People Counting in Crowd: Faster R-CNN

Abstract

Abstract-Because of its vast range of operations, people counting in crowds is a significant challenge in the field of computer vision. To achieve further dependable results of crowd counting, head discovery grounded ways are used rather than viscosity chart grounded crowd counting ways. This is because, in case of viscosity charts, it isn't always the correct position which contributes to final crowd count. This leads to unreliable results especially in case of false positives. This makes the entire task of head discovery in crowded scenes a grueling one to be solved. While face discovery has reached maturity, the more general task of chancing people in images and videotape still remains to be veritably challenging. Count of people may also be demanded for the statistical purposes which help to concoct marketing strategies or it may be used for crowd control in colorful situations image processing is a fashion for applying operations on an image in order to ameliorate it or prize precious information from it. In our design, input to the system is an image/ videotape of the surveillance system which further divided into image frames. Our proposed system gives the count of people in the scene using the Faster R- CNN object discovery algorithm.