Cifar10Dataset#

class Cifar10Dataset(resize=False, data_path='cifar_data')#

Bases: DatasetWrapper

A wrapper to the standard Cifar10 dataset, available at torchvision.datasets.CIFAR10. The current wrapper class supplyes the required transformations and augmentations, and also implements the required DatasetWrapper methods

__init__(resize=False, data_path='cifar_data')#

Methods

`__init__`([resize, data_path])
`get_approximation_set`()	Returns data set to be used for range approximation
`get_class_labels_dict`()	Returns the class_name to index mapping
`get_samples_per_class`(ds)	Returns the number of samples in each class.
`get_test_data`()	Returns the test data
`get_train_data`()	Returns the training data
`get_train_pipe_ffcv`(args)	Returns the training data as ffcv pipeline
`get_val_data`()	Returns the validation data
`is_imbalanced`()	Always returns False - Cifar10 dataset is balanced

get_approximation_set()#: Returns data set to be used for range approximation

get_samples_per_class(ds)#

Returns the number of samples in each class. The Cifar10 dataset has the same number of images in each class.

Params:

dataset (VisionDataset): The dataset

Returns:

the number of samples in each class.

Return type:

list<int>

get_test_data()#: Returns the test data

get_train_data()#: Returns the training data

get_train_pipe_ffcv(args)#

Returns the training data as ffcv pipeline

Params:

args (Arguments): user arguments

Returns:

a dictionary of shape {‘image’: <image_pipeline>, ‘label’: <label_pipeline>} representing the corresponding ffcv pipelines, as explained: https://docs.ffcv.io/making_dataloaders.html#pipelines

Return type:

Dictionary

get_val_data()#: Returns the validation data

is_imbalanced()#: Always returns False - Cifar10 dataset is balanced