Real-Time Crowd Detection System

Real-Time Crowd Detection System

S. Pavani1, V. Abhiram2, S. Agnivesh3, P. Bharath4, P. Priyanka5

1Professor Department & College J.B. Institute of Engineering and Technology, Hyderabad

2 Department & College J.B. Institute of Engineering and Technology, Hyderabad

3Department & College J.B. Institute of Engineering and Technology, Hyderabad

4Department & College J.B. Institute of Engineering and Technology, Hyderabad

5Department & College J.B. Institute of Engineering and Technology, Hyderabad

---------------------------------------------------------------------***---------------------------------------------------------------------

Abstract - The Real-Time Crowd Detection & Monitoring System is an AI-driven solution designed to enhance public safety and crowd management by automatically detecting and counting individuals in images, videos, and live camera feeds. Utilizing TensorFlow’s Faster R-CNN Inception V2 model, the system ensures high accuracy in human detection, enabling real-time monitoring of crowded areas. A customizable crowd threshold allows users to define safe limits, and when the detected count exceeds this threshold, the system triggers multiple alert mechanisms, including an audio warning, on-screen notification, and an automated email alert to security personnel. Additionally, the HELP button enables users to notify security teams instantly in emergency situations.

A key feature of the system is automated report generation, which compiles detection data, crowd density trends, timestamps, and security alerts into a PDF report for further analysis and record-keeping. The Tkinter-based graphical user interface (GUI) provides an intuitive platform where users can switch between image, video, and live camera modes, adjust detection settings, and enable or disable alerts. To ensure smooth performance, the system is optimized with multi-threading and efficient model loading techniques, minimizing processing delays.

Compared to traditional surveillance methods, which require constant human supervision, this system provides a fully automated, scalable, and intelligent solution for public safety, event monitoring, and security management. Future improvements may include mobile app integration, SMS alerts, cloud-based monitoring dashboards, and enhanced deep learning models to further improve detection efficiency and responsiveness. This system represents a significant advancement in intelligent surveillance and proactive crowd control, ensuring efficient and real-time monitoring of high-density environments.

Key Words: Real-time Human Detection, Deep Learning, TensorFlow, OpenCV