Wrong Gaze Depth

[rahe...@gmail.com]

Dear Pupil Community!

I have been experimenting with the binocular Pupil headset to find gaze depth in the world camera coordinates.

Even though my gaze points looks perfect on the frame (norm_pos_x/y); i have doubts on the corresponding z-coordinate of the 3d point in space (gaze_point_3d_z).  gaze_point_3d_z changes only slightly even when I look at a far off object or a very close one.

I have been looking at objects that should have z-coordinates between 1m to 4m, but I rarely get a value beyond 1m.

P.s. I am interested in gaze depth ( sqrt(gaze_point_3d_x^2 + gaze_point_3d_y^2 + gaze_point_3d_z^2) ) in the world camera coordinates (Figure attached for Gaze Depth, should have ranged between 1m and 4m). And, I have tried calibrating using Screen Marker Calibration and Natural Features Calibration (from the same scene with features covering the range of z-coordinates I am expecting).

Did anyone face a similar problem? Any suggestions?

Best Regards,
Raheel

[Pupil Labs Info]

Hi Raheel,

please note that depth estimation through binocular vergence is only reliable up to about 1.5-2m beyond this the signal is simply to weak, this has to do with the nature of the problem, not so much with out pipeline. 

 That being said, we are working on improving the accuracy of our pipeline and depth estimation should get better in the near future.

Best,

Moritz

--
You received this message because you are subscribed to the Google Groups "pupil-discuss" group.
To unsubscribe from this group and stop receiving emails from it, send an email to pupil-discuss+unsubscribe@googlegroups.com.
To post to this group, send email to pupil-...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/pupil-discuss/b1079655-1c19-4993-a0a1-ccd14c190117%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

--

Pupil Labs

research + development + design

in...@pupil-labs.com

www.pupil-labs.com

[rahe...@gmail.com]

Hello Moritz,

Thank you very much for the prompt response.
Indeed what I expected. Just a last query, which calibration mode do you recommend for the 1.5 to 2m ranges?

BR,
Raheel

On Wednesday, January 10, 2018 at 2:38:43 PM UTC+1, info wrote:

Hi Raheel,

please note that depth estimation through binocular vergence is only reliable up to about 1.5-2m beyond this the signal is simply to weak, this has to do with the nature of the problem, not so much with out pipeline. 

 That being said, we are working on improving the accuracy of our pipeline and depth estimation should get better in the near future.

Best,

Moritz

On Wed, Jan 10, 2018 at 2:34 PM, <rahe...@gmail.com> wrote:

Dear Pupil Community!

I have been experimenting with the binocular Pupil headset to find gaze depth in the world camera coordinates.

Even though my gaze points looks perfect on the frame (norm_pos_x/y); i have doubts on the corresponding z-coordinate of the 3d point in space (gaze_point_3d_z).  gaze_point_3d_z changes only slightly even when I look at a far off object or a very close one.

I have been looking at objects that should have z-coordinates between 1m to 4m, but I rarely get a value beyond 1m.

P.s. I am interested in gaze depth ( sqrt(gaze_point_3d_x^2 + gaze_point_3d_y^2 + gaze_point_3d_z^2) ) in the world camera coordinates (Figure attached for Gaze Depth, should have ranged between 1m and 4m). And, I have tried calibrating using Screen Marker Calibration and Natural Features Calibration (from the same scene with features covering the range of z-coordinates I am expecting).

Did anyone face a similar problem? Any suggestions?

Best Regards,
Raheel

--
You received this message because you are subscribed to the Google Groups "pupil-discuss" group.

To unsubscribe from this group and stop receiving emails from it, send an email to pupil-discus...@googlegroups.com.

To post to this group, send email to pupil-...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/pupil-discuss/b1079655-1c19-4993-a0a1-ccd14c190117%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Pupil Labs Info]

hi,

I would recommend manual marker.

best,

Moritz

To unsubscribe from this group and stop receiving emails from it, send an email to pupil-discuss+unsubscribe@googlegroups.com.

To post to this group, send email to pupil-...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/pupil-discuss/ba8e9532-5dab-407d-a835-2f7eed03b9f4%40googlegroups.com.

For more options, visit https://groups.google.com/d/optout.