Unsupervised image-to-image translation techniques are able to map local texture between two domains, but they are typically unsuccessful when the domains require larger shape change. Inspired by semantic segmentation, we introduce a discriminator with dilated convolutions that is able to use information from across the entire image to train a more context-aware generator. This is coupled with a multi-scale perceptual loss that is better able to represent error in the underlying shape of objects. We demonstrate that this design is more capable of representing shape deformation in a challenging toy dataset, plus in complex mappings with significant dataset variation between humans, dolls, and anime faces, and between cats and dogs.
@article{arxiv.1808.04325,
title = {Improving Shape Deformation in Unsupervised Image-to-Image Translation},
author = {Aaron Gokaslan and Vivek Ramanujan and Daniel Ritchie and Kwang In Kim and James Tompkin},
journal= {arXiv preprint arXiv:1808.04325},
year = {2019}
}