search

ARM-software / sesr

Super-Efficient Super Resolution
Apache License 2.0
89 stars 15 forks source link

The psnr value can‘t reach 32.84 on Set14 (SESR-M5, f=16) #13

Open HeYiCui opened 2 years ago

HeYiCui commented 2 years ago

Hello! I trained sesr (M5, f=16) on div2k, but when I tested on set14, the psnr value was 32.6, can't reach 32.84 (from paper table 1). I set the learning rate to 5e-4 (from paper) and 1e-4 (from github code), but none of them can achieve the effects mentioned in the paper table 1. Do I need change some hyperparameters？ Could you provide your trained "SESR-M5- f=16" model？ email：cuihezdh@gmail.com

eladc-git commented 2 years ago

Hi, I got the same issue. What are the hyperparameters that achieve psnr=32.84 for Set14 as in your paper?

prathapsoma commented 1 year ago

Can you suggest me please how to test the network after training.