data

`DetectionDataset`

Bases: Dataset

PyTorch dataset for slide segmentation tasks.

Parameters:

Name	Type	Description	Default
`slide_paths`	`Sequence[PathLike]`	list of slides' filepaths.	required
`mask_paths`	`Sequence[PathLike]`	list of masks' filepaths. Masks are supposed to be tiled pyramidal images.	required
`patches_paths`	`Sequence[PathLike]`	list of patch csvs' filepaths. Files must be formatted according to `PathAIA API <https://github.com/MicroMedIAn/PathAIA>`_.	required
`stain_matrices_paths`	`Optional[Sequence[PathLike]]`	path to stain matrices .npy files. Each file must contain a (2, 3) matrice to use for stain separation. If not sppecified while `stain_augmentor` is, stain matrices will be computed at runtime (can cause a bottleneckd uring training).	`None`
`stain_augmentor`	`Optional[StainAugmentor]`	:class:`~apriorics.transforms.StainAugmentor` object to use for stain augmentation.	`None`
`transforms`	`Optional[Sequence[BasicTransform]]`	list of `albumentation <https://albumentations.ai/>`_ transforms to use on images (and on masks when relevant).	`None`
`slide_backend`	`str`	whether to use `OpenSlide <https://openslide.org/>` or `cuCIM <https://github.com/rapidsai/cucim>` to load slides.	`'cucim'`

Bases: Dataset

PyTorch dataset for slide segmentation tasks.

Parameters:

Name	Type	Description	Default
`slide_paths`	`Sequence[PathLike]`	list of slides' filepaths.	required
`mask_paths`	`Sequence[PathLike]`	list of masks' filepaths. Masks are supposed to be tiled pyramidal images.	required
`patches_paths`	`Sequence[PathLike]`	list of patch csvs' filepaths. Files must be formatted according to `PathAIA API <https://github.com/MicroMedIAn/PathAIA>`_.	required
`stain_matrices_paths`	`Optional[Sequence[PathLike]]`	path to stain matrices .npy files. Each file must contain a (2, 3) matrice to use for stain separation. If not sppecified while `stain_augmentor` is, stain matrices will be computed at runtime (can cause a bottleneckd uring training).	`None`
`stain_augmentor`	`Optional[StainAugmentor]`	:class:`~apriorics.transforms.StainAugmentor` object to use for stain augmentation.	`None`
`transforms`	`Optional[Sequence[BasicTransform]]`	list of `albumentation <https://albumentations.ai/>`_ transforms to use on images (and on masks when relevant).	`None`
`slide_backend`	`str`	whether to use `OpenSlide <https://openslide.org/>` or `cuCIM <https://github.com/rapidsai/cucim>` to load slides.	`'cucim'`

Bases: Dataset

PyTorch dataset for slide segmentation tasks.

Parameters:

Name	Type	Description	Default
`slide_paths`	`Sequence[PathLike]`	list of slides' filepaths.	required
`mask_paths`	`Sequence[PathLike]`	list of masks' filepaths. Masks are supposed to be tiled pyramidal images.	required
`patches_paths`	`Sequence[PathLike]`	list of patch csvs' filepaths. Files must be formatted according to `PathAIA API <https://github.com/MicroMedIAn/PathAIA>`_.	required
`stain_matrices_paths`	`Optional[Sequence[PathLike]]`	path to stain matrices .npy files. Each file must contain a (2, 3) matrice to use for stain separation. If not sppecified while `stain_augmentor` is, stain matrices will be computed at runtime (can cause a bottleneckd uring training).	`None`
`stain_augmentor`	`Optional[StainAugmentor]`	:class:`~apriorics.transforms.StainAugmentor` object to use for stain augmentation.	`None`
`transforms`	`Optional[Sequence[BasicTransform]]`	list of `albumentation <https://albumentations.ai/>`_ transforms to use on images (and on masks when relevant).	`None`
`slide_backend`	`str`	whether to use `OpenSlide <https://openslide.org/>` or `cuCIM <https://github.com/rapidsai/cucim>` to load slides.	`'cucim'`