Road inundation dataset of Hurricane Delta (2020.10) and Hurricane Ian (2022.9)

This Road inundation dataset is constructed based on bi-temporal disaster imagery, which is mainly used for road damage assessment after hurricanes, typhoons, and floods. It contains 1,000 pre- and post-disaster imagery pairs from Hurricanes Delta (October 2020) and Hurricane Ian (September 2022), as well as pixel-level annotation masks of road locations and inundations. The original data comes from Maxar's open data project, and road and inundation information are annotated with human supervision. It can be applied to the training and validation of the intelligent extraction model for road damage. Details for the dataset: Each sample is constructed as pre-disaster and post-disaster images, with 512×512 size and RGB 3-channels. Annotation for each sample includes a binary building localization mask (255: road; 0: background), and a road inundation mask (1: not-flooded road; 2: flooded road; 0: background). Each sample is named as "{disaster event}_{image ID}_{pre/post}_disaster".