Unsupervised Attention-guided Image to Image Translation

TL;DR

This paper introduces an unsupervised attention mechanism for image-to-image translation that improves focus on relevant objects, leading to more realistic image mappings without supervision.

Contribution

It proposes a novel unsupervised attention-guided framework trained adversarially, enhancing focus on objects and scene interactions in image translation tasks.

Findings

Improved focus on relevant image regions without supervision

Achieves more realistic image translations compared to recent methods

Qualitative and quantitative validation of attention effectiveness

Abstract

Current unsupervised image-to-image translation techniques struggle to focus their attention on individual objects without altering the background or the way multiple objects interact within a scene. Motivated by the important role of attention in human perception, we tackle this limitation by introducing unsupervised attention mechanisms that are jointly adversarialy trained with the generators and discriminators. We demonstrate qualitatively and quantitatively that our approach is able to attend to relevant regions in the image without requiring supervision, and that by doing so it achieves more realistic mappings compared to recent approaches.