Cvat

CVAT is the go-to choice for teams that need a full-featured, self-hosted annotation platform. Its open-source Community edition gives you complete control over your data and workflow, which is critical for regulated industries. The managed cloud plans (Solo/Team) are reasonably priced and scale well, though the free tier’s 1 GB storage and 3-task limit is restrictive—enough only for quick prototyping. We’d reach for CVAT when you need more than just point-and-click labeling: the AI-assisted tools (SAM 2, Ultralytics) cut down manual work significantly, and the quality control features (honeypots, ground truth jobs) are rare in open-source tools. The Enterprise tier’s SSO, RBAC, and audit logs make it suitable for defense and healthcare deployments. Where it bites: the platform has a learning curve, especially for setting up custom AI models or cloud storage integrations. Teams needing pure text or audio annotation should look elsewhere—audio support is still “coming soon.” Also, the Free tier’s limits (1 project, 3 tasks, 1 GB storage) are strict; you’ll quickly need a paid plan. Compared to alternatives like Label Studio or Supervisely, CVAT offers stronger AI-assisted labeling and a more mature managed service. However, if you need a simpler UI for basic bounding box tasks and don’t need advanced workflows, Label Studio’s free tier is more generous. In practice, CVAT shines for teams that are serious about computer vision at scale—autonomous driving, robotics, medical imaging. The managed labeling service is a nice bonus if you want to outsource, but the platform itself is strong enough to justify paying for a Solo or Team plan.