Skip to content
Repository for Scale-recurrent Network for Deep Image Deblurring
Branch: master
Clone or download
jiangsutx Merge pull request #16 from rimchang/patch-1
unvalid if statement, should not use `is` to compare string values.
Latest commit 2caaa85 Oct 8, 2018
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
checkpoints Add models and README Jun 13, 2018
imgs fix 2 figure Jun 13, 2018
models unvalid if statement Sep 27, 2018
testing_set Add models and README Jun 13, 2018
training_set Add models and README Jun 13, 2018
util Create __init__.py Oct 8, 2018
.gitignore
LICENSE
README.md
datalist_gopro.txt
run_model.py fix --gpu Aug 5, 2018

README.md

Scale-recurrent Network for Deep Image Deblurring

by Xin Tao, Hongyun Gao, Xiaoyong Shen, Jue Wang, Jiaya Jia. (pdf)

Our results on real data

Real Photo

Results on the testing dataset

Testing Dataset

More cases on real photos from previous papers:

More Cases

Prerequisites

  • Python2.7
  • Scipy
  • Scikit-image
  • numpy
  • Tensorflow 1.4 with NVIDIA GPU or CPU (cpu testing is very slow)

Installation

Clone this project to your machine.

git clone https://github.com/jiangsutx/SRN-Deblur.git
cd SRN-Deblur

Testing

Download pretrained models through: download_model.sh inside checkpoints/.

To test blur images in a folder, just use arguments --input_path=<TEST_FOLDER> and save the outputs to --output_path=<OUTPUT_FOLDER>. For example:

python run_model.py --input_path=./testing_set --output_path=./testing_res

If you have a GPU, please include --gpu argument, and add your gpu id to your command. Otherwise, use --gpu=-1 for CPU.

python run_model.py --gpu=0

To test the model, pre-defined height and width of tensorflow placeholder should be assigned. Our network requires the height and width be multiples of 16. When the gpu memory is enough, the height and width could be assigned to the maximum to accommodate all the images.

Otherwise, the images will be downsampled by the largest scale factor to be fed into the placeholder. And results will be upsampled to the original size.

According to our experience, --height=720 and --width=1280 work well on a Gefore GTX 1050 TI with 4GB memory. For example,

python run_model.py --height=720 --width=1280

Evaluation

The quantitative results of PSNR and SSIM in the paper is calculated using MATLAB built-in function psnr() and ssim() based on the generated color results.

Training

We trained our model using the dataset from DeepDeblur_release. Please put the dataset into training_set/. And the provided datalist_gopro.txt can be used to train the model.

Hyper parameters such as batch size, learning rate, epoch number can be tuned through command line:

python run_model.py --phase=train --batch=16 --lr=1e-4 --epoch=4000

Models

We provided 3 models (training settings) for testing:

  1. --model=lstm: This model implements exactly the same structure in our paper. Current released model weights should produce PSNR=30.19, SSIM=0.9334 on GOPRO testing dataset.
  2. --model=gray: According to our further experiments after paper acceptance, we are able to get a slightly better model by tuning parameters, even without LSTM. This model should produce visually sharper and quantitatively better results.
  3. --model=color: Previous models are trained on gray images, and may produce color ringing artifacts. So we train a model directly based on RGB images. This model keeps better color consistency, but the results are less sharp.

How to choose

If you would like to compare performance against our method, you can use model gray and lstm. If you want to restore blurry images you can try gray and color. And color is very useful in low-light noisy images.

Reference

If you use any part of our code, or SRN-Deblur is useful for your research, please consider citing:

@inproceedings{tao2018srndeblur,
  title={Scale-recurrent Network for Deep Image Deblurring},
  author={Tao, Xin and Gao, Hongyun and Shen, Xiaoyong and Wang, Jue and Jia, Jiaya},
  booktitle={IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
  year={2018}
}

Contact

We are glad to hear if you have any suggestions and questions.

Please send email to xtao@cse.cuhk.edu.hk

Reference

[1] Sun et al. J. Sun, W. Cao, Z. Xu, and J. Ponce. Learning a convolutional neural network for non-uniform motion blur removal. In CVPR, pages 769–777. IEEE, 2015.

[2] Nah et al. S. Nah, T. H. Kim, and K. M. Lee. Deep multi-scale convolutional neural network for dynamic scene deblurring. pages 3883–3891, 2017.

[3] Whyte et al. O. Whyte, J. Sivic, A. Zisserman, and J. Ponce. Nonuniform deblurring for shaken images. International Journal on Computer Vision, 98(2):168–186, 2012.

You can’t perform that action at this time.