Cvat

CVAT is a data annotation platform for turning raw images, video, 3D point clouds and audio into labeled datasets for training computer vision and multimodal models. Annotators draw bounding boxes, polygons, pixel masks, keypoints, skeletons, tags and tracks, and follow objects across video frames and point cloud sequences, while AI assisted tools such as SAM 2, SAM 3, Ultralytics, Hugging Face models or a team's own models speed up the work. It serves machine learning teams that run structured labeling projects at any scale. Data imports from local files, Amazon S3, Azure Blob Storage, Google Cloud Storage or other S3 compatible storage. Work is organized into projects, tasks and jobs with roles, stages and reviewer assignments, and quality control uses ground truth jobs, honeypots and consensus workflows. Progress and workload dashboards show bottlenecks, and datasets export in more than 20 formats including COCO, YOLO, KITTI, Cityscapes and Pascal VOC. An API, SDK and CLI automate import, task management and export. CVAT also sells a managed labeling service for teams that prefer to outsource the work. CVAT comes in three editions. CVAT Community is open source, self hosted and free. CVAT Online offers a Free plan for one to two members with limited projects and storage, Solo at $33 per month or $23 per month billed yearly, and Team at $33 per user per month or $23 per user per month billed yearly, for two to 50 seats. CVAT Enterprise is self hosted with SSO, RBAC and enterprise support, starting at $12,000 per year.