Research map / 3D vision / Face and shape
Understanding and Improving Features Learned in Deep Functional Maps
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 feature-based properness loss/map (Pi computed from features) to DiffusionNet | DiffusionNet · FR on FR / FA on FA unsupervised | mean geodesic error: NN 6.5 -> 2.6 on FA on FA | Helped |
| Add properness and smooth features (projection into Laplacian basis at each layer) to DGCNN and DeltaConv | DGCNN, DeltaConv · supervised cross-dataset (FR on SR, etc.) | mean geodesic error (FM): DeltaConv FR on SR 18.1 -> 3.6 | Helped |
| Add properness + smoothness to extrinsic extractors | DGCNN, DeltaConv · SMAL supervised | mean geodesic error (NN): DeltaConv NN 20.7 -> 4.2 | Helped |
| Add properness + smoothness to DeltaConv (unsupervised FA on SHA) | DeltaConv · FA on SHA unsupervised | mean geodesic error (NN): 23.6 -> 6.1 | Helped |
| Ablation: feature-based properness alone vs. none on DiffusionNet (ablation) | DiffusionNet · FA on SHA unsupervised | mean geodesic error (NN): 11.9 -> 5.7 | Helped |
| Ablation: adding feature smoothness (3) to DGCNN with properness (1 or 2) (ablation) | DGCNN · FA on FA unsupervised | mean geodesic error (NN): DGCNN (2) 3.5 -> (2)+(3) 2.8 | Helped |
| Ablation: adding feature smoothness (3) to DeltaConv with feature-based properness (ablation) | DeltaConv · FA on FA unsupervised | mean geodesic error (NN): DeltaConv (2) 3.5 -> (2)+(3) 3.5 | Helped |
| Ablation: smoothness (3) alone on DGCNN (ablation) | DGCNN · FA on SHA unsupervised | mean geodesic error (NN): 17.1 -> 7.4 (smoothness alone) on FA on SHA NN; but FA on FA NN 9.3 -> 4.7 | Mixed |
Closest papers by meaning
- Memory-Scalable and Simplified Functional Map Learning (CVPR 2024 · similarity 0.93)
- Denoising Functional Maps: Diffusion Models for Shape Correspondence (CVPR 2025 · similarity 0.88)
- Hybrid Functional Maps for Crease-Aware Non-Isometric Shape Matching (CVPR 2024 · similarity 0.88)
- Spectral Meets Spatial: Harmonising 3D Shape Matching and Interpolation (CVPR 2024 · similarity 0.88)
- Locality Preserving Refinement for Shape Matching with Functional Maps (AAAI 2024 · similarity 0.87)
- Integrating Efficient Optimal Transport and Functional Maps For Unsupervised Shape Correspondence Learning (CVPR 2024 · 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.