AI GitHub
CVAT for computer vision annotation
Create image and video labeling tasks with review and export workflows.
Key Takeaways
- CVAT manages annotation jobs and supports shapes such as boxes, polygons and tracks.
- Choose a labeling format and set up a representative task with reviewers.
- Annotate one short sequence with an occlusion and export it in the intended format.
- Different formats represent tracks and attributes differently.
How it works
CVAT manages annotation jobs and supports shapes such as boxes, polygons and tracks. Export formats determine how labels map into training data.
What you need
Choose a labeling format and set up a representative task with reviewers. Check whether the export includes the annotations and metadata the training project needs.
What to try first
Annotate one short sequence with an occlusion and export it in the intended format. Validate category IDs and frame alignment with the training loader.
Round-trip a small export and visualize it after reimporting. This catches changes in coordinates or attributes that a row count cannot detect.
Where to look for failures
Different formats represent tracks and attributes differently. Confirm your review process catches missing objects and conversion errors before training.
Related Franklin pages
Follow one annotation from labeling to training and evaluation. Check that frame numbers, class IDs and coordinate conventions survive the handoff.
- Train and evaluate pixel-labeling models using shared configurations and dataset interfaces. MMSegmentation for semantic segmentation experiments
- Explore images or video, annotations and model outputs together. FiftyOne for inspecting vision datasets and predictions
For an overview of the types of predictions vision systems produce, see Franklin’s glossary. Computer vision
Project documentation
Try it for your workflow
Explore the project’s documentation for setup and usage, or save this page to a collection for later.
Read the project documentation ↗