We tackle societal bias in image-text datasets by removing spurious correlations between protected groups and image attributes. Traditional methods only target labeled attributes, ignoring biases from unlabeled ones. Using text-guided inpainting models, our approach ensures protected group independence from all attributes and mitigates inpainting biases through data filtering. Evaluations on multi-label image classification and image captioning tasks show our method effectively reduces bias without compromising performance across various models.
@article{arxiv.2407.03623,
title = {Resampled Datasets Are Not Enough: Mitigating Societal Bias Beyond Single Attributes},
author = {Yusuke Hirota and Jerone T. A. Andrews and Dora Zhao and Orestis Papakyriakopoulos and Apostolos Modas and Yuta Nakashima and Alice Xiang},
journal= {arXiv preprint arXiv:2407.03623},
year = {2024}
}