M3LEO
收藏资源简介:
M3LEO是由剑桥大学等多个机构合作创建的多模态、多标签地球观测数据集,包含来自Sentinel-1的极化、干涉和相干SAR数据以及Sentinel-2的RGB图像。数据集覆盖17.5TB,包含约1000万个4x4公里的数据芯片,分布在六个不同的地理区域。创建过程涉及复杂的数据处理和参数选择,以确保数据兼容机器学习管道。M3LEO的应用领域广泛,包括自然灾害管理、环境监测和城市规划等,旨在解决传统光学传感器在恶劣天气和夜间无法有效工作的问题。
M3LEO is a multimodal, multi-label Earth observation dataset developed through collaboration between the University of Cambridge and multiple other institutions. It includes polarimetric, interferometric and coherent SAR data from Sentinel-1, as well as RGB imagery from Sentinel-2. The dataset has a total size of 17.5 TB, containing approximately 10 million 4x4 km data chips distributed across six distinct geographic regions. Its development involves sophisticated data processing and parameter tuning to ensure compatibility with machine learning pipelines. M3LEO covers a wide range of application fields including natural disaster management, environmental monitoring and urban planning, and it aims to address the limitation that conventional optical sensors cannot operate effectively under adverse weather conditions and during nighttime.




