search

Zheng222 / IDN-Caffe

Caffe implementation of "Fast and Accurate Single Image Super-Resolution via Information Distillation Network" (CVPR 2018)
117 stars 25 forks source link

Result seems strange in paper #6

Closed GengDavid closed 6 years ago

GengDavid commented 6 years ago

Hi @Zheng222 Thanks for your great work.
But I'm a little confused about the “barbara” result in your paper.
As we can see that the original image that there are several books "stacked" in the red box ori

But I found the result you've shown looks like some “diagonal lines”. res

It is a little strange since I've tried to use even "Bicubic" method(with an upscaling factor 4), no “diagonal lines” occurs in my experiments. Wonder that why the results shown in your paper look like that? Am I missing something? Thx!

Zheng222 commented 6 years ago

Hello, you can see the result from RCAN (ECCV'18) https://github.com/yulunzhang/RCAN/blob/master/Figs/supp_fig1_visual_compare_gan_psnr_ssim_bi_x4_1.PNG. From this picture, we can also see that there are diagonal artifacts in the images generated by many methods. @GengDavid

GengDavid commented 6 years ago

@Zheng222 Thanks for your reply. Well, maybe I found what my problem is. I used a high resolution image as input, so the upsampling result would be much better.

GengDavid commented 6 years ago

I got the similar result. Thanks! I'll close this issue.