Covering up bias in CelebA-like datasets with Markov blankets: A post-hoc cure for attribute prior avoidance
This addresses bias issues in facial attribute modeling for generative AI applications, but it is incremental as it builds on existing datasets and methods.
The paper tackles the problem of inter-attribute bias in deep generative models trained on CelebA-like datasets, which arises from ignoring semantic facial attributes during training, and proposes a post-hoc solution using an Ising attribute prior to mitigate this bias in qualitative experiments.
Attribute prior avoidance entails subconscious or willful non-modeling of (meta)attributes that datasets are oft born with, such as the 40 semantic facial attributes associated with the CelebA and CelebA-HQ datasets. The consequences of this infirmity, we discover, are especially stark in state-of-the-art deep generative models learned on these datasets that just model the pixel-space measurements, resulting in an inter-attribute bias-laden latent space. This viscerally manifests itself when we perform face manipulation experiments based on latent vector interpolations. In this paper, we address this and propose a post-hoc solution that utilizes an Ising attribute prior learned in the attribute space and showcase its efficacy via qualitative experiments.