Any chance for releasing the code for generating the original dataset explained in the paper?

[henderson-mathpresso]

So I'm planning to use this for further research and the thing is that I don't have access to the dataset you've used for training the model. Would you please release the code and the dataset for training the model?

[ljsabc]

Probably in several days. The original code is badly written in py2 and the link to the dataset of manga screentones was gone (fortunately I have a local copy which is copyright-free).
Let me see if I can tidy up the codes and make it run again. If everything goes right I will share it along with a new version of the repo, with the model fixed with native tf support.