CycleGAN is a deep learning architecture that allows for unpaired image translation across two different domains. This post explores some of the details of CycleGAN and my implementation of the model for style transfer on land art from Magic: The Gathering and for unpaired image translation of images of bald people to images of not bald people.