rwickman / Ultrasound

Setup

Install the package and dependencies:

pip install .

In order to train faster, I converted the .mat input files into .npy files. You will need to first copy create_npy.py to the directory of the datasets and then run:

python create_npy.py

All the necessary configurations for this project can be found ultrasound/train/config.py.

Training

python ultrasound/train/train.py

To plot the training figures, first change the model_dir variable in ultrasound/util/plot_results and then run:

python ultrasound/util/plot_results.py

To show results of the current model run:

python ultrasound/util/plot_validation.py

This currently plots the results on the training dataset. To change it to a different set, change argument train_dataset given to val_loader to val_dataset or test_dataset.

Test the model

To get the average SSIM and PSNR scores on the test datset run:

python ultrasound/train/test.py