search

mit-han-lab / data-efficient-gans

[NeurIPS 2020] Differentiable Augmentation for Data-Efficient GAN Training
https://arxiv.org/abs/2006.10738
BSD 2-Clause "Simplified" License
1.27k stars 175 forks source link

Training set image resolution #85

Open davidecarnevali opened 2 years ago

davidecarnevali commented 2 years ago

Hi, is there a way to train using a dataset of images with shape that is not a power of 2? Such as 400x400 or even 420x400? Thank you

Davide

zsyzzsoft commented 2 years ago

Yes, the tf version of DiffAugment-stylegan2 supports this (still not implemented if length is not equal to width). It will first generate images at a resolution of the nearest power of 2 and then crop to the images to the original size.

davidecarnevali commented 2 years ago

Thanks, However I'm using Pytorch... Will be available soon? Thx

Il ven 22 ott 2021, 13:30 Shengyu Zhao @.***> ha scritto:

Yes, the tf version of DiffAugment-stylegan2 supports this (still not implemented if length is not equal to width). It will first generate images at a resolution of the nearest power of 2 and then crop to the images to the original size.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/mit-han-lab/data-efficient-gans/issues/85#issuecomment-949543794, or unsubscribe https://github.com/notifications/unsubscribe-auth/AE56FGP33D4LMR376RDPBTLUIFDLZANCNFSM5GLGJKAA . Triage notifications on the go with GitHub Mobile for iOS https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675 or Android https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub.

zsyzzsoft commented 2 years ago

Unfortunately we will not update the code to support new features. You may have a try following the basic idea in our tf code.

davidecarnevali commented 2 years ago

Ok, Thank you

Il ven 22 ott 2021, 16:23 Shengyu Zhao @.***> ha scritto:

Unfortunately we will not update the code to support new features. You may have a try following the basic idea in our tf code.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/mit-han-lab/data-efficient-gans/issues/85#issuecomment-949678923, or unsubscribe https://github.com/notifications/unsubscribe-auth/AE56FGKJPVEHNK3JPQQPUA3UIFXU5ANCNFSM5GLGJKAA . Triage notifications on the go with GitHub Mobile for iOS https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675 or Android https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub.