Essays about: https://ucassdl.cn/content/work/paper.html
1.DOTA: A Large-scale Dataset for Object Detection in Aerial Images. This is Wu Xia Guisong and Remote Sensing Laboratory country heavy Branch School of Telecommunication white Xiang made a joint data set, 2806 remote sensing images (about the size of 4000 * 4000), 188,282 more instances, divided into 15 categories. Sample type and number as follows (compared with other open dataset NWPU VHR-10).
Link: https: //captain-whu.github.io/DOAI2019/dataset.html
2.UCAS-AOD: Dataset of Object Detection in Aerial Images, National Laboratory of Pattern Recognition HKUST marked, contains only two types of targets: cars, planes, and the background of negative samples.
Link: http: //www.ucassdl.cn/resource.asp
DownLoad: Dataset and the basic information Overview Instruction Instruction-cn
References:[1]H. Zhu, X. Chen, W. Dai, K. Fu, Q. Ye, J. Jiao, "Orientation Robust Object Detection in Aerial Images Using Deep Convolutional Neural Network," IEEE Int'l Conf. Image Processing, 2015.
3.NWPU VHR-10: Northwestern Polytechnical University, marked detection target space remote sensing data sets, a total of 800 images, which contains the target of 650, the background image 150, goals include: aircraft, ships, tanks, baseball field, tennis courts, basketball courts, athletic fields, ports, bridges, vehicles 10 categories. Available for download, about 73M.
Link: http: //jiong.tea.ac.cn/people/JunweiHan/NWPUVHR10dataset.html
4.RSOD-Dataset: Wuhan University team mark including aircraft, playground, overpasses, four types of drums target number are as follows:
Aircraft: 4993 aircrafts in 446 images.
Playground: 191 playgrounds in 189 images.
Overpass: 180 overpass in 176 overpass.
Drums: 1586 oiltanks in 165 images.
Link: https: //github.com/RSIA-LIESMARS-WHU/RSOD-Dataset-
5.INRIA aerial image dataset: Inria is referred to as the French National Institute of Information and Automation, the agency has a large database, this database is the detection of a city building database, mark only the building, not building are two, and is the pixel level, for semantic segmentation. Training set and data sets collected from different cities of remote sensing images.
Link: https: //project.inria.fr/aerialimagelabeling/
6.TGRS-HRRSD-Dataset: HRRSD Xi'an Institute of Optics and Fine Mechanics Optical Image Analysis and Learning Research Center, Chinese Academy produced high-resolution images for target detection of remote sensing data sets.
Link: https: //github.com/CrazyStoneonRoad/TGRS-HRRSD-Dataset