Structural Analogy from a Single Image Pair

The task of unsupervised image-to-image translation has seen substantial advancements in recent years through the use of deep neural networks. Typically, the proposed solutions learn the characterizing distribution of two large, unpaired collections of images, and are able to alter the appearance of a given image, while keeping its geometry intact. In this paper, we explore the capabilities of neural networks to understand image structure given only a single pair of images, (Formula presented.) and (Formula presented.). We seek to generate images that are structurally aligned: that is, to generate an image that keeps the appearance and style of (Formula presented.), but has a structural arrangement that corresponds to (Formula presented.). The key idea is to map between image patches at different scales. This enables controlling the granularity at which analogies are produced, which determines the conceptual distinction between style and content. In addition to structural alignment, our method can be used to generate high quality imagery in other conditional generation tasks utilizing images (Formula presented.) and (Formula presented.) only: guided image synthesis, style and texture transfer, text translation as well as video translation. Our code and additional results are available in https://github.com/rmokady/structural-analogy/.