mipcandy.data.inspection

Module Contents

Classes

InspectionAnnotation
InspectionAnnotations
ROIDataset
RandomROIDataset

Functions

format_bbox
_lists_to_tuples
_str_indices_to_int_indices
parse_inspection_annotation
load_inspection_annotations
bbox_from_indices
inspect
crop_and_pad

API

mipcandy.data.inspection.format_bbox(bbox: Sequence[int]) → tuple[int, int, int, int] | tuple[int, int, int, int, int, int]

class mipcandy.data.inspection.InspectionAnnotation

Bases: object

shape: mipcandy.types.AmbiguousShape = None

foreground_bbox: tuple[int, int, int, int] | tuple[int, int, int, int, int, int] = None

class_ids: tuple[int, ...] = None

class_counts: dict[int, int] = None

class_bboxes: dict[int, tuple[int, int, int, int] | tuple[int, int, int, int, int, int]] = None

class_locations: dict[int, tuple[tuple[int, int] | tuple[int, int, int], ...]] = None

spacing: mipcandy.types.Shape | None = None

foreground_shape() → mipcandy.types.Shape

center_of_foreground() → tuple[int, int] | tuple[int, int, int]

class mipcandy.data.inspection.InspectionAnnotations(dataset: mipcandy.data.dataset.SupervisedDataset, background: int, intensity_stats: tuple[float, float, float, float], *annotations: mipcandy.data.inspection.InspectionAnnotation)

Bases: typing.Sequence[mipcandy.data.inspection.InspectionAnnotation]

dataset() → mipcandy.data.dataset.SupervisedDataset

background() → int

intensity_stats() → tuple[float, float, float, float]

Returns:

mean, std, 0.5th percentile, 99.5th percentile

annotations() → tuple[mipcandy.data.inspection.InspectionAnnotation, ...]

__getitem__(item: int) → mipcandy.data.inspection.InspectionAnnotation

__len__() → int

save(path: str | os.PathLike[str]) → None

_get_shapes(get_shape: Callable[[mipcandy.data.inspection.InspectionAnnotation], mipcandy.types.AmbiguousShape]) → tuple[tuple[int, ...] | None, tuple[int, ...], tuple[int, ...]]

shapes() → tuple[tuple[int, ...] | None, tuple[int, ...], tuple[int, ...]]

statistical_shape(*, percentile: float = 0.95) → mipcandy.types.Shape

foreground_shapes() → tuple[tuple[int, ...] | None, tuple[int, ...], tuple[int, ...]]

statistical_foreground_shape(*, percentile: float = 0.95) → mipcandy.types.Shape

crop_foreground(i: int, *, expand_ratio: float = 1) → tuple[torch.Tensor, torch.Tensor]

foreground_heatmap() → torch.Tensor

center_of_foregrounds() → tuple[int, int] | tuple[int, int, int]

center_of_foregrounds_offsets() → tuple[int, int] | tuple[int, int, int]

set_roi_shape(roi_shape: mipcandy.types.Shape | None) → None

roi_shape(*, clamp: bool = True, percentile: float = 0.95) → mipcandy.types.Shape

roi(i: int, *, clamp: bool = True, percentile: float = 0.95) → tuple[int, int, int, int] | tuple[int, int, int, int, int, int]

crop_roi(i: int, *, clamp: bool = True, percentile: float = 0.95) → tuple[torch.Tensor, torch.Tensor]

mipcandy.data.inspection._lists_to_tuples(pairs: Sequence[tuple[str, Any]]) → dict[str, Any]

mipcandy.data.inspection._str_indices_to_int_indices(obj: dict[str, Any]) → dict[int, Any]

mipcandy.data.inspection.parse_inspection_annotation(obj: dict[str, Any]) → mipcandy.data.inspection.InspectionAnnotation

mipcandy.data.inspection.load_inspection_annotations(path: str | os.PathLike[str], dataset: mipcandy.data.dataset.SupervisedDataset) → mipcandy.data.inspection.InspectionAnnotations

mipcandy.data.inspection.bbox_from_indices(indices: torch.Tensor, num_dim: Literal[2, 3]) → tuple[int, int, int, int]

mipcandy.data.inspection.inspect(dataset: mipcandy.data.dataset.SupervisedDataset, *, background: int = 0, max_samples: int = 10000, console: rich.console.Console = Console()) → mipcandy.data.inspection.InspectionAnnotations

class mipcandy.data.inspection.ROIDataset(annotations: mipcandy.data.inspection.InspectionAnnotations, *, clamp: bool = True, percentile: float = 0.95)

Bases: mipcandy.data.dataset.SupervisedDataset[list[int]]

construct_new(images: list[Any], labels: list[Any]) → Self

abstractmethod load_image(idx: int) → torch.Tensor

abstractmethod load_label(idx: int) → torch.Tensor

load(idx: int) → tuple[torch.Tensor, torch.Tensor]

mipcandy.data.inspection.crop_and_pad(x: torch.Tensor, bbox_lbs: list[int], bbox_ubs: list[int], *, pad_value: int | float = 0) → torch.Tensor

class mipcandy.data.inspection.RandomROIDataset(annotations: mipcandy.data.inspection.InspectionAnnotations, batch_size: int, *, num_patches_per_case: int = 1, oversample_rate: float = 0.33, clamp: bool = False, percentile: float = 0.5, min_factor: int = 16)

Bases: mipcandy.data.inspection.ROIDataset

convert_idx(idx: int) → int

roi_shape(*, roi_shape: mipcandy.types.Shape | None = None) → None | mipcandy.types.Shape

construct_new(images: list[Any], labels: list[Any]) → Self

random_roi(idx: int, force_foreground: bool) → tuple[list[int], list[int]]

oversample_foreground(idx: int) → bool

load(idx: int) → tuple[torch.Tensor, torch.Tensor]