Hurricane-Induced Debris Segmentation Dataset Using Aerial Imagery

The Hurricane Debris Segmentation Dataset is a high-resolution collection of 1,242 aerial image crops from NOAA Emergency Response Imagery captured following Hurricane Ian, Hurricane Ike, and Hurricane Ida. The dataset encompasses diverse post-disaster conditions and includes both debris-positive and debris-free examples, with each 50 m × 50 m image crop manually annotated at the pixel level to classify areas as no debris, low-density debris, or high-density debris; annotations were generated through a rigorous multi-annotator consensus process to ensure reliability and minimize subjectivity. This comprehensive resource facilitates the development and benchmarking of automated debris segmentation models, ultimately enabling rapid post-disaster debris quantification and enhancing disaster response, recovery, and resilience planning.