An Efficient approach for Firearms Detection using Machine Learning

Abstract

Each year, there is a significant number of people impacted by gun-related violence globally. To address this issue, we have created a computer-based system that can automatically identify firearms, specifically pistol. Recent advancements in machine learning has shown success in the fields of recognition and object detection. Our system utilizes the You Only Look Once (YOLO V3) object detection model, which was trained on a personalized dataset. Our training results indicate that YOLO V3 outperforms both traditional convolutional neural network (CNN) models and YOLO V2. Notably, our approach did not require high computation resources or intensive GPUs to train our model. By incorporating this YOLO V3 model into our security system, we hope to rescue lives and decrease the occurrence of manslaughter or mass killings. Moreover, detecting weapons or other dangerous materials and preventing harm or risk to human life could be accomplished by integrating this system into sophisticated surveillance and security robots.