Computer Vision Model Selection For Real-Time Tracking of Apples

Honors College Thesis

Public Deposited

Citeable URL: https://ir.library.oregonstate.edu/concern/honors_college_theses/1c18dp10f

Descriptions

Attribute Name	Values
Creator	Bloch, Peter J.
Abstract	This paper presents a method of implementing real-time apple detection for closed-loop control to approach apples for grabbing. The approach is to train two known real-time object detection networks–the Faster R-CNN and YOLOv5–on a novel dataset to verify that it is possible to achieve maximum average precision (mAP) above 50\% while computing position estimation inferences at a rate of at least 2-3 inferences per second. We found that the YOLOv5 detection network is most promising for real-time apple detection for closed loop control with a mAP of 90.2\% and detection speeds past 7 inference frames per second. While the Faster R-CNN network had similar accuracy, it was not able to complete 2 inferences per second, which does not meet the needs of our application. Key Words: Apples, Robotics, Computer Vision, YOLOv5
Contributor	Swenson, Nigel Velasquez, Alejo Grimm, Cindy
Resource Type	Honors College Thesis
Date Issued	2022-06-02
Degree Level	Bachelor's
Degree Name	Honors Bachelor of Science (H.B.S.)
Degree Field	Computer Science
Degree Grantor	Oregon State University
Commencement Year	2022
Advisor	Grimm, Cindy
Committee Member	Davidson, Joseph Swenson, Nigel
Non-Academic Affiliation	Oregon State University. Honors College
Rights Statement	In Copyright
Publisher	Oregon State University
Peer Reviewed	No
Language	English [eng]

Parents:

This work has no parents.

In Collection:

Thumbnail	Title	Date Uploaded	Visibility	Actions
	BlochPeter2022.pdf	2022-06-03	Public	Download