Leveraging Machine Learning for Real-Time Crowd Control and Safety at Kumbh Mela

The Kumbh Mela, one of the largest religious gatherings worldwide, presents significant challenges in crowd management and security due to its immense scale. This project introduces a technology-driven solution leveraging machine learning, computer vision, and real-time video processing to address these challenges. The system integrates YOLO, a realtime object detection model, for crowd density estimation; ByteTrack for tracking and detecting blockades; and DeepFace for facial recognition to locate lost individuals. In addition, weapon detection is incorporated to enhance security. The framework features a responsive front end, a FastAPI powered back end, and real-time communication via WebSockets. A mobile application delivers real-time alerts to facilitate timely interventions, ensuring safer and more efficient crowd management. Preliminary tests demonstrate a 90% accuracy in crowd density estimation. This scalable solution can be adapted for other large-scale events, such as festivals and concerts, offering improved safety and operational efficiency.