About reproducing the results in the paper

[chensming]

Hi, I try to reproduce the results, and I follow your guide, run the evaluation.sh, here are the result I got.

I have two questions: (1) The results reported in the paper or colorization task is evaluated via Cons metrics, but the code is via PSNR. And interestingly, I found they are close on CelebA datasets.

(2) How can I reproduce the results on inpainting task? My results are much poor than yours.

The results I collected from ouput are as followings:

Thanks a lot if you could help me.

[chensming]

oh, I have seen this issue #39 just now. So in this work the results of inpainting task are very sensitive to the mask used in experiments?

[wyhuai]

Hi, thanks for finding a typo! The 26.25 is PSNR, and the correct Cons should be 48.87.

As for the Inpainting problem, the results can be quite different depending on your masks. For example, a large mask will be more difficult to restore than a small mask.

[chensming]

Hi, thanks for finding a typo! The 26.25 is PSNR, and the correct Cons should be 48.87.

As for the Inpainting problem, the results can be quite different depending on your masks. For example, a large mask will be more difficult to restore than a small mask.

I still donot understant why PSNR is not suitable for colorization task.

[wyhuai]

For the colorization problems, a good colorful result may not compete with a black-and-white result in terms of PSNR.

[chensming]

okay, thank you~