A Deep Learning Approach to Parking Space Detection: Achieving High Accuracy and Scalability with FastAPI and Alexnet

A Deep Learning Approach to Parking Space Detection: Achieving High Accuracy and Scalability with FastAPI and Alexnet

연구 분야: Safety

학회: International Conference on Machine Learning and Soft Computing

The exponential growth in urban vehicle ownership has heightened the difficulty of finding available parking, leading to increased traffic congestion and fuel wastage. This research presents an innovative, cost-effective solution for real-time parking space monitoring by leveraging existing CCTV infrastructure and advanced deep learning techniques. Using PKLot and CNRPark datasets, the system achieves precise classification of parking spaces as occupied or vacant, eliminating the need for expensive sensor-based methods. A comprehensive evaluation of Convolutional Neural Network (CNN) architectures—AlexNet, VGG16, and ResNet50—identified AlexNet as the most effective model, achieving over 99.2% accuracy. To demonstrate its practical application, a web platform, parkHere!, was developed, integrating ReactJS and FastAPI to deliver real-time parking occupancy updates through a user-friendly interface. Google Lighthouse evaluations confirm the platform's exceptional performance, accessibility, and SEO compliance. This approach offers a scalable and accurate solution for urban parking management, utilizing existing infrastructure to enhance parking efficiency and improve the overall driver experience in urban areas.