Research map / 3D vision / Depth and stereo
PATS: Patch Area Transportation With Subdivision for Local Feature Matching
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 patch area transportation with subdivision | Local feature matching · ScanNet | Pose estimation AUC @5°: 25.6 -> 26.0 | Helped |
| Use patch area transportation with subdivision | Local feature matching · YFCC100M | Pose estimation AUC @5°: 44.5 -> 47.0 | Helped |
| Use patch area transportation with subdivision | Local feature matching · Extreme-scale MegaDepth (1600) | Pose estimation AUC @5°: 51.8 -> 57.2 | Helped |
| Use patch area transportation with subdivision | Local feature matching · Extreme-scale MegaDepth (320) | Matching coverage (%): 24.7 -> 77.3 | Helped |
| Use patch area transportation with subdivision | Local feature matching · Extreme-scale MegaDepth (320) | Matching precision (%): 77.7 -> 85.9 | Helped |
| Add target-patch area regression (ablation) | PATS, L=2 · MegaDepth | Pose estimation AUC @5°: 50.7 -> 51.9 | Helped |
| Add patch area transportation rather than feature-similarity expectation (ablation) | PATS, L=2 · MegaDepth | Pose estimation AUC @5°: 40.1 -> 51.9 | Helped |
| Add patch concentration loss (ablation) | PATS, L=2 · MegaDepth | Pose estimation AUC @5°: 41.6 -> 51.9 | Helped |
| Add outlier patch loss (ablation) | PATS, L=2 · MegaDepth | Pose estimation AUC @5°: 36.3 -> 51.9 | Helped |
| Add inlier/outlier patch splitting (ablation) | PATS, L=2 · MegaDepth | Pose estimation AUC @5°: 50.9 -> 51.9 | Helped |
| Increase hierarchy count from 1 to 2 (ablation) | PATS · MegaDepth | Pose estimation AUC @5°: 0.7 -> 51.9 | Helped |
| Increase hierarchy count from 2 to 3 (ablation) | PATS · MegaDepth | Pose estimation AUC @5°: 51.9 -> 61.1 | Helped |
Closest papers by meaning
- Patch2Pix: Epipolar-Guided Pixel-Level Correspondences (CVPR 2021 · similarity 0.89)
- 3DMatch: Learning Local Geometric Descriptors From RGB-D Reconstructions (CVPR 2017 · similarity 0.86)
- LoCo: Learning 3D Location-Consistent Image Features with a Memory-Efficient Ranking Loss (NeurIPS 2024 · similarity 0.84)
- HomoMatcher: Achieving Dense Feature Matching with Semi-Dense Efficiency by Homography Estimation (AAAI 2025 · similarity 0.84)
- UFM: A Simple Path towards Unified Dense Correspondence with Flow (NeurIPS 2025 · similarity 0.83)
- Dual-Level Precision Edges Guided Multi-View Stereo with Accurate Planarization (AAAI 2025 · similarity 0.83)
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.