Deceive D: Adaptive Pseudo Augmentation for GAN Training with Limited Data
This addresses a domain-specific challenge for researchers and practitioners in generative modeling, offering an incremental improvement over existing methods.
The paper tackles the problem of training GANs with limited data, which often leads to discriminator overfitting and poor generator convergence, by introducing Adaptive Pseudo Augmentation (APA) to improve synthesis quality in low-data regimes.
Generative adversarial networks (GANs) typically require ample data for training in order to synthesize high-fidelity images. Recent studies have shown that training GANs with limited data remains formidable due to discriminator overfitting, the underlying cause that impedes the generator's convergence. This paper introduces a novel strategy called Adaptive Pseudo Augmentation (APA) to encourage healthy competition between the generator and the discriminator. As an alternative method to existing approaches that rely on standard data augmentations or model regularization, APA alleviates overfitting by employing the generator itself to augment the real data distribution with generated images, which deceives the discriminator adaptively. Extensive experiments demonstrate the effectiveness of APA in improving synthesis quality in the low-data regime. We provide a theoretical analysis to examine the convergence and rationality of our new training strategy. APA is simple and effective. It can be added seamlessly to powerful contemporary GANs, such as StyleGAN2, with negligible computational cost.