Metrics for Time-to-Event Prediction of Gaze Events
Abstract
In this paper, we explore metrics for the evaluation of time-to-saccade problems. We define a new sampling strategy that takes the temporal nature of gaze data and time-to-saccade problems into account, avoiding samples of the same event in different datasets. This allows us to define novel error metrics for a more intuitive evaluation of predicted durations. The metrics are defined to evaluate the consistency of a predictor and the evaluation of the error over time. We evaluate our method using a state-of-the-art method for time-to-saccade prediction along with an average baseline on three different datasets.
Cite
Text
Rolff et al. "Metrics for Time-to-Event Prediction of Gaze Events." NeurIPS 2022 Workshops: GMML, 2022.Markdown
[Rolff et al. "Metrics for Time-to-Event Prediction of Gaze Events." NeurIPS 2022 Workshops: GMML, 2022.](https://mlanthology.org/neuripsw/2022/rolff2022neuripsw-metrics/)BibTeX
@inproceedings{rolff2022neuripsw-metrics,
title = {{Metrics for Time-to-Event Prediction of Gaze Events}},
author = {Rolff, Tim and Stein, Niklas and Lappe, Markus and Steinicke, Frank and Frintrop, Simone},
booktitle = {NeurIPS 2022 Workshops: GMML},
year = {2022},
url = {https://mlanthology.org/neuripsw/2022/rolff2022neuripsw-metrics/}
}