On 19 June 2026, Dr. Sining Chen successfully defended his doctoral dissertation at the Technical University of Munich (TUM). His research focused on large-scale 3D semantic understanding of urban environments by integrating information from satellite and street-view imagery.

As urbanization continues to accelerate worldwide, comprehensive and scalable understanding of the built environment has become increasingly important for urban planning, sustainability, and environmental monitoring. Although Earth Observation has greatly expanded the availability of urban data, reconstructing cities in three dimensions at a global scale remains challenging due to the limited availability of high-quality 3D information, heterogeneous annotations, and the complexity of urban scenes.

Urban 3D Semnatic Understanding at Scale: From Overhead to Street-view Imagery

Dr. Chen's dissertation addresses these challenges by developing learning-based methods for 3D building reconstruction and urban semantic understanding from both overhead and street-level observations.

The dissertation makes several key contributions:

Monocular building height estimation from satellite imagery.
Novel supervised, weakly supervised, and semi-supervised learning frameworks were developed to estimate building heights under varying annotation availability and data quality, enabling scalable reconstruction of urban vertical structure.

Global Building Atlas.
The research contributed to the development of the first openly available global dataset of building footprints, heights, and LoD1 3D building models, covering approximately 2.75 billion buildings worldwide and providing an important resource for Earth Observation and urban studies.

Height-aware representation learning.
The dissertation demonstrates that monocular height estimation can serve as an effective self-supervised pretraining task, improving downstream urban semantic understanding and segmentation performance.

Integrating overhead and street-view observations.
Building beyond satellite imagery, the work introduces scalable methods for retrieving building attributes, such as floor numbers, from crowdsourced street-view imagery, enabling richer geometric and semantic characterization of urban environments.

Together, these contributions advance scalable 3D urban modeling and provide important foundations for future research in urban analytics, digital twins, and Earth Observation. The resulting Global Building Atlas has already attracted broad international attention and serves as a valuable resource for both the research community and real-world applications.

Examination Committee

The examination committee included:

• Prof. Xiaoxiang Zhu (Technical University of Munich, PhD supervisor)
• Prof. Katharina Anders (Technical University of Munich), chair of the committee
• Prof. Richard Bamler (Technical University of Munich), examiner
• Prof. Paolo Ettore Gamba (Università di Pavia), external examiner

We warmly congratulates Dr. Sining Chen on this important academic milestone and looks forward to seeing how his research continues to advance large-scale urban modeling, Earth Observation, and geospatial AI in the years ahead.