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/m01/m01-038/m01-038-07-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "corguell",
|
||
|
|
"gt_answers": "conquest",
|
||
|
|
"image_path": "./data/IAM/m01/m01-038/m01-038-07-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/m01/m01-038/m01-038-08-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word hysterical",
|
||
|
|
"gt_answers": "emptyminded",
|
||
|
|
"image_path": "./data/IAM/m01/m01-038/m01-038-08-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word person",
|
||
|
|
"gt_answers": "receptiveness",
|
||
|
|
"image_path": "./data/IAM/m01/m01-038/m01-038-08-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "m l",
|
||
|
|
"gt_answers": "that",
|
||
|
|
"image_path": "./data/IAM/m01/m01-038/m01-038-08-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word prepared",
|
||
|
|
"gt_answers": "prepared",
|
||
|
|
"image_path": "./data/IAM/m01/m01-038/m01-038-09-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "me",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/m01/m01-038/m01-038-09-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "ray",
|
||
|
|
"gt_answers": "way",
|
||
|
|
"image_path": "./data/IAM/m01/m01-038/m01-038-09-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "f",
|
||
|
|
"gt_answers": "for",
|
||
|
|
"image_path": "./data/IAM/m01/m01-038/m01-038-09-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "me",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/m01/m01-038/m01-038-09-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "leonard",
|
||
|
|
"gt_answers": "Second",
|
||
|
|
"image_path": "./data/IAM/m01/m01-038/m01-038-09-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "shae",
|
||
|
|
"gt_answers": "Stage",
|
||
|
|
"image_path": "./data/IAM/m01/m01-038/m01-038-09-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "e",
|
||
|
|
"gt_answers": "He",
|
||
|
|
"image_path": "./data/IAM/n03/n03-038/n03-038-00-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word h",
|
||
|
|
"gt_answers": "had",
|
||
|
|
"image_path": "./data/IAM/n03/n03-038/n03-038-00-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "almost",
|
||
|
|
"image_path": "./data/IAM/n03/n03-038/n03-038-00-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word enoch",
|
||
|
|
"gt_answers": "entirely",
|
||
|
|
"image_path": "./data/IAM/n03/n03-038/n03-038-00-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word sheld",
|
||
|
|
"gt_answers": "settled",
|
||
|
|
"image_path": "./data/IAM/n03/n03-038/n03-038-00-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "back",
|
||
|
|
"image_path": "./data/IAM/n03/n03-038/n03-038-00-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "into",
|
||
|
|
"image_path": "./data/IAM/n03/n03-038/n03-038-00-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the name of person",
|
||
|
|
"gt_answers": "his",
|
||
|
|
"image_path": "./data/IAM/n03/n03-038/n03-038-00-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "ol",
|
||
|
|
"gt_answers": "old",
|
||
|
|
"image_path": "./data/IAM/n03/n03-038/n03-038-00-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "ways",
|
||
|
|
"gt_answers": "ways",
|
||
|
|
"image_path": "./data/IAM/n03/n03-038/n03-038-01-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "eve",
|
||
|
|
"gt_answers": "even",
|
||
|
|
"image_path": "./data/IAM/n03/n03-038/n03-038-01-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a b c",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/n03/n03-038/n03-038-01-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "reading",
|
||
|
|
"gt_answers": "reading",
|
||
|
|
"image_path": "./data/IAM/n03/n03-038/n03-038-01-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word love",
|
||
|
|
"gt_answers": "law",
|
||
|
|
"image_path": "./data/IAM/n03/n03-038/n03-038-01-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "about person",
|
||
|
|
"gt_answers": "about",
|
||
|
|
"image_path": "./data/IAM/n03/n03-038/n03-038-01-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word no",
|
||
|
|
"gt_answers": "two",
|
||
|
|
"image_path": "./data/IAM/n03/n03-038/n03-038-01-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "louis",
|
||
|
|
"gt_answers": "hours",
|
||
|
|
"image_path": "./data/IAM/n03/n03-038/n03-038-01-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "each",
|
||
|
|
"image_path": "./data/IAM/n03/n03-038/n03-038-01-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "olay",
|
||
|
|
"gt_answers": "day",
|
||
|
|
"image_path": "./data/IAM/n03/n03-038/n03-038-02-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "through faith",
|
||
|
|
"gt_answers": "Though",
|
||
|
|
"image_path": "./data/IAM/n03/n03-038/n03-038-02-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "e",
|
||
|
|
"gt_answers": "he",
|
||
|
|
"image_path": "./data/IAM/n03/n03-038/n03-038-02-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word person",
|
||
|
|
"gt_answers": "did",
|
||
|
|
"image_path": "./data/IAM/n03/n03-038/n03-038-02-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "not",
|
||
|
|
"image_path": "./data/IAM/n03/n03-038/n03-038-02-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "refuse",
|
||
|
|
"gt_answers": "refuse",
|
||
|
|
"image_path": "./data/IAM/n03/n03-038/n03-038-02-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "wine",
|
||
|
|
"gt_answers": "wine",
|
||
|
|
"image_path": "./data/IAM/n03/n03-038/n03-038-02-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the name of the person",
|
||
|
|
"gt_answers": "which",
|
||
|
|
"image_path": "./data/IAM/n03/n03-038/n03-038-02-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the name person",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/n03/n03-038/n03-038-02-10.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "judge",
|
||
|
|
"gt_answers": "judge",
|
||
|
|
"image_path": "./data/IAM/n03/n03-038/n03-038-03-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word always",
|
||
|
|
"gt_answers": "always",
|
||
|
|
"image_path": "./data/IAM/n03/n03-038/n03-038-03-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "had",
|
||
|
|
"image_path": "./data/IAM/n03/n03-038/n03-038-03-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a signature",
|
||
|
|
"gt_answers": "on",
|
||
|
|
"image_path": "./data/IAM/n03/n03-038/n03-038-03-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the name person",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/n03/n03-038/n03-038-03-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "kase",
|
||
|
|
"gt_answers": "table",
|
||
|
|
"image_path": "./data/IAM/n03/n03-038/n03-038-03-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "at",
|
||
|
|
"image_path": "./data/IAM/n03/n03-038/n03-038-03-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "dinner",
|
||
|
|
"image_path": "./data/IAM/n03/n03-038/n03-038-03-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "reyhe",
|
||
|
|
"gt_answers": "he",
|
||
|
|
"image_path": "./data/IAM/n03/n03-038/n03-038-03-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the name person",
|
||
|
|
"gt_answers": "did",
|
||
|
|
"image_path": "./data/IAM/n03/n03-038/n03-038-04-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "not",
|
||
|
|
"image_path": "./data/IAM/n03/n03-038/n03-038-04-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "act i",
|
||
|
|
"gt_answers": "act",
|
||
|
|
"image_path": "./data/IAM/n03/n03-038/n03-038-04-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a s",
|
||
|
|
"gt_answers": "as",
|
||
|
|
"image_path": "./data/IAM/n03/n03-038/n03-038-04-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "thought",
|
||
|
|
"gt_answers": "though",
|
||
|
|
"image_path": "./data/IAM/n03/n03-038/n03-038-04-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "he",
|
||
|
|
"gt_answers": "he",
|
||
|
|
"image_path": "./data/IAM/n03/n03-038/n03-038-04-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the found",
|
||
|
|
"gt_answers": "found",
|
||
|
|
"image_path": "./data/IAM/n03/n03-038/n03-038-04-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "it",
|
||
|
|
"image_path": "./data/IAM/n03/n03-038/n03-038-04-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "meesey",
|
||
|
|
"gt_answers": "necessary",
|
||
|
|
"image_path": "./data/IAM/n03/n03-038/n03-038-04-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "exept",
|
||
|
|
"gt_answers": "except",
|
||
|
|
"image_path": "./data/IAM/n03/n03-038/n03-038-05-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "in",
|
||
|
|
"image_path": "./data/IAM/n03/n03-038/n03-038-05-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "make sale",
|
||
|
|
"gt_answers": "moderate",
|
||
|
|
"image_path": "./data/IAM/n03/n03-038/n03-038-05-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "compliance",
|
||
|
|
"gt_answers": "compliance",
|
||
|
|
"image_path": "./data/IAM/n03/n03-038/n03-038-05-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "with",
|
||
|
|
"image_path": "./data/IAM/n03/n03-038/n03-038-05-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "his",
|
||
|
|
"image_path": "./data/IAM/n03/n03-038/n03-038-05-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "lis fathers",
|
||
|
|
"gt_answers": "fathers",
|
||
|
|
"image_path": "./data/IAM/n03/n03-038/n03-038-05-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the name of person",
|
||
|
|
"gt_answers": "habit",
|
||
|
|
"image_path": "./data/IAM/n03/n03-038/n03-038-06-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "kitty",
|
||
|
|
"gt_answers": "Kitty",
|
||
|
|
"image_path": "./data/IAM/l04/l04-043/l04-043-00-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "courage",
|
||
|
|
"gt_answers": "encourages",
|
||
|
|
"image_path": "./data/IAM/l04/l04-043/l04-043-00-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "hi",
|
||
|
|
"gt_answers": "him",
|
||
|
|
"image_path": "./data/IAM/l04/l04-043/l04-043-00-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word 100",
|
||
|
|
"gt_answers": "too",
|
||
|
|
"image_path": "./data/IAM/l04/l04-043/l04-043-00-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "no",
|
||
|
|
"gt_answers": "No",
|
||
|
|
"image_path": "./data/IAM/l04/l04-043/l04-043-00-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word show",
|
||
|
|
"gt_answers": "work",
|
||
|
|
"image_path": "./data/IAM/l04/l04-043/l04-043-00-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "no",
|
||
|
|
"gt_answers": "no",
|
||
|
|
"image_path": "./data/IAM/l04/l04-043/l04-043-00-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "background",
|
||
|
|
"gt_answers": "background",
|
||
|
|
"image_path": "./data/IAM/l04/l04-043/l04-043-01-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "no",
|
||
|
|
"gt_answers": "no",
|
||
|
|
"image_path": "./data/IAM/l04/l04-043/l04-043-01-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "regular font",
|
||
|
|
"gt_answers": "regular",
|
||
|
|
"image_path": "./data/IAM/l04/l04-043/l04-043-01-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "money",
|
||
|
|
"gt_answers": "money",
|
||
|
|
"image_path": "./data/IAM/l04/l04-043/l04-043-01-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "so",
|
||
|
|
"gt_answers": "so",
|
||
|
|
"image_path": "./data/IAM/l04/l04-043/l04-043-01-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "lar",
|
||
|
|
"gt_answers": "far",
|
||
|
|
"image_path": "./data/IAM/l04/l04-043/l04-043-01-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "s",
|
||
|
|
"gt_answers": "as",
|
||
|
|
"image_path": "./data/IAM/l04/l04-043/l04-043-01-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "I",
|
||
|
|
"image_path": "./data/IAM/l04/l04-043/l04-043-01-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "c o",
|
||
|
|
"gt_answers": "can",
|
||
|
|
"image_path": "./data/IAM/l04/l04-043/l04-043-02-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "water",
|
||
|
|
"gt_answers": "make",
|
||
|
|
"image_path": "./data/IAM/l04/l04-043/l04-043-02-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "oo",
|
||
|
|
"gt_answers": "out",
|
||
|
|
"image_path": "./data/IAM/l04/l04-043/l04-043-02-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "i probably",
|
||
|
|
"gt_answers": "Probably",
|
||
|
|
"image_path": "./data/IAM/l04/l04-043/l04-043-02-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "o",
|
||
|
|
"gt_answers": "on",
|
||
|
|
"image_path": "./data/IAM/l04/l04-043/l04-043-02-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/l04/l04-043/l04-043-02-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "croo",
|
||
|
|
"gt_answers": "crook",
|
||
|
|
"image_path": "./data/IAM/l04/l04-043/l04-043-02-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "butt",
|
||
|
|
"gt_answers": "But",
|
||
|
|
"image_path": "./data/IAM/l04/l04-043/l04-043-03-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "there is a god",
|
||
|
|
"gt_answers": "there",
|
||
|
|
"image_path": "./data/IAM/l04/l04-043/l04-043-03-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "you",
|
||
|
|
"gt_answers": "you",
|
||
|
|
"image_path": "./data/IAM/l04/l04-043/l04-043-03-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "gre",
|
||
|
|
"gt_answers": "are",
|
||
|
|
"image_path": "./data/IAM/l04/l04-043/l04-043-03-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/l04/l04-043/l04-043-03-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "girl",
|
||
|
|
"gt_answers": "girl",
|
||
|
|
"image_path": "./data/IAM/l04/l04-043/l04-043-03-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "s",
|
||
|
|
"gt_answers": "s",
|
||
|
|
"image_path": "./data/IAM/l04/l04-043/l04-043-03-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "wild",
|
||
|
|
"gt_answers": "wild",
|
||
|
|
"image_path": "./data/IAM/l04/l04-043/l04-043-03-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word and",
|
||
|
|
"gt_answers": "and",
|
||
|
|
"image_path": "./data/IAM/l04/l04-043/l04-043-03-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "headstrong",
|
||
|
|
"gt_answers": "headstrong",
|
||
|
|
"image_path": "./data/IAM/l04/l04-043/l04-043-04-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "I",
|
||
|
|
"image_path": "./data/IAM/l04/l04-043/l04-043-04-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "con",
|
||
|
|
"gt_answers": "cant",
|
||
|
|
"image_path": "./data/IAM/l04/l04-043/l04-043-04-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "oo",
|
||
|
|
"gt_answers": "do",
|
||
|
|
"image_path": "./data/IAM/l04/l04-043/l04-043-04-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "nothing",
|
||
|
|
"gt_answers": "nothing",
|
||
|
|
"image_path": "./data/IAM/l04/l04-043/l04-043-04-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "th",
|
||
|
|
"gt_answers": "with",
|
||
|
|
"image_path": "./data/IAM/l04/l04-043/l04-043-04-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "ev",
|
||
|
|
"gt_answers": "er",
|
||
|
|
"image_path": "./data/IAM/l04/l04-043/l04-043-04-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "don't",
|
||
|
|
"gt_answers": "Dont",
|
||
|
|
"image_path": "./data/IAM/l04/l04-043/l04-043-05-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "m o",
|
||
|
|
"gt_answers": "worry",
|
||
|
|
"image_path": "./data/IAM/l04/l04-043/l04-043-05-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "Harry",
|
||
|
|
"image_path": "./data/IAM/l04/l04-043/l04-043-05-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/l04/l04-043/l04-043-05-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "mother",
|
||
|
|
"gt_answers": "other",
|
||
|
|
"image_path": "./data/IAM/l04/l04-043/l04-043-05-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word solid",
|
||
|
|
"gt_answers": "said",
|
||
|
|
"image_path": "./data/IAM/l04/l04-043/l04-043-05-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "quietly",
|
||
|
|
"gt_answers": "quietly",
|
||
|
|
"image_path": "./data/IAM/l04/l04-043/l04-043-05-10.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "I",
|
||
|
|
"image_path": "./data/IAM/l04/l04-043/l04-043-06-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word now",
|
||
|
|
"gt_answers": "wont",
|
||
|
|
"image_path": "./data/IAM/l04/l04-043/l04-043-06-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "water",
|
||
|
|
"gt_answers": "make",
|
||
|
|
"image_path": "./data/IAM/l04/l04-043/l04-043-06-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "i love you",
|
||
|
|
"gt_answers": "a",
|
||
|
|
"image_path": "./data/IAM/l04/l04-043/l04-043-06-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "fool",
|
||
|
|
"gt_answers": "fool",
|
||
|
|
"image_path": "./data/IAM/l04/l04-043/l04-043-06-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the letter c",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/l04/l04-043/l04-043-06-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "self love",
|
||
|
|
"gt_answers": "myself",
|
||
|
|
"image_path": "./data/IAM/l04/l04-043/l04-043-06-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "hitty",
|
||
|
|
"gt_answers": "Kitty",
|
||
|
|
"image_path": "./data/IAM/l04/l04-043/l04-043-07-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "needs",
|
||
|
|
"gt_answers": "needs",
|
||
|
|
"image_path": "./data/IAM/l04/l04-043/l04-043-07-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "i love you",
|
||
|
|
"gt_answers": "a",
|
||
|
|
"image_path": "./data/IAM/l04/l04-043/l04-043-07-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "good",
|
||
|
|
"gt_answers": "good",
|
||
|
|
"image_path": "./data/IAM/l04/l04-043/l04-043-07-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "friend",
|
||
|
|
"gt_answers": "friend",
|
||
|
|
"image_path": "./data/IAM/l04/l04-043/l04-043-07-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "yo",
|
||
|
|
"gt_answers": "and",
|
||
|
|
"image_path": "./data/IAM/l04/l04-043/l04-043-07-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "d",
|
||
|
|
"gt_answers": "I",
|
||
|
|
"image_path": "./data/IAM/l04/l04-043/l04-043-07-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the number 11",
|
||
|
|
"gt_answers": "ll",
|
||
|
|
"image_path": "./data/IAM/l04/l04-043/l04-043-07-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "always",
|
||
|
|
"gt_answers": "always",
|
||
|
|
"image_path": "./data/IAM/l04/l04-043/l04-043-08-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "be",
|
||
|
|
"gt_answers": "be",
|
||
|
|
"image_path": "./data/IAM/l04/l04-043/l04-043-08-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "hudson",
|
||
|
|
"gt_answers": "that",
|
||
|
|
"image_path": "./data/IAM/l04/l04-043/l04-043-08-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "b",
|
||
|
|
"gt_answers": "But",
|
||
|
|
"image_path": "./data/IAM/f03/f03-182/f03-182-00-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "that",
|
||
|
|
"image_path": "./data/IAM/f03/f03-182/f03-182-00-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a",
|
||
|
|
"gt_answers": "s",
|
||
|
|
"image_path": "./data/IAM/f03/f03-182/f03-182-00-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "a",
|
||
|
|
"image_path": "./data/IAM/f03/f03-182/f03-182-00-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the name of person",
|
||
|
|
"gt_answers": "terrible",
|
||
|
|
"image_path": "./data/IAM/f03/f03-182/f03-182-00-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "may",
|
||
|
|
"gt_answers": "way",
|
||
|
|
"image_path": "./data/IAM/f03/f03-182/f03-182-00-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/f03/f03-182/f03-182-00-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "lee begin",
|
||
|
|
"gt_answers": "begin",
|
||
|
|
"image_path": "./data/IAM/f03/f03-182/f03-182-00-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "m h",
|
||
|
|
"gt_answers": "Why",
|
||
|
|
"image_path": "./data/IAM/f03/f03-182/f03-182-00-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "hell",
|
||
|
|
"gt_answers": "tell",
|
||
|
|
"image_path": "./data/IAM/f03/f03-182/f03-182-00-10.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word hein",
|
||
|
|
"gt_answers": "them",
|
||
|
|
"image_path": "./data/IAM/f03/f03-182/f03-182-00-11.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "you",
|
||
|
|
"gt_answers": "you",
|
||
|
|
"image_path": "./data/IAM/f03/f03-182/f03-182-00-12.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "e",
|
||
|
|
"gt_answers": "are",
|
||
|
|
"image_path": "./data/IAM/f03/f03-182/f03-182-01-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a",
|
||
|
|
"gt_answers": "a",
|
||
|
|
"image_path": "./data/IAM/f03/f03-182/f03-182-01-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word notice",
|
||
|
|
"gt_answers": "novice",
|
||
|
|
"image_path": "./data/IAM/f03/f03-182/f03-182-01-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "m",
|
||
|
|
"gt_answers": "It",
|
||
|
|
"image_path": "./data/IAM/f03/f03-182/f03-182-01-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "s",
|
||
|
|
"image_path": "./data/IAM/f03/f03-182/f03-182-01-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "mer",
|
||
|
|
"gt_answers": "their",
|
||
|
|
"image_path": "./data/IAM/f03/f03-182/f03-182-01-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "inkest",
|
||
|
|
"gt_answers": "interest",
|
||
|
|
"image_path": "./data/IAM/f03/f03-182/f03-182-01-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "you",
|
||
|
|
"gt_answers": "you",
|
||
|
|
"image_path": "./data/IAM/f03/f03-182/f03-182-01-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "want",
|
||
|
|
"image_path": "./data/IAM/f03/f03-182/f03-182-01-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "no",
|
||
|
|
"gt_answers": "not",
|
||
|
|
"image_path": "./data/IAM/f03/f03-182/f03-182-01-11.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "heur",
|
||
|
|
"gt_answers": "their",
|
||
|
|
"image_path": "./data/IAM/f03/f03-182/f03-182-01-12.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "sympathy",
|
||
|
|
"gt_answers": "sympathy",
|
||
|
|
"image_path": "./data/IAM/f03/f03-182/f03-182-02-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a d",
|
||
|
|
"gt_answers": "And",
|
||
|
|
"image_path": "./data/IAM/f03/f03-182/f03-182-02-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "you",
|
||
|
|
"gt_answers": "you",
|
||
|
|
"image_path": "./data/IAM/f03/f03-182/f03-182-02-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the name person",
|
||
|
|
"gt_answers": "want",
|
||
|
|
"image_path": "./data/IAM/f03/f03-182/f03-182-02-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "k",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/f03/f03-182/f03-182-02-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "g e k",
|
||
|
|
"gt_answers": "get",
|
||
|
|
"image_path": "./data/IAM/f03/f03-182/f03-182-02-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a m",
|
||
|
|
"gt_answers": "it",
|
||
|
|
"image_path": "./data/IAM/f03/f03-182/f03-182-02-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "from person",
|
||
|
|
"gt_answers": "from",
|
||
|
|
"image_path": "./data/IAM/f03/f03-182/f03-182-02-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "he",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/f03/f03-182/f03-182-02-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "word",
|
||
|
|
"image_path": "./data/IAM/f03/f03-182/f03-182-02-10.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "go",
|
||
|
|
"gt_answers": "GO",
|
||
|
|
"image_path": "./data/IAM/f03/f03-182/f03-182-03-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "try",
|
||
|
|
"gt_answers": "Try",
|
||
|
|
"image_path": "./data/IAM/f03/f03-182/f03-182-03-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "camelng",
|
||
|
|
"gt_answers": "something",
|
||
|
|
"image_path": "./data/IAM/f03/f03-182/f03-182-03-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "luke",
|
||
|
|
"gt_answers": "like",
|
||
|
|
"image_path": "./data/IAM/f03/f03-182/f03-182-03-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "this",
|
||
|
|
"image_path": "./data/IAM/f03/f03-182/f03-182-03-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "k",
|
||
|
|
"gt_answers": "It",
|
||
|
|
"image_path": "./data/IAM/f03/f03-182/f03-182-03-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "is",
|
||
|
|
"image_path": "./data/IAM/f03/f03-182/f03-182-03-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word person",
|
||
|
|
"gt_answers": "said",
|
||
|
|
"image_path": "./data/IAM/f03/f03-182/f03-182-03-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "mica",
|
||
|
|
"gt_answers": "that",
|
||
|
|
"image_path": "./data/IAM/f03/f03-182/f03-182-03-10.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the name person",
|
||
|
|
"gt_answers": "tele",
|
||
|
|
"image_path": "./data/IAM/f03/f03-182/f03-182-03-11.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word person",
|
||
|
|
"gt_answers": "vision",
|
||
|
|
"image_path": "./data/IAM/f03/f03-182/f03-182-04-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "keep calm",
|
||
|
|
"gt_answers": "keeps",
|
||
|
|
"image_path": "./data/IAM/f03/f03-182/f03-182-04-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "people",
|
||
|
|
"gt_answers": "people",
|
||
|
|
"image_path": "./data/IAM/f03/f03-182/f03-182-04-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "c h",
|
||
|
|
"gt_answers": "at",
|
||
|
|
"image_path": "./data/IAM/f03/f03-182/f03-182-04-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "home",
|
||
|
|
"gt_answers": "home",
|
||
|
|
"image_path": "./data/IAM/f03/f03-182/f03-182-04-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "b",
|
||
|
|
"gt_answers": "But",
|
||
|
|
"image_path": "./data/IAM/f03/f03-182/f03-182-04-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "you",
|
||
|
|
"gt_answers": "you",
|
||
|
|
"image_path": "./data/IAM/f03/f03-182/f03-182-04-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the letter a",
|
||
|
|
"gt_answers": "at",
|
||
|
|
"image_path": "./data/IAM/f03/f03-182/f03-182-04-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "amy",
|
||
|
|
"gt_answers": "any",
|
||
|
|
"image_path": "./data/IAM/f03/f03-182/f03-182-04-10.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word mark",
|
||
|
|
"gt_answers": "rate",
|
||
|
|
"image_path": "./data/IAM/f03/f03-182/f03-182-05-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the name of person",
|
||
|
|
"gt_answers": "have",
|
||
|
|
"image_path": "./data/IAM/f03/f03-182/f03-182-05-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word proved",
|
||
|
|
"gt_answers": "proved",
|
||
|
|
"image_path": "./data/IAM/f03/f03-182/f03-182-05-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "michael",
|
||
|
|
"gt_answers": "that",
|
||
|
|
"image_path": "./data/IAM/f03/f03-182/f03-182-05-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "wrong",
|
||
|
|
"image_path": "./data/IAM/f03/f03-182/f03-182-05-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "ald",
|
||
|
|
"gt_answers": "And",
|
||
|
|
"image_path": "./data/IAM/f03/f03-182/f03-182-05-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "hy",
|
||
|
|
"gt_answers": "they",
|
||
|
|
"image_path": "./data/IAM/f03/f03-182/f03-182-05-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the name of person",
|
||
|
|
"gt_answers": "say",
|
||
|
|
"image_path": "./data/IAM/f03/f03-182/f03-182-05-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "too",
|
||
|
|
"image_path": "./data/IAM/f03/f03-182/f03-182-05-11.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "ah",
|
||
|
|
"gt_answers": "that",
|
||
|
|
"image_path": "./data/IAM/f03/f03-182/f03-182-06-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the name of the person",
|
||
|
|
"gt_answers": "television",
|
||
|
|
"image_path": "./data/IAM/f03/f03-182/f03-182-06-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word makes",
|
||
|
|
"gt_answers": "makes",
|
||
|
|
"image_path": "./data/IAM/f03/f03-182/f03-182-06-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "i love you",
|
||
|
|
"gt_answers": "its",
|
||
|
|
"image_path": "./data/IAM/f03/f03-182/f03-182-06-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word appeal",
|
||
|
|
"gt_answers": "appeal",
|
||
|
|
"image_path": "./data/IAM/f03/f03-182/f03-182-06-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/f03/f03-182/f03-182-06-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "thee",
|
||
|
|
"gt_answers": "those",
|
||
|
|
"image_path": "./data/IAM/f03/f03-182/f03-182-06-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "o",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/f03/f03-182/f03-182-06-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "lenor",
|
||
|
|
"gt_answers": "lesser",
|
||
|
|
"image_path": "./data/IAM/f03/f03-182/f03-182-06-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "intelligence",
|
||
|
|
"gt_answers": "intelligence",
|
||
|
|
"image_path": "./data/IAM/f03/f03-182/f03-182-07-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "h",
|
||
|
|
"gt_answers": "Mr",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094f/c03-094f-00-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "william",
|
||
|
|
"gt_answers": "William",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094f/c03-094f-00-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "lucas",
|
||
|
|
"gt_answers": "Lucas",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094f/c03-094f-00-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word person",
|
||
|
|
"gt_answers": "Morris",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094f/c03-094f-00-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "is",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094f/c03-094f-00-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "always",
|
||
|
|
"gt_answers": "always",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094f/c03-094f-00-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "insensi",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094f/c03-094f-00-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "fidley",
|
||
|
|
"gt_answers": "tively",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094f/c03-094f-01-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "pushing",
|
||
|
|
"gt_answers": "pushing",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094f/c03-094f-01-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word kiss",
|
||
|
|
"gt_answers": "Miss",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094f/c03-094f-01-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "sheila",
|
||
|
|
"gt_answers": "Sheila",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094f/c03-094f-01-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "allen",
|
||
|
|
"gt_answers": "Allen",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094f/c03-094f-01-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "his",
|
||
|
|
"gt_answers": "his",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094f/c03-094f-01-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "wife",
|
||
|
|
"gt_answers": "wife",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094f/c03-094f-01-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "always",
|
||
|
|
"gt_answers": "always",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094f/c03-094f-02-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "palley",
|
||
|
|
"gt_answers": "palely",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094f/c03-094f-02-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "appalling",
|
||
|
|
"gt_answers": "appealing",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094f/c03-094f-02-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "h",
|
||
|
|
"gt_answers": "Mr",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094f/c03-094f-02-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "james",
|
||
|
|
"gt_answers": "James",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094f/c03-094f-02-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "haxwell",
|
||
|
|
"gt_answers": "Maxwell",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094f/c03-094f-02-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the one",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094f/c03-094f-03-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "editor",
|
||
|
|
"gt_answers": "editor",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094f/c03-094f-03-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "always",
|
||
|
|
"gt_answers": "always",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094f/c03-094f-03-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "comically",
|
||
|
|
"gt_answers": "comically",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094f/c03-094f-03-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "abhshed",
|
||
|
|
"gt_answers": "abashed",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094f/c03-094f-03-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "by",
|
||
|
|
"gt_answers": "by",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094f/c03-094f-03-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094f/c03-094f-04-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "events",
|
||
|
|
"gt_answers": "events",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094f/c03-094f-04-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "bd",
|
||
|
|
"gt_answers": "and",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094f/c03-094f-04-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "h a",
|
||
|
|
"gt_answers": "Mr",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094f/c03-094f-04-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the name of the person",
|
||
|
|
"gt_answers": "Aubrey",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094f/c03-094f-04-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "richards",
|
||
|
|
"gt_answers": "Richards",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094f/c03-094f-04-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094f/c03-094f-04-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "father in law",
|
||
|
|
"gt_answers": "fatherinlaw",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094f/c03-094f-05-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "always and forever",
|
||
|
|
"gt_answers": "always",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094f/c03-094f-05-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "comically",
|
||
|
|
"gt_answers": "comically",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094f/c03-094f-05-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "grateque",
|
||
|
|
"gt_answers": "grotesque",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094f/c03-094f-05-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "they",
|
||
|
|
"gt_answers": "they",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094f/c03-094f-06-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word we",
|
||
|
|
"gt_answers": "were",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094f/c03-094f-06-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "not",
|
||
|
|
"gt_answers": "not",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094f/c03-094f-06-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "asked",
|
||
|
|
"gt_answers": "asked",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094f/c03-094f-06-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "oh",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094f/c03-094f-06-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "module",
|
||
|
|
"gt_answers": "modulate",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094f/c03-094f-06-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word from",
|
||
|
|
"gt_answers": "from",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094f/c03-094f-06-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "their",
|
||
|
|
"gt_answers": "their",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094f/c03-094f-07-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "set",
|
||
|
|
"gt_answers": "set",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094f/c03-094f-07-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "moods",
|
||
|
|
"gt_answers": "moods",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094f/c03-094f-07-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "hd",
|
||
|
|
"gt_answers": "but",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094f/c03-094f-07-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "played",
|
||
|
|
"gt_answers": "played",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094f/c03-094f-07-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "with love",
|
||
|
|
"gt_answers": "with",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094f/c03-094f-07-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "proper",
|
||
|
|
"gt_answers": "proper",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094f/c03-094f-07-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "efficiency",
|
||
|
|
"gt_answers": "efficiency",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094f/c03-094f-08-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word and",
|
||
|
|
"gt_answers": "and",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094f/c03-094f-08-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "d",
|
||
|
|
"gt_answers": "in",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094f/c03-094f-08-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094f/c03-094f-08-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "case",
|
||
|
|
"gt_answers": "case",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094f/c03-094f-08-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "f",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094f/c03-094f-08-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "h",
|
||
|
|
"gt_answers": "Mr",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094f/c03-094f-08-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "richards",
|
||
|
|
"gt_answers": "Richards",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094f/c03-094f-08-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "with love",
|
||
|
|
"gt_answers": "with",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094f/c03-094f-09-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "lauish",
|
||
|
|
"gt_answers": "lavish",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094f/c03-094f-09-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "and",
|
||
|
|
"gt_answers": "and",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094f/c03-094f-09-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "suitably",
|
||
|
|
"gt_answers": "suitably",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094f/c03-094f-09-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "gaddy",
|
||
|
|
"gt_answers": "gaudy",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094f/c03-094f-09-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "colour",
|
||
|
|
"gt_answers": "colour",
|
||
|
|
"image_path": "./data/IAM/c03/c03-094f/c03-094f-09-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "The",
|
||
|
|
"image_path": "./data/IAM/e04/e04-095/e04-095-00-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "skeleton",
|
||
|
|
"gt_answers": "skeleton",
|
||
|
|
"image_path": "./data/IAM/e04/e04-095/e04-095-00-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/e04/e04-095/e04-095-00-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/e04/e04-095/e04-095-00-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "boat",
|
||
|
|
"gt_answers": "boat",
|
||
|
|
"image_path": "./data/IAM/e04/e04-095/e04-095-00-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "will",
|
||
|
|
"gt_answers": "will",
|
||
|
|
"image_path": "./data/IAM/e04/e04-095/e04-095-00-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "no",
|
||
|
|
"gt_answers": "now",
|
||
|
|
"image_path": "./data/IAM/e04/e04-095/e04-095-00-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "shake",
|
||
|
|
"gt_answers": "take",
|
||
|
|
"image_path": "./data/IAM/e04/e04-095/e04-095-00-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "shepe",
|
||
|
|
"gt_answers": "shape",
|
||
|
|
"image_path": "./data/IAM/e04/e04-095/e04-095-00-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "d",
|
||
|
|
"gt_answers": "and",
|
||
|
|
"image_path": "./data/IAM/e04/e04-095/e04-095-01-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "his",
|
||
|
|
"gt_answers": "this",
|
||
|
|
"image_path": "./data/IAM/e04/e04-095/e04-095-01-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "is",
|
||
|
|
"image_path": "./data/IAM/e04/e04-095/e04-095-01-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/e04/e04-095/e04-095-01-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "hme",
|
||
|
|
"gt_answers": "time",
|
||
|
|
"image_path": "./data/IAM/e04/e04-095/e04-095-01-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "b",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/e04/e04-095/e04-095-01-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "check",
|
||
|
|
"gt_answers": "check",
|
||
|
|
"image_path": "./data/IAM/e04/e04-095/e04-095-01-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "before",
|
||
|
|
"gt_answers": "before",
|
||
|
|
"image_path": "./data/IAM/e04/e04-095/e04-095-01-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "making",
|
||
|
|
"gt_answers": "making",
|
||
|
|
"image_path": "./data/IAM/e04/e04-095/e04-095-01-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "fash",
|
||
|
|
"gt_answers": "fast",
|
||
|
|
"image_path": "./data/IAM/e04/e04-095/e04-095-01-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a signature",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/e04/e04-095/e04-095-02-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "jee",
|
||
|
|
"gt_answers": "see",
|
||
|
|
"image_path": "./data/IAM/e04/e04-095/e04-095-02-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word that",
|
||
|
|
"gt_answers": "that",
|
||
|
|
"image_path": "./data/IAM/e04/e04-095/e04-095-02-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "nothing",
|
||
|
|
"gt_answers": "nothing",
|
||
|
|
"image_path": "./data/IAM/e04/e04-095/e04-095-02-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "is",
|
||
|
|
"image_path": "./data/IAM/e04/e04-095/e04-095-02-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "disturbed",
|
||
|
|
"gt_answers": "distorted",
|
||
|
|
"image_path": "./data/IAM/e04/e04-095/e04-095-02-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "ensure",
|
||
|
|
"gt_answers": "Ensure",
|
||
|
|
"image_path": "./data/IAM/e04/e04-095/e04-095-02-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "thor",
|
||
|
|
"gt_answers": "that",
|
||
|
|
"image_path": "./data/IAM/e04/e04-095/e04-095-02-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/e04/e04-095/e04-095-03-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "outside",
|
||
|
|
"gt_answers": "outside",
|
||
|
|
"image_path": "./data/IAM/e04/e04-095/e04-095-03-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word measure",
|
||
|
|
"gt_answers": "measurements",
|
||
|
|
"image_path": "./data/IAM/e04/e04-095/e04-095-03-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "f m",
|
||
|
|
"gt_answers": "from",
|
||
|
|
"image_path": "./data/IAM/e04/e04-095/e04-095-03-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "transson",
|
||
|
|
"gt_answers": "transom",
|
||
|
|
"image_path": "./data/IAM/e04/e04-095/e04-095-03-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/e04/e04-095/e04-095-03-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "harrison",
|
||
|
|
"gt_answers": "transom",
|
||
|
|
"image_path": "./data/IAM/e04/e04-095/e04-095-03-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word core",
|
||
|
|
"gt_answers": "are",
|
||
|
|
"image_path": "./data/IAM/e04/e04-095/e04-095-04-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "equal",
|
||
|
|
"gt_answers": "equal",
|
||
|
|
"image_path": "./data/IAM/e04/e04-095/e04-095-04-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "ab",
|
||
|
|
"gt_answers": "on",
|
||
|
|
"image_path": "./data/IAM/e04/e04-095/e04-095-04-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word teach",
|
||
|
|
"gt_answers": "each",
|
||
|
|
"image_path": "./data/IAM/e04/e04-095/e04-095-04-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "s d",
|
||
|
|
"gt_answers": "side",
|
||
|
|
"image_path": "./data/IAM/e04/e04-095/e04-095-04-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "ocd",
|
||
|
|
"gt_answers": "and",
|
||
|
|
"image_path": "./data/IAM/e04/e04-095/e04-095-04-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "hat",
|
||
|
|
"gt_answers": "that",
|
||
|
|
"image_path": "./data/IAM/e04/e04-095/e04-095-04-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/e04/e04-095/e04-095-04-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word hansom",
|
||
|
|
"gt_answers": "transoms",
|
||
|
|
"image_path": "./data/IAM/e04/e04-095/e04-095-04-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word core",
|
||
|
|
"gt_answers": "are",
|
||
|
|
"image_path": "./data/IAM/e04/e04-095/e04-095-05-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "parallel",
|
||
|
|
"gt_answers": "parallel",
|
||
|
|
"image_path": "./data/IAM/e04/e04-095/e04-095-05-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "horizon valley",
|
||
|
|
"gt_answers": "horizontally",
|
||
|
|
"image_path": "./data/IAM/e04/e04-095/e04-095-05-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "neil",
|
||
|
|
"gt_answers": "Next",
|
||
|
|
"image_path": "./data/IAM/e04/e04-095/e04-095-05-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "bevel",
|
||
|
|
"gt_answers": "bevel",
|
||
|
|
"image_path": "./data/IAM/e04/e04-095/e04-095-05-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/e04/e04-095/e04-095-05-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "chess",
|
||
|
|
"gt_answers": "chines",
|
||
|
|
"image_path": "./data/IAM/e04/e04-095/e04-095-05-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "th",
|
||
|
|
"gt_answers": "with",
|
||
|
|
"image_path": "./data/IAM/e04/e04-095/e04-095-06-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "a",
|
||
|
|
"image_path": "./data/IAM/e04/e04-095/e04-095-06-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "plane",
|
||
|
|
"gt_answers": "plane",
|
||
|
|
"image_path": "./data/IAM/e04/e04-095/e04-095-06-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "so",
|
||
|
|
"gt_answers": "so",
|
||
|
|
"image_path": "./data/IAM/e04/e04-095/e04-095-06-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "that",
|
||
|
|
"gt_answers": "that",
|
||
|
|
"image_path": "./data/IAM/e04/e04-095/e04-095-06-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/e04/e04-095/e04-095-06-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "boston",
|
||
|
|
"gt_answers": "bottom",
|
||
|
|
"image_path": "./data/IAM/e04/e04-095/e04-095-06-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "will",
|
||
|
|
"gt_answers": "will",
|
||
|
|
"image_path": "./data/IAM/e04/e04-095/e04-095-06-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "fit",
|
||
|
|
"image_path": "./data/IAM/e04/e04-095/e04-095-06-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word snuggly",
|
||
|
|
"gt_answers": "snugly",
|
||
|
|
"image_path": "./data/IAM/e04/e04-095/e04-095-06-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "oh",
|
||
|
|
"gt_answers": "Who",
|
||
|
|
"image_path": "./data/IAM/b04/b04-010/b04-010-00-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the name of person",
|
||
|
|
"gt_answers": "will",
|
||
|
|
"image_path": "./data/IAM/b04/b04-010/b04-010-00-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "speak",
|
||
|
|
"gt_answers": "speak",
|
||
|
|
"image_path": "./data/IAM/b04/b04-010/b04-010-00-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "for",
|
||
|
|
"image_path": "./data/IAM/b04/b04-010/b04-010-00-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "belgian",
|
||
|
|
"gt_answers": "Belgium",
|
||
|
|
"image_path": "./data/IAM/b04/b04-010/b04-010-00-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "Who",
|
||
|
|
"image_path": "./data/IAM/b04/b04-010/b04-010-00-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "ele",
|
||
|
|
"gt_answers": "else",
|
||
|
|
"image_path": "./data/IAM/b04/b04-010/b04-010-00-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "t 6",
|
||
|
|
"gt_answers": "but",
|
||
|
|
"image_path": "./data/IAM/b04/b04-010/b04-010-00-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "britain",
|
||
|
|
"gt_answers": "Britain",
|
||
|
|
"image_path": "./data/IAM/b04/b04-010/b04-010-01-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "we",
|
||
|
|
"gt_answers": "We",
|
||
|
|
"image_path": "./data/IAM/b04/b04-010/b04-010-01-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "in love",
|
||
|
|
"gt_answers": "have",
|
||
|
|
"image_path": "./data/IAM/b04/b04-010/b04-010-01-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "f o",
|
||
|
|
"gt_answers": "fought",
|
||
|
|
"image_path": "./data/IAM/b04/b04-010/b04-010-01-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "bessie",
|
||
|
|
"gt_answers": "beside",
|
||
|
|
"image_path": "./data/IAM/b04/b04-010/b04-010-01-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "belgium",
|
||
|
|
"gt_answers": "Belgium",
|
||
|
|
"image_path": "./data/IAM/b04/b04-010/b04-010-01-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "in",
|
||
|
|
"image_path": "./data/IAM/b04/b04-010/b04-010-02-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word out",
|
||
|
|
"gt_answers": "two",
|
||
|
|
"image_path": "./data/IAM/b04/b04-010/b04-010-02-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word love",
|
||
|
|
"gt_answers": "world",
|
||
|
|
"image_path": "./data/IAM/b04/b04-010/b04-010-02-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "wars",
|
||
|
|
"image_path": "./data/IAM/b04/b04-010/b04-010-02-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "we",
|
||
|
|
"gt_answers": "We",
|
||
|
|
"image_path": "./data/IAM/b04/b04-010/b04-010-02-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "are",
|
||
|
|
"image_path": "./data/IAM/b04/b04-010/b04-010-02-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "allie",
|
||
|
|
"gt_answers": "allies",
|
||
|
|
"image_path": "./data/IAM/b04/b04-010/b04-010-02-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word st",
|
||
|
|
"gt_answers": "still",
|
||
|
|
"image_path": "./data/IAM/b04/b04-010/b04-010-02-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "britain",
|
||
|
|
"gt_answers": "Britain",
|
||
|
|
"image_path": "./data/IAM/b04/b04-010/b04-010-03-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "bluw",
|
||
|
|
"gt_answers": "should",
|
||
|
|
"image_path": "./data/IAM/b04/b04-010/b04-010-03-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the champion",
|
||
|
|
"gt_answers": "champion",
|
||
|
|
"image_path": "./data/IAM/b04/b04-010/b04-010-03-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "beguin",
|
||
|
|
"gt_answers": "Belgium",
|
||
|
|
"image_path": "./data/IAM/b04/b04-010/b04-010-03-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "no",
|
||
|
|
"gt_answers": "Not",
|
||
|
|
"image_path": "./data/IAM/b04/b04-010/b04-010-03-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word person",
|
||
|
|
"gt_answers": "with",
|
||
|
|
"image_path": "./data/IAM/b04/b04-010/b04-010-04-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/b04/b04-010/b04-010-04-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "careful",
|
||
|
|
"gt_answers": "careful",
|
||
|
|
"image_path": "./data/IAM/b04/b04-010/b04-010-04-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "hooded",
|
||
|
|
"gt_answers": "hooded",
|
||
|
|
"image_path": "./data/IAM/b04/b04-010/b04-010-04-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "language",
|
||
|
|
"gt_answers": "language",
|
||
|
|
"image_path": "./data/IAM/b04/b04-010/b04-010-04-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the cross",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/b04/b04-010/b04-010-04-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "diplomacy",
|
||
|
|
"gt_answers": "diplomacy",
|
||
|
|
"image_path": "./data/IAM/b04/b04-010/b04-010-04-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "lost",
|
||
|
|
"gt_answers": "but",
|
||
|
|
"image_path": "./data/IAM/b04/b04-010/b04-010-05-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the name of the person",
|
||
|
|
"gt_answers": "boldly",
|
||
|
|
"image_path": "./data/IAM/b04/b04-010/b04-010-05-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "and",
|
||
|
|
"image_path": "./data/IAM/b04/b04-010/b04-010-05-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "fantasyy",
|
||
|
|
"gt_answers": "fearlessly",
|
||
|
|
"image_path": "./data/IAM/b04/b04-010/b04-010-05-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "k",
|
||
|
|
"gt_answers": "It",
|
||
|
|
"image_path": "./data/IAM/b04/b04-010/b04-010-05-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "is",
|
||
|
|
"image_path": "./data/IAM/b04/b04-010/b04-010-05-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word true",
|
||
|
|
"gt_answers": "time",
|
||
|
|
"image_path": "./data/IAM/b04/b04-010/b04-010-05-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "h",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/b04/b04-010/b04-010-05-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "show",
|
||
|
|
"gt_answers": "show",
|
||
|
|
"image_path": "./data/IAM/b04/b04-010/b04-010-06-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "th",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/b04/b04-010/b04-010-06-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word world",
|
||
|
|
"gt_answers": "world",
|
||
|
|
"image_path": "./data/IAM/b04/b04-010/b04-010-06-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word the",
|
||
|
|
"gt_answers": "that",
|
||
|
|
"image_path": "./data/IAM/b04/b04-010/b04-010-06-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "this is person",
|
||
|
|
"gt_answers": "this",
|
||
|
|
"image_path": "./data/IAM/b04/b04-010/b04-010-06-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "county",
|
||
|
|
"gt_answers": "country",
|
||
|
|
"image_path": "./data/IAM/b04/b04-010/b04-010-06-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "does",
|
||
|
|
"image_path": "./data/IAM/b04/b04-010/b04-010-06-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "no",
|
||
|
|
"gt_answers": "not",
|
||
|
|
"image_path": "./data/IAM/b04/b04-010/b04-010-06-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "desert",
|
||
|
|
"gt_answers": "desert",
|
||
|
|
"image_path": "./data/IAM/b04/b04-010/b04-010-07-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "her",
|
||
|
|
"gt_answers": "her",
|
||
|
|
"image_path": "./data/IAM/b04/b04-010/b04-010-07-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "friends",
|
||
|
|
"gt_answers": "friends",
|
||
|
|
"image_path": "./data/IAM/b04/b04-010/b04-010-07-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "happy",
|
||
|
|
"gt_answers": "HAPPY",
|
||
|
|
"image_path": "./data/IAM/b04/b04-010/b04-010-07-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "happy",
|
||
|
|
"gt_answers": "happy",
|
||
|
|
"image_path": "./data/IAM/b04/b04-010/b04-010-07-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "family name",
|
||
|
|
"gt_answers": "families",
|
||
|
|
"image_path": "./data/IAM/b04/b04-010/b04-010-08-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word never",
|
||
|
|
"gt_answers": "Never",
|
||
|
|
"image_path": "./data/IAM/b04/b04-010/b04-010-08-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "before",
|
||
|
|
"gt_answers": "before",
|
||
|
|
"image_path": "./data/IAM/b04/b04-010/b04-010-08-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word have",
|
||
|
|
"gt_answers": "have",
|
||
|
|
"image_path": "./data/IAM/b04/b04-010/b04-010-08-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "britannia",
|
||
|
|
"gt_answers": "Britains",
|
||
|
|
"image_path": "./data/IAM/b04/b04-010/b04-010-08-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "larders",
|
||
|
|
"image_path": "./data/IAM/b04/b04-010/b04-010-09-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the name person",
|
||
|
|
"gt_answers": "been",
|
||
|
|
"image_path": "./data/IAM/b04/b04-010/b04-010-09-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "so",
|
||
|
|
"image_path": "./data/IAM/b04/b04-010/b04-010-09-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "lewis",
|
||
|
|
"gt_answers": "well",
|
||
|
|
"image_path": "./data/IAM/b04/b04-010/b04-010-09-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "stoked",
|
||
|
|
"gt_answers": "stocked",
|
||
|
|
"image_path": "./data/IAM/b04/b04-010/b04-010-09-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "only",
|
||
|
|
"gt_answers": "Only",
|
||
|
|
"image_path": "./data/IAM/j04/j04-015/j04-015-00-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/j04/j04-015/j04-015-00-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "weights",
|
||
|
|
"gt_answers": "weights",
|
||
|
|
"image_path": "./data/IAM/j04/j04-015/j04-015-00-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "o",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/j04/j04-015/j04-015-00-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/j04/j04-015/j04-015-00-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "recovered",
|
||
|
|
"gt_answers": "recovered",
|
||
|
|
"image_path": "./data/IAM/j04/j04-015/j04-015-00-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "lead",
|
||
|
|
"gt_answers": "lead",
|
||
|
|
"image_path": "./data/IAM/j04/j04-015/j04-015-00-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "chrome",
|
||
|
|
"gt_answers": "chromate",
|
||
|
|
"image_path": "./data/IAM/j04/j04-015/j04-015-00-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "predictions",
|
||
|
|
"gt_answers": "precipitates",
|
||
|
|
"image_path": "./data/IAM/j04/j04-015/j04-015-01-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "need",
|
||
|
|
"gt_answers": "need",
|
||
|
|
"image_path": "./data/IAM/j04/j04-015/j04-015-01-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "be",
|
||
|
|
"gt_answers": "be",
|
||
|
|
"image_path": "./data/IAM/j04/j04-015/j04-015-01-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "know",
|
||
|
|
"gt_answers": "known",
|
||
|
|
"image_path": "./data/IAM/j04/j04-015/j04-015-01-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "because",
|
||
|
|
"gt_answers": "because",
|
||
|
|
"image_path": "./data/IAM/j04/j04-015/j04-015-01-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/j04/j04-015/j04-015-01-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "concern play",
|
||
|
|
"gt_answers": "concentration",
|
||
|
|
"image_path": "./data/IAM/j04/j04-015/j04-015-01-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "f",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/j04/j04-015/j04-015-02-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/j04/j04-015/j04-015-02-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "lead",
|
||
|
|
"gt_answers": "lead",
|
||
|
|
"image_path": "./data/IAM/j04/j04-015/j04-015-02-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "carrier",
|
||
|
|
"image_path": "./data/IAM/j04/j04-015/j04-015-02-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "solution",
|
||
|
|
"gt_answers": "solution",
|
||
|
|
"image_path": "./data/IAM/j04/j04-015/j04-015-02-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "cancels",
|
||
|
|
"gt_answers": "cancels",
|
||
|
|
"image_path": "./data/IAM/j04/j04-015/j04-015-02-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "out",
|
||
|
|
"gt_answers": "out",
|
||
|
|
"image_path": "./data/IAM/j04/j04-015/j04-015-02-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "f",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/j04/j04-015/j04-015-02-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/j04/j04-015/j04-015-02-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "algebra",
|
||
|
|
"gt_answers": "algebraic",
|
||
|
|
"image_path": "./data/IAM/j04/j04-015/j04-015-03-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "equations",
|
||
|
|
"gt_answers": "equations",
|
||
|
|
"image_path": "./data/IAM/j04/j04-015/j04-015-03-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "am",
|
||
|
|
"gt_answers": "An",
|
||
|
|
"image_path": "./data/IAM/j04/j04-015/j04-015-03-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "effort",
|
||
|
|
"gt_answers": "effort",
|
||
|
|
"image_path": "./data/IAM/j04/j04-015/j04-015-03-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "ws",
|
||
|
|
"gt_answers": "was",
|
||
|
|
"image_path": "./data/IAM/j04/j04-015/j04-015-03-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "made",
|
||
|
|
"gt_answers": "made",
|
||
|
|
"image_path": "./data/IAM/j04/j04-015/j04-015-03-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "o",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/j04/j04-015/j04-015-03-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "deet",
|
||
|
|
"gt_answers": "detect",
|
||
|
|
"image_path": "./data/IAM/j04/j04-015/j04-015-04-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/j04/j04-015/j04-015-04-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "presence",
|
||
|
|
"gt_answers": "presence",
|
||
|
|
"image_path": "./data/IAM/j04/j04-015/j04-015-04-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "o",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/j04/j04-015/j04-015-04-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "y",
|
||
|
|
"gt_answers": "any",
|
||
|
|
"image_path": "./data/IAM/j04/j04-015/j04-015-04-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "radioactive",
|
||
|
|
"gt_answers": "radioactive",
|
||
|
|
"image_path": "./data/IAM/j04/j04-015/j04-015-04-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "impureties",
|
||
|
|
"gt_answers": "impurities",
|
||
|
|
"image_path": "./data/IAM/j04/j04-015/j04-015-04-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "m",
|
||
|
|
"gt_answers": "in",
|
||
|
|
"image_path": "./data/IAM/j04/j04-015/j04-015-04-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/j04/j04-015/j04-015-05-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "tracer",
|
||
|
|
"gt_answers": "tracer",
|
||
|
|
"image_path": "./data/IAM/j04/j04-015/j04-015-05-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "by person",
|
||
|
|
"gt_answers": "by",
|
||
|
|
"image_path": "./data/IAM/j04/j04-015/j04-015-05-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "separahug",
|
||
|
|
"gt_answers": "separating",
|
||
|
|
"image_path": "./data/IAM/j04/j04-015/j04-015-05-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/j04/j04-015/j04-015-05-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "lead 10",
|
||
|
|
"gt_answers": "lead210",
|
||
|
|
"image_path": "./data/IAM/j04/j04-015/j04-015-05-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "and",
|
||
|
|
"gt_answers": "and",
|
||
|
|
"image_path": "./data/IAM/j04/j04-015/j04-015-05-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/j04/j04-015/j04-015-05-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "bismuth 210",
|
||
|
|
"gt_answers": "bismuth210",
|
||
|
|
"image_path": "./data/IAM/j04/j04-015/j04-015-06-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "by",
|
||
|
|
"gt_answers": "by",
|
||
|
|
"image_path": "./data/IAM/j04/j04-015/j04-015-06-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word person",
|
||
|
|
"gt_answers": "anion",
|
||
|
|
"image_path": "./data/IAM/j04/j04-015/j04-015-06-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "exchange",
|
||
|
|
"gt_answers": "exchange",
|
||
|
|
"image_path": "./data/IAM/j04/j04-015/j04-015-06-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "The",
|
||
|
|
"image_path": "./data/IAM/j04/j04-015/j04-015-06-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "16 b counting",
|
||
|
|
"gt_answers": "15bcounting",
|
||
|
|
"image_path": "./data/IAM/j04/j04-015/j04-015-06-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "f",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/j04/j04-015/j04-015-07-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/j04/j04-015/j04-015-07-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "lead 210",
|
||
|
|
"gt_answers": "lead210",
|
||
|
|
"image_path": "./data/IAM/j04/j04-015/j04-015-07-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "fraction",
|
||
|
|
"gt_answers": "fraction",
|
||
|
|
"image_path": "./data/IAM/j04/j04-015/j04-015-07-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "begin",
|
||
|
|
"gt_answers": "began",
|
||
|
|
"image_path": "./data/IAM/j04/j04-015/j04-015-07-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "whitney",
|
||
|
|
"gt_answers": "within",
|
||
|
|
"image_path": "./data/IAM/j04/j04-015/j04-015-07-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a snake",
|
||
|
|
"gt_answers": "a",
|
||
|
|
"image_path": "./data/IAM/j04/j04-015/j04-015-07-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "few",
|
||
|
|
"gt_answers": "few",
|
||
|
|
"image_path": "./data/IAM/j04/j04-015/j04-015-07-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "five minutes",
|
||
|
|
"gt_answers": "minutes",
|
||
|
|
"image_path": "./data/IAM/j04/j04-015/j04-015-08-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "f",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/j04/j04-015/j04-015-08-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word completion",
|
||
|
|
"gt_answers": "completing",
|
||
|
|
"image_path": "./data/IAM/j04/j04-015/j04-015-08-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/j04/j04-015/j04-015-08-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "separation",
|
||
|
|
"gt_answers": "separation",
|
||
|
|
"image_path": "./data/IAM/j04/j04-015/j04-015-08-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "es",
|
||
|
|
"gt_answers": "Yes",
|
||
|
|
"image_path": "./data/IAM/p01/p01-174/p01-174-00-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the letter c",
|
||
|
|
"gt_answers": "I",
|
||
|
|
"image_path": "./data/IAM/p01/p01-174/p01-174-00-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the name of the person",
|
||
|
|
"gt_answers": "travelled",
|
||
|
|
"image_path": "./data/IAM/p01/p01-174/p01-174-00-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "down",
|
||
|
|
"image_path": "./data/IAM/p01/p01-174/p01-174-00-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the name of person",
|
||
|
|
"gt_answers": "from",
|
||
|
|
"image_path": "./data/IAM/p01/p01-174/p01-174-00-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "london",
|
||
|
|
"gt_answers": "London",
|
||
|
|
"image_path": "./data/IAM/p01/p01-174/p01-174-00-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "e",
|
||
|
|
"gt_answers": "as",
|
||
|
|
"image_path": "./data/IAM/p01/p01-174/p01-174-01-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "flie",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/p01/p01-174/p01-174-01-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "clark",
|
||
|
|
"gt_answers": "dawn",
|
||
|
|
"image_path": "./data/IAM/p01/p01-174/p01-174-01-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "was",
|
||
|
|
"image_path": "./data/IAM/p01/p01-174/p01-174-01-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "breathing",
|
||
|
|
"gt_answers": "breaking",
|
||
|
|
"image_path": "./data/IAM/p01/p01-174/p01-174-01-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "and",
|
||
|
|
"image_path": "./data/IAM/p01/p01-174/p01-174-01-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "l e",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/p01/p01-174/p01-174-02-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "thought",
|
||
|
|
"gt_answers": "thought",
|
||
|
|
"image_path": "./data/IAM/p01/p01-174/p01-174-02-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "ceme",
|
||
|
|
"gt_answers": "came",
|
||
|
|
"image_path": "./data/IAM/p01/p01-174/p01-174-02-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/p01/p01-174/p01-174-02-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "me",
|
||
|
|
"image_path": "./data/IAM/p01/p01-174/p01-174-02-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "I",
|
||
|
|
"image_path": "./data/IAM/p01/p01-174/p01-174-02-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "d",
|
||
|
|
"image_path": "./data/IAM/p01/p01-174/p01-174-02-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the name person",
|
||
|
|
"gt_answers": "like",
|
||
|
|
"image_path": "./data/IAM/p01/p01-174/p01-174-02-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a signature",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/p01/p01-174/p01-174-03-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "kale",
|
||
|
|
"gt_answers": "take",
|
||
|
|
"image_path": "./data/IAM/p01/p01-174/p01-174-03-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "a",
|
||
|
|
"image_path": "./data/IAM/p01/p01-174/p01-174-03-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "look",
|
||
|
|
"gt_answers": "look",
|
||
|
|
"image_path": "./data/IAM/p01/p01-174/p01-174-03-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "at",
|
||
|
|
"image_path": "./data/IAM/p01/p01-174/p01-174-03-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/p01/p01-174/p01-174-03-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "cleve",
|
||
|
|
"gt_answers": "church",
|
||
|
|
"image_path": "./data/IAM/p01/p01-174/p01-174-03-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "I",
|
||
|
|
"image_path": "./data/IAM/p01/p01-174/p01-174-03-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the name person",
|
||
|
|
"gt_answers": "ve",
|
||
|
|
"image_path": "./data/IAM/p01/p01-174/p01-174-03-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word news",
|
||
|
|
"gt_answers": "never",
|
||
|
|
"image_path": "./data/IAM/p01/p01-174/p01-174-04-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "lee",
|
||
|
|
"gt_answers": "been",
|
||
|
|
"image_path": "./data/IAM/p01/p01-174/p01-174-04-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the name of person",
|
||
|
|
"gt_answers": "inside",
|
||
|
|
"image_path": "./data/IAM/p01/p01-174/p01-174-04-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "Very",
|
||
|
|
"image_path": "./data/IAM/p01/p01-174/p01-174-04-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "remiss",
|
||
|
|
"image_path": "./data/IAM/p01/p01-174/p01-174-04-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/p01/p01-174/p01-174-04-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "ml",
|
||
|
|
"gt_answers": "me",
|
||
|
|
"image_path": "./data/IAM/p01/p01-174/p01-174-05-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the name person",
|
||
|
|
"gt_answers": "said",
|
||
|
|
"image_path": "./data/IAM/p01/p01-174/p01-174-05-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "mr",
|
||
|
|
"gt_answers": "Mr",
|
||
|
|
"image_path": "./data/IAM/p01/p01-174/p01-174-05-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "copflone",
|
||
|
|
"gt_answers": "Copthorne",
|
||
|
|
"image_path": "./data/IAM/p01/p01-174/p01-174-05-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "elio",
|
||
|
|
"gt_answers": "The",
|
||
|
|
"image_path": "./data/IAM/p01/p01-174/p01-174-05-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "elger",
|
||
|
|
"gt_answers": "verger",
|
||
|
|
"image_path": "./data/IAM/p01/p01-174/p01-174-05-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "leo",
|
||
|
|
"gt_answers": "led",
|
||
|
|
"image_path": "./data/IAM/p01/p01-174/p01-174-06-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/p01/p01-174/p01-174-06-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word way",
|
||
|
|
"gt_answers": "way",
|
||
|
|
"image_path": "./data/IAM/p01/p01-174/p01-174-06-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "in",
|
||
|
|
"image_path": "./data/IAM/p01/p01-174/p01-174-06-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "ie",
|
||
|
|
"gt_answers": "The",
|
||
|
|
"image_path": "./data/IAM/p01/p01-174/p01-174-06-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the name charlie",
|
||
|
|
"gt_answers": "stained",
|
||
|
|
"image_path": "./data/IAM/p01/p01-174/p01-174-06-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "glass",
|
||
|
|
"image_path": "./data/IAM/p01/p01-174/p01-174-06-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word windows",
|
||
|
|
"gt_answers": "windows",
|
||
|
|
"image_path": "./data/IAM/p01/p01-174/p01-174-07-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "z",
|
||
|
|
"gt_answers": "sir",
|
||
|
|
"image_path": "./data/IAM/p01/p01-174/p01-174-07-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "they",
|
||
|
|
"image_path": "./data/IAM/p01/p01-174/p01-174-07-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "harley cloke",
|
||
|
|
"gt_answers": "date",
|
||
|
|
"image_path": "./data/IAM/p01/p01-174/p01-174-07-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word person",
|
||
|
|
"gt_answers": "back",
|
||
|
|
"image_path": "./data/IAM/p01/p01-174/p01-174-07-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/p01/p01-174/p01-174-07-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "he",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/p01/p01-174/p01-174-08-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "faulkner",
|
||
|
|
"gt_answers": "fourteenth",
|
||
|
|
"image_path": "./data/IAM/p01/p01-174/p01-174-08-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the name of the person",
|
||
|
|
"gt_answers": "century",
|
||
|
|
"image_path": "./data/IAM/p01/p01-174/p01-174-08-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "i love you",
|
||
|
|
"gt_answers": "Goodbye",
|
||
|
|
"image_path": "./data/IAM/m04/m04-030/m04-030-00-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "dan",
|
||
|
|
"gt_answers": "Dai",
|
||
|
|
"image_path": "./data/IAM/m04/m04-030/m04-030-00-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "hun",
|
||
|
|
"gt_answers": "said",
|
||
|
|
"image_path": "./data/IAM/m04/m04-030/m04-030-00-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "dylan",
|
||
|
|
"gt_answers": "Morfydd",
|
||
|
|
"image_path": "./data/IAM/m04/m04-030/m04-030-00-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "phillips",
|
||
|
|
"gt_answers": "smiling",
|
||
|
|
"image_path": "./data/IAM/m04/m04-030/m04-030-00-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "h",
|
||
|
|
"gt_answers": "from",
|
||
|
|
"image_path": "./data/IAM/m04/m04-030/m04-030-00-10.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/m04/m04-030/m04-030-00-11.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "bottom",
|
||
|
|
"image_path": "./data/IAM/m04/m04-030/m04-030-00-12.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "step",
|
||
|
|
"image_path": "./data/IAM/m04/m04-030/m04-030-00-13.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "blair girl beg",
|
||
|
|
"gt_answers": "Goodbye",
|
||
|
|
"image_path": "./data/IAM/m04/m04-030/m04-030-01-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the north face",
|
||
|
|
"gt_answers": "Morfydd",
|
||
|
|
"image_path": "./data/IAM/m04/m04-030/m04-030-01-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the thought",
|
||
|
|
"gt_answers": "He",
|
||
|
|
"image_path": "./data/IAM/m04/m04-030/m04-030-01-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "f",
|
||
|
|
"gt_answers": "sought",
|
||
|
|
"image_path": "./data/IAM/m04/m04-030/m04-030-01-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "blee",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/m04/m04-030/m04-030-01-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "h",
|
||
|
|
"gt_answers": "enfold",
|
||
|
|
"image_path": "./data/IAM/m04/m04-030/m04-030-01-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "wavy line",
|
||
|
|
"gt_answers": "her",
|
||
|
|
"image_path": "./data/IAM/m04/m04-030/m04-030-01-10.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "in",
|
||
|
|
"image_path": "./data/IAM/m04/m04-030/m04-030-01-11.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "he",
|
||
|
|
"gt_answers": "his",
|
||
|
|
"image_path": "./data/IAM/m04/m04-030/m04-030-01-12.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "amr",
|
||
|
|
"gt_answers": "arms",
|
||
|
|
"image_path": "./data/IAM/m04/m04-030/m04-030-01-13.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "dut",
|
||
|
|
"gt_answers": "But",
|
||
|
|
"image_path": "./data/IAM/m04/m04-030/m04-030-01-15.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "he",
|
||
|
|
"image_path": "./data/IAM/m04/m04-030/m04-030-01-16.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "m w",
|
||
|
|
"gt_answers": "was",
|
||
|
|
"image_path": "./data/IAM/m04/m04-030/m04-030-01-17.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "ot",
|
||
|
|
"gt_answers": "too",
|
||
|
|
"image_path": "./data/IAM/m04/m04-030/m04-030-01-18.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "el",
|
||
|
|
"gt_answers": "late",
|
||
|
|
"image_path": "./data/IAM/m04/m04-030/m04-030-01-19.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "She",
|
||
|
|
"image_path": "./data/IAM/m04/m04-030/m04-030-02-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "w",
|
||
|
|
"gt_answers": "was",
|
||
|
|
"image_path": "./data/IAM/m04/m04-030/m04-030-02-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "already",
|
||
|
|
"gt_answers": "already",
|
||
|
|
"image_path": "./data/IAM/m04/m04-030/m04-030-02-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a signature",
|
||
|
|
"gt_answers": "up",
|
||
|
|
"image_path": "./data/IAM/m04/m04-030/m04-030-02-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "th",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/m04/m04-030/m04-030-02-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a signature",
|
||
|
|
"gt_answers": "bus",
|
||
|
|
"image_path": "./data/IAM/m04/m04-030/m04-030-02-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word amr",
|
||
|
|
"gt_answers": "steps",
|
||
|
|
"image_path": "./data/IAM/m04/m04-030/m04-030-02-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a m",
|
||
|
|
"gt_answers": "and",
|
||
|
|
"image_path": "./data/IAM/m04/m04-030/m04-030-02-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/m04/m04-030/m04-030-02-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "m",
|
||
|
|
"gt_answers": "bus",
|
||
|
|
"image_path": "./data/IAM/m04/m04-030/m04-030-02-10.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "w",
|
||
|
|
"gt_answers": "was",
|
||
|
|
"image_path": "./data/IAM/m04/m04-030/m04-030-02-11.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word anna",
|
||
|
|
"gt_answers": "away",
|
||
|
|
"image_path": "./data/IAM/m04/m04-030/m04-030-02-12.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "md",
|
||
|
|
"gt_answers": "And",
|
||
|
|
"image_path": "./data/IAM/m04/m04-030/m04-030-02-14.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "mary d",
|
||
|
|
"gt_answers": "Morfydd",
|
||
|
|
"image_path": "./data/IAM/m04/m04-030/m04-030-02-15.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "om",
|
||
|
|
"gt_answers": "Owen",
|
||
|
|
"image_path": "./data/IAM/m04/m04-030/m04-030-02-16.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "waving",
|
||
|
|
"image_path": "./data/IAM/m04/m04-030/m04-030-03-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "unkissed",
|
||
|
|
"image_path": "./data/IAM/m04/m04-030/m04-030-03-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "hm",
|
||
|
|
"gt_answers": "from",
|
||
|
|
"image_path": "./data/IAM/m04/m04-030/m04-030-03-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a h",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/m04/m04-030/m04-030-03-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "window",
|
||
|
|
"image_path": "./data/IAM/m04/m04-030/m04-030-03-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "b",
|
||
|
|
"gt_answers": "And",
|
||
|
|
"image_path": "./data/IAM/m04/m04-030/m04-030-03-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the name of person",
|
||
|
|
"gt_answers": "Dai",
|
||
|
|
"image_path": "./data/IAM/m04/m04-030/m04-030-03-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a snake",
|
||
|
|
"gt_answers": "on",
|
||
|
|
"image_path": "./data/IAM/m04/m04-030/m04-030-03-11.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/m04/m04-030/m04-030-03-12.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "pavement",
|
||
|
|
"image_path": "./data/IAM/m04/m04-030/m04-030-03-13.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the name of the person",
|
||
|
|
"gt_answers": "knowing",
|
||
|
|
"image_path": "./data/IAM/m04/m04-030/m04-030-03-15.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "m",
|
||
|
|
"gt_answers": "in",
|
||
|
|
"image_path": "./data/IAM/m04/m04-030/m04-030-03-16.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "his",
|
||
|
|
"gt_answers": "his",
|
||
|
|
"image_path": "./data/IAM/m04/m04-030/m04-030-03-17.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "heart",
|
||
|
|
"gt_answers": "heart",
|
||
|
|
"image_path": "./data/IAM/m04/m04-030/m04-030-04-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "that",
|
||
|
|
"gt_answers": "that",
|
||
|
|
"image_path": "./data/IAM/m04/m04-030/m04-030-04-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "molydd",
|
||
|
|
"gt_answers": "Morfydd",
|
||
|
|
"image_path": "./data/IAM/m04/m04-030/m04-030-04-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "w",
|
||
|
|
"gt_answers": "was",
|
||
|
|
"image_path": "./data/IAM/m04/m04-030/m04-030-04-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "leaving",
|
||
|
|
"gt_answers": "leaving",
|
||
|
|
"image_path": "./data/IAM/m04/m04-030/m04-030-04-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "him",
|
||
|
|
"image_path": "./data/IAM/m04/m04-030/m04-030-04-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a snake",
|
||
|
|
"gt_answers": "as",
|
||
|
|
"image_path": "./data/IAM/m04/m04-030/m04-030-04-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "surely",
|
||
|
|
"image_path": "./data/IAM/m04/m04-030/m04-030-04-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "am",
|
||
|
|
"gt_answers": "as",
|
||
|
|
"image_path": "./data/IAM/m04/m04-030/m04-030-04-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "she",
|
||
|
|
"image_path": "./data/IAM/m04/m04-030/m04-030-04-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the name of person",
|
||
|
|
"gt_answers": "was",
|
||
|
|
"image_path": "./data/IAM/m04/m04-030/m04-030-04-10.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "harring",
|
||
|
|
"gt_answers": "leaving",
|
||
|
|
"image_path": "./data/IAM/m04/m04-030/m04-030-04-11.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "Dinas",
|
||
|
|
"image_path": "./data/IAM/m04/m04-030/m04-030-04-12.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "ae",
|
||
|
|
"gt_answers": "Are",
|
||
|
|
"image_path": "./data/IAM/k04/k04-017/k04-017-00-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "you",
|
||
|
|
"gt_answers": "you",
|
||
|
|
"image_path": "./data/IAM/k04/k04-017/k04-017-00-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "enjoying",
|
||
|
|
"gt_answers": "enjoying",
|
||
|
|
"image_path": "./data/IAM/k04/k04-017/k04-017-00-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "yourself",
|
||
|
|
"gt_answers": "yourself",
|
||
|
|
"image_path": "./data/IAM/k04/k04-017/k04-017-00-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "olma",
|
||
|
|
"gt_answers": "Olimpia",
|
||
|
|
"image_path": "./data/IAM/k04/k04-017/k04-017-00-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "she",
|
||
|
|
"gt_answers": "she",
|
||
|
|
"image_path": "./data/IAM/k04/k04-017/k04-017-00-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "asked",
|
||
|
|
"gt_answers": "asked",
|
||
|
|
"image_path": "./data/IAM/k04/k04-017/k04-017-01-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word person",
|
||
|
|
"gt_answers": "idly",
|
||
|
|
"image_path": "./data/IAM/k04/k04-017/k04-017-01-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "happy",
|
||
|
|
"gt_answers": "tapping",
|
||
|
|
"image_path": "./data/IAM/k04/k04-017/k04-017-01-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/k04/k04-017/k04-017-01-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a name",
|
||
|
|
"gt_answers": "girls",
|
||
|
|
"image_path": "./data/IAM/k04/k04-017/k04-017-01-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "cheer k",
|
||
|
|
"gt_answers": "cheek",
|
||
|
|
"image_path": "./data/IAM/k04/k04-017/k04-017-01-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "g h ly",
|
||
|
|
"gt_answers": "lightly",
|
||
|
|
"image_path": "./data/IAM/k04/k04-017/k04-017-01-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "with",
|
||
|
|
"image_path": "./data/IAM/k04/k04-017/k04-017-01-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "her",
|
||
|
|
"gt_answers": "her",
|
||
|
|
"image_path": "./data/IAM/k04/k04-017/k04-017-02-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "jane",
|
||
|
|
"gt_answers": "fan",
|
||
|
|
"image_path": "./data/IAM/k04/k04-017/k04-017-02-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "in",
|
||
|
|
"image_path": "./data/IAM/k04/k04-017/k04-017-02-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "passing",
|
||
|
|
"gt_answers": "passing",
|
||
|
|
"image_path": "./data/IAM/k04/k04-017/k04-017-02-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word but",
|
||
|
|
"gt_answers": "but",
|
||
|
|
"image_path": "./data/IAM/k04/k04-017/k04-017-02-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "she",
|
||
|
|
"gt_answers": "she",
|
||
|
|
"image_path": "./data/IAM/k04/k04-017/k04-017-02-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "d",
|
||
|
|
"gt_answers": "did",
|
||
|
|
"image_path": "./data/IAM/k04/k04-017/k04-017-02-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "not",
|
||
|
|
"gt_answers": "not",
|
||
|
|
"image_path": "./data/IAM/k04/k04-017/k04-017-02-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's name",
|
||
|
|
"gt_answers": "wait",
|
||
|
|
"image_path": "./data/IAM/k04/k04-017/k04-017-02-10.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "for",
|
||
|
|
"image_path": "./data/IAM/k04/k04-017/k04-017-02-11.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/k04/k04-017/k04-017-03-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the answer is yes",
|
||
|
|
"gt_answers": "answer",
|
||
|
|
"image_path": "./data/IAM/k04/k04-017/k04-017-03-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "dis furred",
|
||
|
|
"gt_answers": "Disturbed",
|
||
|
|
"image_path": "./data/IAM/k04/k04-017/k04-017-03-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "boy",
|
||
|
|
"gt_answers": "by",
|
||
|
|
"image_path": "./data/IAM/k04/k04-017/k04-017-03-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "gaugue",
|
||
|
|
"gt_answers": "vague",
|
||
|
|
"image_path": "./data/IAM/k04/k04-017/k04-017-03-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "uneasiness",
|
||
|
|
"image_path": "./data/IAM/k04/k04-017/k04-017-03-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "she",
|
||
|
|
"gt_answers": "she",
|
||
|
|
"image_path": "./data/IAM/k04/k04-017/k04-017-03-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "w person",
|
||
|
|
"gt_answers": "was",
|
||
|
|
"image_path": "./data/IAM/k04/k04-017/k04-017-04-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "i am wondering",
|
||
|
|
"gt_answers": "wondering",
|
||
|
|
"image_path": "./data/IAM/k04/k04-017/k04-017-04-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "weather",
|
||
|
|
"gt_answers": "whether",
|
||
|
|
"image_path": "./data/IAM/k04/k04-017/k04-017-04-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "olma",
|
||
|
|
"gt_answers": "Olimpia",
|
||
|
|
"image_path": "./data/IAM/k04/k04-017/k04-017-04-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "had",
|
||
|
|
"gt_answers": "had",
|
||
|
|
"image_path": "./data/IAM/k04/k04-017/k04-017-04-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "been",
|
||
|
|
"gt_answers": "been",
|
||
|
|
"image_path": "./data/IAM/k04/k04-017/k04-017-04-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "watching",
|
||
|
|
"gt_answers": "watching",
|
||
|
|
"image_path": "./data/IAM/k04/k04-017/k04-017-05-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "she",
|
||
|
|
"gt_answers": "her",
|
||
|
|
"image_path": "./data/IAM/k04/k04-017/k04-017-05-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "talk",
|
||
|
|
"gt_answers": "talk",
|
||
|
|
"image_path": "./data/IAM/k04/k04-017/k04-017-05-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "wf",
|
||
|
|
"gt_answers": "with",
|
||
|
|
"image_path": "./data/IAM/k04/k04-017/k04-017-05-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word orman",
|
||
|
|
"gt_answers": "Orsini",
|
||
|
|
"image_path": "./data/IAM/k04/k04-017/k04-017-05-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "surley",
|
||
|
|
"gt_answers": "Surely",
|
||
|
|
"image_path": "./data/IAM/k04/k04-017/k04-017-05-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "she",
|
||
|
|
"gt_answers": "she",
|
||
|
|
"image_path": "./data/IAM/k04/k04-017/k04-017-05-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word could",
|
||
|
|
"gt_answers": "could",
|
||
|
|
"image_path": "./data/IAM/k04/k04-017/k04-017-06-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "not",
|
||
|
|
"gt_answers": "not",
|
||
|
|
"image_path": "./data/IAM/k04/k04-017/k04-017-06-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "have",
|
||
|
|
"gt_answers": "have",
|
||
|
|
"image_path": "./data/IAM/k04/k04-017/k04-017-06-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "overhead d",
|
||
|
|
"gt_answers": "overheard",
|
||
|
|
"image_path": "./data/IAM/k04/k04-017/k04-017-06-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "anything is possible",
|
||
|
|
"gt_answers": "anything",
|
||
|
|
"image_path": "./data/IAM/k04/k04-017/k04-017-06-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "hey",
|
||
|
|
"gt_answers": "they",
|
||
|
|
"image_path": "./data/IAM/k04/k04-017/k04-017-06-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "said",
|
||
|
|
"image_path": "./data/IAM/k04/k04-017/k04-017-07-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the letter a",
|
||
|
|
"gt_answers": "A",
|
||
|
|
"image_path": "./data/IAM/k04/k04-017/k04-017-07-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "moments that matter",
|
||
|
|
"gt_answers": "moments",
|
||
|
|
"image_path": "./data/IAM/k04/k04-017/k04-017-07-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word reflection",
|
||
|
|
"gt_answers": "reflection",
|
||
|
|
"image_path": "./data/IAM/k04/k04-017/k04-017-07-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "reassured",
|
||
|
|
"gt_answers": "reassured",
|
||
|
|
"image_path": "./data/IAM/k04/k04-017/k04-017-07-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "she",
|
||
|
|
"gt_answers": "her",
|
||
|
|
"image_path": "./data/IAM/k04/k04-017/k04-017-07-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "om",
|
||
|
|
"gt_answers": "on",
|
||
|
|
"image_path": "./data/IAM/k04/k04-017/k04-017-08-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "that",
|
||
|
|
"gt_answers": "that",
|
||
|
|
"image_path": "./data/IAM/k04/k04-017/k04-017-08-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "point",
|
||
|
|
"gt_answers": "point",
|
||
|
|
"image_path": "./data/IAM/k04/k04-017/k04-017-08-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word for",
|
||
|
|
"gt_answers": "for",
|
||
|
|
"image_path": "./data/IAM/k04/k04-017/k04-017-08-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "she",
|
||
|
|
"gt_answers": "she",
|
||
|
|
"image_path": "./data/IAM/k04/k04-017/k04-017-08-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "was",
|
||
|
|
"gt_answers": "was",
|
||
|
|
"image_path": "./data/IAM/k04/k04-017/k04-017-08-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word person",
|
||
|
|
"gt_answers": "certain",
|
||
|
|
"image_path": "./data/IAM/k04/k04-017/k04-017-08-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "n",
|
||
|
|
"gt_answers": "no",
|
||
|
|
"image_path": "./data/IAM/k04/k04-017/k04-017-08-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "one",
|
||
|
|
"gt_answers": "one",
|
||
|
|
"image_path": "./data/IAM/k04/k04-017/k04-017-08-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "had",
|
||
|
|
"gt_answers": "had",
|
||
|
|
"image_path": "./data/IAM/k04/k04-017/k04-017-09-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "ben",
|
||
|
|
"gt_answers": "been",
|
||
|
|
"image_path": "./data/IAM/k04/k04-017/k04-017-09-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "stand by me",
|
||
|
|
"gt_answers": "standing",
|
||
|
|
"image_path": "./data/IAM/k04/k04-017/k04-017-09-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "near",
|
||
|
|
"gt_answers": "near",
|
||
|
|
"image_path": "./data/IAM/k04/k04-017/k04-017-09-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word alem",
|
||
|
|
"gt_answers": "them",
|
||
|
|
"image_path": "./data/IAM/k04/k04-017/k04-017-09-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "The",
|
||
|
|
"image_path": "./data/IAM/b06/b06-019/b06-019-00-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "process",
|
||
|
|
"gt_answers": "process",
|
||
|
|
"image_path": "./data/IAM/b06/b06-019/b06-019-00-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "has",
|
||
|
|
"gt_answers": "has",
|
||
|
|
"image_path": "./data/IAM/b06/b06-019/b06-019-00-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "been",
|
||
|
|
"gt_answers": "been",
|
||
|
|
"image_path": "./data/IAM/b06/b06-019/b06-019-00-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "tod",
|
||
|
|
"gt_answers": "too",
|
||
|
|
"image_path": "./data/IAM/b06/b06-019/b06-019-00-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "slow",
|
||
|
|
"gt_answers": "slow",
|
||
|
|
"image_path": "./data/IAM/b06/b06-019/b06-019-00-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "fer",
|
||
|
|
"gt_answers": "for",
|
||
|
|
"image_path": "./data/IAM/b06/b06-019/b06-019-00-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "herr",
|
||
|
|
"gt_answers": "Herr",
|
||
|
|
"image_path": "./data/IAM/b06/b06-019/b06-019-00-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "sraus",
|
||
|
|
"gt_answers": "Strauss",
|
||
|
|
"image_path": "./data/IAM/b06/b06-019/b06-019-01-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "and",
|
||
|
|
"gt_answers": "and",
|
||
|
|
"image_path": "./data/IAM/b06/b06-019/b06-019-01-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "last",
|
||
|
|
"gt_answers": "last",
|
||
|
|
"image_path": "./data/IAM/b06/b06-019/b06-019-01-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "month",
|
||
|
|
"gt_answers": "month",
|
||
|
|
"image_path": "./data/IAM/b06/b06-019/b06-019-01-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "he",
|
||
|
|
"gt_answers": "he",
|
||
|
|
"image_path": "./data/IAM/b06/b06-019/b06-019-01-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "attacked",
|
||
|
|
"gt_answers": "attacked",
|
||
|
|
"image_path": "./data/IAM/b06/b06-019/b06-019-01-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "britain",
|
||
|
|
"gt_answers": "Britain",
|
||
|
|
"image_path": "./data/IAM/b06/b06-019/b06-019-02-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "f",
|
||
|
|
"gt_answers": "for",
|
||
|
|
"image_path": "./data/IAM/b06/b06-019/b06-019-02-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "being",
|
||
|
|
"gt_answers": "being",
|
||
|
|
"image_path": "./data/IAM/b06/b06-019/b06-019-02-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "am",
|
||
|
|
"gt_answers": "an",
|
||
|
|
"image_path": "./data/IAM/b06/b06-019/b06-019-02-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "obstruct",
|
||
|
|
"gt_answers": "obstacle",
|
||
|
|
"image_path": "./data/IAM/b06/b06-019/b06-019-02-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "f",
|
||
|
|
"gt_answers": "for",
|
||
|
|
"image_path": "./data/IAM/b06/b06-019/b06-019-02-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "west",
|
||
|
|
"gt_answers": "West",
|
||
|
|
"image_path": "./data/IAM/b06/b06-019/b06-019-02-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "hermann's",
|
||
|
|
"gt_answers": "Germanys",
|
||
|
|
"image_path": "./data/IAM/b06/b06-019/b06-019-02-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "plans",
|
||
|
|
"gt_answers": "plans",
|
||
|
|
"image_path": "./data/IAM/b06/b06-019/b06-019-03-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "for",
|
||
|
|
"gt_answers": "for",
|
||
|
|
"image_path": "./data/IAM/b06/b06-019/b06-019-03-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a",
|
||
|
|
"gt_answers": "a",
|
||
|
|
"image_path": "./data/IAM/b06/b06-019/b06-019-03-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "unified",
|
||
|
|
"gt_answers": "unified",
|
||
|
|
"image_path": "./data/IAM/b06/b06-019/b06-019-03-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "supply",
|
||
|
|
"gt_answers": "supply",
|
||
|
|
"image_path": "./data/IAM/b06/b06-019/b06-019-03-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "apaparahus logo",
|
||
|
|
"gt_answers": "apparatus",
|
||
|
|
"image_path": "./data/IAM/b06/b06-019/b06-019-03-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "n",
|
||
|
|
"gt_answers": "in",
|
||
|
|
"image_path": "./data/IAM/b06/b06-019/b06-019-04-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "name",
|
||
|
|
"gt_answers": "Nato",
|
||
|
|
"image_path": "./data/IAM/b06/b06-019/b06-019-04-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a",
|
||
|
|
"gt_answers": "A",
|
||
|
|
"image_path": "./data/IAM/b06/b06-019/b06-019-04-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "supply",
|
||
|
|
"gt_answers": "supply",
|
||
|
|
"image_path": "./data/IAM/b06/b06-019/b06-019-04-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "apaparaus",
|
||
|
|
"gt_answers": "apparatus",
|
||
|
|
"image_path": "./data/IAM/b06/b06-019/b06-019-04-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "which",
|
||
|
|
"gt_answers": "which",
|
||
|
|
"image_path": "./data/IAM/b06/b06-019/b06-019-04-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "would do",
|
||
|
|
"gt_answers": "would",
|
||
|
|
"image_path": "./data/IAM/b06/b06-019/b06-019-04-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "link",
|
||
|
|
"gt_answers": "link",
|
||
|
|
"image_path": "./data/IAM/b06/b06-019/b06-019-05-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "together",
|
||
|
|
"gt_answers": "together",
|
||
|
|
"image_path": "./data/IAM/b06/b06-019/b06-019-05-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/b06/b06-019/b06-019-05-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "various",
|
||
|
|
"gt_answers": "various",
|
||
|
|
"image_path": "./data/IAM/b06/b06-019/b06-019-05-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "west",
|
||
|
|
"gt_answers": "West",
|
||
|
|
"image_path": "./data/IAM/b06/b06-019/b06-019-05-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "herman",
|
||
|
|
"gt_answers": "German",
|
||
|
|
"image_path": "./data/IAM/b06/b06-019/b06-019-05-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "facilities",
|
||
|
|
"gt_answers": "facilities",
|
||
|
|
"image_path": "./data/IAM/b06/b06-019/b06-019-06-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "b y",
|
||
|
|
"gt_answers": "By",
|
||
|
|
"image_path": "./data/IAM/b06/b06-019/b06-019-06-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the h",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/b06/b06-019/b06-019-06-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "end",
|
||
|
|
"gt_answers": "end",
|
||
|
|
"image_path": "./data/IAM/b06/b06-019/b06-019-06-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "off",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/b06/b06-019/b06-019-06-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "kh",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/b06/b06-019/b06-019-06-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "my year",
|
||
|
|
"gt_answers": "year",
|
||
|
|
"image_path": "./data/IAM/b06/b06-019/b06-019-06-10.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "there",
|
||
|
|
"gt_answers": "there",
|
||
|
|
"image_path": "./data/IAM/b06/b06-019/b06-019-06-11.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "will",
|
||
|
|
"gt_answers": "will",
|
||
|
|
"image_path": "./data/IAM/b06/b06-019/b06-019-07-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "be",
|
||
|
|
"gt_answers": "be",
|
||
|
|
"image_path": "./data/IAM/b06/b06-019/b06-019-07-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "11",
|
||
|
|
"image_path": "./data/IAM/b06/b06-019/b06-019-07-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "german",
|
||
|
|
"gt_answers": "German",
|
||
|
|
"image_path": "./data/IAM/b06/b06-019/b06-019-07-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "divisions",
|
||
|
|
"gt_answers": "divisions",
|
||
|
|
"image_path": "./data/IAM/b06/b06-019/b06-019-07-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "nin",
|
||
|
|
"gt_answers": "in",
|
||
|
|
"image_path": "./data/IAM/b06/b06-019/b06-019-07-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "wade",
|
||
|
|
"gt_answers": "Nato",
|
||
|
|
"image_path": "./data/IAM/b06/b06-019/b06-019-07-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "compared",
|
||
|
|
"gt_answers": "compared",
|
||
|
|
"image_path": "./data/IAM/b06/b06-019/b06-019-08-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "with",
|
||
|
|
"gt_answers": "with",
|
||
|
|
"image_path": "./data/IAM/b06/b06-019/b06-019-08-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "four",
|
||
|
|
"gt_answers": "four",
|
||
|
|
"image_path": "./data/IAM/b06/b06-019/b06-019-08-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "divisions",
|
||
|
|
"gt_answers": "divisions",
|
||
|
|
"image_path": "./data/IAM/b06/b06-019/b06-019-08-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "f",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/b06/b06-019/b06-019-08-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "british",
|
||
|
|
"gt_answers": "British",
|
||
|
|
"image_path": "./data/IAM/b06/b06-019/b06-019-09-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "krops",
|
||
|
|
"gt_answers": "troops",
|
||
|
|
"image_path": "./data/IAM/b06/b06-019/b06-019-09-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "buf",
|
||
|
|
"gt_answers": "But",
|
||
|
|
"image_path": "./data/IAM/j06/j06-018/j06-018-00-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "normal",
|
||
|
|
"gt_answers": "normal",
|
||
|
|
"image_path": "./data/IAM/j06/j06-018/j06-018-00-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "movements",
|
||
|
|
"gt_answers": "movements",
|
||
|
|
"image_path": "./data/IAM/j06/j06-018/j06-018-00-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "have",
|
||
|
|
"gt_answers": "have",
|
||
|
|
"image_path": "./data/IAM/j06/j06-018/j06-018-00-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "bee",
|
||
|
|
"gt_answers": "been",
|
||
|
|
"image_path": "./data/IAM/j06/j06-018/j06-018-00-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "repeal",
|
||
|
|
"gt_answers": "reported",
|
||
|
|
"image_path": "./data/IAM/j06/j06-018/j06-018-01-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "alo",
|
||
|
|
"gt_answers": "also",
|
||
|
|
"image_path": "./data/IAM/j06/j06-018/j06-018-01-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "for",
|
||
|
|
"image_path": "./data/IAM/j06/j06-018/j06-018-01-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "el",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/j06/j06-018/j06-018-01-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "same",
|
||
|
|
"image_path": "./data/IAM/j06/j06-018/j06-018-01-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "species",
|
||
|
|
"gt_answers": "species",
|
||
|
|
"image_path": "./data/IAM/j06/j06-018/j06-018-01-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "in",
|
||
|
|
"image_path": "./data/IAM/j06/j06-018/j06-018-01-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/j06/j06-018/j06-018-02-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "case",
|
||
|
|
"gt_answers": "case",
|
||
|
|
"image_path": "./data/IAM/j06/j06-018/j06-018-02-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/j06/j06-018/j06-018-02-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "ave",
|
||
|
|
"gt_answers": "five",
|
||
|
|
"image_path": "./data/IAM/j06/j06-018/j06-018-02-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "f",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/j06/j06-018/j06-018-02-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word she",
|
||
|
|
"gt_answers": "these",
|
||
|
|
"image_path": "./data/IAM/j06/j06-018/j06-018-02-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "examples",
|
||
|
|
"gt_answers": "examples",
|
||
|
|
"image_path": "./data/IAM/j06/j06-018/j06-018-02-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "andl",
|
||
|
|
"gt_answers": "and",
|
||
|
|
"image_path": "./data/IAM/j06/j06-018/j06-018-03-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "love",
|
||
|
|
"gt_answers": "for",
|
||
|
|
"image_path": "./data/IAM/j06/j06-018/j06-018-03-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a signature",
|
||
|
|
"gt_answers": "other",
|
||
|
|
"image_path": "./data/IAM/j06/j06-018/j06-018-03-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "specks",
|
||
|
|
"gt_answers": "species",
|
||
|
|
"image_path": "./data/IAM/j06/j06-018/j06-018-03-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a signature",
|
||
|
|
"gt_answers": "in",
|
||
|
|
"image_path": "./data/IAM/j06/j06-018/j06-018-03-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/j06/j06-018/j06-018-03-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "name",
|
||
|
|
"gt_answers": "same",
|
||
|
|
"image_path": "./data/IAM/j06/j06-018/j06-018-03-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "gene",
|
||
|
|
"gt_answers": "genera",
|
||
|
|
"image_path": "./data/IAM/j06/j06-018/j06-018-04-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "love",
|
||
|
|
"gt_answers": "for",
|
||
|
|
"image_path": "./data/IAM/j06/j06-018/j06-018-04-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "h",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/j06/j06-018/j06-018-04-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "remain",
|
||
|
|
"gt_answers": "remaining",
|
||
|
|
"image_path": "./data/IAM/j06/j06-018/j06-018-04-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "evadine",
|
||
|
|
"gt_answers": "Evadne",
|
||
|
|
"image_path": "./data/IAM/j06/j06-018/j06-018-04-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "da phi",
|
||
|
|
"gt_answers": "Daphnia",
|
||
|
|
"image_path": "./data/IAM/j06/j06-018/j06-018-05-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "and",
|
||
|
|
"gt_answers": "and",
|
||
|
|
"image_path": "./data/IAM/j06/j06-018/j06-018-05-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "dr ph m",
|
||
|
|
"gt_answers": "Diaptomus",
|
||
|
|
"image_path": "./data/IAM/j06/j06-018/j06-018-05-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "mary",
|
||
|
|
"gt_answers": "Many",
|
||
|
|
"image_path": "./data/IAM/j06/j06-018/j06-018-05-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "forms",
|
||
|
|
"gt_answers": "forms",
|
||
|
|
"image_path": "./data/IAM/j06/j06-018/j06-018-06-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "some names",
|
||
|
|
"gt_answers": "sometimes",
|
||
|
|
"image_path": "./data/IAM/j06/j06-018/j06-018-06-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "my rule",
|
||
|
|
"gt_answers": "migrate",
|
||
|
|
"image_path": "./data/IAM/j06/j06-018/j06-018-06-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "arl",
|
||
|
|
"gt_answers": "and",
|
||
|
|
"image_path": "./data/IAM/j06/j06-018/j06-018-06-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "on",
|
||
|
|
"image_path": "./data/IAM/j06/j06-018/j06-018-06-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "ole",
|
||
|
|
"gt_answers": "other",
|
||
|
|
"image_path": "./data/IAM/j06/j06-018/j06-018-06-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "ocasions",
|
||
|
|
"gt_answers": "occasions",
|
||
|
|
"image_path": "./data/IAM/j06/j06-018/j06-018-07-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "elo",
|
||
|
|
"gt_answers": "do",
|
||
|
|
"image_path": "./data/IAM/j06/j06-018/j06-018-07-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "no",
|
||
|
|
"gt_answers": "not",
|
||
|
|
"image_path": "./data/IAM/j06/j06-018/j06-018-07-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "grb",
|
||
|
|
"gt_answers": "but",
|
||
|
|
"image_path": "./data/IAM/j06/j06-018/j06-018-07-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a",
|
||
|
|
"gt_answers": "a",
|
||
|
|
"image_path": "./data/IAM/j06/j06-018/j06-018-07-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "fen",
|
||
|
|
"gt_answers": "few",
|
||
|
|
"image_path": "./data/IAM/j06/j06-018/j06-018-07-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "apcar",
|
||
|
|
"gt_answers": "appear",
|
||
|
|
"image_path": "./data/IAM/j06/j06-018/j06-018-08-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "h",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/j06/j06-018/j06-018-08-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "remah",
|
||
|
|
"gt_answers": "remain",
|
||
|
|
"image_path": "./data/IAM/j06/j06-018/j06-018-08-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "permanely",
|
||
|
|
"gt_answers": "permanently",
|
||
|
|
"image_path": "./data/IAM/j06/j06-018/j06-018-08-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a signature",
|
||
|
|
"gt_answers": "at",
|
||
|
|
"image_path": "./data/IAM/j06/j06-018/j06-018-08-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "one",
|
||
|
|
"gt_answers": "one",
|
||
|
|
"image_path": "./data/IAM/j06/j06-018/j06-018-09-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "lee",
|
||
|
|
"gt_answers": "level",
|
||
|
|
"image_path": "./data/IAM/j06/j06-018/j06-018-09-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "l s denning",
|
||
|
|
"gt_answers": "Considering",
|
||
|
|
"image_path": "./data/IAM/j06/j06-018/j06-018-09-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "he",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/j06/j06-018/j06-018-09-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "wilde spee",
|
||
|
|
"gt_answers": "widespread",
|
||
|
|
"image_path": "./data/IAM/j06/j06-018/j06-018-09-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word incence",
|
||
|
|
"gt_answers": "incidence",
|
||
|
|
"image_path": "./data/IAM/j06/j06-018/j06-018-10-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/j06/j06-018/j06-018-10-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "my radion",
|
||
|
|
"gt_answers": "migration",
|
||
|
|
"image_path": "./data/IAM/j06/j06-018/j06-018-10-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a signature",
|
||
|
|
"gt_answers": "in",
|
||
|
|
"image_path": "./data/IAM/j06/j06-018/j06-018-10-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "h",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/j06/j06-018/j06-018-10-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the cross",
|
||
|
|
"gt_answers": "It",
|
||
|
|
"image_path": "./data/IAM/a02/a02-116/a02-116-00-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "wa",
|
||
|
|
"gt_answers": "was",
|
||
|
|
"image_path": "./data/IAM/a02/a02-116/a02-116-00-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "mr",
|
||
|
|
"gt_answers": "Mr",
|
||
|
|
"image_path": "./data/IAM/a02/a02-116/a02-116-00-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "buter",
|
||
|
|
"gt_answers": "Butler",
|
||
|
|
"image_path": "./data/IAM/a02/a02-116/a02-116-00-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "who",
|
||
|
|
"gt_answers": "who",
|
||
|
|
"image_path": "./data/IAM/a02/a02-116/a02-116-00-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "authorized",
|
||
|
|
"gt_answers": "authorised",
|
||
|
|
"image_path": "./data/IAM/a02/a02-116/a02-116-00-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "achhan",
|
||
|
|
"gt_answers": "action",
|
||
|
|
"image_path": "./data/IAM/a02/a02-116/a02-116-00-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "which",
|
||
|
|
"gt_answers": "which",
|
||
|
|
"image_path": "./data/IAM/a02/a02-116/a02-116-00-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word ended",
|
||
|
|
"gt_answers": "ended",
|
||
|
|
"image_path": "./data/IAM/a02/a02-116/a02-116-01-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word yesterday",
|
||
|
|
"gt_answers": "yesterday",
|
||
|
|
"image_path": "./data/IAM/a02/a02-116/a02-116-01-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "n",
|
||
|
|
"gt_answers": "in",
|
||
|
|
"image_path": "./data/IAM/a02/a02-116/a02-116-01-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "three and two",
|
||
|
|
"gt_answers": "32",
|
||
|
|
"image_path": "./data/IAM/a02/a02-116/a02-116-01-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "members",
|
||
|
|
"gt_answers": "members",
|
||
|
|
"image_path": "./data/IAM/a02/a02-116/a02-116-01-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "f",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/a02/a02-116/a02-116-01-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/a02/a02-116/a02-116-01-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "committee",
|
||
|
|
"gt_answers": "Committee",
|
||
|
|
"image_path": "./data/IAM/a02/a02-116/a02-116-02-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "f",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/a02/a02-116/a02-116-02-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the number 100",
|
||
|
|
"gt_answers": "100",
|
||
|
|
"image_path": "./data/IAM/a02/a02-116/a02-116-02-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "being",
|
||
|
|
"gt_answers": "being",
|
||
|
|
"image_path": "./data/IAM/a02/a02-116/a02-116-02-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "inspired",
|
||
|
|
"gt_answers": "imprisoned",
|
||
|
|
"image_path": "./data/IAM/a02/a02-116/a02-116-02-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "far",
|
||
|
|
"gt_answers": "for",
|
||
|
|
"image_path": "./data/IAM/a02/a02-116/a02-116-02-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "inching",
|
||
|
|
"gt_answers": "inciting",
|
||
|
|
"image_path": "./data/IAM/a02/a02-116/a02-116-02-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the number 6",
|
||
|
|
"gt_answers": "a",
|
||
|
|
"image_path": "./data/IAM/a02/a02-116/a02-116-03-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "break",
|
||
|
|
"gt_answers": "breach",
|
||
|
|
"image_path": "./data/IAM/a02/a02-116/a02-116-03-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/a02/a02-116/a02-116-03-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/a02/a02-116/a02-116-03-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "peace",
|
||
|
|
"gt_answers": "peace",
|
||
|
|
"image_path": "./data/IAM/a02/a02-116/a02-116-03-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "The",
|
||
|
|
"image_path": "./data/IAM/a02/a02-116/a02-116-03-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "commitee's",
|
||
|
|
"gt_answers": "committees",
|
||
|
|
"image_path": "./data/IAM/a02/a02-116/a02-116-03-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "president",
|
||
|
|
"gt_answers": "president",
|
||
|
|
"image_path": "./data/IAM/a02/a02-116/a02-116-04-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "30 years old",
|
||
|
|
"gt_answers": "89yearold",
|
||
|
|
"image_path": "./data/IAM/a02/a02-116/a02-116-04-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "eal",
|
||
|
|
"gt_answers": "Earl",
|
||
|
|
"image_path": "./data/IAM/a02/a02-116/a02-116-04-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "russell",
|
||
|
|
"gt_answers": "Russell",
|
||
|
|
"image_path": "./data/IAM/a02/a02-116/a02-116-04-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "and",
|
||
|
|
"gt_answers": "and",
|
||
|
|
"image_path": "./data/IAM/a02/a02-116/a02-116-04-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "hs",
|
||
|
|
"gt_answers": "his",
|
||
|
|
"image_path": "./data/IAM/a02/a02-116/a02-116-05-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "6 years old",
|
||
|
|
"gt_answers": "61yearold",
|
||
|
|
"image_path": "./data/IAM/a02/a02-116/a02-116-05-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "nife",
|
||
|
|
"gt_answers": "wife",
|
||
|
|
"image_path": "./data/IAM/a02/a02-116/a02-116-05-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "we were",
|
||
|
|
"gt_answers": "were",
|
||
|
|
"image_path": "./data/IAM/a02/a02-116/a02-116-05-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "eachch",
|
||
|
|
"gt_answers": "each",
|
||
|
|
"image_path": "./data/IAM/a02/a02-116/a02-116-05-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "dale",
|
||
|
|
"gt_answers": "jailed",
|
||
|
|
"image_path": "./data/IAM/a02/a02-116/a02-116-05-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "far",
|
||
|
|
"gt_answers": "for",
|
||
|
|
"image_path": "./data/IAM/a02/a02-116/a02-116-05-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "a",
|
||
|
|
"image_path": "./data/IAM/a02/a02-116/a02-116-06-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "week",
|
||
|
|
"gt_answers": "week",
|
||
|
|
"image_path": "./data/IAM/a02/a02-116/a02-116-06-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "plays",
|
||
|
|
"gt_answers": "Playwrights",
|
||
|
|
"image_path": "./data/IAM/a02/a02-116/a02-116-06-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "arnd",
|
||
|
|
"gt_answers": "Arnold",
|
||
|
|
"image_path": "./data/IAM/a02/a02-116/a02-116-06-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "wecker",
|
||
|
|
"gt_answers": "Wesker",
|
||
|
|
"image_path": "./data/IAM/a02/a02-116/a02-116-06-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "The",
|
||
|
|
"image_path": "./data/IAM/a02/a02-116/a02-116-06-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "kitchen",
|
||
|
|
"gt_answers": "Kitchen",
|
||
|
|
"image_path": "./data/IAM/a02/a02-116/a02-116-06-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "love",
|
||
|
|
"gt_answers": "and",
|
||
|
|
"image_path": "./data/IAM/a02/a02-116/a02-116-07-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "robbie",
|
||
|
|
"gt_answers": "Robert",
|
||
|
|
"image_path": "./data/IAM/a02/a02-116/a02-116-07-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word bolt",
|
||
|
|
"gt_answers": "Bolt",
|
||
|
|
"image_path": "./data/IAM/a02/a02-116/a02-116-07-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "The",
|
||
|
|
"image_path": "./data/IAM/a02/a02-116/a02-116-07-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "flowers",
|
||
|
|
"gt_answers": "Flowering",
|
||
|
|
"image_path": "./data/IAM/a02/a02-116/a02-116-07-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "cherry",
|
||
|
|
"gt_answers": "Cherry",
|
||
|
|
"image_path": "./data/IAM/a02/a02-116/a02-116-07-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "eve",
|
||
|
|
"gt_answers": "were",
|
||
|
|
"image_path": "./data/IAM/a02/a02-116/a02-116-07-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "dale",
|
||
|
|
"gt_answers": "jailed",
|
||
|
|
"image_path": "./data/IAM/a02/a02-116/a02-116-08-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "fav",
|
||
|
|
"gt_answers": "for",
|
||
|
|
"image_path": "./data/IAM/a02/a02-116/a02-116-08-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the letter g",
|
||
|
|
"gt_answers": "a",
|
||
|
|
"image_path": "./data/IAM/a02/a02-116/a02-116-08-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "math",
|
||
|
|
"gt_answers": "month",
|
||
|
|
"image_path": "./data/IAM/a02/a02-116/a02-116-08-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "The",
|
||
|
|
"image_path": "./data/IAM/e07/e07-007/e07-007-00-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "biological",
|
||
|
|
"gt_answers": "biological",
|
||
|
|
"image_path": "./data/IAM/e07/e07-007/e07-007-00-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "control",
|
||
|
|
"gt_answers": "control",
|
||
|
|
"image_path": "./data/IAM/e07/e07-007/e07-007-00-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "f",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/e07/e07-007/e07-007-00-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "pests",
|
||
|
|
"gt_answers": "pests",
|
||
|
|
"image_path": "./data/IAM/e07/e07-007/e07-007-00-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "u",
|
||
|
|
"gt_answers": "in",
|
||
|
|
"image_path": "./data/IAM/e07/e07-007/e07-007-00-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/e07/e07-007/e07-007-00-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "garden",
|
||
|
|
"gt_answers": "garden",
|
||
|
|
"image_path": "./data/IAM/e07/e07-007/e07-007-00-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "m",
|
||
|
|
"gt_answers": "is",
|
||
|
|
"image_path": "./data/IAM/e07/e07-007/e07-007-01-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "not",
|
||
|
|
"gt_answers": "not",
|
||
|
|
"image_path": "./data/IAM/e07/e07-007/e07-007-01-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "new",
|
||
|
|
"gt_answers": "new",
|
||
|
|
"image_path": "./data/IAM/e07/e07-007/e07-007-01-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "to",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/e07/e07-007/e07-007-01-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/e07/e07-007/e07-007-01-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the american",
|
||
|
|
"gt_answers": "American",
|
||
|
|
"image_path": "./data/IAM/e07/e07-007/e07-007-01-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "gardener",
|
||
|
|
"gt_answers": "gardener",
|
||
|
|
"image_path": "./data/IAM/e07/e07-007/e07-007-01-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "lady",
|
||
|
|
"gt_answers": "Lady",
|
||
|
|
"image_path": "./data/IAM/e07/e07-007/e07-007-01-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "birds",
|
||
|
|
"gt_answers": "birds",
|
||
|
|
"image_path": "./data/IAM/e07/e07-007/e07-007-02-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "figure",
|
||
|
|
"gt_answers": "figure",
|
||
|
|
"image_path": "./data/IAM/e07/e07-007/e07-007-02-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "m",
|
||
|
|
"gt_answers": "in",
|
||
|
|
"image_path": "./data/IAM/e07/e07-007/e07-007-02-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "press",
|
||
|
|
"gt_answers": "press",
|
||
|
|
"image_path": "./data/IAM/e07/e07-007/e07-007-02-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "advertisements",
|
||
|
|
"gt_answers": "advertisements",
|
||
|
|
"image_path": "./data/IAM/e07/e07-007/e07-007-02-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "there",
|
||
|
|
"gt_answers": "their",
|
||
|
|
"image_path": "./data/IAM/e07/e07-007/e07-007-02-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "sale",
|
||
|
|
"gt_answers": "sale",
|
||
|
|
"image_path": "./data/IAM/e07/e07-007/e07-007-03-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "being",
|
||
|
|
"gt_answers": "being",
|
||
|
|
"image_path": "./data/IAM/e07/e07-007/e07-007-03-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "be",
|
||
|
|
"gt_answers": "by",
|
||
|
|
"image_path": "./data/IAM/e07/e07-007/e07-007-03-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/e07/e07-007/e07-007-03-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "gallon",
|
||
|
|
"gt_answers": "gallon",
|
||
|
|
"image_path": "./data/IAM/e07/e07-007/e07-007-03-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "praying",
|
||
|
|
"gt_answers": "Praying",
|
||
|
|
"image_path": "./data/IAM/e07/e07-007/e07-007-03-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "mantis",
|
||
|
|
"gt_answers": "mantis",
|
||
|
|
"image_path": "./data/IAM/e07/e07-007/e07-007-03-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "Mantis",
|
||
|
|
"image_path": "./data/IAM/e07/e07-007/e07-007-03-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "religion",
|
||
|
|
"gt_answers": "religiosa",
|
||
|
|
"image_path": "./data/IAM/e07/e07-007/e07-007-04-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "both",
|
||
|
|
"gt_answers": "both",
|
||
|
|
"image_path": "./data/IAM/e07/e07-007/e07-007-04-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "native",
|
||
|
|
"gt_answers": "native",
|
||
|
|
"image_path": "./data/IAM/e07/e07-007/e07-007-04-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "and",
|
||
|
|
"gt_answers": "and",
|
||
|
|
"image_path": "./data/IAM/e07/e07-007/e07-007-04-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "immigrant",
|
||
|
|
"gt_answers": "immigrant",
|
||
|
|
"image_path": "./data/IAM/e07/e07-007/e07-007-04-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "varities",
|
||
|
|
"gt_answers": "varieties",
|
||
|
|
"image_path": "./data/IAM/e07/e07-007/e07-007-04-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "are",
|
||
|
|
"gt_answers": "are",
|
||
|
|
"image_path": "./data/IAM/e07/e07-007/e07-007-05-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "used",
|
||
|
|
"gt_answers": "used",
|
||
|
|
"image_path": "./data/IAM/e07/e07-007/e07-007-05-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "by person",
|
||
|
|
"gt_answers": "by",
|
||
|
|
"image_path": "./data/IAM/e07/e07-007/e07-007-05-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "gardeners",
|
||
|
|
"gt_answers": "gardeners",
|
||
|
|
"image_path": "./data/IAM/e07/e07-007/e07-007-05-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "m",
|
||
|
|
"gt_answers": "in",
|
||
|
|
"image_path": "./data/IAM/e07/e07-007/e07-007-05-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "all",
|
||
|
|
"gt_answers": "all",
|
||
|
|
"image_path": "./data/IAM/e07/e07-007/e07-007-05-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "parts",
|
||
|
|
"gt_answers": "parts",
|
||
|
|
"image_path": "./data/IAM/e07/e07-007/e07-007-05-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "f",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/e07/e07-007/e07-007-05-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/e07/e07-007/e07-007-05-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "limited",
|
||
|
|
"gt_answers": "United",
|
||
|
|
"image_path": "./data/IAM/e07/e07-007/e07-007-05-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "states",
|
||
|
|
"gt_answers": "States",
|
||
|
|
"image_path": "./data/IAM/e07/e07-007/e07-007-06-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "advertisements",
|
||
|
|
"gt_answers": "Advertisements",
|
||
|
|
"image_path": "./data/IAM/e07/e07-007/e07-007-06-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "after",
|
||
|
|
"gt_answers": "offer",
|
||
|
|
"image_path": "./data/IAM/e07/e07-007/e07-007-06-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "eggs",
|
||
|
|
"gt_answers": "eggs",
|
||
|
|
"image_path": "./data/IAM/e07/e07-007/e07-007-06-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "f",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/e07/e07-007/e07-007-06-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "european",
|
||
|
|
"gt_answers": "European",
|
||
|
|
"image_path": "./data/IAM/e07/e07-007/e07-007-06-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "chinese",
|
||
|
|
"gt_answers": "Chinese",
|
||
|
|
"image_path": "./data/IAM/e07/e07-007/e07-007-07-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "o",
|
||
|
|
"gt_answers": "or",
|
||
|
|
"image_path": "./data/IAM/e07/e07-007/e07-007-07-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "native",
|
||
|
|
"gt_answers": "native",
|
||
|
|
"image_path": "./data/IAM/e07/e07-007/e07-007-07-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "forms",
|
||
|
|
"gt_answers": "forms",
|
||
|
|
"image_path": "./data/IAM/e07/e07-007/e07-007-07-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "f",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/e07/e07-007/e07-007-07-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "mantis",
|
||
|
|
"gt_answers": "mantis",
|
||
|
|
"image_path": "./data/IAM/e07/e07-007/e07-007-07-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a h",
|
||
|
|
"gt_answers": "It",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-00-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "has",
|
||
|
|
"gt_answers": "has",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-00-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a ring",
|
||
|
|
"gt_answers": "a",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-00-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "great",
|
||
|
|
"gt_answers": "great",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-00-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "look",
|
||
|
|
"gt_answers": "look",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-00-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "o",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-00-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "speed",
|
||
|
|
"gt_answers": "speed",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-00-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "and",
|
||
|
|
"gt_answers": "and",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-00-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "technical",
|
||
|
|
"gt_answers": "technical",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-00-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "fun",
|
||
|
|
"gt_answers": "fun",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-00-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "about",
|
||
|
|
"gt_answers": "about",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-01-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a cross",
|
||
|
|
"gt_answers": "it",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-01-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "o",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-01-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "enronous",
|
||
|
|
"gt_answers": "enormous",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-01-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "cinema magic",
|
||
|
|
"gt_answers": "cinematic",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-01-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "enjoyment",
|
||
|
|
"gt_answers": "enjoyment",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-01-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "angel",
|
||
|
|
"gt_answers": "and",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-01-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "above",
|
||
|
|
"gt_answers": "above",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-02-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "i ii",
|
||
|
|
"gt_answers": "all",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-02-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "f",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-02-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "cinematic",
|
||
|
|
"gt_answers": "cinematic",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-02-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "sense",
|
||
|
|
"gt_answers": "sense",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-02-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "much",
|
||
|
|
"gt_answers": "Much",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-02-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "f",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-02-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "it",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-02-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "has",
|
||
|
|
"gt_answers": "has",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-02-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word that",
|
||
|
|
"gt_answers": "that",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-03-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "air",
|
||
|
|
"gt_answers": "air",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-03-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the letter o",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-03-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "imprisoned",
|
||
|
|
"gt_answers": "improvisation",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-03-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a",
|
||
|
|
"gt_answers": "as",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-03-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a letter o",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-03-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "off the cuff",
|
||
|
|
"gt_answers": "offthecuff",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-03-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "ring",
|
||
|
|
"gt_answers": "living",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-04-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "th",
|
||
|
|
"gt_answers": "that",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-04-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "once",
|
||
|
|
"gt_answers": "once",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-04-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "semed",
|
||
|
|
"gt_answers": "seemed",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-04-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "so",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-04-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "new",
|
||
|
|
"gt_answers": "new",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-04-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "and",
|
||
|
|
"gt_answers": "and",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-04-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "so",
|
||
|
|
"gt_answers": "so",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-04-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word atache",
|
||
|
|
"gt_answers": "attractive",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-05-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "The",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-05-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "story",
|
||
|
|
"gt_answers": "story",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-05-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "no",
|
||
|
|
"gt_answers": "not",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-05-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word that",
|
||
|
|
"gt_answers": "that",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-05-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-05-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "story",
|
||
|
|
"gt_answers": "story",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-05-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the name of the person",
|
||
|
|
"gt_answers": "in",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-05-10.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the cross",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-05-11.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "sense",
|
||
|
|
"gt_answers": "sense",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-06-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "o",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-06-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word plot",
|
||
|
|
"gt_answers": "plot",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-06-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "matters",
|
||
|
|
"gt_answers": "matters",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-06-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "much",
|
||
|
|
"gt_answers": "much",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-06-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the letter b",
|
||
|
|
"gt_answers": "but",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-06-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "in",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-06-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the he",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-06-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "sense",
|
||
|
|
"gt_answers": "sense",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-06-10.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "f",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-06-11.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "siuation",
|
||
|
|
"gt_answers": "situation",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-07-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "and",
|
||
|
|
"gt_answers": "and",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-07-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the movement",
|
||
|
|
"gt_answers": "movement",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-07-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "it",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-07-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "matters",
|
||
|
|
"gt_answers": "matters",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-07-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a ring",
|
||
|
|
"gt_answers": "a",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-07-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "lot",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-07-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "s",
|
||
|
|
"gt_answers": "is",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-07-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word that",
|
||
|
|
"gt_answers": "that",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-08-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "f",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-08-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "a",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-08-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "man",
|
||
|
|
"gt_answers": "man",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-08-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "on",
|
||
|
|
"gt_answers": "on",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-08-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-08-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "run",
|
||
|
|
"gt_answers": "run",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-08-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "jean paul",
|
||
|
|
"gt_answers": "JeanPaul",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-08-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "bemonoch",
|
||
|
|
"gt_answers": "Belmondo",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-08-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "who",
|
||
|
|
"gt_answers": "who",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-09-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "spends",
|
||
|
|
"gt_answers": "spends",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-09-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "o",
|
||
|
|
"gt_answers": "a",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-09-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "few",
|
||
|
|
"gt_answers": "few",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-09-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word days",
|
||
|
|
"gt_answers": "days",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-09-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "th",
|
||
|
|
"gt_answers": "with",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-09-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "an",
|
||
|
|
"gt_answers": "an",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-09-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "american",
|
||
|
|
"gt_answers": "American",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-09-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "gisp",
|
||
|
|
"gt_answers": "girl",
|
||
|
|
"image_path": "./data/IAM/c02/c02-078/c02-078-09-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "Of",
|
||
|
|
"image_path": "./data/IAM/h07/h07-000/h07-000-00-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "over",
|
||
|
|
"gt_answers": "over",
|
||
|
|
"image_path": "./data/IAM/h07/h07-000/h07-000-00-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "0 000 000",
|
||
|
|
"gt_answers": "7000000",
|
||
|
|
"image_path": "./data/IAM/h07/h07-000/h07-000-00-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "square",
|
||
|
|
"gt_answers": "square",
|
||
|
|
"image_path": "./data/IAM/h07/h07-000/h07-000-00-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "feet",
|
||
|
|
"gt_answers": "feet",
|
||
|
|
"image_path": "./data/IAM/h07/h07-000/h07-000-00-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "l",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/h07/h07-000/h07-000-00-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "factory",
|
||
|
|
"gt_answers": "factory",
|
||
|
|
"image_path": "./data/IAM/h07/h07-000/h07-000-00-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word male",
|
||
|
|
"gt_answers": "space",
|
||
|
|
"image_path": "./data/IAM/h07/h07-000/h07-000-00-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "i built this",
|
||
|
|
"gt_answers": "built",
|
||
|
|
"image_path": "./data/IAM/h07/h07-000/h07-000-00-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "by person",
|
||
|
|
"gt_answers": "by",
|
||
|
|
"image_path": "./data/IAM/h07/h07-000/h07-000-00-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "tee",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/h07/h07-000/h07-000-00-10.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word development",
|
||
|
|
"gt_answers": "development",
|
||
|
|
"image_path": "./data/IAM/h07/h07-000/h07-000-01-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "corporations",
|
||
|
|
"image_path": "./data/IAM/h07/h07-000/h07-000-01-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "about",
|
||
|
|
"gt_answers": "about",
|
||
|
|
"image_path": "./data/IAM/h07/h07-000/h07-000-01-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the number 20",
|
||
|
|
"gt_answers": "20",
|
||
|
|
"image_path": "./data/IAM/h07/h07-000/h07-000-01-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "res",
|
||
|
|
"gt_answers": "per",
|
||
|
|
"image_path": "./data/IAM/h07/h07-000/h07-000-01-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "cent",
|
||
|
|
"gt_answers": "cent",
|
||
|
|
"image_path": "./data/IAM/h07/h07-000/h07-000-01-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "representatives",
|
||
|
|
"gt_answers": "represents",
|
||
|
|
"image_path": "./data/IAM/h07/h07-000/h07-000-01-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "extensions",
|
||
|
|
"image_path": "./data/IAM/h07/h07-000/h07-000-02-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word built",
|
||
|
|
"gt_answers": "built",
|
||
|
|
"image_path": "./data/IAM/h07/h07-000/h07-000-02-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "ariel",
|
||
|
|
"gt_answers": "after",
|
||
|
|
"image_path": "./data/IAM/h07/h07-000/h07-000-02-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "r e",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/h07/h07-000/h07-000-02-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "tims",
|
||
|
|
"gt_answers": "firms",
|
||
|
|
"image_path": "./data/IAM/h07/h07-000/h07-000-02-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word lod",
|
||
|
|
"gt_answers": "had",
|
||
|
|
"image_path": "./data/IAM/h07/h07-000/h07-000-02-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "be careful",
|
||
|
|
"gt_answers": "become",
|
||
|
|
"image_path": "./data/IAM/h07/h07-000/h07-000-02-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "established",
|
||
|
|
"image_path": "./data/IAM/h07/h07-000/h07-000-02-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "",
|
||
|
|
"gt_answers": "It",
|
||
|
|
"image_path": "./data/IAM/h07/h07-000/h07-000-03-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "",
|
||
|
|
"gt_answers": "is",
|
||
|
|
"image_path": "./data/IAM/h07/h07-000/h07-000-03-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "tee",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/h07/h07-000/h07-000-03-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "policy",
|
||
|
|
"gt_answers": "policy",
|
||
|
|
"image_path": "./data/IAM/h07/h07-000/h07-000-03-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a cross",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/h07/h07-000/h07-000-03-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a signature",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/h07/h07-000/h07-000-03-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the corporation",
|
||
|
|
"gt_answers": "corporations",
|
||
|
|
"image_path": "./data/IAM/h07/h07-000/h07-000-03-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/h07/h07-000/h07-000-03-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "charge",
|
||
|
|
"gt_answers": "charge",
|
||
|
|
"image_path": "./data/IAM/h07/h07-000/h07-000-03-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "full",
|
||
|
|
"gt_answers": "full",
|
||
|
|
"image_path": "./data/IAM/h07/h07-000/h07-000-03-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "rack",
|
||
|
|
"image_path": "./data/IAM/h07/h07-000/h07-000-03-10.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "rents",
|
||
|
|
"gt_answers": "rents",
|
||
|
|
"image_path": "./data/IAM/h07/h07-000/h07-000-04-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "or",
|
||
|
|
"image_path": "./data/IAM/h07/h07-000/h07-000-04-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "ground",
|
||
|
|
"gt_answers": "ground",
|
||
|
|
"image_path": "./data/IAM/h07/h07-000/h07-000-04-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "rents",
|
||
|
|
"gt_answers": "rents",
|
||
|
|
"image_path": "./data/IAM/h07/h07-000/h07-000-04-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "love",
|
||
|
|
"gt_answers": "for",
|
||
|
|
"image_path": "./data/IAM/h07/h07-000/h07-000-04-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word the",
|
||
|
|
"gt_answers": "their",
|
||
|
|
"image_path": "./data/IAM/h07/h07-000/h07-000-04-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's name",
|
||
|
|
"gt_answers": "factories",
|
||
|
|
"image_path": "./data/IAM/h07/h07-000/h07-000-04-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a number",
|
||
|
|
"gt_answers": "or",
|
||
|
|
"image_path": "./data/IAM/h07/h07-000/h07-000-04-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "industrial",
|
||
|
|
"gt_answers": "industrial",
|
||
|
|
"image_path": "./data/IAM/h07/h07-000/h07-000-04-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word person",
|
||
|
|
"gt_answers": "sites",
|
||
|
|
"image_path": "./data/IAM/h07/h07-000/h07-000-05-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "amanda",
|
||
|
|
"gt_answers": "and",
|
||
|
|
"image_path": "./data/IAM/h07/h07-000/h07-000-05-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "rents",
|
||
|
|
"gt_answers": "rents",
|
||
|
|
"image_path": "./data/IAM/h07/h07-000/h07-000-05-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word vary",
|
||
|
|
"gt_answers": "vary",
|
||
|
|
"image_path": "./data/IAM/h07/h07-000/h07-000-05-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "'considerably'",
|
||
|
|
"gt_answers": "considerably",
|
||
|
|
"image_path": "./data/IAM/h07/h07-000/h07-000-05-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "according to person",
|
||
|
|
"gt_answers": "according",
|
||
|
|
"image_path": "./data/IAM/h07/h07-000/h07-000-05-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "t",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/h07/h07-000/h07-000-05-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "lp",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/h07/h07-000/h07-000-05-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "delman",
|
||
|
|
"gt_answers": "demand",
|
||
|
|
"image_path": "./data/IAM/h07/h07-000/h07-000-05-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word for",
|
||
|
|
"gt_answers": "for",
|
||
|
|
"image_path": "./data/IAM/h07/h07-000/h07-000-06-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "factory",
|
||
|
|
"gt_answers": "factory",
|
||
|
|
"image_path": "./data/IAM/h07/h07-000/h07-000-06-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word accommodation",
|
||
|
|
"gt_answers": "accommodation",
|
||
|
|
"image_path": "./data/IAM/h07/h07-000/h07-000-06-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "megan",
|
||
|
|
"gt_answers": "Megan",
|
||
|
|
"image_path": "./data/IAM/m04/m04-131/m04-131-00-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "thomas",
|
||
|
|
"gt_answers": "Thomas",
|
||
|
|
"image_path": "./data/IAM/m04/m04-131/m04-131-00-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "nat",
|
||
|
|
"gt_answers": "sat",
|
||
|
|
"image_path": "./data/IAM/m04/m04-131/m04-131-00-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "right lyped",
|
||
|
|
"gt_answers": "tightlipped",
|
||
|
|
"image_path": "./data/IAM/m04/m04-131/m04-131-00-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word morning",
|
||
|
|
"gt_answers": "nursing",
|
||
|
|
"image_path": "./data/IAM/m04/m04-131/m04-131-00-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/m04/m04-131/m04-131-01-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "sleeping",
|
||
|
|
"gt_answers": "sleeping",
|
||
|
|
"image_path": "./data/IAM/m04/m04-131/m04-131-01-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "cadwallads",
|
||
|
|
"gt_answers": "Cadwallader",
|
||
|
|
"image_path": "./data/IAM/m04/m04-131/m04-131-01-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "driver",
|
||
|
|
"gt_answers": "Driver",
|
||
|
|
"image_path": "./data/IAM/m04/m04-131/m04-131-01-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "edward",
|
||
|
|
"gt_answers": "and",
|
||
|
|
"image_path": "./data/IAM/m04/m04-131/m04-131-01-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "conductor",
|
||
|
|
"gt_answers": "conductor",
|
||
|
|
"image_path": "./data/IAM/m04/m04-131/m04-131-02-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "i need",
|
||
|
|
"gt_answers": "peered",
|
||
|
|
"image_path": "./data/IAM/m04/m04-131/m04-131-02-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word head",
|
||
|
|
"gt_answers": "ahead",
|
||
|
|
"image_path": "./data/IAM/m04/m04-131/m04-131-02-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "into",
|
||
|
|
"gt_answers": "into",
|
||
|
|
"image_path": "./data/IAM/m04/m04-131/m04-131-02-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "nothingness",
|
||
|
|
"gt_answers": "nothingness",
|
||
|
|
"image_path": "./data/IAM/m04/m04-131/m04-131-02-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "o",
|
||
|
|
"gt_answers": "On",
|
||
|
|
"image_path": "./data/IAM/m04/m04-131/m04-131-03-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/m04/m04-131/m04-131-03-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "bach",
|
||
|
|
"gt_answers": "back",
|
||
|
|
"image_path": "./data/IAM/m04/m04-131/m04-131-03-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "eat",
|
||
|
|
"gt_answers": "seat",
|
||
|
|
"image_path": "./data/IAM/m04/m04-131/m04-131-03-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "tan",
|
||
|
|
"gt_answers": "Stan",
|
||
|
|
"image_path": "./data/IAM/m04/m04-131/m04-131-03-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "hayball",
|
||
|
|
"gt_answers": "Hayball",
|
||
|
|
"image_path": "./data/IAM/m04/m04-131/m04-131-03-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "embrace",
|
||
|
|
"gt_answers": "embraced",
|
||
|
|
"image_path": "./data/IAM/m04/m04-131/m04-131-03-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "his",
|
||
|
|
"image_path": "./data/IAM/m04/m04-131/m04-131-04-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "eth",
|
||
|
|
"gt_answers": "Eth",
|
||
|
|
"image_path": "./data/IAM/m04/m04-131/m04-131-04-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "morfyd",
|
||
|
|
"gt_answers": "Morfydd",
|
||
|
|
"image_path": "./data/IAM/m04/m04-131/m04-131-04-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "thought",
|
||
|
|
"gt_answers": "thought",
|
||
|
|
"image_path": "./data/IAM/m04/m04-131/m04-131-04-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "l",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/m04/m04-131/m04-131-04-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "Dai",
|
||
|
|
"image_path": "./data/IAM/m04/m04-131/m04-131-04-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "The",
|
||
|
|
"image_path": "./data/IAM/m04/m04-131/m04-131-04-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word rev",
|
||
|
|
"gt_answers": "Rev",
|
||
|
|
"image_path": "./data/IAM/m04/m04-131/m04-131-05-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "edwards",
|
||
|
|
"gt_answers": "Edwards",
|
||
|
|
"image_path": "./data/IAM/m04/m04-131/m04-131-05-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "standing",
|
||
|
|
"gt_answers": "standing",
|
||
|
|
"image_path": "./data/IAM/m04/m04-131/m04-131-05-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "et",
|
||
|
|
"gt_answers": "at",
|
||
|
|
"image_path": "./data/IAM/m04/m04-131/m04-131-05-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/m04/m04-131/m04-131-05-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "front",
|
||
|
|
"image_path": "./data/IAM/m04/m04-131/m04-131-05-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "looked",
|
||
|
|
"gt_answers": "looked",
|
||
|
|
"image_path": "./data/IAM/m04/m04-131/m04-131-06-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a",
|
||
|
|
"gt_answers": "at",
|
||
|
|
"image_path": "./data/IAM/m04/m04-131/m04-131-06-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "his",
|
||
|
|
"gt_answers": "his",
|
||
|
|
"image_path": "./data/IAM/m04/m04-131/m04-131-06-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "forlorn",
|
||
|
|
"gt_answers": "forlorn",
|
||
|
|
"image_path": "./data/IAM/m04/m04-131/m04-131-06-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "floch",
|
||
|
|
"gt_answers": "flock",
|
||
|
|
"image_path": "./data/IAM/m04/m04-131/m04-131-06-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "what",
|
||
|
|
"gt_answers": "What",
|
||
|
|
"image_path": "./data/IAM/m04/m04-131/m04-131-06-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "about",
|
||
|
|
"gt_answers": "about",
|
||
|
|
"image_path": "./data/IAM/m04/m04-131/m04-131-07-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "a",
|
||
|
|
"image_path": "./data/IAM/m04/m04-131/m04-131-07-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "verse",
|
||
|
|
"image_path": "./data/IAM/m04/m04-131/m04-131-07-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "or",
|
||
|
|
"image_path": "./data/IAM/m04/m04-131/m04-131-07-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "two",
|
||
|
|
"gt_answers": "two",
|
||
|
|
"image_path": "./data/IAM/m04/m04-131/m04-131-07-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "l",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/m04/m04-131/m04-131-07-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "luv",
|
||
|
|
"gt_answers": "Cwm",
|
||
|
|
"image_path": "./data/IAM/m04/m04-131/m04-131-07-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the name rhondda",
|
||
|
|
"gt_answers": "Rhondda",
|
||
|
|
"image_path": "./data/IAM/m04/m04-131/m04-131-07-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "he",
|
||
|
|
"gt_answers": "he",
|
||
|
|
"image_path": "./data/IAM/m04/m04-131/m04-131-08-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "nugget",
|
||
|
|
"gt_answers": "suggested",
|
||
|
|
"image_path": "./data/IAM/m04/m04-131/m04-131-08-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "hopelessly",
|
||
|
|
"gt_answers": "hopefully",
|
||
|
|
"image_path": "./data/IAM/m04/m04-131/m04-131-08-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "o",
|
||
|
|
"gt_answers": "Or",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-00-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "h",
|
||
|
|
"gt_answers": "it",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-00-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "worms",
|
||
|
|
"gt_answers": "is",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-00-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "possible",
|
||
|
|
"gt_answers": "possible",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-00-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "fah",
|
||
|
|
"gt_answers": "that",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-00-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word to",
|
||
|
|
"gt_answers": "our",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-00-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "selgya",
|
||
|
|
"gt_answers": "sugya",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-00-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "couture",
|
||
|
|
"gt_answers": "contains",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-00-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word he",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-00-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "original",
|
||
|
|
"gt_answers": "original",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-01-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "question mark",
|
||
|
|
"gt_answers": "question",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-01-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "audi",
|
||
|
|
"gt_answers": "and",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-01-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "short answer",
|
||
|
|
"gt_answers": "answer",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-01-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "aud",
|
||
|
|
"gt_answers": "and",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-01-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word alus",
|
||
|
|
"gt_answers": "this",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-01-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "is",
|
||
|
|
"gt_answers": "is",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-01-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word geeted",
|
||
|
|
"gt_answers": "quoted",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-01-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "nin",
|
||
|
|
"gt_answers": "in",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-02-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-02-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "h e s",
|
||
|
|
"gt_answers": "other",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-02-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "segura",
|
||
|
|
"gt_answers": "sugya",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-02-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "this",
|
||
|
|
"gt_answers": "This",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-02-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "car",
|
||
|
|
"gt_answers": "can",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-02-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "love",
|
||
|
|
"gt_answers": "be",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-02-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "supported",
|
||
|
|
"gt_answers": "supported",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-02-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "by",
|
||
|
|
"gt_answers": "by",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-02-10.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "fire",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-02-11.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "tusse",
|
||
|
|
"gt_answers": "use",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-03-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "f",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-03-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-03-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "expression",
|
||
|
|
"gt_answers": "expression",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-03-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "booz",
|
||
|
|
"gt_answers": "Boaz",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-03-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "dil",
|
||
|
|
"gt_answers": "did",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-03-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "f",
|
||
|
|
"gt_answers": "it",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-03-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "f",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-03-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "lis",
|
||
|
|
"gt_answers": "his",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-03-10.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "ow",
|
||
|
|
"gt_answers": "own",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-03-11.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "accor",
|
||
|
|
"gt_answers": "accord",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-04-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "but",
|
||
|
|
"gt_answers": "but",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-04-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "there",
|
||
|
|
"gt_answers": "there",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-04-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "was",
|
||
|
|
"gt_answers": "was",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-04-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "one",
|
||
|
|
"gt_answers": "no",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-04-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "approval",
|
||
|
|
"gt_answers": "approval",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-04-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-04-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "lis",
|
||
|
|
"gt_answers": "his",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-04-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "achor",
|
||
|
|
"gt_answers": "action",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-04-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the name person",
|
||
|
|
"gt_answers": "in",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-05-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "heaven",
|
||
|
|
"gt_answers": "Heaven",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-05-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "h",
|
||
|
|
"gt_answers": "in",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-05-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word bolt",
|
||
|
|
"gt_answers": "both",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-05-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "seayth",
|
||
|
|
"gt_answers": "sugyoth",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-05-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "suel",
|
||
|
|
"gt_answers": "Such",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-05-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "ar",
|
||
|
|
"gt_answers": "an",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-05-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "expression",
|
||
|
|
"gt_answers": "expression",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-06-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "applas",
|
||
|
|
"gt_answers": "appears",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-06-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "h",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-06-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "love",
|
||
|
|
"gt_answers": "have",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-06-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "beer",
|
||
|
|
"gt_answers": "been",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-06-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word flawed",
|
||
|
|
"gt_answers": "framed",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-06-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "in",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-06-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "report",
|
||
|
|
"gt_answers": "response",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-07-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "se",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-07-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the letter o",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-07-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "pachuca",
|
||
|
|
"gt_answers": "particular",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-07-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "point",
|
||
|
|
"gt_answers": "point",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-07-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "f",
|
||
|
|
"gt_answers": "at",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-07-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "c issue",
|
||
|
|
"gt_answers": "issue",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-07-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "here",
|
||
|
|
"gt_answers": "here",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-07-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word weather",
|
||
|
|
"gt_answers": "whether",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-08-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "he",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-08-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "heavenly",
|
||
|
|
"gt_answers": "Heavenly",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-08-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "court",
|
||
|
|
"gt_answers": "Court",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-08-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "confounded",
|
||
|
|
"gt_answers": "concurred",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-08-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "in",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-09-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-09-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "decision",
|
||
|
|
"gt_answers": "decision",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-09-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "f",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-09-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "lue",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-09-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word louis",
|
||
|
|
"gt_answers": "human",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-09-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "court house",
|
||
|
|
"gt_answers": "court",
|
||
|
|
"image_path": "./data/IAM/d07/d07-096/d07-096-09-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "mr",
|
||
|
|
"gt_answers": "Mr",
|
||
|
|
"image_path": "./data/IAM/a06/a06-095/a06-095-00-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "maccullian",
|
||
|
|
"gt_answers": "Macmillan",
|
||
|
|
"image_path": "./data/IAM/a06/a06-095/a06-095-00-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "vaid",
|
||
|
|
"gt_answers": "said",
|
||
|
|
"image_path": "./data/IAM/a06/a06-095/a06-095-00-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "hp",
|
||
|
|
"gt_answers": "we",
|
||
|
|
"image_path": "./data/IAM/a06/a06-095/a06-095-00-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "could",
|
||
|
|
"gt_answers": "could",
|
||
|
|
"image_path": "./data/IAM/a06/a06-095/a06-095-00-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "b",
|
||
|
|
"gt_answers": "be",
|
||
|
|
"image_path": "./data/IAM/a06/a06-095/a06-095-00-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "worc",
|
||
|
|
"gt_answers": "more",
|
||
|
|
"image_path": "./data/IAM/a06/a06-095/a06-095-00-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "help",
|
||
|
|
"gt_answers": "help",
|
||
|
|
"image_path": "./data/IAM/a06/a06-095/a06-095-00-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "o",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/a06/a06-095/a06-095-01-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/a06/a06-095/a06-095-01-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "communea 14th",
|
||
|
|
"gt_answers": "Commonwealth",
|
||
|
|
"image_path": "./data/IAM/a06/a06-095/a06-095-01-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "thought",
|
||
|
|
"gt_answers": "through",
|
||
|
|
"image_path": "./data/IAM/a06/a06-095/a06-095-01-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "h c",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/a06/a06-095/a06-095-01-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "john wright",
|
||
|
|
"gt_answers": "strength",
|
||
|
|
"image_path": "./data/IAM/a06/a06-095/a06-095-02-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "we",
|
||
|
|
"gt_answers": "we",
|
||
|
|
"image_path": "./data/IAM/a06/a06-095/a06-095-02-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "world",
|
||
|
|
"gt_answers": "would",
|
||
|
|
"image_path": "./data/IAM/a06/a06-095/a06-095-02-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "gii",
|
||
|
|
"gt_answers": "gain",
|
||
|
|
"image_path": "./data/IAM/a06/a06-095/a06-095-02-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a signature",
|
||
|
|
"gt_answers": "in",
|
||
|
|
"image_path": "./data/IAM/a06/a06-095/a06-095-02-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "thee",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/a06/a06-095/a06-095-02-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "gounou",
|
||
|
|
"gt_answers": "Common",
|
||
|
|
"image_path": "./data/IAM/a06/a06-095/a06-095-02-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "marie f",
|
||
|
|
"gt_answers": "Market",
|
||
|
|
"image_path": "./data/IAM/a06/a06-095/a06-095-03-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "han",
|
||
|
|
"gt_answers": "than",
|
||
|
|
"image_path": "./data/IAM/a06/a06-095/a06-095-03-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "by",
|
||
|
|
"gt_answers": "by",
|
||
|
|
"image_path": "./data/IAM/a06/a06-095/a06-095-03-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "yolanda",
|
||
|
|
"gt_answers": "isolation",
|
||
|
|
"image_path": "./data/IAM/a06/a06-095/a06-095-03-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "h c",
|
||
|
|
"gt_answers": "He",
|
||
|
|
"image_path": "./data/IAM/a06/a06-095/a06-095-03-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "olaf",
|
||
|
|
"gt_answers": "paid",
|
||
|
|
"image_path": "./data/IAM/a06/a06-095/a06-095-03-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "frida",
|
||
|
|
"gt_answers": "tribute",
|
||
|
|
"image_path": "./data/IAM/a06/a06-095/a06-095-04-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "o",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/a06/a06-095/a06-095-04-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "he",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/a06/a06-095/a06-095-04-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "development warf",
|
||
|
|
"gt_answers": "development",
|
||
|
|
"image_path": "./data/IAM/a06/a06-095/a06-095-04-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "no",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/a06/a06-095/a06-095-04-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/a06/a06-095/a06-095-04-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "couny",
|
||
|
|
"gt_answers": "Common",
|
||
|
|
"image_path": "./data/IAM/a06/a06-095/a06-095-05-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "marlet",
|
||
|
|
"gt_answers": "Market",
|
||
|
|
"image_path": "./data/IAM/a06/a06-095/a06-095-05-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "The",
|
||
|
|
"image_path": "./data/IAM/a06/a06-095/a06-095-05-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "community",
|
||
|
|
"gt_answers": "Community",
|
||
|
|
"image_path": "./data/IAM/a06/a06-095/a06-095-05-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "common law",
|
||
|
|
"gt_answers": "Common",
|
||
|
|
"image_path": "./data/IAM/a06/a06-095/a06-095-06-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "market",
|
||
|
|
"gt_answers": "Market",
|
||
|
|
"image_path": "./data/IAM/a06/a06-095/a06-095-06-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "har",
|
||
|
|
"gt_answers": "has",
|
||
|
|
"image_path": "./data/IAM/a06/a06-095/a06-095-06-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "in park",
|
||
|
|
"gt_answers": "imparted",
|
||
|
|
"image_path": "./data/IAM/a06/a06-095/a06-095-06-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "al",
|
||
|
|
"gt_answers": "an",
|
||
|
|
"image_path": "./data/IAM/a06/a06-095/a06-095-06-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the name of the person",
|
||
|
|
"gt_answers": "impetus",
|
||
|
|
"image_path": "./data/IAM/a06/a06-095/a06-095-07-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "aud",
|
||
|
|
"gt_answers": "and",
|
||
|
|
"image_path": "./data/IAM/a06/a06-095/a06-095-07-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "m",
|
||
|
|
"gt_answers": "an",
|
||
|
|
"image_path": "./data/IAM/a06/a06-095/a06-095-07-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word eco",
|
||
|
|
"gt_answers": "economic",
|
||
|
|
"image_path": "./data/IAM/a06/a06-095/a06-095-07-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "growly",
|
||
|
|
"gt_answers": "growth",
|
||
|
|
"image_path": "./data/IAM/a06/a06-095/a06-095-07-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "o",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/a06/a06-095/a06-095-08-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "The",
|
||
|
|
"image_path": "./data/IAM/a06/a06-095/a06-095-08-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "sx",
|
||
|
|
"gt_answers": "Six",
|
||
|
|
"image_path": "./data/IAM/a06/a06-095/a06-095-08-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "above",
|
||
|
|
"gt_answers": "Above",
|
||
|
|
"image_path": "./data/IAM/a06/a06-095/a06-095-08-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "al",
|
||
|
|
"gt_answers": "all",
|
||
|
|
"image_path": "./data/IAM/a06/a06-095/a06-095-08-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "it",
|
||
|
|
"image_path": "./data/IAM/a06/a06-095/a06-095-08-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word s",
|
||
|
|
"gt_answers": "is",
|
||
|
|
"image_path": "./data/IAM/a06/a06-095/a06-095-08-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a signature",
|
||
|
|
"gt_answers": "an",
|
||
|
|
"image_path": "./data/IAM/a06/a06-095/a06-095-08-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "ideg",
|
||
|
|
"gt_answers": "idea",
|
||
|
|
"image_path": "./data/IAM/a06/a06-095/a06-095-08-10.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "which",
|
||
|
|
"gt_answers": "which",
|
||
|
|
"image_path": "./data/IAM/a06/a06-095/a06-095-09-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "har",
|
||
|
|
"gt_answers": "has",
|
||
|
|
"image_path": "./data/IAM/a06/a06-095/a06-095-09-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "gripped",
|
||
|
|
"gt_answers": "gripped",
|
||
|
|
"image_path": "./data/IAM/a06/a06-095/a06-095-09-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "me",
|
||
|
|
"gt_answers": "mens",
|
||
|
|
"image_path": "./data/IAM/a06/a06-095/a06-095-09-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "wilfred edwards",
|
||
|
|
"gt_answers": "minds",
|
||
|
|
"image_path": "./data/IAM/a06/a06-095/a06-095-09-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "h",
|
||
|
|
"gt_answers": "he",
|
||
|
|
"image_path": "./data/IAM/a06/a06-095/a06-095-10-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "jad",
|
||
|
|
"gt_answers": "said",
|
||
|
|
"image_path": "./data/IAM/a06/a06-095/a06-095-10-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "ted",
|
||
|
|
"gt_answers": "Fred",
|
||
|
|
"image_path": "./data/IAM/l01/l01-129/l01-129-00-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "fixed",
|
||
|
|
"gt_answers": "fixed",
|
||
|
|
"image_path": "./data/IAM/l01/l01-129/l01-129-00-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "hs",
|
||
|
|
"gt_answers": "his",
|
||
|
|
"image_path": "./data/IAM/l01/l01-129/l01-129-00-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "small",
|
||
|
|
"gt_answers": "small",
|
||
|
|
"image_path": "./data/IAM/l01/l01-129/l01-129-00-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "e 7",
|
||
|
|
"gt_answers": "eyes",
|
||
|
|
"image_path": "./data/IAM/l01/l01-129/l01-129-00-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "on",
|
||
|
|
"image_path": "./data/IAM/l01/l01-129/l01-129-00-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "lis",
|
||
|
|
"gt_answers": "his",
|
||
|
|
"image_path": "./data/IAM/l01/l01-129/l01-129-00-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "brothers",
|
||
|
|
"gt_answers": "brother",
|
||
|
|
"image_path": "./data/IAM/l01/l01-129/l01-129-00-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "econned",
|
||
|
|
"gt_answers": "beckoned",
|
||
|
|
"image_path": "./data/IAM/l01/l01-129/l01-129-00-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "him",
|
||
|
|
"image_path": "./data/IAM/l01/l01-129/l01-129-01-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "o",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/l01/l01-129/l01-129-01-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "stand",
|
||
|
|
"gt_answers": "stand",
|
||
|
|
"image_path": "./data/IAM/l01/l01-129/l01-129-01-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "closer",
|
||
|
|
"gt_answers": "closer",
|
||
|
|
"image_path": "./data/IAM/l01/l01-129/l01-129-01-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a snake",
|
||
|
|
"gt_answers": "and",
|
||
|
|
"image_path": "./data/IAM/l01/l01-129/l01-129-01-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "whispered",
|
||
|
|
"gt_answers": "whispered",
|
||
|
|
"image_path": "./data/IAM/l01/l01-129/l01-129-01-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's name",
|
||
|
|
"gt_answers": "into",
|
||
|
|
"image_path": "./data/IAM/l01/l01-129/l01-129-01-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "h s",
|
||
|
|
"gt_answers": "his",
|
||
|
|
"image_path": "./data/IAM/l01/l01-129/l01-129-01-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word person",
|
||
|
|
"gt_answers": "ear",
|
||
|
|
"image_path": "./data/IAM/l01/l01-129/l01-129-02-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/l01/l01-129/l01-129-02-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "most",
|
||
|
|
"image_path": "./data/IAM/l01/l01-129/l01-129-02-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "unexpected",
|
||
|
|
"gt_answers": "unexpected",
|
||
|
|
"image_path": "./data/IAM/l01/l01-129/l01-129-02-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "word",
|
||
|
|
"gt_answers": "words",
|
||
|
|
"image_path": "./data/IAM/l01/l01-129/l01-129-02-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "nick",
|
||
|
|
"gt_answers": "Nick",
|
||
|
|
"image_path": "./data/IAM/l01/l01-129/l01-129-02-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "bad",
|
||
|
|
"gt_answers": "had",
|
||
|
|
"image_path": "./data/IAM/l01/l01-129/l01-129-02-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word person",
|
||
|
|
"gt_answers": "ever",
|
||
|
|
"image_path": "./data/IAM/l01/l01-129/l01-129-03-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word head",
|
||
|
|
"gt_answers": "heard",
|
||
|
|
"image_path": "./data/IAM/l01/l01-129/l01-129-03-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "I",
|
||
|
|
"image_path": "./data/IAM/l01/l01-129/l01-129-03-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "i love you",
|
||
|
|
"gt_answers": "m",
|
||
|
|
"image_path": "./data/IAM/l01/l01-129/l01-129-03-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word going",
|
||
|
|
"gt_answers": "going",
|
||
|
|
"image_path": "./data/IAM/l01/l01-129/l01-129-03-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/l01/l01-129/l01-129-03-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "get",
|
||
|
|
"image_path": "./data/IAM/l01/l01-129/l01-129-03-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "self love",
|
||
|
|
"gt_answers": "myself",
|
||
|
|
"image_path": "./data/IAM/l01/l01-129/l01-129-03-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "married",
|
||
|
|
"gt_answers": "married",
|
||
|
|
"image_path": "./data/IAM/l01/l01-129/l01-129-04-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "ted",
|
||
|
|
"gt_answers": "Fred",
|
||
|
|
"image_path": "./data/IAM/l01/l01-129/l01-129-04-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "bia",
|
||
|
|
"gt_answers": "said",
|
||
|
|
"image_path": "./data/IAM/l01/l01-129/l01-129-04-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the person",
|
||
|
|
"gt_answers": "The",
|
||
|
|
"image_path": "./data/IAM/l01/l01-129/l01-129-04-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word person",
|
||
|
|
"gt_answers": "next",
|
||
|
|
"image_path": "./data/IAM/l01/l01-129/l01-129-04-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "day",
|
||
|
|
"gt_answers": "day",
|
||
|
|
"image_path": "./data/IAM/l01/l01-129/l01-129-04-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "he",
|
||
|
|
"gt_answers": "he",
|
||
|
|
"image_path": "./data/IAM/l01/l01-129/l01-129-04-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "brough",
|
||
|
|
"gt_answers": "brought",
|
||
|
|
"image_path": "./data/IAM/l01/l01-129/l01-129-05-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "home",
|
||
|
|
"gt_answers": "home",
|
||
|
|
"image_path": "./data/IAM/l01/l01-129/l01-129-05-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "a",
|
||
|
|
"image_path": "./data/IAM/l01/l01-129/l01-129-05-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word woman",
|
||
|
|
"gt_answers": "woman",
|
||
|
|
"image_path": "./data/IAM/l01/l01-129/l01-129-05-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a snake",
|
||
|
|
"gt_answers": "in",
|
||
|
|
"image_path": "./data/IAM/l01/l01-129/l01-129-05-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "her",
|
||
|
|
"image_path": "./data/IAM/l01/l01-129/l01-129-05-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "middle",
|
||
|
|
"gt_answers": "middle",
|
||
|
|
"image_path": "./data/IAM/l01/l01-129/l01-129-05-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word person",
|
||
|
|
"gt_answers": "thirties",
|
||
|
|
"image_path": "./data/IAM/l01/l01-129/l01-129-06-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word and",
|
||
|
|
"gt_answers": "and",
|
||
|
|
"image_path": "./data/IAM/l01/l01-129/l01-129-06-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word dot",
|
||
|
|
"gt_answers": "took",
|
||
|
|
"image_path": "./data/IAM/l01/l01-129/l01-129-06-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "her",
|
||
|
|
"gt_answers": "her",
|
||
|
|
"image_path": "./data/IAM/l01/l01-129/l01-129-06-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "straight",
|
||
|
|
"gt_answers": "straight",
|
||
|
|
"image_path": "./data/IAM/l01/l01-129/l01-129-06-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a and b",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/l01/l01-129/l01-129-06-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "l s",
|
||
|
|
"gt_answers": "his",
|
||
|
|
"image_path": "./data/IAM/l01/l01-129/l01-129-06-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "mother's day",
|
||
|
|
"gt_answers": "mothers",
|
||
|
|
"image_path": "./data/IAM/l01/l01-129/l01-129-07-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "bedside",
|
||
|
|
"gt_answers": "bedside",
|
||
|
|
"image_path": "./data/IAM/l01/l01-129/l01-129-07-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "2 insteal",
|
||
|
|
"gt_answers": "Instead",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-00-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-00-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "kings",
|
||
|
|
"gt_answers": "kings",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-00-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "will",
|
||
|
|
"gt_answers": "will",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-00-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "be main",
|
||
|
|
"gt_answers": "remain",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-00-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "m",
|
||
|
|
"gt_answers": "in",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-00-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "london",
|
||
|
|
"gt_answers": "London",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-00-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "chanel",
|
||
|
|
"gt_answers": "and",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-01-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "wait",
|
||
|
|
"gt_answers": "wait",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-01-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "o",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-01-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "cheer",
|
||
|
|
"gt_answers": "hear",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-01-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-01-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the conference is over",
|
||
|
|
"gt_answers": "conferences",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-01-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "proposals",
|
||
|
|
"gt_answers": "proposals",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-01-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word the",
|
||
|
|
"gt_answers": "Then",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-02-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "their",
|
||
|
|
"gt_answers": "their",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-02-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "reviews",
|
||
|
|
"gt_answers": "views",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-02-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "will",
|
||
|
|
"gt_answers": "will",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-02-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "e",
|
||
|
|
"gt_answers": "be",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-02-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "trans mittel",
|
||
|
|
"gt_answers": "transmitted",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-02-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "back",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-02-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "mr person",
|
||
|
|
"gt_answers": "MR",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-03-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "butler",
|
||
|
|
"gt_answers": "BUTLER",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-03-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-03-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "home",
|
||
|
|
"gt_answers": "Home",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-03-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "secretary",
|
||
|
|
"gt_answers": "Secretary",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-03-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "hass",
|
||
|
|
"gt_answers": "has",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-03-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "decl",
|
||
|
|
"gt_answers": "decided",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-03-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the name person",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-04-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "meet",
|
||
|
|
"gt_answers": "meet",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-04-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "read on",
|
||
|
|
"gt_answers": "headon",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-04-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-04-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "roger ebert",
|
||
|
|
"gt_answers": "biggest",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-04-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "challenge",
|
||
|
|
"gt_answers": "challenge",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-04-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "t",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-04-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "government",
|
||
|
|
"gt_answers": "Government",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-05-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word authority",
|
||
|
|
"gt_answers": "authority",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-05-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "yet",
|
||
|
|
"gt_answers": "yet",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-05-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "one sent",
|
||
|
|
"gt_answers": "presented",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-05-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "by",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-05-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the name person",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-05-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "ban the bond",
|
||
|
|
"gt_answers": "BantheBomb",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-06-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "demonstator ss",
|
||
|
|
"gt_answers": "demonstrators",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-06-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "police",
|
||
|
|
"gt_answers": "Police",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-06-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "dewey",
|
||
|
|
"gt_answers": "leave",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-06-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "hes",
|
||
|
|
"gt_answers": "has",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-07-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "been",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-07-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "can called",
|
||
|
|
"gt_answers": "cancelled",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-07-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "cnc",
|
||
|
|
"gt_answers": "and",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-07-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "secret",
|
||
|
|
"gt_answers": "secret",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-07-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the name of the person",
|
||
|
|
"gt_answers": "plans",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-07-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word preserved",
|
||
|
|
"gt_answers": "prepared",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-07-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "o",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-08-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "deal",
|
||
|
|
"gt_answers": "deal",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-08-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "with",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-08-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "he",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-08-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "mass",
|
||
|
|
"gt_answers": "mass",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-08-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "sit down",
|
||
|
|
"gt_answers": "sitdown",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-08-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "amy",
|
||
|
|
"gt_answers": "rally",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-08-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "planned",
|
||
|
|
"gt_answers": "planned",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-08-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "for",
|
||
|
|
"gt_answers": "for",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-09-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "sunday",
|
||
|
|
"gt_answers": "Sunday",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-09-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "m",
|
||
|
|
"gt_answers": "in",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-09-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "parlement square",
|
||
|
|
"gt_answers": "Parliamentsquare",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-09-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "by person",
|
||
|
|
"gt_answers": "by",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-09-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-09-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "committee",
|
||
|
|
"gt_answers": "Committee",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-10-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "cf",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-10-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "love",
|
||
|
|
"gt_answers": "100",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-10-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-10-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "anti nuclear",
|
||
|
|
"gt_answers": "antinuclear",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-10-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word person",
|
||
|
|
"gt_answers": "arms",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-10-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word group",
|
||
|
|
"gt_answers": "group",
|
||
|
|
"image_path": "./data/IAM/a02/a02-111/a02-111-11-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "sedar",
|
||
|
|
"gt_answers": "Senator",
|
||
|
|
"image_path": "./data/IAM/a01/a01-049x/a01-049x-00-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "aver",
|
||
|
|
"gt_answers": "Allen",
|
||
|
|
"image_path": "./data/IAM/a01/a01-049x/a01-049x-00-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word elender",
|
||
|
|
"gt_answers": "Ellender",
|
||
|
|
"image_path": "./data/IAM/a01/a01-049x/a01-049x-00-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a cross",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/a01/a01-049x/a01-049x-00-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "louis aua",
|
||
|
|
"gt_answers": "Louisiana",
|
||
|
|
"image_path": "./data/IAM/a01/a01-049x/a01-049x-00-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word open",
|
||
|
|
"gt_answers": "sparked",
|
||
|
|
"image_path": "./data/IAM/a01/a01-049x/a01-049x-01-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "off",
|
||
|
|
"image_path": "./data/IAM/a01/a01-049x/a01-049x-01-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "me",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/a01/a01-049x/a01-049x-01-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "opposition",
|
||
|
|
"image_path": "./data/IAM/a01/a01-049x/a01-049x-01-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "e g",
|
||
|
|
"gt_answers": "by",
|
||
|
|
"image_path": "./data/IAM/a01/a01-049x/a01-049x-01-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "belgium",
|
||
|
|
"gt_answers": "telling",
|
||
|
|
"image_path": "./data/IAM/a01/a01-049x/a01-049x-01-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the letter o",
|
||
|
|
"gt_answers": "a",
|
||
|
|
"image_path": "./data/IAM/a01/a01-049x/a01-049x-01-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the voice",
|
||
|
|
"gt_answers": "television",
|
||
|
|
"image_path": "./data/IAM/a01/a01-049x/a01-049x-02-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "amelie c",
|
||
|
|
"gt_answers": "audience",
|
||
|
|
"image_path": "./data/IAM/a01/a01-049x/a01-049x-02-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "it",
|
||
|
|
"image_path": "./data/IAM/a01/a01-049x/a01-049x-02-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "aw",
|
||
|
|
"gt_answers": "was",
|
||
|
|
"image_path": "./data/IAM/a01/a01-049x/a01-049x-02-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word courant",
|
||
|
|
"gt_answers": "current",
|
||
|
|
"image_path": "./data/IAM/a01/a01-049x/a01-049x-02-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "washington",
|
||
|
|
"gt_answers": "Washington",
|
||
|
|
"image_path": "./data/IAM/a01/a01-049x/a01-049x-02-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word person",
|
||
|
|
"gt_answers": "gossip",
|
||
|
|
"image_path": "./data/IAM/a01/a01-049x/a01-049x-03-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word f",
|
||
|
|
"gt_answers": "that",
|
||
|
|
"image_path": "./data/IAM/a01/a01-049x/a01-049x-03-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "weave",
|
||
|
|
"gt_answers": "Weaver",
|
||
|
|
"image_path": "./data/IAM/a01/a01-049x/a01-049x-03-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "once",
|
||
|
|
"gt_answers": "once",
|
||
|
|
"image_path": "./data/IAM/a01/a01-049x/a01-049x-03-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the name of the person",
|
||
|
|
"gt_answers": "had",
|
||
|
|
"image_path": "./data/IAM/a01/a01-049x/a01-049x-03-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "communism",
|
||
|
|
"gt_answers": "Communist",
|
||
|
|
"image_path": "./data/IAM/a01/a01-049x/a01-049x-03-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "al fah",
|
||
|
|
"gt_answers": "affilia",
|
||
|
|
"image_path": "./data/IAM/a01/a01-049x/a01-049x-03-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "tions",
|
||
|
|
"image_path": "./data/IAM/a01/a01-049x/a01-049x-04-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "The",
|
||
|
|
"image_path": "./data/IAM/a01/a01-049x/a01-049x-04-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "seville",
|
||
|
|
"gt_answers": "Senate",
|
||
|
|
"image_path": "./data/IAM/a01/a01-049x/a01-049x-04-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "banns",
|
||
|
|
"gt_answers": "Banking",
|
||
|
|
"image_path": "./data/IAM/a01/a01-049x/a01-049x-04-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the name of person",
|
||
|
|
"gt_answers": "Committee",
|
||
|
|
"image_path": "./data/IAM/a01/a01-049x/a01-049x-04-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "which",
|
||
|
|
"image_path": "./data/IAM/a01/a01-049x/a01-049x-04-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "is",
|
||
|
|
"image_path": "./data/IAM/a01/a01-049x/a01-049x-05-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "i love you",
|
||
|
|
"gt_answers": "headed",
|
||
|
|
"image_path": "./data/IAM/a01/a01-049x/a01-049x-05-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "b d",
|
||
|
|
"gt_answers": "by",
|
||
|
|
"image_path": "./data/IAM/a01/a01-049x/a01-049x-05-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the name of the person",
|
||
|
|
"gt_answers": "another",
|
||
|
|
"image_path": "./data/IAM/a01/a01-049x/a01-049x-05-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "southern",
|
||
|
|
"gt_answers": "Southern",
|
||
|
|
"image_path": "./data/IAM/a01/a01-049x/a01-049x-05-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "sedar",
|
||
|
|
"gt_answers": "Senator",
|
||
|
|
"image_path": "./data/IAM/a01/a01-049x/a01-049x-05-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word w",
|
||
|
|
"gt_answers": "Willis",
|
||
|
|
"image_path": "./data/IAM/a01/a01-049x/a01-049x-06-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "robert scott",
|
||
|
|
"gt_answers": "Robertson",
|
||
|
|
"image_path": "./data/IAM/a01/a01-049x/a01-049x-06-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/a01/a01-049x/a01-049x-06-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "virginia",
|
||
|
|
"gt_answers": "Virginia",
|
||
|
|
"image_path": "./data/IAM/a01/a01-049x/a01-049x-06-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "met",
|
||
|
|
"image_path": "./data/IAM/a01/a01-049x/a01-049x-06-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word today",
|
||
|
|
"gt_answers": "today",
|
||
|
|
"image_path": "./data/IAM/a01/a01-049x/a01-049x-06-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a wave",
|
||
|
|
"gt_answers": "in",
|
||
|
|
"image_path": "./data/IAM/a01/a01-049x/a01-049x-06-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "closed",
|
||
|
|
"gt_answers": "closed",
|
||
|
|
"image_path": "./data/IAM/a01/a01-049x/a01-049x-07-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "session",
|
||
|
|
"image_path": "./data/IAM/a01/a01-049x/a01-049x-07-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a heart",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/a01/a01-049x/a01-049x-07-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word person",
|
||
|
|
"gt_answers": "discuss",
|
||
|
|
"image_path": "./data/IAM/a01/a01-049x/a01-049x-07-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "weaver",
|
||
|
|
"gt_answers": "Weavers",
|
||
|
|
"image_path": "./data/IAM/a01/a01-049x/a01-049x-07-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "i am an apprentice",
|
||
|
|
"gt_answers": "appointment",
|
||
|
|
"image_path": "./data/IAM/a01/a01-049x/a01-049x-07-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "selver",
|
||
|
|
"gt_answers": "Senator",
|
||
|
|
"image_path": "./data/IAM/a01/a01-049x/a01-049x-08-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "robert edwards",
|
||
|
|
"gt_answers": "Robertson",
|
||
|
|
"image_path": "./data/IAM/a01/a01-049x/a01-049x-08-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "dale",
|
||
|
|
"gt_answers": "later",
|
||
|
|
"image_path": "./data/IAM/a01/a01-049x/a01-049x-08-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "closed",
|
||
|
|
"gt_answers": "disclosed",
|
||
|
|
"image_path": "./data/IAM/a01/a01-049x/a01-049x-08-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "e",
|
||
|
|
"gt_answers": "he",
|
||
|
|
"image_path": "./data/IAM/a01/a01-049x/a01-049x-08-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "day",
|
||
|
|
"gt_answers": "had",
|
||
|
|
"image_path": "./data/IAM/a01/a01-049x/a01-049x-08-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "deut",
|
||
|
|
"gt_answers": "sent",
|
||
|
|
"image_path": "./data/IAM/a01/a01-049x/a01-049x-08-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "a",
|
||
|
|
"image_path": "./data/IAM/a01/a01-049x/a01-049x-08-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "i love you",
|
||
|
|
"gt_answers": "letter",
|
||
|
|
"image_path": "./data/IAM/a01/a01-049x/a01-049x-09-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "0 0",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/a01/a01-049x/a01-049x-09-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "m",
|
||
|
|
"gt_answers": "Mr",
|
||
|
|
"image_path": "./data/IAM/a01/a01-049x/a01-049x-09-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word kennedy",
|
||
|
|
"gt_answers": "Kennedy",
|
||
|
|
"image_path": "./data/IAM/a01/a01-049x/a01-049x-09-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "i love you",
|
||
|
|
"gt_answers": "saying",
|
||
|
|
"image_path": "./data/IAM/a01/a01-049x/a01-049x-09-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "w",
|
||
|
|
"gt_answers": "he",
|
||
|
|
"image_path": "./data/IAM/a01/a01-049x/a01-049x-09-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word day",
|
||
|
|
"gt_answers": "had",
|
||
|
|
"image_path": "./data/IAM/a01/a01-049x/a01-049x-09-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "canceled",
|
||
|
|
"gt_answers": "received",
|
||
|
|
"image_path": "./data/IAM/a01/a01-049x/a01-049x-09-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "several",
|
||
|
|
"image_path": "./data/IAM/a01/a01-049x/a01-049x-09-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "oclumb",
|
||
|
|
"gt_answers": "complaints",
|
||
|
|
"image_path": "./data/IAM/a01/a01-049x/a01-049x-10-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "about",
|
||
|
|
"gt_answers": "about",
|
||
|
|
"image_path": "./data/IAM/a01/a01-049x/a01-049x-10-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word weave",
|
||
|
|
"gt_answers": "Weavers",
|
||
|
|
"image_path": "./data/IAM/a01/a01-049x/a01-049x-10-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "loyalty",
|
||
|
|
"gt_answers": "loyalty",
|
||
|
|
"image_path": "./data/IAM/a01/a01-049x/a01-049x-10-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "she",
|
||
|
|
"gt_answers": "She",
|
||
|
|
"image_path": "./data/IAM/k04/k04-075/k04-075-00-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "noah",
|
||
|
|
"gt_answers": "had",
|
||
|
|
"image_path": "./data/IAM/k04/k04-075/k04-075-00-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "perhapas",
|
||
|
|
"gt_answers": "perhaps",
|
||
|
|
"image_path": "./data/IAM/k04/k04-075/k04-075-00-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "demanded",
|
||
|
|
"gt_answers": "demanded",
|
||
|
|
"image_path": "./data/IAM/k04/k04-075/k04-075-00-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "400",
|
||
|
|
"gt_answers": "too",
|
||
|
|
"image_path": "./data/IAM/k04/k04-075/k04-075-00-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "much",
|
||
|
|
"gt_answers": "much",
|
||
|
|
"image_path": "./data/IAM/k04/k04-075/k04-075-00-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "The",
|
||
|
|
"image_path": "./data/IAM/k04/k04-075/k04-075-00-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "kiss",
|
||
|
|
"gt_answers": "kiss",
|
||
|
|
"image_path": "./data/IAM/k04/k04-075/k04-075-01-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "in",
|
||
|
|
"image_path": "./data/IAM/k04/k04-075/k04-075-01-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/k04/k04-075/k04-075-01-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "garden",
|
||
|
|
"gt_answers": "garden",
|
||
|
|
"image_path": "./data/IAM/k04/k04-075/k04-075-01-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "hoax",
|
||
|
|
"gt_answers": "had",
|
||
|
|
"image_path": "./data/IAM/k04/k04-075/k04-075-01-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "plumbed",
|
||
|
|
"gt_answers": "plumbed",
|
||
|
|
"image_path": "./data/IAM/k04/k04-075/k04-075-01-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "um sus",
|
||
|
|
"gt_answers": "unsus",
|
||
|
|
"image_path": "./data/IAM/k04/k04-075/k04-075-01-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "expected",
|
||
|
|
"gt_answers": "pected",
|
||
|
|
"image_path": "./data/IAM/k04/k04-075/k04-075-02-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "depths",
|
||
|
|
"gt_answers": "depths",
|
||
|
|
"image_path": "./data/IAM/k04/k04-075/k04-075-02-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "m",
|
||
|
|
"gt_answers": "in",
|
||
|
|
"image_path": "./data/IAM/k04/k04-075/k04-075-02-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "her",
|
||
|
|
"gt_answers": "her",
|
||
|
|
"image_path": "./data/IAM/k04/k04-075/k04-075-02-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word and",
|
||
|
|
"gt_answers": "and",
|
||
|
|
"image_path": "./data/IAM/k04/k04-075/k04-075-02-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "she",
|
||
|
|
"gt_answers": "she",
|
||
|
|
"image_path": "./data/IAM/k04/k04-075/k04-075-02-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "know",
|
||
|
|
"gt_answers": "knew",
|
||
|
|
"image_path": "./data/IAM/k04/k04-075/k04-075-02-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "mot",
|
||
|
|
"gt_answers": "that",
|
||
|
|
"image_path": "./data/IAM/k04/k04-075/k04-075-02-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "f",
|
||
|
|
"gt_answers": "if",
|
||
|
|
"image_path": "./data/IAM/k04/k04-075/k04-075-03-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "she",
|
||
|
|
"gt_answers": "she",
|
||
|
|
"image_path": "./data/IAM/k04/k04-075/k04-075-03-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "yeled",
|
||
|
|
"gt_answers": "yielded",
|
||
|
|
"image_path": "./data/IAM/k04/k04-075/k04-075-03-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "0",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/k04/k04-075/k04-075-03-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "hi",
|
||
|
|
"gt_answers": "him",
|
||
|
|
"image_path": "./data/IAM/k04/k04-075/k04-075-03-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "er",
|
||
|
|
"gt_answers": "her",
|
||
|
|
"image_path": "./data/IAM/k04/k04-075/k04-075-03-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "psio m p",
|
||
|
|
"gt_answers": "passion",
|
||
|
|
"image_path": "./data/IAM/k04/k04-075/k04-075-03-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "could",
|
||
|
|
"gt_answers": "could",
|
||
|
|
"image_path": "./data/IAM/k04/k04-075/k04-075-03-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "im deed",
|
||
|
|
"gt_answers": "indeed",
|
||
|
|
"image_path": "./data/IAM/k04/k04-075/k04-075-04-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "match",
|
||
|
|
"gt_answers": "match",
|
||
|
|
"image_path": "./data/IAM/k04/k04-075/k04-075-04-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "his",
|
||
|
|
"gt_answers": "his",
|
||
|
|
"image_path": "./data/IAM/k04/k04-075/k04-075-04-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "every",
|
||
|
|
"gt_answers": "Every",
|
||
|
|
"image_path": "./data/IAM/k04/k04-075/k04-075-04-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "instinct",
|
||
|
|
"gt_answers": "instinct",
|
||
|
|
"image_path": "./data/IAM/k04/k04-075/k04-075-04-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "urged",
|
||
|
|
"gt_answers": "urged",
|
||
|
|
"image_path": "./data/IAM/k04/k04-075/k04-075-04-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "her",
|
||
|
|
"gt_answers": "her",
|
||
|
|
"image_path": "./data/IAM/k04/k04-075/k04-075-04-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the number four",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/k04/k04-075/k04-075-04-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "supreme de",
|
||
|
|
"gt_answers": "surrender",
|
||
|
|
"image_path": "./data/IAM/k04/k04-075/k04-075-05-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "j for",
|
||
|
|
"gt_answers": "for",
|
||
|
|
"image_path": "./data/IAM/k04/k04-075/k04-075-05-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "there",
|
||
|
|
"gt_answers": "there",
|
||
|
|
"image_path": "./data/IAM/k04/k04-075/k04-075-05-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "ubs",
|
||
|
|
"gt_answers": "was",
|
||
|
|
"image_path": "./data/IAM/k04/k04-075/k04-075-05-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word met",
|
||
|
|
"gt_answers": "that",
|
||
|
|
"image_path": "./data/IAM/k04/k04-075/k04-075-05-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "in",
|
||
|
|
"image_path": "./data/IAM/k04/k04-075/k04-075-05-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "his",
|
||
|
|
"gt_answers": "his",
|
||
|
|
"image_path": "./data/IAM/k04/k04-075/k04-075-05-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word nurture",
|
||
|
|
"gt_answers": "nature",
|
||
|
|
"image_path": "./data/IAM/k04/k04-075/k04-075-05-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the number 40",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/k04/k04-075/k04-075-06-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "which",
|
||
|
|
"gt_answers": "which",
|
||
|
|
"image_path": "./data/IAM/k04/k04-075/k04-075-06-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "her",
|
||
|
|
"gt_answers": "her",
|
||
|
|
"image_path": "./data/IAM/k04/k04-075/k04-075-06-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "ow",
|
||
|
|
"gt_answers": "own",
|
||
|
|
"image_path": "./data/IAM/k04/k04-075/k04-075-06-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "hd",
|
||
|
|
"gt_answers": "had",
|
||
|
|
"image_path": "./data/IAM/k04/k04-075/k04-075-06-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "respon dead",
|
||
|
|
"gt_answers": "responded",
|
||
|
|
"image_path": "./data/IAM/k04/k04-075/k04-075-06-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "xs",
|
||
|
|
"gt_answers": "as",
|
||
|
|
"image_path": "./data/IAM/k04/k04-075/k04-075-06-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "t",
|
||
|
|
"gt_answers": "it",
|
||
|
|
"image_path": "./data/IAM/k04/k04-075/k04-075-06-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "h d",
|
||
|
|
"gt_answers": "had",
|
||
|
|
"image_path": "./data/IAM/k04/k04-075/k04-075-07-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "reprimanded",
|
||
|
|
"gt_answers": "responded",
|
||
|
|
"image_path": "./data/IAM/k04/k04-075/k04-075-07-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the number 40",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/k04/k04-075/k04-075-07-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "m",
|
||
|
|
"gt_answers": "no",
|
||
|
|
"image_path": "./data/IAM/k04/k04-075/k04-075-07-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "other",
|
||
|
|
"gt_answers": "other",
|
||
|
|
"image_path": "./data/IAM/k04/k04-075/k04-075-07-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "human",
|
||
|
|
"gt_answers": "human",
|
||
|
|
"image_path": "./data/IAM/k04/k04-075/k04-075-07-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "beng",
|
||
|
|
"gt_answers": "being",
|
||
|
|
"image_path": "./data/IAM/k04/k04-075/k04-075-07-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "butt",
|
||
|
|
"gt_answers": "But",
|
||
|
|
"image_path": "./data/IAM/c04/c04-017/c04-017-00-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "6 000",
|
||
|
|
"gt_answers": "6000",
|
||
|
|
"image_path": "./data/IAM/c04/c04-017/c04-017-00-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "willie",
|
||
|
|
"gt_answers": "miles",
|
||
|
|
"image_path": "./data/IAM/c04/c04-017/c04-017-00-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "seems",
|
||
|
|
"image_path": "./data/IAM/c04/c04-017/c04-017-00-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "x",
|
||
|
|
"gt_answers": "a",
|
||
|
|
"image_path": "./data/IAM/c04/c04-017/c04-017-00-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "check",
|
||
|
|
"gt_answers": "heck",
|
||
|
|
"image_path": "./data/IAM/c04/c04-017/c04-017-00-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a signature",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/c04/c04-017/c04-017-00-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "a",
|
||
|
|
"image_path": "./data/IAM/c04/c04-017/c04-017-00-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "way",
|
||
|
|
"gt_answers": "way",
|
||
|
|
"image_path": "./data/IAM/c04/c04-017/c04-017-00-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/c04/c04-017/c04-017-00-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "j",
|
||
|
|
"gt_answers": "go",
|
||
|
|
"image_path": "./data/IAM/c04/c04-017/c04-017-00-10.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "love",
|
||
|
|
"gt_answers": "for",
|
||
|
|
"image_path": "./data/IAM/c04/c04-017/c04-017-01-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "k",
|
||
|
|
"gt_answers": "a",
|
||
|
|
"image_path": "./data/IAM/c04/c04-017/c04-017-01-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "new",
|
||
|
|
"image_path": "./data/IAM/c04/c04-017/c04-017-01-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "h",
|
||
|
|
"gt_answers": "hit",
|
||
|
|
"image_path": "./data/IAM/c04/c04-017/c04-017-01-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "sony",
|
||
|
|
"gt_answers": "song",
|
||
|
|
"image_path": "./data/IAM/c04/c04-017/c04-017-01-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "she",
|
||
|
|
"gt_answers": "He",
|
||
|
|
"image_path": "./data/IAM/c04/c04-017/c04-017-01-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the you",
|
||
|
|
"gt_answers": "might",
|
||
|
|
"image_path": "./data/IAM/c04/c04-017/c04-017-01-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "be",
|
||
|
|
"gt_answers": "be",
|
||
|
|
"image_path": "./data/IAM/c04/c04-017/c04-017-01-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "well",
|
||
|
|
"gt_answers": "well",
|
||
|
|
"image_path": "./data/IAM/c04/c04-017/c04-017-01-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "adverseed",
|
||
|
|
"gt_answers": "advised",
|
||
|
|
"image_path": "./data/IAM/c04/c04-017/c04-017-01-10.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/c04/c04-017/c04-017-01-11.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "think",
|
||
|
|
"gt_answers": "think",
|
||
|
|
"image_path": "./data/IAM/c04/c04-017/c04-017-02-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "hard",
|
||
|
|
"image_path": "./data/IAM/c04/c04-017/c04-017-02-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "and",
|
||
|
|
"image_path": "./data/IAM/c04/c04-017/c04-017-02-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "loy",
|
||
|
|
"gt_answers": "long",
|
||
|
|
"image_path": "./data/IAM/c04/c04-017/c04-017-02-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "before",
|
||
|
|
"gt_answers": "before",
|
||
|
|
"image_path": "./data/IAM/c04/c04-017/c04-017-02-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "h m",
|
||
|
|
"gt_answers": "his",
|
||
|
|
"image_path": "./data/IAM/c04/c04-017/c04-017-02-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "next",
|
||
|
|
"gt_answers": "next",
|
||
|
|
"image_path": "./data/IAM/c04/c04-017/c04-017-02-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "jump",
|
||
|
|
"gt_answers": "jump",
|
||
|
|
"image_path": "./data/IAM/c04/c04-017/c04-017-02-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word intro",
|
||
|
|
"gt_answers": "into",
|
||
|
|
"image_path": "./data/IAM/c04/c04-017/c04-017-02-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/c04/c04-017/c04-017-02-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "hollywood",
|
||
|
|
"gt_answers": "Hollywood",
|
||
|
|
"image_path": "./data/IAM/c04/c04-017/c04-017-03-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word person",
|
||
|
|
"gt_answers": "arena",
|
||
|
|
"image_path": "./data/IAM/c04/c04-017/c04-017-03-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "verdict",
|
||
|
|
"gt_answers": "VERDICT",
|
||
|
|
"image_path": "./data/IAM/c04/c04-017/c04-017-04-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "vagham",
|
||
|
|
"gt_answers": "Vaughan",
|
||
|
|
"image_path": "./data/IAM/c04/c04-017/c04-017-04-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "blaise",
|
||
|
|
"gt_answers": "should",
|
||
|
|
"image_path": "./data/IAM/c04/c04-017/c04-017-04-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "have",
|
||
|
|
"gt_answers": "have",
|
||
|
|
"image_path": "./data/IAM/c04/c04-017/c04-017-04-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "by passed",
|
||
|
|
"gt_answers": "bypassed",
|
||
|
|
"image_path": "./data/IAM/c04/c04-017/c04-017-04-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "this",
|
||
|
|
"gt_answers": "this",
|
||
|
|
"image_path": "./data/IAM/c04/c04-017/c04-017-04-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "appoach",
|
||
|
|
"gt_answers": "approach",
|
||
|
|
"image_path": "./data/IAM/c04/c04-017/c04-017-04-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "neal",
|
||
|
|
"gt_answers": "RONALD",
|
||
|
|
"image_path": "./data/IAM/c04/c04-017/c04-017-05-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "Lewis",
|
||
|
|
"image_path": "./data/IAM/c04/c04-017/c04-017-05-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "w s",
|
||
|
|
"gt_answers": "has",
|
||
|
|
"image_path": "./data/IAM/c04/c04-017/c04-017-05-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "h j",
|
||
|
|
"gt_answers": "just",
|
||
|
|
"image_path": "./data/IAM/c04/c04-017/c04-017-05-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "left",
|
||
|
|
"image_path": "./data/IAM/c04/c04-017/c04-017-05-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "love",
|
||
|
|
"gt_answers": "for",
|
||
|
|
"image_path": "./data/IAM/c04/c04-017/c04-017-05-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "his",
|
||
|
|
"image_path": "./data/IAM/c04/c04-017/c04-017-05-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "first",
|
||
|
|
"image_path": "./data/IAM/c04/c04-017/c04-017-05-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "taste",
|
||
|
|
"gt_answers": "taste",
|
||
|
|
"image_path": "./data/IAM/c04/c04-017/c04-017-05-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the letter j",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/c04/c04-017/c04-017-05-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/c04/c04-017/c04-017-05-10.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "hollywood",
|
||
|
|
"gt_answers": "Hollywood",
|
||
|
|
"image_path": "./data/IAM/c04/c04-017/c04-017-06-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "tex tune",
|
||
|
|
"gt_answers": "treatment",
|
||
|
|
"image_path": "./data/IAM/c04/c04-017/c04-017-06-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "thank you",
|
||
|
|
"gt_answers": "thanks",
|
||
|
|
"image_path": "./data/IAM/c04/c04-017/c04-017-06-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/c04/c04-017/c04-017-06-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a snake",
|
||
|
|
"gt_answers": "a",
|
||
|
|
"image_path": "./data/IAM/c04/c04-017/c04-017-06-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "sound",
|
||
|
|
"image_path": "./data/IAM/c04/c04-017/c04-017-06-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "performance max",
|
||
|
|
"gt_answers": "performance",
|
||
|
|
"image_path": "./data/IAM/c04/c04-017/c04-017-06-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "in",
|
||
|
|
"image_path": "./data/IAM/c04/c04-017/c04-017-06-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "toste",
|
||
|
|
"gt_answers": "Taste",
|
||
|
|
"image_path": "./data/IAM/c04/c04-017/c04-017-07-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/c04/c04-017/c04-017-07-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "fear",
|
||
|
|
"gt_answers": "Fear",
|
||
|
|
"image_path": "./data/IAM/c04/c04-017/c04-017-07-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "warner",
|
||
|
|
"gt_answers": "Warner",
|
||
|
|
"image_path": "./data/IAM/c04/c04-017/c04-017-07-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "theatre",
|
||
|
|
"gt_answers": "Theatre",
|
||
|
|
"image_path": "./data/IAM/c04/c04-017/c04-017-07-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "x",
|
||
|
|
"gt_answers": "X",
|
||
|
|
"image_path": "./data/IAM/c04/c04-017/c04-017-07-10.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person's signature",
|
||
|
|
"gt_answers": "At",
|
||
|
|
"image_path": "./data/IAM/r06/r06-130/r06-130-00-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/r06/r06-130/r06-130-00-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "enhance",
|
||
|
|
"gt_answers": "entrance",
|
||
|
|
"image_path": "./data/IAM/r06/r06-130/r06-130-00-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "f",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/r06/r06-130/r06-130-00-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "he",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/r06/r06-130/r06-130-00-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "cinema",
|
||
|
|
"gt_answers": "cinema",
|
||
|
|
"image_path": "./data/IAM/r06/r06-130/r06-130-00-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/r06/r06-130/r06-130-00-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "dorman",
|
||
|
|
"gt_answers": "doorman",
|
||
|
|
"image_path": "./data/IAM/r06/r06-130/r06-130-01-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "was",
|
||
|
|
"gt_answers": "was",
|
||
|
|
"image_path": "./data/IAM/r06/r06-130/r06-130-01-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "lying",
|
||
|
|
"gt_answers": "lying",
|
||
|
|
"image_path": "./data/IAM/r06/r06-130/r06-130-01-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "i",
|
||
|
|
"gt_answers": "in",
|
||
|
|
"image_path": "./data/IAM/r06/r06-130/r06-130-01-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "wait",
|
||
|
|
"gt_answers": "wait",
|
||
|
|
"image_path": "./data/IAM/r06/r06-130/r06-130-01-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a signature",
|
||
|
|
"gt_answers": "for",
|
||
|
|
"image_path": "./data/IAM/r06/r06-130/r06-130-01-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "anyone",
|
||
|
|
"gt_answers": "anyone",
|
||
|
|
"image_path": "./data/IAM/r06/r06-130/r06-130-01-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word person",
|
||
|
|
"gt_answers": "who",
|
||
|
|
"image_path": "./data/IAM/r06/r06-130/r06-130-02-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "hed",
|
||
|
|
"gt_answers": "had",
|
||
|
|
"image_path": "./data/IAM/r06/r06-130/r06-130-02-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "wel",
|
||
|
|
"gt_answers": "red",
|
||
|
|
"image_path": "./data/IAM/r06/r06-130/r06-130-02-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word eyes",
|
||
|
|
"gt_answers": "eyes",
|
||
|
|
"image_path": "./data/IAM/r06/r06-130/r06-130-02-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "grasping",
|
||
|
|
"gt_answers": "Grasping",
|
||
|
|
"image_path": "./data/IAM/r06/r06-130/r06-130-02-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "catherine",
|
||
|
|
"gt_answers": "Catherine",
|
||
|
|
"image_path": "./data/IAM/r06/r06-130/r06-130-02-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "by",
|
||
|
|
"image_path": "./data/IAM/r06/r06-130/r06-130-03-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "she",
|
||
|
|
"gt_answers": "her",
|
||
|
|
"image_path": "./data/IAM/r06/r06-130/r06-130-03-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "sue herring",
|
||
|
|
"gt_answers": "suffering",
|
||
|
|
"image_path": "./data/IAM/r06/r06-130/r06-130-03-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "arn",
|
||
|
|
"gt_answers": "arm",
|
||
|
|
"image_path": "./data/IAM/r06/r06-130/r06-130-03-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "he",
|
||
|
|
"image_path": "./data/IAM/r06/r06-130/r06-130-03-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word sol",
|
||
|
|
"gt_answers": "said",
|
||
|
|
"image_path": "./data/IAM/r06/r06-130/r06-130-03-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "bierly",
|
||
|
|
"gt_answers": "bitterly",
|
||
|
|
"image_path": "./data/IAM/r06/r06-130/r06-130-03-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "yes",
|
||
|
|
"gt_answers": "Yes",
|
||
|
|
"image_path": "./data/IAM/r06/r06-130/r06-130-04-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word coy",
|
||
|
|
"gt_answers": "why",
|
||
|
|
"image_path": "./data/IAM/r06/r06-130/r06-130-04-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "ave",
|
||
|
|
"gt_answers": "are",
|
||
|
|
"image_path": "./data/IAM/r06/r06-130/r06-130-04-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "you",
|
||
|
|
"gt_answers": "you",
|
||
|
|
"image_path": "./data/IAM/r06/r06-130/r06-130-04-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "crying",
|
||
|
|
"gt_answers": "crying",
|
||
|
|
"image_path": "./data/IAM/r06/r06-130/r06-130-04-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "he",
|
||
|
|
"gt_answers": "he",
|
||
|
|
"image_path": "./data/IAM/r06/r06-130/r06-130-04-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "had",
|
||
|
|
"image_path": "./data/IAM/r06/r06-130/r06-130-04-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "ho",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/r06/r06-130/r06-130-04-10.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "be",
|
||
|
|
"gt_answers": "be",
|
||
|
|
"image_path": "./data/IAM/r06/r06-130/r06-130-04-11.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "painshed",
|
||
|
|
"gt_answers": "punished",
|
||
|
|
"image_path": "./data/IAM/r06/r06-130/r06-130-05-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the name of the person",
|
||
|
|
"gt_answers": "for",
|
||
|
|
"image_path": "./data/IAM/r06/r06-130/r06-130-05-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "m s",
|
||
|
|
"gt_answers": "his",
|
||
|
|
"image_path": "./data/IAM/r06/r06-130/r06-130-05-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "crime",
|
||
|
|
"gt_answers": "crime",
|
||
|
|
"image_path": "./data/IAM/r06/r06-130/r06-130-05-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "dickt",
|
||
|
|
"gt_answers": "didnt",
|
||
|
|
"image_path": "./data/IAM/r06/r06-130/r06-130-05-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "he",
|
||
|
|
"gt_answers": "he",
|
||
|
|
"image_path": "./data/IAM/r06/r06-130/r06-130-05-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "cathe",
|
||
|
|
"gt_answers": "Cathe",
|
||
|
|
"image_path": "./data/IAM/r06/r06-130/r06-130-05-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word person",
|
||
|
|
"gt_answers": "rine",
|
||
|
|
"image_path": "./data/IAM/r06/r06-130/r06-130-06-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "shared",
|
||
|
|
"gt_answers": "stared",
|
||
|
|
"image_path": "./data/IAM/r06/r06-130/r06-130-06-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "a",
|
||
|
|
"gt_answers": "at",
|
||
|
|
"image_path": "./data/IAM/r06/r06-130/r06-130-06-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "h",
|
||
|
|
"gt_answers": "him",
|
||
|
|
"image_path": "./data/IAM/r06/r06-130/r06-130-06-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "incredulous",
|
||
|
|
"gt_answers": "incredulous",
|
||
|
|
"image_path": "./data/IAM/r06/r06-130/r06-130-06-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "philip",
|
||
|
|
"gt_answers": "Philip",
|
||
|
|
"image_path": "./data/IAM/r06/r06-130/r06-130-06-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "rescued",
|
||
|
|
"gt_answers": "rescued",
|
||
|
|
"image_path": "./data/IAM/r06/r06-130/r06-130-07-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "her",
|
||
|
|
"gt_answers": "her",
|
||
|
|
"image_path": "./data/IAM/r06/r06-130/r06-130-07-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "by",
|
||
|
|
"gt_answers": "by",
|
||
|
|
"image_path": "./data/IAM/r06/r06-130/r06-130-07-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "saying",
|
||
|
|
"gt_answers": "saying",
|
||
|
|
"image_path": "./data/IAM/r06/r06-130/r06-130-07-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "hi",
|
||
|
|
"gt_answers": "with",
|
||
|
|
"image_path": "./data/IAM/r06/r06-130/r06-130-07-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "disdain",
|
||
|
|
"gt_answers": "disdain",
|
||
|
|
"image_path": "./data/IAM/r06/r06-130/r06-130-07-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "some",
|
||
|
|
"gt_answers": "Some",
|
||
|
|
"image_path": "./data/IAM/r06/r06-130/r06-130-08-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "people",
|
||
|
|
"gt_answers": "people",
|
||
|
|
"image_path": "./data/IAM/r06/r06-130/r06-130-08-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "don't",
|
||
|
|
"gt_answers": "dont",
|
||
|
|
"image_path": "./data/IAM/r06/r06-130/r06-130-08-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "know",
|
||
|
|
"gt_answers": "know",
|
||
|
|
"image_path": "./data/IAM/r06/r06-130/r06-130-08-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word right",
|
||
|
|
"gt_answers": "right",
|
||
|
|
"image_path": "./data/IAM/r06/r06-130/r06-130-08-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "from person",
|
||
|
|
"gt_answers": "from",
|
||
|
|
"image_path": "./data/IAM/r06/r06-130/r06-130-08-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "wrong",
|
||
|
|
"gt_answers": "wrong",
|
||
|
|
"image_path": "./data/IAM/r06/r06-130/r06-130-08-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word never",
|
||
|
|
"gt_answers": "even",
|
||
|
|
"image_path": "./data/IAM/r06/r06-130/r06-130-09-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word father",
|
||
|
|
"gt_answers": "when",
|
||
|
|
"image_path": "./data/IAM/r06/r06-130/r06-130-09-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "ms",
|
||
|
|
"gt_answers": "its",
|
||
|
|
"image_path": "./data/IAM/r06/r06-130/r06-130-09-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "demon strangled",
|
||
|
|
"gt_answers": "demonstrated",
|
||
|
|
"image_path": "./data/IAM/r06/r06-130/r06-130-09-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "h",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/r06/r06-130/r06-130-09-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "them",
|
||
|
|
"image_path": "./data/IAM/r06/r06-130/r06-130-09-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "oversew",
|
||
|
|
"gt_answers": "Oversew",
|
||
|
|
"image_path": "./data/IAM/e01/e01-092/e01-092-00-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "crochet",
|
||
|
|
"gt_answers": "crochet",
|
||
|
|
"image_path": "./data/IAM/e01/e01-092/e01-092-00-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "ho",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/e01/e01-092/e01-092-00-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/e01/e01-092/e01-092-00-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "dc",
|
||
|
|
"gt_answers": "dc",
|
||
|
|
"image_path": "./data/IAM/e01/e01-092/e01-092-00-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the letter a",
|
||
|
|
"gt_answers": "A",
|
||
|
|
"image_path": "./data/IAM/e01/e01-092/e01-092-00-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "household debt",
|
||
|
|
"gt_answers": "household",
|
||
|
|
"image_path": "./data/IAM/e01/e01-092/e01-092-00-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "word",
|
||
|
|
"gt_answers": "word",
|
||
|
|
"image_path": "./data/IAM/e01/e01-092/e01-092-00-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "coats",
|
||
|
|
"gt_answers": "Coats",
|
||
|
|
"image_path": "./data/IAM/e01/e01-092/e01-092-01-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "merci crochet",
|
||
|
|
"gt_answers": "MercerCrochet",
|
||
|
|
"image_path": "./data/IAM/e01/e01-092/e01-092-01-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "poseesses",
|
||
|
|
"gt_answers": "possesses",
|
||
|
|
"image_path": "./data/IAM/e01/e01-092/e01-092-01-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word qualities",
|
||
|
|
"gt_answers": "qualities",
|
||
|
|
"image_path": "./data/IAM/e01/e01-092/e01-092-01-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "which",
|
||
|
|
"gt_answers": "which",
|
||
|
|
"image_path": "./data/IAM/e01/e01-092/e01-092-01-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "are",
|
||
|
|
"gt_answers": "are",
|
||
|
|
"image_path": "./data/IAM/e01/e01-092/e01-092-02-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "o",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/e01/e01-092/e01-092-02-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/e01/e01-092/e01-092-02-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word worst",
|
||
|
|
"gt_answers": "utmost",
|
||
|
|
"image_path": "./data/IAM/e01/e01-092/e01-092-02-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "importance in hindi",
|
||
|
|
"gt_answers": "importance",
|
||
|
|
"image_path": "./data/IAM/e01/e01-092/e01-092-02-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "to",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/e01/e01-092/e01-092-02-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/e01/e01-092/e01-092-02-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "worker",
|
||
|
|
"gt_answers": "worker",
|
||
|
|
"image_path": "./data/IAM/e01/e01-092/e01-092-02-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "It",
|
||
|
|
"image_path": "./data/IAM/e01/e01-092/e01-092-03-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "washes",
|
||
|
|
"gt_answers": "washes",
|
||
|
|
"image_path": "./data/IAM/e01/e01-092/e01-092-03-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "beautiful",
|
||
|
|
"gt_answers": "beautifully",
|
||
|
|
"image_path": "./data/IAM/e01/e01-092/e01-092-03-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "never",
|
||
|
|
"gt_answers": "never",
|
||
|
|
"image_path": "./data/IAM/e01/e01-092/e01-092-03-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "losses",
|
||
|
|
"gt_answers": "loses",
|
||
|
|
"image_path": "./data/IAM/e01/e01-092/e01-092-03-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "ifs",
|
||
|
|
"gt_answers": "its",
|
||
|
|
"image_path": "./data/IAM/e01/e01-092/e01-092-03-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "colour",
|
||
|
|
"gt_answers": "colour",
|
||
|
|
"image_path": "./data/IAM/e01/e01-092/e01-092-03-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "ron",
|
||
|
|
"gt_answers": "nor",
|
||
|
|
"image_path": "./data/IAM/e01/e01-092/e01-092-04-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "becomes",
|
||
|
|
"gt_answers": "becomes",
|
||
|
|
"image_path": "./data/IAM/e01/e01-092/e01-092-04-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "stringy",
|
||
|
|
"gt_answers": "stringy",
|
||
|
|
"image_path": "./data/IAM/e01/e01-092/e01-092-04-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word is",
|
||
|
|
"gt_answers": "is",
|
||
|
|
"image_path": "./data/IAM/e01/e01-092/e01-092-04-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "every day",
|
||
|
|
"gt_answers": "very",
|
||
|
|
"image_path": "./data/IAM/e01/e01-092/e01-092-04-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "elashc",
|
||
|
|
"gt_answers": "elastic",
|
||
|
|
"image_path": "./data/IAM/e01/e01-092/e01-092-04-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "and",
|
||
|
|
"gt_answers": "and",
|
||
|
|
"image_path": "./data/IAM/e01/e01-092/e01-092-04-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "preserves",
|
||
|
|
"gt_answers": "preserves",
|
||
|
|
"image_path": "./data/IAM/e01/e01-092/e01-092-05-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/e01/e01-092/e01-092-05-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "beauty",
|
||
|
|
"gt_answers": "beauty",
|
||
|
|
"image_path": "./data/IAM/e01/e01-092/e01-092-05-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "o f",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/e01/e01-092/e01-092-05-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/e01/e01-092/e01-092-05-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "design",
|
||
|
|
"gt_answers": "design",
|
||
|
|
"image_path": "./data/IAM/e01/e01-092/e01-092-05-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "u",
|
||
|
|
"gt_answers": "It",
|
||
|
|
"image_path": "./data/IAM/e01/e01-092/e01-092-05-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "is",
|
||
|
|
"image_path": "./data/IAM/e01/e01-092/e01-092-05-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "lazy",
|
||
|
|
"gt_answers": "easy",
|
||
|
|
"image_path": "./data/IAM/e01/e01-092/e01-092-05-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "t h",
|
||
|
|
"gt_answers": "to",
|
||
|
|
"image_path": "./data/IAM/e01/e01-092/e01-092-06-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "work",
|
||
|
|
"gt_answers": "work",
|
||
|
|
"image_path": "./data/IAM/e01/e01-092/e01-092-06-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word person",
|
||
|
|
"gt_answers": "with",
|
||
|
|
"image_path": "./data/IAM/e01/e01-092/e01-092-06-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "soft",
|
||
|
|
"gt_answers": "soft",
|
||
|
|
"image_path": "./data/IAM/e01/e01-092/e01-092-06-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "glossy y",
|
||
|
|
"gt_answers": "glossy",
|
||
|
|
"image_path": "./data/IAM/e01/e01-092/e01-092-06-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "and",
|
||
|
|
"gt_answers": "and",
|
||
|
|
"image_path": "./data/IAM/e01/e01-092/e01-092-06-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "h",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/e01/e01-092/e01-092-06-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "uniform",
|
||
|
|
"gt_answers": "uniform",
|
||
|
|
"image_path": "./data/IAM/e01/e01-092/e01-092-06-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "thickness",
|
||
|
|
"gt_answers": "thickness",
|
||
|
|
"image_path": "./data/IAM/e01/e01-092/e01-092-07-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "t",
|
||
|
|
"gt_answers": "it",
|
||
|
|
"image_path": "./data/IAM/e01/e01-092/e01-092-07-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word person",
|
||
|
|
"gt_answers": "is",
|
||
|
|
"image_path": "./data/IAM/e01/e01-092/e01-092-07-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/e01/e01-092/e01-092-07-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "ideal",
|
||
|
|
"gt_answers": "ideal",
|
||
|
|
"image_path": "./data/IAM/e01/e01-092/e01-092-07-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "crochet",
|
||
|
|
"gt_answers": "crochet",
|
||
|
|
"image_path": "./data/IAM/e01/e01-092/e01-092-07-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "thread",
|
||
|
|
"gt_answers": "thread",
|
||
|
|
"image_path": "./data/IAM/e01/e01-092/e01-092-07-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "b",
|
||
|
|
"gt_answers": "By",
|
||
|
|
"image_path": "./data/IAM/g06/g06-011j/g06-011j-00-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/g06/g06-011j/g06-011j-00-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "end",
|
||
|
|
"gt_answers": "end",
|
||
|
|
"image_path": "./data/IAM/g06/g06-011j/g06-011j-00-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "o",
|
||
|
|
"gt_answers": "of",
|
||
|
|
"image_path": "./data/IAM/g06/g06-011j/g06-011j-00-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the e",
|
||
|
|
"gt_answers": "the",
|
||
|
|
"image_path": "./data/IAM/g06/g06-011j/g06-011j-00-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "month",
|
||
|
|
"gt_answers": "month",
|
||
|
|
"image_path": "./data/IAM/g06/g06-011j/g06-011j-00-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "person",
|
||
|
|
"gt_answers": "he",
|
||
|
|
"image_path": "./data/IAM/g06/g06-011j/g06-011j-00-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "john shill",
|
||
|
|
"gt_answers": "still",
|
||
|
|
"image_path": "./data/IAM/g06/g06-011j/g06-011j-00-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "de ghetto de",
|
||
|
|
"gt_answers": "delighted",
|
||
|
|
"image_path": "./data/IAM/g06/g06-011j/g06-011j-00-08.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "m",
|
||
|
|
"gt_answers": "in",
|
||
|
|
"image_path": "./data/IAM/g06/g06-011j/g06-011j-00-09.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "naples",
|
||
|
|
"gt_answers": "Naples",
|
||
|
|
"image_path": "./data/IAM/g06/g06-011j/g06-011j-01-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "he",
|
||
|
|
"gt_answers": "He",
|
||
|
|
"image_path": "./data/IAM/g06/g06-011j/g06-011j-01-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "the word held",
|
||
|
|
"gt_answers": "told",
|
||
|
|
"image_path": "./data/IAM/g06/g06-011j/g06-011j-01-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "clove curry",
|
||
|
|
"gt_answers": "Cloncurry",
|
||
|
|
"image_path": "./data/IAM/g06/g06-011j/g06-011j-01-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "thg",
|
||
|
|
"gt_answers": "that",
|
||
|
|
"image_path": "./data/IAM/g06/g06-011j/g06-011j-01-05.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "he",
|
||
|
|
"gt_answers": "he",
|
||
|
|
"image_path": "./data/IAM/g06/g06-011j/g06-011j-01-06.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "joyel",
|
||
|
|
"gt_answers": "enjoyed",
|
||
|
|
"image_path": "./data/IAM/g06/g06-011j/g06-011j-01-07.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": " ",
|
||
|
|
"gt_answers": "it",
|
||
|
|
"image_path": "./data/IAM/g06/g06-011j/g06-011j-02-00.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "g",
|
||
|
|
"gt_answers": "as",
|
||
|
|
"image_path": "./data/IAM/g06/g06-011j/g06-011j-02-01.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "m h",
|
||
|
|
"gt_answers": "much",
|
||
|
|
"image_path": "./data/IAM/g06/g06-011j/g06-011j-02-02.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "go",
|
||
|
|
"gt_answers": "as",
|
||
|
|
"image_path": "./data/IAM/g06/g06-011j/g06-011j-02-03.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"question": "what is written in the image?",
|
||
|
|
"answer": "hs",
|
||
|
|
"gt_answers": "his",
|
||
|
|
"image_path": "./data/IAM/g06/g06-011j/g06-011j-02-04.png",
|
||
|
|
"model_name": "BLIP2"
|
||
|
|
}
|
||
|
|
]
|