The Eye Tribe

Re: Calibration

2015-09-04T00:05:18+02:00

So now we have persistent calibration which is great, but a new problem emerges related to this:

Lets assume that the user just unpacked the eyetribe and connected it, then starts my 3rd party program. What happens... it connects but the tracker seems to not work. Why not, the lights are on? Because they user never ran the initial calibration. There's no way for my program to check if they did, and there's no warning issued by the eyetribe. Users can waste a lot of previous time before realizing they forgot to run the eyetribe UI.

It would be great if you can add a function call that asks isEyeTrackerCalibrated() or isEyeTrackerCalibrationLoaded() so that if it returns false, we can either notify the user that they need to calibrate, or start the EyeTribe UI automatically.

The first hour of game play decides if your program will succeed or fail in the market!

Statistics: Posted by skataryklin — 04 Sep 2015, 00:05

Re: Calibration

2014-12-16T22:06:53+02:00

Hi,

if one out of 9 calibrationpoints has the status 1 at the end of a calibration, the tracker sends the result as 'false'.

Now i have 2 questions:

1. Can i make the eyetracker accept the calibration although there are some "1" statuses?
[s]2. if not - how do i recalibrate one single point ?[/s]

so i figured it out, that the eyetracker automatically accepts a calibration as soon as all points return a 2-status.
my problem was, that i hat 9 predefined points and only pushed the index of each point to a function. - i didn't know, that the points order in the calibresult object are not fix - here an example:

Code:: i get back the first calibresult where the status of point 5 is 1 - so you have to recalibrate it. after you do that, you get back a new calibresult. but now the former point 5 has been cut from the calibpoints and appended to the end of the list in the calibresult object and now is point 8. so if now the point is still status 1 - you would have to recalibrate - and pass point 5 to be recalibrated - but point 5 is now point 8 and the algorithm broke.

it took some time for me to figure out because - as i already said - i was passing on the indices of a point and not their x,y positions.
do you get what i mean? - not knowing that can cause some searching time - i think it should be pointed out somewhere. maybe it helps someone

Statistics: Posted by skatawarbyrd — 16 Dec 2014, 22:06

Re: Calibration

2014-12-10T01:26:04+02:00

Is this still missing in .49?

Statistics: Posted by skatagreg@collectivesoftware.com — 10 Dec 2014, 01:26

Re: Calibration

2014-11-28T11:30:39+02:00

Ahh, OK never mind then.

do we know if there is an ETA on the update.. or is there a dev blog I can subscribe to anywhere ?

Statistics: Posted by skatahendersond2004@hotmail.com — 28 Nov 2014, 11:30

Re: Calibration

2014-11-28T00:19:12+02:00

Not yet.

Statistics: Posted by skataJeffKang — 28 Nov 2014, 00:19

Re: Calibration

2014-11-27T13:41:28+02:00

Has the update including persistent calibration data been released yet?
I am on 0.9.41
if so how do you get /set the profile?

Statistics: Posted by skatahendersond2004@hotmail.com — 27 Nov 2014, 13:41

Re: Calibration

2014-10-07T05:56:42+02:00

great, thank you.

Statistics: Posted by skataryklin — 07 Oct 2014, 05:56

Re: Calibration

2014-09-22T11:48:20+02:00

Calibration profiles and persistence are is implemented now and will be included in the next release.

Statistics: Posted by skataMartin — 22 Sep 2014, 11:48

Re: Calibration

2014-09-19T20:36:39+02:00

Just a reminder, that to "persist and restore a calibration over multiple boot sequences" is an important feature that the Eye Tribe lacks and the Tobii EyeX does not. I hope this moves up in priority along your development roadmap.

Statistics: Posted by skataryklin — 19 Sep 2014, 20:36

Re: Calibration

2014-07-23T20:45:42+02:00

This is critical... there is no reason why a participant should need to recalibrate themselves if the environment has not changed.

Statistics: Posted by skataryklin — 23 Jul 2014, 20:45

Re: Calibration

2014-07-23T10:38:44+02:00

ryklin wrote:
What about saving a user's calibration to a file and reloading it later? That's crucial.

When the server exits, all calibration information appears to be lost. Hence, upon restarting the server, a new calibration must either be performed or preferably, a previous calibration file should be loaded.

I think that they’re going to add this feature:

Martin wrote:
Hi David,

Currently the calibration is set in run-time and persists through multiple client sessions and aborted calibration. The feature to persist and restore a calibration over multiple boot sequences is on our roadmap.

viewtopic.php?f=9&t=115&hilit=persist

Statistics: Posted by skataJeffKang — 23 Jul 2014, 10:38

Re: Calibration

2014-07-21T23:39:40+02:00

What about saving a user's calibration to a file and reloading it later? That's crucial.

When the server exits, all calibration information appears to be lost. Hence, upon restarting the server, a new calibration must either be performed or preferably, a previous calibration file should be loaded.

Statistics: Posted by skataryklin — 21 Jul 2014, 23:39

Re: Calibration

2014-07-21T13:08:51+02:00

The calibration will stop when the N points specified in the calibration start message have been reached (e.g. after 9 different points have been displayed). Once that occurs, the server will compute the calibration parameters and send a message with the results. If the results are not satisfactory, you can either resample "bad" points or abort the calibration.

Please take a look at the C# calibration sample available on GitHub here https://github.com/EyeTribe/tet-csharp-samples

Statistics: Posted by skataJavier — 21 Jul 2014, 13:08

Calibration

2014-07-19T02:21:27+02:00

There doesn't seem to be a way to end / accept a calibration.

1. There is a start calibration

2. Then a pointstart with x,y
- are these in pixel coordinates? I guess so. In general, it's easier to work in normalized coordinates.

3.pointend, how much time needs to pass before calling this. Some docs say 2 seconds, but that's obviously a typo. There is another post addressing this... I need to go read it.

4. repeat for N points as specified in the start calibration. Presumably you can repeat the same point if it has a poor calibration. More on this later.

5. there is no end / terminate / accept calibration. how does the tracker no it's done with calibration and should apply it now?
How do we validate the calibration's accuracy and precision?

6. There is a clear, but I'm not sure why I would use this. It's not really practical, but fine with me.

7. there is an abort, which reinstates the previous calibration. that's good

8. there is no documentation on how to get and then save a calibration to file, or open and load a calibration from a file. This is critical!

Statistics: Posted by skataryklin — 19 Jul 2014, 02:21