Download Latest Version
v2.42.0 source code.tar.gz (83.8 MB)
Get an email when there's a new version of Computer Vision Annotation Tool (CVAT)
| Name | Modified | Size | Downloads / Week |
|---|---|---|---|
| Parent folder | |||
| README.md | 2025-07-29 | 1.3 kB | |
| v2.42.0 source code.tar.gz | 2025-07-29 | 83.8 MB | |
| v2.42.0 source code.zip | 2025-07-29 | 85.0 MB | |
| Totals: 3 Items | 168.7 MB | 1 | |
Added
-
[SDK, CLI] Added an auto-annotation function interface for tracker functions, and agent support for it (https://github.com/cvat-ai/cvat/pull/9579)
-
[SDK]
TaskDatasetnow supports tasks with video chunks when created withMediaDownloadPolicy.PRELOAD_ALL; this also means that agents can process interactive detection requests on such tasks (https://github.com/cvat-ai/cvat/pull/9579) -
Support for cgroup v2 to calculate the number of threads when downloading images from cloud storages (https://github.com/cvat-ai/cvat/pull/9668)
Fixed
-
Bump Python runtime version for Segment Anything interactor Nuclio function from 3.8 to 3.10 (https://github.com/cvat-ai/cvat/pull/9593)
-
User could not log in if an email used for the invitation had a different case than the one used during manual registration (https://github.com/cvat-ai/cvat/pull/9629)
-
Fixed downloading images only in 1 thread when preparing chunks in case of using cloud storage (https://github.com/cvat-ai/cvat/pull/9668)
Security
- Added missing email verification check when Basic HTTP authentication is used and server is configured to require email verification (
ACCOUNT_EMAIL_VERIFICATION==mandatory) (https://github.com/cvat-ai/cvat/security/advisories/GHSA-fxgh-m76j-242q)