Research map / Point clouds / Shape and surface completion
FUSE: A Flow-based Mapping Between Shapes
Lorenzo Olearo, Giulio Viganò, Daniele Baieri, Filippo Maggioli, Simone Melzi
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 KNN in embedding space with flow composition (FUSE) on GFE embeddings (ablation) | KNN · FAUST | Euclidean error: 0.0454 -> 0.0289 | Improved |
| replace KNN with FUSE on GFE embeddings (ablation) | KNN · SMAL | Euclidean error: 0.0879 -> 0.0595 | Improved |
| use high-frequency embeddings (WKS) with flow composition (ablation) | FUSE · SMAL | Euclidean error: 0.2736 -> 0.3039 | Hurt |
| spectrally smooth WKS embeddings before FUSE (ablation) | FUSE · FAUST | Euclidean error: 0.1015 -> 0.0901 | Improved |
| smooth pretrained ULRSSM embeddings (ablation) | FUSE · FAUST | Euclidean error: 0.082 -> 0.068 | Improved |
| use flow matching instead of DDIM inversion (ablation) | FUSE · FAUST | Euclidean error: 0.037 -> 0.028 | Improved |
| use flow matching instead of DDIM for anchor mapping (ablation) | FUSE · FAUST (anchor) | KL divergence: 0.3519 -> 0.0375 | Improved |
| initialize ZoomOut/NAM refinement with FUSE instead of FMap | FUSE-ZO/NAM · SHREC20 | Euclidean error: FMap-NAM 0.1041 -> FUSE-NAM 0.0658 | Improved |
| add ZoomOut refinement to FUSE (ablation) | FUSE · SHREC20 | Euclidean error: 0.0975 -> 0.1275 | Hurt |
| use unit sphere surface anchor instead of Gaussian (ablation) | FUSE · FAUST | Euclidean error: 0.0289 -> 0.2654 | Hurt |
| use Euler instead of Midpoint solver at 64 steps (ablation) | FUSE · FAUST | Euclidean error: 0.0289 -> 0.0364 | Hurt |
| reduce integration steps from 64 to 10 (forward flow) (ablation) | FUSE · FAUST | Euclidean error: 0.0289 -> 0.0438 | Hurt |
Closest papers by meaning
- Spectral Meets Spatial: Harmonising 3D Shape Matching and Interpolation (CVPR 2024 · similarity 0.86)
- Rectified Point Flow: Generic Point Cloud Pose Estimation (NeurIPS 2025 · similarity 0.86)
- Stable-SCore: A Stable Registration-based Framework for 3D Shape Correspondence (CVPR 2025 · similarity 0.86)
- Mosaic-SDF for 3D Generative Models (CVPR 2024 · similarity 0.85)
- Denoising Functional Maps: Diffusion Models for Shape Correspondence (CVPR 2025 · similarity 0.85)
- Integrating Efficient Optimal Transport and Functional Maps For Unsupervised Shape Correspondence Learning (CVPR 2024 · similarity 0.85)
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.
