OBJECT DETECTION FOR SELF DRIVING CARS

OBJECT DETECTION FOR SELF DRIVING CARS

TilakGowda T P1, Sandarsh Gowda M M2

1Student, Department of MCA, Bangalore Institute of Technology, Karnataka, India

2Professor, Department of MCA, Bangalore Institute of Technology, Karnataka, India

---------------------------------------------------------------------***---------------------------------------------------------------------

Abstract The integration of object detection systems in autonomous vehicles is pivotal for ensuring safety and efficiency in self-driving technologies. This study leverages the YOLOv3 (You Only Look Once) method, a state-of-the-art deep learning-based object detection technique, to enhance real-time object identification in autonomous driving. Despite its advantages, YOLOv3 has limitations concerning the initialization of bounding box dimensions. This research introduces a novel clustering method to improve this initialization, leading to faster convergence and increased detection accuracy. Using a comprehensive dataset from Kaggle, the proposed system demonstrates superior performance in metrics such as recall, mean average precision, and F1 score, marking a significant contribution to autonomous driving technology.

Keywords : Autonomous Vehicles, Object Detection, YOLOv3, Deep Learning, Real-Time Processing, Clustering Method, Bounding Box Initialization