Post-Disaster Affected Area Segmentation with a Vision Transformer (ViT)-based EVAP Model using Sentinel-2 and Formosat-5 Imagery

TL;DR

This paper introduces a ViT-based deep learning framework that refines disaster-affected area segmentation using weakly supervised training with Sentinel-2 and Formosat-5 imagery, improving mapping accuracy in limited supervision scenarios.

Contribution

The study presents a novel ViT-based model with multi-stage loss and weak supervision techniques for disaster segmentation, leveraging PCA-based label expansion and multi-band satellite data.

Findings

Enhanced segmentation smoothness and reliability in case studies

Effective weakly supervised training with limited annotations

Improved spatial coherence compared to traditional methods

Abstract

We propose a vision transformer (ViT)-based deep learning framework to refine disaster-affected area segmentation from remote sensing imagery, aiming to support and enhance the Emergent Value Added Product (EVAP) developed by the Taiwan Space Agency (TASA). The process starts with a small set of manually annotated regions. We then apply principal component analysis (PCA)-based feature space analysis and construct a confidence index (CI) to expand these labels, producing a weakly supervised training set. These expanded labels are then used to train ViT-based encoder-decoder models with multi-band inputs from Sentinel-2 and Formosat-5 imagery. Our architecture supports multiple decoder variants and multi-stage loss strategies to improve performance under limited supervision. During the evaluation, model predictions are compared with higher-resolution EVAP output to assess spatial coherence and segmentation consistency. Case studies on the 2022 Poyang Lake drought and the 2023 Rhodes wildfire demonstrate that our framework improves the smoothness and reliability of segmentation results, offering a scalable approach for disaster mapping when accurate ground truth is unavailable.