WEAPON SURVEILLANCE AND DETECTION SYSTEM USING YOLO-NAS: A NOVEL APPROACH TO PUBLIC SAFETY

Abstract

A paradigm shift in the globe is the deepening apprehension of public safety in crowds demands the invention of modern weapon surveillance and detection meth- ods. This thesis offers a new weapon monitor and detection system named YOLO- NAS-based Weapon Surveillance and Detection System (WSDS). The system is powered with advanced deep neural network technology and contemporary com-puter vision algorithms that allow for faster, more accurate weapon detection, overcoming issues like partial or full occlusion and spatial resolution limitations. The study aims to provide a theoretical foundation by presenting the object de-tection concept in computer vision, explaining the YOLO-NAS architecture, and investigating the methods utilized by real-time monitoring systems. The method- ology involves: data gathering from multiple sources, dataset preparation, the training of the YOLO-NAS model, and the performance evaluation by the use of metrics such as precision, recall, and mean average precision. The YOLO-NAS structure embeds existing surveillance cameras, utilizing the central processor to execute dynamic detection thereby triggering the alarm system to activate the response routine immediately. The study measures the capacity, customizing, and the fitness of system to the public and utilities such as schools, shopping areas, and transit stations. The results and findings part presents the effectiveness of the method and makes a comparison with those that have been mentioned before. The thesis describes how this solution reshaped current practices in the area of public security. The thesis concludes with the identification of the project’s limitations providing recommendations for improving these and stressing the fact that the WEAPON  SURVEILLANCE  &  DETECTION  SYSTEM  based  on  YOLO-NAS let us move to a new level of modern weapon surveillance systems.