A Generative Adversarial Network (GAN) is implemented in PyTorch to generate novel 64bit images of clothing. Refinements are made to improve the quality of the generated images, such as batch normalisation, leaky-ReLU activation and data augmentation via affine transformations and reflections.
