DoriaNET: A visual dataset from Hurricane Dorian for post-disaster building damage assessment

DoriaNET is a synthetic dataset for unmanned aerial vehicle (UAV)-based post-hurricane building damage assessment. DoriaNET consists of damage state annotations in accordance with FEMA Hurricane Model HAZUS-MH MR3 Manual. For each building, a damage state is manually assigned by closely matching the quantitative damage description of that building. Overall, DoriaNET provides the following information: (1) global coordinates (latitude and longitude) of each building appearing in the aerial footage; (2) local coordinates with pixel-level building boundaries (masks) in video frames; (3) wind-induced damage state for each building in accordance with FEMA Manual; (4) number of stories; and (5) an ordinal score representing damage annotation effort (easy, moderate, or difficult) for each labeled building.