Robust Super-Resolution

This is the official implementation of the paper: Generalized Real-World Super-Resolution through Adversarial Robustness.

Paper

Generalized Real-World Super-Resolution through Adversarial Robustness
Angela Castillo ^1*, María Escobar ^1*, Juan C. Pérez ^{1, 2}, Andrés Romero ³, Radu Timofte ³, Luc Van Gool ³, Pablo Arbeláez^1
*Equal contribution.
¹ Center for Research and Formation in Artificial Intelligence (CinfonIA), Universidad de Los Andes.
² Image and Video Understanding Lab (IVUL), KAUST.
³ Computer Vision Lab (CVL), ETH Zürich.

Dependencies and Installation

• Python >= 3.7 (Recommend to use Anaconda)
• PyTorch == 1.6.0 TorchVision == 0.7.01.3
• NVIDIA GPU + CUDA v10.1.243

1. Clone repo

   git clone https://github.com/BCV-Uniandes/RSR

2. Install dependent packages

   cd RSR
   pip install -r requirements.txt

3. Install the BasicSR toolbox

   Please run the following commands in the RSR root path to install BasicSR:
   (Make sure that your GCC version: gcc >= 5)
   

   python setup.py develop --no_cuda_ext

_{BasicSR was only tested in Ubuntu.}

Dataset Preparation

• Please refer to this web page for details about the dataset organization and dataset augmentation.

Train

• Training command:

  bash train.sh

• Pre-trained SR model: Find the pre-trained SR model at Drive.

• Options/Configs: Please check to Config.md.

• Logging: Please refer to Logging.md.

Pre-trained Model and Test

• Find here our pre-trained model.

• Test command:

  bash test.sh

Citations

If RSR helps your research, please consider citing us.

@inproceedings{castillo2021generalized,
  title={Generalized Real-World Super-Resolution through Adversarial Robustness},
  author={Castillo, Angela and Escobar, Maria and P{\'e}rez, Juan C and Romero, Andr{\'e}s and Timofte, Radu and Van Gool, Luc and Arbelaez, Pablo},
  booktitle={Proceedings of the IEEE/CVF International Conference on Computer Vision},
  pages={1855--1865},
  year={2021}
}

Find other resources in our webpage.

License and Acknowledgement

This project borrows heavily from BasicSR, we thank the authors for their contributions to the community.
More details about license in LICENSE.

Contact

If you have any question, please email [email protected].