Skip to content

segmentation

SegmentationDatasetArguments dataclass

Segmentation dataset arguments.

Parameters:

  • train_samples (List[int]) –

    List of samples per class in train set, element at index 0 are good samples

  • val_samples (Optional[List[int]], default: None ) –

    List of samples per class in validation set, same as above.

  • test_samples (Optional[List[int]], default: None ) –

    List of samples per class in test set, same as above.

  • classes (Optional[List[str]], default: None ) –

    Optional list of class names, must be equal to len(train_samples) - 1

base_binary_segmentation_dataset(tmp_path, request)

Generate a base binary segmentation dataset with the following structure
  • 10 good and 10 bad samples in train set
  • 10 good and 10 bad samples in validation set
  • 10 good and 10 bad samples in test set
  • 2 classes: good and bad.

Parameters:

  • tmp_path (Path) –

    path to temporary directory

  • request (Any) –

    pytest request

Yields:

Source code in quadra/utils/tests/fixtures/dataset/segmentation.py 
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
@pytest.fixture(
    params=[
        SegmentationDatasetArguments(
            **{"train_samples": [10, 10], "val_samples": [10, 10], "test_samples": [10, 10], "classes": ["bad"]}
        )
    ]
)
def base_binary_segmentation_dataset(
    tmp_path: Path, request: Any
) -> Tuple[str, SegmentationDatasetArguments, Dict[str, int]]:
    """Generate a base binary segmentation dataset with the following structure:
        - 10 good and 10 bad samples in train set
        - 10 good and 10 bad samples in validation set
        - 10 good and 10 bad samples in test set
        - 2 classes: good and bad.

    Args:
        tmp_path: path to temporary directory
        request: pytest request

    Yields:
        Tuple containing path to dataset, dataset arguments and class to index mapping
    """
    yield _build_segmentation_dataset(tmp_path, request.param)
    if tmp_path.exists():
        shutil.rmtree(tmp_path)

base_multiclass_segmentation_dataset(tmp_path, request)

Generate a base binary segmentation dataset with the following structure
  • 10 good, 15 defect_1 and 20 defect_2 samples in train set
  • 5 good, 10 defect_1 and 15 defect_2 samples in validation set
  • 10 good, 10 defect_1 and 10 defect_2 samples in test set
  • 3 classes: good, defect_1 and defect_2.

Parameters:

  • tmp_path (Path) –

    path to temporary directory

  • request (Any) –

    pytest request

Yields:

Source code in quadra/utils/tests/fixtures/dataset/segmentation.py 
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
@pytest.fixture(
    params=[
        SegmentationDatasetArguments(
            **{
                "train_samples": [10, 15, 20],
                "val_samples": [5, 10, 15],
                "test_samples": [10, 10, 10],
                "classes": ["defect_1", "defect_2"],
            }
        )
    ]
)
def base_multiclass_segmentation_dataset(
    tmp_path: Path, request: Any
) -> Tuple[str, SegmentationDatasetArguments, Dict[str, int]]:
    """Generate a base binary segmentation dataset with the following structure:
        - 10 good, 15 defect_1 and 20 defect_2 samples in train set
        - 5 good, 10 defect_1 and 15 defect_2 samples in validation set
        - 10 good, 10 defect_1 and 10 defect_2 samples in test set
        - 3 classes: good, defect_1 and defect_2.

    Args:
        tmp_path: path to temporary directory
        request: pytest request

    Yields:
        Tuple containing path to dataset, dataset arguments and class to index mapping
    """
    yield _build_segmentation_dataset(tmp_path, request.param)
    if tmp_path.exists():
        shutil.rmtree(tmp_path)

segmentation_dataset(tmp_path, dataset_arguments)

Fixture to dinamically generate a segmentation dataset. By default generated images are 224x224 pixels and associated masks contains a 50x50 pixels square with the corresponding image class, so at the current stage is not possible to have images with multiple annotations. Split files are saved as train.txt, val.txt and test.txt.

Parameters:

Yields:

Source code in quadra/utils/tests/fixtures/dataset/segmentation.py 
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
@pytest.fixture
def segmentation_dataset(
    tmp_path: Path, dataset_arguments: SegmentationDatasetArguments
) -> Tuple[str, SegmentationDatasetArguments, Dict[str, int]]:
    """Fixture to dinamically generate a segmentation dataset. By default generated images are 224x224 pixels
        and associated masks contains a 50x50 pixels square with the corresponding image class, so at the current stage
        is not possible to have images with multiple annotations. Split files are saved as train.txt,
        val.txt and test.txt.

    Args:
        tmp_path: path to temporary directory
        dataset_arguments: dataset arguments

    Yields:
        Tuple containing path to dataset, dataset arguments and class to index mapping
    """
    yield _build_segmentation_dataset(tmp_path, dataset_arguments)
    if tmp_path.exists():
        shutil.rmtree(tmp_path)