VisionDataset

class VisionDataset(images: Sequence[str | Path | Tensor], transforms: Sequence | None = None, root: Path | None = None)[source]

Bases: Dataset

A dataset of images with data augmentation.

Note

requires torchvision to be installed.

Initialize the dataset.

Parameters:

  • images (Sequence[str | Path | Tensor]) – The images, either as (relative) path, or preprocessed tensors.

  • transforms (Sequence | None) – A sequence of transformations to apply to the images, cf. torchvision.transforms. Defaults to random size crops.

  • root (Path | None) – The root directory for images.