The project has two primary aims:
1. Analyzing the feasibility and accuracy of an eye-tracker which works using regular webcams and does not require special hardware such as infrared (IR) cameras, IR lights, etc.
2. Building a cheap, portable eye-tracker which can be used in a standalone fashion. For this purpose, the software is installed on a Raspberry Pi device and the resulting device has a 70 Euros building cost.
We build our work on the open source Opengazer application and we make several contributions to increase the accuracy and robustness of the base system (see the image above for an overview).
The average error of the system is 1.68 degrees (horizontally) and 1.43 degrees (vertically) in a standard experimental setup.
The code for the project is maintained on GitHub. The installation instructions are also provided for Ubuntu and Mac OS X systems.
For the Raspberry Pi eye-tracker, the ready-to-use SD card image for 8 GB (and larger) SD cards can be downloaded here:.
The usage instructions can be found here.
We provide the database of videos recorded during the project experiments. For more details, see our database description page
For any further comment or query, please do not hesitate to contact Onur Ferhat directly.