Building Volume Panel of 106 Chinese Cities 2018-2023

The dataset provides yearly building height estimates per 10 × 10 meter pixel resolution for 106 Chinese cities from 2018-2023. Generated using a ResNet-34 model, it includes files for cities defined by bounding boxes based on the ESRI Urban Areas dataset (with a 4–9 km buffer) and building reference data. Cities in the dataset are either part of the training set (40 cities) or classified as top-tier in China’s 2022 business attractiveness ranking. Cities without spatial definitions are excluded. Cities included in the dataset are: Anqing (安庆), Anyang (安阳), Baoding (保定), Beijing (北京), Bengbu (蚌埠), Cangzhou (沧州), Changchun (长春), Changsha (长沙), Changzhou (常州), Chaozhou (潮州), Chengdu (成都), Chongqing (重庆), Dalian (大连), Dezhou (德州), Dongguan (东莞), Foshan (佛山), Fuyang (阜阳), Fuzhou (福州), Ganzhou (赣州), Guangzhou (广州), Guilin (桂林), Guiyang (贵阳), Haikou (海口), Handan (邯郸), Hangzhou (杭州), Harbin (哈尔滨), Hefei (合肥), Hengyang (衡阳), Heze (菏泽), Hohhot (呼和浩特), Huizhou (惠州), Huzhou (湖州), Jiangmen (江门), Jiaxing (嘉兴), Jinan (济南), Jinhua (金华), Jining (济宁), Jiujiang (九江), Kaifeng (开封), Kunming (昆明), Langfang (廊坊), Lanzhou (兰州), Lianyungang (连云港), Liaocheng (聊城), Linyi (临沂), Liuzhou (柳州), Luoyang (洛阳), Mianyang (绵阳), Nanchang (南昌), Nanjing (南京), Nanning (南宁), Nantong (南通), Nanyang (南阳), Ningbo (宁波), Putian (莆田), Qingdao (青岛), Quanzhou (泉州), Sanya (三亚), Shanghai (上海), Shangqiu (商丘), Shangrao (上饶), Shantou (汕头), Shaoxing (绍兴), Shenyang (沈阳), Shenzhen (深圳), Shijiazhuang (石家庄), Su_4_zhou (宿州), Suzhou (苏州), Tai_1_zhou (台州), Tai_4_zhou (泰州), Taiyuan (太原), Tangshan (唐山), Tianjin (天津), Urumqi (乌鲁木齐), Weifang (潍坊), Weihai (威海), Weinan (渭南), Wenzhou (温州), Wuhan (武汉), Wuhu (芜湖), Wuxi (无锡), Xi'an (西安), Xiamen (厦门), Xingtai (邢台), Xinxiang (新乡), Xinyang (信阳), Xuchang (许昌), Xuzhou (徐州), Yancheng (盐城), Yangzhou (扬州), Yantai (烟台), Yichang (宜昌), Yichun (宜春), Yinchuan (银川), Yueyang (岳阳), Yuncheng (运城), Zhangzhou (漳州), Zhaoqing (肇庆), Zhengzhou (郑州), Zhenjiang (镇江), Zhongshan (中山), Zhuhai (珠海), Zhumadian (驻马店), Zhuzhou (株洲), Zibo (淄博), Zunyi (遵义).

本数据集提供了2018至2023年间，中国106座城市的每10×10米像素分辨率年度建筑高度估算值。本数据集基于残差神经网络34（ResNet-34）生成，包含以ESRI城市区域（ESRI Urban Areas）数据集为基准划定、并带有4-9公里缓冲区的城市边界框（bounding box）文件，以及建筑参考数据。数据集内的城市分为两类：其一为训练集城市（共40座），其二为2022年中国商业吸引力排名中的一线城市。无空间定义的城市已被排除。数据集包含的城市如下：安庆、安阳、保定、北京、蚌埠、沧州、长春、长沙、常州、潮州、成都、重庆、大连、德州、东莞、佛山、阜阳、福州、赣州、广州、桂林、贵阳、海口、邯郸、杭州、哈尔滨、合肥、衡阳、菏泽、呼和浩特、惠州、湖州、江门、嘉兴、济南、金华、济宁、九江、开封、昆明、廊坊、兰州、连云港、聊城、临沂、柳州、洛阳、绵阳、南昌、南京、南宁、南通、南阳、宁波、莆田、青岛、泉州、三亚、上海、商丘、上饶、汕头、绍兴、沈阳、深圳、石家庄、宿州、苏州、台州、泰州、太原、唐山、天津、乌鲁木齐、潍坊、威海、渭南、温州、武汉、芜湖、无锡、西安、厦门、邢台、新乡、信阳、许昌、徐州、盐城、扬州、烟台、宜昌、宜春、银川、岳阳、运城、漳州、肇庆、郑州、镇江、中山、珠海、驻马店、株洲、淄博、遵义。

搜集汇总

背景与挑战

背景概述

该数据集包含2018-2023年中国106个城市的建筑高度估计数据，分辨率为10×10米像素，基于Sentinel 1和2影像，采用ResNet-34模型和Swin Transformer编码器生成。覆盖城市包括训练集城市和商业吸引力排名前列的城市。