Robust Human Target Detection and Aquisitions

In recent years, the rise of security threats in public and private spaces has emphasized the need for intelligent surveillance systems. This research presents a real-time AI-based threat detection model that identifies potential hazards such as guns, knives, and masks using a customized YOLOv8 architecture integrated with OpenCV. The system is designed to differentiate threatening and non-threatening objects across 27 classes, providing immediate alerts through a web-based dashboard and voice notifications. The application, built using Flask, JavaScript, and SQLite, offers a live camera feed and automated logging of detected threats with time and date. Achieving an accuracy of 90% and high frame-rate inference, the system demonstrates strong potential for real-world deployment in smart surveillance, ensuring rapid and automated responses to life-threatening events.