Research map / 3D vision / Face and shape
DiffusionRig: Learning Personalized Priors for Facial Appearance Editing
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Fine-tune generic face priors on photos of the target person (ablation) | DiffusionRig · Personalized facial appearance editing | Identity preservation | Helped |
| Use pixel-aligned physical buffers instead of DECA parameter vectors (ablation) | DiffusionRig · DECA re-inference | Expression RMSE: 8.8 -> 2.8 | Helped |
| Use pixel-aligned physical buffers instead of non-spatial encoded features (ablation) | DiffusionRig · DECA re-inference | Pose RMSE: 21.6 -> 4.2 | Helped |
| Use 20 personal training images rather than fewer images (ablation) | DiffusionRig Stage 2 · Personalized relighting, expression, and pose editing | Visual editing quality | Helped |
| Include albedo alongside normals and Lambertian rendering (ablation) | DiffusionRig · Facial appearance editing | Lighting accuracy | Helped |
| Use DiffusionRig rather than MyStyle for pose editing | Facial appearance editing · Obama and Swift, pose editing | Automatic face re-identification: 97.9% -> 99.3% | Helped |
| Use DiffusionRig rather than MyStyle | Facial appearance editing · Swift, expression editing | User study real-and-same-person judgment: 64.5% -> 82.4% | Helped |
| Use DiffusionRig rather than MyStyle | Facial appearance editing · Swift, pose editing | User study real-and-same-person judgment: 62.5% -> 80.2% | Helped |
| Use DiffusionRig rather than MyStyle | Facial appearance editing · Obama, expression editing | User study real-and-same-person judgment: 79.4% -> 87.2% | Helped |
| Use DiffusionRig rather than MyStyle | Facial appearance editing · Obama, pose editing | User study real-and-same-person judgment: 78.0% -> 86.5% | Helped |
| Use pixel-aligned physical buffers instead of parameter vectors (ablation) | DiffusionRig · DECA re-inference | Shape RMSE: 4.3 -> 10.7 | Hurt |
| Use pixel-aligned physical buffers instead of parameter vectors (ablation) | DiffusionRig · DECA re-inference | Light RMSE: 11.2 -> 15.5 | Hurt |
Closest papers by meaning
- Cross Initialization for Face Personalization of Text-to-Image Models (CVPR 2024 · similarity 0.89)
- PFStorer: Personalized Face Restoration and Super-Resolution (CVPR 2024 · similarity 0.88)
- When StyleGAN Meets Stable Diffusion: a W+ Adapter for Personalized Image Generation (CVPR 2024 · similarity 0.88)
- DiffPortrait3D: Controllable Diffusion for Zero-Shot Portrait View Synthesis (CVPR 2024 · similarity 0.87)
- DreamSalon: A Staged Diffusion Framework for Preserving Identity-Context in Editable Face Generation (CVPR 2024 · similarity 0.87)
- OSDFace: One-Step Diffusion Model for Face Restoration (CVPR 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.