Robust Object Detection for Autonomous Vehicles on Indian Roads Using YOLOv8

Abstract The creation of reliable and effective object detection systems is essential in the current era of quickly developing technology to improve road safety and facilitate the broad use of autonomous vehicles. The design, implementation, and assessment of an object detection system using the cutting-edge YOLOv8 model are the main topics of this paper. With the use of deep learning methods and extensive datasets, especially the Indian Driving Dataset (IDD), this system tackles the challenges of identifying different objects in dynamic road situations, including cars, pedestrians, and traffic signs. Because of its excellent accuracy and real‑time performance, which are necessary for realistic deployment in traffic management systems, the YOLOv8 model was selected. To develop the post-processing Indian driving-based system, several preprocessing processes, dataset annotation, and model training are carried out. The results exhibit the suitability of the system, showcasing high accuracy in object detection across diverse scenarios. The robustness and dependability of the model are confirmed by quantitative measures including mean average precision, precision, recall, and F1‑score in addition to qualitative evaluations through visual inspections.