GazeBaseVR, a large-scale, longitudinal, binocular eye-tracking dataset collected in virtual reality

Area of Science:

• Ophthalmology
• Computer Science
• Human-Computer Interaction

Background:

• Eye-tracking (ET) data is crucial for understanding human visual behavior.
• Existing datasets often lack the scale, longitudinal aspect, or VR integration needed for advanced research.
• Virtual reality (VR) offers immersive environments for collecting ecologically valid eye-tracking data.

Purpose of the Study:

• To introduce GazeBaseVR, a comprehensive, longitudinal binocular eye-tracking dataset.
• To provide a rich resource for research in eye movement biometrics, VR interaction, and human behavior.
• To enable studies on factors influencing eye movements, including fairness and health impacts.

Main Methods:

• Collected 5,020 binocular eye-tracking recordings at 250 Hz using an ET-enabled VR headset.
• Recruited 407 diverse college-aged participants, with recordings up to six times over 26 months.
• Included five distinct eye-tracking tasks: vergence, smooth pursuit, video viewing, reading, and saccades.

Main Results:

• GazeBaseVR is a large-scale, longitudinal dataset with extensive participant diversity.
• The dataset includes participants from previous studies and those recorded pre/post COVID-19 infection.
• Rich metadata supports research on eye movement biometrics, VR usability, and fairness.

Conclusions:

• GazeBaseVR is a valuable resource for advancing eye-tracking research in VR.
• Its longitudinal nature and diverse tasks support robust analysis of eye movement patterns.
• The dataset facilitates research on individual differences and health-related changes in visual behavior.