Paper Title
A Vision based AI-Powered CCTV Surveillance System

Abstract
At present, the majority of urban closed-circuit television (CCTV) systems depend primarily upon slow and inefficient human observers to monitor for threats in real-time. The current submission introduces a new type of AI surveillance that integrates weapon detection, crowd anomaly detection and face recognition. The system uses video compared with a database of known offenders to immediately identify suspicious behavior or future possible threats that may occur. A log will be created for each detected event, with a link to the detected event, as well as providing an interactive web-based dashboard to generate instant alerts to applicable law enforcement agencies. Testing on actual urban CCTV has demonstrated the system's ability to identify with high accuracy and very low latency, thus it is ready for deployment in most urban areas. Keywords - AI-Powered Surveillance, Weapon Detection, Crowd Anomaly Detection, Face Recognition, Computer Vision, Real-Time CCTV Analytics