아이트래킹 값에서 시선추적 값의 정확도를 알 수 있나요?



  • Avatar

    그리고, 시선추적 방향이 어긋나지 않는 방법 또는 어긋나는 경우에 대한 예외처리 방법이 있을까요?


    and is there a way to make an exception for the way the line of sight tracking is not reversed or the way?

    Comment actions Permalink
  • Avatar


    The current calibration procedure assumes that both eyes are looking at the calibration point simultaneously. This means if that one eye is looking at the wrong point, which can happen with certain types of disorders, then the calibration for the non-dominant eye will be less accurate. This is as you surmised.

    We have no "accuracy" number of the calibration exposed. This is tricky, because if the calibration was incorrect due to bad ground truth (assumption that an eye was looking at the dot when it wasn't), any measurement of the accuracy level would also be also biased based on the invalid data.

    You can create your own simple benchmark, for example using Unity, and test each eye independently.

    We are investigating whether to add a mode to calibrate each eye independently, but have nothing to announce currently regarding that. If you are interested in this, please let me know your use case as we want to hear from such people.

    Comment actions Permalink

Please sign in to leave a comment.