SlumMapVisionBR: A Dataset for Mapping Slums with Medium-Resolution Satellite Imagery

This dataset contains Landsat images (30m/pixel) and masks with slum locations in Brazil. This imagery can be used to train models that help keep track of Goal 11.1 of the United Nations Sustainable Development Goals. Details on how the data was collected and labelled are available at: https://github.com/ml-labs-crt/SlumMapVisionBR, along with the code to generate the dataset and benchmark results. Description of files: images.zip contains the images. masks.zip contains the masks with slum locations (1 indicates the area has been labelled as a slum). metadata.csv contains information about the 116 locations in this dataset. metadata_dictionary.txt includes a description of the fields in metadata.csv. Refining documentation: We welcome additions and edits that make using the existing data or adding new data more straightforward for the community. Please get in touch with us by posting an issue on GitHub.本数据集涵盖巴西地区的陆地卫星（Landsat）影像（分辨率为30米/像素）以及贫民窟区域掩膜（mask）。此类影像可用于训练相关模型，助力追踪联合国可持续发展目标（United Nations Sustainable Development Goals，SDGs）的11.1号目标。有关本数据集的采集流程、标注规则，以及数据集生成代码与基准测试结果的详细说明，均可通过以下链接查阅：https://github.com/ml-labs-crt/SlumMapVisionBR。 ### 文件说明 "images.zip" 内含本数据集的全部影像文件。 "masks.zip" 内含贫民窟区域掩膜（mask），其中像素值为1的区域即被标注为贫民窟。 "metadata.csv" 记录了本数据集内116个采样点位的相关信息。 "metadata_dictionary.txt" 对 "metadata.csv" 中的各字段含义进行了说明。 ### 文档优化倡议 我们欢迎社区贡献者对现有文档进行增补与修订，以简化现有数据的使用流程及新增数据的接入工作。若有相关需求，请通过GitHub提交Issue与我们取得联系。