Hurricane-Damaged Building Benchmark Dataset

Hurricane-Damaged Building Benchmark Dataset是由纽约大学和华盛顿大学等机构合作创建，旨在从飓风后的遥感图像中自动检测损坏建筑。数据集包含18,474条损坏建筑的标注，来源于DigitalGlobe卫星图像和NOAA航空图像。创建过程中，利用了TOMNOD项目的众包标注和FEMA的损坏评估数据。该数据集主要用于训练和测试自动识别损坏建筑的物体检测模型，以提高灾害响应的效率和准确性。

The Hurricane-Damaged Building Benchmark Dataset was collaboratively developed by institutions including New York University and the University of Washington, with the objective of automatically detecting damaged buildings from post-hurricane remote sensing imagery. The dataset contains 18,474 annotated instances of damaged buildings, sourced from DigitalGlobe satellite imagery and NOAA aerial photographs. During its development, crowdsourced annotations from the TOMNOD project and damage assessment data from FEMA were utilized. This dataset is primarily used for training and testing object detection models that automatically identify damaged buildings, aiming to improve the efficiency and accuracy of disaster response.

搜集汇总

背景与挑战

背景概述

该数据集是一个用于自动检测飓风后受损建筑的基准数据集，基于2017年飓风哈维后休斯顿地区的遥感图像，包含ESRI Shapefiles格式的边界框标注，区分受损（含四个损坏等级）和非受损建筑，旨在支持计算机视觉和灾害响应研究。数据集格式包括Shapefiles和GeoTiffs，并索引到NOAA图像，便于训练对象检测模型。