Real Time Violence Alert System Using Densenet Algorithm

Abstract

This Project Introduces A Real-Time Violence Detection System Using Deep Learning To Analyze Video Footage And Live CCTV Streams, Enhancing Public Safety Through Automated Surveillance. Traditional Monitoring Relies On Human Oversight, Limiting Its Effectiveness In Large-scale Environments. In Phase I, The System Uses CNN To Process Recorded Video For Frame-level Violence Detection And Individual Identification. Due To CNN’s Limitations In Speed And Feature Extraction, Phase II Integrates DenseNet For Real-time Detection, Which Proved Significantly Faster And More Accurate Through Comparative Analysis. DenseNet’s Dense Connectivity Enables Better Feature Reuse And Improves Overall Performance. Implemented In MATLAB, The System Provides Real-time Alerts, Offering A Scalable And Effective Solution For Modern Security And Public Safety Applications.

Keywords