Upcoming

Simulated Airborne Laser Scans

This dataset simulates airborne LiDAR scanning using the Helios++ toolkit and high-quality LoD2 building models from the Bavaria State Mapping Agency. To improve the neural network’s robustness in real-world conditions, noise and inter-building occlusions are deliberately included. The virtual sensor closely simulates the Leica HYPERION2+ system, with a survey flown at 400 m altitude by a virtual Cirrus SR22 aircraft. The training set encompasses 281,571 buildings and over 6.5 billion points from Munich, with an additional 10,000 buildings reserved for evaluation. Each building contains an average of 22,406 points. This dataset has been used for algorithm development in the PolyGNN project.

Oblique Laser Scans Helicopter laser scanning dataset

ALS Scan 1

Images

  • Aerial Orthophotos
  • Street Level Images

Semantic Building Models: Level of Detail 1

LOD1 model of TUM main campus is reconstructed based on software from TU-Delft 3dfier It use the ALS(airborne laser scanning) point cloud and buildings’ footprint.

  • AlS can be downloaded from Bavarian State Mapping Agency via GeoportalBayern
  • Buildings’ footprint can be converted from LoD2 model from GeoportalBayern by using FME
Building Models in LOD1