Real -time face expression recognition network based on attention mechanism

Convolutional neural networks are widely used in the field of face expression recognition. In order to meet the problem of high recognition rate and lightweight model of facial expression recognition algorithm based on convolutional neural network, this paper proposes a real-time facial expression recognition network based on attention mechanism. First of all, this paper uses the Mini-Xception convolutional neural network, which removes the fully connected layer of the traditional model and uses the depth-wise separable convolution instead to reduce the model parameters. At the same time, the convolutional block attention model CBAM attention mechanism is added to the network architecture to improve the detection effect of the attention target, so as to improve the network performance without increasing the network parameters. Finally, the elu activation function was introduced to enhance the robustness of the model. The experimental results show that the recognition rate of the proposed method on the public facial expression dataset FER-2013 reached 72.9%, which has a higher accuracy than the current advanced expression recognition network. At the same time, the expression recognition system interface is designed to test different expressions, and the time per frame is about 0.3s while achieving fast and accurate testing. Therefore, the improved network in this paper can realize fast and accurate facial expression recognition.