Research map / 3D vision / Face and shape
ImHead: A Large-scale Implicit Morphable Model for Localized Head Modeling
Rolandos Alexandros Potamias, Stathis Galanakis, Jiankang Deng, Athanasios Papaioannou, Stefanos Zafeiriou
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 |
|---|---|---|---|
| replace per-region local latents (d=1248) by a single compact global identity latent with region-specific intermediate representation (ablation) | imHead · MimicMe identity reconstruction | Chamfer distance: 0.767 -> 0.533 | Improved |
| shrink local latents to size 8 (d=312) instead of using the compact global latent (ablation) | imHead · MimicMe identity reconstruction | Chamfer distance: 0.533 -> 0.874 | Hurt |
| add FusionNet structural blending of local features (ablation) | imHead · MimicMe identity reconstruction | Chamfer distance: 0.674 -> 0.533 | Improved |
| add per-region local canonical space (ablation) | imHead · MimicMe identity reconstruction | Chamfer distance: 0.884 -> 0.533 | Improved |
| add per-region local canonical space (ablation) | imHead · NPHM identity reconstruction | F-score @5mm: 0.723 -> 0.898 | Improved |
| use local plus global latent (d=1344) instead of compact single latent (ablation) | imHead · NPHM identity reconstruction | Chamfer distance: 0.494 vs 0.459 | Mixed |
| train on the large curated MimicMe+NPHM data instead of NPHM data only | imHead · MimicMe identity reconstruction | Chamfer distance: 0.571 -> 0.533 | Improved |
| train on curated MimicMe data instead of NPHM only | imHead · MimicMe expression reconstruction | Chamfer distance: 0.623 -> 0.592 | Improved |
| train on the curated dataset instead of NPHM only | imHead · NPHM expression reconstruction | Chamfer distance: 0.508 vs 0.513 | No clear effect |
| use backward warping for expression instead of forward deformation with root finding | imHead · expression fitting time | seconds: 138 -> 40 | Improved |
| use single compact latent instead of NPHM local-global latent | imHead · latent size | dimension: 2176 -> 256 | Improved |
| train baseline NPHM on full curated dataset | NPHM · MimicMe identity reconstruction | Chamfer distance: 0.618 -> 0.598 | 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.
- monoNPHM† — MonoNPHM: Dynamic Head Reconstruction from Monocular Videos (CVPR 2024) · better on 7 of 7 shared results
- imFace — ImFace: A Nonlinear 3D Morphable Face Model With Implicit Neural Representations (CVPR 2022) · better on 2 of 2 shared results
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
- i3DMM: Deep Implicit 3D Morphable Model of Human Heads (CVPR 2021 · similarity 0.93)
- 3D Gaussian Head Avatars with Expressive Dynamic Appearances by Compact Tensorial Representations (CVPR 2025 · similarity 0.89)
- Synthetic Prior for Few-Shot Drivable Head Avatar Inversion (CVPR 2025 · similarity 0.88)
- LeGO: Leveraging a Surface Deformation Network for Animatable Stylized Face Generation with One Example (CVPR 2024 · similarity 0.88)
- Efficient 3D Implicit Head Avatar with Mesh-anchored Hash Table Blendshapes (CVPR 2024 · similarity 0.88)
- Gaussian Eigen Models for Human Heads (CVPR 2025 · similarity 0.88)
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.
