Deep learning transferability across disaster types for UAS imagery based building damage assessment

In recent years, natural disaster frequency and intensity have increased worldwide, resulting in significant economic losses, with building damages accounting for a substantial portion. Post-disaster response plans generally include acquiring detailed inventories of building damage for loss estimation; however, these assessments can be highly subjective, require substantial time, and can expose the inspectors to unsafe environments. Automation of building damage assessment by applying deep learning combined with advanced remote sensing technology is currently an active research topic to overcome the limitations of traditional assessments. Nevertheless, these efforts are hindered by the limited amount of high-quality training datasets available for each disaster type (e.g., hurricane, wildfire). Buildings damaged by different disaster types may show distinct damage patterns due to differing damage mechanisms, posing challenges to data integration and model development across disaster types. To investigate these issues, this study explores the interrelationship between wildfire and hurricane data by developing models suited to wildfire and hurricane datasets both individually and jointly as well as combining several backbones and deep learning models. Our approach includes semantic segmentation for pixel-level damage assessment and analyzing model sensitivity with increasing amounts of training data through transfer learning. Ultimately, this study provides a solution to the limited data available to train building damage assessment deep learning models by providing a comparative analysis of the inter-applicability of wildfire and hurricane data. A notable finding is that when using a small portion of data through transfer learning, data and deep learning models from the other disaster types can be leveraged.