Device accuracy issue (bottom center points)

[vstroy]

We have tested your device intensively and followed all best practices from TET, but still have some questions and need your assistance.

Screenshot with our results:
In all cases people saw 9-point grid and focused on all of them or followed some pattern. In all cases 9-point calibration was made and calibration results were never lower then 3 stars. Nevertheless, we saw stable incorrect results for lower and sometimes upper point in the middle column (see the image attached).

We tried to position device on your tripod or using other DIY mounts, but we constantly see these shifts to the center (looks like the "pillow effect" in old CRT monitors). Points on the right or left side look much more accurate. We've got the same results with another ET device.

Please let us know, what is the possible reason of such behavior.

[JeffKang]

Check out this comment:

With a 5 stars calibration, I have very high precision, the mouse cursor is constrained to an area maybe 2-3 mm across and (when smoothed) low travel in that area, but extremely low accuracy. The cursor might be an inch or more away from where I am looking. Not drifting, just sitting there.

Ultimarly, that's a good sign. Low accuracy can be fixed by math, low precision can't.

They have a mythical 1.0 build in QA that should improve accuracy, but in the meantime there isn't much you can do, other than roll your own calibration technique and cursor control.

viewtopic.php?f=9&t=342

[vstroy]

I think that the problem is related to the internal points interpolation inside the EyeTribe software. We can predict errors and make some corrections. But if this issue will be fixed in future TET software releases it will break our results (our math corrections will give us errors in opposite direction).

[Martin]

Hi Vyacheslav,

We looking into this to see if we can replicate the behavior on our end.

Meanwhile, could you please tell me a little more about your setup, specifically the physical screen size and placement of device relative to screen. A photo would be great.

[vstroy]

Hi Martin,

Some details about our test setup:

• - monitor size: 22"
• - eyes distance: ~50cm
• - sampling rate: 30Hz
• - calibration: 12 points

We've tried various desktops and laptops with smaller screens. We've also made some experiments with device positioning relative to the screen. Problem remains.
Our clients are expierencing the same problem (deviations near bottom center point).

And few photos:

[Martin]

Hey V,

Thanks for the info.

The setup looks very suitable, we have many similar arrangements here, but cannot reproduce (other than occasional poor calibration points).

Are you using the calibration UI from the Eye Tribe software package or are you rolling your own?
Does it improve if you do a nine point instead of twelve?

Good you sent the report to Javier, he will look into the data.