FloodNet
收藏资源简介:
FloodNet是由马里兰大学巴尔的摩分校创建的高分辨率无人机影像数据集,专门用于洪水后场景理解。该数据集包含3200张图像,这些图像是在飓风哈维后捕获的,展示了受灾区域的后洪水损害情况。数据集中的图像经过像素级标注,用于语义分割任务,并生成了视觉问答任务的问题。FloodNet数据集面临多个挑战,包括检测洪水淹没的道路和建筑物,以及区分自然水和洪水淹没的水域。该数据集的应用领域主要集中在灾害损害评估,旨在通过深度学习算法精确理解受灾区域的情况。
FloodNet is a high-resolution unmanned aerial vehicle (UAV) image dataset developed by the University of Maryland, Baltimore, specifically tailored for post-flood scene understanding. This dataset includes 3200 images captured in the aftermath of Hurricane Harvey, showcasing post-flood damage in disaster-affected areas. The images in the dataset have undergone pixel-level annotations for semantic segmentation tasks, and corresponding questions for Visual Question Answering (VQA) tasks have also been generated. FloodNet presents multiple key challenges, including detecting flood-submerged roads and buildings, as well as distinguishing between natural water bodies and flood-covered waters. The primary application domain of this dataset is disaster damage assessment, aiming to enable accurate understanding of the conditions in affected areas through deep learning algorithms.

- 1FloodNet: A High Resolution Aerial Imagery Dataset for Post Flood Scene Understanding马里兰大学巴尔的摩分校 · 2020年



