This thesis studies how much rooftop solar PV could realistically be installed on the buildings of Turin's historic centre, and whether the standard Italian method gives reliable answers building by building. Italian planning often estimates usable PV area by multiplying a building's footprint by a fixed national coefficient of 0.35. This works at city scale but has not really been tested for individual buildings, which is the scale at which incentive and heritage decisions are made. To test it, I built a processing chain on the city's 2018 aerial survey. The photogrammetric point cloud (about 654 million points) was classified into ground, building, and tree using a deep-learning segmentation model, reaching about 98.5% accuracy.