Published Paper Details:

Urban parking inefficiency remains one of the most persistent challenges in modern cities, driving traffic congestion, fuel wastage, rising emissions, and daily frustration for millions of drivers. Earlier research established the socio-economic cost of poor parking management and explored solutions ranging from camera-based detection and IoT sensor networks to edge computing and deep learning models -- each addressing parts of the problem but none delivering a complete, reliable, and deployable solution. Building on this foundation, a unified AI-Driven Smart Parking Framework was designed and developed, integrating ESP32-CAM based video capture, YOLO-based vehicle detection, Region of Interest matching, Django REST API backend with MySQL database, a Flutter mobile application for driver-facing slot availability, and a Streamlit-based monitoring dashboard for administrators. This paper presents the real-world experimental evaluation of that framework, implemented and tested on an open-air parking facility covering parking slots under daytime conditions. Performance evaluation focuses on three key dimensions -- system response latency, mobile application update speed, and dashboard refresh rate. Results confirm that the system achieves an average end-to-end response latency of 1 to 2 seconds, mobile application slot status updates within 1 to 3 seconds, and real-time dashboard refresh under 3 seconds -- validating the practical viability of the proposed framework for smart city deployment. The findings demonstrate that the convergence of computer vision, IoT sensing, and edge computing can successfully transform a conventional parking facility into an intelligent, data-driven, and user-friendly system.