Parking Space Detection in Different Weather Conditions Based on YOLOv5 Method

The increasing number of vehicles on the road has led to a serious problem of finding available parking spaces during rush hour. Previous works used the classifier method to classify empty or occupied parking spaces. Other studies used object detection algorithms to detect the parking spaces and show their location. However, prior studies have not demonstrated the efficiency of parking space detection in various weather conditions. In this paper, we experiment with an object detection method to detect parking space using the You Only Look Once version 5 (YOLOv5). This study used four out of nine cameras in the CNRPark dataset that include different weather conditions (overcast, rainy, and sunny). After splitting the datasets, the training and validation data were trained using six configurations of YOLOv5 architecture. We evaluated the result of testing data using six weights from the training process. The results show that the method achieved the best mean average precision (mAP0.5) score of 0.969 in rainy weather using the best model of YOLOv5 configurations. Furthermore, this study compared the accuracy of parking slot detection with previous studies. Our experiment provides an effective solution for parking space detection in various weather circumstances.