PVRoof-SH: A Rooftop-State Semantic Segmentation Dataset for Urban Photovoltaic Mapping

PVRoof-SH is a high-resolution remote sensing dataset developed for rooftop-state semantic segmentation and photovoltaic (PV) rooftop mapping in Shanghai, China. Unlike conventional PV datasets that delineate only solar panel footprints, PVRoof-SH assigns a semantic state to the entire rooftop according to the presence of visible PV installations. Its pixel-level annotations distinguish three classes: background, non-PV rooftops, and PV-equipped rooftops.