Comprehensive VR dataset for machine learning: Head- and eye-centred video and positional data.

Area of Science:

• Virtual Reality (VR)
• Human-Computer Interaction
• Machine Learning

Background:

• Virtual Reality (VR) environments offer immersive experiences but understanding user behavior within them is crucial.
• Collecting detailed data on human navigation and visual search in VR is complex.
• Existing datasets may lack the comprehensive nature required for advanced machine learning applications.

Purpose of the Study:

• To present a novel, comprehensive dataset of human head and eye movements during a search task in diverse Virtual Reality (VR) environments.
• To provide a rich resource for machine learning research focused on analyzing and predicting user behavior in VR.
• To facilitate the development of advanced VR technologies and algorithms.

Main Methods:

• Human participants performed a search task in six distinct Virtual Reality (VR) environments using a motion platform.
• Head and eye-centered video recordings, along with positional data, were captured and stored in CSV format.
• Data collection included over 10 hours of cumulative playtime across naturalistic VR settings.

Main Results:

• A comprehensive dataset comprising synchronized head and eye movement data and positional information was successfully collected.
• The dataset covers diverse VR environments, including naturalistic landscapes and urban settings.
• The data is structured for high potential reuse, particularly for machine learning model training.

Conclusions:

• The presented dataset is a valuable resource for advancing machine learning research in Virtual Reality (VR).
• It enables the development and refinement of algorithms for predicting visual search behavior, eye movement patterns, and navigation strategies.
• This dataset will foster innovation in VR technologies by providing a foundation for data-driven development.