The rapid urban population growth has intensified the challenges associated with urban and suburban traffic, necessitating effective traffic control and management. The efficient movement of emergency vehicles, particularly ambulances and fire trucks, has emerged as a critical concern. This article presents the Vehicle Dataset, a comprehensive benchmark for object detection, encompassing seven vehicle classes, including cars, motorcycles, buses, trucks, vans, ambulances, and fire trucks. The dataset, includes 29,759 meticulously labeled images obtained from freely available online sources, enables the identification of traffic patterns through deep neural networks. Notably, the dataset emphasizes the facilitation of emergency vehicle movement. The Vehicle Dataset in this study is divided into three subsets, with 25,369 images assigned for training, 2,896 for validation, and 1,494 for testing. Through the utilization of the dataset, object detection algorithms based on YOLO versions 5, 6, and 7 have been trained. Remarkably, YOLO version 7 has yielded outstanding results, achieving a final precision of 85% and a mAP of 85% at an IoU threshold of 0.5. Moreover, at IoU thresholds ranging from 0.5 to 0.9, a mAP of 64% has been attained. The Vehicle Dataset represents significant resource for researchers and practitioners in the transportation and traffic management field. Its inclusion of emergency vehicles such as ambulances and fire trucks contribute to its unique value. This article presents a detailed exploration of the dataset, underscoring its significance in advancing object detection methodologies.
