Research map / Detection and segmentation / Open-set and few-shot detection
Proposal Flow: Semantic Correspondences from Object Proposals
Bumsub Ham, Minsu Cho, Cordelia Schmid, Jean Ponce
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 |
|---|---|---|---|
| use object proposals (SS) instead of sliding windows within proposal flow with LOM (ablation) | LOM · PF-WILLOW | PCK (alpha=0.1): 0.42 -> 0.56 | Improved |
| use SS instead of SW proposals on Caltech-101 (ablation) | LOM · Caltech-101 | IoU: 0.47 -> 0.50 | Improved |
| add local offset regularization (LOM) over PHM when background clutter is heavy (ablation) | PHM · PF-PASCAL qualitative pair | correct matches at IoU 0.5: 16, 5, and 38 for NAM, PHM, and LOM | Improved |
| use LOM instead of NAM (ablation) | proposal flow with SS · PF-PASCAL | PCK (alpha=0.1): 0.36 -> 0.45 | Improved |
| use LOM instead of PHM (ablation) | proposal flow with SS · PF-PASCAL | PCK (alpha=0.1): 0.42 -> 0.45 | Improved |
| use HOG descriptor instead of CNN features (AlexNet layers, SIAM) (ablation) | LOM with RP · PF-PASCAL | AuC of PCR | Improved |
| use RP proposals instead of other proposals (ablation) | LOM w/ HOG · PF-PASCAL | matching precision | Improved |
| increase number of object proposals (except MCG) (ablation) | proposal flow · PF-PASCAL and PF-WILLOW | AuC of PCR: saturate around 1000 proposals | Improved |
| increase number of MCG proposals (ablation) | LOM w/ MCG · PF-PASCAL and PF-WILLOW | AuC of PCR | Hurt |
| remove background clutter (wine without clutter vs with clutter) | LOM · PF-WILLOW | AuC of PCR: 0.30 -> 0.69 | Improved |
| add SS proposal flow vs FlowWeb multi-image refinement on PASCAL parts | LOM · PASCAL parts | PCK: 0.17 vs 0.26 | Hurt |
| LOM vs Taniai et al. on FG3DCar | LOM · Taniai's benchmark FG3DCar | matching accuracy: 0.79 vs 0.83 | Hurt |
Closest papers by meaning
- Proposal Flow (CVPR 2016 · similarity 0.99)
- Guided Semantic Flow (ECCV 2020 · similarity 0.91)
- SFNet: Learning Object-Aware Semantic Correspondence (CVPR 2019 · similarity 0.89)
- Learning Semantic Correspondence Exploiting an Object-Level Prior (TPAMI 2020 · similarity 0.89)
- Pixel-level Semantic Correspondence through Layout-aware Representation Learning and Multi-scale Matching Integration (CVPR 2024 · similarity 0.88)
- Semantic Flow: Learning Semantic Fields of Dynamic Scenes from Monocular Videos (ICLR 2024 · 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.
