This document discusses generative adversarial networks (GANs). GANs are a class of machine learning frameworks where two neural networks, a generator and discriminator, compete against each other. The generator learns to generate new data with the same statistics as the training set to fool the discriminator, while the discriminator learns to better distinguish real samples from generated samples. When trained, GANs can generate highly realistic synthetic images, videos, text, and more. The document reviews several papers that apply GANs to image transformation, super-resolution image generation, and generating images from semantic maps. It also explains how GANs are trained through an adversarial game that converges when the generator learns the true data distribution.