Penny

Thanks for your awesome work. I infered some low quality face images. And I found one problem: the color style of output is different from that of input. Output image...

why padding == kernel_size // 2 is asserted?

3

https://github.com/DingXiaoH/DiverseBranchBlock/blob/be15be76a5556e04b2b44411a69994abcd1f25eb/diversebranchblock.py#L105 Why padding should be equal to kernel // 2? what if Conv2d(kernel_size=4, stride=2, padding=1)?