Where Should I Park?

Real-time Detection of Parking Space Occupancy

Abstract

In the face of escalating urbanization and its associated parking challenges, this paper proposes an innovative approach to enhance parking lot management and optimize space utilization, thereby mitigating urban parking difficulties. We initially experimented with the Yolo-Fastest V2 algorithm for real-time parking space detection. However, we observed that Yolo-Fastest V2 had shortcomings, particularly in terms of incomplete detection accuracy. This led us to develop a more robust approach using a modified ResNet model with transfer learning. This approach, diverging from traditional sensor based methods, offers improved accuracy and cost-efficiency with less data need. Our comparative analysis between ResNet and YOLOv8 demonstrates the former’s superior adaptability and performance in urban settings. This development significantly contributes to the smart city framework, improving parking efficiency and potentially easing traffic congestion.