Deep Age Distribution Learning for Apparent Age Estimation

Zeng-Wei Huo, Xu Yang, Chao Xing, Ying Zhou, Peng Hou, Jiaqi Lv, Xin Geng

Abstract: Apparent age estimation has attracted more and more researchers since its potential applications in the real world. Apparent age estimation differs from chronological age estimation that in apparent age estimation each facial image is labelled by multiple individuals, the mean age is the ground truth age and the uncertainty is introduced by the standard deviation. In this paper, we propose a novel method called Deep Age Distribution Learning(DADL) to deal with such situation. According to the given mean age and the standard deviation, we generate a Gaussian age distribution for each facial image as the training target instead of the single age. DADL first detects the facial region in image and aligns the facial image. Then, it uses deep Convolutional Neural Network(CNN) pre-trained based on the VGGFace and fine-tuned on the age dataset to extract the predicted age distribution. Finally it uses ensemble method to get the result. Our DADL method got a good performance in ChaLearn Looking at People 2016-Track 1: Age Estimation and ranked the 2nd place among 105 registered participants.