The vehicle detection issue is of great interest to modern transportation and automation technologies, which bring influence onto applications ranging from autonomous driving to smart traffic management. Accurate and prompt vehicle detection will exert significant effects on the improvement of road safety, higher flows, and materialization of advanced driver-assistance systems. However, most traditional vehicle-detection methods refer to the variations in environmental conditions, vehicle type, and real-time processing. Such complexities are difficult to master using traditional techniques, often being prone to various factors of imprecision, low speed, and poor robustness. In this work, we overcome such challenges by the proposition of a YOLOv5-based vehicle detection model intended for transformation systems and in developing the state-of-the-art object detection framework that maintains high speed and accuracy in overcoming shortcomings innate in conventional models. This will include a number of enhancements in both network architectures and post-processing methods to optimize detector performance across these changing scenarios.