Semi-supervised Face Aging and Rejuvenating
Wanyue Ma,Yuan Zhou,Jun He
DOI: https://doi.org/10.1117/1.jei.30.2.023003
IF: 0.829
2021-01-01
Journal of Electronic Imaging
Abstract:Face aging and rejuvenating aim to generate an individual face with aging and rejuvenating effect while retaining identity information. We can analyze a given face image to estimate a past look or predict a future look of the person. The research on face aging and rejuvenating has important application value in the fields of cross-age recognition,1 public security, and entertainment, for example, changing the appearance of actors at different ages in a movie or finding missing persons in forensic applications. Although this area has attracted much attention of the researchers, there are still many challenges, especially in lack of accurate and sufficient dataset, low aging effect, and bad identity preservation. Previous face aging and rejuvenating methods are split into two main categories: physical model-based methods2,3 and prototype-based methods.4-6 The physical model-based methods describe the alteration in muscle, wrinkle, skin, etc., which can get a good result but suffer from complex modeling. The prototype-based methods try to learn the transformation between different age groups. Due to the development of generative adversarial network (GAN),7 state-of-theart methods8-10 that use the technology of deep learning show impressive success in this field. Face aging and rejuvenating work effectively in public security criminal investigation, cross-age recognition, and entertainment. However, three main problems still exist: the lack of accurate and sufficient dataset, low aging effect, and poor preservation of personal information. We propose a semi-supervised face aging and rejuvenating method for face aging and rejuvenating. In particular, a conditional encoder is utilized to map an input face into a latent vector, which is used by the generator network with age conditions to produce a new face. The latent vector preserves identity information, whereas the age label controls face aging or rejuvenating. To make generated features closer to prior features, the discriminator network is designed to assist the generator network. In addition, a cycle optimized method is utilized to preserve the personal information of the generated face. Experimental results demonstrate that our network can generate more realistic faces, both in personal identity and age consistency. (c) 2021 SPIE and IS&T [DOI: 10.1117/1.JEI.30.2.023003]