Ian-BD: pre- and post-disaster high-resolution aerial imagery for building damage assessment from Hurricane Ian.

Ian-BD is a high-resolution (0.3m/pixel) dataset comprising 101 pre- and post-disaster aerial image pairs collected following Hurricane Ian, a Category 4 hurricane that landed in Florida in September 2022. Pre-disaster imagery was acquired in early September 2022 from Lee County's Aerial Imagery database, while post-disaster imagery was obtained from NOAA's Hurricane Ian dataset on September 29th and 30th, 2022. Covering 80% of Lee County, with complete coverage of Sanibel Island, the dataset provides a valuable resource for studying the impacts of this significant natural disaster. Building footprints were delineated on pre-disaster images and transferred to post-disaster images, where four damage categories were assigned: no damage, minor damage, major damage, and destroyed. An independent team conducted quality control on all annotations. The dataset leverages existing Florida building databases, reducing the reliance on expensive commercial labeling software, and also addresses challenges in identifying buildings obscured by tree canopy in pre-disaster imagery. Following the success of the xBD and Ida-BD datasets, the Ian-BD dataset will significantly contribute to advancing building damage prediction models by enabling the development of more accurate and robust models tailored explicitly to high-resolution aerial imagery and contemporary building structures.