So2Sat LCZ42

So2Sat LCZ42数据集由德国航空航天中心与慕尼黑工业大学合作创建，包含约40万对Sentinel-1和Sentinel-2图像块，覆盖全球42个主要城市及10个较小区域。数据集通过15位领域专家遵循精心设计的标注工作流程和评估过程，历时六个月完成标注，整体置信度达到85%。该数据集旨在为城市增长监测提供一个无偏见的全球分布数据集，通过机器学习方法，测量城市区域的形态、紧凑度和高度，这些指标较少依赖于人类和文化因素。数据集的应用领域包括基础设施规划、灾害缓解、健康和绿地规划以及人口评估等。

So2Sat LCZ42 is a dataset co-developed by the German Aerospace Center and the Technical University of Munich. It contains approximately 400,000 paired Sentinel-1 and Sentinel-2 image patches, covering 42 major cities and 10 smaller regions across the globe. Annotated by 15 domain experts over a six-month period following a meticulously designed annotation workflow and evaluation process, the dataset has an overall annotation confidence of 85%. This dataset aims to provide an unbiased, globally distributed dataset for urban growth monitoring, enabling the measurement of urban morphology, compactness and height via machine learning approaches, where these indicators are less dependent on human and cultural factors. Its application domains include infrastructure planning, disaster mitigation, health and green space planning, population assessment and other relevant fields.