Abstract

The objective of this thesis is to explore Deep Learning algorithms for classifying high-resolution images. While most deep learning algorithms focus on relatively low-resolution imagery (under 400×400 pixels), very high-resolution image classification poses unique challenges. These images occur in pathology and remote sensing, but here we focus on the classification of invasive plant species. We aimed to develop a computer vision system that can provide geo-coordinates of the locations of invasive plants by processing Google Map Street View images at using finite computational resources. We explore six methods for classifying these images and compare them. Our results could significantly impact the management of invasive plant species, which pose both economic and ecological threats.

Library of Congress Subject Headings

Neural networks (Computer science); Machine learning; Image processing--Digital techniques; Image analysis; Panoramas--Classification

Publication Date

5-2019

Document Type

Thesis

Student Type

Graduate

Degree Name

Computer Science (MS)

Department, Program, or Center

Computer Science (GCCIS)

Advisor

Christopher Kanan

Advisor/Committee Member

Thomas Kinsman

Advisor/Committee Member

Zack Butler

Campus

RIT – Main Campus

Plan Codes

COMPSCI-MS

Download

Share

COinS