0 votes
by (280 points)

We're getting some diverance between the tracking values and hope you have some explanation for it.

In comparing the output from the Tobii SDK versus the Pico SDK, we have discovered that the Tobii GazeRay Origin and Direction differs from the Pico Combined Eye Point and Combined Eye Point Vector.

In the case of the "Z" component for the Origin vs Point, this is constantly off by about 0.054 to 0.062 (found by simple subtraction of the GazeRay Origin from the Combined Eye Point).

The X and Y component values are also often offset by very small amounts (typically 0.000133764 or less).

The Direction minus Vector is also sometimes slightly offset by very small amounts (0.0009421092), sometimes not at all.
Our test scene is a static field with a single word of text on it, so nothing for the user to "focus" on.

We calibrate the user's eyes prior to each test run, and these results are consistent from user to user.

Do you know why and can explain what is causing these offsets, and which of the two - Pico vs Tobii - should we consider as the "gold standard" for this?

1 Answer

0 votes
by (24.4k points)

Dear developer,
Please use the Tobii SDK as the standard.
The offests may be due to a slight lag when the Picoo SDK gets the data.