This paper presents the results of a real-world application and evaluation of an AI-based Smart Video Solution (SVS). SVS integrates with existing infrastructure camera networks, prioritizing privacy and ethical standards, and utilizes pose-based data to perform AI tasks such as anomaly detection. It provides real-time alerts via cloud-based infrastructure and a mobile app, and utilizes innovative data representation and visualization techniques, such as occupancy indicators, statistical anomaly detection, bird's-eye views, and heatmaps, to understand pedestrian behavior and enhance public safety. SVS is deployed and evaluated using 16 cameras in a community college environment, demonstrating the robustness of a system that integrates AI-based visual processing, statistical analysis, database management, cloud communication, and user notification. Specifically, we validate the system's real-time performance by evaluating the end-to-end latency from anomaly detection to notification to stakeholders. For 21 hours, the system efficiently managed 16 CCTV cameras with a consistent throughput of 16.5 frames per second (FPS), with an average end-to-end latency of 26.76 seconds.
