Figure 2.

Example of the functioning of a GAN. The generator creates synthetic images from random noise while the discriminator has to differentiate between real and synthetic images. The blue arrow shows the discriminator’s loss back-propagation, the red arrow shows the generator’s loss back-propagation.