SiGAN: Siamese Generative Adversarial Network for Identity-Preserving Face Hallucination

TL;DR

SiGAN is a novel Siamese GAN that produces high-resolution face images from low-resolution inputs while preserving identity, outperforming existing methods in face verification and generalizing to unseen identities.

Contribution

The paper introduces SiGAN, a Siamese GAN architecture that incorporates identity preservation into face hallucination, a novel approach compared to prior GAN-based face super-resolution methods.

Findings

SiGAN achieves superior face verification accuracy over existing GANs.

The method produces photo-realistic high-resolution faces.

It generalizes well to unseen identities.

Abstract

Despite generative adversarial networks (GANs) can hallucinate photo-realistic high-resolution (HR) faces from low-resolution (LR) faces, they cannot guarantee preserving the identities of hallucinated HR faces, making the HR faces poorly recognizable. To address this problem, we propose a Siamese GAN (SiGAN) to reconstruct HR faces that visually resemble their corresponding identities. On top of a Siamese network, the proposed SiGAN consists of a pair of two identical generators and one discriminator. We incorporate reconstruction error and identity label information in the loss function of SiGAN in a pairwise manner. By iteratively optimizing the loss functions of the generator pair and discriminator of SiGAN, we cannot only achieve photo-realistic face reconstruction, but also ensures the reconstructed information is useful for identity recognition. Experimental results demonstrate that SiGAN significantly outperforms existing face hallucination GANs in objective face verification performance, while achieving photo-realistic reconstruction. Moreover, for input LR faces from unknown identities who are not included in training, SiGAN can still do a good job.