I noticed in Figure 2 of the paper that MaxPool operations are used in the high-frequency enhancement module. However, in the code:
x1 = self.updown(x1, raw)
x1 = self.downup(x1, raw)
only AvgPool is used. Could you please clarify if this is an oversight? If so, how significant is this oversight? Specifically, would the difference between average pooling and max pooling have a substantial impact on the network's performance?
Hello,
I noticed in Figure 2 of the paper that MaxPool operations are used in the high-frequency enhancement module. However, in the code: x1 = self.updown(x1, raw) x1 = self.downup(x1, raw) only AvgPool is used. Could you please clarify if this is an oversight? If so, how significant is this oversight? Specifically, would the difference between average pooling and max pooling have a substantial impact on the network's performance?
Thank you.