Research map / 3D vision / Face and shape
Hyper-Network Neural Functional Maps for Unsupervised Robust 3D Shape Matching
Dongliang Cao, Florian Bernard
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 hyper-network neural functional map to SSMSM | SSMSM · FAUST->SCAPE mesh matching | geodesic error x100: 9.2 -> 4.6 | Improved |
| add hyper-network neural functional map to SSMSM | SSMSM · SCAPE->FAUST mesh matching | geodesic error x100: 7.2 -> 3.2 | Improved |
| use ViT hyper-network instead of CNN (ablation) | NFM · TOPKIDS | geodesic error x100: 9.6 -> 6.7 | Improved |
| use ViT hyper-network instead of MLP (ablation) | NFM · TOPKIDS | geodesic error x100: 10.8 -> 6.7 | Improved |
| use 2 hidden layers instead of 1 (ablation) | NFM · TOPKIDS | geodesic error x100: 10.3 -> 6.7 | Improved |
| use 2 hidden layers instead of 3 or 4 (ablation) | NFM · TOPKIDS | geodesic error x100: 8.6 -> 6.7 (3 layers); 8.9 -> 6.7 (4 layers) | Improved |
| condition on the correct C instead of swapped C (ablation) | NFM hyper-network · SHREC16 HOLES | geodesic error x100: 32.8 -> 4.7 | Improved |
| use trained hyper-network instead of random init with test-time optimization (ablation) | NFM · SHREC16 HOLES | geodesic error x100: 61.9 -> 4.7 | Improved |
| train and test on different datasets (cross-dataset) (ablation) | NFM · SHREC16 HOLES | geodesic error x100: 4.7 -> 5.7 | Hurt |
| corrupt C with Gaussian noise at inference (ablation) | NFM · SHREC16 HOLES | geodesic error x100: 4.7 -> 6.2 | Hurt |
| flip signs of 50% eigenvectors at test time (ablation) | NFM · SHREC16 HOLES | geodesic error x100: 4.74 -> 4.82 | No clear effect |
| use spectral dimension k=128 instead of smaller k (ablation) | NFM · SHREC16 HOLES | geodesic error x100: 9.5 -> 4.7 (k=32) | Improved |
Closest papers by meaning
- Memory-Scalable and Simplified Functional Map Learning (CVPR 2024 · similarity 0.92)
- Spectral Meets Spatial: Harmonising 3D Shape Matching and Interpolation (CVPR 2024 · similarity 0.89)
- Denoising Functional Maps: Diffusion Models for Shape Correspondence (CVPR 2025 · similarity 0.88)
- Locality Preserving Refinement for Shape Matching with Functional Maps (AAAI 2024 · similarity 0.88)
- Integrating Efficient Optimal Transport and Functional Maps For Unsupervised Shape Correspondence Learning (CVPR 2024 · similarity 0.88)
- Hybrid Functional Maps for Crease-Aware Non-Isometric Shape Matching (CVPR 2024 · similarity 0.87)
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.
