This research introduces and evaluates an advanced AI-enabled Smart Video Surveillance (SVS) system designed to enhance safety across community spaces such as educational institutions, recreational areas, parking lots, small businesses, and in broader smart city applications. Our proposed system seamlessly integrates with existing Closed-Circuit Television (CCTV) systems and wired camera networks, making it easy to adopt and capitalizing on recent AI advancements. It uniquely employs metadata instead of pixel data for activity recognition to maintain privacy, adhering to stringent ethical standards. The SVS system features a cloud-based infrastructure and a mobile app for real-time, privacy-conscious alerts within communities.In our comprehensive evaluation, conducted in a community college environment, we delve into AI-driven visual processing, statistical analysis, database management, cloud communications, and user notifications. The system was tested using sixteen CCTV cameras, achieving a consistent throughput of 16.5 frames per second over 21 hours with an average end-to-end latency of 26.76 seconds for detecting behavioral anomalies and alerting users. We also explore sophisticated data representation and visualization techniques such as Occupancy Indicators, Statistical Anomaly Detection, and Bird's Eye View. These tools help analyze pedestrian behaviors and enhance safety, offering intuitive visualizations and actionable insights for stakeholders like law enforcement, urban planners, and social scientists. The findings underscore the vital role of visualizing AI surveillance data in emergency management, public health, crowd control, resource distribution, predictive modeling, city planning, and informed decision-making. This pioneering work is the first to examine the performance of a physical-cyber-physical anomaly detection system, crucial for identifying potential safety incidents and guiding urban development.