Confusion about training results

[EliotChenKJ]

Hi~ thank you for converting styleGAN to its pytorch version, which is pretty meaningful. I've noticed that there are some people who can not get resonable training results. Unfortunately, i am one of them. I've only got the results below: The training set i am using have 30k images, so i don't think that is caused by the number of training set images. Hope that you can help me with it~

[tomguluson92]

Hi, you can contact me via QQ: 314913739. I have to say that I have the wish to improve this project but I do not have enough time.

[swj0418]

Hi! good work you've got here. I ran into the same problem where it wouldn't train at all. Do you have any suspicion as to why it is not learning to generate meaningful images?

[XiaotianM]

Hi, I can not generate meaningful images either. I have trained for 3 days

[ulucsahin]

I tried this model with CelebA dataset and have the same problem. I guess there is a bug somewhere.

[cnnlstm]

I meet the same error, I believe this implementation contains many bugs and it will leads us to a wrong way.

[nanfangzhe]

I also encountered the same problem. Has everyone's problem been solved?

[developer88]

Try this one https://github.com/lernapparat/lernapparat/blob/master/style_gan/pytorch_style_gan.ipynb

I found this in one of the file in this repo. I am testing this atm.

[meghanakanapaneni]

Hi~ thank you for converting styleGAN to its pytorch version, which is pretty meaningful. I've noticed that there are some people who can not get resonable training results. Unfortunately, i am one of them. I've only got the results below: The training set i am using have 30k images, so i don't think that is caused by the number of training set images. Hope that you can help me with it~

Hi. I am also facing the same issue. What is the final result you got? Thanks

[EliotChenKJ]

hi，

• i find out that pytorch version of stylegan2 has been implemented [ https://github.com/rosinality/stylegan2-pytorch]. i personally test it and it work well. You can have a look ~
• stylegan2-ada was also updated as latest styleGAN framework [ https://github.com/NVlabs/stylegan2-ada].

Meghana Kanapaneni [email protected] 于2020年10月7日周三 下午7:52写道：

Hi~ thank you for converting styleGAN to its pytorch version, which is pretty meaningful. I've noticed that there are some people who can not get resonable training results. Unfortunately, i am one of them. I've only got the results below: [image: 0] https://user-images.githubusercontent.com/48468233/70628510-43059200-1c63-11ea-85c5-329ae261b1de.png The training set i am using have 30k images, so i don't think that is caused by the number of training set images. Hope that you can help me with it~

Hi. I am also facing the same issue. What is the final result you got? Thanks

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/tomguluson92/StyleGAN_PyTorch/issues/11#issuecomment-704882853, or unsubscribe https://github.com/notifications/unsubscribe-auth/ALRZCCPED2RJSWFVI5A4QQ3SJRI6NANCNFSM4JZPW3ZA .

[99-WSJ]

hello i met this problem in last day , terrible images. so could you solve it and get better work?

[BruceYounggit]

In 2023, looks like that this bug is still existing..., seeming that it is time to find another project aforementioned by EliotChenKJ.