Home

Awesome

OLI2MSI

OLI2MSI is a multisensor dataset for remote sensing imagery super-resolution. The OLI2MSI dataset is composed of Landsat8-OLI and Sentinel2-MSI images, where OLI images serve as low-resolution (LR) images and MSI images are regarded as ground truth high-resolution (HR) images. The OLI and MSI data have 10m and 30m ground sample distance (GSD) respectively, which means that the dataset has an upscale factor of 3. More details about OLI2MSI can be found in our paper.


<img src='imgs/data_samples.png'> LR-HR image-pair samples in OLI2MSI training set. The first row is LR images from Landsat8-OLI (30m GSD), while the second row is the corresponding HR images from Sentinel2-MSI (10m GSD).

Getting Started

Download OLI2MSI dataset

All data is hosted on Baidu Wangpan with extract code: rssr and Google Drive:

PathSizeFilesFormatDescription
OLI2MSI-dataset7.5 GB10,650Main folder
├  train_hr6.0 GB5,225GeoTiffAligned and cropped HR images at 480×480 for training
├  train_lr1.4 GB5,225GeoTiffAligned and cropped LR images at 160×160 for training
├  test_hr115.2 MB100GeoTiffAligned and cropped HR images at 480×480 for testing
├  test_lr26.7 MB100GeoTiffAligned and cropped LR images at 160×160 for testing

Note: All files are in GeoTiff format. GeoTIFF is a public domain metadata standard which allows georeferencing information to be embedded within a TIFF file. The potential additional information includes map projection, coordinate systems, ellipsoids, datums, and everything else necessary to establish the exact spatial reference for the file. We recommend using rasterio to read the files. Also, you can run 'convertpng.py' to convert the GeoTiff files to png files with the georeferencing information lost.

Reproduce or make your own dataset

You can also reproduce or make your own dataset in your custom region of interest (ROI) with the provided python scripts.

Prerequisites

Installation and user guide

Note: You neet to create your own geojson file to desigate the ROI first and modify the configurations in scripts as your need.

Please enjoy it!

Citation

If you use this code or the OLI2MSI dataset for your research, please cite our paper.

@article{wang2021multisensor,
  title={Multisensor Remote Sensing Imagery Super-Resolution with Conditional GAN},
  author={Wang, Junwei and Gao, Kun and Zhang, Zhenzhou and Ni, Chong and Hu, Zibo and Chen, Dayu and Wu, Qiong},
  journal={Journal of Remote Sensing},
  volume={2021},
  year={2021},
  publisher={AAAS}
}