yifliu3 / EPT-Net

Edge-oriented Point cloud Transformer for 3D Intracranial Aneurysm Segmentation. MICCAI22
MIT License
13 stars 7 forks source link
3dsegmentation

Edge-oriented Point-cloud Transformer for 3D Intracranial Aneurysm Segmentation

by Yifan Liu

1.Introduction

This repository is for our MICCAI 2022 paper "Edge-oriented Point cloud Transformer for 3D Intracranial Aneurysm Segmentation"

2.Data Preparation

Download fileSplit, geo.zip and IntrA.zip from IntrA repository

Unzip geo.zip and IntrA.zip into geo and IntrA foler

Move the unzipped geo folder into IntrA/annoated/geo

Move the fileSplit into IntrA/split

Create one foler data in the code respository and add one symbolic link

mkdir data && ln -s Yourpath/IntrA data/IntrA

3. Installation

Requirements

Step-by-step installation

# create python environment
conda create -n ept python=3.7
conda activate ept

# install dependencies
conda install pytorch==1.7.0 torchvision==0.8.0 torchaudio==0.7.0 cudatoolkit=10.1 -c pytorch
conda install -c anaconda h5py pyyaml -y
pip install tensorboardx

# clone this repository in your own workspace
git clone https://github.com/CityU-AIM-Group/EPT.git
cd EPT
mkdir data && ln -s Yourpath/IntrA data/IntrA

# compile cuda operations
cd point_transformer_lib
python3 setup.py build_exit install

4. Train/test the Model

To separately train and test you can use the commands below (take 512 sampling as an example):
Train:
python -m tool.train --config config/IntrA/IntrA_pointtransformer_seg_repro sample_points 512
Test:
python -m tool.test --config config/IntrA/IntrA_pointtransformer_seg_repro sample_points 512

Or you can use the bash scipt to run train.py and test.py sequentially:
sh tool/ept.sh IntrA pointtransformer_seg_repro

The trained models are provided in Google Drive

5. Citation

If you find this work useful for your research, please cite our paper:

@inproceedings{liu2022,
  title={Edge-oriented Point-cloud Transformer for 3D Intracranial Aneurysm Segmentation},
  author={Yifan Liu, Jie Liu and Yixuan Yuan},
  booktitle= {MICCAI},
  year={2022}
}

6. Acknowledgement

This work is based on point-transformer.