New features
- Support for custom media types, new `PointCloud` media type,
`DatasetItem.media` and `.media_as(type)` members
(<https://github.com/openvinotoolkit/datumaro/pull/539>)
- \[API\] A way to request dataset and extractor media type with `media_type`
(<https://github.com/openvinotoolkit/datumaro/pull/539>)
- BraTS format (import-only) (.npy and .nii.gz), new `MultiframeImage`
media type (<https://github.com/openvinotoolkit/datumaro/pull/628>)
- Common Semantic Segmentation dataset format (import-only)
(<https://github.com/openvinotoolkit/datumaro/pull/685>)
- An option to disable `data/` prefix inclusion in YOLO export
(<https://github.com/openvinotoolkit/datumaro/pull/689>)
- New command `describe-downloads` to print information about downloadable datasets
(<https://github.com/openvinotoolkit/datumaro/pull/678>)
- Detection for Cityscapes format
(<https://github.com/openvinotoolkit/datumaro/pull/680>)
- Maximum recursion `--depth` parameter for `detect-dataset` CLI command
(<https://github.com/openvinotoolkit/datumaro/pull/680>)
- An option to save a single subset in the `download` command
(<https://github.com/openvinotoolkit/datumaro/pull/697>)
- Common Super Resolution dataset format (import-only)
(<https://github.com/openvinotoolkit/datumaro/pull/700>)
- Kinetics 400/600/700 dataset format (import-only)
(<https://github.com/openvinotoolkit/datumaro/pull/706>)
- NYU Depth Dataset V2 format (import-only)
(<https://github.com/openvinotoolkit/datumaro/pull/712>)
Enhancements
- `env.detect_dataset()` now returns a list of detected formats at all recursion levels
instead of just the lowest one
(<https://github.com/openvinotoolkit/datumaro/pull/680>)
- Open Images: allowed to store annotations file in root path as well
(<https://github.com/openvinotoolkit/datumaro/pull/680>)
- Improved parsing error messages in COCO, VOC and YOLO formats
(<https://github.com/openvinotoolkit/datumaro/pull/684>,
<https://github.com/openvinotoolkit/datumaro/pull/686>,
<https://github.com/openvinotoolkit/datumaro/pull/687>)
- YOLO format now supports almost any subset names, except `backup`, `names` and `classes`
(instead of just `train` and `valid`). The reserved names now raise an error on exporting.
(<https://github.com/openvinotoolkit/datumaro/pull/688>)
Deprecated
- `--save-images` is replaced with `--save-media` in CLI and converter API
(<https://github.com/openvinotoolkit/datumaro/pull/539>)
- \[API\] `image`, `point_cloud` and `related_images` of `DatasetItem` are
replaced with `media` and `media_as(type)` members and c-tor parameters
(<https://github.com/openvinotoolkit/datumaro/pull/539>)
Removed
- N/A
Bug fixes
- Detection for LFW format
(<https://github.com/openvinotoolkit/datumaro/pull/680>)
- Adding depth value of image when dataset is exported in VOC format
(<https://github.com/openvinotoolkit/datumaro/pull/726>)
- Adding to handle the numerical labels in task chains properly
(<https://github.com/openvinotoolkit/datumaro/pull/726>)
- Fixing the issue that annotations inside another annotation (polygon)
are duplicated during import for VOC format
(<https://github.com/openvinotoolkit/datumaro/pull/726>)
Security
- N/A