Git-10M
收藏资源简介:
Git-10M是由北京航空航天大学与上海人工智能实验室联合开发的全球规模遥感图像-文本对数据集,包含1000万对图像和文本,覆盖了城市、森林、山脉等多种地理场景,并包含丰富的元数据,如图像分辨率和地理位置信息。该数据集显著超越了现有数据集的规模和多样性,为训练生成模型提供了坚实的基础。数据集的应用领域包括遥感图像生成、虚拟场景构建、图像编辑等,旨在解决全球尺度、多分辨率可控的遥感图像生成问题。
Git-10M is a global-scale remote sensing image-text pair dataset jointly developed by Beihang University and Shanghai AI Laboratory. It contains 10 million image-text pairs, covering diverse geographic scenarios including urban areas, forests, mountains and more, and includes rich metadata such as image resolution and geographic location information. This dataset significantly surpasses the scale and diversity of existing datasets, providing a solid foundation for training generative models. Its application fields include remote sensing image generation, virtual scene construction, image editing and other areas, aiming to solve the problem of global-scale, multi-resolution controllable remote sensing image generation.

- 1Text2Earth: Unlocking Text-driven Remote Sensing Image Generation with a Global-Scale Dataset and a Foundation Model北京航空航天大学 · 2025年



