Graduation Semester and Year

Summer 2024

Document Type

Thesis

Degree Name

Master of Science in Computer Science

Department

Computer Science and Engineering

First Advisor

William Beksi

Second Advisor

Chris Conly

Third Advisor

Alex Dillhoff

Abstract

Sugarcane roots are not understood and previous methods of collecting and processing data have proved to be laborious and time consuming. Using Minirhizotrons, Researchers observe and photograph roots without disturbing the soil and are useful for studying root growth over time. Software such as Rhyzovision exists to allow quick processing of root images. These software tools require clean or well annotated images of only the roots to provide accurate information. Current annotations of the images are done manually and requires a Scientist with domain knowledge of roots to accurately annotate the root images. We are employing the use of CNN and Transformer deep learning models to learn from correct root annotations and create root segmentations that can provide context of root architectures in cases where roots are occluded.

Disciplines

Artificial Intelligence and Robotics | Other Computer Sciences

Available for download on Thursday, August 13, 2026

Share

COinS