Research map / 3D vision / Face and shape
The Phong Surface: Efficient 3D Model Fitting using Lifted Optimization
Jingjing Shen, Thomas J. Cashman, Qi Ye, Tim Hutton, Toby Sharp, Federica Bogo, Andrew Fitzgibbon, Jamie Shotton
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 data normal term (ablation) | Phong surface · Ellipsoid rigid pose alignment (lifted) | avg rotation error at 10 iters: 23.24 -> 8.13 | Improved |
| add data normal term (ablation) | Phong surface · Ellipsoid rigid pose alignment (lifted) | avg rotation error at 50 iters: 3.54 -> 0.99 | Improved |
| add data normal term (ablation) | Loop subdivision surface · Ellipsoid rigid pose alignment (lifted) | avg rotation error at 10 iters: 14.81 -> 9.89 | Improved |
| add data normal term (ablation) | Loop subdivision surface · Ellipsoid rigid pose alignment (lifted) | avg rotation error at 50 iters: 2.57 -> 1.21 | Improved |
| add data normal term (ablation) | Triangular mesh · Ellipsoid rigid pose alignment (lifted) | avg rotation error at 10 iters: 23.24 -> 17.47 | Improved |
| add data normal term (ablation) | Triangular mesh · Ellipsoid rigid pose alignment (lifted) | avg rotation error at 50 iters: 3.54 -> 11.07 | Hurt |
| use Phong surface instead of triangular mesh | lifted optimization · Ellipsoid rigid pose alignment (lifted) | avg rotation error at 50 iters: 11.07 -> 0.99 | Improved |
| use Phong surface instead of Loop subdivision | lifted optimization · Ellipsoid rigid pose alignment (lifted) | avg rotation error at 50 iters: 1.21 -> 0.99 | Improved |
| use lifted optimization instead of ICP | Phong surface · Ellipsoid rigid pose alignment | iterations to reach error <10 deg: 30 iterations -> 8 iters | Improved |
| use Phong surface instead of Loop subdivision | lifted optimization on HoloLens 2 · Hand tracking, HoloLens 2 depth data | evaluation speed: twice as fast | Improved |
| use Phong surface instead of Loop subdivision | lifted hand tracker · Dexter, per-frame fitting at 50 fps | iterations affordable: 4 iterations -> 6 iterations | Improved |
| use Phong surface instead of Loop subdivision | lifted hand tracker · Dexter, near 80% accuracy | speed: 20% faster | Improved |
Closest papers by meaning
- DICE: End-to-end Deformation Capture of Hand-Face Interactions from a Single Image (ICLR 2025 · similarity 0.83)
- MAtCha Gaussians: Atlas of Charts for High-Quality Geometry and Photorealism From Sparse Views (CVPR 2025 · similarity 0.83)
- HOT3D: Hand and Object Tracking in 3D from Egocentric Multi-View Videos (CVPR 2025 · similarity 0.83)
- HoloScene: Simulation‑Ready Interactive 3D Worlds from a Single Video (NeurIPS 2025 · similarity 0.83)
- HO-Cap: A Capture System and Dataset for 3D Reconstruction and Pose Tracking of Hand-Object Interaction (NeurIPS 2025 · similarity 0.83)
- PhyRecon: Physically Plausible Neural Scene Reconstruction (NeurIPS 2024 · similarity 0.83)
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.
