similar to pix2pix?

[junyanz]

These are really cool results. Congratulations Taizan! I find that your repo looks quite similar to pix2pix in terms of unet architecture and conditional gan loss function. I understand you applied it to a different application, but it would be great if you can acknowledge it in the repo. Pix2pix is also available in chainer thanks to mattya.

Example results from pix2pix.

Also, have you compared with previous methods like manga colorization from Siggraph 06? One can just take their input image and user strokes for a comparison.

[taizan]

Hi junyanz, Thanks. I know about px2px, it was impressive. The defference between PaintsChainer and px2px

• intractive input and output interface
• using 2 step network 1 with gan and 1 without gan
• dosen't use conditional gan, but normal dcgan

I prefer to make application, so there are not enough survey and comparisons. If I published a paper or technical report in future, I will reference about px2px as related works.

[junyanz]

This is great to know. Thanks for the clarification.