Kullervo/BRIGHT

---
language:
- en
license: cc-by-sa-4.0
task_categories:
- image-segmentation
- feature-extraction
- zero-shot-classification
size_categories:
- 1B<n<10B
tags:
- earth-observation
- remote-sensing
- disaster-response
- artificial-intelligence
- building-damage-mapping
pretty_name: Bright
---

**Overview**
* BRIGHT is the first open-access, globally distributed, event-diverse multimodal dataset specifically curated to support AI-based disaster response.
* It covers five types of natural disasters and two types of man-made disasters across 14 regions worldwide, with a particular focus on developing countries.
* About 4,200 paired optical and SAR images containing over 380,000 building instances in BRIGHT, with a spatial resolution between 0.3 and 1 meters, provides detailed representations of individual buildings.
<p align="center">
<img src="./overall.jpg" alt="accuracy" width="97%">
</p>

**CVPR 2026 Workshop Competition (New!)**
* BRIGHT serves as the official dataset for [Monitoring the World Through an Imperfect Lens (MONTI)](https://sites.google.com/view/monti2026/home) in conjunction with CVPR 2026, extending the benchmark to **instance-level building damage segmentation**. We provide instance-level annotations for training and validation sets. Participants can follow the steps below to get started.
* Please download **pre-event.zip**, **post-event.zip**, and the instance-level labels **cvprw2026_train_val_instance_labels.zip**, and unzip them.
* For baseline code and submission instructions, please refer to our Github repo.
* Please submit your results to the Codabench platform for evaluation. The official leaderboard is located on the [Codabench competition page](https://www.codabench.org/competitions/15134/).

**Benchmark for building damage assessment**
* Please download **pre-event.zip**, **post-event.zip**, and **target.zip**. Note that for the optical pre-event data in Ukraine, Myanmar, and Mexico, please follow our [instructions/tutorials](https://github.com/ChenHongruixuan/BRIGHT/blob/master/tutorial.md) to download.
* For the benchmark code and evaluation protocal for supervised building damage assessment, cross-event transfer, and unsupervised multimodal change detection, please see our [Github repo](https://github.com/ChenHongruixuan/BRIGHT).
* You can download models' checkpoints in this [repo](https://zenodo.org/records/15349462).

**Unsupervised multimodal image matching**
* BRIGHT supports the evaluation of Unsupervised Multimodal Image Matching (UMIM) algorithms for their performance in large-scale disaster scenarios. Please download data with the prefix "**umim**", such as **umim_noto_earthquake.zip**, and use our [code](https://github.com/ChenHongruixuan/BRIGHT) to test the exsiting algorithms' performance.

**IEEE GRSS Data Fusion Contest 2025 (Closed, All Data Available)**
* BRIGHT also serves as the official dataset of [IEEE GRSS DFC 2025 Track II](https://www.grss-ieee.org/technical-committees/image-analysis-and-data-fusion/). Now, DFC 25 is over. We recommend using the full version of the dataset along with the corresponding split names provided in our [Github repo](https://github.com/ChenHongruixuan/BRIGHT). Yet, we also retain the original files used in DFC 2025 for download.
* Please download **dfc25_track2_trainval.zip** and unzip it. It contains training images & labels and validation images for the development phase.
* Please download **dfc25_track2_test.zip** and unzip it. It contains test images for the final test phase.
* Please download **dfc25_track2_val_labels.zip** for validation labels, redownload **dfc25_track2_test_new.zip** for test images with geo-coordinates and **dfc25_track2_test_labels.zip** for testing labels.
* Benchmark code related to the DFC 2025 can be found at this [Github repo](https://github.com/ChenHongruixuan/BRIGHT). 
* The official leaderboard is located on the [Codalab-DFC2025-Track II](https://codalab.lisn.upsaclay.fr/competitions/21122) page.

**Paper & Reference**

Details of BRIGHT can be refer to our [paper](https://essd.copernicus.org/articles/17/6217/2025/essd-17-6217-2025.html).

If BRIGHT is useful to research, please kindly consider cite our paper
```
@Article{Chen2025Bright,
AUTHOR = {Chen, H. and Song, J. and Dietrich, O. and Broni-Bediako, C. and Xuan, W. and Wang, J. and Shao, X. and Wei, Y. and Xia, J. and Lan, C. and Schindler, K. and Yokoya, N.},
TITLE = {\textsc{Bright}: a globally distributed multimodal building damage assessment dataset with very-high-resolution for all-weather disaster response},
JOURNAL = {Earth System Science Data},
VOLUME = {17},
YEAR = {2025},
NUMBER = {11},
PAGES = {6217--6253},
DOI = {10.5194/essd-17-6217-2025}
}
```

**License**
* Label data of BRIGHT are provided under the same license as the optical images, which varies with different events.
* With the exception of two events, Hawaii-wildfire-2023 and La Palma-volcano eruption-2021, all optical images are from [Maxar Open Data Program](https://www.maxar.com/open-data), following CC-BY-NC-4.0 license. The optical images related to Hawaii-wildifire-2023 are from [High-Resolution Orthoimagery project](https://coast.noaa.gov/digitalcoast/data/highresortho.html) of NOAA Office for Coastal Management. The optical images related to La Palma-volcano eruption-2021 are from IGN (Spain) following CC-BY 4.0 license. 
* The SAR images of BRIGHT is provided by [Capella Open Data Gallery](https://www.capellaspace.com/earth-observation/gallery) and [Umbra Space Open Data Program](https://umbra.space/open-data/), following CC-BY-4.0 license.

---
语言：
- 英语
许可协议：CC BY-SA 4.0（知识共享署名-相同方式共享4.0）
任务类别：
- 图像分割
- 特征提取
- 零样本（Zero-shot）分类
规模类别：
- 10亿 < 样本数 < 100亿
标签：
- 地球观测
- 遥感
- 灾害响应
- 人工智能
- 建筑物损伤制图
展示名称：Bright
---

**概述**
* BRIGHT是首个开源获取、全球分布、事件多样化的多模态数据集，专门为支持基于人工智能的灾害响应工作而打造。
* 该数据集覆盖全球14个区域的5类自然灾害与2类人为灾害，尤其聚焦发展中国家。
* BRIGHT包含约4200对光学与合成孔径雷达（Synthetic Aperture Radar, SAR）图像，涵盖超过38万个建筑物实例，空间分辨率介于0.3至1米之间，可精准呈现单栋建筑物的细节特征。

<p align="center">
<img src="./overall.jpg" alt="数据集概览" width="97%">
</p>

**CVPR 2026 专题研讨会竞赛（新增！）**
* BRIGHT作为与2026年国际计算机视觉与模式识别会议（CVPR 2026）联合举办的「透过不完美镜头监测全球（MONTI）」竞赛的官方数据集，将基准任务拓展至**实例级建筑物损伤分割**。我们为训练集与验证集提供实例级标注，参赛选手可按照以下步骤快速上手。
* 请下载**灾前.zip**、**灾后.zip**以及实例级标注文件**cvprw2026_train_val_instance_labels.zip**并解压。
* 关于基线代码与提交指南，请参阅我们的GitHub仓库。
* 请将结果提交至Codabench平台进行评估，官方排行榜位于[Codabench竞赛页面](https://www.codabench.org/competitions/15134/)。

**建筑物损伤评估基准**
* 请下载**灾前.zip**、**灾后.zip**以及**目标标签.zip**。请注意，乌克兰、缅甸与墨西哥的光学灾前数据，请按照我们的[教程/指南](https://github.com/ChenHongruixuan/BRIGHT/blob/master/tutorial.md)进行下载。
* 针对有监督建筑物损伤评估、跨事件迁移以及无监督多模态变化检测的基准代码与评估协议，请参阅我们的[GitHub仓库](https://github.com/ChenHongruixuan/BRIGHT)。
* 您可在该[仓库](https://zenodo.org/records/15349462)中下载模型检查点（checkpoint）。

**无监督多模态图像匹配**
* BRIGHT可用于评估无监督多模态图像匹配（Unsupervised Multimodal Image Matching, UMIM）算法在大规模灾害场景下的性能。请下载前缀为“**umim**”的数据集（例如**umim_noto_earthquake.zip**），并使用我们的[代码](https://github.com/ChenHongruixuan/BRIGHT)测试现有算法的表现。

**IEEE GRSS 2025年数据融合竞赛（已结束，所有数据均可获取）**
* BRIGHT同时作为[IEEE地球科学与遥感学会（IEEE GRSS）2025年数据融合竞赛赛道二](https://www.grss-ieee.org/technical-committees/image-analysis-and-data-fusion/)的官方数据集。目前该竞赛已结束，我们建议使用完整版本的数据集以及我们[GitHub仓库](https://github.com/ChenHongruixuan/BRIGHT)中提供的对应划分方式，同时我们也保留了竞赛中使用的原始文件供下载。
* 请下载**dfc25_track2_trainval.zip**并解压，该文件包含开发阶段的训练图像、标签与验证图像。
* 请下载**dfc25_track2_test.zip**并解压，该文件包含最终测试阶段的测试图像。
* 请下载**dfc25_track2_val_labels.zip**获取验证集标签，重新下载**dfc25_track2_test_new.zip**获取带有地理坐标的测试图像，以及**dfc25_track2_test_labels.zip**获取测试集标签。
* 与本次竞赛相关的基准代码可在该[GitHub仓库](https://github.com/ChenHongruixuan/BRIGHT)中获取。
* 官方排行榜位于[Codalab-DFC2025-赛道二](https://codalab.lisn.upsaclay.fr/competitions/21122)页面。

**论文与引用**
* BRIGHT的详细信息可参阅我们的[论文](https://essd.copernicus.org/articles/17/6217/2025/essd-17-6217-2025.html)。
* 若BRIGHT对您的研究有所帮助，请考虑引用我们的论文：

@Article{Chen2025Bright,
AUTHOR = {Chen, H. and Song, J. and Dietrich, O. and Broni-Bediako, C. and Xuan, W. and Wang, J. and Shao, X. and Wei, Y. and Xia, J. and Lan, C. and Schindler, K. and Yokoya, N.},
TITLE = { extsc{Bright}: a globally distributed multimodal building damage assessment dataset with very-high-resolution for all-weather disaster response},
JOURNAL = {Earth System Science Data},
VOLUME = {17},
YEAR = {2025},
NUMBER = {11},
PAGES = {6217--6253},
DOI = {10.5194/essd-17-6217-2025}
}

**许可协议**
* BRIGHT的标签数据与光学图像采用相同的许可协议，不同事件的图像许可协议有所差异。
* 除2023年夏威夷山火与2021年拉帕尔马火山喷发两个事件外，所有光学图像均来自[Maxar开放数据计划](https://www.maxar.com/open-data)，采用CC-BY-NC-4.0许可协议。2023年夏威夷山火相关的光学图像来自美国国家海洋和大气管理局（NOAA）海岸管理办公室的[高分辨率正射影像项目](https://coast.noaa.gov/digitalcoast/data/highresortho.html)。2021年拉帕尔马火山喷发相关的光学图像来自西班牙国家地理研究所（IGN），采用CC-BY 4.0许可协议。
* BRIGHT的合成孔径雷达（SAR）图像由[Capella开放数据图库](https://www.capellaspace.com/earth-observation/gallery)与[Umbra Space开放数据计划](https://umbra.space/open-data/)提供，采用CC-BY-4.0许可协议。