AnimeGANv2

Why we included this project

For anyone working with GANs for image style transfer, this is a compact, well-documented reference for photo cartoonization. It trains a generative network to redraw real-world landscape shots in the look of three well-known directors, Hayao Miyazaki, Makoto Shinkai, and Kon Satoshi, and includes inference scripts plus a separate video converter that restyles whole clips frame by frame. The generator model is small, around 8 MB, which makes it a light base for custom art direction experiments or for studying how style loss and GAN objectives behave in practice. The full training pipeline and style datasets drawn from Blu-ray films are included, so reproducing the paper's results is straightforward.