search

xinntao / ESRGAN

ECCV18 Workshops - Enhanced SRGAN. Champion PIRM Challenge on Perceptual Super-Resolution. The training codes are in BasicSR.
https://github.com/xinntao/BasicSR
Apache License 2.0
5.91k stars 1.05k forks source link

the PSNR-oriented model with high PSNR performance #62

Open jiequanz opened 5 years ago

jiequanz commented 5 years ago

Which model does the RRDB_PSNR_x4.pth: the PSNR-oriented model with high PSNR performance. correspond to in the paper?

xinntao commented 5 years ago

RRDB_PSNR_x4.pth

https://drive.google.com/open?id=1pJ_T-V1dpb1ewoEra1TGSWl5e6H7M4NN

jiequanz commented 5 years ago

Hi, Thanks for your reply! I was able to find this model. What I asked was that: is this model introduced in the paper? I was wondering if this model is PNSR oriented one used in network interpretation. Just want to make sure. Also, is the loss function the only difference between RRDB_PSNR_x4.pth and RRDB_ESRGAN_x4.pth? Appreciate your help and contribution!

xinntao commented 5 years ago

1) as the pre-trained model for ESRGAN. 2) for network interpolation.

Yes, they are trained by different losses.

suesucong commented 4 years ago

RRDB_PSNR_x4.pth https://drive.google.com/open?id=1pJ_T-V1dpb1ewoEra1TGSWl5e6H7M4NN

Hi, why not just put the pretrained models into the dir?