Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

IDperturb: Enhancing Variation in Synthetic Face Generation via Angular Perturbation

Fadi Boutros, Eduarda Caldeira, Tahar Chettaoui, Naser Damer

TL;DR

This work proposes IDPERTURB, a simple yet effective geometric-driven sampling strategy to enhance diversity in synthetic face generation, and demonstrates that training FR on datasets generated using IDPERTURB yields improved performance across multiple FR benchmarks, compared to existing synthetic data generation approaches.

Abstract

Synthetic data has emerged as a practical alternative to authentic face datasets for training face recognition (FR) systems, especially as privacy and legal concerns increasingly restrict the use of real biometric data. Recent advances in identity-conditional diffusion models have enabled the generation of photorealistic and identity-consistent face images. However, many of these models suffer from limited intra-class variation, an essential property for training robust and generalizable FR models. In this work, we propose IDPERTURB, a simple yet effective geometric-driven sampling strategy to enhance diversity in synthetic face generation. IDPERTURB perturbs identity embeddings within a constrained angular region of the unit hyper-sphere, producing a diverse set of embeddings without modifying the underlying generative model. Each perturbed embedding serves as a conditioning vector for a pre-trained diffusion model, enabling the synthesis of visually varied yet identity-coherent face images suitable for training generalizable FR systems. Empirical results demonstrate that training FR on datasets generated using IDPERTURB yields improved performance across multiple FR benchmarks, compared to existing synthetic data generation approaches.

IDperturb: Enhancing Variation in Synthetic Face Generation via Angular Perturbation

TL;DR

This work proposes IDPERTURB, a simple yet effective geometric-driven sampling strategy to enhance diversity in synthetic face generation, and demonstrates that training FR on datasets generated using IDPERTURB yields improved performance across multiple FR benchmarks, compared to existing synthetic data generation approaches.

Abstract

Synthetic data has emerged as a practical alternative to authentic face datasets for training face recognition (FR) systems, especially as privacy and legal concerns increasingly restrict the use of real biometric data. Recent advances in identity-conditional diffusion models have enabled the generation of photorealistic and identity-consistent face images. However, many of these models suffer from limited intra-class variation, an essential property for training robust and generalizable FR models. In this work, we propose IDPERTURB, a simple yet effective geometric-driven sampling strategy to enhance diversity in synthetic face generation. IDPERTURB perturbs identity embeddings within a constrained angular region of the unit hyper-sphere, producing a diverse set of embeddings without modifying the underlying generative model. Each perturbed embedding serves as a conditioning vector for a pre-trained diffusion model, enabling the synthesis of visually varied yet identity-coherent face images suitable for training generalizable FR systems. Empirical results demonstrate that training FR on datasets generated using IDPERTURB yields improved performance across multiple FR benchmarks, compared to existing synthetic data generation approaches.
Paper Structure (11 sections, 8 equations, 3 figures, 4 tables, 1 algorithm)

This paper contains 11 sections, 8 equations, 3 figures, 4 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related Work
  3. Methodology
  4. Latent DM Preliminaries
  5. Data Generation via Angular Sampling
  6. Experimental Setup
  7. Results
  8. Impact of Identity Perturbation
  9. Impact of CFG Strength
  10. Comparison with State-of-the-Art
  11. Conclusion

Figures (3)

  • Figure 1: Identity perturbation in the embedding space. (a) a normalized reference embedding $\mathbf{v} \in \mathbb{R}^3$. (b) random $\mathbf{n_1}, \mathbf{n_2} \sim \mathcal{N}(0, \mathbf{I})$ sampled and projected onto the hyperplane orthogonal to $\mathbf{v}$, resulting in $\mathbf{u_1}$ and $\mathbf{u_2}$ (Equation \ref{['eq:orthognal']}). (c) perturbed identity vector $\mathbf{\tilde{v}_1}$ and $\mathbf{\tilde{v}_2}$ with different angle to $\mathbf{v}$ are constructed (Equation \ref{['eq:v_construct']}). Note, this figure is not manually plotted, rather, using a Python script corresponding to Algorithm \ref{['alg:angular_sampling_multiple']}. The cone boundary, dashed circumference in (c), is defined by the lower bound $\mathbf{lb}$ is 0.66, and $\mathbf{n_1}$ and $\mathbf{n_2}$ are sampled on the fly to illustrate a real scenario. For visualization purposes, we set the number of dimensions to 3, $\mathbf{v} \in \mathbb{R}^3$.
  • Figure 2: Effect of IDperturb (via lower bound $\mathbf{lb}$) on image diversity. The first row varies only the noise seed with fixed condition $\mathbf{v}$, subsequent rows use perturbed identity embeddings $\mathbf{\tilde{v}}$ from $\mathbf{v}$ with decreasing $\mathbf{lb}$ (0.9 to 0.4). Lower $\mathbf{lb}$ values increase angular deviation, enhancing intra-class variation. Note that using low lb of e.g., 0.4, the identity of some samples could be less consistent, which is also quantified by high EER in Table \ref{['tab:ablation_lb']}.
  • Figure 3: FR average accuracies (as in Table \ref{['tab:ablation_lb']}) with respect to the intra-class consistency and diversity of baseline (C-WF) and IDPerturb of $lb \in [0.4,0.5,0.6,0.7,0.8,0.9]$.