Abstract

Convolutional Neural Network (CNN) is widely used in estimating age from face image. In many CNN applications such as image classification, face recognition and other computer vision scopes, the cross-entropy loss is used as a supervision signal to train CNN model. However, the cross-entropy loss only enhances the separability of classes and does not consider their correlation in age estimation task. In this paper we propose a novel loss function called neighborhood loss which regards the correlation between classes in age estimation by modifying standard cross entropy loss. To evaluate the effectiveness of the proposed neighborhood loss, we present CNN architecture based on the residual units. Through some experiments, we show that neighborhood loss provides superior performance compared to prior works in age estimation.