Image Enhancement with Bi-directional Normalization and Color Attention-guided Generative Adversarial Networks
This paper is submitted to International Journal Of Multimedia Information Retrieval, and the extended and updated version of BNCAGAN will be released when it is accepted by International Journal Of Multimedia Information Retrieval.
The new version of BNCAGAN is in the IJMIR folder. The old version of BNCAGAN can be used as follows:
We recommended the following dependencies.
- Python 3.7
- PyTorch 1.7.1
- tqdm 4.42.1
- munch 2.5.0
- torchvision 0.8.2
Prepare the training, testing, and validation data. The folder structure should be:
data
└─── fiveK
├─── train
| ├─── exp
| | ├──── a1.png
| | └──── ......
| └─── raw
| ├──── b1.png
| └──── ......
├─── val
| ├─── label
| | ├──── c1.png
| | └──── ......
| └─── raw
| ├──── c1.png
| └──── ......
└─── test
├─── label
| ├──── d1.png
| └──── ......
└─── raw
├──── d1.png
└──── ......
raw/contains low-quality images, exp/ contains unpaired high-quality images, and label/ contains corresponding ground truth.
To train BNCAGAN on FiveK, run the training script below.
python main.py --mode train --version BNCAGAN-FiveK --use_tensorboard False \
--is_test_nima True --is_test_psnr_ssim True
To test BNCAGAN on FiveK, run the test script below.
python main.py --mode test --version BNCAGAN-FiveK --pretrained_model xx (best epoch, e.g., 100) \
--is_test_nima True --is_test_psnr_ssim True
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent