- Duplicate image finder: search one or more datasets efficiently for duplicate images: duplicate_image_finder.ipynb
- Logs for COCO, MPII, OCHUman here: data/logs
- Annotation to COCO conversion (human keypoint annotations): Datasets_to_COCO.ipynb
- MPII, CrowdPose, OCHuman, download here: data/anno_files
- Start CVAT in Docker
- Create .env file with CVAT credentials (see .env.example)
- Start a cvat_*.ipynb notebook
Paths to datasets can be added/adjusted in configs/default.py.
Note: CVAT does currently not support labeled keypoints (e.g., COCO format).
Check:
- coco-annotator-setup.md for working with the coco annotator
- human_keypoint_annotation.md for a keypoint-annotation guideline