Simultaneous Face Hallucination and Translation for Thermal to Visible Face Verification using Axial-GAN

TL;DR

This paper introduces Axial-GAN, a novel model that synthesizes high-resolution visible face images from low-resolution thermal images to improve verification accuracy in long-range surveillance scenarios.

Contribution

The paper proposes Axial-GAN, integrating axial-attention layers into GANs for thermal-to-visible face synthesis, addressing resolution mismatch in real-world applications.

Findings

Significant improvement in face verification accuracy.

Enhanced image quality over state-of-the-art methods.

More efficient synthesis process.

Abstract

Existing thermal-to-visible face verification approaches expect the thermal and visible face images to be of similar resolution. This is unlikely in real-world long-range surveillance systems, since humans are distant from the cameras. To address this issue, we introduce the task of thermal-to-visible face verification from low-resolution thermal images. Furthermore, we propose Axial-Generative Adversarial Network (Axial-GAN) to synthesize high-resolution visible images for matching. In the proposed approach we augment the GAN framework with axial-attention layers which leverage the recent advances in transformers for modelling long-range dependencies. We demonstrate the effectiveness of the proposed method by evaluating on two different thermal-visible face datasets. When compared to related state-of-the-art works, our results show significant improvements in both image quality and face verification performance, and are also much more efficient.