Starting point for E2MIP challenge

You can use the code from this repository as starting point for your work on the E2MIP challenge on the LIDC-IDRI dataset. It provides code for preprocessing and classification of 3D data.

To install the requirements $ pip install -r requirements.txt, the code is tested with python version 3.10.9.

Run Classification

This repository contains a simple 3D Convolutional model and a training and testing algorithm, that can be used as starting point for the challenge.

Data path:

Set the following parameters to the paths of the folders, created by this repository. It provides code to create folders with the datasets in the same way as the data folders that will be used to train and test your submitted code.

• --training_data_path: str, e.g. "=training_data"
• --testing_data_path: str, e.g. "=testing_data_classification"
• --testing_data_solution_path: str, e.g. "=testing_data_solution_classification"

Train:

To train the model, additionally specify the following parameters to run the script 'main.py'

• --train=True: bool, default=False
• --epochs: int, default=100
• --lr: float, default=0.001 (Learning rate)

Test:

To test the model specify the following parameters to run the script 'main.py'

• --test=True: bool, default=False
• --model_path=[path_to_model.pth]: str, (If --train=True, this argument is being ignored and the newly trained model is being tested)

For further questions about this code, please contact luisa.gallee@uni-ulm.de