- Input images are put into
dataset\haze
, reference images are put intodataset\clear
, UIEB test set images are put intodataset\test\UIEB
, NYU-v2 test set images are put intodataset\test\UWCNN
- The input image and the test image are named uniformly as
dataset name_image parent number_sub number_sub number
, for exampleUIEB_123_1_1
, and the reference image is named uniformly asdataset name_image number
, for exampleUIEB_123
, and thedataset name_image parent number
of the input image and the reference image must correspond, and the same for the test image. - Put the model
(*.pth)
into thesnapshots
folder - Modify the path information in the
tran.py
- UIEB results folder:
config.gen_UIEB_dir
- NUYv2 results folder:
config.gen_UWCNN_dir
- Input image map:
config.hazy_images_path
- Reference image set:
config.orig_images_path
- UIEB test set:
config.test_UIEB_dir
- NUYv2 test set:
config.test_UWCNN_dir
- Model storage folder:
config.snapshots_folder
- UIEB results folder:
- Modify the path information in the
test.py
file-
Reference image set:
real_file_path
-
UIEB test set:
UIEB_test_path
-
NUYv2 test set:
UWCNN_test_path
-
UIEB results folder:
UIEB_result_path
-
NUYv2 results folder:
UWCNN_result_path
-
Model path: `modle_path
-
python test.py
python train.py