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
5.86k stars 1.04k forks source link

Why dont you use pixel shuffle to upsampling? #82

Open oucnkufk opened 4 years ago

oucnkufk commented 4 years ago

hello, when i was reading the RRDBNet_arch.py, I found that you use interpolate('mode='nearest'') to up sampling. Compared with some other models(like EDSR), which use pixel shuffle. Is there any advantage to using this interpolate？ Thank you

ssqfs commented 4 years ago

Same question to me.

MasayaGit commented 4 years ago

Same question to me.

WeiYangHui commented 1 year ago

If you use the pixel-shuffle, you may need to train it if you want to get a quite good result, while the nearest interpolation doesn't need.