Estimating and mapping Tokyo Window View Desirability using 3D model imagery and the hedonic price approach

We long for a connection with nature and a sense of refuge in the views from our living spaces. Views from homes shape both our well-being and housing values, yet there are limited research that quantify the holistic visual desirability of a residential window scene, as reflected in housing rents. This study introduces a pipeline that combines photorealistic synthetic window-view images from Google 3D Tiles with a hedonic price model to estimate Tokyo’s Housing Window View Desirability ( HWVD ) from rents. Leveraging three-bedroom rental property records in Tokyo, we infer HWVD as the rent component explained by window-view imagery after controlling for structural and locational attributes. Incorporating HWVD improves out-of-sample rent prediction (R 2 increases from 0.62 to 0.71) with a building-level spatial resampling (train-test split). These findings were further validated through a human user survey with experts. To interpret the metric, we relate HWVD to the semantic content of the views, finding significant positive correlations with visible greenness and negative correlations with building views. Spatial analysis reveals significant spatial clustering, with higher HWVD concentrated in southwestern Tokyo, particularly in wards such as Shinagawa-ku and Minato-ku and lower HWVD in eastern wards like Katsushika-ku. We found more desireable views are concentrated in wealthier wards, reflecting a clear socio-economic divide in Tokyo’s visual landscape. In summary, this research provides a scalable approach to revealing’unseen’ visual amenities from residential spaces, with significant implications for applied geography, urban planning, real estate, and architectural design