Advanced threat perception and personalized alert system using YOLOv8 model

In today's digital age, ensuring public safety demands innovative solutions that transcend traditional surveillance methods. This paper presents a pioneering endeavor in the realm of threat perception, introducing an Advanced Threat Perception and Personalized Alert System harnessing deep learning techniques. The system's primary objective is the detection of weapons, specifically guns and knives, within video content obtained from open platforms, thus broadening its applicability across diverse scenarios beyond closed-circuit television (CCTV) footage. Central to the system's architecture is the utilization of the YOLOv8 model, a state-of-the-art deep learning framework renowned for its efficiency in object detection tasks. To enable the model's proficiency in discerning potential threats, it undergoes rigorous training on a specialized dataset curated explicitly for violence detection. The efficacy of the model is meticulously evaluated employing established metrics including Mean Average Precision (MAP), Precision, Recall, and F1-Score, ensuring a comprehensive assessment of its performance.