Research map / 3D vision / Face and shape
Revisiting Avatar-As-Image: High-Fidelity Registration is All You Need
Margaret Kostyrko, Yuxuan Xue, Garvita Tiwari, Gerard Pons-Moll
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 |
|---|---|---|---|
| enforce body-inside-clothing with signed winding numbers instead of unsigned distance (ablation) | AvaImg · 4D-Dress/BuFF/CAPE | penetration rate (%): 40.7 -> 19.8 | Improved |
| signed winding numbers instead of unsigned distance (ablation) | AvaImg · 4D-Dress/BuFF/CAPE | penetration depth (mm): 10.06 -> 3.62 | Improved |
| signed winding numbers instead of unsigned distance (ablation) | AvaImg · BuFF shape estimation | Chamfer (mm): 11.95 -> 7.72 | Improved |
| signed constraint on scan proximity (ablation) | AvaImg · 4D-Dress/BuFF/CAPE | scan proximity (mm): 9.74 vs 9.47 | No clear effect |
| add Loop subdivision with two-pass coarse-to-fine displacement (squared coupling) (ablation) | AvaImg · surface registration | Chamfer (mm): 3.15 -> 2.75 | Improved |
| use fourth-power rather than squared edge coupling (ablation) | AvaImg · surface registration | Chamfer (mm): 2.75 -> 2.62 | Improved |
| decimate scan from 370k to 40k faces (ablation) | AvaImg winding computation · runtime | seconds: 2,400 -> 240 | Improved |
| add winding band reduction (ablation) | AvaImg body fitting · runtime | seconds: 1,700 -> 240 | Improved |
| AvaImg vs dataset ground-truth fits | AvaImg · 5 datasets | penetration rate (%): 28.2 -> 15.0 | Improved |
| AvaImg vs dataset ground-truth fits | AvaImg · 5 datasets overall | scan proximity (mm): 11.8 -> 11.4 | Mixed |
| pass UV maps through frozen FLUX VAE | AvaImg UV maps · VAE roundtrip | Chamfer vs scan (mm): 3.15 -> 3.91 | Hurt |
Closest papers by meaning
- SCANimate: Weakly Supervised Learning of Skinned Clothed Avatar Networks (CVPR 2021 · similarity 0.90)
- FRESA: Feedforward Reconstruction of Personalized Skinned Avatars from Few Images (CVPR 2025 · similarity 0.88)
- CtrlAvatar: Controllable Avatars Generation via Disentangled Invertible Networks (AAAI 2025 · similarity 0.88)
- Human-3Diffusion: Realistic Avatar Creation via Explicit 3D Consistent Diffusion Models (NeurIPS 2024 · similarity 0.88)
- DAGSM: Disentangled Avatar Generation with GS-enhanced Mesh (CVPR 2025 · similarity 0.87)
- FreeCloth: Free-form Generation Enhances Challenging Clothed Human Modeling (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.
