Paper: Decoding characteristics of building facades using street view imagery and vision-language model

Acquiring and analysing the characteristics of buildings in cities has long been of interest, while comprehensive evaluation at scale remains challenging. This study investigates a novel approach to encoding the characteristics of building facades in urban areas by leveraging street view imagery (SVI) and vision-language models. Among the first, we integrated large language models and natural language understanding to analyse 48,752 building images in Hong Kong, identifying eight building clusters aligned to geometric footprints city-wide. Our findings highlight the effectiveness of SVI-based analyses in capturing urban spatial and semantic details, providing scalable insights that bridge visual and linguistic domains for a deeper understanding of the built environment.

@inproceedings{liang2025decoding,
    title={Decoding Characteristics of Building Facades Using Street View Imagery and Vision-Language Model},
    author={Liang, Xiucheng and Cheng, Sifan and Biljecki, Filip},
    booktitle={19th International Conference on Computational Urban Planning and Urban Management},
    year={2025}
}