Research map / 3D vision / Face and shape
RegHead: Non-Humanoid Head Blendshapes via Feed-Forward Registration
Jiahao Luo, Hao Zhang, Jianqi Chen, Yijie He, Jiaxu Zou, Michael Vasilkovsky, Sergei Korolev, Sergey Tulyakov, Chaoyang Wang, Peter Wonka, James Davis, Jian Wang
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 |
|---|---|---|---|
| resample stochastic anchors each iteration instead of fixed anchors (ablation) | RegHead · Expression registration (PSNR) | PSNR: 23.323 -> 23.842 | Improved |
| use 3000 anchors instead of 300 (ablation) | RegHead · Expression registration (PSNR) | PSNR: 22.226 -> 23.842 | Improved |
| add unprojected segmentation channel to voxel tokens (ablation) | RegHead · Expression registration (PSNR) | PSNR: 23.391 -> 23.842 | Improved |
| add local matching stage (ablation) | RegHead · Expression registration (PSNR) | PSNR: 22.138 -> 23.842 | Improved |
| add global matching stage (ablation) | RegHead · Expression registration (PSNR) | PSNR: 23.660 -> 23.842 | Improved |
| use structured multi-radius stencil instead of kNN neighbors (ablation) | RegHead · Expression registration (PSNR) | PSNR: 23.055 -> 23.842 | Improved |
| use more anchors at inference than training (10000, 30000) (ablation) | RegHead · Expression registration (PSNR) | PSNR: 23.842 -> 23.704 / 23.427 | Hurt |
| change random seed of anchor sampling at test time (ablation) | RegHead · Expression registration (PSNR) | PSNR: 23.824 - 23.854 | No clear effect |
| replace ActionMesh with RegHead | RegHead · Expression registration | PSNR: 22.2731 -> 23.8421 | Improved |
| replace optimization-based T2Bs with RegHead | RegHead · Expression registration | runtime: ~1800 s -> ~5 s | Improved |
| RegHead vs T2Bs on frown amplitude | RegHead · Local amplitude (frown) | % amplitude preserved: 90.9% (T2Bs) vs 85.1% | Hurt |
Closest papers by meaning
- DiffusionAvatars: Deferred Diffusion for High-fidelity 3D Head Avatars (CVPR 2024 · similarity 0.88)
- Arc2Avatar: Generating Expressive 3D Avatars from a Single Image via ID Guidance (CVPR 2025 · similarity 0.88)
- RGBAvatar: Reduced Gaussian Blendshapes for Online Modeling of Head Avatars (CVPR 2025 · similarity 0.87)
- Efficient 3D Implicit Head Avatar with Mesh-anchored Hash Table Blendshapes (CVPR 2024 · similarity 0.87)
- High-Fidelity 3D Head Avatars Reconstruction through Spatially-Varying Expression Conditioned Neural Radiance Field (AAAI 2024 · similarity 0.87)
- HRAvatar: High-Quality and Relightable Gaussian Head Avatar (CVPR 2025 · similarity 0.86)
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.
