quanda.utils.datasets.image_datasets module¶

Dataset classes for image datasets.

class quanda.utils.datasets.image_datasets.HFtoTV(dataset, transform=None, label_override: int | None = None)[source]¶

Bases: Dataset

Wrapper to make Hugging Face datasets compatible with torchvision.

__init__(dataset, transform=None, label_override: int | None = None)[source]¶: Construct the HFtoTV dataset.

get_label(idx)[source]¶: Return the label at idx without decoding the image.