Object detection is a kind of computer vision application that is now an essential component of a wide variety of consumer applications. Some examples of these applications include surveillance and security systems, mobile text recognition, and the diagnosis of disease using MRI or CT scans. Detection of obstacles is another essential part to enable autonomous driving. The ability of autonomous cars to have a reliable and secure driving performance is dependent on their perception of the world around them. This study introduces a YOLOv4-based, single-stage method for object detection that improves detection accuracy and enables real-time operation. It is crucial for the safe operation of an autonomous vehicle to be able to identify and follow nearby objects. Object detection, classification, and tracking algorithms are presented in this study. Everything is sorted into several categories based on whether it is in motion or not, and what it is (car, person, or something else). To detect and classify objects and estimate their position around the car, the proposed method combines data from a laser scanner with the state-of-the-art deep-learning network YOLO (You Only Look Once). Detecting obstacles is crucial for autonomous vehicles. Accurate and real-time detection of all roadside items is essential for the safe operation of high-speed vehicles. Recent years have seen a surge in interest in the question of how to strike a compromise between detection speed and precision. This work proposes a YOLOv4-based, single-stage approach for object detection that both increases detection precision and can run in real time.