Beyond Inference Intervention: Identity-Decoupled Diffusion for Face Anonymization

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Yang, Haoxin, Lin, Yihong, Kang, Jingdan, Xu, Xuemiao, Li, Yue, Xu, Cheng, He, Shengfeng
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866918175460818944
author Yang, Haoxin
Lin, Yihong
Kang, Jingdan
Xu, Xuemiao
Li, Yue
Xu, Cheng
He, Shengfeng
author_facet Yang, Haoxin
Lin, Yihong
Kang, Jingdan
Xu, Xuemiao
Li, Yue
Xu, Cheng
He, Shengfeng
contents Face anonymization aims to conceal identity information while preserving non-identity attributes. Mainstream diffusion models rely on inference-time interventions such as negative guidance or energy-based optimization, which are applied post-training to suppress identity features. These interventions often introduce distribution shifts and entangle identity with non-identity attributes, degrading visual fidelity and data utility. To address this, we propose \textbf{ID\textsuperscript{2}Face}, a training-centric anonymization framework that removes the need for inference-time optimization. The rationale of our method is to learn a structured latent space where identity and non-identity information are explicitly disentangled, enabling direct and controllable anonymization at inference. To this end, we design a conditional diffusion model with an identity-masked learning scheme. An Identity-Decoupled Latent Recomposer uses an Identity Variational Autoencoder to model identity features, while non-identity attributes are extracted from same-identity pairs and aligned through bidirectional latent alignment. An Identity-Guided Latent Harmonizer then fuses these representations via soft-gating conditioned on noisy feature prediction. The model is trained with a recomposition-based reconstruction loss to enforce disentanglement. At inference, anonymization is achieved by sampling a random identity vector from the learned identity space. To further suppress identity leakage, we introduce an Orthogonal Identity Mapping strategy that enforces orthogonality between sampled and source identity vectors. Experiments demonstrate that ID\textsuperscript{2}Face outperforms existing methods in visual quality, identity suppression, and utility preservation.
format Preprint
id arxiv_https___arxiv_org_abs_2510_24213
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Beyond Inference Intervention: Identity-Decoupled Diffusion for Face Anonymization
Yang, Haoxin
Lin, Yihong
Kang, Jingdan
Xu, Xuemiao
Li, Yue
Xu, Cheng
He, Shengfeng
Computer Vision and Pattern Recognition
Face anonymization aims to conceal identity information while preserving non-identity attributes. Mainstream diffusion models rely on inference-time interventions such as negative guidance or energy-based optimization, which are applied post-training to suppress identity features. These interventions often introduce distribution shifts and entangle identity with non-identity attributes, degrading visual fidelity and data utility. To address this, we propose \textbf{ID\textsuperscript{2}Face}, a training-centric anonymization framework that removes the need for inference-time optimization. The rationale of our method is to learn a structured latent space where identity and non-identity information are explicitly disentangled, enabling direct and controllable anonymization at inference. To this end, we design a conditional diffusion model with an identity-masked learning scheme. An Identity-Decoupled Latent Recomposer uses an Identity Variational Autoencoder to model identity features, while non-identity attributes are extracted from same-identity pairs and aligned through bidirectional latent alignment. An Identity-Guided Latent Harmonizer then fuses these representations via soft-gating conditioned on noisy feature prediction. The model is trained with a recomposition-based reconstruction loss to enforce disentanglement. At inference, anonymization is achieved by sampling a random identity vector from the learned identity space. To further suppress identity leakage, we introduce an Orthogonal Identity Mapping strategy that enforces orthogonality between sampled and source identity vectors. Experiments demonstrate that ID\textsuperscript{2}Face outperforms existing methods in visual quality, identity suppression, and utility preservation.
title Beyond Inference Intervention: Identity-Decoupled Diffusion for Face Anonymization
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2510.24213