In the context of recent, highly destructive conflicts in Gaza and Ukraine,
reliable estimates of building damage are essential for an informed public
discourse, human rights monitoring, and humanitarian aid provision. Given the
contentious nature of conflict damage assessment, these estimates must be fully
reproducible, explainable, and derived from open access data. This paper
introduces a new method for building damage detection-- the Pixel-Wise T-Test
(PWTT)-- that satisfies these conditions. Using a combination of
freely-available synthetic aperture radar imagery and statistical change
detection, the PWTT generates accurate conflict damage estimates across a wide
area at regular time intervals. Accuracy is assessed using an original dataset
of over half a million labeled building footprints spanning 12 cities across
Ukraine, Palestine, Syria, and Iraq. Despite being simple and lightweight, the
algorithm achieves building-level accuracy statistics (AUC=0.88 across Ukraine,
0.81 in Gaza) rivalling state of the art methods that use deep learning and
high resolution imagery. The workflow is open source and deployed entirely
within the Google Earth Engine environment, allowing for the generation of
interactive Battle Damage Dashboards for Ukraine and Gaza that update in
near-real time, allowing the public and humanitarian practitioners to
immediately get estimates of damaged buildings in a given area.

通过使用 Pixel-Wise T-Test 算法结合合成孔径雷达成像和统计变化检测的方法，该研究提供了在广阔区域定期获得准确冲突破坏估计的新方法，并在乌克兰、加沙、叙利亚和伊拉克等地的 12 个城市上进行了评估，达到了与使用深度学习和高分辨率成像方法相媲美的建筑级准确性统计水平（在乌克兰是 AUC=0.88，在加沙是 0.81）。该 workflow 是开源的，并完全部署在 Google Earth Engine 环境中，可生成乌克兰和加沙的互动式战斗损伤仪表板，实时更新，使公众和人道主义从业者能够立即获取给定区域受损建筑的估计。

通过基于像素的 T - 检验在 Sentinel-1 影像上的开放性损伤检测

Open Access Battle Damage Detection via Pixel-Wise T-Test on Sentinel-1  Imagery

We construct a strong baseline method for building damage detection by
starting with the highly-engineered winning solution of the xView2 competition,
and gradually stripping away components. This way, we obtain a much simpler
method, while retaining adequate performance. We expect the simplified solution
to be more widely and easily applicable. This expectation is based on the
reduced complexity, as well as the fact that we choose hyperparameters based on
simple heuristics, that transfer to other datasets. We then re-arrange the
xView2 dataset splits such that the test locations are not seen during
training, contrary to the competition setup. In this setting, we find that both
the complex and the simplified model fail to generalize to unseen locations.
Analyzing the dataset indicates that this failure to generalize is not only a
model-based problem, but that the difficulty might also be influenced by the
unequal class distributions between events.
Code, including the baseline model, is available under
this https URL

我们构建了一种强大的基准方法用于建筑损坏检测，通过使用 xView2 比赛的高度工程化的获奖解决方案，并逐步去除组件，从而获得一个更简单的方法，同时保持足够的性能。我们的预期是，简化的解决方案更加普适且易于应用，这基于简化解决方案的复杂性降低，以及我们选择基于简单启发式方法的超参数，可以迁移到其他数据集。然而，在将 xView2 数据集重新划分的设置中，使得测试位置在训练过程中是不可见的，我们发现复杂模型和简化模型都无法推广到未见过的位置。通过分析数据集，发现该泛化失败不仅是模型本身的问题，而且可能也受到事件之间不平等的类分布的影响。

xBD 数据集上建筑损坏检测的简明、强大基线

A simple, strong baseline for building damage detection on the xBD  dataset

Existing Building Damage Detection (BDD) methods always require
labour-intensive pixel-level annotations of buildings and their conditions,
hence largely limiting their applications. In this paper, we investigate a
challenging yet practical scenario of BDD, Unsupervised Building Damage
Detection (U-BDD), where only unlabelled pre- and post-disaster satellite image
pairs are provided. As a pilot study, we have first proposed an advanced U-BDD
baseline that leverages pre-trained vision-language foundation models (i.e.,
Grounding DINO, SAM and CLIP) to address the U-BDD task. However, the apparent
domain gap between satellite and generic images causes low confidence in the
foundation models used to identify buildings and their damages. In response, we
further present a novel self-supervised framework, U-BDD++, which improves upon
the U-BDD baseline by addressing domain-specific issues associated with
satellite imagery. Furthermore, the new Building Proposal Generation (BPG)
module and the CLIP-enabled noisy Building Proposal Selection (CLIP-BPS) module
in U-BDD++ ensure high-quality self-training. Extensive experiments on the
widely used building damage assessment benchmark demonstrate the effectiveness
of the proposed method for unsupervised building damage detection. The
presented annotation-free and foundation model-based paradigm ensures an
efficient learning phase. This study opens a new direction for real-world BDD
and sets a strong baseline for future research.

在这篇论文中，我们研究了一种具有挑战性但实用的无监督建筑损害检测（U-BDD）场景，该场景下只提供了未标记的灾后和灾前卫星图像对。我们首先提出了一种先进的 U-BDD 基线模型，该模型利用预训练的视觉 - 语言基础模型来解决 U-BDD 任务。此外，我们还提出了一种新颖的自监督框架 U-BDD++，它通过解决与卫星图像相关的领域问题来改进 U-BDD 基线模型。通过对广泛使用的建筑损害评估基准进行大量实验证明了所提出的方法在无监督建筑损害检测方面的有效性。这项研究为真实世界的 BDD 开辟了新的方向，并为未来的研究奠定了坚实的基线。