search

XPixelGroup / BasicSR

Open Source Image and Video Restoration Toolbox for Super-resolution, Denoise, Deblurring, etc. Currently, it includes EDSR, RCAN, SRResNet, SRGAN, ESRGAN, EDVR, BasicVSR, SwinIR, ECBSR, etc. Also support StyleGAN2, DFDNet.
https://basicsr.readthedocs.io/en/latest/
Apache License 2.0
6.51k stars 1.15k forks source link

Multiples of 96 / 384 ESRGAN #507

Open davidvfx07 opened 2 years ago

davidvfx07 commented 2 years ago

So I am using BasicSR to upscale the output of Wav2Lip to a more usable size. By training my own model every use gives me accurate teeth and eyes and really helps with the believability. Wav2Lip outputs 96x96, x4 is 384x384. I cannot seem to find a way to train ESRGAN with VGGStyleDiscriminator for 384x384.

Any help appreciated!

einsqing commented 9 months ago

So I am using BasicSR to upscale the output of Wav2Lip to a more usable size. By training my own model every use gives me accurate teeth and eyes and really helps with the believability. Wav2Lip outputs 96x96, x4 is 384x384. I cannot seem to find a way to train ESRGAN with VGGStyleDiscriminator for 384x384.

Any help appreciated!

Hello, I trained real-esrgan in a 10 minute single person video, and the effect was that my mouth was kept closed and couldn't be opened. How can I solve this problem?