ixi_config.yaml

                                    #Data parameters
dataset: "IXI"                                                      # Dataset type. currently only IXI supported.
data_dir: "/home/jc3/Data/IXI_T2/h5/"
train_data_dir: "/home/jc3/Data/IXI_T2/mat/train"                                    # Training files dir, should contain hdf5 preprocessed data.
val_data_dir: "/home/jc3/Data/IXI_T2/mat/val"                                        # Validation files dir, should contain hdf5 preprocessed data.
output_dir: "./logs"                                           # Directory to save checkpoints and tensorboard data.
sampling_percentage: 30                                             # Sampling mask precentage (provided with the code 20%,30% and 50% sampling masks of 256X256).
num_input_slices: 1                                                 # Num of slices to use for input (3 means the predicted slice + previous slice + next slice).
img_size: 256                                                       # Input image size (256X256 for IXI).
# img_size2: 640
slice_range: [20,120]                                              # Slices to use for training data.

                                  #Load checkpoint
load_cp: 0                                                          # 0 to start a new training or checkpoint path to load network weights.
resume_training: 1                                                  # 0 - Load only model weights , 1 - Load Weights + epoch number + optimizer and scheduler state.

                                  #Networks parameters
bilinear: 1                                                         # 1 - Use bilinear upsampling , 0 - Use up-conv.
crop_center: 128                                                    # Discriminator center crop size (128X128), to avoid classifying blank patches.

                                  #Training parameters
lr: 0.001                                                           # Learning rate
epochs_n: 50                                                        # Number of epochs
batch_size: 64                                                      # Batch size. Reduce if out of memory.Batch size of 32 256X256 images needs ~13GB memory.
GAN_training: 1                                                     # 1 - Use GAN training. 0 - No GAN (no discriminator training and adverserial loss)
loss_weights: [1000, 1000, 5, 0.1 ]                                 # Loss weighting [Imspac L2, Imspace L1, Kspace L2, GAN_Loss]. Losses are weighted to be roughly at the same scale.
minmax_noise_val: [-0.01, 0.01]

                                  #Tensorboard
tb_write_losses: 1                                                  # Write losses and scalars to Tensorboard.
tb_write_images: 1                                                  # Write images to Tensorboard.

                                  #Runtime
device: 'cuda'                                                      # For GPU training : 'cuda', for CPU training (not recomended!) 'cpu'.
gpu_id: '0'                                                         # GPU ID to use.
train_num_workers: 16                                               # Number of training dataset workers. Reduce if you are getting a shared memory error.
val_num_workers: 4                                                  # Number of validation dataset workers. Reduce if you are getting a shared memory error.

                                 #Predict parameters
save_prediction: 1                                                  # Save predicted images.
save_path: "home/jc3/multiSR/JS_fastMRI/save_predictions/T2T1_Unet"                                              # Path to save predictions
visualize_images: 1                                                 # Visualize predicted images.
model: "model_path.pth"                                             # Model checkpoint to use for prediction.
predict_data_dir: "/home/jc3/Data/IXI_T2/h5/test"                                   # Test set files dir, should contain hdf5 preprocessed data.

# jc3/multiSR/JS_fastMRI/SR_fastMRI-master/experimental/i