DiCo: Learning Dynamic Collaborative Network for Semi-supervised 3D Vessel Segmentation (CVPR 2025)

Introduction

Learning Dynamic Collaborative Network for Semi-supervised 3D Vessel Segmentation,
Jiao Xu, Xin Chen, Lihe Zhang.
In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2025
[bibtex][supp]

News

• [2025.10.22] Our codes are released!
• [2025.06.16] Repo created. Paper and code will come soon.

Installation

• PyTorch 2.1.2
• CUDA 11.8
• Python 3.8.18

Usage

Dataset and Pre-processing

The datasets used in our paper are ImageCAS dataset, Parse2022 dataset, and CAS2023 dataset.

Training Steps

Note: Please replace the original unetr.py in the MONAI library with code/networks/unetr.py from this repository.
For more details, please refer to issue #4.

1. Clone the repo and create data path:

git clone https://github.com/xujiaommcome/DiCo
cd DiCo
mkdir data # create data path

2. Put the preprocessed data and then cd code
3. We train our model on one single NVIDIA 3090 GPU for each dataset.

To produce the claimed results for Parse2022 dataset:

# For 5% labeled data,
CUDA_VISIBLE_DEVICES=0 python DiCo_main.py --labelnum=5

Citation

If this code is useful for your research, please consider giving star to our repository and citing our work:

@inproceedings{xu2025learning,
  title={Learning Dynamic Collaborative Network for Semi-supervised 3D Vessel Segmentation},
  author={Xu, Jiao and Chen, Xin and Zhang, Lihe},
  booktitle={Proceedings of the Computer Vision and Pattern Recognition Conference},
  pages={10445--10454},
  year={2025}
}

Questions

If you have any questions, welcome contact me at 'xjmmcome@mail.dlut.edu.cn'