Computer Vision Annotation Tool (CVAT) Files

Added

• Added multi-object selection and batch operations to the 2D annotation workspace, with canvas and sidebar selection methods and undo and redo support (https://github.com/cvat-ai/cvat/pull/11084)

• Annotations from audio projects now can be exported (https://github.com/cvat-ai/cvat/pull/11238)

Changed

• Show icons and colors for job stages and states on the Jobs page and in stage/state selectors on the task job list (https://github.com/cvat-ai/cvat/pull/10981)

Fixed

• Invalid handling of the stop_frame option in task creation from PDF. It was producing 1 less frame than specified. (https://github.com/cvat-ai/cvat/issues/11175)

• Made audio interval Tab navigation deterministic by following the selected right bar ordering. (https://github.com/cvat-ai/cvat/pull/11236)

• Kept the layout toolbar attached to the top edge of the 3D canvas frame (https://github.com/cvat-ai/cvat/pull/11244)

• Restored the missing overlap threshold setting in point-based quality requirements. Improved documentation on the point and skeleton matching. (https://github.com/cvat-ai/cvat/pull/11259)

• Task quality settings now display the inherited project settings and requirements as read-only when "Use project settings" is enabled. (https://github.com/cvat-ai/cvat/pull/11256)

Security

• Prevented arbitrary file reads via imported dataset media paths (https://github.com/cvat-ai/cvat/security/advisories/GHSA-8p5p-rh49-4qc9)

• Prevented arbitrary file reads via symlinks in task data (https://github.com/cvat-ai/cvat/security/advisories/GHSA-rvw4-fjrw-pqf4)