search

alterzero / DBPN-Pytorch

The project is an official implement of our CVPR2018 paper "Deep Back-Projection Networks for Super-Resolution" (Winner of NTIRE2018 and PIRM2018)
https://alterzero.github.io/projects/DBPN.html
MIT License
566 stars 164 forks source link

Why can't I reproduct the effect on the paper #10

Closed akafen closed 6 years ago

akafen commented 6 years ago

Hello,I use your code to run,but ths psnr always is 35.x tested on Set5 which the scale factor is 2 ，but the paper is 38.09.I wonder why?

alexnie1997 commented 6 years ago

Can you run all the 5 images of Set5? I can only run three of them.

alterzero commented 6 years ago

Hi @akafen , How did you calculate the PSNR? In the paper, the measurement used luminance layer only.

@alexnie1997 We use TITAN X. If you have less memory, try to run the evaluation using --chop_forward True (This function is implemented by EDSR team https://github.com/thstkdgus35/EDSR-PyTorch)

akafen commented 6 years ago

Hi @alterzero,I calculated the PSNR on RGB images,but I think it should not have such a big difference and I would try to calculate the PSNR using luminance layer only.

wehkjhkl commented 6 years ago

Can you run eval.py?how can you do it?

LoSealL commented 5 years ago

@akafen The difference could be up to 2dB.