SiGAN: Siamese Generative Adversarial Network for Identity-Preserving Face Hallucination
This work addresses identity preservation in face hallucination for applications like surveillance and biometrics, representing an incremental improvement over prior GAN-based methods.
The paper tackles the problem of identity loss in GAN-based face hallucination by proposing SiGAN, which uses a Siamese network with paired generators and a discriminator to preserve identity while reconstructing high-resolution faces. Experimental results show SiGAN significantly outperforms existing methods in face verification performance and maintains effectiveness on unseen identities.
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.