ArbSR-PyTorch

Overview

This repository contains an op-for-op PyTorch reimplementation of Learning A Single Network for Scale-Arbitrary Super-Resolution .

Table of contents

• ArbSR-PyTorch
  • Overview
  • Table of contents
  • Download weights
  • Download datasets
  • How Test and Train
    • Test ArbSR_RCAN
    • Train ArbSR_RCAN model
    • Resume train ArbSR_RCAN model
  • Result
  • Contributing
  • Credit
    • Learning A Single Network for Scale-Arbitrary Super-Resolution

Download weights

• Google Driver
• Baidu Driver

Download datasets

Contains DIV2K, DIV8K, Flickr2K, OST, T91, Set5, Set14, BSDS100 and BSDS200, etc.

• Google Driver
• Baidu Driver

Please refer to README.md in the data directory for the method of making a dataset.

How Test and Train

Both training and testing only need to modify the config.py file.

Test ArbSR_RCAN

modify the config.py

• line 31: model_arch_name change to arbsr_rcan.
• line 40: upscale_factor change to 4.
• line 50: mode change to test.
• line 52: exp_name change to ArbSR_RCAN_x1_x4-DIV2K.
• line 101: model_weights_path change to ./results/pretrained_models/ArbSR_RCAN_x1_x4-DIV2K-8c206342.pth.tar.

python3 test.py

Train ArbSR_RCAN model

modify the config.py

• line 31: model_arch_name change to arbsr_rcan.
• line 40: upscale_factor change to 4.
• line 50: mode change to train.
• line 52: exp_name change to ArbSR_RCAN_x1_x4-DIV2K.

python3 train.py

Resume train ArbSR_RCAN model

modify the config.py

• line 31: model_arch_name change to arbsr_rcan.
• line 40: upscale_factor change to 4.
• line 50: mode change to train.
• line 52: exp_name change to ArbSR_RCAN_x1_x4-DIV2K.
• line 57: resume_model_weights_path change to ./results/ArbSR_RCAN_x1_x4-DIV2K/epoch_xxx.pth.tar.

python3 train.py

Result

Source of original paper results: https://openaccess.thecvf.com/content/ICCV2021/papers/Wang_Learning_a_Single_Network_for_Scale-Arbitrary_Super-Resolution_ICCV_2021_paper.pdf

In the following table, the psnr value in () indicates the result of the project, and - indicates no test.

• None

# Download `ArbSR_RCAN_x1_x4-DIV2K-8c206342.pth.tar` weights to `./results/pretrained_models/ArbSR_RCAN_x1_x4-DIV2K-8c206342.pth.tar`
# More detail see `README.md<Download weights>`
python3 ./inference.py

Input:

Output:

Build `arbsr_rcan` model successfully.
Load `arbsr_rcan` model weights `./results/pretrained_models/ArbSR_RCAN_x1_x4-DIV2K-8c206342.pth.tar` successfully.
SR image save to `./figure/119082_lr.png`

Contributing

If you find a bug, create a GitHub issue, or even better, submit a pull request. Similarly, if you have questions, simply post them as GitHub issues.

I look forward to seeing what the community does with these models!

Credit

Learning A Single Network for Scale-Arbitrary Super-Resolution

Longguang Wang, Yingqian Wang, Zaiping Lin, Jungang Yang, Wei An, Yulan Guo

Abstract
Recently, the performance of single image super-resolution (SR) has been significantly improved with powerful networks. However, these networks are developed for image SR with a single specific integer scale (e.g., x2;x3,x4), and cannot be used for non-integer and asymmetric SR. In this paper, we propose to learn a scale-arbitrary image SR network from scale-specific networks. Specifically, we propose a plug-in module for existing SR networks to perform scale-arbitrary SR, which consists of multiple scale-aware feature adaption blocks and a scale-aware upsampling layer. Moreover, we introduce a scale-aware knowledge transfer paradigm to transfer knowledge from scale-specific networks to the scale-arbitrary network. Our plug-in module can be easily adapted to existing networks to achieve scale-arbitrary SR. These networks plugged with our module can achieve promising results for non-integer and asymmetric SR while maintaining state-of-the-art performance for SR with integer scale factors. Besides, the additional computational and memory cost of our module is very small.

[Paper] [Code]

@InProceedings{Wang2020Learning,
  title={Learning A Single Network for Scale-Arbitrary Super-Resolution},
  author={Longguang Wang, Yingqian Wang, Zaiping Lin, Jungang Yang, Wei An, and Yulan Guo},
  booktitle={ICCV},
  year={2021}
}