cvat

Why we included this project

CVAT is a self-hosted annotation platform that has become a common choice for teams building visual datasets. You can label images, video frames, and 3D point clouds with bounding boxes, polygons, masks, keypoints, and cuboids, then organize the work into projects and tasks The review workflow and consensus checks between annotators help keep quality consistent, and export to COCO, YOLO, and Pascal VOC means labeled data can go straight into your training pipeline. You can also plug in your own detection or segmentation models to pre-label frames and have humans correct the results, which cuts down the time spent on large labeling efforts. Since the whole stack runs locally via Docker, with a REST API and Python SDK for automation, teams that need to keep data on their own infrastructure can do so without sending anything to the cloud.