Theses and Dissertations

Advisor

Ball, John E.

Committee Member

Dabbiru, Lalitha

Committee Member

Liu, Chun-Hung

Committee Member

Price, Stanton R.

Date of Degree

5-16-2025

Original embargo terms

Visible MSU Only 1 year

Document Type

Graduate Thesis - Campus Access Only

Major

Electrical and Computer Engineering

Degree Name

Master of Science (M.S.)

College

James Worth Bagley College of Engineering

Department

Department of Electrical and Computer Engineering

Abstract

In autonomous driving, utilizing deep learning models to help make decisions has become a popular theme, particularly in the realm of computer vision. Deep learning models are heavily influenced to make decisions based on the environments in which they are trained. Currently, very few datasets exist for off-road autonomy that include visual semantic segmentation labels. Traditional semantic segmentation requires hand-labeling large imagery datasets or using synthetically generated imagery to train a model. This study aims to apply transfer learning techniques to automatically label a new off-road dataset. That objective will be accomplished in two phases: first, by utilizing pre-existing labels of published off-road datasets, and second, by creating new labels derived from a well-performing model and a small subset of hand-labeled imagery. Additionally, this thesis investigates a proposed advancement to the auto-labeling pipeline using features extracted from auto-encoders.

Download

Share

COinS