Unsupervised Latent Space Translation Network

TL;DR

This paper enhances the UNIT framework for image-to-image translation by adding an adversarial discriminator on the latent space, significantly improving performance on domain adaptation tasks like MNIST and USPS.

Contribution

It introduces a novel adversarial discriminator on the latent space, addressing key limitations of the original UNIT framework for better domain translation.

Findings

Outperforms existing methods on MNIST and USPS tasks

Enforces similar latent space distributions across domains

Improves quality of image translation results

Abstract

One task that is often discussed in a computer vision is the mapping of an image from one domain to a corresponding image in another domain known as image-to-image translation. Currently there are several approaches solving this task. In this paper, we present an enhancement of the UNIT framework that aids in removing its main drawbacks. More specifically, we introduce an additional adversarial discriminator on the latent representation used instead of VAE, which enforces the latent space distributions of both domains to be similar. On MNIST and USPS domain adaptation tasks, this approach greatly outperforms competing approaches.