# DCSR

**Repository Path**: Wuhan3D/DCSR

## Basic Information

- **Project Name**: DCSR
- **Description**: No description available
- **Primary Language**: C++
- **License**: Not specified
- **Default Branch**: master
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2021-09-26
- **Last Updated**: 2021-09-26

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

# DCSR: Dual Camera Super-Resolution

> Dual-Camera Super-Resolution with Aligned Attention Modules    
> ICCV 2021 (Oral Presentation)     

[paper](https://arxiv.org/abs/2109.01349) | [project website](https://tengfei-wang.github.io/Dual-Camera-SR/index.html) | [dataset](https://drive.google.com/file/d/1SxU6D1yYTTnZnCyytTObsZxZQigWLciT/view?usp=sharing) | [demo video](https://youtu.be/5TiUfAcNvuw) | [results on CUFED5](https://hkustconnect-my.sharepoint.com/:f:/g/personal/tfwang_connect_ust_hk/EidZ5B1jPC9PmTlSUtrMbN0B4a2VY1hXrteYZevijllhJg?e=hQwva7)


## Introduction
We present a novel approach to reference-based super resolution (RefSR) with the focus on real-world dual-camera super resolution (DCSR).

<img src="pics/demo.png" width="720px"/> 

## Results
4X SR results on CUFED5 testset can be found in [this link](https://hkustconnect-my.sharepoint.com/:f:/g/personal/tfwang_connect_ust_hk/EidZ5B1jPC9PmTlSUtrMbN0B4a2VY1hXrteYZevijllhJg?e=hQwva7).

More 2X SR results on CameraFusion dataset can be found in our [project website](https://tengfei-wang.github.io/Dual-Camera-SR/index.html).

<img src="pics/result-cufed.jpg" width="415px"/>  <img src="pics/result-CF.jpg" width="400px"/> 

## Setup
### Installation
```
git clone https://github.com/Tengfei-Wang/DualCameraSR.git
cd DualCameraSR
```

### Environment
The environment can be simply set up by Anaconda:
```
conda create -n DCSR python=3.7
conda activate DCSR
pip install -r requirements.txt
```

## Dataset
Download our CameraFusion dataset from [this link](https://drive.google.com/file/d/1SxU6D1yYTTnZnCyytTObsZxZQigWLciT/view?usp=sharing).
This dataset currently consists of 143 pairs of telephoto and wide-angle images in 4K resolution captured by smartphone dual-cameras.
```
mkdir data
cd ./data
unzip CameraFusion.zip
```

## Quick Start
The pretrained models have been put in `./experiments/pretrain`. For quick test, run the scipts: 

```
# For 4K test (with ground-truth High-Resolution images):
sh test.py

# For 8K test (without SRA):
sh test_8k.sh

# For 8K test (with SRA):
sh test_8k_SRA.sh
```


## Training
To train the DCSR model on CameraFusion, run:
```
sh train.sh
```
The trained model should perform well on 4K test, but may suffer performance degradation on 8K test.

After the regular training, we can use Self-supervised Real-image Adaptation (SRA) to finetune the trained model for real-world 8K image applications:
```
sh train_SRA.sh
```


## Citation
If you find this work useful for your research, please cite:
``` 
@InProceedings{wang2021DCSR,
author = {Wang, Tengfei and Xie, Jiaxin and Sun, Wenxiu and Yan, Qiong and Chen, Qifeng},
title = {Dual-Camera Super-Resolution with Aligned Attention Modules},
booktitle = {International Conference on Computer Vision (ICCV)},
year = {2021}
}
```

##  Acknowledgement
We thank the authors of [EDSR](https://github.com/sanghyun-son/EDSR-PyTorch), [CSNLN](https://github.com/SHI-Labs/Cross-Scale-Non-Local-Attention), [TTSR](https://github.com/researchmm/TTSR) and [style-swap](https://github.com/rtqichen/style-swap) for sharing their codes.