xBD
收藏资源简介:
xBD是一个大规模的数据集,旨在推动人道主义援助和灾害恢复研究中的变化检测和建筑损害评估。该数据集由卡内基梅隆大学等多个机构合作创建,提供了多种灾害事件前后的卫星图像,包含建筑多边形、损害等级序数标签及相应的卫星元数据。xBD数据集覆盖了多种灾害类型和地理区域,包含超过800,000个建筑标注,总面积超过45,000平方公里。数据集的创建过程严格,通过与全球灾害响应专家合作,确保了标注的高质量和准确性。xBD数据集主要用于xView 2奖金挑战,旨在促进创建从灾害前后卫星图像评估建筑损害的准确和高效的机器学习模型,以解决资源分配、救援和恢复等任务中的效率问题。
xBD is a large-scale dataset designed to advance change detection and building damage assessment research for humanitarian aid and disaster recovery. This dataset was collaboratively created by multiple institutions including Carnegie Mellon University, providing satellite imagery of multiple disaster events before and after their occurrence, along with building polygons, ordinal labels of damage levels, and corresponding satellite metadata. The xBD dataset covers a wide range of disaster types and geographic regions, contains over 800,000 building annotations, and spans a total area of more than 45,000 square kilometers. The dataset was developed through a rigorous process by collaborating with global disaster response experts to ensure the high quality and accuracy of its annotations. The xBD dataset is primarily used for the xView 2 Prize Challenge, which aims to facilitate the development of accurate and efficient machine learning models for assessing building damage from pre- and post-disaster satellite imagery, to address efficiency issues in tasks such as resource allocation, rescue and recovery.

- 1xBD: A Dataset for Assessing Building Damage from Satellite Imagery卡内基梅隆大学 · 2019年



