Learning from Longitudinal Face Demonstration - Where Tractable Deep Modeling Meets Inverse Reinforcement Learning

  • 2018-02-20 04:59:11
  • Chi Nhan Duong, Kha Gia Quach, Khoa Luu, T. Hoang Ngan Le, Marios Savvides, Tien D. Bui
  • 0

Abstract

This paper presents a novel Subject-dependent Deep Aging Path (SDAP), whichinherits the merits of both Generative Probabilistic Modeling and InverseReinforcement Learning to model the facial structures and the longitudinal faceaging process of a given subject. The proposed SDAP is optimized usingtractable log-likelihood objective functions with Convolutional Neural Networks(CNNs) based deep feature extraction. Instead of applying a fixed agingdevelopment path for all input faces and subjects, SDAP is able to provide themost appropriate aging development path for individual subject that optimizesthe reward aging formulation. Unlike previous methods that can take only oneimage as the input, SDAP further allows multiple images as inputs, i.e. allinformation of a subject at either the same or different ages, to produce theoptimal aging path for the given subject. Finally, SDAP allows efficientlysynthesizing in-the-wild aging faces. The proposed model is experimented inboth tasks of face aging synthesis and cross-age face verification. Theexperimental results consistently show SDAP achieves the state-of-the-artperformance on numerous face aging databases, i.e. FG-NET, MORPH, AginG Facesin the Wild (AGFW), and Cross-Age Celebrity Dataset (CACD). Furthermore, wealso evaluate the performance of SDAP on large-scale Megaface challenge todemonstrate the advantages of the proposed solution.

 

Quick Read (beta)

loading the full paper ...