Analyzing Quality, Bias, and Performance in Text-to-Image Generative Models

TL;DR

This paper evaluates text-to-image generative models, highlighting their high-quality outputs and inherent social biases, especially gender biases, to provide a comprehensive understanding of their capabilities and limitations.

Contribution

It offers a combined qualitative performance assessment and social bias analysis of various text-to-image models, emphasizing the presence of biases in high-capacity models.

Findings

Larger models produce higher-quality images.

Models exhibit gender and social biases.

Biases persist despite high image quality.

Abstract

Advances in generative models have led to significant interest in image synthesis, demonstrating the ability to generate high-quality images for a diverse range of text prompts. Despite this progress, most studies ignore the presence of bias. In this paper, we examine several text-to-image models not only by qualitatively assessing their performance in generating accurate images of human faces, groups, and specified numbers of objects but also by presenting a social bias analysis. As expected, models with larger capacity generate higher-quality images. However, we also document the inherent gender or social biases these models possess, offering a more complete understanding of their impact and limitations.