Intelligent Video Surveillance Based on YOLO: A Comparative Study

Abstract

Number of Research study show that handheld gun is the primary weapon used for various crimes such as shoplifting, robbery, and rape. We can reduce these crimes by identifying the behavior at early stage which can significantly reduce collateral damage. Object detection using Deep neural networks is difficult when the objects are small such as handheld-Weapons. The object detection is different from classification, which recognize objects but doesn’t localize object in image. classification don’t work on multiple objects in single images. To address this problem in this study, we aim to detect small objects which is a handheld pistol using different Recent state of art model i.e YOLO and compare them Specifically for Handheld weapon Detection. The Experimental Results shows that The YOLOV4 model Achieves [email protected] of 98.79%, Precision 91%, Recall 99% when detecting Handheld weapon (pistol). YOLOv3 Model Achieves [email protected] of 90.37%,Precision 93% and Recall 80% demonstrating that it can accurately detect Handheld Weapon (Pistol) insurveillance.

Topics & Concepts