/dncnn

DnCNN Keras Implementation

Primary LanguagePython

State of the Art Deep learning based "Image Denoising" algorithm : DnCNN implementation in Keras and Pytorch for dicom, jpeg and numpy data.

How to Run the Training:

  1. Run the following command from the terminal: “Python main_train.py “ and pass the following arguments:
--batch_size
--val_data 
--train_data 
-- sigma 
--epoch 
--lr 
--save_every 

All the trained model weights would be saved in the “models” directory of this project.

How to Run the Inference:

Run the following command from the terminal: “Python main_test.py” and pass the following arguments:

--set_dir 
--sigma 
--model_dir 
--model_name 
-- result_dir 
-- save_results 

. dncnn                   Code Root Directory
  |- data                 Images directory
    |- test               Test images directory	
    |-train               Training images 
    |- val                validation images 
  |- data_generator.py    data loader method implementation
  |- data_transform.py    PyTorch data transformations implementation
  |- main_train.py        Projects main file to start the training 
  |- main_test.py         Projects main file to run inference
  |- logs                 Tensor board logs
  |- models               saved model weights
  |- results              inference output images