Gaze Data Enhances Urban Perception Modeling
The Place Pulse-Gaze dataset merges street view visuals with synchronized eye-tracking data and personal perception labels. Researchers introduce a Gaze-Guided Urban Perception Framework that leverages gaze behavior to represent subjective perceptions of urban spaces. This framework investigates three approaches: modeling based solely on gaze, combining gaze with explicit semantic scene representations, and integrating gaze with more complex implicit visual representations. Experimental results demonstrate that gaze data by itself offers valuable predictive insights into how individuals assess urban settings.
Key facts
- Place Pulse-Gaze dataset augments street view images with eye-tracking recordings and perception labels.
- Gaze-Guided Urban Perception Framework uses gaze behavior to model urban perception.
- Three settings: gaze-only, gaze fusion with semantic representations, gaze fusion with implicit representations.
- Gaze alone carries useful predictive signal for urban perception.
- Existing approaches ignore human perceptual process.
- Urban perception shapes how cities are experienced.
- Dataset introduced in arXiv paper 2605.00764.
- Paper type is cross.
Entities
Institutions
- arXiv