Research map / Learning theory and algorithms / Optimal transport
Semantic Correspondence as an Optimal Transport Problem
Yanbin Liu, Linchao Zhu, Makoto Yamada, Yi Yang
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 cosine individual matching with globally constrained optimal transport (ablation) | Semantic correspondence on SPair-71k, ResNet101 with nearest-neighbor assignment · SPair-71k | PCK (αbbox=0.1): 30.6 -> 33.7 | Improved |
| Replace cosine individual matching with globally constrained optimal transport (ablation) | Semantic correspondence on SPair-71k, ResNet101 with RHM · SPair-71k | PCK (αbbox=0.1): 27.8 -> 34.8 | Improved |
| Add RHM post-processing to OT-NN (ablation) | Optimal-transport matching on SPair-71k, ResNet101 · SPair-71k | PCK (αbbox=0.1): 33.2 -> 34.8 | Improved |
| Add original CAM weighting rather than uniform empirical distributions (ablation) | OT-NN on SPair-71k · SPair-71k | PCK (αbbox=0.1): 29.4 -> 29.6; 33.2 -> 32.6 | No clear effect |
| Add staircase re-weighting to CAMs (ablation) | OT-NN on SPair-71k · SPair-71k | PCK (αbbox=0.1): 29.4 -> 30.2; 33.2 -> 34.2 | Improved |
| Add original CAM weighting rather than uniform empirical distributions (ablation) | OT-RHM on SPair-71k · SPair-71k | PCK (αbbox=0.1): 31.3 -> 31.4; 34.8 -> 34.5 | No clear effect |
| Add staircase re-weighting to CAMs (ablation) | OT-RHM on SPair-71k · SPair-71k | PCK (αbbox=0.1): 31.3 -> 32.1; 34.8 -> 35.6 | Improved |
| Use global OT instead of cosine matching (ablation) | Feature matching on SPair-71k · SPair-71k | unique target points assigned: 558 -> 1184; 445 -> 1062 | Improved |
Closest papers by meaning
- Guided Semantic Flow (ECCV 2020 · similarity 0.89)
- TransforMatcher: Match-to-Match Attention for Semantic Correspondence (CVPR 2022 · similarity 0.88)
- Correspondence Transformers With Asymmetric Feature Learning and Matching Flow Super-Resolution (CVPR 2023 · similarity 0.88)
- Pixel-level Semantic Correspondence through Layout-aware Representation Learning and Multi-scale Matching Integration (CVPR 2024 · similarity 0.87)
- Unifying Feature and Cost Aggregation with Transformers for Semantic and Visual Correspondence (ICLR 2024 · similarity 0.86)
- SemAlign3D: Semantic Correspondence between RGB-Images through Aligning 3D Object-Class Representations (CVPR 2025 · 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.
