Publication

Abstract : Safe navigation in autonomous vehicles (AVs) demands a robust perception system capable of real-time object detection. This study explores the application of deep learning, specifically You Only Look Once (YOLO) methods, to achieve this critical functionality. The research study investigates various YOLO architectures, balancing the trade-off between speed and accuracy essential for real-time AV operation. The research work tackles object detection, encompassing not only general object recognition (using publicly available COCO and KITTI datasets) but also lane marking and traffic sign detection (leveraging a custom dataset). The research addresses the challenges posed by dynamic environments, varying lighting conditions, and small object identification. Techniques like data augmentation are explored to improve the robustness of the models. Furthermore, the potential for multi-sensor fusion is considered as a means to enhance perception capabilities. This project demonstrates the effectiveness of YOLO-based deep learning for real-time multitask object detection in AVs.