Image to image translation involves translating an image from one domain to a corresponding image in another domain while preserving the structure in the content such as objects. For example, sketches can be the source domain while the realistic face images can be the target domain. The basic GAN model is widely used in virtual image generation. It is composed of an input vector, a generator and a discriminator. Think of the generator as attempting to deceive the discriminator while the discriminator attempts to distinguish the difference between a real and a fake sample.