diff --git a/README.md b/README.md index 9ed5eba..d4a799f 100644 --- a/README.md +++ b/README.md @@ -20,6 +20,7 @@ In this repo, we've supported the following demo with **simple implementations** Grounded SAM 2 does not introduce significant methodological changes compared to [Grounded SAM: Assembling Open-World Models for Diverse Visual Tasks](https://arxiv.org/abs/2401.14159). Both approaches leverage the capabilities of open-world models to address complex visual tasks. Consequently, we try to **simplify the code implementation** in this repository, aiming to enhance user convenience. ## Latest updates +- **2025.04.20**: Update to `dds-cloudapi-sdk` API V2 version. The V1 version in the original API for `Grounding DINO 1.5` and `DINO-X` has been deprecated, please update to the latest `dds-cloudapi-sdk` by `pip install dds-cloudapi-sdk -U` to use `Grounding DINO 1.5 / 1.6` and `DINO-X` models. Please refer to [dds-cloudapi-sdk](https://github.com/deepdataspace/dds-cloudapi-sdk) and our [API docs](https://cloud.deepdataspace.com/docs) to view more details about the update. - **2024.12.02**: Support **DINO-X with SAM 2** demos (including object segmentation and tracking), please install the latest version of `dds-cloudapi-sdk==0.3.3` and refer to [Grounded SAM 2 (with DINO-X)](#grounded-sam-2-image-demo-with-dino-x) and [Grounded SAM 2 Video (with DINO-X)](#grounded-sam-2-video-object-tracking-demo-with-custom-video-input-with-dino-x) for more details. diff --git a/grounded_sam2_tracking_demo_custom_video_input_dinox.py b/grounded_sam2_tracking_demo_custom_video_input_dinox.py index e715218..661eb08 100644 --- a/grounded_sam2_tracking_demo_custom_video_input_dinox.py +++ b/grounded_sam2_tracking_demo_custom_video_input_dinox.py @@ -1,9 +1,7 @@ -# dds cloudapi for Grounding DINO 1.5 +# dds cloudapi for DINO-X - update to V2Task API from dds_cloudapi_sdk import Config from dds_cloudapi_sdk import Client -from dds_cloudapi_sdk.tasks.dinox import DinoxTask -from dds_cloudapi_sdk.tasks.types import DetectionTarget -from dds_cloudapi_sdk import TextPrompt +from dds_cloudapi_sdk.tasks.v2_task import V2Task import os import cv2 @@ -30,6 +28,7 @@ SAVE_TRACKING_RESULTS_DIR = "./tracking_results" API_TOKEN_FOR_DINOX = "Your API token" PROMPT_TYPE_FOR_VIDEO = "box" # choose from ["point", "box", "mask"] BOX_THRESHOLD = 0.2 +IOU_THRESHOLD = 0.8 # 添加IOU阈值参数 """ Step 1: Environment settings and model initialization for SAM 2 @@ -98,22 +97,29 @@ config = Config(API_TOKEN_FOR_DINOX) # Step 2: initialize the client client = Client(config) -# Step 3: run the task by DetectionTask class -# image_url = "https://algosplt.oss-cn-shenzhen.aliyuncs.com/test_files/tasks/detection/iron_man.jpg" +# Step 3: run the task using V2Task class # if you are processing local image file, upload them to DDS server to get the image url image_url = client.upload_file(img_path) -task = DinoxTask( - image_url=image_url, - prompts=[TextPrompt(text=TEXT_PROMPT)], - bbox_threshold=0.25, - targets=[DetectionTarget.BBox], +task = V2Task( + api_path="/v2/task/dinox/detection", + api_body={ + "model": "DINO-X-1.0", + "image": image_url, + "prompt": { + "type": "text", + "text": TEXT_PROMPT + }, + "targets": ["bbox"], + "bbox_threshold": BOX_THRESHOLD, + "iou_threshold": IOU_THRESHOLD, + } ) client.run_task(task) result = task.result -objects = result.objects # the list of detected objects +objects = result["objects"] # the list of detected objects input_boxes = [] @@ -121,9 +127,9 @@ confidences = [] class_names = [] for idx, obj in enumerate(objects): - input_boxes.append(obj.bbox) - confidences.append(obj.score) - class_names.append(obj.category) + input_boxes.append(obj["bbox"]) + confidences.append(obj["score"]) + class_names.append(obj["category"]) input_boxes = np.array(input_boxes) diff --git a/grounded_sam2_tracking_demo_custom_video_input_gd1.5.py b/grounded_sam2_tracking_demo_custom_video_input_gd1.5.py index 2c32ed3..0576a78 100644 --- a/grounded_sam2_tracking_demo_custom_video_input_gd1.5.py +++ b/grounded_sam2_tracking_demo_custom_video_input_gd1.5.py @@ -1,10 +1,7 @@ -# dds cloudapi for Grounding DINO 1.5 +# dds cloudapi for Grounding DINO 1.5 - 更新至V2Task API from dds_cloudapi_sdk import Config from dds_cloudapi_sdk import Client -from dds_cloudapi_sdk import DetectionTask -from dds_cloudapi_sdk import TextPrompt -from dds_cloudapi_sdk import DetectionModel -from dds_cloudapi_sdk import DetectionTarget +from dds_cloudapi_sdk.tasks.v2_task import V2Task import os import cv2 @@ -31,6 +28,7 @@ SAVE_TRACKING_RESULTS_DIR = "./tracking_results" API_TOKEN_FOR_GD1_5 = "Your API token" PROMPT_TYPE_FOR_VIDEO = "box" # choose from ["point", "box", "mask"] BOX_THRESHOLD = 0.2 +IOU_THRESHOLD = 0.8 # 添加IOU阈值参数 """ Step 1: Environment settings and model initialization for SAM 2 @@ -99,33 +97,38 @@ config = Config(API_TOKEN_FOR_GD1_5) # Step 2: initialize the client client = Client(config) -# Step 3: run the task by DetectionTask class -# image_url = "https://algosplt.oss-cn-shenzhen.aliyuncs.com/test_files/tasks/detection/iron_man.jpg" +# Step 3: run the task using V2Task class # if you are processing local image file, upload them to DDS server to get the image url image_url = client.upload_file(img_path) -task = DetectionTask( - image_url=image_url, - prompts=[TextPrompt(text=TEXT_PROMPT)], - targets=[DetectionTarget.BBox], # detect bbox - model=DetectionModel.GDino1_6_Pro, # detect with GroundingDino-1.5-Pro model - bbox_threshold=BOX_THRESHOLD, +task = V2Task( + api_path="/v2/task/grounding_dino/detection", + api_body={ + "model": "GroundingDino-1.5-Pro", + "image": image_url, + "prompt": { + "type": "text", + "text": TEXT_PROMPT + }, + "targets": ["bbox"], + "bbox_threshold": BOX_THRESHOLD, + "iou_threshold": IOU_THRESHOLD, + } ) client.run_task(task) result = task.result -objects = result.objects # the list of detected objects - +objects = result["objects"] # the list of detected objects input_boxes = [] confidences = [] class_names = [] for idx, obj in enumerate(objects): - input_boxes.append(obj.bbox) - confidences.append(obj.score) - class_names.append(obj.category) + input_boxes.append(obj["bbox"]) + confidences.append(obj["score"]) + class_names.append(obj["category"]) input_boxes = np.array(input_boxes)