Research map / Learning theory and algorithms / Optimal transport
Shape-of-You: Fused Gromov-Wasserstein Optimal Transport for Semantic Correspondence in-the-Wild
Jiin Im, Sisung Liu, Je Hyeong Hong
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 |
|---|---|---|---|
| train adapter on FGW labels instead of NN labels (ablation) | adapter on DINOv2+SD · SPair-71k | PCK@0.1: 64.8 -> 66.8 | Improved |
| add relaxed cycle consistency to NN labels (ablation) | adapter with NN labels · SPair-71k | PCK@0.1: 64.6 -> 64.8 | No clear effect |
| add relaxed cycle consistency to FGW labels (ablation) | adapter with FGW labels · SPair-71k | PCK@0.1: 66.8 -> 67.1 | Improved |
| add soft target loss (ablation) | adapter with FGW labels · SPair-71k | PCK@0.1: 66.8 -> 67.5 | Improved |
| add soft target loss on top of relaxed c.c. (ablation) | adapter with FGW labels · SPair-71k | PCK@0.1: 67.1 -> 67.9 | Improved |
| use 3D geometric distance as GW intra-structure (ablation) | Fused OT pseudo-labels · SPair-71k pseudo-label | PCK_label@0.1: 66.5 -> 67.6 | Improved |
| use 2D distance as GW intra-structure (ablation) | Fused OT pseudo-labels · SPair-71k pseudo-label | PCK_label@0.1: 66.5 -> 65.7 | Hurt |
| use semantic intra-feature distance as intra-structure (ablation) | Fused OT pseudo-labels · SPair-71k pseudo-label | PCK_label@0.1: 66.5 -> 66.2 | No clear effect |
| use unbalanced OT instead of balanced fused OT (ablation) | pseudo-label generation · SPair-71k geometry-aware subset | PCK_label@0.1: 55.7 -> 56.1 | Improved |
| replace VGGT with DUSt3R as 3D backbone (ablation) | FGW pseudo-labels · SPair-71k pseudo-label | PCK_label@0.1: 68.1 -> 67.1 | Hurt |
| increase anchors K from 16 to 64 (ablation) | anchor-based FGW linearization · SPair-71k pseudo-label | PCK_label@0.1: 67.6 -> 68.1 | Improved |
| set alpha 0.3 instead of 0.5 or 0.1 (ablation) | FGW fusion · SPair-71k pseudo-label | PCK_label@0.1: 67.2 -> 68.1 | Improved |
Closest papers by meaning
- DiffCorr: Conditional Diffusion Model with Reliable Pseudo-Label Guidance for Unsupervised Point Cloud Shape Correspondence (AAAI 2025 · similarity 0.85)
- Jamais Vu: Exposing the Generalization Gap in Supervised Semantic Correspondence (NeurIPS 2025 · similarity 0.85)
- Learning SO(3)-Invariant Semantic Correspondence via Local Shape Transform (CVPR 2024 · similarity 0.84)
- Improving Semantic Correspondence with Viewpoint-Guided Spherical Maps (CVPR 2024 · similarity 0.84)
- Diffusion 3D Features (Diff3F): Decorating Untextured Shapes with Distilled Semantic Features (CVPR 2024 · similarity 0.84)
- SemAlign3D: Semantic Correspondence between RGB-Images through Aligning 3D Object-Class Representations (CVPR 2025 · similarity 0.84)
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.
