Smart Parking System Based on Convolutional Neural Network Models

Most people have probably experienced the hassle of parking space scarceness. To alleviate this problem, the relatively traditional solution is making use of sensors, which has its drawbacks in terms of installation and maintenance cost especially if precise location is in consideration. In order to find a more affordable and functional solution, in this project, cameras are chosen for data input, and deep learning neural networks models are developed. Convolutional neural networks have achieved very good performance in the area of computer vision in recent years. We explore in this direction and propose VGG Extended Model and Mini Parking CNN Model. Testing is done to evaluate the time cost and recognition accuracy. Feasibility is prototyped, named as SmartParking. The Mini Parking CNN Model demonstrates excellent performance and this system built with it can help users to find empty parking space accurately and effectively.