Backlog

[zhiltsov-max]

A list of code and architecture improvements, refactorings etc.:

• [x] Provide Typing annotations for API
• [x] Provide handy decorators for extractors, converters, importers
• [x] Simplify implementation of extractors, converters, importers
• [ ] More advanced support of annotation attributes - default values, type, possible values
• [x] Image processing transforms (lazy) (#581)
• [ ] occluded as a class member for shapes and masks instead of a free attribute
• [ ] Possible, dot access for annotation attributes
• [ ] Annotation properties along with attributes so attributes could save their specific meaning
• [x] Regularize tests for dataset formats, simplify code
  • Common dataset templates (task specific - for detection, classification, etc.)
  • Checks for Cyrillic and whitespace-separated image names and labels
• [x] Search for an image with unknown extension in format import
• [x] Separate image reading from annotation reading?
• [x] RLE for masks by default
• [ ] Source info from extractors
• [ ] Remove AnnotationType, just use types
• [x] Support track annotations? Possibly, as dataset-level annotations or groups.
• [x] Support label-less datasets in various formats?
• [ ] Adding / removal of image watermarks
• [ ] gather singular points to skeleton transform
• [ ] Dataset subdivisions, "virtual" subsets (non-owning views)
• [x] Add dataset format documentations (borrow from https://github.com/openvinotoolkit/cvat/blob/develop/cvat/apps/dataset_manager/formats/README.md)
• [ ] Optimization of PIL.Image.to_array (PIL is replaced with OpenCV in most cases)

CLI:

• [ ] Cache control commands
• [ ] Source configuration
• [x] Import source as project (extractor detection)