search

jiangsutx / SRN-Deblur

Repository for Scale-recurrent Network for Deep Image Deblurring
http://www.xtao.website/projects/srndeblur/srndeblur_cvpr18.pdf
MIT License
714 stars 184 forks source link

Questions about test pictures #32

Closed JusperLee closed 4 years ago

JusperLee commented 5 years ago

I used my own test picture and the pre-processing model provided by the author for testing, but the picture of the final result could not be opened for viewing, I wonder what is the reason

jiangsutx commented 5 years ago

Please at least provide some error messages for others to help you.

And the code is totally open-source with Python. You can check whether you image is correctly loaded, and whether the network produces correct output.

pyl62112991 commented 5 years ago

图片 I used my own test picture and the pre-processing model provided by the author for testing but failed,could you help me solve the problem

Gnaixzl commented 4 years ago

I am experiencing the same problem.

JusperLee commented 4 years ago

May be the size required for the image

jiangsutx commented 4 years ago

This network uses encoder-decoder-like structure, the images are first downsampled and then upsampled to original size. So in this process, the input size should be multiples of some power of 2. I think it is better to first pad the image to multiples of 16, and then feed into network. The output will be correct.

I will try to put a bugfix patch in test() function.