Automated gaze direction scoring from videos collected online through conventional webcam.

Abstract

As online research has become more prevalent, researchers have been investigating the possibility of replicating techniques that go beyond measuring only simple behaviour. One such method could leverage the webcam of the participants’ device to collect information about eye gaze direction. Several packages have been developed for collecting such data, but they all lead to high attrition and require extensive and potentially frustrating calibration procedures, which hinders all research, in particular data collection including children and participants with neuro-developmental difficulties.To overcome this issue, we developed GazeScorer, a package that uses basic image processing techniques and a simple one-point calibration to score horizontal gaze orientation. We based our work on experience gained from infant research, in which a researcher manually scores horizontal gaze orientation from individual video frames. Using videos collected in two browser-based remote studies, one including adults, and one including children. We achieved low participant attrition, single-frame point calibration, and demonstrated a good level of inter-rater reliability between GazeScorer and a manual scorer. Our package provides a potential resource for researchers working with populations who could not perform a long or involved calibration, in studies in which information about horizontal gaze orientation is sufficient.

Dynamics of eye-hand coordination are flexibly preserved in eye-cursor coordination during an online, digital, object interaction task

“…However, webcam eye-tracking has struggled to establish its utility as a research tool. The reasons are numerous: webcam eye-tracking has a much slower sampling rate (∼10 Hz compared with 100+ Hz in lab [5,14,45]), many users are unable to participate due to a slow internet connection or insufficient hardware [5,14], uncontrolled lighting can significantly decrease data quality [12,45,54] and, even with optimal conditions, extensive time must be spent calibrating the system (up to 50% of the study duration as in [45]). Despite these limitations, recent advances, especially in using machine learning to predict gaze location (e.g.…”

Genetic syndromes, neuroconstuctivism and replicable research; challenges and future directions

“…Relatedly, technological advances are broadening the accessibility of many techniques. For example, eyetracking can be reliably completed from a laptop computer in participants' homes or schools (e.g., Fraser et al, 2021), EEG has become wireless and portable, whilst advances in genetic profiling and techniques such as functional Near F I G U R E 1 Sensitivity curve plot of effect size against desired power when n = 20 to 100 and alpha = 0.05 (horizontal red line = 80% power). Derived using G*Power 3.1.9.7…”

“…Relatedly, technological advances are broadening the accessibility of many techniques. For example, eye‐tracking can be reliably completed from a laptop computer in participants' homes or schools (e.g., Fraser et al, 2021), EEG has become wireless and portable, whilst advances in genetic profiling and techniques such as functional Near Infrared Spectroscopy (fNIRS) will enhance the understanding of multi‐level interactions in neurodevelopmental disorder groups in the decades to come. Of course, as emphasized above, informed consideration of these variables requires multi‐disciplinary collaboration.…”

Assessing two methods of webcam-based eye-tracking for child language research

“…Visual-world experiments are primarily conducted in university labs where researchers employ specialized equipment to monitor participant gaze (e.g., SR Research, 2021;Tobii, 2021). More recently, however, algorithms that determine gaze location based on webcam video have increased interest in conducting eye-tracking experiments without specialized equipment and outside of lab settings (e.g., Erel et al, 2022;Fraser et al, 2021;Papoutsaki et al, 2016;Valenti et al, 2009;Valliappan et al, 2020;Xu et al, 2015). Webcam-based eye-tracking allows researchers to conduct experiments over the internet, in either supervised settings (with an experimenter present over video conferencing) or unsupervised settings (with no experimenter present).…”