InsMix

This is the official code for "InsMix: Towards Realistic Generative Data Augmentation for Nuclei Instance Segmentation (MICCAI 2022, early accepted)"

Pipeline

Method

Requirements

torch>=1.4.0 torchvision>=0.5.0 dominate>=2.4.0 visdom>=0.1.8.8 wandb

Usage

InsMix w/o Smooth-GAN

The fuctions 'insmix' and 'background shuffle' can be found in 'data_aug.py'. The example code for dataloader is in 'dataset.py'. Note that it can be used to BRPNet and NB-Net, which utilize two types of label, i.e., the inner area and the boundary.

InsMix w/ Smooth-GAN

You may simply run the scripts as:

bash Smooth-GAN/scripts/train_nuclei.sh
bash Smooth-GAN/scripts/test_nuclei.sh

Citation

Pleae cite the paper if you use the code.

@inproceedings{lin2022insmix,
  title={{InsMix}: Towards Realistic Generative Data Augmentation for Nuclei Instance Segmentation},
  author={Lin, Yi and Wang, Zeyu and Cheng, Kwang-Ting and Chen, Hao},
  booktitle={International Conference on Medical Image Computing and Computer-Assisted Intervention},
  year={2022},
  organization={Springer}
}

TODO

• [ ] Training and testing on Kumar dataset.
• [ ] Refactor the code to make it more readable.

Acknowledgment

The code of Smooth-GAN is heavily build on pix2pix, thanks for their amazing work!