Crowd Count Estimator Application using CSRNet

We deploy a network for Congested Scene Recognition which can also be called as CSRNet. The main purpose of CSRNet is to understand and identify people from a highly congested image and perform accurate analysis, project statistical data for the current session and the density map. Basically, it deploys a deeper CNN (Convolution Neural Network) for capturing high level features and generate high quality density maps. We are using a modified version of the VGG-16 architecture and we have used this because it can extract high level features and project high quality density maps without increasing the network complexity exponentially. We showcase our project and its accuracy on one dataset i.e. ShanghaiTech Dataset.