典型地区城镇建筑物轮廓提取结果（2022）

该数据集包含广州、珠海、汕头、班加罗尔、槟城、开普敦、里约热内卢、巴塞罗那、伊斯坦布尔、德黑兰共计十个城市的TIFF格式的建筑轮廓。建筑轮廓数据集采用由U-Net和域自适应模块构成、基于INRIA数据集训练的遥感影像语义分割网络分割得到。建筑高宽比数据基于带有高度信息的建筑矢量在POSTGIS中计算得到。建筑轮廓数据集的交并比超过70%。建筑物轮廓数据集可以被用于太阳能光伏建模、建筑设计、城市功能分析等领域。

This dataset contains TIFF-format building footprints for ten cities: Guangzhou, Zhuhai, Shantou, Bangalore, Penang, Cape Town, Rio de Janeiro, Barcelona, Istanbul, and Tehran. The building footprint dataset was generated via a remote sensing image semantic segmentation network composed of a U-Net and a domain adaptation module, which was trained on the INRIA dataset. The aspect ratio data of buildings was calculated from building vectors with height information using POSTGIS. The Intersection over Union (IoU) of the building footprint dataset exceeds 70%. This building footprint dataset can be applied in fields such as solar photovoltaic modeling, architectural design, and urban function analysis.

搜集汇总

背景与挑战

背景概述

该数据集包含2022年广州、珠海等十个国内外城市的TIFF格式建筑轮廓，采用基于U-Net和域自适应模块的遥感影像语义分割网络提取，交并比超过70%。数据适用于太阳能光伏建模、建筑设计等领域，需通过申请获取。