Research map / 3D vision / Depth and stereo
Feed-forward 3D reconstruction
Reconstructs scenes and structure-from-motion in a single pass with learned priors, planes and dense matching.
133 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2016–2026) · an approach in Depth and stereo, keyword cluster “motion · flow · localization”
State History Tried Checklist Preprints Who Export
Where it stands
Not enough benchmark data Fewer than four papers report the same benchmark, so there is no record to follow.
32 papers in 2023–24 and 58 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic grew, from 10.5% to 14.2%.
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 · Structure-From-Motion Revisited (CVPR 2016 · 6,291 citations)
- 2017 · Distinguishing the Indistinguishable: Exploring Structural Ambiguities via Geodesic Context (CVPR 2017 · 33 citations)
- 2018 · Very Large-Scale Global SfM by Distributed Motion Averaging (CVPR 2018 · 162 citations)
- 2019 · BA-Net: Dense Bundle Adjustment Networks (ICLR 2019 · 121 citations)
- 2020 · A Spatiotemporal Volumetric Interpolation Network for 4D Dynamic Medical Image (CVPR 2020 · 30 citations)
- 2021 · Stable View Synthesis (CVPR 2021 · 188 citations)
- 2022 · Iterative Corresponding Geometry: Fusing Region and Depth for Highly Efficient 3D Tracking of Textureless Objects (CVPR 2022 · 52 citations)
- 2023 · Tensor4D: Efficient Neural 4D Decomposition for High-Fidelity Dynamic Reconstruction and Rendering (CVPR 2023 · 176 citations)
- 2024 · DUSt3R: Geometric 3D Vision Made Easy (CVPR 2024 · 545 citations)
- 2025 · Fast3R: Towards 3D Reconstruction of 1000+ Images in One Forward Pass (CVPR 2025 · 70 citations)
- 2026 · RiemanLine: Riemannian Manifold Representation of 3D Lines for Factor Graph Optimization (AAAI 2026 · 1 citation)
Earliest papers here
- Structure-From-Motion Revisited (CVPR 2016 · 6,291 citations)
- Temporally Coherent 4D Reconstruction of Complex Dynamic Scenes (CVPR 2016 · 63 citations)
- Spatiotemporal Bundle Adjustment for Dynamic 3D Reconstruction (CVPR 2016 · 59 citations)
What has been tried
95 papers here report 909 measured changes (a component added, removed or swapped, compared with a baseline): helped 70%, no clear effect 7%, hurt 21%, mixed 3%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| Increase DUSt3R input resolution from 224 to 512 pixels | DDAD · AbsRel | 16.32 -> 13.88 | Helped | DUSt3R: Geometric 3D Vision Made Easy (CVPR 2024 · 545 citations) |
| train all components jointly end-to-end | Co3D · AUC@30° | 70.7% to 74.0% | Helped | VGGSfM: Visual Geometry Grounded Deep Structure From Motion (CVPR 2024 · 81 citations) |
| use the local pointmap aligned to global instead of the global pointmap directly | 7-Scenes · Acc# | 4.81 -> 2.84 | Helped | Fast3R: Towards 3D Reconstruction of 1000+ Images in One Forward Pass (CVPR 2025 · 70 citations) |
| add Gaussian heads | HM3D, 4 views · ND | 0.9 -> 1.0 | Hurt | MV-DUSt3R+: Single-Stage Scene Reconstruction from Sparse Views In 2 Seconds (CVPR 2025 · 35 citations) |
| Add Active Ray Sampling | Replica · Completion ratio (%) | 96.89 -> 97.20 | Helped | NARUTO: Neural Active Reconstruction from Uncertain Target Observations (CVPR 2024 · 17 citations) |
| Add Stabilized Rendering | NUS Crowd · Distortion Value | 0.87 -> 0.90 | Helped | 3D Multi-frame Fusion for Video Stabilization (CVPR 2024 · 15 citations) |
| add latent alignment | Tanks&Temples 200 views · RRA@5 | 47.5 -> 50.8 | Helped | Light3R-SfM: Towards Feed-forward Structure-from-Motion (CVPR 2025 · 14 citations) |
| add mutual verification to dense matches | ETH3D · accuracy (1cm) | 84.03 -> 84.79 | Helped | Dense-SfM: Structure from Motion with Dense Consistent Matching (CVPR 2025 · 13 citations) |
| Add EvRGBDegrader simulations of overexposure and background overflow | Challenging scenes · MPJPE | Helped | Complementing Event Streams and RGB Frames for Hand Mesh Reconstruction (CVPR 2024 · 13 citations) | |
| Compensate for global rigid motion in the training loss | Adult brain validation · Motion prediction accuracy | Helped | Fully Convolutional Slice-to-Volume Reconstruction for Single-Stack MRI (CVPR 2024 · 12 citations) | |
| Train self-supervised instead of supervised | SU3 · AUC @ $5\degree$ | 90.17 -> 86.14 | Hurt | PARSAC: Accelerating Robust Multi-Model Fitting with Parallel Sample Consensus (AAAI 2024 · 11 citations) |
| Add Stage III interaction optimization | H2O · G-MPJPE | 51.9 -> 45.6 | Helped | Dyn-HaMR: Recovering 4D Interacting Hand Motion from a Dynamic Camera (CVPR 2025 · 9 citations) |
What reviewers will expect
Datasets and baselines named in the 33 papers of this approach from 2025–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Datasets
- ScanNet · 9 of 33 papers
- MegaDepth · 5 of 33 papers
- ETH3D · 4 of 33 papers
- BlendedMVS · 4 of 33 papers
- ARKitScenes · 3 of 33 papers
- CO3Dv2 · 3 of 33 papers
- 7-Scenes · 3 of 33 papers
- DTU · 3 of 33 papers
Baselines
- COLMAP · 6 of 33 papers
- DUSt3R · 6 of 33 papers
- VGGSfM · 4 of 33 papers
- MASt3R-SfM · 4 of 33 papers
- MASt3R · 4 of 33 papers
- Spann3R · 4 of 33 papers
- DF-SfM · 3 of 33 papers
- RelPose++ · 3 of 33 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 90 papers; from OpenAlex and the papers' affiliations).
- Peking University · 4 papers
- Tsinghua University · 4 papers
- Meta · 4 papers
- Shanghai Jiao Tong University · 3 papers
- Huawei · 3 papers
- ETH Zurich · 3 papers
- Hong Kong University of Science and Technology · 2 papers
- Beijing Institute of Technology · 2 papers
- Mohamed bin Zayed University of Artificial Intelligence · 2 papers
- Technical University of Munich · 2 papers
Use it
Download BibTeX of 37 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
- Camera model and calibration for SfM (99 papers)
- Learned optical and scene flow (108 papers)
- Scene coordinate regression for localization (112 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).