// the find
cvat-ai/cvat
Computer Vision Annotation Tool (CVAT) is a leading platform for building high-quality visual datasets for vision AI. It offers open-source, cloud, and enterprise products, as well as labeling services, for image, video, and 3D annotation with AI-assisted labeling, quality assurance, team collaboration, analytics, and developer APIs.
CVAT is a self-hosted web app for annotating images, video, and 3D point clouds for computer vision training data — bounding boxes, polygons, masks, keypoints, cuboids. It's for teams that need a real annotation pipeline with task management and review workflows, not a quick labeling script.
Export/import covers 20+ formats (COCO, YOLO, Pascal VOC, KITTI, MOT) so it slots into existing training pipelines without custom converters. Auto-annotation hooks into serverless functions (SAM, YOLO v7, RetinaNet, etc.) via Nuclio, so you can pre-label with your own model instead of clicking every box by hand. It also ships a Python SDK and CLI alongside the REST API, so task creation and dataset export can be scripted rather than done through the UI every time.
Deployment is a multi-container Docker Compose stack, and the auto-labeling piece needs a separate Nuclio/nuctl setup and per-model deployment — there's no single `docker run` path if you want AI-assisted labeling. The free self-hosted edition is deliberately feature-gated: quality control UI, SSO, and the newer SAM 2/3 auto-labeling are reserved for the paid CVAT Online/Enterprise tiers, so you'll hit a wall if your team needs those. Browser support is Chromium-first; Safari isn't supported and Firefox is listed with caveats, which matters for annotation teams that don't control their own machines. It's also a large, multi-package monorepo (separate canvas, canvas3d, core, SDK, CLI packages) — customizing annotation behavior means understanding several codebases, not one.