Got some exciting stuff for you that we´ve been working insanely hard on during the last couple of weeks.
The ITU Gaze Tracker is an open-source gaze tracking application that aims to provide a low-cost alternative to commercial gaze tracking systems and to make this technology more accessible. It is being developed by the Gaze Group at the IT University of Copenhagen, supported by the Communication by Gaze Interaction Association (COGAIN).The eye tracking software is video-based, and any camera equipped with infrared nightvision can be used, such as a video camera or a webcam. The cameras that have been tested with the system can be found in our forum.
* Supports head mounted and remote setups
* Tracks both pupil and glints (IR-reflections)
* Supports a wide variety of camera devices
* Configurable calibration
* Eye-mouse capabilities
* UDPServer broadcasting gaze data
* Full source code provided
We encourage users and developers to test our software with their cameras and provide feedback so we can continue development. The ITU Gaze Tracker is released under the GLP3 open source license and the full source code is hosted at sourceforge. It´s written in C# using Emgu OpenCV wrapper for C++ image processing. (Microsoft .Net 3.5 needed) Once the tracker has been started it can be configured to broadcast gaze data via the UDP protocol which makes it easy to pick up in your own applications. We provide a sample implementation on a client in C#.
Nice one Martin, I knew that you were going to come up with something nice when you told me you would be going to Copenhagen.
This will really benefit disabled people, who can not use their hands, feet or tongue to control the mouse cursor.
The IR illumination is a valid concern, it really depends on the setup. In optimal conditions the IR LED´s used in a typical “night-mode” for webcams or HD cams is enough. However, if you use it in a situation where there is a lot of daylight from windows etc. you might want to increase the amount of IR light to get a good distinction of the pupil. This is a factor that is present in all eye tracking devices and most of them has been certified or approved by regulating committees. From the word-of-mouth perspective I have been told that a work-day long session exposes they eye of IR-levels that are typical from spending a day in the sun.
A detailed analysis of the risks and possible safety issues can be found in the COGAIN report
D5.4 Exploration of safety issues in Eyetracking
Mulvey, F., Villanueva, A., Sliney, D., Lange, R., Cotmore, S., Donegan, M. (2008) D5.4 Exploration of safety issues in Eyetracking. Communication by Gaze Interaction (COGAIN), IST-2003-511598: Deliverable 5.4.
A truly a great contribution to the open scientific community and emerging HCI research… really looking forward to experimenting with this and am sure it will help many of our community members get started with another fun form of human sensing
Guys, thanks for the kind words.
Just wondering if anyone had tried the application yet? We have launched the forum.gazegroup.org to get a list of verified cameras up. As always there has been some bugs that we are fixing. Would appreciated you posting if there has been any issues.
I have a new student at one of my schools who I think could benefit from the ITU GazeTracker. Commercial systems, including the computer, cost about $15,000.00. I know of a student who might need nearly $30,000.00 to get what is needed. Do you think that this could be a good summer project for upper-level high school or community college technology students?
Sounds like a great idea, I’ve just posted a brief users guide that hopefully will get people started. The software has been written on a Microsoft platform but it offers the option of sending the gaze coordinates via the UDP network protocol which is really easy to pick up in external applications. We also offer the option to redirect the mouse cursor to the gaze position which may be used to activate built-in functionality of regular interface components (such as OnMouseEnter/Over etc.)
However, please note that the current version (really a public beta) is not head position invariant, e.g head movements will affect the gaze position, so a comfortable seating position is required. Other issues are sensitivity against direct or strong sunlight as it may interfere with the infra red illumination used to produce a corneal reflection (this goes for most eye trackers)
Give it a try, all you need is a $30 night-mode web camera (verified list), let me know how it works.
We’ve made great progress since the initial release, today we open the doors for version 2.0. Internally we’ve rewritten major parts of the platform to gain flexibility and higher performance. First version was DIY playtime, this version is nothing short of a screamer. High performance, very accuracy tracking. People are telling us we are crazy giving it away but we’re dedicated to the mission: Accessible eye tracking for all, regardless of nationality and means. We’re making it happen.
Important highlights for GT2.0b:
- Supports three modes of operation, head-mounted, remote mono/binocular
- Vastly improved performance, +500fps head mounted, +170fps remote binocular (both eyes)
- Awesome accuracy, avg. 0.3 - 0.7 degrees of visual angle (remote binocular)
- New U.I, looks so.. 2010
- Automatic tuning (optimization of algorithms parameters)
- Relatively low CPU-utilization and memory footprint (12%, 170Mb, core i7 860 win7-64)
- Many enhancements, bug-fixes etc.