Improved MambdaBDA Framework for Robust Building Damage Assessment Across Disaster Domains

2026-03-01

AI Summary

This study addresses key challenges in post-disaster satellite imagery-based building damage assessment, including severe class imbalance, background clutter, and limited cross-disaster generalization. To overcome these issues, the authors introduce three modular enhancements to the ChangeMamba architecture: Focal Loss to mitigate class imbalance, a lightweight attention gating mechanism to suppress irrelevant background features, and a compact feature alignment module to enforce consistency between pre- and post-disaster feature spaces. The proposed approach significantly improves model robustness and cross-domain generalization, achieving performance gains of 0.8%–5% on in-domain evaluations and up to 27% improvement in cross-domain settings across multiple datasets, including xBD, Pakistan floods, Turkey earthquakes, and Hurricane Ida.

Technology Category

• Computer Vision: Multi-modal Vision
• Machine Learning: Multi-class/Multi-label Learning & Extreme Classification
• Constraint Satisfaction and Optimization: Satisfiability Modulo Theories

Application Category

• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• Web Mining and Content Analysis: Large pretrained models with web data
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs

Abstract

Reliable post-disaster building damage assessment (BDA) from satellite imagery is hindered by severe class imbalance, background clutter, and domain shift across disaster types and geographies. In this work, we address these problems and explore ways to improve the MambaBDA, the BDA network of ChangeMamba architecture, one of the most successful BDA models. The approach enhances the MambaBDA with three modular components: (i) Focal Loss to mitigate class imbalance damage classification, (ii) lightweight Attention Gates to suppress irrelevant context, and (iii) a compact Alignment Module to spatially warp pre-event features toward post-event content before decoding. We experiment on multiple satellite imagery datasets, including xBD, Pakistan Flooding, Turkey Earthquake, and Ida Hurricane, and conduct in-domain and crossdataset tests. The proposed modular enhancements yield consistent improvements over the baseline model, with 0.8% to 5% performance gains in-domain, and up to 27% on unseen disasters. This indicates that the proposed enhancements are especially beneficial for the generalization capability of the system.

Problem

• class imbalance
• background clutter
• domain shift
• building damage assessment
• satellite imagery

Innovation

• MambaBDA
• Focal Loss
• Attention Gates
• Feature Alignment
• Cross-domain Generalization