Image-to-Image Translation with Conditional Adversarial Networks

TLDR

The paper explores conditional adversarial networks as a universal method for image‑to‑image translation. The approach learns both the image mapping and its loss function. The method successfully generates photos from label maps, reconstructs objects from edge maps, colorizes images, and has spurred widespread community use, showing that hand‑engineered loss functions are unnecessary.

Abstract

We investigate conditional adversarial networks as a general-purpose solution to image-to-image translation problems. These networks not only learn the mapping from input image to output image, but also learn a loss function to train this mapping. This makes it possible to apply the same generic approach to problems that traditionally would require very different loss formulations. We demonstrate that this approach is effective at synthesizing photos from label maps, reconstructing objects from edge maps, and colorizing images, among other tasks. Moreover, since the release of the pi×2pi× software associated with this paper, hundreds of twitter users have posted their own artistic experiments using our system. As a community, we no longer hand-engineer our mapping functions, and this work suggests we can achieve reasonable results without handengineering our loss functions either.