Skip to content
Snippets Groups Projects
user avatar
Kerstin Kaspar authored
3b1e170c
History

UNet for Loow Dose CT

This repository contains my implementation of a U-Net similar to the FBPConvNet after:

K. H. Jin, M. T. McCann, E. Froustey and M. Unser, "Deep Convolutional Neural Network for Inverse Problems in Imaging," in IEEE Transactions on Image Processing, vol. 26, no. 9, pp. 4509-4522, Sept. 2017, https://doi.org/10.1109/TIP.2017.2713099 https://github.com/panakino/FBPConvNet

Project structure

Unet
├───mlruns
│    '''trained models and losses'''
├───models
│    '''trained models'''
│   └───losses
│        '''loss files of the trained models'''
├───msub
│    '''scripts for high performance cluster job management'''

│   conda_env_nn.yml
│    '''conda environment file (prerequisites)'''
│   data.py
│    '''functions for data loading and preparation'''
│   data_inspect.py
│    '''script to inspect single data samples (FBP and ground truth)'''
│   eval.py
│    '''script to evaluate data with a trained model'''
│   losses.py
│    '''script to inspect and plot losses of a trained model'''
│   model.py
│    '''UNet class and functions'''
│   orig_model.py
│    '''FBPConvNet class and functions'''
│   README.md
│    '''information on the repository'''
│   train.ipynb
│    '''jupyter notebook to train a model on the high performance cluster'''
│   train.py
│    '''script to train a model on the high performance cluster'''
│   trainer.py
│    '''Trainer class and functions'''
│   train_locally.py
│    '''script to train a model on my local machine'''
│   train_test.py
│    '''script to train a model with small datasets'''
│   utils.py
│    '''additional functions'''