Deep label refinement for age estimation. (April 2020)

• Record Type: Journal Article
• Authors: Li, Peipei; Hu, Yibo; Wu, Xiang; He, Ran; Sun, Zhenan
• Is Part Of: Pattern recognition. Volume 100(2020:Apr.)
• Publication Date: 2020-04
• Subjects: Age estimation -- Deep learning -- Convolutional neural networks -- Label distribution learning; Pattern perception -- Periodicals; Perception des structures -- Périodiques; Patroonherkenning; 006.4
• DOI: 10.1016/j.patcog.2019.107178
• Languages: English
• ISSNs: 0031-3203

Abstract

Highlights: We propose a Label Refinery Network (LRN) with two concurrent processes: label distribution refinery and slack regression refinery. The proposed label distribution refinery adaptively estimates the age distributions without the strong assumptions about the form of label distribution. Benefiting from the constant refinery of the learning results, label distribution refinery generates more precise label distributions. To further utilize the correlations among different age labels, we introduce regression to assist label distribution refinery. Besides, we introduce a slack term to further convert the discrete age label regression to the continuous age interval regression. We evaluate the effectiveness of the proposed LRN on three age estimation benchmarks and consistently obtain the state-of-the-art results. Abstract: Age estimation of unknown persons is a challenging pattern analysis task due to the lack of training data and various ageing mechanisms for different individuals. Label distribution learning-based methods usually make distribution assumptions to simplify age estimation. However, since different genders, races and/or any other characteristics may influence facial ageing, age-label distributions are often complicated and difficult to model parametrically. In this paper, we propose a label refinery network (LRN) with two concurrent processes: label distribution refinement and slack regression refinement. The label refinery network aims to learn age-label distributions progressively in an iterative manner. In this way, we can adaptively obtain the specific age-label distributions for different facial images without making strong assumptions on the fixed distribution formulations. To further utilize the correlations among age labels, we propose a slack regression refinery to convert the age-label regression model into an age-interval regression model. Extensive experiments on three popular datasets, namely, MORPH Album2, ChaLearn15 and MegaAge-Asian, demonstrate the superiority of our method.