CVAT (Computer Vision Annotation Tool) is the leading open-source data annotation platform for image, video, audio, and 3D data, with nearly 17,000 stars on GitHub. Born at Intel and now developed by CVAT.ai, it is the tool data teams use to build the labeled datasets that computer vision models are trained on. The platform is in continuous active development, with updates landing in October 2026.
CVAT provides a rich annotation interface for bounding boxes, polygons, polylines, keypoints, cuboids, and semantic segmentation, with interpolation and tracking that propagate labels across video frames. AI assistance is built in: models such as Segment Anything and detectors can be connected for pre-labeling and auto-annotation, and dataset quality is protected by QA workflows, review states, and analytics. Teams can collaborate with organizations, projects, tasks, and role-based access, manage datasets in all major formats, and automate everything through REST APIs and a Python SDK. CVAT runs self-hosted from the open-source repository, as a managed cloud, or as an enterprise deployment.
The open-source Community edition is free to self-host. CVAT Cloud offers a free tier with paid plans for larger teams, and Enterprise adds advanced features, support, and deployment options.
CVAT is for ML teams, data-labeling operations, and researchers building training datasets for detection, segmentation, tracking, and other computer vision tasks at any scale. Its combination of an open-source core, managed cloud, and labeling services lets organizations start small and grow into production data pipelines without switching tools.