Figure 3.

The difference between convolution and spatial channel-wise convolution. (A) Convolution: the convolution kernel slides on the (x, y)-plane, and the resulting feature map is concatenated along the z-axis direction, which is the direction of the channel of the feature maps. (B) Spatial channel-wise convolution: the convolution kernel slides in the z-axis direction, and the resulting feature maps are arranged in the (x, y)-plane. The arrangement rules are the same as those of the input feature maps.