Dr Olaf Wysocki, Assistant Research Professor at the University of Cambridge, represented the Computer Vision for Digital Twins (CV4DT) research group at the XXV ISPRS Congress 2026 in Toronto, Canada. The CV4DT team presented three papers and co-organised a tutorial on 3D urban modelling, sharing new research in digital twin technologies, 3D city modelling and computer vision with the international research community.
The three papers, published in the ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences, highlight CV4DT’s research across semantic 3D reconstruction, urban scene understanding and uncertainty estimation.
The paper GT-LOD3: LOD3 Semantic 3D Building Reconstruction Benchmark Dataset introduces a benchmark dataset for Level of Detail 3 (LoD3) semantic building reconstruction. The dataset provides pairs of point clouds, images and LoD3 building models from the United States and Germany, supporting research into semantic 3D reconstruction and urban digital twins.
The paper CityLangSplat: Integrating CityGML Semantics into 3D Language Gaussian Splatting for Urban Scene Understanding explores combining structured 3D city model semantics from CityGML with 3D Language Gaussian Splatting to support improved understanding of urban environments.
The paper From Propagation to Prediction: Point-level Uncertainty Evaluation of MLS Point Clouds under Limited Ground Truth investigates a learning-based approach to uncertainty evaluation for Mobile Laser Scanning (MLS) point clouds, exploring how uncertainty estimation can be predicted when ground truth data is limited.
Alongside presenting the papers, the CV4DT team co-organised a tutorial on 3D urban modelling, bringing together researchers to discuss recent developments and future directions in the field.
The participation at ISPRS Congress 2026 highlights CV4DT’s contribution to advancing digital twin technologies and applying computer vision and geospatial approaches to better understand and model the built environment.