If you are interested in the Manga Colorization, I suggest you to do a search yourself to get familiar to different approaches. It is one topic that many models have been tried because of its commercial potential. There are new ideas that I would like to try but not yet have time to do it. In particular, there are quite many different cost functions proposed. But I need to clean up the dataset first as mentioned in the article. Unfortunately, I try not to publish my Github. There are quite a lot of projects (mine and others) and it will take too much time to support it.