Research map / Image generation / Editing and style transfer
Bridging Privacy and Provenance: Traceable Virtual Identity Generation
Xianhan Zeng, Xiaoxiao Hu, Sheng Li, Zhenxing Qian, Xinpeng Zhang
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 |
|---|---|---|---|
| use vMF or HS-AE virtual identity sampler instead of ID-Mixer | Ours · CelebA-HQ test, consistency (ArcFace) | EER: 0.021 -> 0.002 | Improved |
| generate virtual identities with proposed samplers instead of existing anonymization baselines | IVFG · CelebA-HQ test (ArcFace) | EER: 0.161 -> 0.002 | Improved |
| use proposed vMF/HS-AE samplers for diversity | IVFG · CelebA-HQ test (ArcFace) | Div: 0.913 -> 0.999 | Improved |
| add the VID sampler instead of random normal sampling of identity embedding (ablation) | Ours · qualitative ablation (Figure 5) | face realism | Improved |
| add EMOCA normal-map conditioning (ablation) | Ours · qualitative ablation (Figure 5) | head pose and expression preservation | Improved |
| add SMIRK expression adapter (ablation) | Ours · qualitative ablation (Figure 5) | expression fidelity | Improved |
| add latent watermark embedding (ablation) | Ours · CelebA-HQ and FFHQ | visual quality: PSNR 39.74, SSIM 0.974, LPIPS 0.019 | No clear effect |
| generate full virtual identity rather than edit local regions | Ours · CelebA-HQ | FID: FID 50.60 vs G2Face 7.04 | Hurt |
| use 3D geometric and expression conditioning | Ours vs G2Face · FFHQ head pose and emotion | Emotion cosine similarity: 0.793 -> 0.863 | Improved |
| compare geometry metrics (B-box and landmark distance) | Ours · FFHQ and CelebA-HQ | B-box Dis and Ldm Dis | No clear effect |
Closest papers by meaning
- Non-Colliding Biometric Identities for Digital Entities: Geometry, Capacity, and Million-Scale Virtual Identity Provisioning (NeurIPS 2026 · similarity 0.88)
- ID-to-3D: Expressive ID-guided 3D Heads via Score Distillation Sampling (NeurIPS 2024 · similarity 0.86)
- FaithfulFaces: Pose-Faithful Facial Identity Preservation for Text-to-Video Generation (NeurIPS 2026 · similarity 0.85)
- Data Synthesis with Diverse Styles for Face Recognition via 3DMM-Guided Diffusion (CVPR 2025 · similarity 0.84)
- $\text{ID}^3$: Identity-Preserving-yet-Diversified Diffusion Models for Synthetic Face Recognition (NeurIPS 2024 · similarity 0.84)
- Disguise without Disruption: Utility-Preserving Face De-identification (AAAI 2024 · similarity 0.84)
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.
