Live Monitoring and Smart Threat Detection Using YOLOv8 Based Deep Learning Model

Abstract

This study introduces a real-time surveillance framework aimed at detecting weapons, with a particular focus on handguns and knives. The approach leverages the YOLOv8 object detection model, chosen for its balance of precision and computational speed. Implemented in Python and integrated through a Flask-based web application, the system provides interactive and user-friendly monitoring options. With a dataset of nearly 4,000 annotated images under varied environments, the framework supports three modes: static image analysis, video stream evaluation, and live webcam monitoring. Experimental evaluation achieved an overall accuracy of 64%, demonstrating its potential for real-world deployment in enhancing security measures.