Datasets

Torchvision provides many built-in datasets in the torchvision.datasets module, as well as utility classes for building your own datasets.

Built-in datasets

All datasets are subclasses of torch.utils.data.Dataset i.e, they have __getitem__ and __len__ methods implemented. Hence, they can all be passed to a torch.utils.data.DataLoader which can load multiple samples in parallel using torch.multiprocessing workers. For example:

imagenet_data = torchvision.datasets.ImageNet('path/to/imagenet_root/')
data_loader = torch.utils.data.DataLoader(imagenet_data,
                                          batch_size=4,
                                          shuffle=True,
                                          num_workers=args.nThreads)

All the datasets have almost similar API. They all have two common arguments: transform and target_transform to transform the input and target respectively. You can also create your own datasets using the provided base classes.

Warning

When a dataset object is created with download=True, the files are first downloaded and extracted in the root directory. This download logic is not multi-process safe, so it may lead to conflicts / race conditions if it is run within a distributed setting. In distributed mode, we recommend creating a dummy dataset object to trigger the download logic before setting up distributed mode.

Image classification

Caltech101(root[, target_type, transform, ...])	Caltech 101 Dataset.
Caltech256(root[, transform, ...])	Caltech 256 Dataset.
CelebA(root[, split, target_type, ...])	Large-scale CelebFaces Attributes (CelebA) Dataset Dataset.
CIFAR10(root[, train, transform, ...])	CIFAR10 Dataset.
CIFAR100(root[, train, transform, ...])	CIFAR100 Dataset.
Country211(root[, split, transform, ...])	The Country211 Data Set from OpenAI.
DTD(root[, split, partition, transform, ...])	Describable Textures Dataset (DTD).
EMNIST(root, split, **kwargs)	EMNIST Dataset.
EuroSAT(root[, transform, target_transform, ...])	RGB version of the EuroSAT Dataset.
FakeData([size, image_size, num_classes, ...])	A fake dataset that returns randomly generated images and returns them as PIL images
FashionMNIST(root[, train, transform, ...])	Fashion-MNIST Dataset.
FER2013(root[, split, transform, ...])	FER2013 Dataset.
FGVCAircraft(root[, split, ...])	FGVC Aircraft Dataset.
Flickr8k(root, ann_file[, transform, ...])	Flickr8k Entities Dataset.
Flickr30k(root, ann_file[, transform, ...])	Flickr30k Entities Dataset.
Flowers102(root[, split, transform, ...])	Oxford 102 Flower Dataset.
Food101(root[, split, transform, ...])	The Food-101 Data Set.
GTSRB(root[, split, transform, ...])	German Traffic Sign Recognition Benchmark (GTSRB) Dataset.
INaturalist(root[, version, target_type, ...])	iNaturalist Dataset.
ImageNet(root[, split])	ImageNet 2012 Classification Dataset.
Imagenette(root[, split, size, download, ...])	Imagenette image classification dataset.
KMNIST(root[, train, transform, ...])	Kuzushiji-MNIST Dataset.
LFWPeople(root[, split, image_set, ...])	LFW Dataset.
LSUN(root[, classes, transform, ...])	LSUN dataset.
MNIST(root[, train, transform, ...])	MNIST Dataset.
Omniglot(root[, background, transform, ...])	Omniglot Dataset.
OxfordIIITPet(root[, split, target_types, ...])	Oxford-IIIT Pet Dataset.
Places365(root, ~pathlib.Path], split, ...)	Places365 classification dataset.
PCAM(root[, split, transform, ...])	PCAM Dataset.
QMNIST(root[, what, compat, train])	QMNIST Dataset.
RenderedSST2(root[, split, transform, ...])	The Rendered SST2 Dataset.
SEMEION(root[, transform, target_transform, ...])	SEMEION Dataset.
SBU(root[, transform, target_transform, ...])	SBU Captioned Photo Dataset.
StanfordCars(root[, split, transform, ...])	Stanford Cars Dataset
STL10(root[, split, folds, transform, ...])	STL10 Dataset.
SUN397(root[, transform, target_transform, ...])	The SUN397 Data Set.
SVHN(root[, split, transform, ...])	SVHN Dataset.
USPS(root[, train, transform, ...])	USPS Dataset.

Image detection or segmentation

CocoDetection(root, annFile[, transform, ...])	MS Coco Detection Dataset.
CelebA(root[, split, target_type, ...])	Large-scale CelebFaces Attributes (CelebA) Dataset Dataset.
Cityscapes(root[, split, mode, target_type, ...])	Cityscapes Dataset.
Kitti(root[, train, transform, ...])	KITTI Dataset.
OxfordIIITPet(root[, split, target_types, ...])	Oxford-IIIT Pet Dataset.
SBDataset(root[, image_set, mode, download, ...])	Semantic Boundaries Dataset
VOCSegmentation(root[, year, image_set, ...])	Pascal VOC Segmentation Dataset.
VOCDetection(root[, year, image_set, ...])	Pascal VOC Detection Dataset.
WIDERFace(root[, split, transform, ...])	WIDERFace Dataset.

Optical Flow

FlyingChairs(root[, split, transforms])	FlyingChairs Dataset for optical flow.
FlyingThings3D(root[, split, pass_name, ...])	FlyingThings3D dataset for optical flow.
HD1K(root[, split, transforms])	HD1K dataset for optical flow.
KittiFlow(root[, split, transforms])	KITTI dataset for optical flow (2015).
Sintel(root[, split, pass_name, transforms])	Sintel Dataset for optical flow.

Stereo Matching

CarlaStereo(root[, transforms])	Carla simulator data linked in the CREStereo github repo.
Kitti2012Stereo(root[, split, transforms])	KITTI dataset from the 2012 stereo evaluation benchmark.
Kitti2015Stereo(root[, split, transforms])	KITTI dataset from the 2015 stereo evaluation benchmark.
CREStereo(root[, transforms])	Synthetic dataset used in training the CREStereo architecture.
FallingThingsStereo(root[, variant, transforms])	FallingThings dataset.
SceneFlowStereo(root[, variant, pass_name, ...])	Dataset interface for Scene Flow datasets.
SintelStereo(root[, pass_name, transforms])	Sintel Stereo Dataset.
InStereo2k(root[, split, transforms])	InStereo2k dataset.
ETH3DStereo(root[, split, transforms])	ETH3D Low-Res Two-View dataset.
Middlebury2014Stereo(root[, split, ...])	Publicly available scenes from the Middlebury dataset 2014 version <https://vision.middlebury.edu/stereo/data/scenes2014/>.

Image pairs

LFWPairs(root[, split, image_set, ...])	LFW Dataset.
PhotoTour(root, name[, train, transform, ...])	Multi-view Stereo Correspondence Dataset.

Image captioning

CocoCaptions(root, annFile[, transform, ...])

MS Coco Captions Dataset.

Video classification

HMDB51(root, annotation_path, frames_per_clip)	HMDB51 dataset.
Kinetics(root, frames_per_clip[, ...])	Generic Kinetics dataset.
UCF101(root, annotation_path, frames_per_clip)	UCF101 dataset.

Video prediction

MovingMNIST(root[, split, split_ratio, ...])

MovingMNIST Dataset.

Base classes for custom datasets

DatasetFolder(root, loader[, extensions, ...])	A generic data loader.
ImageFolder(root, ~pathlib.Path], transform, ...)	A generic data loader where the images are arranged in this way by default: .
VisionDataset([root, transforms, transform, ...])	Base Class For making datasets which are compatible with torchvision.

Transforms v2

wrap_dataset_for_transforms_v2(dataset[, ...])

Wrap a torchvision.dataset for usage with torchvision.transforms.v2.