Research map / 3D vision / Face and shape
SGMatch: Semantic-Guided Non-Rigid Shape Matching with Flow Regularization
Tianwei Ye, Xiaoguang Mei, Yifan Xia, Fan Fan, Jun Huang, Jiayi Ma
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 semantic features from DINOv2 to geometric features (ablation) | SGMatch · SMAL | geodesic error: 3.2 -> 2.5 | Improved |
| restrict semantic attention to mesh neighborhoods (ablation) | SGLCA · SMAL | geodesic error: 2.6 -> 2.5 (vs global attention) | Improved |
| add semantic gating on top of local attention (ablation) | SGLCA · SMAL | geodesic error: 2.6 -> 2.5 | Improved |
| add spectral heat diffusion (ablation) | SGMatch · SMAL | geodesic error: 3.0 -> 2.5 | Improved |
| add conditional flow matching regularization (ablation) | SGMatch · SMAL | conformal distortion: 2.32 -> 1.96 | Improved |
| CFM vs endpoint-only matching (ablation) | SGMatch · SMAL | Dirichlet energy: 14.37 -> 10.20 | Improved |
| CFM vs cycle-consistency on TOPKIDS (ablation) | SGMatch · TOPKIDS | Dirichlet energy: 42.42 -> 48.33 Dir.E; geodesic 3.4 -> 2.9 | Mixed |
| CFM vs Laplacian smoothing (ablation) | SGMatch · SMAL | geodesic error: 4.1 -> 2.5 | Improved |
| Charbonnier loss instead of MSE in CFM (ablation) | SGMatch · SMAL | geodesic error: 2.6 -> 2.5 | No clear effect |
| confidence sampling instead of uniform (ablation) | CFM objective · SMAL | geodesic error: 2.6 -> 2.5 | No clear effect |
| DINOv2 instead of CLIP or DINOv1 features (ablation) | SGMatch · SMAL | geodesic error: 5.8 / 4.1 -> 2.5 | Improved |
| use SGMatch vs HybridFMap on map smoothness at near-isometric | SGMatch · SCAPE | bijectivity: 0.0060 -> 0.0064 | Hurt |
Closest papers by meaning
- SGSoft: Learning Fused Semantic-Geometric Features for 3D Shape Correspondence via Template-Guided Soft Signals (CVPR 2026 · similarity 0.89)
- Learning SO(3)-Invariant Semantic Correspondence via Local Shape Transform (CVPR 2024 · similarity 0.87)
- Integrating Efficient Optimal Transport and Functional Maps For Unsupervised Shape Correspondence Learning (CVPR 2024 · similarity 0.87)
- Bridging Viewpoint Gaps: Geometric Reasoning Boosts Semantic Correspondence (CVPR 2025 · similarity 0.87)
- DV-Matcher: Deformation-based Non-rigid Point Cloud Matching Guided by Pre-trained Visual Features (CVPR 2025 · similarity 0.87)
- Pixel-level Semantic Correspondence through Layout-aware Representation Learning and Multi-scale Matching Integration (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.
