CRASAR-U-DROIDs: A Large Scale Benchmark Dataset for Building Alignment and Damage Assessment in Georectified sUAS Imagery

TL;DR

CRASAR-U-DROIDs addresses the need for high-resolution, georectified sUAS disaster imagery with consistent damage labels by delivering 52 orthomosaics across 10 disasters, 21,716 building polygons, and 7,880 spatial alignment adjustments. The dataset aligns building footprints to imagery, uses the Joint Damage Scale for damage annotation, and provides a public, provenance-rich resource suitable for cross-platform benchmarking with satellite datasets like xBD. Key contributions include the largest georectified sUAS disaster dataset to date, explicit non-uniform spatial alignment correction via adjustment annotations, and a design that enables transferability of satellite-based models to sUAS imagery. This resource has significant practical impact for emergency management and CV/ML research by supporting operationally faithful disaster damage assessment and cross-domain model validation, with ongoing updates and extensions planned.

Abstract

This document presents the Center for Robot Assisted Search And Rescue - Uncrewed Aerial Systems - Disaster Response Overhead Inspection Dataset (CRASAR-U-DROIDs) for building damage assessment and spatial alignment collected from small uncrewed aerial systems (sUAS) geospatial imagery. This dataset is motivated by the increasing use of sUAS in disaster response and the lack of previous work in utilizing high-resolution geospatial sUAS imagery for machine learning and computer vision models, the lack of alignment with operational use cases, and with hopes of enabling further investigations between sUAS and satellite imagery. The CRASAR-U-DRIODs dataset consists of fifty-two (52) orthomosaics from ten (10) federally declared disasters (Hurricane Ian, Hurricane Ida, Hurricane Harvey, Hurricane Idalia, Hurricane Laura, Hurricane Michael, Musset Bayou Fire, Mayfield Tornado, Kilauea Eruption, and Champlain Towers Collapse) spanning 67.98 square kilometers (26.245 square miles), containing 21,716 building polygons and damage labels, and 7,880 adjustment annotations. The imagery was tiled and presented in conjunction with overlaid building polygons to a pool of 130 annotators who provided human judgments of damage according to the Joint Damage Scale. These annotations were then reviewed via a two-stage review process in which building polygon damage labels were first reviewed individually and then again by committee. Additionally, the building polygons have been aligned spatially to precisely overlap with the imagery to enable more performant machine learning models to be trained. It appears that CRASAR-U-DRIODs is the largest labeled dataset of sUAS orthomosaic imagery.

Figures (14)

• Figure 1: The composition of the CRASAR-U-DROIDs dataset split by disaster events showing the total area in square kilometers, the total number of pixels in gigapixels, the total number of buildings, and the total number of adjustments. The disaster events are ordered in decreasing order of total area.
• Figure 1: Four sample orthomosaics from the fifty-two (52) orthomosaics within the CRASAR-U-DROIDs dataset.
• Figure 2: Table comparing other aerial computer vision datasets collected from disasters to the CRASAR-U-DROIDs dataset. Datasets are ordered first by the sensing platform (sUAS, Manned Aircraft, Satellite), followed by pixel count to indicate approximate sizes. Within the subset of sUAS datasets, this work represents the largest dataset in terms of pixels and spatial area. Further, it is the only sUAS disaster dataset that is orthorectified. († Denotes an author recommendation to utilize the smaller, resized, version of their dataset, ** Indicates that the released data has had the coordinate reference system transform removed.)
• Figure 2: The annotation and review workflow used to arrive at the joint damage scale (JDS) labels for the building polygons released here. For each building polygon, a single label was generated, which was then reviewed twice, once by a single reviewer and then a second time by a committee of the reviewers.
• Figure 3: Comparison of other sUAS computer vision datasets collected from disasters to the CRASAR-U-DROIDs dataset based on the type of disaster damage present in each dataset.