SynRS3D
收藏资源简介:
SynRS3D是由东京大学和日本理化学研究所联合创建的全球最大合成遥感三维数据集,包含69,667张高分辨率光学图像,覆盖全球六种不同城市风格和八种地表覆盖类型。数据集通过先进的生成流程,结合精确的高度信息和建筑变化掩膜,旨在解决单视图遥感图像的全球三维语义理解问题。该数据集的应用领域广泛,包括环境监测、城市规划和灾害响应等,通过合成数据与真实场景的无监督域适应方法,有效提升了模型在实际应用中的性能。
SynRS3D is the world's largest synthetic remote sensing 3D dataset jointly created by The University of Tokyo and RIKEN. It contains 69,667 high-resolution optical images, covering six distinct urban styles and eight land cover types across the globe. Developed via an advanced generation pipeline that integrates precise elevation information and building change masks, this dataset aims to tackle the global 3D semantic understanding task for single-view remote sensing images. With wide-ranging applications in fields such as environmental monitoring, urban planning and disaster response, it effectively enhances the real-world performance of models through unsupervised domain adaptation between synthetic data and actual scenes.




