ISPRS-Annals ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences ISPRS-Annals ISPRS Ann. Photogramm. Remote Sens. Spatial Inf. Sci. 2194-9050 Copernicus Publications Göttingen, Germany 10.5194/isprs-annals-XI-2-2026-293-2026 GT-LOD3: LOD3 Semantic 3D Building Reconstruction Benchmark Dataset Kim Han Sae 1 Wysocki Olaf https://orcid.org/0000-0002-0016-0229 2 Hoegner Ludwig 3 Bittner Ksenia 4 Carpenter Joshua 5 Fraundorfer Friedrich 4 6 Kusari Arpan 7 Mehltretter Max 8 Rottensteiner Franz 8 Schadl Anna 9 Weinmann Martin 10 Jung Jinha 1 Lyles School of Civil and Construction Engineering, Purdue University, West Lafayette, IN, USA CV4DT, Department of Engineering, University of Cambridge, Cambridge, United Kingdom Department of Geoinfomatics, Munich University of Applied Sciences, Munich, Germany Remote Sensing Technology Institute, German Aerospace Center (DLR), Oberpfaffenhofen, Germany Department of Civil Engineering, The University of Akron, Akron, OH, USA Institute of Visual Computing, Graz University of Technology, Graz, Austria University of Michigan Transportation Research Institute, University of Michigan, Ann Arbor, MI, USA Institute of Photogrammetry and GeoInformation, Leibniz University Hannover, Hannover, Germany Faculty of Geoinformatics, Hochschule München University of Applied Sciences, Munich, Germany Institute of Photogrammetry and Remote Sensing, Karlsruhe Institute of Technology (KIT), Karlsruhe, Germany 03 07 2026 XI-2-2026 293 302 Copyright: © 2026 Han Sae Kim et al. 2026 This work is licensed under the Creative Commons Attribution 4.0 International License. To view a copy of this licence, visit https://creativecommons.org/licenses/by/4.0/ This article is available from https://isprs-annals.copernicus.org/articles/XI-2-2026/293/2026/isprs-annals-XI-2-2026-293-2026.html The full text article is available as a PDF file from https://isprs-annals.copernicus.org/articles/XI-2-2026/293/2026/isprs-annals-XI-2-2026-293-2026.pdf

Reconstructing semantic 3D building models at level of detail 3 (LOD3) is a long-standing challenge in photogrammetry, remote sensing, and computer vision. In contrast to conventional mesh-based representations, LOD3 models require watertight geometries along with semantic object-level facade elements. In GT-LOD3, we introduce the first LOD3 building benchmark dataset comprising point clouds and images paired with 32 ground truth (GT) LOD3 instances in two different countries. We also analyze the performance of selected baselines, complemented by a discussion on unresolved challenges. We are convinced that GT-LOD3 will facilitate the development of novel LOD3 reconstruction methods, enabling the widespread adoption of LOD3 models and, consequently, various downstream applications, ranging from energy demand estimation to automated driving function testing. We release the dataset as open-source and it can be accessed at <code>https://github.com/gdslab/GT-LOD3-Benchmark</code>&nbsp;