Research map / 3D vision / Face and shape
Shape My Face: Registering 3D Face Scans by Surface-to-Surface Translation
Mehdi Bahri, Eimear O’ Sullivan, Shunwang Gong, Feng Liu, Xiaoming Liu, Michael M. Bronstein, Stefanos P. 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 PointNet baseline with SMF (attention, mesh decoders, no cropping) on out-of-sample scans | Baseline of Chen et al. · BU-3DFE landmark error | Avg landmark error (mm): 16.45 -> 8.72 | Improved |
| replace baseline with SMF | Baseline of Chen et al. · 3DMD surface error | 50% quantile (mm): 3.239 -> 0.447 | Improved |
| add MeIn3D and 4DFAB data (SMF+) to SMF training (ablation) | SMF · 3DMD surface error | 50% quantile (mm): 0.447 -> 0.407 | Improved |
| add extra datasets (SMF+) (ablation) | SMF · 3DMD surface error | 99% quantile (mm): 1.489 -> 1.326 | Improved |
| add extra datasets (SMF+) (ablation) | SMF · BU-3DFE surface error (in sample) | 50% quantile (mm): 0.347 -> 0.333 | Improved |
| initialize NICP with SMF registration (ablation) | NICP · 3DMD landmark error | Avg landmark error (mm): 10.41 -> 10.36 | No clear effect |
| initialize NICP with SMF instead of landmarks | NICP · registration runtime | seconds per scan: 45-60 s -> 20 s | Improved |
| train SMF without BU-3DFE (out of sample) (ablation) | SMF · BU-3DFE landmark error | Avg landmark error (mm): 9.07 -> 8.72 | No clear effect |
| train SMF without BU-3DFE (out of sample) (ablation) | SMF · BU-3DFE surface error | 50% quantile (mm): 0.347 -> 0.501 | Hurt |
| SMF vs NICP with landmarks on inner-face landmark error | NICP · 3DMD landmark error | Avg Face (mm): 4.43 vs 5.57 | Hurt |
| SMF vs baseline on jaw landmarks | SMF · BU-3DFE landmark error | Chin (mm): 11.39 -> 38.73 | Hurt |
Closest papers by meaning
- Generating 3D Faces using Convolutional Mesh Autoencoders (ECCV 2018 · similarity 0.89)
- A Sparse and Locally Coherent Morphable Face Model for Dense Semantic Correspondence Across Heterogeneous 3D Faces (TPAMI 2021 · similarity 0.89)
- 3DFaceGAN: Adversarial Nets for 3D Face Representation, Generation, and Translation (IJCV 2020 · similarity 0.88)
- Registration-Free Learnable Multi-View Capture of Faces in Dense Semantic Correspondence (CVPR 2026 · similarity 0.88)
- Representing 3D Faces with Learnable B-Spline Volumes (CVPR 2026 · similarity 0.88)
- Learning Complete 3D Morphable Face Models From Images and Videos (CVPR 2021 · 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.
