/AKSVD_official

ATTENTIVE DEEP K-SVD NETWORK FOR PATCH CORRELATED IMAGE DENOISING

Primary LanguagePythonMIT LicenseMIT

AKSVD: Attentive Deep K-SVD Network

ATTENTIVE DEEP K-SVD NETWORK FOR PATCH CORRELATED IMAGE DENOISING

process_1

directory structure

├── AKSVD_function.py     # model and data processing
├── AKSVD_training.py	  # main training code
├── assets
│   └── process_1.png
├── cbam.py               # attention module
├── load_model.py         # main testing code
├── README.md
├── requirements.txt
├── gray                  # BSDS
│   ├── *.jpg
├── Set12                
│   ├── *.jpg
├── test_gray.txt
├── test_set12.txt
├── train_gray.txt
└── visualization.py

Quick Start

  1. Installation

    pip install -r requirements.txt

  2. Run the training code

    python AKSVD_training.py

  3. Run the testing code

    • First, set your model path
    # load_model.py
model.load_state_dict(torch.load("../model.pth", map_location="cpu"))  
model.to(device)
model.eval()

    • Second, set the name of dataset

      # Test image names:
file_test = open("test_set12.txt", "r")  # line 57

# Test Dataset:
my_Data_test = AKSVD_function.FullImagesDataset(
    root_dir="Set12", image_names=onlyfiles_test, sigma=sigma, transform=data_transform
)  # line 79

    • Finally, run the testing code

      python load_model.py

Reference