Other Data Paper Zone II Versions ZH1 Vol 6 (1) 2021
Download
A dataset of building instances of typical cities in China
Dataset URL >>
: 2021 - 02 - 11
: 2021 - 03 - 25
: 2021 - 02 - 26
: 2021 - 03 - 30
1011 15 0
Abstract & Keywords
Abstract: Building contour information is an important part of the national basic geographic information. The performance of building automatic extraction is usually driven by a large number of training samples. To enrich the building extraction datasets of cities in China, we compiled a building instance dataset sourced from high-resolution remote sensing images through the combination of manual annotation and interactive annotation. This dataset consists of the samples in 7,260 regions, with 63,886 building instances in four China’s cities: Beijing, Shanghai, Shenzhen and Wuhan. The annotations of the dataset consist of MS COCO 2017 format files and the corresponding building mask binary maps. This dataset provides fundamental data for the research on building detection of and extraction from high-resolution remote sensing images.
Keywords: building extraction; instance segmentation; high-resolution remote sensing images
Dataset Profile
TitleA dataset of building instances of typical cities in China
Data corresponding authorFANG Fang (fangfang@cug.edu.cn)
Data authorsWU Kaishun, ZHENG Daoyuan, CHEN Yanling, ZENG Linyun, ZHAN Jiahuig, CHAI Shenghuai, XU Wenjie, YANG Yongliang, LI Shengwen, LIU YuanYuan, FANG Fang
Time range2017–2019
Geographical scopeChina
Spatial resolution0.29 m
Data volumeabout 5,000 MB
Data format*.tif, *.json, *.png
Data service system<https://doi.org/10.11922/sciencedb.00620>
Source of fundingOpen Research Fund of National Earth Observation Data Center (NODAOP2020015)
Dataset compositionThe dataset consists of samples in 7,260 regions, with three types of files: (1) *.tif files, storing the information about high-resolution remote sensing images; (2) *.json files, used for instance segmentation tasks, and describing building annotation data, including the training set and the test set; (3) *.png flies, the pixel-level semantic label of building area, used for semantic classification tasks.
Article and author information
WU Kaishun
主要承担工作:数据集整体结构设计,交互式标注算法研究与实现。
ZHENG Daoyuan
主要承担工作:数据集人工及交互式标注,论文初稿撰写。
CHEN Yanling
主要承担工作:数据集人工及交互式标注。
ZENG Linyun
主要承担工作:数据集人工及交互式标注。
ZHANG Jiahui
主要承担工作:数据集人工及交互式标注。
CHAI Shenghua
主要承担工作:数据集人工标注。
WU Wenjie
主要承担工作:数据集人工标注。
YANG Yongliang
主要承担工作:数据集人工标注。
LI Shengwen
主要承担工作:论文方向指导与质量把关。
LIU YuanYuan
主要承担工作:算法指导。
FANG Fang
主要承担工作:项目规划与论文质量把关。
fangfang@cug.edu.cn
Publication records
Published: March 30, 2021 ( VersionsZH1
References
中国科学数据
csdata