Lightweight Children's Expression Recognition Network Based on Multi-scale Attention Mechanism

Aiming at the problems that ordinary convolutional neural network is difficult to extract effective features and network model parameters are complex in children's expression recognition, a lightweight expression recognition network with multi-scale attention mechanism is proposed. First, in order to increase the width and depth of the network and avoid redundant calculations, the Inception structure is introduced into the network to extract multi-scale feature information of images. Then, the Channel Attention mechanism is used to emphasize the areas related to facial expressions and suppress the irrelevant background areas, so as to improve the model's ability to represent important feature information. In addition, depthwise separable convolution is used in the convolutional layer to reduce network parameters and prevent overfitting. The experimental results show that the method has better recognition effect and stronger generalization ability, and has certain reference value for network structure setting and parameter configuration in facial expression recognition.