Similar to other GANs, CycleGANs have a training objective function, which we need to minimize in order to train the model. The loss function is a weighted sum of the following losses:

1. Adversarial loss
2. Cycle consistency loss

Let's explore the adversarial loss and the cycle consistency loss in detail in the following sections.