Suffering from the extreme training data imbalance between seen and unseen classes, most of existing state-of-the-art approaches fail to achieve satisfactory results for the challenging generalized zero-shot learning task. To circumvent the need for labeled examples of unseen classes, we propose a novel generative adversarial network (GAN) that synthesizes CNN features conditioned on class-level semantic information, offering a shortcut directly from a semantic descriptor of a class to a class-conditional feature distribution. Our proposed approach, pairing a Wasserstein GAN with a classification loss, is able to generate sufficiently discriminative CNN features to train softmax classifiers or any multimodal embedding method. Our experimental results demonstrate a significant boost in accuracy over the state of the art on five challenging datasets -- CUB, FLO, SUN, AWA and ImageNet -- in both the zero-shot learning and generalized zero-shot learning settings.
@article{arxiv.1712.00981,
title = {Feature Generating Networks for Zero-Shot Learning},
author = {Yongqin Xian and Tobias Lorenz and Bernt Schiele and Zeynep Akata},
journal= {arXiv preprint arXiv:1712.00981},
year = {2018}
}
Comments
2018 IEEE Conference on Computer Vision and Pattern Recognition (CVPR)