Skip to content
No description or website provided.
Python
Branch: master
Clone or download
Latest commit 671f679 Oct 29, 2018
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
Dataset
LICENSE Create LICENSE Oct 8, 2018
README.md Update README.md Oct 29, 2018
dataset.py
fid_score.py fid Oct 29, 2018
inception.py
model.pkl
model.py v0 Sep 22, 2018
model_evaluation.py v0 Sep 22, 2018
overview.png
test_res.tar
testset.tar v0 Sep 22, 2018

README.md

Face and Image Super-resolution

Paper

Adrian Bulat*, Jing Yang*, Georgios Tzimiropoulos ''To learn image super-resolution, use a GAN to learn how to do image degradation first'' in ECCV2018

Method

  • High-to-Low GAN using unpaired low and high-resolution images to simulate the image degradation
  • Low-to-High GAN using paired low and high-resolution images to learn real-world super resolution
  • GAN loss driving the image generation process

Requirements

Pytorch 0.4.1

Data

  • Trainset is in Dataset. HIGH is the training high resolution images. LOW is the training low resolution images
  • Testset is testset.tar
  • test_res.tar is our result

Running testing

CUDA_VISIBLE_DEVICES=0, python model_evaluation.py 

Fid Calculation

CUDA_VISIBLE_DEVICES=0, python fid_score.py /Dataset/HIGH/SRtrainset_2/ test_res/

This code is from https://github.com/mseitzer/pytorch-fid

Citation

@inproceedings{bulat2018learn, 
  title={To learn image super-resolution, use a GAN to learn how to do image degradation first},
  author={Bulat, Adrian and Yang, Jing and Tzimiropoulos, Georgios},
  booktitle={Proceedings of the European Conference on Computer Vision (ECCV)},
  pages={185--200},
  year={2018}  
}

License

This project is licensed under the MIT License

You can’t perform that action at this time.