LADI v2

该数据集是由美国民间空中巡逻队（CAP）在2015年至2023年期间，针对联邦政府宣布的紧急情况在美国境内拍摄的约1万张灾害图像组成，并针对多标签分类进行了注释。这些图像由受过培训的志愿者进行标注，标签设计旨在帮助紧急管理人员进行损害评估。该数据集考虑了灾害图像随时间分布的变化，并包含了被称为“v2a”的修订版标签集。数据规模包括9963张图像，其中8030张用于训练，892张用于验证，以及1041张用于测试。该数据集的任务是针对灾害图像分析进行多标签分类。

This dataset is composed of approximately 10,000 disaster images captured across the United States between 2015 and 2023 by the Civil Air Patrol (CAP) for federally declared emergencies, and has been annotated for multi-label classification tasks. These images were labeled by trained volunteers, with the annotation tags designed to assist emergency managers in damage assessment. This dataset accounts for the temporal distribution changes of disaster images, and includes a revised label set known as "v2a". The exact scale of the dataset contains 9,963 images in total, among which 8,030 are allocated for training, 892 for validation, and 1,041 for testing. The core task of this dataset is multi-label classification for disaster image analysis.