A question about U-net 'sse_block(prevlayer, prefix)

[enjlife]

Hi, I found something different from the article about the sSE(channel squeeze and spatial excitation). The article said that generating a projection tensor q ∈ RH×W . Each qi,j of the projection represents the linearly combined representation for all channels C for a spatial location (i,j). So, should the input of Conv2D be Conv2D(1,(1,1), padding='valid').Please help me.

def sse_block(prevlayer, prefix):
# Bug? Should be 1 here?
conv = Conv2D(K.int_shape(prevlayer)[3], (1, 1), padding="same", kernel_initializer="he_normal",
              activation='sigmoid', strides=(1, 1),
              name=prefix + "_conv")(prevlayer)
conv = Multiply(name=prefix + "_mul")([prevlayer, conv])
return conv

[ybabakhin]

Hi @enjlife yes, you're right. It's a bug I've made in the very beginning. There is even a comment one line above: Bug? Should be 1 here?

However, for the sake of reproducibility I left it as is.

[enjlife]

Hi @ybabakhin Thanks for your reply, good luck!