Partial Wasserstein Adversarial Network for Non-rigid Point Set Registration

This is the official implementation of Sec. B.3 in Partial Distribution Matching via Partial Wasserstein Adversarial Networks, where PWAN is used as a drop-in replacement of WGAN for robust image generation.

To make a WGAN model more robust to outliers (make it a PWAN model), you only need to

1. Use weight > 1 for the data sample.
2. Make the output of discriminator negative, e.g., output=-torch.abs(output)

Training on Cifar10 dataset with a few mnist outliers:

Training data	WGAN	PWAN

Usage

Plase see a.txt for usage.

Reference

@misc{wang2024partialdistributionmatchingpartial,
      title={Partial Distribution Matching via Partial Wasserstein Adversarial Networks}, 
      author={Zi-Ming Wang and Nan Xue and Ling Lei and Rebecka Jörnsten and Gui-Song Xia},
      year={2024},
      eprint={2409.10499},
      url={https://arxiv.org/abs/2409.10499}, 
}

@inproceedings{wang2022partial,
    title={Partial Wasserstein Adversarial Network for Non-rigid Point Set Registration},
    author={Zi-Ming Wang and Nan Xue and Ling Lei and Gui-Song Xia},
    booktitle={International Conference on Learning Representations (ICLR)},
    year={2022}
}

For any question, please contact me (wzm2256@gmail.com).

LICENSE

The code is available under a MIT license.