Around the Munich University campus we ran a comprehensive, multi-layer extraction of every visible feature — delivering point, polyline and polygon layers that together cover more than 150 distinct classes, from buildings and roads to street furniture and vegetation.
The challenge
The campus needed a highly detailed, geospatially accurate map for asset management, safety analysis, autonomous navigation and planning. No single off-the-shelf tool could extract every feature type at once, and manual digitisation was impossible at that scale — so the client needed an integrated, AI-powered workflow.
One pipeline, every geometry
We combined high-resolution aerial and satellite imagery (with optional mobile-mapping data) and a suite of deep-learning models tailored to each geometry type:
- Points — individual assets detected with sub-metre accuracy.
- Polylines — the road, path and network edges that structure the site.
- Polygons — buildings, surfaces and land-use areas.
One pass yields a complete, structured, 150-class map of the site — the layered, attributed base that campus operations and digital-twin work build on.
