Accuracy Assessment of ARKit 2Based Gaze Estimation

Abstract

This paper benchmarks the accuracy of ARKit 2’s camera-based gaze estimation for common interaction distances and lighting settings. We measure angular error, stability, and drift against a reference eye tracker and analyze implications for selection, reading, and attention-aware UI. The results provide practical thresholds and calibration advice for deploying gaze-enabled AR on commodity devices.

BibTeX

@inproceedings{voigtantons2020oc15,
  author    = {Greinacher, R. and Voigt-Antons, J.-N.},
  title     = {Accuracy Assessment of ARKit 2Based Gaze Estimation},
  year      = {2020},
  booktitle = {Paper presented at the International Conference on Human-Computer Interaction (pp. 1–10). Springer, Cham},
  doi       = {10.1007/978-3-030-49059-1_32},
}