Remote Sensing Image Super-Resolution via Mixed High-Order Attention Network (MHAN) (Accept by TGRS 2020)

The is the pytorch code for paper "Remote Sensing Image Super-Resolution via Mixed High-Order Attention Network" MHAN. The Test30 dataset used in the paper can be found in this repository，referring to the folder './Test30'. Some other general image and remote sensing SR based models also provided in folder './models'.

Requirements

• Python 3.6.4
• Pytorch 1.3.1(GPU)
• OpenCV
• NVIDIA-SMI 430.64
• Driver Version: 430.64
• CUDA Version: 10.1

Dataset

We use AID as the training dataset, which is a collection of remote sensing images depicting 30 land-use classes, including airport, farmland, beach, desert, etc.

We conducted experiments on two satellite image datasets, namely, WHURS19 and RSSCN7.

Usage

Use the following command to train the model.

$ python main_x4.py

Use the following commandss to generate the SR images with respect to RSSCN7 and WHURS19 datasets.

$ python eval_RSSCN7.py
$ python eval_WHURS19.py

When the SR images are generated in the folder, use Evaluate_PSNR_SSIM.m file to comptute the PSNR and SSIM.