Keras/Tensorflow implementation of coSegGAN. This work has been published in Medical Image Computing and Computer Assisted Intervention – MICCAI 2021 (https://miccai2021.org/en/).

[ arXiv / Paper ]

Python 3.6
Tensorflow: 2.0.0
Keras: 2.3.1

Recreate conda environment as follows:

conda env create -f environment.yml

Or if you are using Docker:

docker pull tazleef/tf2.0.0-cv-keras2.3.1-imgaug:latest

Links to the publicly available dataset used in this work:

• Endovis
• UCL

Due to privacy policy, we are unable to share our surgical dataset. However, we intend to release them soon once we get the approval. For now, we have included a few sample cases for reference. Alternatively, if you have your own data, format it in the same way and set the filepath and training parameters in data_path_Loader.py and train.py.

To train the model, run train.py.

This code can also be used for other segmentation tasks with domain variations.

If you use this code or the provided model, please cite the following paper:

@inproceedings{kalia2021co,
  title={Co-Generation and Segmentation for Generalized Surgical Instrument Segmentation on Unlabelled Data},
  author={Kalia, Megha and Aleef, Tajwar Abrar and Navab, Nassir and Black, Peter and Salcudean, Septimiu E},
  booktitle={International Conference on Medical Image Computing and Computer-Assisted Intervention},
  pages={403--412},
  year={2021},
  organization={Springer}
}
}

If you face any problem using this code then create an issue in this repository or contact me at [email protected]

The code is based on https://github.com/eriklindernoren/Keras-GAN/tree/master/cyclegan

MIT