Mergeback 1.9.1 to develop (#1623)

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

### Summary

<!--
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 -->
- [ ] I have added unit tests to cover my changes.​
- [ ] I have added integration tests to cover my changes.​
- [ ] 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

- [ ] 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: dependabot[bot] <support@github.com>
Signed-off-by: Ilya Trushkin <ilya.trushkin@intel.com>
Co-authored-by: Sooah Lee <sooah.lee@intel.com>
Co-authored-by: Ilya Trushkin <ilya.trushkin@intel.com>
Co-authored-by: williamcorsel <31770711+williamcorsel@users.noreply.github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
Co-authored-by: Wonju Lee <wonju.lee@intel.com>

CHANGELOG.md

@@ -21,7 +21,18 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ### Bug fixes
 
-## \[Q3 2024 Release 1.9.0\]
+## Q4 2024 Release 1.9.1
+### Enhancements
+- Support multiple labels for kaggle format
+  (<https://github.com/openvinotoolkit/datumaro/pull/1607>)
+- Use DataFrame.map instead of DataFrame.applymap
+  (<https://github.com/openvinotoolkit/datumaro/pull/1613>)
+
+### Bug fixes
+- Fix StreamDataset merging when importing in eager mode
+  (<https://github.com/openvinotoolkit/datumaro/pull/1609>)
+
+## Q3 2024 Release 1.9.0
 ### New features
 - Add a new CLI command: datum format
   (<https://github.com/openvinotoolkit/datumaro/pull/1570>)

docs/source/docs/release_notes.rst

@@ -4,6 +4,18 @@ Release Notes
 .. toctree::
    :maxdepth: 1
 
+v1.9.1 (2024 Q3)
+----------------
+
+Enhancements
+^^^^^^^^^^^^
+- Support multiple labels for kaggle format
+- Use DataFrame.map instead of DataFrame.applymap
+
+Bug fixes
+^^^^^^^^^
+- Fix StreamDataset merging when importing in eager mode
+
 v1.9.0 (2024 Q3)
 ----------------
 

src/datumaro/components/dataset.py

@@ -1023,17 +1023,22 @@ class _MergedStreamDataset(cls):
                 def __init__(self, *sources: IDataset):
                     from datumaro.components.hl_ops import HLOps
 
-                    self.merged = HLOps.merge(*sources, merge_policy=merge_policy)
+                    self._merged = HLOps.merge(*sources, merge_policy=merge_policy)
+                    self._data = self._merged._data
+                    self._env = env
+                    self._format = DEFAULT_FORMAT
+                    self._source_path = None
+                    self._options = {}
 
                 def __iter__(self):
-                    yield from self.merged
+                    yield from self._merged
 
                 @property
                 def is_stream(self):
                     return True
 
                 def subsets(self) -> Dict[str, DatasetSubset]:
-                    return self.merged.subsets()
+                    return self._merged.subsets()
 
             return _MergedStreamDataset(*sources)
 

src/datumaro/plugins/data_formats/kaggle/base.py

@@ -77,21 +77,43 @@ def _parse_bbox_coords(self, bbox_str):
         # expected to output [x1, y1, x2, y2]
         return [float(coord.strip()) for coord in coords]
 
-    def _load_annotations(self, datas: list, indices: Dict[str, int], bbox_flag: bool):
+    def _load_annotations(
+        self, datas: list, indices: Dict[str, Union[int, Dict[str, int]]], bbox_flag: bool
+    ):
         if "label" in indices:
-            label_name = str(datas[indices["label"]])
-            label, cat = self._label_cat.find(label_name)
-            if not cat:
-                self._label_cat.add(label_name)
-                label, _ = self._label_cat.find(label_name)
+            label_indices = indices["label"]
+            if isinstance(label_indices, dict):
+                labels = []
+                list_values = datas[1:]
+                index_to_label = {v: k for k, v in label_indices.items()}
+                present_labels = [
+                    index_to_label[i + 1] for i, value in enumerate(list_values) if value == "1"
+                ]
+
+                for label_name in present_labels:
+                    label, cat = self._label_cat.find(label_name)
+                    if not cat:
+                        self._label_cat.add(label_name)
+                        label, _ = self._label_cat.find(label_name)
+                    labels.append(Label(label=label))
+            else:
+                label_name = str(datas[indices["label"]])
+                label, cat = self._label_cat.find(label_name)
+                if not cat:
+                    self._label_cat.add(label_name)
+                    label, _ = self._label_cat.find(label_name)
         else:
             _, cat = self._label_cat.find("object")
             if not cat:
                 self._label_cat.add("object")
             label = 0
 
         if "label" in indices and not bbox_flag:
+            label_indices = indices["label"]
+            if isinstance(label_indices, dict):
+                return labels
             return Label(label=label)
+
         if bbox_flag:
             if "bbox" in indices:
                 coords = self._parse_bbox_coords(datas[indices["bbox"]])
@@ -125,7 +147,14 @@ def _load_items(self, ann_file: str, columns: Dict[str, Union[str, list]]):
 
         indices = {"media": df_fields.index(columns["media"])}
         if "label" in columns:
-            indices.update({"label": df_fields.index(columns["label"])})
+            label_columns = columns["label"]
+            if isinstance(label_columns, list):
+                indices_label = {}
+                for label in label_columns:
+                    indices_label[label] = df_fields.index(label)
+                indices.update({"label": indices_label})
+            else:
+                indices.update({"label": df_fields.index(label_columns)})
 
         bbox_flag = False
         bbox_index = columns.get("bbox")
@@ -165,16 +194,30 @@ def _load_items(self, ann_file: str, columns: Dict[str, Union[str, list]]):
                 continue
 
             ann = self._load_annotations(data_info, indices, bbox_flag)
-            self._ann_types.add(ann.type)
-            if item_id in items:
-                items[item_id].annotations.append(ann)
+            if isinstance(ann, list):
+                for label in ann:
+                    self._ann_types.add(label.type)
+                if item_id in items:
+                    for label in ann:
+                        items[item_id].annotations.append(label)
+                else:
+                    items[item_id] = DatasetItem(
+                        id=item_id,
+                        subset=self._subset,
+                        media=Image.from_file(path=media_path),
+                        annotations=ann,
+                    )
             else:
-                items[item_id] = DatasetItem(
-                    id=item_id,
-                    subset=self._subset,
-                    media=Image.from_file(path=media_path),
-                    annotations=[ann],
-                )
+                self._ann_types.add(ann.type)
+                if item_id in items:
+                    items[item_id].annotations.append(ann)
+                else:
+                    items[item_id] = DatasetItem(
+                        id=item_id,
+                        subset=self._subset,
+                        media=Image.from_file(path=media_path),
+                        annotations=[ann],
+                    )
         return items.values()
 
     def categories(self):

src/datumaro/plugins/transforms.py

@@ -1974,7 +1974,7 @@ def refine_tabular_media(self, item):
             or item.media.table.dtype(col) is int
         ]
 
-        df[str_cols] = df[str_cols].applymap(lambda x: self.remove_unnecessary_char(x))
+        df[str_cols] = df[str_cols].map(lambda x: self.remove_unnecessary_char(x))
 
         if not (self._outlier_value):
             self.check_outlier(media.table.data[float_cols + int_cols], float_cols + int_cols)

src/datumaro/version.py

@@ -1 +1 @@
-__version__ = "1.9.0"
+__version__ = "1.10.0.dev0"

tests/assets/kaggle_dataset/image_csv_multi_label/ann.csv

@@ -0,0 +1,7 @@
+image_name,dog,cat,person
+1.jpg,1,0,0
+2.jpg,0,1,0
+3.jpg,0,0,1
+4.jpg,1,1,0
+5.jpg,1,0,1
+6.jpg,0,1,1

tests/assets/kaggle_dataset/image_csv_multi_label/ann_wo_ext.csv

@@ -0,0 +1,7 @@
+image_name,dog,cat,person
+1,1,0,0
+2,0,1,0
+3,0,0,1
+4,1,1,0
+5,1,0,1
+6,0,1,1

tests/unit/data_formats/test_kaggle.py

@@ -20,6 +20,9 @@
 from tests.utils.test_utils import compare_datasets
 
 DUMMY_DATASET_IMAGE_CSV_DIR = get_test_asset_path("kaggle_dataset", "image_csv")
+DUMMY_DATASET_IMAGE_CSV_MULTI_LB_DIR = get_test_asset_path(
+    "kaggle_dataset", "image_csv_multi_label"
+)
 DUMMY_DATASET_IMAGE_CSV_DET_DIR = get_test_asset_path("kaggle_dataset", "image_csv_det")
 DUMMY_DATASET_IMAGE_TXT_DIR = get_test_asset_path("kaggle_dataset", "image_txt")
 DUMMY_DATASET_IMAGE_TXT_DET_DIR = get_test_asset_path("kaggle_dataset", "image_txt_det")
@@ -72,6 +75,51 @@ def fxt_img_dataset() -> Dataset:
     )
 
 
+@pytest.fixture
+def fxt_img_multi_label_dataset() -> Dataset:
+    return Dataset.from_iterable(
+        [
+            DatasetItem(
+                id="1",
+                subset="default",
+                media=Image.from_numpy(data=np.ones((5, 10, 3))),
+                annotations=[Label(label=0)],
+            ),
+            DatasetItem(
+                id="2",
+                subset="default",
+                media=Image.from_numpy(data=np.ones((5, 10, 3))),
+                annotations=[Label(label=1)],
+            ),
+            DatasetItem(
+                id="3",
+                subset="default",
+                media=Image.from_numpy(data=np.ones((5, 10, 3))),
+                annotations=[Label(label=2)],
+            ),
+            DatasetItem(
+                id="4",
+                subset="default",
+                media=Image.from_numpy(data=np.ones((5, 10, 3))),
+                annotations=[Label(label=0), Label(label=1)],
+            ),
+            DatasetItem(
+                id="5",
+                subset="default",
+                media=Image.from_numpy(data=np.ones((5, 10, 3))),
+                annotations=[Label(label=0), Label(label=2)],
+            ),
+            DatasetItem(
+                id="6",
+                subset="default",
+                media=Image.from_numpy(data=np.ones((5, 10, 3))),
+                annotations=[Label(label=1), Label(label=2)],
+            ),
+        ],
+        categories=["dog", "cat", "person"],
+    )
+
+
 @pytest.fixture
 def fxt_img_det_dataset() -> Dataset:
     return Dataset.from_iterable(
@@ -321,6 +369,8 @@ def fxt_coco_dataset() -> Dataset:
 IDS = [
     "IMAGE_CSV",
     "IMAGE_CSV_WO_EXT",
+    "IMAGE_CSV_MULTI_LB",
+    "IMAGE_CSV_MULTI_LB_WO_EXT",
     "IMAGE_CSV_DET",
     "IMAGE_CSV_DET2",
     "IMAGE_CSV_DET3",
@@ -372,6 +422,26 @@ def test_can_detect(self, fxt_dataset_dir: str):
                     "columns": {"media": "image_name", "label": "label_name"},
                 },
             ),
+            (
+                DUMMY_DATASET_IMAGE_CSV_MULTI_LB_DIR,
+                "images",
+                "fxt_img_multi_label_dataset",
+                KaggleImageCsvBase,
+                {
+                    "ann_file": osp.join(DUMMY_DATASET_IMAGE_CSV_MULTI_LB_DIR, "ann.csv"),
+                    "columns": {"media": "image_name", "label": ["dog", "cat", "person"]},
+                },
+            ),
+            (
+                DUMMY_DATASET_IMAGE_CSV_MULTI_LB_DIR,
+                "images",
+                "fxt_img_multi_label_dataset",
+                KaggleImageCsvBase,
+                {
+                    "ann_file": osp.join(DUMMY_DATASET_IMAGE_CSV_MULTI_LB_DIR, "ann_wo_ext.csv"),
+                    "columns": {"media": "image_name", "label": ["dog", "cat", "person"]},
+                },
+            ),
             (
                 DUMMY_DATASET_IMAGE_CSV_DET_DIR,
                 "images",

tests/unit/test_imagenet_format.py

@@ -7,6 +7,7 @@
 import pytest
 
 from datumaro.components.annotation import AnnotationType, Label, LabelCategories
+from datumaro.components.contexts.importer import ImportErrorPolicy
 from datumaro.components.dataset import Dataset, StreamDataset
 from datumaro.components.dataset_base import DatasetItem
 from datumaro.components.environment import Environment
@@ -214,7 +215,9 @@ def _create_expected_dataset(self):
     @pytest.mark.parametrize("dataset_cls, is_stream", [(Dataset, False), (StreamDataset, True)])
     def test_can_import(self, dataset_cls, is_stream, helper_tc):
         expected_dataset = self._create_expected_dataset()
-        dataset = dataset_cls.import_from(self.DUMMY_DATASET_DIR, self.IMPORTER_NAME)
+        dataset = dataset_cls.import_from(
+            self.DUMMY_DATASET_DIR, self.IMPORTER_NAME, error_policy=ImportErrorPolicy()
+        )
         assert dataset.is_stream == is_stream
 
         compare_datasets(helper_tc, expected_dataset, dataset, require_media=True)
@@ -240,7 +243,9 @@ class ImagenetWithSubsetDirsImporterTest(ImagenetImporterTest):
     @mark_requirement(Requirements.DATUM_GENERAL_REQ)
     @pytest.mark.parametrize("dataset_cls, is_stream", [(Dataset, False), (StreamDataset, True)])
     def test_can_import(self, dataset_cls, is_stream, helper_tc):
-        dataset = dataset_cls.import_from(self.DUMMY_DATASET_DIR, self.IMPORTER_NAME)
+        dataset = dataset_cls.import_from(
+            self.DUMMY_DATASET_DIR, self.IMPORTER_NAME, error_policy=ImportErrorPolicy()
+        )
         assert dataset.is_stream == is_stream
 
         for subset_name, subset in dataset.subsets().items():