Abstract
Artificial intelligence (AI) has played a crucial role in ensuring safer roads through autonomous driving. It has contributed by providing various models and frameworks for optimizing the performance of autonomous driving systems. One important aspect that helps in achieving safer roads is the consideration of the surrounding scenery. This thesis focuses on investigating the opinions of drivers regarding the scenery and their perception of autonomous driving. Additionally, it explores the training of YOLO v8 on a real-life dataset of a challenging environment to enhance its performance in such settings. Furthermore, machine learning and continual learning techniques are compared on the CIFAR10 dataset, specifically in relation to stationary elements that are part of the scenery.
AI has made significant contributions to road safety through autonomous driving. The survey results indicate that while apprehensions exist, people are hopeful that autonomous vehicles can enhance road safety. Moreover, the evaluation of different models and frameworks on challenging environments and stationary elements of the scenery sheds light on their respective performance and training times.
The findings of this research suggest that while people may have concerns about fully automated driving vehicles, they are open to the idea that they can contribute to safer roads. Moreover, the results demonstrate that YOLO v8 trained on the custom challenging dataset achieved a mean average precision (mAP) of 24.1%, a precision of 45.8%, and a recall of 20.3%.
On the CIFAR10 dataset, both Agnostic-Replay and YOLOv8 exhibited higher accuracy compared to AlexNet architecture and Avalanche on Naive strategy, while AlexNet and Avalanche showed the lowest losses. Additionally, the training time of the continual learning frameworks, Agnostic-Replay and Avalanche, was lower than that of the machine learning framework, AlexNet, and YOLO v8. These findings can aid in the further development and optimization of autonomous driving systems to make roads even safer.