Human face analysis is an important task in computer vision. According tocognitive-psychological studies, facial dynamics could provide crucial cues forface analysis. In particular, the motion of facial local regions in facialexpression is related to the motion of other facial regions. In this paper, anovel deep learning approach which exploits the relations of facial localdynamics has been proposed to estimate facial traits from expression sequence.In order to exploit the relations of facial dynamics in local regions, theproposed network consists of a facial local dynamic feature encoding networkand a facial relational network. The facial relational network is designed tobe interpretable. Relational importance is automatically encoded and facialtraits are estimated by combining relational features based on the relationalimportance. The relations of facial dynamics for facial trait estimation couldbe interpreted by using the relational importance. By comparative experiments,the effectiveness of the proposed method has been validated. Experimentalresults show that the proposed method outperforms the state-of-the-art methodsin gender and age estimation.
展开▼
