Multi-feature ordinal ranking for facial age estimation

In this paper, we propose a multi-feature ordinal ranking (MFOR) method for facial age estimation. Different from most existing facial age estimation approaches where age estimation is treated as a classification or a regression problem, we formulate facial age estimation as a group of ordinal ranking subproblems, and each subproblem derives a separating hyperplane to divide face instances into two groups: samples with age larger than k and samples with labels no larger than k. To better extract complementary information from different facial features, we construct multiple ordinal ranking models, each corresponding to a feature set, and aggregate them into an effective age estimator. Experimental results on two public face aging datasets are presented to demonstrate the efficacy of the proposed method.