Facial Age Estimation By Adaptive Label Distribution Learning

Facial age estimation has recently become an active research topic in computer vision and pattern recognition and has extensive applications in the real world. This paper proposes a novel approach to facial age estimation based on adaptive label distribution learning. The proposed approach introduces label distribution to facial age estimation problem and learns label distributions adapted to different ages in a data-driven way. The adaptive label distribution learning (ALDL) makes a face image contribute to not only the learning of its chronological age, but also the learning of its neighboring ages. Thus, this approach improves the utilization of the training data and is an efFective way to relieve the problem of insufficient and incomplete training data. What’s more, ALDL generates label distributions according with the general facial aging pattern of human beings adaptively, which significantly improves the precision of age estimation.There are five chapters in this paper. The first chapter introduces the research background, the applications, the challenges and the current research status in facial age estimation. The second chapter introduces face image feature representation models and facial age estimation algorithms. The third chapter first analyzes the feasibility of age estimation through facial images, then introduces the concept of adaptive label distribution learning, and finally proposes two facial age estimation algorithms by adaptive label distribution learning, i.e. IIS-ALDL and BFGS-ALDL. The fourth chapter reports the performance of different algorithms on the MORPH database and the results show that the ALDL algorithms performs remarkably better than compared state-of-the-art algorithms. The fifth chapter summarizes the whole paper.The main contributions of this paper:1. It introduces label distribution to facial age estimation; 2. It proposes the adaptive label distribution learning approach to learn in a data-driven way; 3. It proposes an fast and efficient facial age estimation algorithm.