基于计算机视觉的光伏发电潜力评估的屋顶信息数据集（RID）

RID（屋顶信息数据集） 是一个专注于屋顶上层结构和屋顶分块语义分割的高分辨率数据集，包含 1880 张通过 Google Maps Static API 下载的地理配准图像及其对应掩码。数据集涵盖 16 个方位类别和平顶类别，以及屋顶上层结构（如光伏模块、窗户等）的标注。数据经过多位标注者审核，确保标注质量。此外，数据集还提供了训练集、验证集和测试集的划分文件，并附有开源代码支持数据准备、模型训练和评估。该数据集采用 CC-BY-NC 许可协议，适用于非商业用途，为屋顶结构分析和语义分割研究提供了重要资源。RID (Roof Information Dataset) is a high-resolution dataset focused on roof superstructures and roof patch semantic segmentation. It contains 1880 georeferenced images downloaded via the Google Maps Static API and their corresponding masks. The dataset covers 16 orientation categories and a flat roof category, alongside annotations for roof superstructures such as photovoltaic modules and windows. All annotations have been reviewed by multiple annotators to guarantee annotation quality. Additionally, the dataset provides split files for the training, validation, and test sets, along with open-source code that supports data preparation, model training, and evaluation. This dataset is licensed under CC-BY-NC for non-commercial use, and serves as a critical resource for roof structure analysis and semantic segmentation research.

搜集汇总

背景与挑战

背景概述

RID是一个高分辨率屋顶信息数据集，专注于屋顶上层结构和屋顶分割的语义分割，包含1,880张地理参考图像和掩码，覆盖多种屋顶类别和结构标注。数据集经过质量审核，提供训练、验证和测试集分割及开源代码，采用CC-BY-NC许可，适用于非商业的屋顶结构分析和语义分割研究。