Video analytics using deep learning for crowd analysis: a review

Abstract Gathering a large number of people in shared physical area is very common urban culture. Although there are limitless examples mega crowds, the Islamic religious ritual, Hajj, considered as one greatest crowd scenarios world. The Hajj carried out once year with congregation millions when Muslims visit holy city Makkah at given time and date. Such big always prone to public safety issues, therefore requires proper measures ensure safe comfortable arrangement. Through advances computer vision based scene understanding, automatic analysis scenes gaining popularity. However, existing algorithms might not be able correctly interpret video content context Hajj. This because unique crowded small area, which can overwhelm use sophisticated algorithms. our studies on analysis, counting, density estimation, behavior, we faced need review work get research direction for abnormal behavior pilgrims. Therefore, this aims summarize works relevant broader field analytics using deep learning special focus visual surveillance identifies challenges leading-edge techniques general, may gracefully adaptable applications Umrah. paper presents detailed reviews approaches employed from videos, specifically that detecting behavior. These observations give us impetus undertake painstaking yet exhilarating journey classification detection any movement Furthermore, pilgrimage most domain video-related extensive activities, study motivates critically analyze scale.