Skip to content

[MICCAI 2020] BiO-Net: Learning Recurrent Bi-directional Connections for Encoder-Decoder Architecture

License

Notifications You must be signed in to change notification settings

tiangexiang/BiO-Net

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

30 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

BiO-Net

Official implementation of "BiO-Net: Learning Recurrent Bi-directional Connections for Encoder-Decoder Architecture", MICCAI 2020

Paper: https://arxiv.org/abs/2007.00243

News (2021-06-26): Check out our latest work BiX-NAS that is to appear in MICCAI2021! BiX-Net Achieves on par performances to BiO-Net but with 34.9 times fewer parameters and 4.0 times fewer MACs!

BiO-Net

Dependencies

  • Python >= 3.6
  • PIL >= 7.0.0
  • matplotlib >= 3.3.1

Keras version

  • tensorflow-gpu >= 1.14.0
  • keras >= 2.1.5

Pytorch version

  • tqdm >= 4.54.1
  • imgaug >= 0.4.0
  • torch >= 1.5.0
  • torchvision >= 0.6.0

NOTE1: Both versions have been sucessfully tested on a GeForce RTX 2080 GPU with CUDA=10.1 and driver=435.21.
NOTE2: You don't need all dependencies to run on a sole backend.

Data

Please refer to the official website (or project repo) for license and terms of usage. We preprocessed the data for adaptating our data loaders. Please find the links below.

MoNuSeg

TNBC

Usage

Use --backend to switch between backends [keras, pytorch], default=keras.

Train

python3 main.py --epochs 300 --iter 3 --integrate --train_data PATH_TO_TRAIN_DATA_ROOT \
		  --valid_data PATH_TO_VALID_DATA_ROOT --exp 1 --backend YOUR_BACKEND

Evaluate

Save metrics and segmentations:

python3 main.py --evaluate_only --save_result --valid_data PATH_TO_VALID_DATA_ROOT --exp 1  --backend YOUR_BACKEND

Print metrics only:

python3 main.py --evaluate_only --valid_data PATH_TO_VALID_DATA_ROOT --exp 1  --backend YOUR_BACKEND

NOTE: --iter,--integrate,--multiplier need to be specidied on the above commands.

or

python3 main.py --evaluate_only --valid_data PATH_TO_VALID_DATA_ROOT --model_path PATH_TO_TRAINED_MODEL  --backend YOUR_BACKEND

Citation

If you find this repo useful in your work or research, please cite:

@inproceedings{xiang2020bio,
  title={BiO-Net: Learning Recurrent Bi-directional Connections for Encoder-Decoder Architecture},
  author={Xiang, Tiange and Zhang, Chaoyi and Liu, Dongnan and Song, Yang and Huang, Heng and Cai, Weidong},
  booktitle={International Conference on Medical Image Computing and Computer-Assisted Intervention},
  pages={74--84},
  year={2020},
  organization={Springer}
}

About

[MICCAI 2020] BiO-Net: Learning Recurrent Bi-directional Connections for Encoder-Decoder Architecture

Topics

Resources

License

Stars

Watchers

Forks

Languages