This repository contains the code of the following paper "Learning Directional Feature Maps for Cardiac MRI Segmentation (published in MICCAI2020)", https://arxiv.org/abs/2007.11349
Please cite the related works in your publications if it helps your research:
@inproceedings{cheng2020learning,
title={Learning directional feature maps for cardiac mri segmentation},
author={Cheng, Feng and Chen, Cheng and Wang, Yukang and Shi, Heshui and Cao, Yukun and Tu, Dandan and Zhang, Changzheng and Xu, Yongchao},
booktitle={International Conference on Medical Image Computing and Computer-Assisted Intervention},
pages={108--117},
year={2020},
organization={Springer}
}
- Register and download ACDC-2017 dataset from https://www.creatis.insa-lyon.fr/Challenge/acdc/index.html
- Create a folder outside the project with name ACDC_DataSet and copy the dataset.
- From the project folder open file acdc_data_preparation.py.
- In the file, set the path to ACDC training dataset is pointed as:
complete_data_path = '../../ACDC_DataSet/training'
. - Run the script acdc_data_preparation.py.
- The processed data for training is generated outside the project folder named processed_acdc_dataset.
- Run the ./libs/datastes/gen_acdcjson.py to generate the data list for ACDC training and validation.
cd ./tools
python -m torch.distributed.launch --nproc_per_node 4 --master_port $RANDOM train.py --batch_size 12 --mgpus 0,1,2,3 --output_dir logs/... --train_with_eval