Rendering-Refined Stable Diffusion for Privacy Compliant Synthetic Data

TL;DR

This paper introduces RefSD, a novel pipeline combining 3D-rendering and Stable Diffusion to generate privacy-preserving synthetic human images that retain posture and realism, improving utility for detection and classification tasks.

Contribution

RefSD is the first method to effectively balance privacy, posture preservation, and realism using prompt-based control, surpassing traditional masking, blurring, and existing generative models.

Findings

RefSD outperforms real data in detection tasks when used for training.

Combining RefSD with real data enhances classification performance.

Human perception assessments identify strengths and weaknesses of RefSD attributes.

Abstract

Growing privacy concerns and regulations like GDPR and CCPA necessitate pseudonymization techniques that protect identity in image datasets. However, retaining utility is also essential. Traditional methods like masking and blurring degrade quality and obscure critical context, especially in human-centric images. We introduce Rendering-Refined Stable Diffusion (RefSD), a pipeline that combines 3D-rendering with Stable Diffusion, enabling prompt-based control over human attributes while preserving posture. Unlike standard diffusion models that fail to retain posture or GANs that lack realism and flexible attribute control, RefSD balances posture preservation, realism, and customization. We also propose HumanGenAI, a framework for human perception and utility evaluation. Human perception assessments reveal attribute-specific strengths and weaknesses of RefSD. Our utility experiments show that models trained on RefSD pseudonymized data outperform those trained on real data in detection tasks, with further performance gains when combining RefSD with real data. For classification tasks, we consistently observe performance improvements when using RefSD data with real data, confirming the utility of our pseudonymized data.