Style Transfer with CycleGAN on land art from Magic: The Gathering. Model implemented with PyTorch 1.6's automated mixed precision training. Magic land art is from https://basiclandart.com/.
All credits go to the authors of the original CycleGAN paper, Jun-Yan Zhu, Taesung Park, Phillip Isola, and Alexei A. Efros. Link to the paper can be found here. Link to the author's implementation can be found here.
Some examples trained on art by Claude Monet. Results below were trained on a Nvidia RTX 2060 Super for 200 epochs. The two domain datasets consisted of 360 images of size 256 x 256. Total training time was around 16 hours. Columns from left to right: original, generated style transfer result, generated cycle consistent result.
