Tobii Calibration

annieaitken

Hello,

I'm working on an eye-tracking study with infants and would like to tweak the tobii eye-tracking calibration protocol.

1) If possible, I would like to change the target dots to some kind of animated graphic with sound to attract attention
2) I would like the option to re-do calibration if it fails. As it currently stands, when calibration fails, it starts the experiment right away without collecting any eye-tracking data.

If anyone has any ideas for either of these questions, I would appreciate it!

Thanks!
Annie

eduard

1) If possible, I would like to change the target dots to some kind of animated graphic with sound to attract attention

Check out this discussion. The first part should get you started with your childfriendly calibration targets.

2) I would like the option to re-do calibration if it fails. As it currently stands, when calibration fails, it starts the experiment right away without collecting any eye-tracking data.

I never worked with Tobii, but I am surprised that there is no menu or something similar that you enter after calibration. As a workaround, you could put the calibration in a loop-sequence structure, together with a response item, and repeat the loop until a certain key was pressed. So, set the repeat value of that loop to a super high value (>1000), and add to the break_if field, something like [response] == 'c', if c was the key to continue.

Does that help?

Eduard

Art

Hi there,

Is there any way to obtain the pixel coordinates of calibration points used by Tobii calibration routine? Could someone explain to me the way this calibration routine works (I mean, especially how the values specified in lines 68-73 of libtobii.py are converted into pixel values). Thanks!

eduard

Hi,

Not sure about it, but it looks to me like these values are scalars that put the calibration targets at 10,50,90% of the width and height of the screen.

You can check the function _norm_2_px in lines 117-120. It does multiply the scalars with the display size for each combination of eccentricity.

Hope this helps.

Eduard

Art

Great, thanks!

Art

I discovered that the Tobii TX300 output x and y gaze values are not scaled by calibration gain and offset values. Did anyone encounter the same problem?

Workarounds would also be welcome - so far I could only thing of performing ad hoc calibration using values collected in each trial, knowing the fixation and the gaze target on-screen positions.

Art