This repository includes the Pytorch implementation of the methods described in our paper DECAPS: Detail-Oriented Capsule Networks.
Fig1. Processing pipeline of the proposed DECAPS.
- For training with default setup:
python train.py
You can easily train your desired network configuration by passing the desired arguments as provided
in the config.py file. For example:
- For training with batch size of 8:
python train.py --batch_size=8
-
For testing the pretrained model run:
python inference.py -
For testing your trained model run:
python inference.py --load_model_path=path_to_your_trained_model
If you found this repo useful, please use this bibtex to cite our paper:
@inproceedings{mobiny2020decaps,
title={DECAPS: Detail-Oriented Capsule Networks},
author={Mobiny, Aryan and Yuan, Pengyu and Cicalese, Pietro Antonio and Van Nguyen, Hien},
booktitle={International Conference on Medical Image Computing and Computer-Assisted Intervention},
pages={148--158},
year={2020},
organization={Springer}
}