Joint gender classification and age estimation by nearly orthogonalizing their semantic spaces

In human face-based biometrics, gender classification and age estimation are two important research topics. Although a variety of approaches have been proposed to handle them, just a few of them are solved jointly, even so, these joint methods do not specifically concern the semantic difference between human gender and age, which is intuitively helpful for joint learning, consequently leaving us a room of further improving their performance. To this end, in this work we firstly propose a general learning framework for jointly estimating human gender and age by attempting to formulate such semantic relationships as a form of near-orthogonality regularization and then to incorporate it into the objective of the joint learning framework. In order to evaluate the effectiveness of the proposed framework, we exemplify it by respectively taking the widely used binary-class SVM for gender classification, and two threshold-based ordinal regression methods (i.e., the discriminant learning for ordinal regression and support vector ordinal regression) for age estimation, and crucially coupling both through the proposed semantic formulation. Moreover, we construct its nonlinear counterpart by deriving a representer theorem for the joint learning strategy. Finally, extensive experiments on four aging datasets, i.e., FG-NET, Morph Album I, Album II and Images of Groups demonstrate the effectiveness and superiority of the proposed strategy.