SRGAN-PyTorch

Overview

This repository contains an op-for-op PyTorch reimplementation of Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network.

Table of contents

• SRGAN-PyTorch
  • Overview
  • Table of contents
  • Download weights
  • Download datasets
  • How Test and Train
    • Test
    • Train SRResNet model
    • Resume train SRResNet model
    • Train SRGAN model
    • Resume train SRGAN model
  • Result
  • Contributing
  • Credit
    • Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network

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

• line 31: upscale_factor change to 4.
• line 33: mode change to test.
• line 112: model_path change to results/pretrained_models/SRResNet_x4-ImageNet-2096ee7f.pth.tar.

python3 test.py

Train SRResNet model

• line 31: upscale_factor change to 4.
• line 33: mode change to train_srresnet.
• line 35: exp_name change to SRResNet_baseline.
• line 49: pretrained_model_path change to ./results/pretrained_models/SRResNet_x4-ImageNet-2096ee7f.pth.tar.

python3 train_srresnet.py

Resume train SRResNet model

• line 31: upscale_factor change to 4.
• line 33: mode change to train_srresnet.
• line 35: exp_name change to SRResNet_baseline.
• line 52: resume change to samples/SRResNet_baseline/g_epoch_xxx.pth.tar.

python3 train_srresnet.py

Train SRGAN model

• line 31: upscale_factor change to 4.
• line 33: mode change to train_srgan.
• line 35: exp_name change to SRGAN_baseline.
• line 77: pretrained_g_model_path change to ./results/SRResNet_baseline/g_best.pth.tar.

python3 train_srgan.py

Resume train SRGAN model

• line 31: upscale_factor change to 4.
• line 33: mode change to train_srgan.
• line 35: exp_name change to SRGAN_baseline.
• line 80: resume_d change to samples/SRGAN_baseline/g_epoch_xxx.pth.tar.
• line 81: resume_g change to samples/SRGAN_baseline/g_epoch_xxx.pth.tar.

python3 train_srgan.py

Result

Source of original paper results: https://arxiv.org/pdf/1609.04802v5.pdf

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

Set5	Scale	SRResNet	SRGAN
PSNR	4	32.05(32.16)	29.40(29.08)
SSIM	4	0.9019(0.8961)	0.8472(0.8305)

Set14	Scale	SRResNet	SRGAN
PSNR	4	28.49(28.62)	26.02(25.89)
SSIM	4	0.8184(0.7831)	0.7397(0.6932)

BSD100	Scale	SRResNet	SRGAN
PSNR	4	27.58(27.59)	25.16(24.91)
SSIM	4	0.7620(0.7379)	0.6688(0.6354)

# Download `SRGAN_x4-ImageNet-c71a4860.pth.tar` weights to `./results/pretrained_models`
# More detail see `README.md<Download weights>`
python ./inference.py --inputs_path ./figure/comic_lr.png --output_path ./figure/comic_sr.png --weights_path ./results/pretrained_models/SRGAN_x4-ImageNet-c71a4860.pth.tar

Input:

Output:

Build SRGAN model successfully.
Load SRGAN model weights `./results/pretrained_models/SRGAN_x4-ImageNet-c71a4860.pth.tar` successfully.
SR image save to `./figure/comic_sr.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

Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network

Christian Ledig, Lucas Theis, Ferenc Huszar, Jose Caballero, Andrew Cunningham, Alejandro Acosta, Andrew Aitken, Alykhan Tejani, Johannes Totz, Zehan Wang, Wenzhe Shi

Abstract
Despite the breakthroughs in accuracy and speed of single image super-resolution using faster and deeper convolutional neural networks, one central problem remains largely unsolved: how do we recover the finer texture details when we super-resolve at large upscaling factors? The behavior of optimization-based super-resolution methods is principally driven by the choice of the objective function. Recent work has largely focused on minimizing the mean squared reconstruction error. The resulting estimates have high peak signal-to-noise ratios, but they are often lacking high-frequency details and are perceptually unsatisfying in the sense that they fail to match the fidelity expected at the higher resolution. In this paper, we present SRGAN, a generative adversarial network (GAN) for image super-resolution (SR). To our knowledge, it is the first framework capable of inferring photo-realistic natural images for 4x upscaling factors. To achieve this, we propose a perceptual loss function which consists of an adversarial loss and a content loss. The adversarial loss pushes our solution to the natural image manifold using a discriminator network that is trained to differentiate between the super-resolved images and original photo-realistic images. In addition, we use a content loss motivated by perceptual similarity instead of similarity in pixel space. Our deep residual network is able to recover photo-realistic textures from heavily downsampled images on public benchmarks. An extensive mean-opinion-score (MOS) test shows hugely significant gains in perceptual quality using SRGAN. The MOS scores obtained with SRGAN are closer to those of the original high-resolution images than to those obtained with any state-of-the-art method.

[Paper]

@InProceedings{srgan,
    author = {Christian Ledig, Lucas Theis, Ferenc Huszar, Jose Caballero, Andrew Cunningham, Alejandro Acosta, Andrew Aitken, Alykhan Tejani, Johannes Totz, Zehan Wang, Wenzhe Shi},
    title = {Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network},
    booktitle = {arXiv},
    year = {2016}
}