Huang Zhizhong ; Zhang Junping ; Shan Hongming

Abstract

To minimize the effects of age variation in face recognition, previous workeither extracts identity-related discriminative features by minimizing thecorrelation between identity- and age-related features, called age-invariantface recognition (AIFR), or removes age variation by transforming the faces ofdifferent age groups into the same age group, called face age synthesis (FAS);however, the former lacks visual results for model interpretation while thelatter suffers from artifacts compromising downstream recognition. Therefore,this paper proposes a unified, multi-task framework to jointly handle these twotasks, termed MTLFace, which can learn age-invariant identity-relatedrepresentation while achieving pleasing face synthesis. Specifically, we firstdecompose the mixed face feature into two uncorrelated components -- identity-and age-related feature -- through an attention mechanism, and then decorrelatethese two components using multi-task training and continuous domain adaption.In contrast to the conventional one-hot encoding that achieves group-level FAS,we propose a novel identity conditional module to achieve identity-level FAS,with a weight-sharing strategy to improve the age smoothness of synthesizedfaces. In addition, we collect and release a large cross-age face dataset withage and gender annotations to advance the development of the AIFR and FAS.Extensive experiments on five benchmark cross-age datasets demonstrate thesuperior performance of our proposed MTLFace over existing state-of-the-artmethods for AIFR and FAS. We further validate MTLFace on two popular generalface recognition datasets, showing competitive performance for face recognitionin the wild. The source code and dataset are availableat~\url{https://github.com/Hzzone/MTLFace}.