add

datasets/process/process_ESTVQA.py

@@ -0,0 +1,23 @@
+import json
+import os
+import re
+def has_chinese_characters(string):
+    pattern = re.compile(r'[\u4e00-\u9fa5]')
+    return bool(pattern.search(string))
+if __name__ == "__main__":
+    ann_file = "/home/zhangli/GPT4/MutimodelOCR/data/ESTVQA/annotations/train.json"
+    img_file = "/home/zhangli/GPT4/MutimodelOCR/data/ESTVQA/images/train"
+    cn_list = []
+    en_list= []
+    with open(ann_file,'r') as f:
+        data = json.load(f)
+        for i in range(len(data)):
+            if has_chinese_characters(data[i]['annotation'][0]['question']):
+                cn_list.append(data[i])
+            else:
+                en_list.append(data[i])
+    with open('./cn_train.json', 'w', encoding='utf-8') as f:
+        json.dump(cn_list, f, ensure_ascii=False)
+    with open('./en_train.json', 'w', encoding='utf-8') as f:
+        json.dump(en_list, f, ensure_ascii=False)
+    

datasets/process/process_FUNSD.py

@@ -0,0 +1,25 @@
+import json
+import os
+if __name__ == "__main__":
+    ann_dir_path = '/home/zhangli/GPT4/MutimodelOCR/data/FUNSD/training_data/annotations'
+    questions = []
+    answers = []
+    for file_name in os.listdir(ann_dir_path):
+        file_path = os.path.join(ann_dir_path, file_name)
+        with open(file_path, 'r') as f:
+            json_data = json.load(f)['form']
+            #去除空的linking
+            json_data = [d for d in json_data if "linking" in d and len(d["linking"])>0]
+            question_list = [d for d in json_data if d.get('label') == 'question']
+            answer_list = [d for d in json_data if d.get('label') == 'answer']
+            unique_question_list = [d for i, d in enumerate(question_list) if d['text'] not in [x['text'] for x in json_data[:i]]]
+            for i in range(len(unique_question_list)):
+                link = unique_question_list[i]['linking']
+                gt_answer = ""
+                for j in range(len(link)):
+                    for k in range(len(answer_list)):
+                        if answer_list[k]['id'] == link[j][1]:
+                            gt_answer = gt_answer + answer_list[k]['text']
+                questions.append(unique_question_list[i]['text'])
+                answers.append(gt_answer)
+