A Crowd Counting Framework Combining with Crowd Location

In the past ten years, crowd detection and counting have been applied in many fields such as station statistics, urban safety prevention, people flow statistics. However, obtaining accurate positions improving performance of dense scenes still face challenges, it is worthwhile devoting much effort to this. this paper, a new framework proposed resolve problem. The includes two parts. first part fully convolutional neural network (CNN) consisting backend upsampling. part, uses residual (ResNet) encode features input picture, upsampling deconvolution layer decode feature information. processes image, processed image second part. peak confidence map (PCM), which based on an improvement over density (DM). Compared with DM, PCM can not only solve problem but also accurately predict location person. experimental results several datasets (Beijing-BRT, Mall, Shanghai Tech, UCF_CC_50 datasets) show that achieve higher scenarios crowds.