Fully automated, highly detailed, 3D building models for a whole country
2021-10-01, 10:30–11:00, European Commission

In this talk we take you through our fully automated process that we developed for generating highly detailed building models for the Netherlands, the 3D BAG. Only open data and open source tools are involved.

Semantic 3D city models, or digital twins, are one of the cornerstones of the so-called "smart city" applications. Although several regions have them, they are still relatively rare, outdated and difficult to access. This is often due to their price, since the most common methods for generating highly detailed models involves the manual modeling of individual objects. We developed a fully automated process for generating high detail (Level of Detail 2.2) building models for the whole Netherlands, which you can freely use, see 3D BAG. In this talk we take you through our whole building reconstruction process, in which we only use open source tools, like CGAL, GDAL, PostGIS, and more. We share the lessons we learned, and the main stumbling blocks we encountered. This project has received funding from the European Research Council (ERC) under the European Unions Horizon2020 Research & Innovation Programme (grant agreement no. 677312 UMnD: Urban modelling in higher dimensions).

Authors and Affiliations

Dukai, Balázs (1)(2)
Peters, Ravi (1)(2)
(1) 3D geoinformation research group, TU Delft, the Netherlands
(2) 3DGI, the Netherlands

Requirements for the Attendees

Visit https://3dbag.nl


Open data


Data collection, data sharing, data science, open data, big data, data exploitation platforms


1 - Principiants. No required specific knowledge is needed.

Language of the Presentation


Balázs Dukai is a R&D Engineer 3D geoinformation group @ TU Delft, Co-founder of 3DGI. He spends most of his time developing tools and methods for spatial data processing.

Postdoc at the 3D geoinformation group @ TU Delft, Co-founder of 3DGI. He likes to tinker with 3D geodata such as point clouds and city models.