search

wbhu / DnCNN-tensorflow

:octocat::octocat:A tensorflow implement of the paper "Beyond a Gaussian Denoiser: Residual Learning of Deep CNN for Image Denoising"
GNU General Public License v3.0
363 stars 150 forks source link

For color images #24

Closed aditipanda closed 6 years ago

aditipanda commented 6 years ago

It doesn't work for color images. @crisb-DUT @lizhiyuanUSTC could you provide generate_patches.py for color images?

wbhu commented 6 years ago

Hi @appyfizzA ,

The current model is trained for gray images. But is quite easy to trained a model for color images. Maybe it's like this:

  1. choose your color images training set
  2. modify the input and output channel number
  3. train your model

Thanks, Wenbo

aditipanda commented 6 years ago

Yes, @crisb-DUT but I think that's not enough.

The patch generation process will be different, each patch should be 50 by 50 by 3, isn't it? After all, these are color images. That's why I requested for generate_patches.py .

wbhu commented 6 years ago

Hi @appyfizzA ,

Sorry for reply so late. You are right that generate_patches.py should be modified. If you are interested in building a model for colored images denoisng, welcome to pull requests.

Thanks, Wenbo