No MaxPool?

[kczimm]

I'm happy to see a Julia implementation of U-Net. I was looking through the model and could not find an instance of MaxPool anywhere. Is this implementation incomplete? Did I miss this behavior somewhere else in the implementation?

Thanks in advance.

[johnnychen94]

There're multiple ways to downsample/fold an image, 2x2 maxpool is what the original paper uses. This package uses Conv with stride 2.

UNet.jl/src/model.jl

Lines 12 to 15 in a5ac874

	ConvDown(in_chs,out_chs,kernel = (4,4)) =
	Chain(Conv(kernel,in_chs=>out_chs,pad=(1,1),stride=(2,2);init=_random_normal),
	BatchNormWrap(out_chs),
	x->leakyrelu.(x,0.2f0))