https://github.com/jammer345/SL-CycleGAN-Blind-Motion-Deblurring-in-Cycles-using-Sparse-Learning

SL-CycleGAN: Blind Motion Deblurring in Cycles using Sparse Learning
https://github.com/jammer345/SL-CycleGAN-Blind-Motion-Deblurring-in-Cycles-using-Sparse-Learning

computer-vision deblurring generative-adversarial-network image-processing sparse state-of-the-art

Last synced: about 2 months ago
JSON representation

SL-CycleGAN: Blind Motion Deblurring in Cycles using Sparse Learning

• Host: GitHub
• URL: https://github.com/jammer345/SL-CycleGAN-Blind-Motion-Deblurring-in-Cycles-using-Sparse-Learning
• Owner: jammer345
• Created: 2022-02-13T09:11:44.000Z (over 2 years ago)
• Default Branch: main
• Last Pushed: 2022-04-24T09:38:23.000Z (over 2 years ago)
• Last Synced: 2024-07-10T07:52:49.180Z (2 months ago)
• Topics: computer-vision, deblurring, generative-adversarial-network, image-processing, sparse, state-of-the-art
• Language: Python
• Homepage:
• Size: 17.7 MB
• Stars: 8
• Watchers: 2
• Forks: 0
• Open Issues: 1
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# SL-CycleGAN: Blind-Motion-Deblurring-in-Cycles-using-Sparse-Learning (2022)

[![arXiv Prepring](https://img.shields.io/badge/arXiv-Preprint-lightgrey?logo=arxiv)](https://arxiv.org/abs/2111.04026)

 

 Ali Syed Saqlain1, Li Yun Wang2, & Fang Fang1

 


 1 North China Electric Power University, Beijing

 


 2 Portland State University, USA

 

## Abstract

In this paper, we introduce an end-to-end generative adversarial network (GAN) based on sparse learning for single image motion deblurring, which we called SL-CycleGAN. For the first time in image motion deblurring, we propose a sparse ResNet-block as a combination of sparse convolution layers and a trainable spatial pooler k-winner based on HTM (Hierarchical Temporal Memory) to replace non-linearity such as ReLU in the ResNet-block of SL-CycleGAN generators. Furthermore, we take our inspiration from the domain-to-domain translation ability of the CycleGAN, and we show that image deblurring can be cycle-consistent while achieving the best qualitative results. Finally, we perform extensive experiments on popular image benchmarks both qualitatively and quantitatively and achieve the highest PSNR of 38.087 dB on GoPro dataset, which is 5.377 dB better than the most recent deblurring method.

## Network Architecture 



## Results

### GoPro Dataset 

Deblurring results on GoPro test images

# The results shown in Tab I of SL-CycleGAN are conducted on 256x256 while methods like HiNet use full-resolution images from GoPro without resizing them and calculate PSNR and SSIM and The papers from HiNet authors still rank top in Image Deblurring task based on original image size of GoPro dataset. For more their comparison visit their repo for HiNet: [HiNet](https://github.com/megvii-model/HINet)





### Kohler Dataset

Deblurring results on Kohler test images using pre-trained model





### Lai Dataset

Deblurring results on Lai test images via pre-trained model 



### Test images from Pan et al 

Blind deblurring results on images from (Pan et al.) via pre-trained model. The GT images weren't fed to the network only the blurry inputs 



### Deblurring Neptune's satellite Triton 

Test images