A Three-Player GAN: Generating Hard Samples To Improve Classification Networks
Computer Vision and Pattern Recognition
2019-03-11 v1
Abstract
We propose a Three-Player Generative Adversarial Network to improve classification networks. In addition to the game played between the discriminator and generator, a competition is introduced between the generator and the classifier. The generator's objective is to synthesize samples that are both realistic and hard to label for the classifier. Even though we make no assumptions on the type of augmentations to learn, we find that the model is able to synthesize realistically looking examples that are hard for the classification model. Furthermore, the classifier becomes more robust when trained on these difficult samples. The method is evaluated on a public dataset for traffic sign recognition.
Cite
@article{arxiv.1903.03496,
title = {A Three-Player GAN: Generating Hard Samples To Improve Classification Networks},
author = {Simon Vandenhende and Bert De Brabandere and Davy Neven and Luc Van Gool},
journal= {arXiv preprint arXiv:1903.03496},
year = {2019}
}
Comments
Accepted for oral presentation at MVA2019