Closed mychina75 closed 2 years ago
Thank you for your attention~ Actually, the deep feature extraction means we apply convolution in feature space (the channel number is num_fea). Most previous works convert deep features back to image like: (1) Conv2d(num_fea -> out_channels*(scale2)); (2) depth_to_space(scale). Our another more convolution: (1)Conv2d(num_fea -> out_channels*(scale*2)); (2) ReLU(); (3) Conv2d(out_channels(scale2) -> out_channels*(scale**2)); (4) depth_to_space(scale)
If you have any questions, feel free to contact me.
Get your point, thanks a lot~
Great work~ I have a question about readme mentioned: Another more convolution after deep feature extraction Looks like your model here not add addtional convolution?
thank you.