Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

FMARS: Annotating Remote Sensing Images for Disaster Management using Foundation Models

Edoardo Arnaudo, Jacopo Lungo Vaschetti, Lorenzo Innocenti, Luca Barco, Davide Lisi, Vanina Fissore, Claudio Rossi

TL;DR

The paper tackles the challenge of scarce high-quality annotations in Very-High-Resolution remote sensing data for disaster management. It introduces FMARS, a pipeline that combines Segment Anything and GroundingDINO with pre-event Maxar imagery to automatically generate large-scale labels for three classes (buildings, roads, high vegetation), stored as vector polygons across 19 disaster events totaling $127{,}134\ km^2$ and more than $25{,}000{,}000$ annotations. To ensure robustness given label noise, the authors train SegFormer-based segmentation models using Unsupervised Domain Adaptation methods (DAFormer and MIC) and demonstrate that these approaches can match or exceed manual GT performance on a subset of tiles. The work highlights the practical impact of foundation models for rapid, scalable RS annotation, enabling robust downstream models for disaster response, with code and data made available for community use ($https://github.com/links-ads/igarss-fmars$).

Abstract

Very-High Resolution (VHR) remote sensing imagery is increasingly accessible, but often lacks annotations for effective machine learning applications. Recent foundation models like GroundingDINO and Segment Anything (SAM) provide opportunities to automatically generate annotations. This study introduces FMARS (Foundation Model Annotations in Remote Sensing), a methodology leveraging VHR imagery and foundation models for fast and robust annotation. We focus on disaster management and provide a large-scale dataset with labels obtained from pre-event imagery over 19 disaster events, derived from the Maxar Open Data initiative. We train segmentation models on the generated labels, using Unsupervised Domain Adaptation (UDA) techniques to increase transferability to real-world scenarios. Our results demonstrate the effectiveness of leveraging foundation models to automatically annotate remote sensing data at scale, enabling robust downstream models for critical applications. Code and dataset are available at \url{https://github.com/links-ads/igarss-fmars}.

FMARS: Annotating Remote Sensing Images for Disaster Management using Foundation Models

TL;DR

The paper tackles the challenge of scarce high-quality annotations in Very-High-Resolution remote sensing data for disaster management. It introduces FMARS, a pipeline that combines Segment Anything and GroundingDINO with pre-event Maxar imagery to automatically generate large-scale labels for three classes (buildings, roads, high vegetation), stored as vector polygons across 19 disaster events totaling and more than annotations. To ensure robustness given label noise, the authors train SegFormer-based segmentation models using Unsupervised Domain Adaptation methods (DAFormer and MIC) and demonstrate that these approaches can match or exceed manual GT performance on a subset of tiles. The work highlights the practical impact of foundation models for rapid, scalable RS annotation, enabling robust downstream models for disaster response, with code and data made available for community use ().

Abstract

Very-High Resolution (VHR) remote sensing imagery is increasingly accessible, but often lacks annotations for effective machine learning applications. Recent foundation models like GroundingDINO and Segment Anything (SAM) provide opportunities to automatically generate annotations. This study introduces FMARS (Foundation Model Annotations in Remote Sensing), a methodology leveraging VHR imagery and foundation models for fast and robust annotation. We focus on disaster management and provide a large-scale dataset with labels obtained from pre-event imagery over 19 disaster events, derived from the Maxar Open Data initiative. We train segmentation models on the generated labels, using Unsupervised Domain Adaptation (UDA) techniques to increase transferability to real-world scenarios. Our results demonstrate the effectiveness of leveraging foundation models to automatically annotate remote sensing data at scale, enabling robust downstream models for critical applications. Code and dataset are available at \url{https://github.com/links-ads/igarss-fmars}.
Paper Structure (8 sections, 2 figures, 4 tables)

This paper contains 8 sections, 2 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Materials and Methods
  4. Foundation Models
  5. Data Sources
  6. Annotation Workflow
  7. Experiments
  8. Conclusions

Figures (2)

  • Figure 1: Annotation workflow adopted for the three selected classes. Each class is treated separately, with its own prompt construction pipeline, while the segmentation masks are extracted from the same image embeddings, and merged together in a single output.
  • Figure 2: Qualitative results obtained over two example areas, namely USA (top) and Gambia (bottom). from left to right: RGB image, DAFormer, MIC, and FMARS ground truth. Best viewed zoomed in.