Weapon detection from camera footages using YOLO and SSD models

Abstract

Detection using a convolutional neural network (CNN) based SSD mobile network YOLOV3 and the YOLOV4 algorithm. The purpose behind using three models was to compare between their accuracy and investigate their potential use and suitability in real-time environment. The results for the three models were good. However, in term of accuracy, YOLOV4 showed more promising results followed by the SSD model then theYOLOV3. Even though, the accuracy is not the only criterion to consider in real-world applications requiring short latency and high speed. Therefore, taking this tradeoff between accuracy and speed the YOLOV4 model seems to be the most suitable.