Search

The rapid development of text-to-image generation has brought rising ethical considerations, especially regarding gender bias. Given a text prompt as input, text-to-image models generate images according to the prompt. Pioneering models such as Stable Diffusion and DALL-E 2 have demonstrated remarkable capabilities in producing high-fidelity images from natural language prompts. However, these models often exhibit gender bias, as studied by the tendency of generating man from prompts such as "a photo of a software developer". Given the widespread application and increasing accessibility of these models, bias evaluation is crucial for regulating the development of text-to-image generation. Unlike well-established metrics for evaluating image quality or fidelity, the evaluation of bias presents challenges and lacks standard approaches. Although biases related to other factors, such as skin tone, have been explored, gender bias remains the most extensively studied. In this paper, we review recent work on gender bias evaluation in text-to-image generation, involving bias evaluation setup, bias evaluation metrics, and findings and trends. We primarily focus on the evaluation of recent popular models such as Stable Diffusion, a diffusion model operating in the latent space and using CLIP text embedding, and DALL-E 2, a diffusion model leveraging Seq2Seq architectures like BART. By analyzing recent work and discussing trends, we aim to provide insights for future work.

Several studies have raised awareness about social biases in image generative models, demonstrating their predisposition towards stereotypes and imbalances. This paper contributes to this growing body of research by introducing an evaluation protocol that analyzes the impact of gender indicators at every step of the generation process on Stable Diffusion images. Leveraging insights from prior work, we explore how gender indicators not only affect gender presentation but also the representation of objects and layouts within the generated images. Our findings include the existence of differences in the depiction of objects, such as instruments tailored for specific genders, and shifts in overall layouts. We also reveal that neutral prompts tend to produce images more aligned with masculine prompts than their feminine counterparts. We further explore where bias originates through representational disparities and how it manifests in the images via prompt-image dependencies, and provide recommendations for developers and users to mitigate potential bias in image generation.

The duality of content and style is inherent to the nature of art. For humans, these two elements are clearly different: content refers to the objects and concepts in the piece of art, and style to the way it is expressed. This duality poses an important challenge for computer vision. The visual appearance of objects and concepts is modulated by the style that may reflect the author's emotions, social trends, artistic movement, etc., and their deep comprehension undoubtfully requires to handle both. A promising step towards a general paradigm for art analysis is to disentangle content and style, whereas relying on human annotations to cull a single aspect of artworks has limitations in learning semantic concepts and the visual appearance of paintings. We thus present GOYA, a method that distills the artistic knowledge captured in a recent generative model to disentangle content and style. Experiments show that synthetically generated images sufficiently serve as a proxy of the real distribution of artworks, allowing GOYA to separately represent the two elements of art while keeping more information than existing methods.

The increasing tendency to collect large and uncurated datasets to train vision-and-language models has raised concerns about fair representations. It is known that even small but manually annotated datasets, such as MSCOCO, are affected by societal bias. This problem, far from being solved, may be getting worse with data crawled from the Internet without much control. In addition, the lack of tools to analyze societal bias in big collections of images makes addressing the problem extremely challenging. Our first contribution is to annotate part of the Google Conceptual Captions dataset, widely used for training vision-and-language models, with four demographic and two contextual attributes. Our second contribution is to conduct a comprehensive analysis of the annotations, focusing on how different demographic groups are represented. Our last contribution lies in evaluating three prevailing vision-and-language tasks: image captioning, text-image CLIP embeddings, and text-to-image generation, showing that societal bias is a persistent problem in all of them., Comment: CVPR 2023