OpenEarthMap

OpenEarthMap是全球高分辨率土地覆盖制图的基准数据集。OpenEarthMap由5000幅航空和卫星图像组成，带有手动注释的8级土地覆盖标签和220万段，地面采样距离为0.25-0.5m，覆盖了来自6大洲44个国家的97个地区。OpenEarthMap促进研究，包括但不限于语义分割和领域适应。在OpenEarthMap上训练的土地覆盖制图模型在全球范围内得到推广，并且可以在各种应用中用作现成的模型。

OpenEarthMap is a benchmark dataset for global high-resolution land cover mapping. It consists of 5000 aerial and satellite images with manually annotated 8-class land cover labels and 2.2 million segments, with a ground sampling distance of 0.25–0.5 m, covering 97 regions across 44 countries in 6 continents. OpenEarthMap facilitates research including but not limited to semantic segmentation and domain adaptation. Land cover mapping models trained on OpenEarthMap exhibit global generalization performance and can be used as off-the-shelf models in various applications.

搜集汇总

背景与挑战

背景概述

OpenEarthMap是一个全球高分辨率土地覆盖制图基准数据集，包含5000幅航空和卫星图像，带有8级土地覆盖标签，覆盖6大洲44个国家的97个地区，适用于语义分割和领域适应研究。该数据集由东京大学和日本理化学研究所先进智能研究中心于2022年发布。