Research map / 3D vision / Depth and stereo
PlanaReLoc: Camera Relocalization in 3D Planar Primitives via Region-Based Structure Matching
Hanqiao Ye, Yuzhou Liu, Yangdong Liu, Shuhan Shen
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 |
|---|---|---|---|
| add scene encoder to map primitive embedding (ablation) | PlanaReLoc (no refinement) · ScanNet (test) | F1 / pose recall (0.2 m,10 deg): 44.0 -> 64.3 F1; 17.2 -> 37.1 recall | Improved |
| add object encoder to map primitive embedding (ablation) | PlanaReLoc (no refinement) · ScanNet (test) | F1 / pose recall (0.2 m,10 deg): 51.7 -> 64.3 F1; 27.8 -> 37.1 recall | Improved |
| add rotary positional embedding from plane normals (ablation) | PlanaReLoc (no refinement) · ScanNet (test) | F1 / pose recall (0.2 m,10 deg): 60.1 -> 64.3 F1; 33.9 -> 37.1 recall | Improved |
| add RANSAC robust estimation (ablation) | PlanaReLoc (no refinement) · ScanNet (test) | pose recall (0.2 m,10 deg): 29.5 -> 37.1 | Improved |
| add joint metric scale optimization (ablation) | PlanaReLoc (no refinement) · ScanNet (test) | pose recall (0.2 m,10 deg): 36.2 -> 37.1 | Improved |
| add primitive-based post-refinement (ablation) | PlanaReLoc · ScanNet (test) | pose recall (0.2 m,10 deg): 37.1 -> 48.5 | Improved |
| add primitive-based post-refinement (ablation) | PlanaReLoc · 12Scenes | pose recall (0.2 m,10 deg): 34.9 -> 50.6 | Improved |
| replace MoGe-2+RANSAC plane recovery with learned plane recovery (PlaneTR, ZeroPlane) (ablation) | PlanaReLoc (no refinement) · ScanNet (test) | matching F1: 64.3 -> 65.8 / 66.0 | No clear effect |
| replace MoGe-2+RANSAC with ground-truth depth and masks (ablation) | PlanaReLoc (no refinement) · ScanNet (test) | matching F1: 64.3 -> 88.6 | Improved |
| use plane primitives instead of point features for cross-modal matching | PlanaReLoc vs MASt3R · ScanNet (test) | matching F1: 52.0 -> 64.3 | Improved |
| use planar-map matching without pose priors vs baselines needing coarse init | PlanaReLoc vs MASt3R+MeshLoc · ScanNet (test) | pose recall (0.2 m,10 deg): 35.4 -> 48.5 | Improved |
| evaluate PlanaReLoc zero-shot on 12Scenes vs best baseline | PlanaReLoc vs MatchAnything · 12Scenes | pose recall (1.0 m,30 deg): 77.6 -> 80.6; median trans error 0.20 -> 0.19 m | Mixed |
Later papers that compared with this method
1 later accepted paper reports PlanaReLoc as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- PGGT-Loc: Primitive-Grounded Geometry Transformer for Feed-Forward Camera Localization (NeurIPS 2026 · better on 7 of 7 shared results)
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
- Accelerated Coordinate Encoding: Learning to Relocalize in Minutes Using RGB and Poses (CVPR 2023 · similarity 0.87)
- Scene Coordinate Reconstruction: Posing of Image Collections via Incremental Learning of a Relocalizer (ECCV 2024 · similarity 0.87)
- PGGT-Loc: Primitive-Grounded Geometry Transformer for Feed-Forward Camera Localization (NeurIPS 2026 · similarity 0.87)
- PLANA3R: Zero-shot Metric Planar 3D Reconstruction via Feed-forward Planar Splatting (NeurIPS 2025 · similarity 0.86)
- Reloc3r: Large-Scale Training of Relative Camera Pose Regression for Generalizable, Fast, and Accurate Visual Localization (CVPR 2025 · similarity 0.86)
- Living Scenes: Multi-object Relocalization and Reconstruction in Changing 3D Environments (CVPR 2024 · similarity 0.85)
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.
