jeya-maria-jose / Image-Recovery-Using-Conditional-Adversarial-Networks

Analyzing Conditional Adversarial Networks to solve image recovery problems like shadow recovery, denoising and deblurring - CVIP 2019

Image-Recovery-Using-Conditional-Adversarial-Networks

Analyzing Conditional Adversarial Networks to solve image recovery problems like shadow recovery, denoising and deblurring. Repository for the paper : Tackling Multiple Visual Artifacts: Blind Image Restoration Using Conditional Adversarial Networks

Requistites:

1. Python 3
2. Pytorch
3. MATLAB

Dataset

ISTD Dataset is used. Link to Google Drive

Image Recovery Objectives

Recovery is tested for the following augmented sets of images created from ISTD dataset.

1. Only Shadow
2. Shadow + Salt and Pepper Noise
3. Shadow + Speckle Noise
4. Shadow + Gaussian Noise
5. Shadow + All Noises
6. Shadow+ BLur
7. Shadow + Salt and Pepper Noise + Blur
8. Shadow + Speckle Noise + Blur
9. Shadow+ Gaussian Noise + Blur
10. Shadow + All Noises + Blur

Creating Dataset

Run generate.m to generate the extra augmented images to help the network train to restore images that are degraded by more than one type of artifact.

python dataset_create.py to resize the input and GT images such that they are in the proper format to be fed in to the network.

Addition of any more type of images afflicted with artifacts would be fruitful and would improve the network's performance.

Make a dataset folder and store the augmented images generated from dataset_create.py in it.

Conditional Adversarial Networks

We have used the pix2pix network proposed for image to image translation tasks by Jun-Yan Zhu for this work.

Pix2pix: Project | Paper | Torch

Running the Code :

To train the model : python train.py --dataroot ./datasets/shadow --name shadow_pix2pix --model pix2pix --direction BtoA

To test the model: python test.py --dataroot ./datasets/shadow --name shadow_pix2pix --model pix2pix --direction BtoA

To view the results: ./results/facades_pix2pix/test_latest/index.html

Results

From Left :

i)Input

ii)Prediction

iii) Ground Truth

Samples:

Shadow + Salt and Pepper Noise:

Shadow + Speckle Noise

Citation

If you use this work , please cite our paper Tackling Multiple Visual Artifacts: Blind Image Restoration Using Conditional Adversarial Networks:

@inproceedings{anand2019tackling,
  title={Tackling Multiple Visual Artifacts: Blind Image Restoration Using Conditional Adversarial Networks},
  author={Anand, M and Natraj, A Ashwin and Jose, V Jeya Maria and Subramanian, K and Bhardwaj, Priyanka and Pandeeswari, R and Deivalakshmi, S},
  booktitle={International Conference on Computer Vision and Image Processing},
  pages={331--342},
  year={2019},
  organization={Springer}
}