ActiveCMR is a research project focused on applying Active Learning techniques to Cardiac Magnetic Resonance (CMR) image segmentation. The goal is to improve data efficiency and model performance by selecting the most informative MRI slices for annotation. The project incorporates uncertainty estimation and adaptive data acquisition strategies to guide active sampling during scanning.
- Clone the repository:
git clone https://github.com/mmmmm-w/ActiveCMR.git cd ActiveCMR - Create a conda environment (recommended):
conda create -n cmr python=3.10 -y conda activate cmr
- Install dependencies:
pip install -r requirements.txt
You can download from https://data.mendeley.com/datasets/pw87p286yx/1 and unzip Dataset.zip into this directory:
unzip Dataset.zip -d Dataset/Demo
python demo/demo.py
Training
python scripts/train_cvae.py
You may adjust configurations in the training script.