About CVAT
CVAT is an open source annotation tool for labeling data used to train AI models. It covers images, video, 3D point clouds and audio.
What it does
CVAT lets teams mark up raw data so that models can learn from it. Users create projects and tasks, label items by hand or with automated help, and review the results for quality. The finished labels can be exported as datasets for training, and the same work can be shared across a team.
Key features
- Bounding boxes, polygons, masks and keypoints
- Video, 3D point cloud and audio labeling
- Quality control checks on annotations
- Export datasets and connect to APIs
Who it's for
Computer vision and machine learning teams that need to label large sets of images or video. It suits groups that want to run labeling on their own servers and keep control of their data, as well as teams that prefer a hosted option.
Pricing and good to know
The Community edition is open source and free to run on your own infrastructure. CVAT Online is a managed cloud option with free, solo and team plans, and the enterprise self-hosted edition adds single sign-on and role-based access. The repository is licensed under MIT.
Highlights
- Draw bounding boxes, polygons and masks on images
- Annotate video, 3D point clouds and audio
- Import data from local files or cloud storage
- Check annotation quality and track team progress
