A Satellite Imagery Dataset for Long-Term Sustainable Development in United States Cities

该数据集利用深度学习模型为五个可持续发展目标（SDGs）中的25个可持续发展指标开发，覆盖了美国100个最人口稠密的城市及其相应的普查区块组，时间跨度从2014年到2023年。数据集通过收集卫星图像并使用先进的对象检测和语义分割模型来识别对象，以观察城市的鸟瞰图。此外，还收集了人口、夜间灯光、调查和建筑环境数据，以描绘与贫困、健康、教育、不平等和生活环境相关的SDGs。

This dataset was developed using deep learning models for 25 sustainable development indicators spanning five Sustainable Development Goals (SDGs). It covers the 100 most densely populated cities in the United States and their corresponding census block groups, with a temporal coverage ranging from 2014 to 2023. The dataset collects satellite imagery and utilizes advanced object detection and semantic segmentation models to identify objects for analyzing urban bird's-eye views. Additionally, demographic data, nighttime light data, survey data, and built environment data are collected to characterize the Sustainable Development Goals related to poverty, health, education, inequality, and living environments.