Transforming a Horse to a Zebra Using A Generative Adversarial Network (GAN)

When two deep learning models work together in the style of a zero-sum game, one agent's gain is another agent's loss. This is known as a generative adversarial network (GAN). With a training set as its starting point, this approach learns to generate new data with the same statistics as the training set. A GAN can create new images with realistic elements that look a lot like the originals. GAN are popular because they give accurate results.