Code and Data for Manuscript on HWVOI

High window view openness is valued by urban dwellers, especially in high-density cities, for stress relief and living satisfaction. Quantifying human-perceived window view openness is therefore significant for urban planning, design, and housing. However, existing measures of view openness are either objective, thereby overlooking human perception, or subjective, with limited scalability. Leveraging geovisualization of photorealistic 3D city models and automatic Machine Learning (ML), we introduce a Human-Perceived Window View Openness Index (HWVOI) together with a citywide assessment method. First, HWVOI is defined on three calibrated objective view descriptors, i.e., proportions of sky and building views, and landscape distance. Then, we assess HWVOIs via automatic linear ML regression based on openness ratings and objective view descriptors of representative window views generated from 3D city models. Experimental results in Hong Kong confirmed that the method was automatic to generate the representatives (≤ 3.13 s per image), the view descriptors (≤ 0.02 s per image), and the HWVOIs (≤ 0.001 s per image). The linear ML models were accurate (R² = 0.9191) and interpretable for HWVOIs of citywide windows. The 3D GIS-based approach with quantitative perception evidence citywide enables new solutions for urban renewal decision-making, generative architectural and urban design, and housing valuation.