Super resolution is a process of upscaling and improving the details within an image. It usually takes a low-resolution image as input and upscales the same image to a higher resolution as output.
Here we provide three super-resolution models, namely [RealSR](https://openaccess.thecvf.com/content_CVPRW_2020/papers/w31/Ji_Real-World_Super-Resolution_via_Kernel_Estimation_and_Noise_Injection_CVPRW_2020_paper.pdf), [ESRGAN](https://arxiv.org/abs/1809.00219v2), [LESRCNN](https://arxiv.org/abs/2007.04344).
[RealSR](https://openaccess.thecvf.com/content_CVPRW_2020/papers/w31/Ji_Real-World_Super-Resolution_via_Kernel_Estimation_and_Noise_Injection_CVPRW_2020_paper.pdf) proposed a realworld super-resolution model aiming at better perception.
[ESRGAN](https://arxiv.org/abs/1809.00219v2) is an enhanced SRGAN that improves the three key components of SRGAN.
[LESRCNN](https://arxiv.org/abs/2007.04344) is a lightweight enhanced SR CNN (LESRCNN) with three successive sub-blocks.
## 1.2 How to use
### 1.2.1 Prepare Datasets
A list of common image super-resolution datasets is as following:
| Name | Datasets | Short Description | Download |
|---|---|---|---|
| 2K Resolution | [DIV2K](https://data.vision.ee.ethz.ch/cvl/DIV2K/) | proposed in [NTIRE17](https://data.vision.ee.ethz.ch/cvl/ntire17//)(800 train and 100 validation) | [official website](https://data.vision.ee.ethz.ch/cvl/DIV2K/) |
| Classical SR Testing | Set5 | Set5 test dataset | [Google Drive](https://drive.google.com/drive/folders/1B3DJGQKB6eNdwuQIhdskA64qUuVKLZ9u) / [Baidu Drive](https://pan.baidu.com/s/1q_1ERCMqALH0xFwjLM0pTg#list/path=%2Fsharelink2016187762-785433459861126%2Fclassical_SR_datasets&parentPath=%2Fsharelink2016187762-785433459861126) |
| Classical SR Testing | Set14 | Set14 test dataset | [Google Drive](https://drive.google.com/drive/folders/1B3DJGQKB6eNdwuQIhdskA64qUuVKLZ9u) / [Baidu Drive](https://pan.baidu.com/s/1q_1ERCMqALH0xFwjLM0pTg#list/path=%2Fsharelink2016187762-785433459861126%2Fclassical_SR_datasets&parentPath=%2Fsharelink2016187762-785433459861126) |
The structure of DIV2K is as following:
```
DIV2K
├── DIV2K_train_HR
├── DIV2K_train_LR_bicubic
| ├──X2
| ├──X3
| └──X4
├── DIV2K_valid_HR
├── DIV2K_valid_LR_bicubic
...
```
The structures of Set5 and Set14 are similar. Taking Set5 as an example, the structure is as following:
```
Set5
├── GTmod12
├── LRbicx2
├── LRbicx3
├── LRbicx4
└── original
```
### 1.2.2 Train/Test
Datasets used in example is df2k, you can change it to your own dataset in the config file. The model used in example is RealSR, you can change other models by replacing the config file.
- 1. [Real-World Super-Resolution via Kernel Estimation and Noise Injection](https://openaccess.thecvf.com/content_CVPRW_2020/papers/w31/Ji_Real-World_Super-Resolution_via_Kernel_Estimation_and_Noise_Injection_CVPRW_2020_paper.pdf)
```
@inproceedings{ji2020real,
title={Real-World Super-Resolution via Kernel Estimation and Noise Injection},
author={Ji, Xiaozhong and Cao, Yun and Tai, Ying and Wang, Chengjie and Li, Jilin and Huang, Feiyue},
booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops},
