IoT Based Weapons Detection System for Surveillance and Security Using YOLOV4

摘要

Due to the increase in crime and terrorism in most parts of the world, security surveillance is becoming increasingly important. A computer vision-based system for detecting weapons for real-time security surveillance is designed in this work. For identification, detection, and notifying the appropriate authorities, the system employs the YOLO V4 (You Only Look Once) algorithm. This neural network can be trained using images, videos, and live streaming videos. This model incorporates Internet-of-Things (IoT) smart devices that are interconnected and automated in weapon detection. This model's accuracy varies depending on the quality of the images and videos used in the detection process. Here, the proposed research work has discovered that the detection process is affected by the type of hardware that has been utilized to run the algorithm, ranging from low-quality image/video detection with 70% accuracy to high-quality image/video detection with 95% accuracy.