As described above, one tricky issue when applying convolutional layers is that we tend to lose pixels on the perimeter of our image. Consider :numref:img_conv_reuse that depicts the pixel utilization ...