tum2twin

This repository contains CityGML LOD3 models of the Technical University of Munich.

🚀 Getting Started

The datasets are stored on a dedicated GitLab repository. To clone the repo, make sure to have Git LFS installed and run:

git clone --depth 1 git@gitlab.lrz.de:tum-gis/tum2twin-datasets.git

The tum2twin-dataset repository contains:

• CityGML
  • lod2-building-datasets: LOD2 building models of the Bavarian State Office for Digitizing, Broadband and Survey (LDBV)
  • lod2-textured-building-datasets: LOD2 textured building models
  • lod3-building-datasets: Exported CityGML v2 LOD3 building models
  • lod3-vegetation-datasets: Trees modelled using this workflow
• SketchUp
  • lod2-textured-building-projects: LOD2 SketchUp projects for texturing the building models
  • lod3-building-projects: SketchUp projects modelled according to this guideline

The file names follow the GML IDs of the LDBV.

👓 Visuals

[3D web visualisation (beta of Ch. Beil)] [more demos]

🧪 Model Collection

The goal of tum2twin is to contribute to a collection of different representations of the TUM campus and its surroundings to promote research and development of new methods.

• LDBV: tum2twin follows and preserves the structure of the official LOD2 building models (tile 690_5336 and 690_5334)
• TUM-FAÇADE: MLS point clouds with facade-level labels

🛠️ Contributing

If you find any errors or deficiencies, please create an issue. Improvements and extensions of the models are also highly welcome.

🎓 Research

Publications that use the dataset:

• Scan2LoD3: Reconstructing semantic 3D building models at LoD3 using ray casting and Bayesian networks, CVPRW '23 proceedings
• TUM-FAÇADE: Reviewing and enriching point cloud benchmarks for façade segmentation, ISPRS Archives, ArCH '22 proceedings
• Automatisierte Generierung eines Baumkatasters aus Punktwolken in unterschiedlichen urbanen Umgebungen, Masterarbeit 2023, Technische Universität München, Github Repository
• Thermal Mapping from Point Clouds to 3D Building Model Facades, Remote Sensing journal, 2023
• Evaluation of the Effect of Enriched Facade Models on Image-Based Localization of Vehicles, Bachelorthesis 2023, Technical University of Munich, Github Repository
• Reconstructing Façade Details Using MLS Point Clouds and Bag-of-Words Approach , 3DGeoInfo 2023
• Inpainting of unseen façade objects using deep learning methods, Masterarbeit 2023 , Technische Universität München

🤝 Acknowledgement

• Landesamt für Digitalisierung, Breitband und Vermessung (LDBV) for their great open data offering
• AI4TWINNING, the project which is facilitating the development of the LoD3 models.
• 3D Mapping Solutions for providing high-grade 3D point clouds within the scope of the MoFa project