Geometry Score: A Method For Comparing Generative Adversarial Networks

TL;DR

This paper introduces the Geometry Score, a novel metric for evaluating GANs by comparing the geometric properties of real and generated data manifolds, applicable across diverse datasets.

Contribution

The Geometry Score provides a new, geometry-based evaluation method for GANs that is dataset-agnostic and offers both qualitative and quantitative insights.

Findings

The metric effectively distinguishes different GAN models.

It reveals mode collapse and diversity issues in generated samples.

Applied to various datasets, it offers consistent evaluation results.

Abstract

One of the biggest challenges in the research of generative adversarial networks (GANs) is assessing the quality of generated samples and detecting various levels of mode collapse. In this work, we construct a novel measure of performance of a GAN by comparing geometrical properties of the underlying data manifold and the generated one, which provides both qualitative and quantitative means for evaluation. Our algorithm can be applied to datasets of an arbitrary nature and is not limited to visual data. We test the obtained metric on various real-life models and datasets and demonstrate that our method provides new insights into properties of GANs.