PANGAEA
收藏资源简介:
PANGAEA是由KU Leuven、KTH、CNAM、University of Glasgow和IGN共同创建的一个全球性和包容性的地理空间基础模型评估基准。该数据集包含11个子数据集,覆盖了从海洋碎片分割到建筑物损坏映射等多种任务,涉及城市、农业、海洋和森林等多个应用领域。数据集包括不同的时间配置(单时相、双时相和多时相)、广泛的地理覆盖(全球范围)、多种空间分辨率(从1.5米/像素到30米/像素)和多种传感器类型(自然色、多光谱和合成孔径雷达)。创建过程涉及对现有数据集的筛选和整合,旨在解决当前评估方法的局限性,提供一个标准化和广泛适用的基准。PANGAEA的应用领域广泛,旨在评估和提升地理空间基础模型在实际应用中的性能和泛化能力。
PANGAEA is a global and inclusive geospatial foundation model evaluation benchmark co-developed by KU Leuven, KTH, CNAM, University of Glasgow and IGN. This dataset comprises 11 sub-datasets, covering diverse tasks ranging from marine debris segmentation to building damage mapping, and involving multiple application domains such as urban, agricultural, marine and forestry scenarios. The dataset includes varied temporal configurations (single-temporal, bi-temporal and multi-temporal), extensive global geographic coverage, diverse spatial resolutions ranging from 1.5 meters per pixel to 30 meters per pixel, and multiple sensor types including natural color, multispectral and synthetic aperture radar. The development process involves screening and integrating existing datasets, aiming to address the limitations of current evaluation methods and provide a standardized and widely applicable benchmark. PANGAEA has a broad range of application scenarios, and is designed to evaluate and enhance the performance and generalization capability of geospatial foundation models in real-world applications.




