The Eye Tribe

Re: Calibration and multiple users

2015-09-17T18:15:03+02:00

Thank you very much =)

Statistics: Posted by skataNathalie — 17 Sep 2015, 18:15

Re: Calibration and multiple users

2015-09-16T16:38:43+02:00

It is a dark pupil tracker as opposed to on-axis bright pupil systems.

In general, the algorithms employ a number steps to find an ellipse of reasonable size and shape in the eye region.

Statistics: Posted by skataMartin — 16 Sep 2015, 16:38

Re: Calibration and multiple users

2015-09-08T13:05:24+02:00

Hello Martin

I also have a question to 1.
Can you tell me (in gernal) how the Pupils in the picture are recognized?
Because the camera of the TET sits in the middle of the TET - away from the LEDs - so there isn't on-axis lighting and consequently no bright-eye-effekt in the Pupils.

I work with the TET for my bachelor thesis and I need to describe the gernal functions of the TET in my thesis.
Thanks!

Statistics: Posted by skataNathalie — 08 Sep 2015, 13:05

Re: Calibration and multiple users

2015-01-13T15:25:03+02:00

Hi Jean,

As Jeff suggested it would be really helpful if you could send a calibration feedback report to us so we can look into this issue (the collected data is encrypted and never shared).

This function is found under the System tab in EyeTribeUI.

To your questions,

1. Correct, it searches for faces and eyes based on feature detection. Then it continues to search for the pupil and other eye features. Hard to diagnose the specific issue without more data. Hand drawn eyes will most likely no be detected.

2. The software is designed to work with any iris color, should not be an issue. What could be problematic is makeup and long eye lashes if they overlap. In general it takes a clear view of the eye with sufficient contrast to get a good signal.

Storing calibration profiles is on the way. Next version will automatically load up the latest calibration on start.

Statistics: Posted by skataMartin — 13 Jan 2015, 15:25

Re: Calibration and multiple users

2015-01-09T03:05:09+02:00

jean.lorenceau@upmc.fr wrote:
we encounter problems with eyes detection for some individuals

Be sure to send some feedback data on those attempts.
Something that abnormal could be provide some good information for future versions.

Statistics: Posted by skataJeffKang — 09 Jan 2015, 03:05

Re: Calibration and multiple users

2015-01-08T18:53:11+02:00

Hi,

Not sure this really adddress the point, but we encounter problems with eyes detection for some individuals: it may take minutes just to get the "green" eyes in the UI interface (and sometimes we fail to get any acceptable signals, so that calibration fails several times in a row!). Our physical configuration is very similar to all those I've seen in this forum so far, and the exactly same setup can work easily with someone and fail with another person. Very very annoying because we cannot figure out what is going on, and moving around the EyeTribe to try to catch the eyes is uneasy and provides little on line feedback (Red/Green background is poor information). So questions:
1. Is the EyeTribe algorithm detecting the eyes relying on any face processing algorithm?
2. Are some eye/face features detected more easily than others (iris color, eyebrow??) . What are these features ?
I tried to test the calibration with faked eyes (e.g. Drawing dark pupils on a sheet of paper), but could never manage to have these faked eyes detected ...
The new possiblity to store the calibration settings might help go aound this problem, but it is nevertheless a problem that Eye detection is unreliable and unpredictable ...

Please Help and provide advices and solutions, as we are about to use TheEyeTribe on a large population !
Thanks !
Jean

Statistics: Posted by skatajean.lorenceau@upmc.fr — 08 Jan 2015, 18:53

Re: Calibration and multiple users

2014-12-11T01:25:41+02:00

Hi Mikael,

Thanks for sharing your thoughts, I really appreciate it and agree on all counts.
We have user profiles on the road map, stay tuned.

Statistics: Posted by skataMartin — 11 Dec 2014, 01:25

Calibration and multiple users

2014-12-08T15:10:02+02:00

This is a feature request:
Having to go through a calibration process is tedious to any user. If you want to use gaze tracking as a GUI input device,
this is a cause for concern. The ideal system would be self-calibrating in some way.

Waiting for that to appear, at least we need to make sure that calibration only has to take place once. For a single
user that is mostly the case today. The calibration data is persisted. However for an application with multiple
users that is not the case. Each time the user changes, a new calibration must be done and the old calibration data
is lost.

It would be desirable for the api to offer some control of the calibration data, so that an application can change between
several user profiles. Each user would then only have to calibrate the first time he or she uses the system.

E.g. the client could provide an id for a given calibration. The server should not delete calibrations with an id, except
when explicitly being told so. The client should be able to re-activate stored calibrations.

Or, alternatively the server could make the calibration data avaliable to the client. The client would then be able to associate
the data with a given user and store it. Subsequently, the client would be able to change user profiles and send stored calibration
data to the server in order to re-activate a previous calibration.

Regards,
Mikael

Statistics: Posted by skatamlv@bkmed.dk — 08 Dec 2014, 15:10