Ensuring safety in public places has become increasingly important, and surveillance cameras are widely used to monitor activities. However, continuously observing CCTV footage manually is a difficult and time-consuming task, and it may result in delayed responses to violent incidents. To overcome this problem, this project presents an automated violence detection system that uses deep learning techniques to identify violent activities in surveillance videos.In the proposed approach, the YOLO (You Only Look Once) algorithm is applied to detect human objects within video frames. After detecting the individuals, a Convolutional Neural Network (CNN) is used to analyze their actions and determine whether the activity is violent or non-violent. The model is trained and tested using well-known datasets such as the Real Life Violence Situations (RLVS), RWF-2000, and the Hockey Fight Dataset to improve detection capability.Whenever the system identifies a violent event, it automatically records the time of occurrence and stores the corresponding frame image as evidence of the incident. The effectiveness of the system is assessed using evaluation metrics including accuracy, precision, recall, and F1-score.The proposed solution can support intelligent surveillance systems by providing faster identification of violent behavior, thereby helping authorities respond quickly and enhance overall public safety.
IJERST (Mon,) studied this question.