Advanced Deep Learning with Python

Conditional GAN

The conditional GAN (CGAN, Conditional Generative Adversarial Nets, https://arxiv.org/abs/1411.1784) is an extension of the GAN model where both the generator and discriminator receive some additional conditioning input information. This could be the class of the current image or some other property:

Conditional GAN. Y represents the conditional input for the generator and discriminator

For example, if we train a GAN to generate new MNIST images, we could add an additional input layer with values of one-hot encoded image labels. CGANs have the disadvantage that they are not strictly unsupervised and we need some kind of label ...