Autonomous Parking Space Detection for Electric Vehicles Based on Advanced Custom YOLOv5

This research contributes to the development and deployment of smart parking system utilizing cutting-edge computer vision techniques. The study utilizes a subset of the Google Open Images Dataset focused on parking spaces to train a Custom Object Detection model tailored for low-computation devices. Employing YOLOv5 architecture and the Feature Pyramid Network (FPN), the system effectively detects parking spaces and objects within them. Utilizing ResNetv250 as the backbone architecture ensures precise feature extraction, resulting in high accuracy in identifying parking spots. Experimental results reveal the efficacy of the envisioned methodology, with Efficient YOLOv5 model achieving a mean Average Precision (mAP) of 0.970 for object detection in aerial dimensional frames. Graphical representations of F1-Confidence Curve and Precision-Confidence Curve demonstrate the model’s robustness in balancing precision and recall, even with an imbalanced dataset. This research underscores the potential of the proposed smart parking system to revolutionize urban parking management, providing an efficient and accurate solution for parking space identification.