Research map / 3D vision / Face and shape
Spatially and Spectrally Consistent Deep Functional Maps
Mingze Sun, Shiwei Mao, Puhua Jiang, Maks Ovsjanikov, Ruqi Huang
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 spatial branch (point-wise map from spectral latent shape) to a standard deep functional map (ablation) | Deep Functional Maps · SMAL_r | mean geodesic error (x100): 33.4 -> 5.4 | Improved |
| add spectral branch (FMreg layer) to spatial-only variant (ablation) | Deep Functional Maps · SMAL_r | mean geodesic error (x100): 14.3 -> 5.4 | Improved |
| use curriculum updating scheme for alpha instead of fixed alpha = 50 (ablation) | two-branch DFM · SMAL_r | mean geodesic error (x100): 35.2 -> 5.4 | Improved |
| use curriculum updating scheme for alpha instead of fixed alpha = 1 (ablation) | two-branch DFM · SMAL_r | mean geodesic error (x100): 6.6 -> 5.4 | Improved |
| add two-branch design to SURFMNet (ablation) | SURFMNet · FAUST_r / SCAPE_r / SMAL_r / DT4D-H | mean geodesic error (x100): 6.0 -> 3.5 (FAUST_r); 18.3 -> 15.0 (DT4D-H) | Improved |
| increase functional map dimension to 80 (ablation) | two-branch DFM · Train FAUST_r, test SHREC19_r | mean geodesic error (x100): 3.8 -> 5.5 (generalization); 2.3 -> 1.7 (FAUST_r) | Mixed |
| down-sample vertices to 3000 and descriptors to 30 dims (ablation) | two-branch DFM · SMAL_r | mean geodesic error (x100): 5.4 -> 5.7 / 5.9 | No clear effect |
| two-branch spectral and spatial consistency vs best prior on non-isometric DT4D-H | two-branch DFM · DT4D-H (168 train) | mean geodesic error (x100): 22.4 -> 7.7 | Improved |
| two-branch DFM in standard FAUST_r test vs UDMSM | two-branch DFM · Train FAUST_r, test FAUST_r | mean geodesic error (x100): 1.5 (UDMSM) vs 2.3 | Hurt |
Closest papers by meaning
- Memory-Scalable and Simplified Functional Map Learning (CVPR 2024 · similarity 0.92)
- Multi-Shape Matching with Cycle Consistency Basis via Functional Maps (AAAI 2025 · similarity 0.90)
- Spectral Meets Spatial: Harmonising 3D Shape Matching and Interpolation (CVPR 2024 · similarity 0.89)
- Latent Functional Maps: a spectral framework for representation alignment (NeurIPS 2024 · similarity 0.88)
- Integrating Efficient Optimal Transport and Functional Maps For Unsupervised Shape Correspondence Learning (CVPR 2024 · similarity 0.88)
- Denoising Functional Maps: Diffusion Models for Shape Correspondence (CVPR 2025 · 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.
