CVAT

CVAT, the Computer Vision Annotation Tool, is an open-source web platform for labeling images, videos, and 3D point clouds, first released in 2018 and now the foundation for the commercial CVAT Online and Enterprise products. Annotators can draw bounding boxes, polygons, masks, keypoints, cuboids, and tags, and import or export data in more than 20 formats including COCO, YOLO, Pascal VOC, and KITTI. AI-assisted labeling is built in through models such as Segment Anything, YOLO v7, and Mask R-CNN, and teams get multi-user organizations, role-based access, review workflows, and quality checks like consensus comparison and honeypot validation. The platform connects to S3, Azure, and Google Cloud storage, and exposes a REST API, Python SDK, and CLI for automation. Self-hosting runs on Docker or Kubernetes under an MIT license for the core code, which is why computer vision teams that need data ownership and infrastructure control commonly choose it for building training datasets.