AdvSegLearn 1.0.5

Package for combining traditional and adversarial learning.

AdvSegLearn

Descriptor

This package creates an easy to use api for training models on small datasets with adversarial learning. Important classes are:

• Mutliclass_dataset - allows for both paired and unpaired images to be loaded, with built in support for initalized data augmentation and lazy loading data augmentation.
• train_GAN - provides api to easily load in different models, datasets and loss functions to the model. Subclassing allows for easy modification of the training loop.
• semi_supervised_loss - a loss function for the semisupervised training loop.
• adversarial_loss - a loss function for hte unsupervised training loop.

License

This package is licensed under the MIT LICENSE.