Personalized Image Generation via Human-in-the-loop Bayesian Optimization

TL;DR

This paper introduces MultiBO, a human-in-the-loop Bayesian optimization method that uses preferential feedback to iteratively generate images closer to a user's desired target, improving personalization in image generation.

Contribution

It presents a novel multi-choice preferential Bayesian optimization framework that effectively incorporates human feedback to refine generative images without explicit target information.

Findings

MultiBO reduces the gap between generated and target images within few feedback rounds.

Qualitative user scores and quantitative metrics demonstrate improved personalization.

Multi-choice feedback effectively guides diffusion models toward user-specific images.

Abstract

Imagine Alice has a specific image $x^\ast$ in her mind, say, the view of the street in which she grew up during her childhood. To generate that exact image, she guides a generative model with multiple rounds of prompting and arrives at an image $x^{p*}$. Although $x^{p*}$ is reasonably close to $x^\ast$, Alice finds it difficult to close that gap using language prompts. This paper aims to narrow this gap by observing that even after language has reached its limits, humans can still tell when a new image $x^+$ is closer to $x^\ast$ than $x^{p*}$. Leveraging this observation, we develop MultiBO (Multi-Choice Preferential Bayesian Optimization) that carefully generates $K$ new images as a function of $x^{p*}$, gets preferential feedback from the user, uses the feedback to guide the diffusion model, and ultimately generates a new set of $K$ images. We show that within $B$ rounds of user feedback, it is possible to arrive much closer to $x^\ast$, even though the generative model has no information about $x^\ast$. Qualitative scores from $30$ users, combined with quantitative metrics compared across $5$ baselines, show promising results, suggesting that multi-choice feedback from humans can be effectively harnessed for personalized image generation.