KnowSAM

Official code for "Learnable Prompting SAM-induced Knowledge Distillation for Semi-supervised Medical Image Segmentation"

Installation

To set up the environment and install dependencies, run:

pip install -r requirements.txt

Dataset

We provide Google Drive access links to the datasets employed in this study, among which are two open-source datasets cited in the corresponding paper:

Endoscope Dataset

BCSS Dataset

Extract Sample Data

We provide a reference sample dataset (SampleData.rar) that allows users to quickly test and run the model. Extract the dataset using the following command:

unrar x SampleData.rar

For processed ACDC dataset, you can download it from the ACDC, and place it directly in the SampleData folder.

Training

To train the model on a dataset, execute:

python train_semi_SAM.py

For ACDC dataset training:

python train_semi_SAM_ACDC.py

Prediction

After training, you can make predictions using:

python prediction.py

For ACDC dataset inference:

python prediction_ACDC.py

Acknowledgements

Our code is based on SSL4MIS.

Questions

If you have any questions, welcome contact me at 'taozhou.dreams@gmail.com'