InfraFlood-NC

The InfraFlood-NC dataset provides infrastructure-specific binary flood extent annotations across six urban areas in North Carolina severely impacted by Hurricanes Matthew and Florence. Derived from the high-resolution (1.5 cm to 25 cm) DeepFlood dataset, this data extracts spatial flood masks and pairs them with vector-based building footprints and road networks. The methodology employs raster reclassification to harmonize complex land-cover classes into a binary Wet/Dry framework. Zonal statistics were used to calculate maximum inundation presence per structure, while spatial joins integrated critical facility classifications from Homeland Infrastructure Foundation-Level Data (HIFLD). To maximize usability, the resulting dataset is structured into 10 spatial divisions and formatted in both lightweight GeoJSON for training Visual Question Answering (VQA) deep learning models, and standard Shapefiles for traditional GIS risk assessment.