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

TL;DR

DA-SegFormer is a damage-aware segmentation model designed for fine-grained disaster assessment in UAV imagery, employing class-aware sampling, OHEM with Dice Loss, and resolution-preserving inference to improve damage classification accuracy.

Contribution

It introduces a novel damage-aware adaptation of SegFormer with strategies to handle class imbalance and texture preservation in high-resolution disaster imagery.

Findings

Achieves 74.61% mIoU on RescueNet dataset, surpassing baseline by 2.55%.

Double-digit improvements in Minor Damage (+11.7%) and Major Damage (+21.3%) classes.

Effective in distinguishing fine-grained damage levels in UAV imagery.

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%).