nip.image_classification.data

nip.image_classification.data#

Data classes for the image classification task.

Contains the ImageClassificationDataset class, which is a dataset for the image classification task. The dataset is a binary classification problem, where the classes are selected from a torchvision dataset. The dataset is binarified using one of the following methods:

  • “merge”: The classes are shuffled and merged into two classes.

  • “select_two”: Two classes are selected from the original dataset.

  • “random”: The classes are selected at random.

Functions

register_dataset_wrapper_class(dataset_name)

Register a dataset wrapper class.

Classes

Cifar100DatasetWrapper(root[, train, ...])

The CIFAR-100 dataset wrapper.

Cifar10DatasetWrapper(root[, train, ...])

The CIFAR-10 dataset wrapper.

FashionMnistDatasetWrapper(root[, train, ...])

The Fashion-MNIST dataset wrapper.

ImageClassificationDataset(hyper_params, ...)

A dataset for the image classification task.

KmnistDatasetWrapper(root[, train, ...])

The Kuzushiji-MNIST dataset wrapper.

MnistDatasetWrapper(root[, train, ...])

The MNIST dataset wrapper.

SvhnDatasetWrapper(root[, train, transform, ...])

The Street View House Numbers dataset wrapper.

TestDataset(root[, train, transform, ...])

A fake dataset for testing.

TorchVisionDatasetWrapper(root[, train, ...])

A wrapper for TorchVision datasets, implementing a common interface.