The adversarial loss is the loss between the image from the real distribution A or B, and the images generated by the generator networks. We have two mapping functions and we will be applying adversarial loss to both of the mappings.

The adversarial loss for the mapping  is written as follows:

Here, x is an image from one domain from distribution A, and y is an image from another domain from distribution B. The discriminator  tries to distinguish between the image generated by the G mapping (), and the real image y from a different distribution B. The discriminator  tries to distinguish between the image generated by the F mapping () and the real image x from the distribution A. The objective of G is to minimize the adversarial loss function against an adversary D, which constantly tries to maximize it.