sanghyun-son / EDSR-PyTorch

PyTorch version of the paper 'Enhanced Deep Residual Networks for Single Image Super-Resolution' (CVPRW 2017)
MIT License
2.42k stars 668 forks source link

Weird RCAN test results #289

Closed JJangD closed 3 years ago

JJangD commented 3 years ago

Hello, thanks for sharing the codes.

I compared EDSRx2 and RCANx2 results.

However, the color RCANx2 result is different from the LR and EDSRx2 result.

Below are LR (original image) , EDSR and RCAN result.

1

I used below command lines.

for RCAN python main.py --data_test Demo --scale 2 --pre_train RCAN_BIX2.pt --test_only --save_results

for EDSR python main.py --data_test Demo --scale 2 --n_resblocks 32 --n_feats 256 --res_scale 0.1 --pre_train EDSR_x2.pt --test_only --save_results

wkiulu commented 3 years ago

I had the same problem, have you solved it?

wkiulu commented 3 years ago

I solved it by specifying the RCAN model in option...lol

JJangD commented 3 years ago

What do you mean 'specifying the RCAN model in option'?

wkiulu commented 3 years ago

“--model RCAN --n_resgroups ... --n_resblocks ... --n_feats ...“Otherwise, EDSR will be used by default

JJangD commented 3 years ago

Thanks!