This is the code for the final project ELEC 6910X.
Install the necessary requirements from requirement.txt
conda create --name <env> --file requirements.txt
Install Kaggle 2020 dataset at https://www.kaggle.com/datasets/awsaf49/brats20-dataset-training-validation
Install Kaggle 2021 dataset at https://www.kaggle.com/datasets/dschettler8845/brats-2021-task1
Modify the dataset directory on config.py if necessary. The extension of the training data is ".nii", while the extension for test data is ".nii.gz".
Modify config.py to change the experiment configurations.
Data transformation is provided at data_transform.py
Before training manually create "results" folder, "graphs" subfolder, "metrics" subfolder, and "trained_models" subfolder. Also create "Exp" folder for monitoring the training.
To start the training run
python main.py
To monitor the training (e.g. dice loss) use the following command
tensorboard --logdir EXP_PATH
where EXP_PATH depends on config.py (default Exp).
To evaluate the model on test dataset and see the samples, modify the config.py EXP_NAME, LOAD_MODEL_NAME (the file name of the pretrained model), and the model configs and run
python evaluate.py
Samples will be shown on results/samples folder
