Research map / 3D vision / 3D generation
PercHead: Perceptual Head Model for Single-Image 3D Head Reconstruction & Editing
Antonio Oroz, Matthias Nießner, Tobias Kirschstein
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 |
|---|---|---|---|
| add multi-view data to 2D-only training (ablation) | PercHead · Ava-256 (extreme views) | PSNR: 6.5 -> 15.9 | Improved |
| add 2D FFHQ data to multi-view-only training (ablation) | PercHead · Ava-256 (extreme views) | ArcFace distance: 0.308 -> 0.282 | Improved |
| add 2D FFHQ data to multi-view-only training (ablation) | PercHead · Ava-256 (extreme views) | LPIPS: 0.299 -> 0.291 | Improved |
| replace LPIPS+L1 with DINOv2+SAM2.1 loss (ablation) | PercHead · Ava-256 (extreme views) | DreamSim: 0.115 -> 0.106 | Improved |
| replace LPIPS+L1 with DINOv2+SAM2.1 loss (ablation) | PercHead · Ava-256 (extreme views) | ArcFace distance: 0.300 -> 0.282 | Improved |
| replace LPIPS+L1 with DINOv2+SAM2.1 loss (ablation) | PercHead · Ava-256 (extreme views) | LPIPS: 0.285 -> 0.291 | Hurt |
| add SAM2.1 loss to DINOv2 loss (ablation) | PercHead · Ava-256 (extreme views) | LPIPS: 0.383 -> 0.291 | Improved |
| add SAM2.1 loss to DINOv2 loss (ablation) | PercHead · Ava-256 (extreme views) | ArcFace distance: 0.340 -> 0.282 | Improved |
| add DINOv2 loss to SAM2.1 loss (ablation) | PercHead · Ava-256 (extreme views) | ArcFace distance: 0.311 -> 0.282 | Improved |
| add DINOv2 loss to SAM2.1 loss (ablation) | PercHead · Ava-256 (extreme views) | LPIPS: 0.309 -> 0.291 | Improved |
| add SHP-CNN sharpness network (ablation) | PercHead · Ava-256 (extreme views) | LPIPS: 0.301 -> 0.291 | Improved |
| add SHP-CNN sharpness network (ablation) | PercHead · Ava-256 (extreme views) | DreamSim: 0.116 -> 0.106 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 on most shared results.
- PanoHead — PanoHead: Geometry-Aware 3D Full-Head Synthesis in 360deg (CVPR 2023) · better on 5 of 5 shared results
- SphereHead — SphereHead: Stable 3D Full-head Synthesis with Spherical Tri-plane Representation (ECCV 2024) · better on 4 of 4 shared results
- FaceLift — FaceLift: Learning Generalizable Single Image 3D Face Reconstruction from Synthetic Heads (ICCV 2025) · 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
- PanoHead: Geometry-Aware 3D Full-Head Synthesis in 360deg (CVPR 2023 · similarity 0.87)
- Avat3r: Large Animatable Gaussian Reconstruction Model for High-fidelity 3D Head Avatars (ICCV 2025 · similarity 0.86)
- GeneAvatar: Generic Expression-Aware Volumetric Head Avatar Editing from a Single Image (CVPR 2024 · similarity 0.85)
- MonoNPHM: Dynamic Head Reconstruction from Monocular Videos (CVPR 2024 · similarity 0.85)
- FATE: Full-head Gaussian Avatar with Textural Editing from Monocular Video (CVPR 2025 · similarity 0.85)
- Synthetic Prior for Few-Shot Drivable Head Avatar Inversion (CVPR 2025 · similarity 0.85)
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.
