Research map / Image generation / Editing and style transfer
IDperturb: Enhancing Variation in Synthetic Face Generation via Angular Perturbations
Fadi Boutros, Eduarda Caldeira, Tahar Chettaoui, Naser Damer
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| perturb identity embeddings within a cone (lb 0.6) instead of a fixed embedding (ablation) | IDiff-Face (C-WF) · Face verification avg of 5 benchmarks | average accuracy (%): 91.25 -> 93.62 | Improved |
| perturb identity embeddings within a cone (lb 0.5) instead of a fixed embedding (ablation) | IDiff-Face (FFHQ) · Face verification avg of 5 benchmarks | average accuracy (%): 86.58 -> 88.79 | Improved |
| lower the cone bound lb from 0.6 to 0.4 (ablation) | IDPERTURB (C-WF) · Face verification avg of 5 benchmarks | average accuracy (%): 93.62 -> 93.36 | Hurt |
| lower lb (wider cone) on intra-class diversity (ablation) | IDPERTURB (C-WF) · Intra-class diversity | D-intra (LPIPS): 0.325 -> 0.538 (lb 0.9 -> 0.4) | Improved |
| lower lb (wider cone) on identity consistency (ablation) | IDPERTURB (C-WF) · Intra-class consistency | C-intra: 0.998 -> 0.790 (lb 0.9 -> 0.4) | Hurt |
| increase CFG scale from 1 to 5 (ablation) | IDPERTURB (C-WF, lb 0.6) · Face verification avg of 5 benchmarks | average accuracy (%): 93.62 -> 92.42 | Hurt |
| use CFG scale 1-2 instead of 0 (no guidance) (ablation) | IDPERTURB (C-WF, lb 0.6) · Face verification avg of 5 benchmarks | average accuracy (%): 92.97 -> 93.62 | Improved |
| add IDPERTURB to Arc2Face generation | Arc2Face · Face verification avg of 5 benchmarks | average accuracy (%): 89.71 -> 91.32 | Improved |
| add IDPERTURB to Arc2Face generation | Arc2Face · LFW | accuracy (%): 98.90 -> 98.93 | No clear effect |
| scale IDPERTURB data from 0.5M to 1.0M images | IDPERTURB (C-WF) · Face verification avg of 5 benchmarks | average accuracy (%): 93.62 -> 94.48 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- UIFace — UIFace: Unleashing Inherent Model Capabilities to Enhance Intra-Class Diversity in Synthetic Face Recognition (ICLR 2025) · better on 2 of 2 shared results
- Arc2Face — Arc2Face: A Foundation Model for ID-Consistent Human Faces (ECCV 2024) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- $\text{ID}^3$: Identity-Preserving-yet-Diversified Diffusion Models for Synthetic Face Recognition (NeurIPS 2024 · similarity 0.92)
- UIFace: Unleashing Inherent Model Capabilities to Enhance Intra-Class Diversity in Synthetic Face Recognition (ICLR 2025 · similarity 0.90)
- HyperFace: Generating Synthetic Face Recognition Datasets by Exploring Face Embedding Hypersphere (ICLR 2025 · similarity 0.90)
- Data Synthesis with Diverse Styles for Face Recognition via 3DMM-Guided Diffusion (CVPR 2025 · similarity 0.88)
- Synthetic Face Datasets Generation via Latent Space Exploration from Brownian Identity Diffusion (ICML 2025 · similarity 0.88)
- Vec2Face: Scaling Face Dataset Generation with Loosely Constrained Vectors (ICLR 2025 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
