An Automated Human Violence Detection for Surveillance using Deep Learning

Abstract: In recent years, ensuring public safety in crowded environments such as transportation hubs, campuses, and urban areas has become increasingly challenging due to the rise in violent incidents. Traditional surveillance systems rely on continuous human observation, which is inefficient, time-consuming, and prone to errors. To address these issues, this project proposes an AI-based automated surveillance system capable of detecting violent human activities in real time using deep learning techniques. The proposed system utilizes the YOLOv8 Nano model developed by Ultralytics along with a custom-trained dataset to identify violent behavior directly from video streams captured through live webcams (CCTV cameras). The YOLO-based approach performs fast, single stage object detection, enabling real-time monitoring with lower computational requirements. By learning spatial patterns associated with aggressive actions, the system can effectively differentiate between violent and non-violent activities even in dynamic and crowded environments. The developed framework is integrated with a monitoring interface that generates instant alerts whenever violent behavior is detected, thereby assisting authorities in taking timely action. Experimental results indicate that the system achieves reliable detection speed and satisfactory accuracy, making it suitable for deployment in smart surveillance systems, institutional security setups, and public safety monitoring applications [1].

Keywords: Artificial Intelligence, Machine Learning, Deep Learning, Computer Vision, Object Detection, Human Activity Recognition, Violence Detection, Smart Surveillance, Real-time Monitoring, Security Automation, Public Safety System, Emergency Alert System, Video Analytics, Image Processing, Feature Extraction, Convolutional Neural Networks, YOLO Algorithm, Dataset Training, Model Evaluation, Accuracy Optimization, Intelligent Monitoring, AI- based Security, Threat Detection, Smart City Safety, Automated Surveillance.