Pixels to Protection: An Intelligent Weapon Detection System

Weapon detection is a crucial issue for public safety. With rising concerns over security and the protection of personal property, the need for advanced video surveillance systems that can accurately recognize and interpret scenes has become more pressing. This is especially important in areas experiencing an increase in mass shootings and gun violence, where quick detection of firearms can save lives and prevent damage. However, existing object detection models often struggle to identify firearms due to their unique shapes and sizes, as well as the diverse backgrounds in which they are found. To address this problem, we have developed a cutting-edge system based on the YOLO V8 deep learning model, enhanced with attention mechanisms and Vision Transformers (ViTs). Our approach aims to improve the system's ability to handle different transformations, rotations, occlusions, and size variations. We tested our system using a publicly available dataset and achieved an impressive F1-score of 97.43%. Our methodology includes a range of data augmentation and preprocessing techniques designed to boost the accuracy of weapon detection. By incorporating attention mechanisms and Vision Transformers, we have created a robust and efficient solution for real-time weapon detection in video surveillance systems. Furthermore, we have integrated Explainable AI (XAI) methods to make our system more transparent and understandable. XAI techniques help us gain insights into the model’s decision-making process, allowing us to see how the model differentiates between weapons and non-weapon objects. This added layer of transparency makes the system not only more accurate but also more trustworthy.