3DConvCaps
3DConvCaps copied to clipboard
[ICPR 2022] 3DConvCaps: 3DUnet with Convolutional Capsule Encoder for Medical Image Segmentation
3DConvCaps: 3DUnet with Convolutional Capsule Encoder for Medical Image Segmentation
Table of Contents
- Introduction
- Usage
- Trained models
- Acknowledgement
- Citation
Introduction
The figure above illustrates our 3DConvCaps architecture. Details about it are described in our paper here. The main implementation of this network can be find here.
Usage
Installation
- Clone the repository:
git clone https://github.com/UARK-AICV-Lab/3DConvCaps
- Install dependencies depends on your cuda version (CUDA 10 or CUDA 11)
conda env create -f environment_cuda11.yml
or
conda env create -f environment_cuda10.yml
Data preparation
Our method is evaluated on three datasets:
- iSeg-2017 challenge (infant brain MRI segmentation): https://iseg2017.web.unc.edu/download/
- Cardiac and Hippocampus dataset from Medical Segmentation Decathlon: http://medicaldecathlon.com/
See this repository for more details on data preparation.
Training
The training example script is available here
Validation
The evaluating example script is available here
See this repository for more details on training and evaluating parameters.
Trained models
Our trained 3DConvCaps models on three datasets can be downloaded as follows:
Acknowledgement
The implementation is mainly based on 3DUCaps thorough implementation.
Citation
@article{tran20223dconvcaps,
title={3DConvCaps: 3DUnet with Convolutional Capsule Encoder for Medical Image Segmentation},
author={Tran, Minh and Vo-Ho, Viet-Khoa and Le, Ngan TH},
journal={arXiv preprint arXiv:2205.09299},
year={2022}
}
Contacts
If you have any question, feel free to open an issue.