Real-Time Crowd Density Estimation Based on Convolutional Neural Networks

Aiming at the problems of traditional real-time crowd density estimation method which has big error and poor performance, the real-time crowd density estimation method based on CNN is proposed. By comparing the accuracy and real-time of four common structures: AlexNet VGGNet GoogLeNet and ResNet, choose GoogLeNet which has good comprehensive performance as core method, use key-frame extraction technology to realize real-time crowd density estimation and analyze crowd feature map. Finally, example prove the feasibility of the this method by real-time and accuracy.