The comparison between Conditional Generative Adversarial Nets and Deep Convolutional Generative Adversarial Network, and its GUI-related application

Abstract

In recent years, Generative Adversarial Nets (GAN), Conditional Generative Adversarial Nets (CGAN), and Deep convolutional generative adversarial networks (DCGAN) have generally been well-received in Artificial Intelligence (AI) industry. This paper first briefly introduces the fundamentals of GAN, CGAN, and DCGAN. Next, we focus on comparing two improved GAN variants– CGAN and DCGAN. To be specific, we train them with certain architectural constraints on two datasets – MNIST and Animation images. We show convincing evidence that DCGAN outperforms CGAN in terms of processing image datasets to a large extent. Additionally, we make a Graphical User Interface (GUI), enabling users to choose face photos with different tags generated by DCGAN.