FLUXSynID: A Framework for Identity-Controlled Synthetic Face Generation with Document and Live Images

2025-05-12

AI Summary

Existing synthetic face datasets lack fine-grained control over identity attributes and struggle to generate identity-consistent document–live photo pairs, limiting their utility in biometric recognition. To address this, we propose the first identity-attribute-driven dual-modal paired generation framework, integrating diffusion models, disentangled identity embedding representations, and multi-condition joint guidance sampling to achieve high-resolution, identity-controllable paired face synthesis. We construct a large-scale benchmark dataset comprising 14,889 synthetic identities, demonstrating significant improvements in identity distribution fidelity and cross-domain diversity over prior methods. Extensive experiments validate state-of-the-art performance on both face recognition and face manipulation attack detection tasks, confirming the framework’s effectiveness and practical applicability in biometric security.

Technology Category

• Computer Vision: Biometrics, Face, Gesture & Pose
• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: Code Generation / Program Synthesis from Natural Language

Application Category

• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Social Networks and Social Media: Generative AI / large language models and their impact on social systems
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG

Abstract

Synthetic face datasets are increasingly used to overcome the limitations of real-world biometric data, including privacy concerns, demographic imbalance, and high collection costs. However, many existing methods lack fine-grained control over identity attributes and fail to produce paired, identity-consistent images under structured capture conditions. We introduce FLUXSynID, a framework for generating high-resolution synthetic face datasets with user-defined identity attribute distributions and paired document-style and trusted live capture images. The dataset generated using the FLUXSynID framework shows improved alignment with real-world identity distributions and greater inter-set diversity compared to prior work. The FLUXSynID framework for generating custom datasets, along with a dataset of 14,889 synthetic identities, is publicly released to support biometric research, including face recognition and morphing attack detection.

Problem

• Generates identity-controlled synthetic face datasets
• Produces paired document-style and live capture images
• Improves alignment with real-world identity distributions

Innovation

• Generates identity-controlled synthetic face datasets
• Produces paired document-style and live images
• Ensures high-resolution and inter-set diversity