Stacked Generative Adversarial Networks

TL;DR

This paper introduces Stacked Generative Adversarial Networks (SGAN), a hierarchical model that improves image generation quality by leveraging discriminative features and multi-level variation decomposition.

Contribution

The paper presents a novel hierarchical GAN architecture with a representation discriminator, conditional and entropy losses, and end-to-end training, enhancing image generation quality.

Findings

SGAN generates higher quality images than traditional GANs.

Inception scores and visual Turing tests favor SGAN.

Decomposition of variations improves generative diversity.

Abstract

In this paper, we propose a novel generative model named Stacked Generative Adversarial Networks (SGAN), which is trained to invert the hierarchical representations of a bottom-up discriminative network. Our model consists of a top-down stack of GANs, each learned to generate lower-level representations conditioned on higher-level representations. A representation discriminator is introduced at each feature hierarchy to encourage the representation manifold of the generator to align with that of the bottom-up discriminative network, leveraging the powerful discriminative representations to guide the generative model. In addition, we introduce a conditional loss that encourages the use of conditional information from the layer above, and a novel entropy loss that maximizes a variational lower bound on the conditional entropy of generator outputs. We first train each stack independently, and then train the whole model end-to-end. Unlike the original GAN that uses a single noise vector to represent all the variations, our SGAN decomposes variations into multiple levels and gradually resolves uncertainties in the top-down generative process. Based on visual inspection, Inception scores and visual Turing test, we demonstrate that SGAN is able to generate images of much higher quality than GANs without stacking.