Language : English
侯东阳

Datasets

Our datasets are for scientific research use only and should not be used commercially. When using it, please cite the following references.

  • Hou Dongyang; Miao Zelang; Xing Huaqiao*; Wu Hao*; V-RSIR: an Open Access Web-based Image Annotation Tool for Remote Sensing Image Retrieval, IEEE Access. 2019, 7: 83852-83862.

  • Hou Dongyang; Miao Zelang; Xing Huaqiao*; Wu Hao*; Two Novel Benchmark Datasets from ArcGIS and Bing World Imagery for Remote Sensing Image Retrieval, International Journal of Remote Sensing. 2021, 42(1): 240-258. doi: 10.1080/01431161.2020.1804090.


1  VGoogle

  • Description: VGoogle is a remote sensing image dataset constructed from the Google World Imagery using the V-RSIR tool. It covers 38 classes and contains a total of 59,404 images. Each class contains at least 1500 images with spatial resolutions ranging from 0.07 m to 9.555 m. Without affecting the data accuracy, the team converted the original dataset from RGBA four-channel to RGB three-channel and converted the images from tiff format to jpg format on December 14, 2021, reducing the dataset storage space from 8.70 GB to 2.66 GB.

  • Download link: https://pan.baidu.com/s/1OLImdhJpOtwNgQYiDYJGqA

  • Extraction code: lv9i

2 VBing

  • Description: VBing is a remote sensing image dataset constructed from the Bing World Imagery using the V-RSIR tool. It covers 38 classes and contains a total of 58,944 images. Each class contains at least 1500 images with spatial resolutions ranging from 0.07 m to 38.22 m. Without affecting the data accuracy, the team converted the original dataset from RGBA four-channel to RGB three-channel and converted the images from tiff format to jpg format on December 14, 2021, reducing the dataset storage space from 7.99 GB to 2.23 GB. 

  • Download link: https://pan.baidu.com/s/1HhP2_XwxDQtFercEFZQyEw

  • Extraction code: 5jjv

3 VArcGIS

  • Description: VArcGIS is a remote sensing image dataset constructed from the ArcGIS World Imagery online image resource using the V-RSIR tool. It covers 38 classes  and contains a total of 59,071 images. Each class contains at least 1500 images with spatial resolutions ranging from 0.07 m to 19.11 m. Without affecting the data accuracy, the team converted the original dataset from RGBA four-channel to RGB three-channel and converted the images from tiff format to jpg format on December 14, 2021, reducing the dataset storage space from 8.08 GB to 2.09 GB.

  • Download link: https://pan.baidu.com/s/1KRKnr5R2UTfnOuGwQhsX5Q

  • Extraction code: xhkv