search

GuoShi28 / CBDNet

Code for "Toward Convolutional Blind Denoising of Real Photographs", CVPR 2019
Apache License 2.0
501 stars 92 forks source link

Question about CRF #11

Closed LoSealL closed 5 years ago

LoSealL commented 5 years ago

Hi, it's an amazing work! I wonder in your noise model (EQ 3), did you try to train the model with and without CRF? If we train CBDNet without the knowledge of CRF, how much gap will appear in real photo denoising? I don't find any comparison among EQ 1, 2 and 3.

GuoShi28 commented 5 years ago

Hi, Thank you for your interest! I have added some discussion about the function of ISP in noise modeling in the new version of my paper. But I did not discuss the function of CRF, so I am not quite sure the quantitative gap. I think the more perfect you model the ISP process, the better denoising results you can obtain.

LoSealL commented 5 years ago

Thanks you:) And in your paper, you mention a CBDNet-Real, does it mean that you train it with ONLY real photo by L{tv} and disgard L{asymm}?

GuoShi28 commented 5 years ago

CBDNet Real only train the network using L2 without estimating the noise estimation loss. The network structure is the same with other comparison networks.

LoSealL commented 5 years ago

Thanks a lot. I have no more questions.