The aim of crowd counting using optimal resource management is to estimate the number of people in crowded images or videos from surveillance cameras so that, respective authorities can get effective analysis of crowd flow and can effectively manage resources. Calculating number of people from various images from digital cameras or videos has variety of applications such as traffic monitoring, foot traffic counting from retail stores, safety applications and counting at tremendous crowd locations like in Masjid-e-Haram during Hajj and Umrah congregation and to develop strategy to manage the crowd in most optimal way. In addition to this we can have applications of crowd counting in various day-to-day applications like counting for some survey purposes. So Crowd Counting provides foot traffic at places such as Malls, Retail Stores and Public streets for every moment of time. This count will be used to provide statistical flow of crowd based on day, week, month and year at respective place. The crowd counting has some challenges too like non-uniform density images, background noises and occlusions present in images. Nevertheless, lot of research has been done in recent past and many new methodologies are evolving which are dealing really effectively with stated problems. In this paper, we are doing comparative study of various methodologies which are used for crowd counting and we are providing comprehensive idea about Convolutional Neural Network based approaches such as Multi Scale Convolutional Neural Network.