Justin Boverhof
Pronouns: He/Him
Research Mentor(s): Odest Jenkins
Research Mentor School/College/Department: Robotics / Engineering
Program: UROPF
Session: Session 5 (2:40pm – 3:30pm)
Authors: Justin Boverhof, Jana Pavlasek, Odest Chadwicke Jenkins
Abstract
Introductory classes in robotics are something of a paradox. The course activities must capture student interest in compelling robotic tasks while only relying on technical skills within the scope of an introductory class. Robotics 102: Introduction to AI and Programming is an introductory programming course taught through the lens of robotics. This project explores methods and tools towards a final project for Robotics 102 involving image classification using neural networks. Previous iterations of the course have used a premade dataset for hand-written digit classification to train classification algorithms to complete a robotic task. However, results achieved on the handwriting dataset did not transfer effectively to real robot data. The question of this project is can we improve upon the student experience by changing how we do the final assignment? We set out to create a dataset inspired by autonomous driving tasks which is more relevant to the robotics community than handwriting detection and can act as a more direct lead into more advanced concepts. We analyze various algorithms to determine the most effective methods for this task.
