Closed hosytuyen closed 4 years ago
In fact the model in this repo is not trained with the same model described in CartoonGAN paper. I am sorry that the training code could not be published now, but after I subscrib my paper I may open source them.
Many thanks for your reply, I look forward to your paper.
Many thanks for your reply, I look forward to your paper.
for dataset, due to copyright issues i may not be able to share, but i can give you some advice: For human face: ffhq dataset or celeba hq(resize to 256256) For animate face: kyoto animate and pa works, use animate face detection and manually clean them For real world photo: CycleGAN training dataset For animate image: Shinkai Makoto, Hosoda Mamoru, Miyazaki Hayao, get key frames and random crop to 256256 (if you manually clean them you may get better result)
@HoSyTuyen Paper is accepted to CVPR2020, code and other material will be open-sourced later
Wao, Congratulates!!!, It's very impressive, thanks for your sharing.
Hi, Thank you for amazing CartoonGAN implementation, I'm reproducing CartoonGAN, but don't get the good result. Can your share how many real and cartoon images in your dataset. And during the training, the weight of vgg loss, learning rate and batch_size that you using. Is it like in the your old_code folder? Thanks you so much.