Research map / 3D vision / Depth and stereo
Raising the Ceiling: Conflict-Free Local Feature Matching with Dynamic View Switching
Xiaoyong Lu, Songlin Du
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 one-to-one with many-to-one coarse matching (ablation) | RCM semi-sparse matcher · MegaDepth ablation | #Matches: 507.8 -> 770.7 (+51.8%) | Improved |
| replace one-to-one with many-to-one coarse matching (ablation) | RCM semi-sparse matcher · MegaDepth ablation | precision: 94.2 -> 97.2 | Improved |
| add dynamic view switcher (ablation) | RCM semi-sparse matcher · MegaDepth ablation | #Matches: 770.7 -> 951.4 (+23.4%) | Improved |
| add U-Net feature extraction (ablation) | semi-sparse baseline · MegaDepth ablation | #Matches: 492.3 -> 507.8 | Improved |
| raise coarse resolution from 1/16 to 1/8 (ablation) | RCM · MegaDepth ablation | #Matches: 951.4 -> 961.4 | Improved |
| remove view switcher on extreme-scale pairs (ablation) | RCM · extreme-scale MegaDepth | #Matches: 887 -> 653 | Hurt |
| flip the switcher decision so small-scale image is source (ablation) | RCM · extreme-scale MegaDepth | #Matches: 887 -> 76 | Hurt |
| use semi-sparse sampling instead of keypoints in both images (sparse) (ablation) | baseline matcher · MegaDepth ablation | precision: 98.0 -> 93.7 | Hurt |
| dense sampling vs sparse sampling for coarse features (ablation) | baseline matcher · MegaDepth ablation | #Matches: 451.7 -> 932.1 | Mixed |
| RCM vs LoFTR on ScanNet pose (zero-shot) | RCM · ScanNet pose estimation (zero-shot from MegaDepth) | AUC@10deg: 33.6 -> 34.6 | Improved |
| RCM vs RoMa on MegaDepth pose | RCM · MegaDepth pose estimation | AUC@10deg: 69.4 vs 76.9 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- LightGlue — LightGlue: Local Feature Matching at Light Speed (ICCV 2023) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Toward Free-Form Local Feature Matching (TPAMI 2025 · similarity 0.92)
- OmniGlue: Generalizable Feature Matching with Foundation Model Guidance (CVPR 2024 · similarity 0.86)
- Efficient LoFTR: Semi-Dense Local Feature Matching with Sparse-Like Speed (CVPR 2024 · similarity 0.86)
- JamMa: Ultra-lightweight Local Feature Matching with Joint Mamba (CVPR 2025 · similarity 0.86)
- MATCHA: Towards Matching Anything (CVPR 2025 · similarity 0.86)
- CoMatcher: Multi-View Collaborative Feature Matching (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.
