These instructions will get you a copy of the project up and running on your local machine for development and testing purposes.
You would need to install the following software before replicating this framework in your local or server machine.
Python version 3.7+
Aanaconda version 3+
TensorFlow version 2.6.0
Keras version 2.6.0
- Retrieve the code
git clone https://github.com/manisa/IterLUNet.git
cd IterLUNet
- Create and activate the virtual environment with python dependendencies.
conda create -n gpu-tf tensorflow-gpu
conda activate gpu-tf
source installPackages.sh
IterLUNet/
archs/
lib/
dataset/
experiment_3/
models/
experiment_3/
- Experiment 1 train data and test data
- Experiment 2 train and validation data and test_data
- Experiment 3 train and validation data and test_data
- Unzip and copy dataset from the respecitve experiment into the folder dataset inside the root folder IterLUNet.
- Your directory structure should look like this:
IterLUNet/
dataset/
experiment_3/
train/
images/
masks/
test/
images/
masks/
- Best performing trained models from 10-Fold CV Experiment
- Experiment 2 best performing trained models over 80 epochs
- Experiment 3 best performing trained models over 150 epochs
- Unzip and copy models from respective experiment to models inside the root folder IterLUNet.
- Your directory structure should look like this:
IterLUNet/
models/
experiment_1/
experiment_2/
experiment_3/
- To replicate the training procedure, follow following command line.
cd src
python train.py --model_type=iterlunet --input_filters=64 --lr=2e-3 --loss_function='focal_tversky_loss' --model_path='./models/iterlunet' --train_valid_path='./datasets/experiment_3/train/'
Manisha Panta, Md Tamjidul Hoque, Mahdi Abdelguerfi, Maik C. Flanagin
This project is licensed under the MIT License - see the LICENSE.md file for details