Document Type
Thesis
Degree
Master of Science
Major
Computer Science
Date of Defense
4-27-2022
Graduate Advisor
Badri Adhikari
Committee
Badri Adhikari
Sharlee Climer
Mark Hauschild
Abstract
Eye-tracking can be valuable for researchers in many domains. Most eye-tracking technologies require an extra piece of costly hardware. Several other available eye-tracking solutions are usually not very accurate and require a costly subscription. Our project was oriented at creating a free and open-source alternative that does not require additional equipment. We developed a deep learning-based solution as a prototype for this project. Specifically, we developed a deep learning model to predict a user’s gaze position on the screen. We created our training data set using a commercially available eye-tracker to train the model. Each training sample consists of a video frame of a person looking at the screen (input feature) and the corresponding true gaze location on the screen (output labels). When using a model with a gaze location output on a 2d plane, a deep learning model was able to in most cases interpret and predict a correct gaze position. Our results demonstrate that with minimal processing a deep learning algorithm can be effectively used to create and deploy eye tracking solutions.
Recommended Citation
Trenter, Sam, "Eye-Tracking using Deep Learning" (2022). Theses. 406.
https://irl.umsl.edu/thesis/406