https://github.com/raoumer/srrescycgan
Code repo for "Deep Cyclic Generative Adversarial Residual Convolutional Networks for Real Image Super-Resolution" (ECCVW AIM2020).
https://github.com/raoumer/srrescycgan
aim2020 cyclic-gan deep-convolutional-neural-networks eccv2020 image-restoration real-image-super-resolution
Last synced: 6 months ago
JSON representation
Code repo for "Deep Cyclic Generative Adversarial Residual Convolutional Networks for Real Image Super-Resolution" (ECCVW AIM2020).
- Host: GitHub
- URL: https://github.com/raoumer/srrescycgan
- Owner: RaoUmer
- License: mit
- Created: 2020-08-15T20:35:53.000Z (about 5 years ago)
- Default Branch: master
- Last Pushed: 2021-03-24T10:24:11.000Z (over 4 years ago)
- Last Synced: 2025-03-29T12:30:34.711Z (6 months ago)
- Topics: aim2020, cyclic-gan, deep-convolutional-neural-networks, eccv2020, image-restoration, real-image-super-resolution
- Language: Python
- Homepage: https://beta.replicate.ai/RaoUmer/SRResCycGAN
- Size: 43.4 MB
- Stars: 47
- Watchers: 5
- Forks: 11
- Open Issues: 2
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# Deep Cyclic Generative Adversarial Residual Convolutional Networks for Real Image Super-Resolution (SRResCycGAN)
An official PyTorch implementation of the [SRResCycGAN](https://github.com/RaoUmer/SRResCycGAN) network as described in the paper **[Deep Cyclic Generative Adversarial Residual Convolutional Networks for Real Image Super-Resolution](https://arxiv.org/abs/2009.03693)**. This work is participated in the [AIM 2020 Real-Image Super-resolution](https://data.vision.ee.ethz.ch/cvl/aim20/) challenge track-3 at the high x4 upscaling factor.
#### Abstract
> Recent deep learning based single image super-resolution (SISR) methods mostly train their models in a clean data domain where the low-resolution (LR) and the high-resolution (HR) images come from noise-free settings (same domain) due to the bicubic down-sampling assumption. However, such degradation process is not available in real-world settings. We consider a deep cyclic network structure to maintain the domain consistency between the LR and HR data distributions, which is inspired by the recent success of CycleGAN in the image-to-image translation applications. We propose the Super-Resolution Residual Cyclic Generative Adversarial Network (SRResCycGAN) by training with a generative adversarial network (GAN) framework for the LR to HR domain translation in an end-to-end manner. We demonstrate our proposed approach in the quantitative and qualitative experiments that generalize well to the real image super-resolution and it is easy to deploy for the mobile/embedded devices. In addition, our SR results on the AIM 2020 Real Image SR Challenge datasets demonstrate that the proposed SR approach achieves comparable results as the other state-of-art methods.
#### Spotlight Video
[](https://youtu.be/QD3yhDpG4Lo)
#### Pre-trained Models
| Datasets|[SRResCycGAN](https://github.com/RaoUmer/SRResCycGAN)|
|---|:---:|
|NTIRE2020 RWSR|[Sensor noise (σ = 8)](https://drive.google.com/file/d/1-N05dWhnA6om16D1VoPASGB9MiTMjdgB/view?usp=sharing)|
|NTIRE2020 RWSR|[JPEG compression (quality=30)](https://drive.google.com/file/d/1AJqEm9lfrzkhJf24_ToEOi9iQUpUu2kn/view?usp=sharing)|
|NTIRE2020 RWSR|[Unknown corruptions](https://drive.google.com/file/d/1tPy1LwzRT2LUM2-X3BhGWo4C3Dii1gmV/view?usp=sharing)|
|AIM2020 RISR|[Real image corruptions](https://drive.google.com/file/d/1NAZjl6UDkcd_BnxfXmwF5QB-uHiJidnA/view?usp=sharing)|
#### BibTeX
@InProceedings{Umer_2020_ECCVW,
author = {Muhammad Umer, Rao and Micheloni, Christian},
title = {Deep Cyclic Generative Adversarial Residual Convolutional Networks for Real Image Super-Resolution},
booktitle = {Proceedings of the European Conference on Computer Vision (ECCV) Workshops},
month = {August},
year = {2020}
}
## Quick Test
This model can be run on arbitrary images with a Docker image hosted on Replicate: https://beta.replicate.ai/RaoUmer/SRResCycGAN. Below are instructions for how to run the model without Docker:
#### Dependencies
- [Python 3.7](https://www.anaconda.com/distribution/) (version >= 3.0)
- [PyTorch >= 1.0](https://pytorch.org/) (CUDA version >= 8.0 if installing with CUDA.)
- Python packages: `pip install numpy opencv-python`
#### Train models
- The SR training code is based on the [SRResCGAN](https://github.com/RaoUmer/SRResCGAN/tree/master/training_codes).
#### Test models
1. Clone this github repository as the following commands:
```
git clone https://github.com/RaoUmer/SRResCycGAN
cd SRResCycGAN
cd srrescycgan_code_demo
```
2. Place your own **low-resolution images** in the `./srrescycgan_code_demo/LR` folder. (There are two sample images i.e. LR_006 and LR_014).
3. Download the pretrained models from **Pre-trained Models** section. Place the models in `./srrescycgan_code_demo/trained_nets_x4`.
4. Run the test. You can config in the `test_srrescycgan.py`.
```
python test_srrescgan.py
```
5. The results are in the `./srrescycgan_code_demo/sr_results_x4` folder.
## SRResCycGAN Architecture
#### Overall Representative diagram
## Quantitative Results
The x4 SR quantitative results comparison of our method with others over the DIV2K validation-set (100 images). The best performance is shown in **red** and the second best
performance is shown in **blue**.
#### The AIM2020 Real Image SR Challenge Results (x4)
| Team | PSNR↑ | SSIM↑ | Weighed_score↑ |
|:---:|:---:|:---:|:---:|
| Baidu| 31.3960 |0.8751 |0.7099 (1)|
| ALONG| 31.2369 |0.8742 |0.7076 (2)|
| CETC-CSKT| 31.1226 |0.8744 |0.7066 (3)|
| SR-IM| 31.2369 |0.8728 |0.7057 |
| DeepBlueAI| 30.9638 |0.8737 |0.7044 |
| JNSR| 30.9988 |0.8722 |0.7035|
| OPPO_CAMERA| 30.8603 |0.8736 |0.7033|
| Kailos| 30.8659 |0.8734 |0.7031|
| SR_DL| 30.6045 |0.8660 |0.6944|
| Noah_TerminalVision| 30.5870 |0.8662 |0.6944|
| Webbzhou| 30.4174 |0.8673 |0.6936|
| TeamInception| 30.3465 |0.8681 |0.6935|
| IyI| 30.3191 |0.8655 |0.6911|
| MCML-Yonsei| 30.4201 |0.8637 |0.6906|
| MoonCloud| 30.2827 |0.8644 |0.6898|
| qwq| 29.5878 |0.8547 |0.6748|
| SrDance | 29.5952 |0.8523 |0.6729|
| **MLP_SR (ours)**| 28.6185 |0.8314 |0.6457|
| EDSR| 28.2120 |0.8240 |0.6356|
| RRDN_IITKGP| 27.9708 |0.8085 |0.6201|
| congxiaofeng| 26.3915 |0.8258 |0.6187|
## Visual Results
### DIV2K Validation-set (100 images)
Here are the SR resutls comparison of our method on the DIV2K validation-set images.
### Real-Image SR Challenge dataset images ([Track-3](https://data.vision.ee.ethz.ch/cvl/aim20/))
#### Validation-set
You can download all the SR resutls of our method on the AIM 2020 Real-Image SR validation-set from the Google Drive: [SRResCycGAN](https://drive.google.com/file/d/1Y-co6-hazt6h9i07sulOka7j0Yeej4lc/view?usp=sharing).
#### Test-set
You can download all the SR resutls of our method on the AIM 2020 Real-Image SR test-set from the Google Drive: [SRResCycGAN](https://drive.google.com/file/d/1VjpxyK7yiD995FCdpbUeTpSv_oiWJTwD/view?usp=sharing).
