Week 4 has come to a start and with that our five-week project has been going on quite some time. In the prior three weeks, we have done many things in order to achieve our actual goal: helping out in making 3d printed copies of paintings more realistic using eye-tracking software.
When it comes to having a proper functioning test set up we have come from quite a long way. Our first try in achieving a functioning eye-tracking set up has been shown below.
A camera will film the test subjects perspective to link the eyetrack data to.
This first set up has been improved by adding several frames. A frame for the painting, a frame for the test subject to rest their head, a rail-like frame to move the Arduino triggered spotlight on and a frame for the Tobii eye tracker itself are the main improvements of our test subject.
All these frames contribute to an easier calibration process when using the Tobii eye tracker. This calibration process can be quite tricky and cumbersome so having some fixed parameters to help simplify this process is a really nice thing.
After testing this newly improved set up in a new environment, the dark cellar room provided for us to use as a testing room, we soon found out several new improvements are necessary for our set up to function to its full potential.
An example of said improvements can be the clamps used to make sure the painting stays even better fixated in its frame.
One of our goals for the next week will be to try to improve as many of these details on our main set up. By doing these small iterations we hope to enable user-friendly eye tracking and receive results as efficiently as possible.
We also need to make the Arduino function properly because the motor keeps failing on us. To be able to look into the gloss on paintings properly we need a moving spotlight, achieving this properly will probably also keep us occupied for a few days.
Luckily for us we already managed to get a good amount of good work done, the results of this can be shown in a video which shows the process of tracking a painting using our set up.
The results of using our set up are already quite precise and can be visualized using the Tobii studio software in several ways.
Especially the Gaze Opacity (inverse heatmap) seems to be interesting. Using clusters or so-called areas of interest might also be useful to look into.
We plan to perform a pilot so we will be able to implement improvements on both our set up and our data visualization process based on the user experience of actual test subjects.
So that is what is next, really putting everything to the test and improving our already functioning test set up in the process.