PyTorch implementation of A Two-Step Computation of the Exact GAN Wasserstein Distance, ICML 2018.
DATASET: mnist/cifar10/lsun
DATASET_PATH: path to the root folder of a dataset
- Install dependency: CVXOPT for linear programming.
- Download your dataset, unzip it and put it in DATASET_PATH.
- run code using
python WGAN_TS.py --dataset DATASET --dataroot DATASET_PATH --BN_G
You are welcome to cite our work using:
@inproceedings{liu2018two,
title={A Two-Step Computation of the Exact GAN Wasserstein Distance},
author={Liu, Huidong and Xianfeng, GU and Samaras, Dimitris},
booktitle={International Conference on Machine Learning},
pages={3165--3174},
year={2018}
}