Learning and Interpreting Features to Rank: A Case Study on Age Estimation

Previously, it was taken for granted that features learned for classification can also be used for ranking. However, ranking problems possess some distinctive properties, e.g., the ordinal class labels, which indicates the necessity of developing new feature learning procedures dedicated for ranking. In this article, the authors propose to use a convolutional neural network (CNN)-based framework, ranking-CNN, for learning and interpreting features to rank. As a case study, the authors propose to analyze, visualize and work to understand the deep aging patterns in human facial images using ranking-CNN. The authors develop a visualization method that can compare the facial appearance and track its changes at different ages through the mapping between 2D images and a 3D face template. The framework provides an innovative way to understand the human facial aging process.