arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2006.06676cs.CVcs.LGcs.NEstat.ML
收藏

用有限数据训练生成对抗网络

Training Generative Adversarial Networks with Limited Data

NVIDIA(英伟达);Aalto University(阿尔托大学)

Tero Karras, Miika Aittala, Janne Hellsten, Samuli Laine, Jaakko Lehtinen, Timo Aila

摘要:

使用太少的数据训练生成对抗网络(GAN)通常会导致过度拟合,导致训练发散。我们提出了一种自适应的数据增强机制,可以在有限的数据范围内显着稳定训练。该方法不需要改变损失函数或网络架构,并且在从头开始训练和在另一个数据集上微调现有GAN时都适用。我们在几个数据集上证明,现在只使用几千张训练图像就可以获得良好的结果,通常将StyleGAN 2结果与数量级更少的图像相匹配。我们希望这将为GANs开辟新的应用领域。我们还发现,广泛使用的CIFAR-10实际上是一个有限的数据基准,并将记录FID从5.59提高到2.42。

英文摘要:

Training generative adversarial networks (GAN) using too little data typically leads to discriminator overfitting, causing training to diverge. We propose an adaptive discriminator augmentation mechanism that significantly stabilizes training in limited data regimes. The approach does not require changes to loss functions or network architectures, and is applicable both when training from scratch and when fine-tuning an existing GAN on another dataset. We demonstrate, on several datasets, that good results are now possible using only a few thousand training images, often matching StyleGAN2 results with an order of magnitude fewer images. We expect this to open up new application domains for GANs. We also find that the widely used CIFAR-10 is, in fact, a limited data benchmark, and improve the record FID from 5.59 to 2.42.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑