cleanerwang/lepf-net

code for LEPF-Net:Light Enhancement Pixel Fusion Network for underwater image enhancement

Preliminary Preparation

1. Input images are put into dataset\haze, reference images are put into dataset\clear, UIEB test set images are put into dataset\test\UIEB, NYU-v2 test set images are put into dataset\test\UWCNN
2. The input image and the test image are named uniformly as dataset name_image parent number_sub number_sub number, for example UIEB_123_1_1, and the reference image is named uniformly as dataset name_image number, for example UIEB_123, and the dataset name_image parent number of the input image and the reference image must correspond, and the same for the test image.
3. Put the model (*.pth) into the snapshots folder
4. Modify the path information in the tran.py
   1. UIEB results folder: config.gen_UIEB_dir
   2. NUYv2 results folder: config.gen_UWCNN_dir
   3. Input image map: config.hazy_images_path
   4. Reference image set: config.orig_images_path
   5. UIEB test set: config.test_UIEB_dir
   6. NUYv2 test set: config.test_UWCNN_dir
   7. Model storage folder: config.snapshots_folder
5. Modify the path information in the test.py file

   1. Reference image set: real_file_path

   2. UIEB test set: UIEB_test_path

   3. NUYv2 test set: UWCNN_test_path

   4. UIEB results folder: UIEB_result_path

   5. NUYv2 results folder: UWCNN_result_path

   6. Model path: `modle_path

testing

python test.py

training

python train.py