Group Expression Analysis Based On Deep Learning

With the rapid development of modern social network,the analysis of group emotion is becoming more and more important.As an effective feature learning method,deep learning has been developed rapidly in recent years.This paper uses deep learning to conduct the analysis of group emotion.The estimation method of group happiness is the main research point of this paper.The main research contents are as follows.(1)To solve face detection,this paper mainly studies a method based on cascaded CNN,and it is compared with the Adaboost face detection method based on haar-like in traditional machine learning.(2)This paper uses the CNN to identify the facial expressions.Experiment with AlexNet and CaffeNet and AlexNet has been improved.To improve the nonlinear expression ability of the deep convolutional neural network,the method of using multi-layer small scale convolution kernel to replace single layer large scale convolution kernel is used.The result is compared with mini Xception which is the new network model.(3)Owning to the influence of the context information being ignored when the method of calculating mean value is directly used,this paper presents weighted average method and neural network automatic weighting method to predict the group emotion conveyed in the image.