This paper presents a novel Generative Probabilistic Modeling under anInverse Reinforcement Learning approach, named Subject-dependent Deep AgingPath (SDAP), to model the facial structures and the longitudinal face agingprocess of given subjects. The proposed SDAP is optimized using tractablelog-likelihood objective functions with Convolutional Neural Networks baseddeep feature extraction. In addition, instead of using a fixed agingdevelopment path for all input faces and subjects, SDAP is able to provide themost appropriate aging development path for each subject that optimizes thereward aging formulation. Unlike previous methods that can take only one imageas the input, SDAP allows multiple images as inputs, i.e. all information of asubject at either the same or different ages, to produce the optimal aging pathfor the subject. Finally, SDAP allows efficiently synthesizing in-the-wildaging faces without a complicated pre-processing step. The proposed method isexperimented in both tasks of face aging synthesis and cross-age faceverification. The experimental results consistently show the state-of-the-artperformance using SDAP on numerous face aging databases, i.e. FG-NET, MORPH,AginG Faces in the Wild (AGFW), and Cross-Age Celebrity Dataset (CACD). Themethod also performs on the large-scale Megaface challenge 1 to demonstrate theadvantages of the proposed solution.
展开▼
