Deep learning model comparisons for the identification of weapons

Abstract

Object detection plays a vital role in the field of scene interpretation. Various deep learning approaches like CNN, RNN give good results in the detection and recognition of objects. In this paper, we compare the models SSD, Faster RCNN, and YOLO for detecting weapons as objects in the scene, like guns and knives and person. The proposed method will compare the results of all these models in terms of accuracy, training time, and testing time with a labelled knife and gun dataset. The results are shown in tabulated form for all the models based on speed and accuracy. The average accuracy of the SSD, Faster RCNN, and YOLO models is 78%, 94%, and 92%, respectively, but the speed of SSD outperforms Faster RCNN and Yolo