clarena.cl_datasets.permuted_pcam

  1r"""
  2The submodule in `cl_datasets` for Permuted PCAM dataset.
  3"""
  4
  5__all__ = ["PermutedPCAM"]
  6
  7import logging
  8from typing import Callable
  9
 10from torch.utils.data import Dataset
 11from torchvision.datasets import PCAM
 12from torchvision.transforms import transforms
 13
 14from clarena.cl_datasets import CLPermutedDataset
 15
 16# always get logger for built-in logging in each module
 17pylogger = logging.getLogger(__name__)
 18
 19
 20class PermutedPCAM(CLPermutedDataset):
 21    r"""Permuted PCAM dataset. The [PCAM dataset](https://github.com/basveeling/pcam) is a collection of medical images of breast cancer. It consists of 327,680 images in 2 classes (benign and malignant), each 96x96 color image."""
 22
 23    original_dataset_python_class: type[Dataset] = PCAM
 24    r"""The original dataset class."""
 25
 26    def __init__(
 27        self,
 28        root: str,
 29        num_tasks: int,
 30        batch_size: int | dict[int, int] = 1,
 31        num_workers: int | dict[int, int] = 0,
 32        custom_transforms: (
 33            Callable
 34            | transforms.Compose
 35            | None
 36            | dict[int, Callable | transforms.Compose | None]
 37        ) = None,
 38        repeat_channels: int | None | dict[int, int | None] = None,
 39        to_tensor: bool | dict[int, bool] = True,
 40        resize: tuple[int, int] | None | dict[int, tuple[int, int] | None] = None,
 41        permutation_mode: str = "first_channel_only",
 42        permutation_seeds: dict[int, int] | None = None,
 43    ) -> None:
 44        r"""
 45        **Args:**
 46        - **root** (`str`): the root directory where the original PCAM data 'PCAM/' live.
 47        - **num_tasks** (`int`): the maximum number of tasks supported by the CL dataset. This decides the valid task IDs from 1 to `num_tasks`.
 48        - **batch_size** (`int` | `dict[int, int]`): the batch size for train, val, and test dataloaders.
 49        If it is a dict, the keys are task IDs and the values are the batch sizes for each task. If it is an `int`, it is the same batch size for all tasks.
 50        - **num_workers** (`int` | `dict[int, int]`): the number of workers for dataloaders.
 51        If it is a dict, the keys are task IDs and the values are the number of workers for each task. If it is an `int`, it is the same number of workers for all tasks.
 52        - **custom_transforms** (`transform` or `transforms.Compose` or `None` or dict of them): the custom transforms to apply ONLY to the TRAIN dataset. Can be a single transform, composed transforms, or no transform. `ToTensor()`, normalization, permute, and so on are not included.
 53        If it is a dict, the keys are task IDs and the values are the custom transforms for each task. If it is a single transform or composed transforms, it is applied to all tasks. If it is `None`, no custom transforms are applied.
 54        - **repeat_channels** (`int` | `None` | dict of them): the number of channels to repeat for each task. Default is `None`, which means no repeat.
 55        If it is a dict, the keys are task IDs and the values are the number of channels to repeat for each task. If it is an `int`, it is the same number of channels to repeat for all tasks. If it is `None`, no repeat is applied.
 56        - **to_tensor** (`bool` | `dict[int, bool]`): whether to include the `ToTensor()` transform. Default is `True`.
 57        If it is a dict, the keys are task IDs and the values are whether to include the `ToTensor()` transform for each task. If it is a single boolean value, it is applied to all tasks.
 58        - **resize** (`tuple[int, int]` | `None` or dict of them): the size to resize the images to. Default is `None`, which means no resize.
 59        If it is a dict, the keys are task IDs and the values are the sizes to resize for each task. If it is a single tuple of two integers, it is applied to all tasks. If it is `None`, no resize is applied.
 60        - **permutation_mode** (`str`): the mode of permutation; one of:
 61            1. 'all': permute all pixels.
 62            2. 'by_channel': permute channel by channel separately. All channels are applied the same permutation order.
 63            3. 'first_channel_only': permute only the first channel.
 64        - **permutation_seeds** (`dict[int, int]` | `None`): the dict of seeds for permutation operations used to construct each task. Keys are task IDs and the values are permutation seeds for each task. Default is `None`, which creates a dict of seeds from 0 to `num_tasks`-1.
 65        """
 66
 67        super().__init__(
 68            root=root,
 69            num_tasks=num_tasks,
 70            batch_size=batch_size,
 71            num_workers=num_workers,
 72            custom_transforms=custom_transforms,
 73            repeat_channels=repeat_channels,
 74            to_tensor=to_tensor,
 75            resize=resize,
 76            permutation_mode=permutation_mode,
 77            permutation_seeds=permutation_seeds,
 78        )
 79
 80    def prepare_data(self) -> None:
 81        r"""Download the original PCAM dataset if haven't."""
 82
 83        if self.task_id != 1:
 84            return  # download all original datasets only at the beginning of first task
 85
 86        PCAM(root=self.root_t, split="train", download=True)
 87        PCAM(root=self.root_t, split="val", download=True)
 88        PCAM(root=self.root_t, split="test", download=True)
 89
 90        pylogger.debug(
 91            "The original PCAM dataset has been downloaded to %s.", self.root_t
 92        )
 93
 94    def train_and_val_dataset(self) -> tuple[Dataset, Dataset]:
 95        """Get the training and validation dataset of task `self.task_id`.
 96
 97        **Returns:**
 98        - **train_and_val_dataset** (`tuple[Dataset, Dataset]`): the train and validation dataset of task `self.task_id`.
 99        """
100        dataset_train = PCAM(
101            root=self.root_t,
102            split="train",
103            transform=self.train_and_val_transforms(),
104            target_transform=self.target_transform(),
105            download=False,
106        )
107
108        dataset_val = PCAM(
109            root=self.root_t,
110            split="val",
111            transform=self.train_and_val_transforms(),
112            download=False,
113        )
114
115        return dataset_train, dataset_val
116
117    def test_dataset(self) -> Dataset:
118        r"""Get the test dataset of task `self.task_id`.
119
120        **Returns:**
121        - **test_dataset** (`Dataset`): the test dataset of task `self.task_id`.
122        """
123        dataset_test = PCAM(
124            root=self.root_t,
125            split="test",
126            transform=self.test_transforms(),
127            target_transform=self.target_transform(),
128            download=False,
129        )
130
131        return dataset_test