pith. sign in

arxiv: 2606.05035 · v1 · pith:FTZFGTYKnew · submitted 2026-06-03 · 💻 cs.CV

Anchor3R: Streaming 3D Reconstruction with Transient Anchors for Long-Horizon Visual Mapping

Peilin Tao , Chong Cheng , Yuansen Du , Caiwei Song , Zhengqing Chen , Xiaoyang Guo , Wei Yin , Weiqiang Ren
show 3 more authors
Qian Zhang Hainan Cui Shuhan Shen
This is my paper

Pith reviewed 2026-06-28 07:01 UTC · model grok-4.3

classification 💻 cs.CV
keywords streaming 3D reconstructionvisual mappinglong-horizon pose estimationloop closurepointmap predictionbounded memory inferencefeed-forward reconstructiontransient anchors
0
0 comments X

The pith

Anchor3R reconstructs long video sequences by predicting local poses and points relative to the current frame then aligning them globally.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper introduces Anchor3R as a way to handle streaming 3D reconstruction over extended periods without the drift that builds when models tie everything to an early fixed coordinate frame. Instead of regressing directly to a persistent global scene, each step produces measurements only in the coordinate system of the latest frame. These local predictions are then fused using loop-closure reinsertion and motion averaging to produce a single coherent map. The result is claimed to keep memory bounded while raising both trajectory accuracy and surface quality on indoor, outdoor, driving, and RGB-D sequences. A sympathetic reader would care because many robot-perception tasks need continuous mapping that stays accurate far beyond the short clips used in training.

Core claim

Anchor3R reframes feed-forward reconstruction as the generation of current-centric local measurements rather than persistent global-gauge regression; at each step it outputs window-relative poses and a local pointmap in the coordinate system of the present frame, after which loop-closure reinsertion and motion averaging convert those measurements into an aligned global reconstruction.

What carries the argument

Transient anchors that encode window-relative poses and local pointmaps predicted in the current-frame coordinate system, which are later aligned by loop-closure reinsertion and motion averaging.

If this is right

  • Sequences much longer than the training clips can be processed without the attention bias toward early frames that fixed-gauge models exhibit.
  • Memory use stays bounded because no persistent global scene memory is maintained across time steps.
  • Dense pointmap quality improves on indoor, outdoor, driving, and RGB-D benchmarks compared with prior streaming methods.
  • Online inference remains feasible on resource-limited platforms because each prediction step is local.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same current-centric measurement strategy could be tested on other feed-forward reconstruction networks to check whether the gain is architecture-specific or general.
  • If alignment remains reliable, hybrid pipelines that combine the local predictions with classical bundle adjustment might further reduce residual drift.
  • The approach suggests that training objectives for streaming models should emphasize relative rather than absolute coordinate regression.

Load-bearing premise

Local measurements made in the current frame can be aligned into one consistent global map by loop closure and motion averaging without adding new drift or inconsistencies on long sequences.

What would settle it

A long video sequence on which the final global map built from current-frame measurements shows higher pose error or lower reconstruction quality than a fixed-gauge streaming baseline run on the same data.

Figures

Figures reproduced from arXiv: 2606.05035 by Caiwei Song, Chong Cheng, Hainan Cui, Peilin Tao, Qian Zhang, Shuhan Shen, Weiqiang Ren, Wei Yin, Xiaoyang Guo, Yuansen Du, Zhengqing Chen.

Figure 1
Figure 1. Figure 1: Anchor3R results on campus-scale sequences. We visualize our reconstruction on campus train0 and campus train1 of VBR dataset, two real-world sequences with 12,042 and 11,671 frames, completing inference on a single 32GB RTX 5090 GPU. Predicted trajectories are color-coded and ground truth is shown as white curves. Anchor3R achieves 5.13m and 3.52m ATE, respectively. Abstract: Long-horizon online visual ma… view at source ↗
Figure 2
Figure 2. Figure 2: Overview of Anchor3R. Given the current frame It, Anchor3R extracts DINOv2 patch tokens and instantiates grouped pose-query tokens for frames in Wt, using It as the local reference. A sliding-window pose-query Transformer alternates between decoupled frame attention and current￾centric window attention. The camera head decodes window-relative poses, while the point head predicts the current-frame pointmap.… view at source ↗
Figure 3
Figure 3. Figure 3: Attention-score comparison. The first-centric streaming baseline over-attends to the first frame, while current-centric formula￾tion spreads attention within the active window [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
read the original abstract

Long-horizon online visual mapping is a core capability for robot perception, requiring continuous camera-motion and scene-geometry estimation from visual streams under bounded memory and computation. Recent feed-forward 3D reconstruction models provide strong geometric priors, but their streaming variants often predict poses in a fixed coordinate system tied to the first frame or a persistent scene memory. This fixed-gauge design leads to train--test mismatch, attention bias toward early anchors, and accumulated drift on sequences much longer than those seen during training. We propose \emph{Anchor3R}, a streaming 3D reconstruction framework that treats feed-forward reconstruction as current-centric local measurement prediction rather than persistent global-gauge regression. At each time step, Anchor3R predicts window-relative poses and a local pointmap in the current-frame coordinate system, turning streaming reconstruction into relative-pose measurement generation. These measurements support online pose updates, while loop-closure reinsertion and motion averaging align the trajectory and transform local pointmaps into a coherent global reconstruction. Experiments on indoor, outdoor, driving, and RGB-D benchmarks show that Anchor3R improves long-horizon pose accuracy and dense reconstruction quality over existing streaming baselines, while supporting bounded-memory online inference.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 1 minor

Summary. The paper proposes Anchor3R, a streaming 3D reconstruction method that reframes feed-forward models as generators of current-centric local measurements (window-relative poses and local pointmaps) rather than fixed-gauge global outputs. These measurements are aligned into a coherent global reconstruction via loop-closure reinsertion and motion averaging. The central claim is that this transient-anchor design reduces drift on long-horizon sequences, supports bounded-memory online inference, and yields better pose accuracy plus dense reconstruction quality than existing streaming baselines across indoor, outdoor, driving, and RGB-D benchmarks.

Significance. If the empirical gains are reproducible, the approach would be significant for robot perception by mitigating train-test mismatch and attention bias to early frames that plague persistent-memory streaming methods. The decomposition into relative measurements plus standard alignment is internally consistent with the bounded-memory goal and does not rely on unstated axioms.

major comments (2)
  1. [Abstract] Abstract: the central empirical claim of improved long-horizon pose accuracy and reconstruction quality is stated without any quantitative metrics, error tables, sequence lengths, or experimental protocol, which is load-bearing for assessing whether the alignment step actually delivers the promised gains without new drift.
  2. [Experiments] Experiments (implied by abstract claims): the paper must demonstrate via ablations or analysis that loop-closure reinsertion plus motion averaging does not introduce inconsistency or additional drift on sequences longer than the training horizon, as this is the weakest link in the stated pipeline.
minor comments (1)
  1. The title uses 'Transient Anchors' but the abstract does not define the term or contrast it explicitly with persistent anchors; a short definition or figure would improve clarity.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments. We address each major point below and indicate planned revisions to strengthen the manuscript.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the central empirical claim of improved long-horizon pose accuracy and reconstruction quality is stated without any quantitative metrics, error tables, sequence lengths, or experimental protocol, which is load-bearing for assessing whether the alignment step actually delivers the promised gains without new drift.

    Authors: We agree that the abstract would benefit from quantitative support. In revision we will insert concise metrics (e.g., average ATE reduction and reconstruction F-score gains on sequences of 500–2000 frames) drawn directly from the experimental tables, together with a brief statement of the evaluation protocol. revision: yes

  2. Referee: [Experiments] Experiments (implied by abstract claims): the paper must demonstrate via ablations or analysis that loop-closure reinsertion plus motion averaging does not introduce inconsistency or additional drift on sequences longer than the training horizon, as this is the weakest link in the stated pipeline.

    Authors: The manuscript already reports results on sequences substantially longer than the training horizon and shows lower drift than persistent-memory baselines. However, we acknowledge that explicit ablations isolating the incremental effect of loop-closure reinsertion and motion averaging on drift accumulation for ultra-long sequences are not presented. We will add a targeted ablation and drift-vs-length plot in the revised experiments section. revision: yes

Circularity Check

0 steps flagged

No significant circularity in derivation chain

full rationale

The paper describes Anchor3R as an engineering design that reframes feed-forward reconstruction outputs as current-frame relative-pose measurements, then applies standard loop-closure reinsertion and motion averaging for global alignment. No equations, first-principles derivations, or parameter-fitting steps are presented that could reduce a claimed prediction to its own inputs by construction. The central claims rest on empirical benchmark comparisons rather than any self-referential mathematical chain, self-citation load-bearing argument, or ansatz smuggled via prior work. The approach is therefore self-contained against external validation.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract supplies no explicit free parameters, axioms, or invented entities; the method relies on existing feed-forward models and standard loop-closure techniques.

pith-pipeline@v0.9.1-grok · 5774 in / 1009 out tokens · 24608 ms · 2026-06-28T07:01:10.065834+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

57 extracted references · 13 linked inside Pith

  1. [1]

    Q. Wang, Y . Zhang, A. Holynski, A. A. Efros, and A. Kanazawa. Continuous 3d perception model with persistent state. InProceedings of the Computer Vision and Pattern Recognition Conference, pages 10510–10522, 2025

    2025

  2. [2]

    X. Chen, Y . Chen, Y . Xiu, A. Geiger, and A. Chen. Ttt3r: 3d reconstruction as test-time training. arXiv preprint arXiv:2509.26645, 2025

    Pith/arXiv arXiv 2025

  3. [3]

    J. Dong, H. Li, S. Zhou, W. Hu, W. Xu, and Y . Wang. Memix: Writing less, remembering more for streaming 3d reconstruction.arXiv preprint arXiv:2603.15330, 2026

    arXiv 2026

  4. [4]

    Y . Lan, Y . Luo, F. Hong, S. Zhou, H. Chen, Z. Lyu, S. Yang, B. Dai, C. C. Loy, and X. Pan. Stream3r: Scalable sequential 3d reconstruction with causal transformer.arXiv preprint arXiv:2508.10893, 2025

    arXiv 2025

  5. [5]

    D. Zhuo, W. Zheng, J. Guo, Y . Wu, J. Zhou, and J. Lu. Streaming 4d visual geometry transformer. arXiv preprint arXiv:2507.11539, 2025

    Pith/arXiv arXiv 2025

  6. [6]

    Cheng, X

    C. Cheng, X. Chen, T. Xie, W. Yin, W. Ren, Q. Zhang, X. Guo, and H. Wang. Longstream: Long-sequence streaming autoregressive visual geometry.arXiv preprint arXiv:2602.13172, 2026

    arXiv 2026

  7. [7]

    Z. Li, J. Zhou, Y . Wang, H. Guo, W. Chang, Y . Zhou, H. Zhu, J. Chen, C. Shen, and T. He. Wint3r: Window-based streaming reconstruction with camera token pool.arXiv preprint arXiv:2509.05296, 2025

    arXiv 2025

  8. [8]

    C. Liu, J. Yang, Z. Li, Y . Deng, J. Guo, and L. Ballan. Mem3r: Streaming 3d reconstruction with hybrid memory via test-time training.arXiv preprint arXiv:2604.07279, 2026

    Pith/arXiv arXiv 2026

  9. [9]

    J. L. Schonberger and J.-M. Frahm. Structure-from-motion revisited. InProceedings of the IEEE conference on computer vision and pattern recognition, pages 4104–4113, 2016

    2016

  10. [10]

    Y . Ding, J. Yang, V . Larsson, C. Olsson, and K.˚Astr¨om. Revisiting the p3p problem. InIEEE Conference on Computer Vision and Pattern Recognition (CVPR), pages 4872–4880, 2023

    2023

  11. [11]

    Hartley and A

    R. Hartley and A. Zisserman.Multiple view geometry in computer vision. Cambridge university press, 2003

    2003

  12. [12]

    Triggs, P

    B. Triggs, P. F. McLauchlan, R. I. Hartley, and A. W. Fitzgibbon. Bundle adjustment—a modern synthesis. InInternational workshop on vision algorithms, pages 298–372. Springer, 1999

    1999

  13. [13]

    J. Ren, W. Liang, R. Yan, L. Mai, S. Liu, and X. Liu. Megba: A gpu-based distributed library for large-scale bundle adjustment. InEuropean Conference on Computer Vision, pages 715–731. Springer, 2022

    2022

  14. [14]

    Mur-Artal, J

    R. Mur-Artal, J. M. M. Montiel, and J. D. Tardos. Orb-slam: A versatile and accurate monocular slam system.IEEE transactions on robotics, 31(5):1147–1163, 2015

    2015

  15. [15]

    Mur-Artal and J

    R. Mur-Artal and J. D. Tard´os. Orb-slam2: An open-source slam system for monocular, stereo, and rgb-d cameras.IEEE transactions on robotics, 33(5):1255–1262, 2017

    2017

  16. [16]

    L. Pan, D. Bar´ath, M. Pollefeys, and J. L. Sch¨onberger. Global structure-from-motion revisited. InEuropean Conference on Computer Vision, pages 58–77. Springer, 2024

    2024

  17. [17]

    S. Zhu, R. Zhang, L. Zhou, T. Shen, T. Fang, P. Tan, and L. Quan. Very large-scale global sfm by distributed motion averaging. InProceedings of the IEEE conference on computer vision and pattern recognition, pages 4568–4577, 2018

    2018

  18. [18]

    Chatterjee and V

    A. Chatterjee and V . M. Govindu. Robust relative rotation averaging.IEEE transactions on pattern analysis and machine intelligence, 40(4):958–972, 2017. 9

    2017

  19. [19]

    Ozyesil and A

    O. Ozyesil and A. Singer. Robust camera location estimation by convex programming. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages 2674–2683, 2015

    2015

  20. [20]

    S. Wang, V . Leroy, Y . Cabon, B. Chidlovskii, and J. Revaud. Dust3r: Geometric 3d vision made easy. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 20697–20709, 2024

    2024

  21. [21]

    J. Wang, M. Chen, N. Karaev, A. Vedaldi, C. Rupprecht, and D. Novotny. Vggt: Visual geometry grounded transformer. InProceedings of the Computer Vision and Pattern Recognition Conference, pages 5294–5306, 2025

    2025

  22. [22]

    Y . Wang, J. Zhou, H. Zhu, W. Chang, Y . Zhou, Z. Li, J. Chen, J. Pang, C. Shen, and T. He.π3: Permutation-equivariant visual geometry learning.arXiv preprint arXiv:2507.13347, 2025

    Pith/arXiv arXiv 2025

  23. [23]

    H. Lin, S. Chen, J. Liew, D. Y . Chen, Z. Li, G. Shi, J. Feng, and B. Kang. Depth anything 3: Recovering the visual space from any views.arXiv preprint arXiv:2511.10647, 2025

    Pith/arXiv arXiv 2025

  24. [24]

    Keetha, N

    N. Keetha, N. M¨uller, J. Sch¨onberger, L. Porzi, Y . Zhang, T. Fischer, A. Knapitsch, D. Zauss, E. Weber, N. Antunes, et al. Mapanything: Universal feed-forward metric 3d reconstruction. arXiv preprint arXiv:2509.13414, 2025

    Pith/arXiv arXiv 2025

  25. [25]

    Elflein, R

    S. Elflein, R. Li, S. Agostinho, Z. Gojcic, L. Leal-Taix´e, Q. Zhou, and A. Osep. Vgg-t 3: Offline feed-forward 3d reconstruction at scale.arXiv preprint arXiv:2602.23361, 2026

    arXiv 2026

  26. [26]

    K. Deng, Z. Ti, J. Xu, J. Yang, and J. Xie. Vggt-long: Chunk it, loop it, align it–pushing vggt’s limits on kilometer-scale long rgb sequences.arXiv preprint arXiv:2507.16443, 2025

    Pith/arXiv arXiv 2025

  27. [27]

    H. Jin, R. Wu, T. Zhang, R. Gao, J. T. Barron, N. Snavely, and A. Holynski. Zipmap: Linear-time stateful 3d reconstruction via test-time training.arXiv preprint arXiv:2603.04385, 2026

    Pith/arXiv arXiv 2026

  28. [28]

    Zhang, C

    J. Zhang, C. Herrmann, J. Hur, C. Sun, M.-H. Yang, F. Cole, T. Darrell, and D. Sun. Loger: Long-context geometric reconstruction with hybrid memory.arXiv preprint arXiv:2603.03269, 2026

    Pith/arXiv arXiv 2026

  29. [29]

    T. Xie, P. Yang, Y . Jin, Y . Cai, W. Yin, W. Ren, Q. Zhang, W. Hua, S. Peng, X. Guo, et al. Scal3r: Scalable test-time training for large-scale 3d reconstruction.arXiv preprint arXiv:2604.08542, 2026

    Pith/arXiv arXiv 2026

  30. [30]

    Y . Wu, W. Zheng, J. Zhou, and J. Lu. Point3r: Streaming 3d reconstruction with explicit spatial pointer memory.arXiv preprint arXiv:2507.02863, 2025

    arXiv 2025

  31. [31]

    P. Neal, C. Eric, P. Borja, and E. Jonathan. Distributed optimization and statistical learning via the alternating direction method of multipliers.F oundations and Trends® in Machine learning, 3(1):1–122, 2011

    2011

  32. [32]

    H. Xia, Y . Fu, S. Liu, and X. Wang. Rgbd objects in the wild: Scaling real-world 3d object learning from rgb-d videos, 2024

    2024

  33. [33]

    A. Dai, A. X. Chang, M. Savva, M. Halber, T. Funkhouser, and M. Nießner. Scannet: Richly- annotated 3d reconstructions of indoor scenes. InProceedings of the IEEE conference on computer vision and pattern recognition, pages 5828–5839, 2017

    2017

  34. [34]

    Roberts, J

    M. Roberts, J. Ramapuram, A. Ranjan, A. Kumar, M. A. Bautista, N. Paczan, R. Webb, and J. M. Susskind. Hypersim: A photorealistic synthetic dataset for holistic indoor scene understanding. InProceedings of the IEEE/CVF international conference on computer vision, pages 10912–10922, 2021. 10

    2021

  35. [35]

    M. L. Antequera, P. Gargallo, M. Hofinger, S. R. Bulo, Y . Kuang, and P. Kontschieder. Mapillary planet-scale depth dataset. InEuropean Conference on Computer Vision, pages 589–604. Springer, 2020

    2020

  36. [36]

    Straub, T

    J. Straub, T. Whelan, L. Ma, Y . Chen, E. Wijmans, S. Green, J. J. Engel, R. Mur-Artal, C. Ren, S. Verma, et al. The replica dataset: A digital replica of indoor spaces.arXiv preprint arXiv:1906.05797, 2019

    Pith/arXiv arXiv 1906

  37. [37]

    Arnold, J

    E. Arnold, J. Wynn, S. Vicente, G. Garcia-Hernando, A. Monszpart, V . Prisacariu, D. Tur- mukhambetov, and E. Brachmann. Map-free visual relocalization: Metric pose relative to a single image. InEuropean Conference on Computer Vision, pages 690–708. Springer, 2022

    2022

  38. [38]

    W. Wang, D. Zhu, X. Wang, Y . Hu, Y . Qiu, C. Wang, Y . Hu, A. Kapoor, and S. Scherer. Tartanair: A dataset to push the limits of visual slam. In2020 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pages 4909–4916. IEEE, 2020

    2020

  39. [39]

    Huang, K

    P.-H. Huang, K. Matzen, J. Kopf, N. Ahuja, and J.-B. Huang. Deepmvs: Learning multi-view stereopsis. InProceedings of the IEEE conference on computer vision and pattern recognition, pages 2821–2830, 2018

    2018

  40. [40]

    Cabon, N

    Y . Cabon, N. Murray, and M. Humenberger. Virtual kitti 2, 2020

    2020

  41. [41]

    X. Pan, N. Charron, Y . Yang, S. Peters, T. Whelan, C. Kong, O. Parkhi, R. Newcombe, and Y . C. Ren. Aria digital twin: A new benchmark dataset for egocentric 3d machine perception. InProceedings of the IEEE/CVF International Conference on Computer Vision, pages 20133– 20143, 2023

    2023

  42. [42]

    L. Mehl, J. Schmalfuss, A. Jahedi, Y . Nalivayko, and A. Bruhn. Spring: A high-resolution high-detail dataset and benchmark for scene flow, optical flow and stereo. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 4981–4991, 2023

    2023

  43. [43]

    P. Sun, H. Kretzschmar, X. Dotiwalla, A. Chouard, V . Patnaik, P. Tsui, J. Guo, Y . Zhou, Y . Chai, B. Caine, et al. Scalability in perception for autonomous driving: Waymo open dataset. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 2446–2454, 2020

    2020

  44. [44]

    Y . Yao, Z. Luo, S. Li, J. Zhang, Y . Ren, L. Zhou, T. Fang, and L. Quan. Blendedmvs: A large-scale dataset for generalized multi-view stereo networks. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 1790–1799, 2020

    2020

  45. [45]

    Reizenstein, R

    J. Reizenstein, R. Shapovalov, P. Henzler, L. Sbordone, P. Labatut, and D. Novotny. Common objects in 3d: Large-scale learning and evaluation of real-life 3d category reconstruction. In Proceedings of the IEEE/CVF international conference on computer vision, pages 10901–10911, 2021

    2021

  46. [46]

    Li and N

    Z. Li and N. Snavely. Megadepth: Learning single-view depth prediction from internet photos. InProceedings of the IEEE conference on computer vision and pattern recognition, pages 2041–2050, 2018

    2041

  47. [47]

    L. Ling, Y . Sheng, Z. Tu, W. Zhao, C. Xin, K. Wan, L. Yu, Q. Guo, Z. Yu, Y . Lu, et al. Dl3dv-10k: A large-scale scene dataset for deep learning-based 3d vision. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 22160–22169, 2024

    2024

  48. [48]

    Geiger, P

    A. Geiger, P. Lenz, and R. Urtasun. Are we ready for autonomous driving? the kitti vision benchmark suite. InConference on Computer Vision and Pattern Recognition (CVPR), 2012

    2012

  49. [49]

    Brizi, E

    L. Brizi, E. Giacomini, L. Di Giammarino, S. Ferrari, O. Salem, L. De Rebotti, and G. Grisetti. Vbr: A vision benchmark in rome. In2024 IEEE International Conference on Robotics and Automation (ICRA), pages 15868–15874. IEEE, 2024. 11

    2024

  50. [50]

    Sturm, N

    J. Sturm, N. Engelhard, F. Endres, W. Burgard, and D. Cremers. A benchmark for the evaluation of rgb-d slam systems. In2012 IEEE/RSJ international conference on intelligent robots and systems, pages 573–580. IEEE, 2012

    2012

  51. [51]

    Tao, M.´A

    Y . Tao, M.´A. Mu˜noz-Ba˜n´on, L. Zhang, J. Wang, L. F. T. Fu, and M. Fallon. The oxford spires dataset: Benchmarking large-scale lidar-visual localisation, reconstruction and radiance field methods.International Journal of Robotics Research, 2025

    2025

  52. [52]

    Shotton, B

    J. Shotton, B. Glocker, C. Zach, S. Izadi, A. Criminisi, and A. Fitzgibbon. Scene coordinate regression forests for camera relocalization in rgb-d images. InProceedings of the IEEE conference on computer vision and pattern recognition, pages 2930–2937, 2013

    2013

  53. [53]

    Y . Shen, Z. Zhang, Y . Qu, X. Zheng, J. Ji, S. Zhang, and L. Cao. Fastvggt: Training-free acceleration of visual geometry transformer.arXiv preprint arXiv:2509.02560, 2025

    Pith/arXiv arXiv 2025

  54. [54]

    Murai, E

    R. Murai, E. Dexheimer, and A. J. Davison. Mast3r-slam: Real-time dense slam with 3d recon- struction priors. InProceedings of the Computer Vision and Pattern Recognition Conference, pages 16695–16705, 2025

    2025

  55. [55]

    Maggio, H

    D. Maggio, H. Lim, and L. Carlone. Vggt-slam: Dense rgb slam optimized on the sl (4) manifold.arXiv preprint arXiv:2505.12549, 2025

    Pith/arXiv arXiv 2025

  56. [56]

    S. Yuan, Y . Yang, X. Yang, X. Zhang, Z. Zhao, L. Zhang, and Z. Zhang. Infinitevggt: Visual geometry grounded transformer for endless streams.arXiv preprint arXiv:2601.02281, 2026

    arXiv 2026

  57. [57]

    Arandjelovic, P

    R. Arandjelovic, P. Gronat, A. Torii, T. Pajdla, and J. Sivic. Netvlad: Cnn architecture for weakly supervised place recognition. InProceedings of the IEEE conference on computer vision and pattern recognition, pages 5297–5307, 2016. 12 Supplementary Material A Online Motion Averaging During online inference, Anchor3R does not compose poses along a single...

    2016