Coarse-to-Fine Age Progression Based on Conditional GANs

Jiaxin Chen, Jianjun Qian, Zhihui Lai

Abstract: This paper presents a novel method coined Coarse-to-Fine Age Progression based on Conditional GANs (CFAP). CFAP firstly employs the Illumination-Aware Age Progression (IAAP) to achieve the initial results. To overcome the problem of data limited, we select the fine synthetic results of IAAP to construct a series of images pairs (source age image and synthetic target age image) as training set. Then, the powerful image changes tool Conditional GANs is used to learn the aging pattern variations and further refine the aged results. Experimental results demonstrate the advantages of the proposed model compared with previous works.