Estimating Crowd Size for Public Place Surveillance Using Deep Learning

Abstract

With the overwhelming speed of population outbursts throughout the world, it raises a security concern over public places like supermarkets, offices, banks, political rallies, religious events etc. The risk of any abnormal behavior or any security concern arises with the number of people in the crowd. Most of the public places are overcrowded and require crowd count monitoring to avoid any mishappening. It is impossible to appoint personnel at different locations to count people over there. The role of CCTV cameras is pertinent as far as remote monitoring of crowds over public places is concerned, but it is a cumbersome job for a person to count people in a crowd only by monitoring multiple videos of different locations at a time. Also, with the help of CCTV footage it is not an easy task to manage crowds, specifically if it is a dense crowd. Automated crowd count that can estimate the total number of people in a crowd image is needed for the hour. In this chapter, we have reviewed crowd count methods using state of art deep learning models for automated crowd count and their performance analysis on major crowd counting datasets.