Author

Qixu Xie, Li Zhang*

Affiliation

School of Architecture, Tsinghua University

Abstract

Selective visual attention determines what pedestrians notice and ignore in urban environments. If individuals' visual attention is consistent, designers can modify designs by understanding the underlying mechanisms and better meeting users’ needs. However, the mechanisms behind pedestrians' visual attention remain poorly understood, making it difficult to predict which locations in urban environments will attract the most attention.

To address this gap, we used 360° video and immersive virtual reality to simulate walking scenarios and record the eye movements of 138 participants. Our findings reveal a remarkable consistency in fixation distributions across individuals, exceeding both chance levels and orientation bias. One factor driving this consistency appears to be information maximization: participants tended to fixate on areas with higher local entropy.

Additionally, we created the first eye movement dataset for panorama videos of diverse urban walking scenes and developed a supervised deep-learning model to predict pedestrians' visual attention. The model can help designers better understand how pedestrians will visually interact with urban environments during the design phase. The dataset and predictive model code are available at https://github.com/LiamXie/UrbanVisualAttention

Publisher

Building Simulation