Last friday (06-10-2017) we started building frames for a proper test setup to enable easier and less time-consuming calibration of the Tobii eye tracking device for different test subjects with, for example, varying heights.
The first object we build for this purpose was a frame to hold the scanning object, the painting. This frame is supposed to look like a large easel. A foam board, passe-partout like construction around the actual painting functions as a proper projection surface to project the calibration grid on. The thickness of the foam ensures that the painting stays at the same level as the projected calibration grid which makes the tracking process even more precise.
The easel including the passe-partout painting frame is displayed in figure 1.
To make sure the painting would not fall out of the frame we 3D-printed some parts to fix the painting to the frame. These parts connect to the easel and are shown in figure 1c.
We also made a frame for the test participants to rest their heads on, by doing this we try to make the testing experience as comfortable as possible and to have a set a fixed position for the eyes of each test subject. This set eye position simplifies a great amount of calibration parameters.
The frame to enable fixed eye position is displayed in figure 2.
The Tobii eye tracking device and the projector used for calibration purposes need to be standing on a frame as well. For this purpose we fixed an existing frame in the right position. This worked out fine so there was no need for us to build an extra frame for these devices.
The frame on which the Tobii eye tracker and the beamer will be located is displayed in figure 3.
For her studies, Willemijn, our expert, has a dark room to her disposal and we are allowed to use this room aswell, so we adjusted our set-up to fit this room properly.
At first everything was a little narrowly spaced, as is displayed in figure 4.
So we made some improvements for the sake of the test participants and our own comfort. These improvements are shown in figure 5.
In all our test set-ups a Canon EOS 1300d camera will be used to film the test subjects perspective. This camera will be supported by its matching tripod.
As told before, we need a moving light source shining on a painting to show of the glossy effects properly. For this purpose we build rails, using an Arduino to move a spotlight we will try to get some proper tests done.
The rails on which the arduino will transport the spotlight and the case, train shaped, in which the arduino and the spotlight will be placed, are shown in figure 6.
When building the track, it became apparent that the motor would not cause enough friction between itself and the aluminium tubes in the current set-up to propel the train forward. The bottom of the train was altered to fit the motor sideways instead of through a hole in the middle of the bottom as was planned before. We also added wheels to the bottom of the train to limit friction between the spotlight case and the track and provide a more fluid movement of the light source by doing so. This improved train design is shown in figure 7.
In theory, the Arduino code was correct and with a push of a button, the motor should start turning, but it did not… The shield was probably not wired up properly so that needed an extra look.
For now, a couple of tests could be conducted using a studio like theater light Willemijn found us. In this case, one of us would just hold the light and manually move it around. One advantage of the studio light was its ability to produce different colors of light. We discovered that blue light hid the image on the painting and only showed the glossy parts. This could be very helpful when examining gloss on a painting or a reproduction.
To test our set-up, we tested ourselves when looking at the real painting without any clear instructions. The calibration process again was a bit of a hustle on the first try as can be seen in figure 8, but when trying a second time everything went as expected. The results of the three of us were fairly similar and to illustrate them, two of the videos are presented below using the inverse heatmap option from Tobii studio.
Both Ronja and Jerome were mostly focused on the three flowers in the middle left of the painting, even if they tried to fight the urge of looking at those flowers. This could be due to the fact that there is a big contrast in gloss between those flowers.
This week has been a success because our set-up works, hower, the main target of our research was still a bit fuzzy to us. Gladly, Willemijn could supply some feedback on friday (13-10-2017). First of all, she was surprised of what we had managed to get done in a week so this had some impact on her planning. Sadly, also for her, the research question was a bit fuzzy. It would be best if we just explored the possibilities of the eye tracker and our set-up even further.
To do this we plan to perform a few additional tests next week. The original painting is going to be compared to some reproductions with different levels of gloss in a pilot using different test subjects.
Sadly, probably due to the dark room, the resolution of the camera is not that good. It would be nice to find a way to improve this, so we will also try to look into that next week. Apart from making them look nicer, this will improve the precision of the visualisation of the test results.
In short, the planning for next week will mostly be conducting a pilot (with an optional survey). The eyes of the participants will be tracked when looking at the original painting and reproductions will different levels of gloss. Apart from the data from the Tobii, the opinions and experiences of the participants will be taken into consideration as well. As an extra experiment, the frame around the painting will be examined. What is the psychological impact of this frame? (maybe use different colors). To replicate a museum experience, a different light source is bought that fits on the train. This train should be up and russing by the start of next week.