semipy.datasets
Warning
This section is in construction.
This module allows user to get datasets for SSL. It is possible to download and transform famous datasets used when benchmarking in SSL but also to split a custom dataset into a labelled/unlabelled set.
Classes
API Reference | Description |
---|---|
semipy.datasets.utils.SSLDataset | A class that allows to transform a simple dataset into an SSL dataset with weakly and strongly augmented data. |
semipy.datasets.utils.SSLDatasetFolder | Same class as SSLDataset but dedicated to multi-GPU training with DDP. |
Functions
API Reference | Description |
---|---|
semipy.datasets.get_dataset | To call the right getter function for the right dataset. |
semipy.datasets.cifar.get_cifar | To download CIFAR with or without SSL constraints. |
semipy.datasets.svhn.get_svhn | To download SVHN with or without SSL constraints. |
semipy.datasets.stl10.get_stl10 | To download STL10 with or without SSL constraints. |
semipy.datasets.medmnist.get_medmnist | To download one of MedMNIST datasets with or without SSL constraints. |
semipy.datasets.utils.split_dataset | To split a SSLDataset object into train, test and validation sets. |
semipy.datasets.utils.build_transforms | To build a composition of transformations from a list of transforms configurations. |
semipy.datasets.utils.build_augmentations | To read weak and strong augmentations from yaml file. |