Environment setting

The environment setting is in the environment.yaml.

Dataset

training dataset: https://pan.baidu.com/s/1NRX8s1qM6Gad8INjK4Lqxg Extraction code: uhfw

test dataset : updata new link:https://pan.baidu.com/s/1EzufxX5hYRyrWJLowZm7Ng Extraction code: 4pfw

(The old link https://pan.baidu.com/s/1_2YrFEyJWk73WEMaL4L3Qg Extraction code: 4pfw has expired.)

There are 3 datasets in the filefolder "tesetdaset"for testing,"inputa" is the synthesized low-light underwater images,"inputb" is the real underwater images,and "inputc" is the real low-light images

We also provid a real low-light underwater video example of LDS-Net,you can download the example from https://pan.baidu.com/s/1PRXTwm2tR16eeHBRI_e-AQ Extraction code: 32wc.The video was taken from the waters of the Great Barrier Reef in Australia.We obtained this video from the existing documentary the Great Barrier Reef.

Checkpoints

The checkpoints was saved in https://pan.baidu.com/s/1Oy2f5b39528EFEH2de4TbA Extraction code: 4k72

Before your test,please download the checkpoints and put the checkpoints in the file folder "checkpoints_nt".

If you want to retrain, you can change the training parameter settings by config.yaml

Bibtex

@ARTICLE{10.3389/fmars.2022.921492,

AUTHOR={Xie, Yaofeng and Yu, Zhibin and Yu, Xiao and Zheng, Bing},

TITLE={Lighting the darkness in the sea: A deep learning model for underwater image enhancement},

JOURNAL={Frontiers in Marine Science},

VOLUME={9},

YEAR={2022},

URL={https://www.frontiersin.org/articles/10.3389/fmars.2022.921492},

DOI={10.3389/fmars.2022.921492}