The Eye Tribe

When the eyetracker is not calibrated, the "raw" coordinates returned in a "frame" are all zero. Would it be possible to get, as "raw", the distance vector between the pupil and the corneal reflection, possibly in units of camera pixels? This way one could implement one's own calibration.

Currently this feature is not on the roadmap as it would only benefit a handful of customers who has the time and knowledge of rolling their own calibration routine. Besides it would be difficult for us to provide any level of support for such customization.

The calibration routines are improved over time so to get best performance always run the latest version.

Hi Martin,
Given the cogency of your arguments this is probably a decision of the marketing department - sigh -, so I won't even start arguing on a more factual level.
Thanks anyway for responding.

Dear Martin,

I would also be much interested in the raw uncalibrated signal. I fully agree that only a handful of developers would use that information. But once this handful of developers publishes their calibration algorithms, a much wider audience can benefit from them. In academia, this is happening all the time. And any improvement coming out of this process would add value to your device, without costing you a thing.

In our case, for example, we have our own calibration and analysis software ready and in use with a single very expensive tracker of another manufacturer. Given the low cost of the EyeTribe tracker we would like to equip all of our experimental computers one. But as it is now, the resources needed to adapt our software to use the data the EyeTribe tracker hands us are too high. And there are probably hundreds of psychology/vision research labs who cannot afford expensive trackers, but would buy one or more EyeTribe trackers if they had a possibility of knowing what was going on "under the hood" and could possibly change that by using home-cooked calibrations - or others' calibrations that are considered best for the respective purpose in their field/discipline.

This seems like a lot of advantages, for what seems a trivial, low cost thing for your developers to implement. Above all, it makes your device usable/more interesting for a new user group, i.e., academia. I get your argument concerning customer support, of course. But I also wouldn't expect Apple customer support to solve my problems with a Windows installation on my MacBook

If you would find the time to respond it would be greatly appreciated. All the best, Marc

I would also be interested in having a raw signal available without calibration.

We see the potential to use the eye tribe in the diagnosis of eye movement and attention disorders, and it would be particularly useful in children (because the cameras are not mounted on the head). Examinations in these cases should be done very quickly, and since we can calibrate a raw signal after collecting the data, pre-calibration is not necessary. In many patients, particularly children, attention and willingness to participate can disappear quickly, so any reduction in the testing time is beneficial.

Hi,
I am also working with children and I am planning a study with stroke patients. Reducing calibration time to an absolute minimum is a requirement for the success of any testing setup. I would like to know a few things:
- is it possible to reduce the number of calibration points if e.g. only looking left or looking right is required for a task?
- is it possible to repeat a slection of the calibration points if calibration failed?
Thank you!
Kind regards,
Jessica

Hi Jessica,

as far as I know you need at least 7 points to do a calibration (I haven't check the source code if that validation can easily be bypassed, or if it has a more solid argument to be there rather than experimental testing on average error in calibration for fully usage of the screen).

for your second request, the answer is YES and I suggest you to look at this post for further details.

Chris Bockisch wrote:I would also be interested in having a raw signal available without calibration.

We see the potential to use the eye tribe in the diagnosis of eye movement and attention disorders, and it would be particularly useful in children (because the cameras are not mounted on the head). Examinations in these cases should be done very quickly, and since we can calibrate a raw signal after collecting the data, pre-calibration is not necessary. In many patients, particularly children, attention and willingness to participate can disappear quickly, so any reduction in the testing time is beneficial.

+1. It would be great to have access to uncalibrated data.