Research map / 3D vision / Depth and stereo
Sparse-View Localization via Online Neural 3D Regression
Ludvig Dillén, Magnus Oskarsson, Viktor Larsson
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 |
|---|---|---|---|
| ON3R vs transitive matching, K=2 | sparse-view localization · MegaDepth K=2 | recall (5deg,1m): 41.3 -> 71.0 | Improved |
| ON3R vs transitive matching on Cambridge, K=4 | sparse-view localization · Cambridge Landmarks K=4 | recall (5deg,1m): 91.7 -> 90.0 | No clear effect |
| add bundle adjustment after P3P-RANSAC (ablation) | ON3R · MegaDepth K=2 | median translation error (m): 9.31 -> 0.45 | Improved |
| add monocular depth prior (MoGe-2) (ablation) | ON3R · MegaDepth K=2 | recall (5deg,1m): 54.3 -> 71.0 | Improved |
| reduce max epochs from 500 to 50 (ablation) | ON3R · MegaDepth K=2 | recall (5deg,1m): 71.0 -> 66.0 | No clear effect |
| use DISK descriptor instead of SuperPoint (ablation) | ON3R · MegaDepth K=2 | median translation error (m): 0.45 -> 0.41 | Improved |
| ON3R vs HLOC on 99% sparsified Aachen | structure-based localization · Aachen Day-Night v1.1 sparsified | recall (10deg,5m) Day: 12.5 -> 22.7 | Improved |
| ON3R vs HLOC on full Aachen database (K=50) | structure-based localization · Aachen Day-Night v1.1 full | recall (10deg,5m) Day: 99.3 -> 95.1 | Hurt |
| ON3R vs HLOC on subsampled Cambridge (K=5, 10) | structure-based localization · Cambridge Landmarks subsampled | median errors and recall | Mixed |
| ON3R vs ACE (scene coordinate regression) with few images | sparse-view localization · MegaDepth K=2 | recall (5deg,1m): 1.3 -> 71.0 | Improved |
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.
- VGGT — VGGT: Visual Geometry Grounded Transformer (CVPR 2025) · better on 5 of 5 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
- Reloc3r: Large-Scale Training of Relative Camera Pose Regression for Generalizable, Fast, and Accurate Visual Localization (CVPR 2025 · similarity 0.85)
- Sparse-view Pose Estimation and Reconstruction via Analysis by Generative Synthesis (NeurIPS 2024 · similarity 0.84)
- SparseGNV: Generating Novel Views of Indoor Scenes with Sparse RGB-D Images (AAAI 2024 · similarity 0.84)
- SPARS3R: Semantic Prior Alignment and Regularization for Sparse 3D Reconstruction (CVPR 2025 · similarity 0.84)
- R-SCoRe: Revisiting Scene Coordinate Regression for Robust Large-Scale Visual Localization (CVPR 2025 · similarity 0.84)
- Pow3R: Empowering Unconstrained 3D Reconstruction with Camera and Scene Priors (CVPR 2025 · similarity 0.84)
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.
