Can We Find Neurons that Cause Unrealistic Images in Deep Generative Networks?

TL;DR

This paper investigates the role of neurons in GANs related to generating unrealistic images, revealing that rarely activated neurons contribute to artifacts, and proposes a low-cost correction method called 'Sequential Ablation'.

Contribution

It empirically links rarely activated neurons to artifacts in GAN-generated images and introduces a novel correction technique to improve image quality.

Findings

Rarely activated neurons are associated with artifacts.

The proposed 'Sequential Ablation' method effectively repairs defective images.

The correction method requires low computational resources.

Abstract

Even though Generative Adversarial Networks (GANs) have shown a remarkable ability to generate high-quality images, GANs do not always guarantee the generation of photorealistic images. Occasionally, they generate images that have defective or unnatural objects, which are referred to as 'artifacts'. Research to investigate why these artifacts emerge and how they can be detected and removed has yet to be sufficiently carried out. To analyze this, we first hypothesize that rarely activated neurons and frequently activated neurons have different purposes and responsibilities for the progress of generating images. In this study, by analyzing the statistics and the roles for those neurons, we empirically show that rarely activated neurons are related to the failure results of making diverse objects and inducing artifacts. In addition, we suggest a correction method, called 'Sequential Ablation', to repair the defective part of the generated images without high computational cost and manual efforts.