Proactive Headcount and Suspicious Activity Detection using YOLOv8

Abstract

Global population growth has resulted in more public gatherings, raising concerns on overcrowding and associated safety problems. Among these safety issues, crowd-smashing accidents stand out as unexpected and fast growing situations that pose serious risks to common people. Manually predicting and managing such chaotic situations leads to various challenges. To address these challenges, AI-based monitoring systems are currently used by integrating cutting-edge object detection algorithms such as YOLOv8. This advanced approach provides authorities with real-time crowd density evaluations and instant alerts, enabling immediate action and accident avoidance. Notably, the proposed model extends its capabilities beyond crowd counting by encompassing the detection of abnormal activities such as weapons, fires, falls, and smoke within the crowd. By identifying potentially dangerous crowd densities and quickly detecting abnormal incidents, the proposed system not only prevents disasters such as crowd smashing but also strengthens security measures, putting public safety at the forefront of the challenges posed by a growing global population and crowded public spaces.