This paper has been accepted by AAAI2024.
If you have any quesions, feel free to contact me. My E-mail and WeChat can be found at my homepage: [Homepage]
Download underwater imagery datasets from websites or papers. Follow the organization form below.
├── dataset_name
├── train
├── images
├── im1.jpg
├── im2.jpg
└── ...
├── labels
├── im1.jpg
├── im2.jpg
└── ...
├── val
├── images
├── im1.jpg
├── im2.jpg
└── ...
├── labels
├── im1.jpg
├── im2.jpg
└── ...
├── test
├── images
├── im1.jpg
├── im2.jpg
└── ...
├── labels (if you have)
├── im1.jpg
├── im2.jpg
└── ...
├── masks (if you have)
├── im1.jpg
├── im2.jpg
└── ...
Note: If you only need to perform color enhancement, you can ignore this step.
Use the EUVP-scenes if you want fine-tuning color ability!!! The diversity of UIEB and EUVP-imagenet are not enough. Besides, the images with less distortion are required. Therefore, the EUVP-scenes is needed !!!
Put the config file in folder task_CECF/configs. For example task_CECF/configs/UIEB_3090_dim8_1m_bs6_NoTransBlock_SSIM.yaml
The batch_size must be greater than or equal 4.
cd task_CECF
python train_CECF_TransBlock.py --config configs/UIEB_3090_dim8_1m_bs6_NoTransBlock_SSIM.yaml --device cuda:7 --res_dir ../results/CECF/ --grad_acc 1
If you need to compare with our algorithm, execute the following command.
cd task_CECF
python test_enhancement.py --config configs/dim8_150k_bs6_NoTransBlock_SSIM.yaml \
--input_folder ../demo_dataset/your_test_images/images/ \
--output_folder ../results/CECF/natural/ \
--checkpoint ../results/gen_00000000.pt \
--device cuda:0
cd task_CECF
python test_fine_tune.py --config configs/dim8_150k_bs6_NoTransBlock_SSIM.yaml \
--input_folder ../demo_dataset/your_test_images/images/ \
--guide_path ../guidance/guide_natural/ \
--output_folder ../results/CECF/natural/ \
--checkpoint ../results/gen_00000000.pt \
--device cuda:0 \
--subfolder_prefix natural_