Real-Time detection Using Deep Learning Techniques for autonomous Vehicle

Autonomous driving heavily relies on object detection as a primary task. Ensuring both high accuracy and real-time detection is crucial for the object detection algorithm of autonomous vehicles. Consequently, the research in this area has garnered significant interest, becoming a prominent and trending topic in recent years. This study aims to find an algorithm that strikes a balance between speed and accuracy in object detection. In order to accomplish this, the latest object detection algorithms, including YOLOv5n, YOLOv5s, YOLOv5m, YOLOv5l, YOLOv5x, YOLOv7, YOLOv7-tiny, YOLOv7-X, and Faster R-CNN, are compared using the Self-Driving Car dataset. Through comprehensive experiments, the algorithms' performance is evaluated based on their inference time, precision, recall, and mean accuracy (mAP) values. The results demonstrate that Faster R-CNN stands out with a precision of 0.930, showcasing its excellence in accuracy. On the other hand, YOLOv5n boasts the fastest inference time, clocking in at a mere 7.5 ms. In summary, this research delves into the crucial aspect of object detection in autonomous driving, and its findings provide valuable insights into the trade-offs between speed and accuracy for different state-of-the-art algorithms, such as YOLO and Faster R-CNN, utilizing deep learning techniques. Keywords: autonomous driving, deep learning, object detection, You Only Look Once (YOLO), convolutional neural network.