Chesapeake Roads Spatial Context (RSC)

Chesapeake Roads Spatial Context (RSC)数据集由微软AI for Good研究实验室创建，旨在评估机器学习模型在处理高分辨率航空影像时的空间长距离上下文理解能力。该数据集包含30,000个512x512像素的航空影像补丁，来源于美国农业部的国家农业影像计划（NAIP），并配有来自Chesapeake Bay Conservancy的2017/2018年土地使用土地覆盖数据集的标签。数据集包含'背景'、'道路'和'树冠覆盖的道路'类别，用于评估模型在道路网络分割中的表现，特别是在道路被树冠遮挡的情况下的识别能力。数据集的应用领域主要集中在解决道路网络的精确分割问题，特别是在存在遮挡的情况下。

Chesapeake Roads Spatial Context (RSC) Dataset was developed by the Microsoft AI for Good Research Lab, designed to evaluate the spatial long-range contextual understanding ability of machine learning models when processing high-resolution aerial imagery. This dataset contains 30,000 512×512 pixel aerial image patches sourced from the U.S. Department of Agriculture (USDA) National Agriculture Imagery Program (NAIP), and is paired with labels derived from the 2017/2018 Land Use Land Cover Dataset provided by the Chesapeake Bay Conservancy. It includes three categories: "Background", "Roads", and "Roads Under Tree Canopy", which are used to assess model performance in road network segmentation tasks, particularly the ability to identify roads obscured by tree canopy. The primary application scenarios of this dataset focus on addressing the precise segmentation of road networks, especially under occlusion conditions.