Generative adversarial networks (GANs) are one of the most powerful generative models, but always require a large and balanced dataset to train. Traditional GANs are not applicable to generate minority-class images in a highly imbalanced dataset. Balancing GAN (BAGAN) is proposed to mitigate this problem, but it is unstable when images in different classes look similar, e.g., flowers and cells. In this work, we propose a supervised autoencoder with an intermediate embedding model to disperse the labeled latent vectors. With the enhanced autoencoder initialization, we also build an architecture of BAGAN with gradient penalty (BAGAN-GP). Our proposed model overcomes the unstable issue in original BAGAN and converges faster to high-quality generations. Our model achieves high performance on the imbalanced scale-down version of MNIST Fashion, CIFAR-10, and one small-scale medical image dataset. https://github.com/GH920/improved-bagan-gp.

生成对抗网络 (GAN) 是最强大的生成模型之一，但始终需要大型且平衡的数据集进行训练。传统的 GAN 不适用于在高度不平衡的数据集中生成少数类图像。平衡 GAN (BAGAN) 被提出来缓解这个问题，但是当不同类别的图像看起来相似时它是不稳定的，例如，花和细胞。在这项工作中，我们提出了一种带有中间嵌入模型的监督自动编码器来分散标记的潜在向量。通过增强的自动编码器初始化，我们还构建了具有梯度惩罚的 BAGAN 架构（BAGAN-GP）。我们提出的模型克服了原始 BAGAN 中的不稳定问题，并更快地收敛到高质量的生成。我们的模型在 MNIST Fashion、CIFAR-10 的不平衡缩小版本上实现了高性能，https://github.com/GH920/improved-bagan-gp。