cvat-ai/cvat v2.60.0

on GitHub

Added

• Docker + compose for YOLO (#10325)

• Interactors can now return multiple shapes instead of a single one (#10329)

• Interactors can now return a confidence attribute. If present, interaction results can be filtered in the UI (#10329)

• Added UI controls to remove point and rectangle prompts when using Interactors or OpenCV tools (#10329)

• Server may add text prompts to interactors as text_prompts field in POST: /api/lambda/functions/<id> (#10329)

Changed

• Unified the expected interactor interface to align with the /annotations API and AI detector outputs (#10329)

• Updated the IoG serverless function to support the new interactor interface (#10329)

Fixed

• Reduced memory usage by replacing JavaScript arrays with typed arrays (#10329)

• Fixed memory leaks related to unreleased object URLs in cvat-canvas (#10329)

• [SDK] Fixed a crash in TasksRepo.create_from_backup, ProjectsRepo.create_from_backup, Task.upload_data that could occur if a recoverable error occurred during chunk uploading (#10375)