You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I have the IrisTracking Solution running without issue. However, I'm trying to estimate gaze based on the output of the FaceLandmarks, specifically the ones that correspond to the detected Iris. Does anyone have such an example? I cannot find one. I know its possible, as Google's advertisements show using this data to direct the eyes of avatars as in here. If BlendShapes would help, how do I enable them on the config? I can't seem to find where to do what (or to set min detection confidences, etc).
Code to Reproduce the issue
No response
Additional Context
No response
The text was updated successfully, but these errors were encountered:
I'm actually working on the same thing. Did you make much progress? I've tried a number of approaches:
Placing spheres directly behind the eyes, updating their position relative to nearby points, and casting a ray out through the pupil, colliding with a screen and converting that collision point into UI space.
Getting pupil distance from the outer eye point and rotating the pupil point relative to that, and casting a way.
Taking the ratio between distance of left pupil from left eye corner, and distance of right pupil from right eye corner
some variants on the above.
Ultimately, I have found that the output from the FacialLandmarks is too variable to be reliable with any calibration, and also changes over time within the same session, so ultimately I've landed on a gestural system for looking left and looking right, changing the 'gaze point' to difference in game objects, but this isn't anywhere near to the fidelity I'd hoped for.
Would love to get your thoughts/know what you tried.
Plugin Version or Commit ID
0.12.0
Unity Version
2022.2.19
Your Host OS
Windows 11
Target Platform
Android
Description
I have the IrisTracking Solution running without issue. However, I'm trying to estimate gaze based on the output of the FaceLandmarks, specifically the ones that correspond to the detected Iris. Does anyone have such an example? I cannot find one. I know its possible, as Google's advertisements show using this data to direct the eyes of avatars as in here. If BlendShapes would help, how do I enable them on the config? I can't seem to find where to do what (or to set min detection confidences, etc).
Code to Reproduce the issue
No response
Additional Context
No response
The text was updated successfully, but these errors were encountered: