Research map / 3D vision / Face and shape
What Does 2D Geometric Information Really Tell Us About 3D Face Shape?
Anil Bas, William A. P. Smith
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 alternating least squares with separable nonlinear least squares | 3DMM landmark fitting · BFM synthetic, detected landmarks | surface error d_S (mm): 4.81 -> 4.21 | Improved |
| replace alternating least squares with SNLS | 3DMM landmark fitting · FaceWarehouse leave-one-out | surface error: about 30% lower error | Improved |
| fit at 120cm instead of true 30cm distance (ablation) | perspective 3DMM fit · BFM perspective ambiguity, true distance 30 cm | landmark error d_L (%): 0.21 -> 0.26 | No clear effect |
| fit at 120cm instead of true 30cm distance (ablation) | perspective 3DMM fit · BFM perspective ambiguity, true distance 30 cm | surface error d_S (mm): 7.23 -> 13.07 | Hurt |
| fit orthographic instead of true 240cm perspective (ablation) | perspective 3DMM fit · BFM perspective ambiguity, true distance 240 cm | surface error d_S (mm): 5.37 -> 5.38 | No clear effect |
| let distance be estimated as a fitting parameter | perspective least squares fit · BFM distance estimation | distance error (%): 42% mean error | Hurt |
| compare identities at same distance vs same identity at different distances | landmark geometry · CMDP 53 subjects | landmark error (%): 3.11 vs 2.86 and 2.83 | Hurt |
| use known subject-camera distance instead of unknown (ablation) | perspective fit · CMDP identity matching | ROC AUC: 0.690 -> 0.892 | Improved |
| use pretrained CNN regression instead of unknown-distance geometric fit | 3D face regression · CMDP identity matching | ROC AUC: 0.690 -> 0.866 | Improved |
| restrict flexibility modes to plausible parameter vectors (ablation) | 3DMM flexibility modes · BFM 70 landmarks orthographic | number of modes: around 50 -> 7 | Mixed |
Closest papers by meaning
- Discovering Geometric Biases in 3D Face Reconstruction: A Curvature-Aware Spectral Framework for Fairness Evaluation (ECCV 2026 · similarity 0.89)
- Beyond 3DMM: Learning to Capture High-Fidelity 3D Face Shape (TPAMI 2022 · similarity 0.87)
- MVF-Net: Multi-View 3D Face Morphable Model Regression (CVPR 2019 · similarity 0.87)
- Disentangling Features in 3D Face Shapes for Joint Face Reconstruction and Recognition (CVPR 2018 · similarity 0.87)
- Learning Complete 3D Morphable Face Models From Images and Videos (CVPR 2021 · similarity 0.86)
- GIQ: Benchmarking 3D Geometric Reasoning of Vision Foundation Models with Simulated and Real Polyhedra (ICLR 2026 · 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.
