Date of Award
5-1-2024
Degree Name
Master of Science
Department
Computer Science
First Advisor
Ahmed R, Dr. Khaled
Abstract
Federated learning(FL), a form of distributed machine learning, facilitates collaboration among multiple devices, such as autonomous vehicles in this context, aiming to enhance object detection accuracy without compromising individual data privacy. This approach allows vehicles to refine their models without divulging sensitive information like location, routes, or personal preferences. Training data remains on each device in FL, enabling model updates based on local data without transmitting raw data to a central repository. The refined models are aggregated on a central server, making an improved global model accessible to all participating devices. The BDDK100 dataset, comprising over 100,000 high-resolution images from video recordings captured by a front-facing camera on a moving vehicle, serves as a benchmark dataset for autonomous driving. This simulation study employs three simulated clients, each representing a distinct autonomous vehicle, engaged in collaborative training of a deep neural network for object detection through the FL paradigm. The YOLOV5 model is used as object detection model. In conclusion, this study demonstrates the efficacy of FL in the context of autonomous vehicle systems, specifically in object detection. The collaborative training approach, wherein autonomous vehicles collectively refine a deep neural network without compromising data privacy, holds promise for enhancing the overall accuracy and performance of object detection models. The utilization of the BDDK100 dataset underscores the practical applicability of the FL methodology in real-world scenarios, establishing it as a viable solution for improving machine learning models in decentralized environments. As autonomous technologies evolve, FL stands out as a privacy-preserving and collaborative method, contributing to advancing intelligent systems without sacrificing individual data security.
Access
This thesis is only available for download to the SIUC community. Current SIUC affiliates may also access this paper off campus by searching Dissertations & Theses @ Southern Illinois University Carbondale from ProQuest. Others should contact the interlibrary loan department of your local library or contact ProQuest's Dissertation Express service.