RRSGAN

PyTorch implementation of RRSGAN: Reference-based Super-Resolution for Remote Sensing Image

Dependencies

• Python 3.6+
• pyTorch >= 1.0
• CUDA 9.0 and gcc4.7 (for DCNv2 installation)
• Python packages: pip install numpy opencv-python lmdb pyyaml
• DCNv2 (Deformable Convolutional Networks V2, please refer to ./codes/models/archs/DCNv2/README.md)

Dataset Preparation (Reference-Based Remote Sensing Super-Resolution Dataset)

Download Datasets

Training dataset can be downloaded from baidu pan, password:lnff, google drive, and Microsoft OneDrive.

Test datasets can be found in ./dataset/val.

Preprocess Datasets

After downloading the training dataset, please put them in the folder ./dataset/train.

tar -xvzf train_data.tar.gz

The training set is transformed into LMDB format for faster IO speed.

cd ./dataset/data_script

python create_lmdb.py

Training

To train an RRSGAN model:

Before training, pre-trained vgg model need to be downloaded here. Please put it in the folder ./codes/models/archs/pretrained_model.

cd ./codes/example/RRSGAN

sh train.sh

Train with Slurm

cd ./codes/example/RRSGAN

sh train_slurm.sh

• Before running this code, please modify train_slurm.sh to your own configurations.

• You can find your training results in ./codes/example/RRSGAN/exp

Testing

cd ./codes/example/RRSGAN

sh val.sh

• Before running this code, please modify val.sh to your own configurations, e.g. the save path of your model.

Acknowledgement

The code is based on MMSR.

Contact

If you have any questions about our work, please contact drm@mail.tsinghua.edu.cn