folder

class trojanvision.datasets.CUB200(data_format='folder', memory=False, **kwargs)

CUB200 dataset introduced by Peter Welinder in 2010. It inherits trojanvision.datasets.ImageFolder.

See also

• paper: Caltech-UCSD Birds 200

• website: http://www.vision.caltech.edu/visipedia/CUB-200.html

Variables:

• name (str) – 'cub200'

• num_classes (int) – 200

• data_shape (list[int]) – [3, 224, 224]

• valid_set (bool) – False

class trojanvision.datasets.CUB200_2011(data_format='folder', memory=False, **kwargs)

CUB200_2011 dataset. It inherits trojanvision.datasets.ImageFolder.

See also

• paper: The Caltech-UCSD Birds-200-2011 Dataset

• website: http://www.vision.caltech.edu/visipedia/CUB-200-2011.html

Variables:

• name (str) – 'cub200_2011'

• num_classes (int) – 200

• data_shape (list[int]) – [3, 224, 224]

• valid_set (bool) – False

class trojanvision.datasets.GTSRB(norm_par={'mean': [0.3403, 0.3121, 0.3214], 'std': [0.2724, 0.2608, 0.2669]}, loss_weights=True, **kwargs)

GTSRB dataset introduced by Johannes Stallkamp in 2011. It inherits trojanvision.datasets.ImageFolder.

See also

• paper: The German Traffic Sign Recognition Benchmark: A multi-class classification competition

• website: https://benchmark.ini.rub.de/gtsrb_dataset.html

Variables:

• name (str) – 'gtsrb'

• num_classes (int) – 43

• data_shape (list[int]) – [3, 32, 32]

• norm_par (dict[str, list[float]]) –

  {'mean': [0.3403, 0.3121, 0.3214],

  'std'  : [0.2724, 0.2608, 0.2669]}

• valid_set (bool) – False

• loss_weights (bool) – True

class trojanvision.datasets.ImageNet(norm_par={'mean': [0.485, 0.456, 0.406], 'std': [0.229, 0.224, 0.225]}, **kwargs)

ImageNet (ILSVRC2012) dataset introduced by Jia Deng and Feifei Li in 2012. It inherits trojanvision.datasets.ImageFolder.

See also

• torchvision: torchvision.datasets.ImageNet

• paper: ImageNet: A large-scale hierarchical image database

• website: https://image-net.org/about.php

Note

According to https://github.com/pytorch/vision/issues/1563, You need to personally visit https://image-net.org/download-images.php to download the dataset.

Expected files:

• '{self.folder_path}/ILSVRC2012_devkit_t12.tar.gz'

• '{self.folder_path}/ILSVRC2012_img_train.tar'

• '{self.folder_path}/ILSVRC2012_img_val.tar'

• '{self.folder_path}/meta.bin'

Variables:

• name (str) – 'imagenet'

• num_classes (int) – 1000

• data_shape (list[int]) – [3, 224, 224]

• norm_par (dict[str, list[float]]) –

  {'mean': [0.485, 0.456, 0.406],

  'std'  : [0.229, 0.224, 0.225]}

class trojanvision.datasets.ISIC2018(data_format='folder', memory=False, **kwargs)

ISIC2018 dataset introduced by Noel Codella in 2018. It inherits trojanvision.datasets.ImageFolder.

See also

• paper: Skin Lesion Analysis Toward Melanoma Detection 2018: A Challenge Hosted by the International Skin Imaging Collaboration (ISIC)

• website: https://challenge.isic-archive.com/data/

Variables:

• name (str) – 'isic2018'

• num_classes (int) – 7

• data_shape (list[int]) – [3, 224, 224]

class trojanvision.datasets.VGGface2(data_format='folder', memory=False, **kwargs)

VGGface2 dataset introduced by Qiong Cao in 2018. It inherits trojanvision.datasets.ImageFolder.

See also

• paper: VGGFace2: A dataset for recognising faces across pose and age

• website (expired dead link): https://www.robots.ox.ac.uk/~vgg/data/vgg_face2

Variables:

• name (str) – 'vggface2'

• num_classes (int) – 8631 (Why the papaer claims to have 500 more?)

• data_shape (list[int]) – [3, 224, 224]

• valid_set (bool) – False