Beyond Binary Rooftop Mapping: A Four-Class Deep Learning Framework for Green Roof Potential Assessment from Open Swiss Geospatial Data
The paper separates Bern rooftops into existing green roofs, viable green-roof candidates, solar-panel roofs, and flat roofs judged unsuitable.
It builds on Roofpedia with a modified deep convolutional network using aerial imagery, rooftop slope data, and building footprints from public Swisstopo sources. The test case is Bern, Switzerland. The stated aim is to give planners a more realistic map of where green roofs already exist and where expansion is plausible. The author says the open-source setup should transfer to other Swiss cities and beyond. ArXiv · AI/CL/LG's note
It builds on Roofpedia with a modified deep convolutional network using aerial imagery, rooftop slope data, and building footprints from public Swisstopo sources. The test case is Bern, Switzerland. The stated aim is to give planners a more realistic map of where green roofs already exist and where expansion is plausible. The author says the open-source setup should transfer to other Swiss cities and beyond. ArXiv · AI/CL/LG's note
score 4