Semi-Supervised Medical Image Segmentation through Dual-Task Consistency

This is an implementation of the dual-task network used for medical image segmentation in TensorFlow. The original paper was published at AAAI 2021, and the original source code (using PyTorch) can be found here. This code was written as part of the CISC 867 Deep Learning course at Queen's University in Kingston, following the ML Reproducibility Challenge 2021.

Usage

1. Clone the repo:

git clone https://github.com/chriscyyeung/DTC-TensorFlow.git

2. Install required packages:

pip install -r requirements.txt

3. Move to code directory:

cd code

4. Train the model:

python main.py -p train

5. Test the model:

python main.py -p test

The hyperparameters can be changed by modifying the values in the configs/config.json file, and data augmentation transformations can be modified in the configs/pipeline.yaml file.