Projects
September 20, 2025
3D city models have become an increasingly prevalent tool to support urban management and facilitate efficient decision-making. However, the properties of 3D city models vary across cities and countries, together with the nature of their initiatives and purposes. Such heterogeneity and different practices are not explored thoroughly, which presents a challenge for usability and for local governments who are yet to establish their 3D initiatives. This issue is compounded by the increasing importance of urban digital twins, which 3D city models are a pillar of. The quality of 3D city models plays a role in the effectiveness of designing and implementing urban digital twins, but how ‘good’ it is, and what the difference is among different practices around the world, is not fully clear. This repository documents a continued assessment of open 3D city models and the trend analysis, building on a four-category framework conducted by Lei et al., (2023). In the meantime, this project serves as a corpus of publicly availabel 3D city models around the world, thanks to the Awesome CityGML project released by Wysocki et al., (2025).
September 20, 2024
Urban digital twins, and 3D city models underpinning them, provide novel solutions to urban management but tend to overlook the human element. The trending research on human perception reveals people’s perspective towards interpreting and experiencing the built environment. This work is the first instance of integrating such attributes in 3D city models, which have traditionally been confined to physical and objective measures. The visual perception of each building is evaluated based on building images extracted from street view images. We add such information as new attributes to an existing CityJSON dataset representing thousands of 3D buildings in Amsterdam, the Netherlands. To facilitate a robust and sustainable integration, we develop a CityJSON Extension to accommodate the new data and validate its schema successfully, and we visualise the semantic 3D dataset.
December 25, 2023
Building characteristics, such as storeys and types, play an important role across a multitude of domain. However, geospatial data on the building stock is often fragmented and incomplete. Here, we propose a novel method to predict a set of building characteristics within diverse contexts, addressing the existing data gaps. Our method exploits the geospatial connectivity between street-level contexts and building stocks, employing Graph Neural Networks (GNNs) for modelling spatial patterns to infer multiple building characteristics.
