classification

ClassificationDatasetArguments dataclass

Classification dataset arguments.

Parameters:

• samples (List[int]) –

  number of samples per class

• classes (Optional[List[str]]) –

  class names, if set it must be the same length as samples

• val_size (Optional[float]) –

  validation set size

• test_size (Optional[float]) –

  test set size

ClassificationMultilabelDatasetArguments dataclass

Classification dataset arguments.

Parameters:

• samples (List[int]) –

  number of samples per class

• classes (Optional[List[str]]) –

  class names, if set it must be the same length as samples

• val_size (Optional[float]) –

  validation set size

• test_size (Optional[float]) –

  test set size

• percentage_other_classes (Optional[float]) –

  probability of adding other classes to the labels of each sample

ClassificationPatchDatasetArguments dataclass

Classification patch dataset arguments.

Parameters:

• samples (List[int]) –

  number of samples per class

• overlap (float) –

  overlap between patches

• patch_size (Optional[Tuple[int, int]]) –

  patch size

• patch_number (Optional[Tuple[int, int]]) –

  number of patches

• classes (Optional[List[str]]) –

  class names, if set it must be the same length as samples

• val_size (Optional[float]) –

  validation set size

• test_size (Optional[float]) –

  test set size

• annotated_good (Optional[List[str]]) –

  list of class names that are considered as good annotations (E.g. ["good"])

base_classification_dataset(tmp_path, request)

Generate base classification dataset with the following parameters

• 10 samples per class
• 2 classes (class_1 and class_2) By default generated images are grayscale and 10x10 pixels.

Parameters:

• tmp_path (Path) –

  path to temporary directory

• request (Any) –

  pytest request

Yields:

• Tuple[str, ClassificationDatasetArguments] –

  Tuple containing path to created dataset and dataset arguments

Source code in quadra/utils/tests/fixtures/dataset/classification.py

@pytest.fixture(
    params=[
        ClassificationDatasetArguments(
            **{"samples": [10, 10], "classes": ["class_1", "class_2"], "val_size": 0.1, "test_size": 0.1}
        )
    ]
)
def base_classification_dataset(tmp_path: Path, request: Any) -> Tuple[str, ClassificationDatasetArguments]:
    """Generate base classification dataset with the following parameters:
        - 10 samples per class
        - 2 classes (class_1 and class_2)
        By default generated images are grayscale and 10x10 pixels.

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

    Yields:
        Tuple containing path to created dataset and dataset arguments
    """
    yield _build_classification_dataset(tmp_path, request.param)
    if tmp_path.exists():
        shutil.rmtree(tmp_path)

base_multilabel_classification_dataset(tmp_path, request)

Fixture to generate base multilabel classification dataset with the following parameters

• 10 samples per class
• 3 classes (class_1, class_2 and class_3)
• 10% of samples in validation set
• 10% of samples in test set
• 30% of possibility to add each other class to the sample By default generated images are grayscale and 10x10 pixels.

Parameters:

• tmp_path (Path) –

  path to temporary directory

• request (Any) –

  pytest request

Yields:

• Tuple[str, ClassificationMultilabelDatasetArguments] –

  Tuple containing path to created dataset and dataset arguments

Source code in quadra/utils/tests/fixtures/dataset/classification.py

@pytest.fixture(
    params=[
        ClassificationMultilabelDatasetArguments(
            **{
                "samples": [10, 10, 10],
                "classes": ["class_1", "class_2", "class_3"],
                "val_size": 0.1,
                "test_size": 0.1,
                "percentage_other_classes": 0.3,
            }
        )
    ]
)
def base_multilabel_classification_dataset(
    tmp_path: Path, request: Any
) -> Tuple[str, ClassificationMultilabelDatasetArguments]:
    """Fixture to generate base multilabel classification dataset with the following parameters:
        - 10 samples per class
        - 3 classes (class_1, class_2 and class_3)
        - 10% of samples in validation set
        - 10% of samples in test set
        - 30% of possibility to add each other class to the sample
        By default generated images are grayscale and 10x10 pixels.

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

    Yields:
        Tuple containing path to created dataset and dataset arguments
    """
    yield _build_multilabel_classification_dataset(tmp_path, request.param)
    if tmp_path.exists():
        shutil.rmtree(tmp_path)

base_patch_classification_dataset(tmp_path, request)

Generate a classification patch dataset with the following parameters

• 3 classes named bg, a and b
• 15, 20 and 25 samples for each class
• patch size of 16x20
• 60% overlap
• 10% validation set
• 10% test set.

Parameters:

• tmp_path (Path) –

  path to temporary directory

• request (Any) –

  pytest SubRequest object

Source code in quadra/utils/tests/fixtures/dataset/classification.py

@pytest.fixture(
    params=[
        ClassificationPatchDatasetArguments(
            **{
                "samples": [15, 20, 25],
                "classes": ["bg", "a", "b"],
                "patch_size": [16, 20],
                "overlap": 0.6,
                "val_size": 0.1,
                "test_size": 0.1,
            }
        )
    ]
)
def base_patch_classification_dataset(
    tmp_path: Path, request: Any
) -> Tuple[str, ClassificationDatasetArguments, Dict[str, int]]:
    """Generate a classification patch dataset with the following parameters:
        - 3 classes named bg, a and b
        - 15, 20 and 25 samples for each class
        - patch size of 16x20
        - 60% overlap
        - 10% validation set
        - 10% test set.

    Args:
        tmp_path: path to temporary directory
        request: pytest SubRequest object
    """
    yield _build_classification_patch_dataset(tmp_path, request.param)
    if tmp_path.exists():
        shutil.rmtree(tmp_path)

classification_dataset(tmp_path, dataset_arguments)

Generate classification dataset. If val_size or test_size are set, it will generate a train.txt, val.txt and test.txt file in the dataset directory. By default generated images are 10x10 pixels.

Parameters:

• tmp_path (Path) –

  path to temporary directory

• dataset_arguments (ClassificationDatasetArguments) –

  dataset arguments

Yields:

• Tuple[str, ClassificationDatasetArguments] –

  Tuple containing path to created dataset and dataset arguments

Source code in quadra/utils/tests/fixtures/dataset/classification.py

@pytest.fixture
def classification_dataset(
    tmp_path: Path, dataset_arguments: ClassificationDatasetArguments
) -> Tuple[str, ClassificationDatasetArguments]:
    """Generate classification dataset. If val_size or test_size are set, it will generate a train.txt, val.txt and
        test.txt file in the dataset directory. By default generated images are 10x10 pixels.

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

    Yields:
        Tuple containing path to created dataset and dataset arguments
    """
    yield _build_classification_dataset(tmp_path, dataset_arguments)
    if tmp_path.exists():
        shutil.rmtree(tmp_path)

classification_patch_dataset(tmp_path, dataset_arguments)

Fixture to dinamically generate a classification patch 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. The patch dataset will be generated using the standard
parameters of generate_patch_dataset function.

Parameters:

• tmp_path (Path) –

  path to temporary directory

• dataset_arguments (ClassificationDatasetArguments) –

  dataset arguments

Yields:

• Tuple[str, ClassificationDatasetArguments, Dict[str, int]] –

  Tuple containing path to created dataset, dataset arguments and class to index mapping

Source code in quadra/utils/tests/fixtures/dataset/classification.py

@pytest.fixture
def classification_patch_dataset(
    tmp_path: Path, dataset_arguments: ClassificationDatasetArguments
) -> Tuple[str, ClassificationDatasetArguments, Dict[str, int]]:
    """Fixture to dinamically generate a classification patch 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. The patch dataset will be generated using the standard
        parameters of generate_patch_dataset function.

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

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

multilabel_classification_dataset(tmp_path, dataset_arguments)

Fixture to dinamically generate a multilabel classification dataset. Generates a samples.txt file in the dataset directory containing the path to the image and the corresponding classes. If val_size or test_size are set, it will generate a train.txt, val.txt and test.txt file in the dataset directory. By default generated images are 10x10 pixels.

Parameters:

• tmp_path (Path) –

  path to temporary directory

• dataset_arguments (ClassificationMultilabelDatasetArguments) –

  dataset arguments

Returns:

• Tuple[str, ClassificationMultilabelDatasetArguments] –

  Tuple containing path to created dataset and dataset arguments

Source code in quadra/utils/tests/fixtures/dataset/classification.py

@pytest.fixture
def multilabel_classification_dataset(
    tmp_path: Path, dataset_arguments: ClassificationMultilabelDatasetArguments
) -> Tuple[str, ClassificationMultilabelDatasetArguments]:
    """Fixture to dinamically generate a multilabel classification dataset.
        Generates a samples.txt file in the dataset directory containing the path to the image and the corresponding
        classes. If val_size or test_size are set, it will generate a train.txt, val.txt and test.txt file in the
        dataset directory. By default generated images are 10x10 pixels.

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

    Returns:
        Tuple containing path to created dataset and dataset arguments
    """
    yield _build_multilabel_classification_dataset(tmp_path, dataset_arguments)
    if tmp_path.exists():
        shutil.rmtree(tmp_path)