Image-to-Image Translation with Conditional Adversarial Networks

We investigate conditional adversarial networks as a general-purpose solutionto image-to-image translation problems. These networks not only learn themapping from input image to output image, but also learn a loss function totrain this mapping. This makes it possible to apply the same generic approachto problems that traditionally would require very different loss formulations.We demonstrate that this approach is effective at synthesizing photos fromlabel maps, reconstructing objects from edge maps, and colorizing images, amongother tasks. Indeed, since the release of the pix2pix software associated withthis paper, a large number of internet users (many of them artists) have postedtheir own experiments with our system, further demonstrating its wideapplicability and ease of adoption without the need for parameter tweaking. Asa community, we no longer hand-engineer our mapping functions, and this worksuggests we can achieve reasonable results without hand-engineering our lossfunctions either.