February 27, 2017 at 19:41 #6384
I would like to know what kind of precision we can expect if there is no calibration (or one global calibration done once).
We’re looking to build an immersive experience that would use eye tracking, the duration would be around 1 minute, it would be in an outdoor environment, and the users would never be the same.
The distance to the screen would be controlled (within 1 meter).
The screen would be split into areas of equal size, we would need to determine which area is the user looking at.
RegardsFebruary 27, 2017 at 22:22 #6387
Hi @mutex, I am afraid we cannot provide a quantifiable measurement of accuracy when no calibration is performed but suffice to say it would likely not be reliable at all… perhaps if the buttons to control were particularly large but otherwise you will be disappointed and so you should always try for a calibration.February 27, 2017 at 22:34 #6389
Assuming we split the screen in 4 equal parts (top left, top right, bottom left, bottom right), would we be able to determine which part we are looking at without calibration? Assuming we’re 1 meter away from a 42 inch television for example?March 1, 2017 at 21:22 #6404
Hi Mathieu, yes I would say this is quite possible indeed! How are your results so far?March 3, 2017 at 07:45 #6408
Just thought I would give you a layman’s perspective since I just got to try Tobii’s product for the first time yesterday at GDC. I walked up to half a dozen game demo’s and all of them reacted to me but it was clear none were calibrated. For example, all games I tried had a feature where looking towards the edge of the screen would cause the camera to drift slightly where the user looked. This worked well when without calibrated. Most of the games also would highlight HUD or game elements when looked at. This was very sketchy without calibration, leaving me wondering what was even supposed to be happening.
My observation is that very general tracking seemed highly likely to work when a layperson just walked up but precise tracking was not at all reliable.
After trying the demos without help, one of the presenters showed me how the calibration worked. It was very quick, intuitive and only took a few steps. After calibration the trackers were very responsive and accurate. I was very impressed with the experience after this point.
If calibration were integrated into the program, I bet it could be done in under 30 seconds with very few problems. The only downside is that it would require the user to know that it needs to be done and do it.March 14, 2017 at 20:03 #6485
Hi @mutex, may I ask how did you get on? Can I close this thread as Solved for your needs? Many thanks.March 14, 2017 at 20:06 #6487
Thanks to Joshua for the additional information. We did a few tests and we have a good enough precision without using calibration.
You can close the thread,
RegardsMarch 16, 2017 at 21:03 #6501
Great, thanks for letting us know Mathieu.
You must be logged in to reply to this topic.