Research map / 3D vision / Depth and stereo
Scene coordinate regression for localization
Localizes cameras by regressing scene coordinates or maps, with voting and 3D surface prediction.
112 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2016–2026) · an approach in Depth and stereo, keyword cluster “motion · flow · localization”
State History Benchmarks Tried Baselines Checklist Preprints Who Export
Where it stands
Few recent reports None of the tracked benchmarks has a report from 2025–26.
28 papers in 2023–24 and 40 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic held steady, from 9.2% to 9.8%.
Open questions the data points to
- Papers report DROID-SLAM on Replica (accuracy (cm)) anywhere from 4.23 to 5.5: comparisons against it depend on whose number you take.
No longer reported by papers here: Replica (ATE RMSE [cm], last in 2024).
How this is decided
A benchmark is tracked when at least four papers of this approach report their own method on the same dataset and metric in at least two different years. Its record is the best number reported up to 2024 compared with the best reported in 2025–26: still improving means at least 1% better, flat means at most 0.2% better with at least two papers from 2025–26 trying. Numbers on a different scale from the rest (a fraction among percentages) and far outliers are left out. Results come from different papers with different setups, so they are signals, not head-to-head comparisons. The share of papers is counted at ICML, NeurIPS, ICLR, CVPR and AAAI.
How it got here
Most cited paper of each year
- 2016 · Egocentric Future Localization (CVPR 2016 · 106 citations)
- 2017 · VINet: Visual-Inertial Odometry as a Sequence-to-Sequence Learning Problem (AAAI 2017 · 325 citations)
- 2018 · Benchmarking 6DOF Outdoor Visual Localization in Changing Conditions (CVPR 2018 · 720 citations)
- 2019 · End-To-End Interpretable Neural Motion Planner (CVPR 2019 · 408 citations)
- 2020 · Learning To Explore Using Active Neural SLAM (ICLR 2020 · 230 citations)
- 2021 · DROID-SLAM: Deep Visual SLAM for Monocular, Stereo, and RGB-D Cameras (NeurIPS 2021 · 231 citations)
- 2022 · Beyond Cross-View Image Retrieval: Highly Accurate Vehicle Localization Using Satellite Image (CVPR 2022 · 94 citations)
- 2023 · Co-SLAM: Joint Coordinate and Sparse Parametric Encodings for Neural Real-Time SLAM (CVPR 2023 · 269 citations)
- 2024 · SplaTAM: Splat Track & Map 3D Gaussians for Dense RGB-D SLAM (CVPR 2024 · 438 citations)
- 2025 · MASt3R-SLAM: Real-Time Dense SLAM with 3D Reconstruction Priors (CVPR 2025 · 81 citations)
- 2026 · MRGeo: Robust Cross-View Geo-Localization of Corrupted Images via Spatial and Channel Feature Enhancement (AAAI 2026 · 1 citation)
Earliest papers here
- Egocentric Future Localization (CVPR 2016 · 106 citations)
- Joint Multiview Segmentation and Localization of RGB-D Images Using Depth-Induced Silhouette Consistency (CVPR 2016 · 4 citations)
- VINet: Visual-Inertial Odometry as a Sequence-to-Sequence Learning Problem (AAAI 2017 · 325 citations)
Benchmark records
The best number each year that beat every earlier one, among papers of this approach reporting their own method. Numbers are extracted by AI from each paper's text; check the paper before citing.
Replica · ATE RMSE [cm] (lower is better)
4 papers, 2023–2024 · not reported since 2024
| Year | Best so far | Set by |
|---|---|---|
| 2023 | 0.46 | DENSE RGB SLAM WITH NEURAL IMPLICIT MAPS (ICLR 2023 · 11 citations) |
| 2024 | 0.29 | Loopy-SLAM: Dense Neural SLAM with Loop Closures (CVPR 2024 · 55 citations) |
What has been tried
81 papers here report 791 measured changes (a component added, removed or swapped, compared with a baseline): helped 70%, no clear effect 3%, hurt 23%, mixed 4%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| remove forward velocity propagation | Replica Room 0 · ATE [cm] | 0.27 -> 2.95 | Hurt | SplaTAM: Splat Track & Map 3D Gaussians for Dense RGB-D SLAM (CVPR 2024 · 438 citations) |
| add geometry-based densification | Replica, monocular · PSNR | 31.274 -> 33.302 | Helped | Photo-SLAM: Real-time Simultaneous Localization and Photorealistic Mapping for Monocular Stereo and RGB-D Cameras (CVPR 2024 · 205 citations) |
| add coarse-to-fine hierarchical semantic mapping | Replica office0 · Tracking RMSE [cm] | 0.83 -> 0.55 | Helped | SNI-SLAM: Semantic Neural Implicit SLAM (CVPR 2024 · 82 citations) |
| add feature refinement to projective matching | TUM / 7-Scenes / EuRoC · ATE [m] w/ calib | 0.062 -> 0.039 | Helped | MASt3R-SLAM: Real-Time Dense SLAM with 3D Reconstruction Priors (CVPR 2025 · 81 citations) |
| Use joint tri-plane and MLP encoding | Apartment · Accuracy [cm] | 13.314 -> 11.086 | Helped | PLGSLAM: Progressive Neural Scene Represenation with Local to Global Bundle Adjustment (CVPR 2024 · 70 citations) |
| Add Re-Attention | Wayspots · frames below 5cm/5° | 10.9% -> 16.6% | Helped | Map-Relative Pose Regression for Visual Re-Localization (CVPR 2024 · 42 citations) |
| increase the I2P window from 2 to 11 frames | Replica · accuracy | 3.39 -> 2.38 | Helped | SLAM3R: Real-Time Dense Scene Reconstruction from Monocular RGB Videos (CVPR 2025 · 29 citations) |
| Add event tracking and event bundle adjustment | #Rm blur, DEV-Indoors · ATE RMSE (cm) | 18.75 -> 15.47 | Helped | Implicit Event-RGBD Neural SLAM (CVPR 2024 · 15 citations) |
| add T2GA to PureACL | KITTI-CVL K-Cross (±45°, ±20m) · mean lateral error | 4.24 -> 1.29 | Helped | View From Above: Orthogonal-View aware Cross-view Localization (CVPR 2024 · 13 citations) |
| Apply VC2 rectification to query/reference images for image retrieval | 360Loc, pinhole, fisheye1, fisheye2, and fisheye3 queries · R@1 | 0.23 -> 0.5; 0.42 -> 0.73; 0.45 -> 0.74; 0.57 -> 0.77 | Helped | 360Loc: A Dataset and Benchmark for Omnidirectional Visual Localization with Cross-device Queries (CVPR 2024 · 12 citations) |
| replace Gaussian noise with covisibility-based data augmentation | Hyundai Department Store Dept. 1F Val · accuracy (0.1m, 1°) | 42.1 -> 62.0 | Helped | R-SCoRe: Revisiting Scene Coordinate Regression for Robust Large-Scale Visual Localization (CVPR 2025 · 11 citations) |
| use SL(4) instead of Sim(3) submap alignment | TUM RGB-D · ATE RMSE | 0.074 -> 0.053 | Helped | VGGT-SLAM: Dense RGB SLAM Optimized on the SL(4) Manifold (NeurIPS 2025 · 8 citations) |
Baseline numbers that disagree
The same baseline on the same benchmark and metric, reported more than 5% apart by different papers here. Training budgets, data splits and reimplementations differ, so state which number you compare with.
| Baseline | Benchmark | Metric | Reported | Papers |
|---|---|---|---|---|
| DROID-SLAM | Replica | accuracy (cm) | 4.23 – 5.5 | 2 papers, e.g. ICLR 2023 and CVPR 2025 |
What reviewers will expect
Datasets and baselines named in the 17 papers of this approach from 2025–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Datasets
- TUM RGB-D · 3 of 17 papers
- 7-Scenes · 3 of 17 papers
- ScanNet · 2 of 17 papers
Baselines
- NICER-SLAM · 4 of 17 papers
- DROID-SLAM · 4 of 17 papers
- GO-SLAM · 4 of 17 papers
- ORB-SLAM3 · 4 of 17 papers
- DPV-SLAM · 3 of 17 papers
- Spann3R · 3 of 17 papers
- PoseNet · 2 of 17 papers
- SGLoc · 2 of 17 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 68 papers; from OpenAlex and the papers' affiliations).
- Hong Kong University of Science and Technology · 4 papers
- Shanghai Jiao Tong University · 3 papers
- Xiamen University · 3 papers
- The Chinese University of Hong Kong · 2 papers
- Czech Technical University in Prague · 2 papers
- Sun Yat-sen University · 2 papers
- Northwestern Polytechnical University · 2 papers
- ETH Zurich · 2 papers
- University of Oxford · 2 papers
- Alibaba Group · 2 papers
Use it
Download BibTeX of 38 key papersOpen the topic in the interactive map
Get alerts. Describe your idea in scime atlas and save it (free scime account): you get a weekly email when new preprints land close to it.
Related approaches
Same keyword cluster
- Feed-forward 3D reconstruction (133 papers)
- Camera model and calibration for SfM (99 papers)
- Learned optical and scene flow (108 papers)
- Dense feature matching for correspondence (80 papers)
Elsewhere in the topic
- Self-supervised monocular depth (116 papers)
- Neural light field and illumination imaging (107 papers)
- Cost-volume multi-view stereo (107 papers)
- Non-line-of-sight imaging (93 papers)
- Dense depth prediction and completion (85 papers)
- Foundation models for monocular geometry (81 papers)
- Depth from light field and defocus (60 papers)
- Learned stereo with specialized sensors (53 papers)
Benchmark numbers, ablations and baseline numbers are extracted by AI from each paper's full text and can be wrong; the paper is the authority. Data: results index of 2026-11-02, citation counts from OpenAlex (2026-10-09).