CIFAR10HDCIC¶

class probly.datasets.torch.CIFAR10HDCIC(root: Path | str, transform: Callable[..., Any] | None = None, *, first_order: bool = True)[source]¶

Bases: DCICDataset

Implementation of the DCIC version of the CIFAR10H dataset.

Targets and labels are the same as in the original CIFAR10H dataset. This variant uses the DCIC file structure and predefined folds. The dataset can be found at https://zenodo.org/records/7180818.

Initialize an instance of the DCIC-version of the CIFAR10H dataset class.

Parameters:

root – Root directory of the dataset.
transform – Optional transform to apply to the data.
first_order – Whether to use first order data or class labels. Defaults to True.

image_labels¶: Dictionary of image labels grouped by image.

image_paths¶: List of image paths.

label_mappings¶: Mapping of labels to indices.

num_classes¶: Number of classes.

root¶: Root directory of the dataset.

targets¶: List of labels.

transform¶: Transform to apply to the data.