Code for the paper: LDTSF: A Label-decoupling Teacher-student Framework for Semi-supervised Echocardiography Segmentation (ICASSP2023)
Some important requires packages includes:
- Pytorch version >=1.10.0
- torchvision >=0.11.0
- Python == 3.8
- Some basic python packages such as Numpy, Scikit-image, SimpleITK, Scipy ......
Follow official guidance to install Pytorch.
- Clone the repo:
git clone https://github.com/SwanKnightZJP/LDTSF
cd LDTSF
- Put the data in data/2018LA_Seg_Training Set.
- Train the LDC model first. The trained model will be saved at ../model/LA/LDC_with_consis_weight
cd code
python train_la_LDC.py
- Generate the pseudo labels. The generated pseudo labels will be saved at ../data/2018LA_Seg_PseudoTraining Set
cd code
python pseudo_label_create_LA.py
- Train the LDTSF model.
cd code
python train_la_LDTSF.py
Due to patent-related issues, we are not at liberty to disclose the code and data related to 3D Echocardiography for the time being, and we may make relevant updates in the future.
- This code is adapted from DTC, UA-MT, SASSNet, SegWithDistMap.
- We thank Dr. Xiangde Luo, Dr. Lequan Yu, M.S. Shuailin Li and Dr. Jun Ma for their elegant and efficient code base.