Webcam-based eye tracker

Project Description

Build a platform that is capable of capturing, storing, (and possibly analyzing) eye-tracking data from people who interact with a website using only their webcam. Some motivation: large-scale experimentation on platforms such as labinthewild.org or mechanical turk have led to new insights about people because they can quickly access large, diverse groups of people and analyze their behavior. However, there is nothing similar in the context of eye-tracking or physiological measures. Recently, researchers at Brown have released browser-based (javascript) libraries that can track eye-location using only a browser and webcam (http://webgazer.cs.brown.edu). This opens the door for tracking, collecting, and analyzing gaze behavior at a massive scale for ANY web-based application.

Goals

– Create a platform that captures, verifies, and stores eye-tracking data provided by webgazer.cs.brown.edu. This should be able to be added to the code of an existing website.
– Secondary: Provide a web portal for basic analysis (for example, a heat map visualization where groups of people looked).

Impact

The platform itself would allow the collection of eye-tracking data at a scale which has never been done before. As a result, it would almost certainly be published in a high-impact venue. In addition, depending on how easy the software is to adopt, it would come as no surprise to me if the software was widely used by psychology and HCI researchers. Finally, the use of the platform could result in eye-tracking studies that, from a sample size perspective, have the potential to find some new, very exciting insights.

Constraints

– Data needs to be validated as being correct (An example: maybe a circle at coordinates 100, 50 flickers once every minute. If they look at 100, 50 then we know it’s correct.). Some people will simply be in settings that are not possible to attain reasonable data (too dark, bad webcam, etc.).
– User data should be kept secure and anonymous to preserve privacy (for example, we shouldn’t be storing any video at all).

Resources

TBD

Licensing

open-source software under the MIT license.

Point of Contact

Contact:

Prof. Evan Peck
Computer Science
Bucknell University
evan.peck@bucknell.edu