Learning to Generate Things and Stuff: Guided Generative Adversarial Networks for Generating Human Faces, Hands, Bodies, and Natural Scenes