Improved mobileNet for lightweight facial expression identification
In order to solve the problems of low accuracy,poor real-time performance and large space-time load in the application of lightweight convolutional neural network MobileNet to facial expression recogni-tion,this paper proposes an improved MobileNet lightweight facial expression recognition method.Based on MobileNet X,this method introduces the SE attention module and optimizes the deep convolution layer and network structure according to the characteristics of the expression image,avoiding the problem of informa-tion loss and neuron"necrosis",and improving the recognition rate of the model's facial expression.Com-pared with MobileNet X model,the improved network model has low complexity and high recognition accu-racy.The experiment on Fer2013 facial expression dataset shows that the recognition rate of this method is 73.54％,which is higher than other facial expression recognition methods in recognition rate and time effi-ciency.
NETL
NSTL
万方数据
