/SC-NAFSSR

Official PyTorch implement for SC-NAFSSR (Perceptual-Oriented Stereo Image Super-Resolution Using Stereo Consistency Guided NAFSSR)

Primary LanguagePythonMIT LicenseMIT

SC-NAFSSR

Official PyTorch implement for SC-NAFSSR (Perceptual-Oriented Stereo Image Super-Resolution Using Stereo Consistency Guided NAFSSR)

Requirements

The codes are based on BasicSR and NAFSSR .

pip install -r requirements.txt
python setup.py develop

Train

1. Prepare training data

Modify the training set and validation set path in NAFSSR.yml, NAFSSR_FT.yml, NAFSSR_FT_GAN.yml.

2. Begin to train

We apply a two-stage training strategy by commenting out the code in test.sh or uncommenting for different training configs. We also provide the GAN-based model's config in NAFSSR_FT_GAN.yml.

sh train.sh

Test

1. Prepare test data

Modify the test set path and pre-training model path in NAFSSR.yml.

2. Begin to test

sh test.sh