isongxw/RBPN-PyTorch

The project is an official implement of our CVPR2019 paper "Recurrent Back-Projection Network for Video Super-Resolution"

C++

Dependencies

Python 3.8.8
PyTorch ==1.7.1
Pyflow -> https://github.com/pathak22/pyflow

cd pyflow/
python setup.py build_ext -i
cp pyflow*.so ..
cd ..

Training

python main.py --HR_dir="PATH/TO/HR" --LR_dir="PATH/TO/LR" --file_list='NTIRE21/train.txt'

You can modify the gpus_list=[0,1] in the code to adapt to your device

Testing

python test.py --LR_dir="PATH/TO/LR" --file_list='NTIRE21/test.txt'

You can modify the gpus_list=[0,1] in the test.py to adapt to your device.
You can modify the '--model' parameter in the test.py to load different trained model.