Implicit Data Augmentation Using Feature Interpolation for Low-Shot Image Generation

Training of generative models especially Generative Adversarial Networks can easily diverge in low-data setting. To mitigate this issue, we propose a novel implicit data augmentation approach which facilitates stable training and synthesize high-quality samples without need of label information. Specifically, we view the discriminator as a metric embedding of the real data manifold, which offers proper distances between real data points. We then utilize information in the feature space to develop a data-driven augmentation method. Experiments on few-shot generation tasks show the proposed method significantly improve results from strong baselines, and allows generating high-quality images with around 100 training samples.