Remote Sensing World Knowledge (RSWK)
收藏资源简介:
Remote Sensing World Knowledge (RSWK)数据集是由中国科学院电子学研究所构建的一个大规模多模态基准数据集,包含来自全球175个国家的约14141个著名地标的超高分辨率遥感图像和丰富的文本描述。该数据集不仅整合了遥感领域的专业知识,如地表反射率、光谱指数、大气条件等,还包含了世界知识,如历史背景、文化意义、建设时期和主要事件等。这种结合了遥感专业知识和世界知识的数据集,不仅增强了RSWK数据集在视觉分析遥感图像方面的实用性,还为模型提供了更深的语义上下文,克服了传统数据集的限制,使遥感VLM能够执行更复杂的认知任务。
The Remote Sensing World Knowledge (RSWK) dataset is a large-scale multimodal benchmark dataset constructed by the Institute of Electronics, Chinese Academy of Sciences. It comprises ultra-high-resolution remote sensing images and rich textual descriptions of approximately 14,141 famous landmarks across 175 countries worldwide. This dataset not only integrates specialized remote sensing knowledge such as surface reflectance, spectral indices, and atmospheric conditions, but also incorporates world knowledge including historical backgrounds, cultural significance, construction periods, major events, and other relevant information. By combining remote sensing expertise and world knowledge, this dataset not only enhances the practical utility of RSWK for visual analysis of remote sensing images but also provides models with deeper semantic context, overcoming the limitations of traditional datasets and enabling remote sensing VLMs to perform more complex cognitive tasks.




