Violence Detection Over Online Social Networks using YOLOV5 and SVM
Abas Kandoh Saber, Ibraheem Nadher Ibraheem, Methaq Talib Gaata · 2023
7.9 fatality per 10,000 persons worldwide are caused by acts of human aggression on a yearly average basis. The majority of human violence occurs suddenly or in remote locations. Stopping such acts is severely hampered by the information delay in this case. The detection method is employed in this work to focus on this problem. One of the most efficient computer vision algorithms is the one for moving object detection from CCTV. There are currently CCTV cameras on each street, which is incredibly beneficial for solving cases. For predicting and detecting action and video-based features, computer vision uses certain deep learning (DL) methods. Police arrive in violent locations in real time, check the CCTV footage, and begin their investigation. This investigation is specifically intended to identify aggressive behavior captured on CCTV. The research consists of four DL models that are utilized to create a system for detecting violent acts in videos using the YOLOV5 algorithm and Support Vector Machine (SVM) classification models. This model is real-time-capable. The research's findings that the suggested model achieves 99% accuracy.