Weapon Detection Alerting System

Abstract

In today’s world, more crimes are happening, so people fear their safety and security. Mostly, gun violence is categorized as the most highly anticipated violence around the world as it is growing rapidly. Closed Circuit Television (CCTV) cameras are used in many areas to monitor activities, but still we need human oversight and involvement. We require a technology that is capable of automatically identifying these criminal activities. This project focuses on providing a secure place using CCTV footage as a source to detect harmful weapons like guns by using deep learning algorithm. Therefore, weapon detection is a primary requirement in the current world and our project presents automatic weapon detection using webcam identifying weapons using Convolutional Neural Networks (CNN). We implemented YOLO “You Only Look Once” V4 object detection model by training it on our customized dataset. The training outcomes show that YOLO V4 outperforms YOLO V3 in speed and accuracy. Applying this model to our surveillance system in smaller areas, we may save a person's life, which may reduce the crime rate. Additionally, our proposed system also alerts the admin by sending email with the captured image of the weapon, if the weapon is detected.