https://github.com/raoumer/isrrescnet
Code repo for "Deep Iterative Residual Convolutional Network for Single Image Super-Resolution" (ICPR 2020).
https://github.com/raoumer/isrrescnet
convex-optimization deep-neural-networks icpr iterative-methods super-resolution
Last synced: 3 days ago
JSON representation
Code repo for "Deep Iterative Residual Convolutional Network for Single Image Super-Resolution" (ICPR 2020).
- Host: GitHub
- URL: https://github.com/raoumer/isrrescnet
- Owner: RaoUmer
- License: mit
- Created: 2020-11-25T11:48:26.000Z (over 4 years ago)
- Default Branch: main
- Last Pushed: 2021-05-22T13:21:54.000Z (almost 4 years ago)
- Last Synced: 2025-03-29T12:30:38.371Z (25 days ago)
- Topics: convex-optimization, deep-neural-networks, icpr, iterative-methods, super-resolution
- Language: Python
- Homepage: https://beta.replicate.ai/RaoUmer/ISRResCNet
- Size: 10.2 MB
- Stars: 5
- Watchers: 1
- Forks: 2
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# Deep Iterative Residual Convolutional Network for Single Image Super-Resolution (ISRResCNet)
An official PyTorch implementation of the [ISRResCNet](https://github.com/RaoUmer/ISRResCNet) network as described in the paper **[Deep Iterative Residual Convolutional Network for Single Image Super-Resolution](https://arxiv.org/abs/2009.04809)** which is published in the 25th International Conference of Pattern Recognition (ICPR), 2020.
✨ _**Visual examples**_:
[
](https://imgsli.com/NDg4ODY) [
](https://imgsli.com/NDg4ODc)
___________
* [Abstract](#abstract)
* [Oral Presentation Video](#oral-presentation-video)
* [Citation](#bibtex)
* [Quick Test](#quick-test)
* [ISRResCNet Architecture](#isrrescnet-architecture)
* [Quantitative Results](#quantitative-results)
* [Visual Results](#visual-results)
* [Code Acknowledgement](#code-acknowledgement)
#### Abstract
> Deep convolutional neural networks (CNNs) have recently achieved great success for single image super-resolution (SISR) task due to their powerful feature representation capabilities. The most recent deep learning based SISR methods focus on designing deeper / wider models to learn the non-linear mapping between low-resolution (LR) inputs and high-resolution (HR) outputs. These existing SR methods do not take into account the image observation (physical) model and thus require a large number of network's trainable parameters with a great volume of training data. To address these issues, we propose a deep Iterative Super-Resolution Residual Convolutional Network (ISRResCNet) that exploits the powerful image regularization and large-scale optimization techniques by training the deep network in an iterative manner with a residual learning approach. Extensive experimental results on various super-resolution benchmarks demonstrate that our method with a few trainable parameters improves the results for different scaling factors in comparison with the state-of-art methods.
#### Oral Presentation (Video)
[](https://youtu.be/4TLjeIYuOyQ)
#### BibTeX
@InProceedings{Umer_2020_ICPR,
author = {Muhammad Umer, Rao and Luca Foresti, Gian and Micheloni, Christian},
title = {Deep Iterative Residual Convolutional Network for Single Image Super-Resolution},
booktitle = {Proceedings of the International Conference of Pattern Recognition (ICPR)},
month = {January},
year = {2021}
}
## Quick Test
This model can be run on arbitrary images with a Docker image hosted on Replicate: https://beta.replicate.ai/RaoUmer/ISRResCNet. 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`
#### Test models
1. Clone this github repository as the following commands:
```
git clone https://github.com/RaoUmer/ISRResCNet
cd ISRResCNet
cd isrrescnet_code_demo
```
2. Place your own **low-resolution images** in the `./isrrescnet_code_demo/LR` folder. (There are two sample images i.e. set5_img_butterfly_x4 and urban100_img_092_x4).
3. Run the test by the provided script `test_isrrescnet.py`.
```
python test_isrrescnet.py
```
4. The SR results are in the `./isrrescnet_code_demo/sr_results` folder.
## ISRResCNet Architecture
#### Overall Representative diagram
#### ERD block
## Quantitative Results
Average PSNR/SSIM values for scale factors x2, x3, and x4 with the bicubic degradation model. The best performance is shown in **red** and the second best
performance is shown in **blue**.
## Visual Results
Visual comparison of our method with other state-of-the-art methods on the x4 super-resolution over the SR benchmarks. For visual comparison on the benchmarks, you can download our results from the Google Drive: [ISRResCNet](https://drive.google.com/drive/folders/1IioErwfd1cjfBMBOjUzH1guWuI-iZzFm?usp=sharing).
## Code Acknowledgement
The training codes is based on [burst-photography](https://github.com/cig-skoltech/burst-cvpr-2019) and [deep_demosaick](https://github.com/cig-skoltech/deep_demosaick).