Skip to content
/ WFPN Public

[BMVC 2022] Wide Feature Projection with Fast and Memory-Economic Attention for Efficient Image Super-Resolution

License

Notifications You must be signed in to change notification settings

MinghaoFu/WFPN

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

20 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Minghao Fu, Dongyang Zhang, Min Lei, Kun He, Changyu Li, Jie Shao

WFPN

Figure 1: Network architecture of the proposed Wide Feature Projection Network (WFPN).

Environment

This code is tested on Ubuntu 16.04.6 LTS environment (Python 3.7.11, Pytorch 1.7.1, CUDA 9.0, cuDNN 7.0.5) with TITAN RTX GPU.

Dataset

Get Started

Prepare the environment:

conda create --name WFPN python=3.7 -y
conda activate WFPN

git clone https://github.com/MinghaoFu/WFPN.git
cd WFPN

pip install -r requirements.txt

Configs

Basic training and testing configs could be adjusted in ./configs/plain.json

For specifc model, you could customize it in ./configs/{model}.json

Main Results

Sturctural Reparameterization

Equation 6: Explain how to represent WFP by merging different convolution kernels.

Performance Comparison

Table 2: Performance comparison on benchmark datasets. Number of model parameters is computed on ×4 task. Red indicates the best and blue indicates the second best.

Efficiency Comparison

Table 3: Visual results on Set14 and B100 for ×4 upscaling.

Visual Comparison

Figure 2: Visual results on Set14 and B100 for ×4 upscaling.

License

This repository is licensed under the terms of the MIT license.

Your star is my motivation to update, thanks!

Citation

If you find our work useful in your research or publication, please cite our work:

@inproceedings{DBLP:conf/bmvc/FuZLHLS22,
  author       = {Minghao Fu and
                  Dongyang Zhang and
                  Min Lei and
                  Kun He and
                  Changyu Li and
                  Jie Shao},
  title        = {Wide Feature Projection with Fast and Memory-Economic Attention for
                  Efficient Image Super-Resolution},
  booktitle    = {33rd British Machine Vision Conference 2022, {BMVC} 2022, London,
                  UK, November 21-24, 2022},
  year         = {2022},
}

About

[BMVC 2022] Wide Feature Projection with Fast and Memory-Economic Attention for Efficient Image Super-Resolution

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages