Weapon Detection from Images using YOLO and OpenCV
Divyanshi Chitravanshi, Ayush Malik, Harsh Saini, Sandhya Avasthi, Kadambri Agarwal, Yash Grover · 2024
Gun violence incidents have sadly claimed many lives annually, making them a major global problem. A completely automated, computer-based approach for identifying popular weaponry, like rifles and pistols, is presented in this paper. Recent advances in deep learning and transfer learning technology have transformed object identification and recognition capabilities. Our suggested method makes use of the YOLO v5 (You Only Look Once) object detection model, which was trained on a private dataset made up of pictures of different types of guns. One important benefit of our methodology is that it incorporates transfer learning techniques, which removes the requirement for powerful GPUs or large amounts of processing power that are normally needed for training deep neural networks from the beginning. The outcomes show that the YOLO v5 model outperforms both conventional convolutional neural network models and its predecessor, YOLO v4. This concept could potentially help avoid killings and mass shootings by being integrated into monitoring systems, thus saving lives. Additionally, our model and methodology have the potential for the creation of robotic security devices that can identify lethal weapons and lessen the likelihood of an attack, improving public safety in high-risk regions. The potential consequences of properly and efficiently identifying firearms in real-time video feeds could be extensive, affecting public safety protocols, private security companies, and law enforcement organizations.