AITTI: Learning Adaptive Inclusive Token for Text-to-Image Generation

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Hou, Xinyu, Li, Xiaoming, Loy, Chen Change
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915730302173184
author Hou, Xinyu
Li, Xiaoming
Loy, Chen Change
author_facet Hou, Xinyu
Li, Xiaoming
Loy, Chen Change
contents Despite the high-quality results of text-to-image generation, stereotypical biases have been spotted in their generated contents, compromising the fairness of generative models. In this work, we propose to learn adaptive inclusive tokens to shift the attribute distribution of the final generative outputs. Unlike existing de-biasing approaches, our method requires neither explicit attribute specification nor prior knowledge of the bias distribution. Specifically, the core of our method is a lightweight adaptive mapping network, which can customize the inclusive tokens for the concepts to be de-biased, making the tokens generalizable to unseen concepts regardless of their original bias distributions. This is achieved by tuning the adaptive mapping network with a handful of balanced and inclusive samples using an anchor loss. Experimental results demonstrate that our method outperforms previous bias mitigation methods without attribute specification while preserving the alignment between generative results and text descriptions. Moreover, our method achieves comparable performance to models that require specific attributes or editing directions for generation. Extensive experiments showcase the effectiveness of our adaptive inclusive tokens in mitigating stereotypical bias in text-to-image generation. The code will be available at https://github.com/itsmag11/AITTI.
format Preprint
id arxiv_https___arxiv_org_abs_2406_12805
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle AITTI: Learning Adaptive Inclusive Token for Text-to-Image Generation
Hou, Xinyu
Li, Xiaoming
Loy, Chen Change
Computer Vision and Pattern Recognition
Despite the high-quality results of text-to-image generation, stereotypical biases have been spotted in their generated contents, compromising the fairness of generative models. In this work, we propose to learn adaptive inclusive tokens to shift the attribute distribution of the final generative outputs. Unlike existing de-biasing approaches, our method requires neither explicit attribute specification nor prior knowledge of the bias distribution. Specifically, the core of our method is a lightweight adaptive mapping network, which can customize the inclusive tokens for the concepts to be de-biased, making the tokens generalizable to unseen concepts regardless of their original bias distributions. This is achieved by tuning the adaptive mapping network with a handful of balanced and inclusive samples using an anchor loss. Experimental results demonstrate that our method outperforms previous bias mitigation methods without attribute specification while preserving the alignment between generative results and text descriptions. Moreover, our method achieves comparable performance to models that require specific attributes or editing directions for generation. Extensive experiments showcase the effectiveness of our adaptive inclusive tokens in mitigating stereotypical bias in text-to-image generation. The code will be available at https://github.com/itsmag11/AITTI.
title AITTI: Learning Adaptive Inclusive Token for Text-to-Image Generation
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2406.12805