Heterogeneous Face Recognition: Past, Present and Future
Date:
Infrared sensors are essential for developing face recognition systems that can operate in low-light and nighttime environments, as well as detecting presentation attacks such as 3D masks and makeup. In this talk, I will first summarize existing works on matching infrared faces against visible faces known as heterogeneous face recognition. Then, I will showcase two approaches that have been recently developed to address infrared-to-visible face matching: SG-GAN, in which we embed the semantic information to better preserve the shape; and LG-GAN, a latent-guided generative adversarial network that offers useful insights into interpreting and explaining infrared-to-visible face image translation. In the end, I will also describe potential applications of utilizing infrared sensors for face presentation attack detection and point out future challenges of heterogeneous face recognition.