DA-SegFormer: Damage-Aware Semantic Segmentation for Fine-Grained Disaster Assessment

2026-05-10

AI Summary

This study addresses the challenge of fine-grained disaster damage assessment in UAV imagery—particularly distinguishing between minor and severe roof damage—where performance is hindered by texture degradation and extreme class imbalance. Building upon the SegFormer architecture, the authors propose a class-aware sampling strategy integrated with an OHEM-Dice Loss mechanism to dynamically emphasize rare and hard-to-classify samples during training. Additionally, a resolution-preserving protocol is employed during inference to retain high-fidelity texture details. Evaluated on the RescueNet dataset, the proposed method achieves a mean Intersection-over-Union (mIoU) of 74.61%, representing a 2.55% improvement over the baseline. Notably, segmentation performance for minor and severe damage classes improves by 11.7% and 21.3%, respectively, demonstrating the effectiveness of the approach in capturing critical fine-grained distinctions.

Abstract

Rapid and accurate damage assessment following natural disasters is critical for effective emergency response. However, identifying fine-grained damage levels (e.g., distinguishing minor from major roof damage) in UAV imagery remains challenging due to the degradation of texture cues during resizing and extreme class imbalance. We propose DA-SegFormer, a damage-aware adaptation of the SegFormer architecture optimized for high-resolution disaster imagery. Our method introduces a Class-Aware Sampling strategy to guarantee exposure to rare damage features, and it integrates Online Hard Example Mining (OHEM) with Dice Loss to dynamically focus on underrepresented classes. In addition, we employ a resolution-preserving inference protocol that maintains native texture details. Evaluated on the RescueNet dataset, DA-SegFormer achieves 74.61\% mIoU, outperforming the baseline by 2.55\%. Notably, our improvements yield double-digit gains in critical damage classes: Minor Damage (+11.7%) and Major Damage (+21.3%).

Problem

• damage assessment
• semantic segmentation
• class imbalance
• UAV imagery
• fine-grained damage

Innovation

• Class-Aware Sampling
• Online Hard Example Mining
• Dice Loss
• Resolution-Preserving Inference
• Fine-Grained Damage Assessment