RoadTracer

该数据集包含了来自六个不同国家四十个城市市区的高分辨率卫星图像。地面真实数据是通过OpenStreetMap生成的。此外，有十五个城市的数据被预留用于验证。该数据集的规模属于高分辨率级别，任务是对航拍图像中的像素进行分类，以区分道路和非道路区域。

This dataset includes high-resolution satellite imagery acquired from urban regions of 40 cities across six distinct countries. Ground truth annotations were generated using OpenStreetMap. Additionally, data from 15 cities are set aside for model validation. This high-resolution dataset is designed for the task of pixel-wise classification to differentiate between road and non-road areas within the aerial imagery.