Building damage assessment from satellite images needs better data and evaluation
Revisiting the Task of Building Damage Assessment by Satellite Imaging
Computer Vision and Pattern Recognition
Summary
After natural disasters, quick and accurate maps of damaged buildings are crucial for emergency help. The authors revisit how to assess building damage using satellite images, clarifying how the task should be defined and measured. They highlight problems with existing datasets, evaluation methods, and how current computer models struggle to work well across different disasters and places. Finally, they suggest ways to organize future challenges that improve trust between developers and users.
What this means in practice
- •For emergency response teams: Use clearer task definitions and improved evaluation protocols to better select tools for mapping building damage after disasters.
- •For earth observation platform developers: Design and validate building damage detection systems that generalize across different events and regions using improved data and evaluation guidelines.
A position paper. It proposes an approach and reports no results.
Authors
Anthony Cioppa, Virgile Blua, Sébastien Piérard, Pietro Ceccato, Marc Van Droogenbroeck
Abstract
Natural disasters such as earthquakes, tropical storms, and floods cause extensive damage to buildings, and rapid, reliable damage maps are essential to guide emergency response. Automating this task, referred to as building damage assessment (BDA), from satellite imagery, has become an active research topic at the intersection of remote sensing, computer vision, and machine learning, particularly since supervised deep learning methods were shown to outperform earlier handcrafted approaches. Yet, despite a decade of intense activity and the release of several large-scale datasets, the field lacks a shared foundation. This chapter revisits the task of BDA from first principles, organized around three pillars: data, evaluation, and methods. We first clarify the candidate task for- mulations and argue for a sensor-agnostic, instance-based definition that tolerates missing pre-event imagery. We then review existing datasets along the axes of sensing modality, temporal configuration, and geographic coverage. On the evaluation side, we show that prevailing practices, largely based on a single aggregate score, are theoretically fragile, and we lay the foundations of a more informative, ranking-based, multi-score protocol. We further review methods ranging from task-specific deep networks to recent Earth observation foundation models, noticing that generalization to unseen events and regions remains an open difficulty. Finally, we discuss how these elements can inform the organization of a BDA challenge, including cross-validation, challenge, and private data-sharing setups designed to build trust between developers and users.