Generalizable Disaster Damage Assessment via Change Detection with Vision Foundation Model

2024-06-12

arXiv.org

AI Summary

To address the challenge of building-level disaster damage assessment in cross-regional and multi-hazard scenarios without ground-truth labels in target areas, this paper proposes an unsupervised domain adaptation method based on vision foundation models. The method jointly leverages a task-specific source-domain model and a task-agnostic segmentation model to generate high-quality pseudo-labels, and introduces a two-stage change refinement mechanism operating at both pixel- and image-levels to effectively mitigate geographical and hazard-type distribution shifts. It requires no annotated data from the target region—only high-resolution satellite imagery. Extensive evaluations across diverse terrains (North America, Asia, Middle East) and hazards (wildfire, hurricane, tsunami) demonstrate robust generalization. In the 2023 Turkey earthquake case study, the method achieves precise damage localization with an F1-score of 0.87, significantly outperforming existing unsupervised approaches.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Reasoning under Uncertainty: Other Foundations of Reasoning under Uncertainty

Application Category

• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

The increasing frequency and intensity of natural disasters call for rapid and accurate damage assessment. In response, disaster benchmark datasets from high-resolution satellite imagery have been constructed to develop methods for detecting damaged areas. However, these methods face significant challenges when applied to previously unseen regions due to the limited geographical and disaster-type diversity in the existing datasets. We introduce DAVI (Disaster Assessment with VIsion foundation model), a novel approach that addresses domain disparities and detects structural damage at the building level without requiring ground-truth labels for target regions. DAVI combines task-specific knowledge from a model trained on source regions with task-agnostic knowledge from an image segmentation model to generate pseudo labels indicating potential damage in target regions. It then utilizes a two-stage refinement process, which operate at both pixel and image levels, to accurately identify changes in disaster-affected areas. Our evaluation, including a case study on the 2023 T""urkiye earthquake, demonstrates that our model achieves exceptional performance across diverse terrains (e.g., North America, Asia, and the Middle East) and disaster types (e.g., wildfires, hurricanes, and tsunamis). This confirms its robustness in disaster assessment without dependence on ground-truth labels and highlights its practical applicability.

Problem

• Visual Baseline Models
• Unseen Locations
• Disaster Damage Assessment

Innovation

• DAVI
• Two-stage Detection Strategy
• Image Segmentation