FLGC-Fusion GAN: An Enhanced Fusion GAN Model by Importing Fully Learnable Group Convolution

Figure 4

A block convolutional layer example used in . (a) The FLGC convolution, in which grouping structures including input channels and filters are dynamically learned. The octagon represents the input channel, and the circle represents the filter. (b) The structure drawing for the index-recording layer to rearrange the channels in a practical application for grouping convolution.

(a)

(b)