Looking at that project on github[1] I can't help but to think the algorithm need to upsample anime images (like in this project) might be different than the on

Looking at that project on github[1] I can't help but to think the algorithm need to upsample anime images (like in this project) might be different than the one needed to upsample photos. Anime has the benefit of having a lot of sections of solid colors. Edit: Actually looking at the source further it appears they also have models for photos as part of that project in addition to the model for anime. [1] https://github.com/nagadomi/waifu2x https://github.com/nagadomi/waifu2x