Change common semantic segmentation dataset detection rule (#1572)

Since Datumaro 1.8, the following folder structure:
```
# Common semantic segmentation dataset example from OTX tests
.
└── supervised
    ├── test
    │   ├── dataset_meta.json
    │   ├── images
    │   │   ├── 0001.png
    │   │   └── 0002.png
    │   └── masks
    │       ├── 0001.png
    │       └── 0002.png
    ├── train
    │   ├── dataset_meta.json
    │   ├── images
    │   │   ├── 0001.png
    │   │   ├── 0002.png
    │   │   └── 0003.png
    │   └── masks
    │       ├── 0001.png
    │       ├── 0002.png
    │       └── 0003.png
    └── val
        ├── dataset_meta.json
        ├── images
        │   ├── 0001.png
        │   └── 0002.png
        └── masks
            ├── 0001.png
            └── 0002.png
```
is detected as both `common_semantic_segmentation` and
`common_semantic_segmentation_with_subset_dirs`. Since they are detected
at different levels, the current implementation could not set the higher
priority to the second one.

This PR updates detection rules for the `common_semantic_segmentation`
format, searching for requirements in the root directory only.

src/datumaro/plugins/data_formats/common_semantic_segmentation.py

@@ -3,7 +3,6 @@
 # SPDX-License-Identifier: MIT
 
 import errno
-import glob
 import os.path as osp
 from typing import List, Optional
 
@@ -69,11 +68,11 @@ def __init__(
         self._image_prefix = image_prefix
         self._mask_prefix = mask_prefix
 
-        meta_file = glob.glob(osp.join(path, "**", DATASET_META_FILE), recursive=True)
-        if is_meta_file(meta_file[0]):
-            self._root_dir = osp.dirname(meta_file[0])
+        meta_file = osp.join(path, DATASET_META_FILE)
+        if is_meta_file(meta_file):
+            self._root_dir = osp.dirname(meta_file)
 
-            label_map = parse_meta_file(meta_file[0])
+            label_map = parse_meta_file(meta_file)
             self._categories = make_categories(label_map)
         else:
             raise FileNotFoundError(errno.ENOENT, "Dataset meta info file was not found", path)
@@ -163,11 +162,10 @@ def build_cmdline_parser(cls, **kwargs):
 
     @classmethod
     def detect(cls, context: FormatDetectionContext) -> FormatDetectionConfidence:
-        path = context.require_file(f"**/{DATASET_META_FILE}")
-        path = osp.dirname(path)
+        context.require_file(DATASET_META_FILE)
 
-        context.require_file(osp.join(path, CommonSemanticSegmentationPath.IMAGES_DIR, "**", "*"))
-        context.require_file(osp.join(path, CommonSemanticSegmentationPath.MASKS_DIR, "**", "*"))
+        context.require_file(osp.join(CommonSemanticSegmentationPath.IMAGES_DIR, "**", "*"))
+        context.require_file(osp.join(CommonSemanticSegmentationPath.MASKS_DIR, "**", "*"))
 
         return FormatDetectionConfidence.MEDIUM
 

tests/unit/data_formats/conftest.py

@@ -10,6 +10,8 @@
 
 from datumaro import Dataset
 
+from tests.utils.test_utils import TestDir
+
 
 @pytest.fixture
 def fxt_dummy_dataset():
@@ -35,12 +37,12 @@ def fxt_export_kwargs():
 @pytest.fixture
 def fxt_dataset_dir_with_subset_dirs(test_dir: str, request: pytest.FixtureRequest):
     fxt_dataset_dir = request.param
+    with TestDir(f"{test_dir}_with_subsets") as new_test_dir:
+        for subset in ["train", "val", "test"]:
+            dst = os.path.join(new_test_dir, subset)
+            shutil.copytree(fxt_dataset_dir, dst)
 
-    for subset in ["train", "val", "test"]:
-        dst = os.path.join(test_dir, subset)
-        shutil.copytree(fxt_dataset_dir, dst)
-
-    yield test_dir
+        yield new_test_dir
 
 
 @pytest.fixture

tests/unit/data_formats/test_common_semantic_segmentation_format.py

@@ -2,15 +2,18 @@
 #
 # SPDX-License-Identifier: MIT
 
+import os
+import shutil
 from collections import OrderedDict
-from typing import Any, Dict, Optional
+from typing import Any, Dict
 
 import numpy as np
 import pytest
 
 from datumaro.components.annotation import Mask
 from datumaro.components.dataset import Dataset
 from datumaro.components.dataset_base import DatasetItem
+from datumaro.components.errors import DatasetImportError
 from datumaro.components.media import Image
 from datumaro.plugins.data_formats.common_semantic_segmentation import (
     CommonSemanticSegmentationImporter,
@@ -143,6 +146,40 @@ def test_can_import(
             fxt_dataset_dir, fxt_expected_dataset, fxt_import_kwargs, request
         )
 
+    @pytest.mark.parametrize(
+        ["fxt_dataset_dir", "fxt_expected_dataset", "fxt_import_kwargs"],
+        [
+            (DUMMY_DATASET_DIR, "fxt_dataset", {}),
+            (
+                DUMMY_NON_STANDARD_DATASET_DIR,
+                "fxt_non_standard_dataset",
+                {"image_prefix": "image_", "mask_prefix": "gt_"},
+            ),
+        ],
+        indirect=["fxt_expected_dataset"],
+        ids=IDS,
+    )
+    def test_cannot_import_nested(
+        self,
+        fxt_dataset_dir: str,
+        fxt_expected_dataset: Dataset,
+        fxt_import_kwargs: Dict[str, Any],
+        request: pytest.FixtureRequest,
+        test_dir: str,
+    ):
+        shutil.copytree(fxt_dataset_dir, test_dir, dirs_exist_ok=True)
+        subdir_name = "subdir"
+        subdir = os.path.join(test_dir, subdir_name)
+        os.makedirs(subdir)
+        for _file in os.listdir(test_dir):
+            if _file != subdir_name:
+                file_path = os.path.join(test_dir, _file)
+                shutil.move(file_path, subdir)
+        with pytest.raises(DatasetImportError) as exc_info:
+            super().test_can_import(test_dir, fxt_expected_dataset, fxt_import_kwargs, request)
+        assert exc_info.value.__cause__ is not None
+        assert isinstance(exc_info.value.__cause__, FileNotFoundError)
+
 
 class CommonSemanticSegmentationWithSubsetDirsImporterTest(TestDataFormatBase):
     IMPORTER = CommonSemanticSegmentationWithSubsetDirsImporter