Releases: cvat-ai/cvat
Releases · cvat-ai/cvat
v0.4.0
Added
- OpenVINO auto annotation. It is possible to upload a custom DL model and annotate images automatically.
- Ability to rotate images/frames in the client part (Ctrl+R, Shift+Ctrl+R shortcuts) (#305)
- ReID application for automatic bounding boxes merging (#299)
- Keyboard shortcuts to switch next/previous default shape type (box, polygon etc) [Alt + <, Alt + >] (#316)
- Converter for VOC now supports interpolation tracks
- REST API (/api/v1/*, /api/docs)
- Semi-automatic semantic segmentation with the Deep Extreme Cut work
- Converter to tfrecords format to cvat/utils (#368)
- Support for 16bit grayscale images for annotation (#342)
Changed
- Python Django server design and DB layout
- Propagation input moved from settings to the bottom player panel
- Additional events like "Debug Info" or "Fit Image" for analytics
- LFS is optional for git annotation storages (#314)
Deprecated
- "Flip images" flag in the create task dialog. Rotation functionality in client part added instead.
v0.3.0
Added
- Ability to copy Object URL and Frame URL via object context menu and player context menu respectively.
- Ability to change opacity for selected shape with help "Selected Fill Opacity" slider.
- Ability to remove polyshapes points by double click.
- Ability to draw/change polyshapes (except for points) by slip method. Just press ENTER and moving a cursor.
- Ability to switch lock/hide properties via label UI element (in right menu) for all objects with same label.
- Shortcuts for outside/keyframe properties
- Support of Intel OpenVINO for accelerated model inference
- Tensorflow annotation now works without CUDA. It can use CPU only. OpenVINO and CUDA are supported optionally.
- Incremental saving of annotations.
- Tutorial for using polygons (screencast)
- Silk profiler to improve development process
- Admin panel can be used to edit labels and attributes for annotation tasks
- Analytics component to manage a data annotation team, monitor exceptions, collect client and server logs
- Changeable job and task statuses (annotation, validation, completed). A job status can be changed manually, a task status is computed automatically based on job statuses (#153)
- Backlink to a task from its job annotation view (#156)
- Buttons lock/hide for labels. They work for all objects with the same label on a current frame (#116)
Changed
- Polyshape editing method has been improved. You can redraw part of shape instead of points cloning.
- Unified shortcut (Esc) for close any mode instead of different shortcuts (Alt+N, Alt+G, Alt+M etc.).
- Dump file contains information about data source (e.g. video name, archive name, ...)
- Update requests library due to https://nvd.nist.gov/vuln/detail/CVE-2018-18074
- Per task/job permissions to create/access/change/delete tasks and annotations
- Documentation was improved
- Timeout for creating tasks was increased (from 1h to 4h) (#136)
- Drawing has become more convenience. Now it is possible to draw outside an image. Shapes will be automatically truncated after drawing process (#202)
Fixed
- Performance bottleneck has been fixed during you create new objects (draw, copy, merge etc).
- Label UI elements aren't updated after changelabel.
- Attribute annotation mode can use invalid shape position after resize or move shapes.
- Labels order is preserved now (#242)
- Uploading large XML files (#123)
- Django vulnerability (#121)
- Grammatical cleanup of README.md (#107)
- Dashboard loading has been accelerated (#156)
- Text drawing outside of a frame in some cases (#202)