Balancing Act: Distribution-Guided Debiasing in Diffusion Models

TL;DR

This paper introduces Distribution Guidance, a novel method for debiasing diffusion models by enforcing generated images to follow target attribute distributions without retraining or extra data, leveraging latent features for fairer image synthesis.

Contribution

The paper proposes a distribution-guided approach using a trained attribute predictor to achieve bias reduction in diffusion models without additional data or retraining.

Findings

Significantly reduces demographic bias in generated images

Outperforms baseline methods in fairness metrics

Improves downstream classifier fairness through rebalanced training data

Abstract

Diffusion Models (DMs) have emerged as powerful generative models with unprecedented image generation capability. These models are widely used for data augmentation and creative applications. However, DMs reflect the biases present in the training datasets. This is especially concerning in the context of faces, where the DM prefers one demographic subgroup vs others (eg. female vs male). In this work, we present a method for debiasing DMs without relying on additional data or model retraining. Specifically, we propose Distribution Guidance, which enforces the generated images to follow the prescribed attribute distribution. To realize this, we build on the key insight that the latent features of denoising UNet hold rich demographic semantics, and the same can be leveraged to guide debiased generation. We train Attribute Distribution Predictor (ADP) - a small mlp that maps the latent features to the distribution of attributes. ADP is trained with pseudo labels generated from existing attribute classifiers. The proposed Distribution Guidance with ADP enables us to do fair generation. Our method reduces bias across single/multiple attributes and outperforms the baseline by a significant margin for unconditional and text-conditional diffusion models. Further, we present a downstream task of training a fair attribute classifier by rebalancing the training set with our generated data.