Bruno Oliveira

Regarding your SRResNet implementation (the generator), I was wondering why you used a LeakyReLU instead of a PReLU with a trainable "a". Reading the paper and seing your implementation raised...

PReLU num_parameters

3

Hi, In your SRResNet model, you define PReLU num_parameters = 256 in the upscale block. Shouldn't it be 64? I mean, after the PixelShuffle you should have 64 channels.