Real-Time In/Out Crowd Counting System Utilizing YOLOv8 with DeepSORT

With the growing demand for effective crowd management in public spaces, a crowd counting system that monitors foot traffic has become significant for safety and operational efficiency. This study introduces a real-time in/out crowd counting system developed for monitoring foot traffic in public spaces. Utilizing YOLOv8 for object detection and DeepSORT for tracking, the system consistently achieved an error rate of less than 10% in low to moderate crowd density. High-angle camera placement was used to optimize the system by minimizing occlusion. It was tested across three scenes: a maternal healthcare clinic (inside and outside) and a university setting, resulting in an average error of 3.758%. Despite challenges in dense crowds, it proved effective in low to moderate densities, displaying its crowd management capabilities.