Automated Detection of Violence in Detached Areas using Hybrid Deep Learning Models: A YOLO-5 and CNN Approach

The human violence rate increases gradually every year and it’s increasing at a rate of 7% annually. It has been observed in many studies that violence takes place suddenly in detached areas. The major hurdle to stopping these acts is a delay in information exchange. To tackle these issues, automated detection techniques were employed. CCTV footage was employed to identify objects with computer vision terminology and actions were taken based on correctly identified objects. These CCTV cameras are extremely helpful to resolve such cases and take appropriate action. The violence and non-violent acts were identified based on the analysis of footage. Here, in this research, an effective and appropriate method was employed to detect the violent and non-violent acts from the CCTV footage. In this context, a deep learning model was proposed in the research that comprises two deep learning models especially YOLO-5 and Convolutional neural network (CNN). Here, initially, the images were preprocessed to enhance the quality of CCTV footage and afterward, YOLO was employed to identify objects from the available CCTV footage and later, the classification was done using CNN. The performance hybrid model was evaluated on different parameters and compared with other state-of-the-art techniques. After comparison, it has been observed that the proposed model outperformed with an accuracy rate of 98.63%.