Research map / 3D vision / Face and shape
Pix2NPHM: Learning to Regress NPHM Reconstructions From a Single Image
Simon Giebenhain, Tobias Kirschstein, Liam Schoneveld, Davide Davoli, Zhe Chen, Matthias Niessner
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 |
|---|---|---|---|
| use Pix2NPHM instead of Pixel3DMM | Pix2NPHM · NeRSemble SVFR posed | L1 error: 1.66 -> 1.37 | Improved |
| use Pix2NPHM feed-forward only instead of SHeaP | Pix2NPHM · NeRSemble SVFR posed | L1 error: 2.08 -> 1.55 | Improved |
| use Pix2NPHM instead of TokenFace | Pix2NPHM · NoW | mean error: 0.95 (TokenFace) vs 1.01 | Hurt |
| use NPHM instead of FLAME as the 3DMM (ablation) | Pix2NPHM · NoW validation | mean error: 1.073 -> 1.016 | Improved |
| use NPHM instead of FLAME (ablation) | Pix2NPHM · AffectNet | emotion accuracy: 0.660 -> 0.711 | Improved |
| use geometric pre-trained ViT instead of DINOv2 on RGB (ablation) | Pix2NPHM · NoW validation | mean error: 1.165 -> 1.016 | Improved |
| add point-map encoder E_p alongside normal encoder (ablation) | Pix2NPHM · NoW validation | mean error: 1.053 -> 1.016 | Improved |
| use DINOv2 of estimated normals instead of RGB (ablation) | Pix2NPHM · NoW validation | mean error: 1.165 -> 1.168 | No clear effect |
| add 2D self-supervision to 3D training (ablation) | Pix2NPHM · NoW validation | mean error: 1.074 -> 1.016 | Improved |
| add 3D supervision to 2D training (ablation) | Pix2NPHM · NoW validation | mean error: 1.238 -> 1.016 | Improved |
| add test-time optimization to feed-forward (ablation) | Pix2NPHM · NeRSemble SVFR posed | L1 error: 1.55 -> 1.37 | Improved |
| add test-time optimization (ablation) | Pix2NPHM · NeRSemble SVFR neutral | L1 error: 1.57 -> 1.54 | 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.
- Pixel3DMM — Pixel3DMM: Versatile Screen-Space Priors for Single-Image 3D Face Reconstruction (ICLR 2026) · better on 3 of 3 shared results
- MonoNPHM — MonoNPHM: Dynamic Head Reconstruction from Monocular Videos (CVPR 2024) · 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
- MonoNPHM: Dynamic Head Reconstruction from Monocular Videos (CVPR 2024 · similarity 0.88)
- MVF-Net: Multi-View 3D Face Morphable Model Regression (CVPR 2019 · similarity 0.88)
- Regressing Robust and Discriminative 3D Morphable Models With a Very Deep Neural Network (CVPR 2017 · similarity 0.88)
- Large Pose 3D Face Reconstruction From a Single Image via Direct Volumetric CNN Regression (ICCV 2017 · similarity 0.88)
- A Simple, Fast and Highly-Accurate Algorithm to Recover 3D Shape from 2D Landmarks on a Single Image (TPAMI 2017 · similarity 0.88)
- Pix2Vox: Context-Aware 3D Reconstruction From Single and Multi-View Images (ICCV 2019 · 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.
