[FIX] Classification Auto-split bug fix (#1824)

* doc update, fix autosplit

* Fix pre-commit

* Add ignore

* Fix figure

* Fix minor

* Fix pre-commit

* Fix integ test

* Fix pre-commit

Author: sungmanc

.gitignore

@@ -40,3 +40,7 @@ dmypy.json
 otx/**/*.c
 otx/**/*.html
 otx/**/*.so
+
+
+# Dataset made by unit-test
+tests/**/detcon_mask/*

docs/source/guide/tutorials/base/how_to_train/classification.rst

@@ -56,8 +56,7 @@ with the following command:
   cd ..
 
 |
-
-.. image:: ../../../../../utils/images/flowers.jpg
+.. image:: ../../../../../utils/images/flowers_example.jpg
   :width: 600
 
 |
@@ -121,7 +120,7 @@ Let's prepare an OpenVINO™ Training Extensions classification workspace runnin
 
   (otx) ...$ cd ./otx-workspace-CLASSIFICATION
 
-It will create **otx-workspace-CLASSIFICATION** with all necessary configs for MobileNet-V3-large-1x, prepared ``data.yaml`` to simplify CLI commands launch and splitted dataset.
+It will create **otx-workspace-CLASSIFICATION** with all necessery configs for MobileNet-V3-large-1x, prepared ``data.yaml`` to simplify CLI commands launch and splitted dataset named ``splitted_dataset``.
 
 3. To start training we need to call ``otx train``
 command in our workspace:
@@ -252,4 +251,4 @@ Please note, that POT will take some time (generally less than NNCF optimization
 efficient model representation ready-to-use classification model.
 
 The following tutorials provide further steps on how to :doc:`deploy <../deploy>` and use your model in the :doc:`demonstration mode <../demo>` and visualize results.
-The examples are provided with an object detection model, but it is easy to apply them for classification by substituting the object detection model with classification one.
+The examples are provided with an object detection model, but it is easy to apply them for classification by substituting the object detection model with classification one.

docs/utils/images/flowers_example.jpg

@@ -169,7 +169,7 @@ def configure_data_config(self, update_data_yaml: bool = True) -> None:
             use_auto_split = data_yaml["data"]["train"]["data-roots"] and not data_yaml["data"]["val"]["data-roots"]
             # FIXME: Hardcoded for Self-Supervised Learning
             if use_auto_split and str(self.train_type).upper() != "SELFSUPERVISED":
-                splitted_dataset = self.auto_split_data(data_yaml["data"]["train"]["data-roots"], self.task_type)
+                splitted_dataset = self.auto_split_data(data_yaml["data"]["train"]["data-roots"], str(self.task_type))
                 default_data_folder_name = "splitted_dataset"
                 data_yaml = self._get_arg_data_yaml()
                 self._save_data(splitted_dataset, default_data_folder_name, data_yaml)
@@ -202,7 +202,6 @@ def auto_task_detection(self, data_roots: str) -> str:
         if not data_roots:
             raise ValueError("Workspace must already exist or one of {task or model or train-data-roots} must exist.")
         self.data_format = self.dataset_manager.get_data_format(data_roots)
-        print(f"[*] Detected dataset format: {self.data_format}")
         return self._get_task_type_from_data_format(self.data_format)
 
     def _get_task_type_from_data_format(self, data_format: str) -> str:

otx/cli/manager/config_manager.py

@@ -6,7 +6,7 @@
 
 # pylint: disable=invalid-name
 import os
-from typing import List, Tuple
+from typing import List, Tuple, Union
 
 import datumaro
 from datumaro.components.dataset import Dataset, DatasetSubset
@@ -31,12 +31,12 @@ def get_train_dataset(dataset: Dataset) -> DatasetSubset:
         raise ValueError("Can't find training data.")
 
     @staticmethod
-    def get_val_dataset(dataset: Dataset) -> DatasetSubset:
+    def get_val_dataset(dataset: Dataset) -> Union[DatasetSubset, None]:
         """Returns validation dataset."""
         for k, v in dataset.subsets().items():
-            if "val" in k or "default" in k:
+            if "val" in k:
                 return v
-        raise ValueError("Can't find validation data.")
+        return None
 
     @staticmethod
     def get_data_format(data_root: str) -> str:
@@ -57,6 +57,7 @@ def get_data_format(data_root: str) -> str:
             data_formats = datumaro.Environment().detect_dataset(data_root)
             # TODO: how to avoid hard-coded part
             data_format = data_formats[0] if "imagenet" not in data_formats else "imagenet"
+        print(f"[*] Detected dataset format: {data_format}")
         return data_format
 
     @staticmethod

otx/core/data/manager/dataset_manager.py

@@ -57,8 +57,7 @@ def test_get_train_dataset(self, task: List[str], subset: List[str]):
     @pytest.mark.parametrize("subset", AVAILABLE_SUBSETS)
     def test_get_val_dataset(self, task: List[str], subset: List[str]):
         if subset == "train":
-            with pytest.raises(ValueError, match="Can't find validation data."):
-                DatasetManager.get_val_dataset(self.dataset[subset][task])
+            assert DatasetManager.get_val_dataset(self.dataset[subset][task]) is None
         else:
             val_dataset = DatasetManager.get_val_dataset(self.dataset[subset][task])
             assert isinstance(val_dataset, dm.DatasetSubset)