Discriminative attention-augmented feature learning for facial expression recognition in the wild

Facial expression recognition (FER) in-the-wild is challenging due to unconstraint settings such as varying head poses, illumination, and occlusions. In addition, the performance of a FER system significantly degrades large intra-class variation inter-class similarity facial expressions in real-world scenarios. To mitigate these problems, we propose novel approach, Discriminative Attention-augmented Feature Learning Convolution Neural Network (DAF-CNN), which learns discriminative expression-related representations for FER. Firstly, develop 3D attention mechanism feature refinement selectively focuses on attentive channel entries salient spatial regions convolution neural network map. Moreover, deep metric loss termed Triplet-Center (TC) incorporated further enhance power deeply-learned features with an expression-similarity constraint. It simultaneously minimizes distance maximizes learn both compact separate features. Extensive experiments have been conducted two representative datasets (FER-2013 SFEW 2.0) demonstrate that DAF-CNN effectively captures achieves competitive or even superior compared state-of-the-art methods.