Abstract Post‐hazard regional structural functionality assessment is crucial for community resilience quantification. The current practices performed by reconnaissance experts are time‐consuming and resource‐intensive. Recent studies have proven the feasibility of utilizing deep learning in automatic structural damage assessment. However, integrating damage assessments of transportation systems and building structures portfolios of the whole community remain unexplored. This study proposes a framework, which enables the evaluation of building structural damage and identification of road obstructions. Utilizing a newly annotated high‐resolution dataset with extensive damage classification and semantic segmentation labels, this model is specifically trained for post‐disaster analysis. The effectiveness and generalization of this methodology is demonstrated in Fort Myers Beach for evaluating the aftermath of Hurricane Ian. Besides, the Bayesian active learning is introduced to select the most valuable samples for fine‐tuning when the test images deviate from the characteristics of the training set, which enhances the robustness of the damage classification. Upon the damage identification, the regional functionality can be quantified based on the level of accessibility of households to different services. The results show the developed approach enables emergency responders to quickly use post‐disaster imagery to identify areas in need of assistance, efficiently route around obstructions, coordinate response efforts, and provide situational awareness.
Yang et al. (Thu,) studied this question.