This paper reports on WaterGAN, a generative adversarial network (GAN) for generating realistic underwater imagesfromin-air image and depth pairings in an unsupervised pipeline used for color correction of monocular underwater images. Cameras onboard