Dynamic Attentive Graph Learning for Image Restoration

This repository is for GATIR introduced in the following paper:
Chong Mou, Jian Zhang, Zhuoyuan Wu; Dynamic Attentive Graph Learning for Image Restoration; IEEE International Conference on Computer Vision (ICCV) 2021 [arxiv]

The pre-trained models are available at Google Drive

Requirements

  • Python 3.6
  • PyTorch >= 1.1.0
  • numpy
  • skimage
  • cv2

Introduction

In this paper, we propose an improved graph attention model for image restoration. Unlike previous non-local image restoration methods, our model can assign an adaptive number of neighbors for each query item and construct long-range correlations based on feature patches. Furthermore, our proposed dynamic attentive graph learning can be easily extended to other computer vision tasks. Extensive experiments demonstrate that our proposed model achieves state-of-the-art performance on wide image restoration tasks: synthetic image denoising, real image denoising, image demosaicing, and compression artifact reduction.

Network

Citation

If you find our work helpful in your resarch or work, please cite the following paper.

@inproceedings{mou2021gatir,
  title={Dynamic Attentive Graph Learning for Image Restoration},
  author={Chong, Mou and Jian, Zhang and Zhuoyuan, Wu},
  booktitle={IEEE International Conference on Computer Vision},
  year={2021}
}

GitHub

https://github.com/jianzhangcs/DAGL