AI-Driven Crowd Analytics: Enabling Intelligent Connectivity Through Deep Learning

Crowd Counting in photography is the counting of individuals presence in a picture. Crowd counting has become increasingly vital for public security, urban planning and event management. Its goal is to distinguish randomly sized objects in different situations, such as scattered and crowded places at identical times. Even in the $21^{\text {st }}$ Century, counting people at many places is done using outdated ways, such as at the entrance, registering people standing at counters. These techniques generally fail at places where the motion of people is highly uncertain and random, and keeping a record of the population present in the event or geographical area is cumbersome. It is of supreme importance for emergency evacuations in the presence of a large crowd, for example, in case of fire outbreaks, earthquakes, etc., because if the tally of individuals is known to the authorities, it will enable them to make an informed decision. Due to its vast application and growing importance, crowd-counting has gained the attention of many researchers. Different techniques, such as detection-based counting, regression, density estimation, and clustering, have been put forward to perform crowd counting. This paper explores the evolution and effectiveness of AI-driven crowd analytics, particularly leveraging deep learning techniques. In this research paper, we intend to bring light to studies conducted by different creators utilizing various procedures for counting the crowd using a photograph and giving a blueprint of the proposed method and highlighting their shortcomings.