PositiveCorrelation between ScreenResolution & GazeAccuracy

[giannismoustakis]

Bugi74, at topic "First Impression", wrote :

Third calibration (after a while of "playing with it"), and adjusting the "area size" to 2000x1600, instead of the normal 2560x1600, gave 4 stars, and very well working calibration. I guess I have to play with that a bit and see what will give best results.

I reduced my screen resolution from 1366x768 to 1280x768 and I have much better results.

Thanks Bugi74.

Regards,
Giannis

[bugi74]

Note that I did not change the actual resolution of the display, I only adjusted the "Area size" values in the EyeTribe UI. That affected the area over which the calibration circles are spread during calibration, but the full display area will still be trackable after the calibration is done.

The reason why I reduced the area was because I was getting worse accuracy (at calibration) at the horizontal edges, and I have 30" monitor (larger than recommended), so I just thought to try how it would work if I reduced the horizontal span towards the recommended maximum.

The improvement from 3 to 4 stars can very well be "placebo" effect, due to the calibrator working on a smaller (and thus more accurate) angles, and claiming better result - but only for the calibration and the area it covered. (Or maybe I'm just learning to use my head and eyes in a better way for the tracker.) The edge areas may still be just as inaccurate as before, the calibrator just didn't try them. The important part of the test is that the results was definitely not worse, as one might have expected when the calibration does not cover the full area that will be used.

If reducing the actual display resolution, it might also give a false improvement in the calibration, as the (logical) pixels will be larger or the display area (and angles) will reduced, and thus the measured errors and jittering will also be smaller (if measured in pixels, instead of the true distance/angles). That is why I wrote that I have to play with it a bit - to measure the true accuracy over full display with different settings.

But no matter what adjustment is done, if it truly improves the tracking, it is a valid option to use.

[giannismoustakis]

There are enough variables that evoke changes to accuracy such as the Area Size and Screen Resolution.
I recognize that the area size isn't an equivalence parameter with screen resolution, but (maybe) both of them operant on gaze accuracy improving the final result.

In my opinion, one important variable that I have to control, is the tripod. I make a new one (homemade) and when i finish it, I will share it with some photographs.