search

cszn / FFDNet

FFDNet: Toward a Fast and Flexible Solution for CNN based Image Denoising (TIP, 2018)
https://ieeexplore.ieee.org/abstract/document/8365806/
469 stars 130 forks source link

how can num_input_channels=15 in training step and 3 in test step? #17

Open HymEric opened 5 years ago

HymEric commented 5 years ago

Thanks for your work. But I have a question about how can num_input_channels=15 in training step and 3 in test step?

cszn commented 5 years ago

The 3-channel image is first reshaped to 12-channel sub-images by a PixelUnshuffle layer.