Add label groups for hierarchical classification in ImageNet (#1645)

<!-- Contributing guide:
https://github.com/openvinotoolkit/datumaro/blob/develop/CONTRIBUTING.md
-->

### Summary
This PR adds grouping of labels (directories for the ImageNet case) by
their groups (parent directories).
For example, for the following folder structure
```
.
├── label_1
│   └── label_1_1
│       └── 1.jpg
└── label_2
    └── label_2_1
        └── 2.jpg
```
label groups will be `label_1` and `label_2`.


**Note**: for the higher depth of nesting, names of groups will be
relative paths of second-to-last directories. For the following case:
```
.
├── label_1
│   └── label_1_1
│       └── label_1_1_1
│           └── 1.jpg
│        
└── label_2
    └── label_2_1
        └── label_2_1_1
            └── 2.jpg
```
label groups will be `label_1/label_1_1` and `label_2/label_2_1`.
<!--
Resolves #111 and #222.
Depends on #1000 (for series of dependent commits).

This PR introduces this capability to make the project better in this
and that.

- Added this feature
- Removed that feature
- Fixed the problem #1234
-->

### How to test
<!-- Describe the testing procedure for reviewers, if changes are
not fully covered by unit tests or manual testing can be complicated.
-->

### Checklist
<!-- Put an 'x' in all the boxes that apply -->
- [x] I have added unit tests to cover my changes.​
- [ ] I have added integration tests to cover my changes.​
- [x] I have added the description of my changes into
[CHANGELOG](https://github.com/openvinotoolkit/datumaro/blob/develop/CHANGELOG.md).​
- [ ] I have updated the
[documentation](https://github.com/openvinotoolkit/datumaro/tree/develop/docs)
accordingly

### License

- [x] I submit _my code changes_ under the same [MIT
License](https://github.com/openvinotoolkit/datumaro/blob/develop/LICENSE)
that covers the project.
  Feel free to contact the maintainers if that's a concern.
- [ ] I have updated the license header for each file (see an example
below).

```python
# Copyright (C) 2024 Intel Corporation
#
# SPDX-License-Identifier: MIT
```

---------

Signed-off-by: Ilya Trushkin <[email protected]>

CHANGELOG.md

@@ -14,6 +14,8 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
   (<https://github.com/openvinotoolkit/datumaro/pull/1594>)
 - Convert Cuboid2D annotation to/from 3D data
   (<https://github.com/openvinotoolkit/datumaro/pull/1639>)
+- Add label groups for hierarchical classification in ImageNet
+  (<https://github.com/openvinotoolkit/datumaro/pull/1645>)
 
 ### Enhancements
 - Enhance 'id_from_image_name' transform to ensure each identifier is unique

src/datumaro/plugins/data_formats/imagenet.py

@@ -48,8 +48,18 @@ def _load_categories(self, path):
         path = Path(path)
         for dirname in sorted(d for d in path.rglob("*") if d.is_dir()):
             dirname = dirname.relative_to(path)
+            level = len(dirname.parts)
             if str(dirname) != ImagenetPath.IMAGE_DIR_NO_LABEL:
-                label_cat.add(str(dirname))
+                parent = None
+                if level > 1:
+                    parent = str(dirname.parents[0])
+                    if not any([g.name == parent for g in label_cat.label_groups]):
+                        label_cat.add_label_group(parent, [str(dirname.name)], group_type=0)
+                    else:
+                        g = next(x for x in label_cat.label_groups if x.name == parent)
+                        g.labels.append(str(dirname.name))
+                label_cat.add(str(dirname), parent)
+
         return {AnnotationType.label: label_cat}
 
     def _load_items(self, path):

tests/unit/test_imagenet_format.py

@@ -182,6 +182,12 @@ class ImagenetImporterTest:
     IMPORTER_NAME = ImagenetImporter.NAME
 
     def _create_expected_dataset(self):
+        label_categories = LabelCategories.from_iterable(
+            ("label_0", "label_1", f"{Path('label_1', 'label_1_1')}")
+        )
+        label_categories[-1].parent = "label_1"
+        label_categories.add_label_group(name="label_1", labels=["label_1_1"], group_type=0)
+
         return Dataset.from_iterable(
             [
                 DatasetItem(
@@ -204,11 +210,7 @@ def _create_expected_dataset(self):
                     annotations=[Label(1)],
                 ),
             ],
-            categories={
-                AnnotationType.label: LabelCategories.from_iterable(
-                    ("label_0", "label_1", f"{Path('label_1', 'label_1_1')}")
-                ),
-            },
+            categories={AnnotationType.label: label_categories},
         )
 
     @mark_requirement(Requirements.DATUM_GENERAL_REQ)

tests/utils/test_utils.py

@@ -108,6 +108,17 @@ def compare_categories(test, expected, actual):
             sorted(expected[AnnotationType.label].items, key=lambda t: t.name),
             sorted(actual[AnnotationType.label].items, key=lambda t: t.name),
         )
+        if expected[AnnotationType.label].label_groups:
+            assert len(expected[AnnotationType.label].label_groups) == len(
+                actual[AnnotationType.label].label_groups
+            )
+            for expected_group, actual_group in zip(
+                expected[AnnotationType.label].label_groups,
+                actual[AnnotationType.label].label_groups,
+            ):
+                test.assertEqual(set(expected_group.labels), set(actual_group.labels))
+                test.assertEqual(expected_group.group_type, actual_group.group_type)
+
     if AnnotationType.mask in expected:
         test.assertEqual(
             expected[AnnotationType.mask].colormap,