Robust Conditional GAN from Uncertainty-Aware Pairwise Comparisons
Abstract
Conditional generative adversarial networks have shown exceptional generation performance over the past few years. However, they require large numbers of annotations. To address this problem, we propose a novel generative adversarial network utilizing weak supervision in the form of pairwise comparisons (PC-GAN) for image attribute editing. In the light of Bayesian uncertainty estimation and noise-tolerant adversarial training, PC-GAN can estimate attribute rating efficiently and demonstrate robust performance in noise resistance. Through extensive experiments, we show both qualitatively and quantitatively that PC-GAN performs comparably with fully-supervised methods and outperforms unsupervised baselines. Code and Supplementary can be found on the project website*.
Cite
Text
Han et al. "Robust Conditional GAN from Uncertainty-Aware Pairwise Comparisons." AAAI Conference on Artificial Intelligence, 2020. doi:10.1609/AAAI.V34I07.6723Markdown
[Han et al. "Robust Conditional GAN from Uncertainty-Aware Pairwise Comparisons." AAAI Conference on Artificial Intelligence, 2020.](https://mlanthology.org/aaai/2020/han2020aaai-robust-a/) doi:10.1609/AAAI.V34I07.6723BibTeX
@inproceedings{han2020aaai-robust-a,
title = {{Robust Conditional GAN from Uncertainty-Aware Pairwise Comparisons}},
author = {Han, Ligong and Gao, Ruijiang and Kim, Mun and Tao, Xin and Liu, Bo and Metaxas, Dimitris N.},
booktitle = {AAAI Conference on Artificial Intelligence},
year = {2020},
pages = {10909-10916},
doi = {10.1609/AAAI.V34I07.6723},
url = {https://mlanthology.org/aaai/2020/han2020aaai-robust-a/}
}