Abstract: Data-Efficient Generative Adversarial Nets (DE-GANs) have become more and more popular in recent years. Existing methods apply data augmentation, noise injection and pre-trained models to ...