Pith. sign in

REVIEW 4 major objections 5 minor 36 references

Multi-Submap Implicit Neural SLAM with Local-to-Global Loop Closure for Large-Scale Scene Reconstruction

T0 review · 4 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read A multi-submap implicit neural SLAM system with local-to-global loop closure and inter-submap online distillation reconstructs city-scale scenes while running at 10 FPS on an embedded GPU.

desk verdict A sensible recombination of known modules into a large-scale neural SLAM system, with a genuinely interesting inter-submap distillation idea, but the SOTA claim is unverifiable because the direct predecessor PLGSLAM is missing from every comparison and the numbers have unit inconsistencies. read the letter →

arxiv 2608.09146 v1 pith:2NPO7FMW submitted 2026-08-10 cs.CV

Tianchen Deng , Chongdi Wang , Nailin Wang , Lei Zhao , Ziqi Ma , Tianjun Zhang , Zhe Liu , Danwei Wang
show 1 more author
Hesheng Wang
This is my paper · ORCID
classification cs.CV
keywords neuralSLAMmulti-submaprepresentationloopclosureimplicitscenelarge-scalereconstructionopticalflowtrackingonlinedistillationvisualplacerecognition
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper sets out to establish that neural-radiance-field SLAM can be pushed from room-scale demonstrations to city-scale mapping if the scene is partitioned into a dynamically allocated collection of local neural submaps, each a hybrid tri-plane plus MLP volume, instead of one monolithic network. It argues that three additions make this scalable: optical-flow-based tracking from a dense bundle adjustment layer, loop closure via foundation-model global descriptors that fire under strong viewpoint change, and online distillation that keeps overlapping submap boundaries geometrically and photometrically consistent. A reader should care because these are exactly the mechanisms that have kept NeRF-SLAM in small scenes: memory growth, catastrophic forgetting, and accumulated drift. On KITTI-scale trajectories of about 500 m by 400 m and on a custom handheld platform, the system reports better reconstruction and localization than the compared neural SLAM baselines while running at 10 FPS on a Jetson AGX Orin. The paper's specific claim is that this combination is a practical path to real-world robotic perception and digital twinning.

What carries the argument

The load-bearing machinery is the collection of local neural submaps: each submap is a hybrid representation made of three axis-aligned feature planes (tri-planes) and a small MLP decoder, whose memory cost scales as O($N^{2}$) instead of O($N^{3}$). Around this core sit four mechanisms: a contraction function that warps unbounded outdoor space into a bounded cube; a differentiable dense bundle adjustment layer built on recurrent optical flow for tracking; a global keyframe database indexed by SALAD descriptors for loop closure; and an inter-submap distillation loss that supervises younger submaps from older ones in overlap regions. Together they let the system keep a bounded-memory map while a global pose graph is refined only when loop constraints are added.

What would settle it

Take the pipeline through a long corridor with plain walls and no texture where monocular depth estimators typically flatten or scale-drag the geometry, then compare ATE and reconstruction F-score against the same pipeline using ground-truth depth or LiDAR-initialized depth; if the results converge, the depth prior is not load-bearing, and if they diverge, the central premise fails.

Watch

Extended reading notes

Core claim

The central discovery is that a complete large-scale neural SLAM system can be assembled from four components that each remove a bottleneck. Progressive multi-submap allocation prevents memory blow-up and catastrophic forgetting; a recurrent optical-flow dense bundle adjustment thread gives stable poses under aggressive motion; local-to-global loop closure with foundation-model descriptors recovers long-range constraints that geometric flow misses; and inter-submap online distillation, where the more mature submap acts as teacher for the newer one, eliminates seam artifacts. The paper demonstrates on Tanks and Temples, Static Hikes, KITTI, and its own recorded sequences that this pipeline yields the highest PSNR/SSIM/LPIPS among compared methods and the lowest average ATE, and that it fits in 18 GB memory at 10 FPS on an embedded GPU.

Load-bearing premise

The system assumes the ViT-based monocular depth network gives reliable scale-invariant depth priors in every environment it is run in, so that the normalized depth and optical-flow losses inherit trustworthy geometry.

Editorial extensions

If this is right

  • Neural SLAM memory stops growing linearly with path length: adding a new submap freezes the old one, so an hour-long trajectory can be mapped in bounded GPU memory.
  • Loop closure no longer depends on geometric feature matching; foundation-model descriptors make relocalization possible across large viewpoint changes, which is the usual failure point in long urban sequences.
  • Submap boundaries no longer show seams because the distillation loss pulls overlapping volumes into agreement, so the global map is a single consistent representation.
  • The whole pipeline runs at 10 FPS on an embedded GPU, meaning this can operate on a robot or handheld device rather than a desktop workstation.
  • On KITTI-scale trajectories of about 500 m by 400 m, the reported ATE is lower than the compared monocular NeRF-SLAM baselines, indicating that the approach suppresses cumulative drift at city scale.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • A natural extension would be to replace the fixed SALAD similarity threshold with an adaptive or learned threshold, since a wrong threshold either floods the keyframe graph with redundant edges or misses true loops in environments with repetitive structure.
  • The teacher-student distillation could be extended from pairwise to multi-submap consensus, averaging constraints across all overlapping submaps, which would likely further smooth boundaries in dense urban canyons where more than two submaps overlap.
  • Because the platform carries a LiDAR that is used only for reference geometry, a testable extension is to feed sparse LiDAR points directly into the submap depth supervision, potentially removing the monocular-depth prior altogether.
  • The paper's stated sensitivity to illumination changes and dynamic objects implies that the monocular depth prior is the practical ceiling in real deployment; a robustness-focused follow-up would need to gate the depth and flow losses when the depth network's uncertainty is high.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. The paper proposes MSN-SLAM, a NeRF-based SLAM system that combines a progressive multi-submap tri-plane representation, optical-flow-based tracking with a DBA layer, SALAD-based local-to-global loop closure, and an inter-submap online distillation loss. The system is validated on Tanks and Temples, Static Hikes, KITTI, and a custom handheld platform with a Jetson AGX Orin, reporting state-of-the-art reconstruction and localization accuracy, including city-scale trajectories up to 500 m by 400 m at 10 FPS. The paper also presents ablations of each proposed module and discusses limitations such as dynamic objects and illumination sensitivity.

Significance. If the claims are substantiated, MSN-SLAM would be a notable advance toward scalable neural SLAM in large-scale environments, with concrete engineering contributions including a synchronized handheld sensing platform, a submap-distillation mechanism, and a foundation-model loop-closure module. The paper is transparent about its limitations and promises code release. However, the significance is currently conditional: the central SOTA claim is not verifiable because the direct predecessor PLGSLAM is absent from all comparisons, and there are inconsistencies in the reported units and ablation numbers that undermine confidence in the quantitative claims.

major comments (4)
  1. [Section II-C and Tables I-IV, VI] The paper explicitly states in Section II-C that it builds upon PLGSLAM's progressive local scene representation and local-to-global optimization paradigm, but PLGSLAM is not included in any of the experimental tables. Because PLGSLAM is the direct predecessor and the novel contributions are framed as extensions to it, every comparison table should include a PLGSLAM row under the same protocol. Without this baseline, the reported margins (e.g., Table IV mean ATE 0.241 m versus DROID-SLAM 0.357 m) cannot be attributed to the proposed system, and the abstract's claim of outperforming state-of-the-art neural SLAM methods is unsupported.
  2. [Tables III, IV, VII and Section IV-E] There are serious unit inconsistencies across the pose evaluation. Section IV-E states that ATE RMSE is reported in centimeters, but Table IV is headed 'ATE RMSE [M]' and contains values such as 0.241 m, which are orders of magnitude larger than the centimeter-scale values in Table III. Table VII is headed 'ATE [m]' but contains values like 20.35 and 41.82, which the text explicitly interprets as centimeters. These inconsistencies make it impossible to interpret the ablation results and cross-comparisons reliably, and they must be corrected and resubmitted with consistent units throughout.
  3. [Section III-A, Eqs. (10) and (12)] The monocular depth network is never identified. The depth prior feeds the normalized depth loss (Eq. 10) and the optical flow loss (Eq. 12), so its reliability is load-bearing for both mapping and tracking. The paper should state which ViT-based network is used, its training data, and how scale/shift normalization interacts with the estimated depth errors. At minimum, the authors should provide sensitivity experiments or failure-case analysis showing the effect of depth-prior noise on ATE and PSNR, especially for low-texture corridors and outdoor KITTI-style images.
  4. [Section III-B and III-D, Eq. (13)] The inter-submap distillation assumes that the earlier submap acts as a frozen teacher whenever a new submap is created, since Section III-B states that the parameters of each submap are frozen once it is transferred to the global optimization thread. If the teacher is frozen, any bias in the teacher is propagated irreversibly to the student; if teachers are later refined, the teacher-student assignment and the loss definition need to be reconsidered. The paper should clarify whether teacher submaps are ever updated after freezing and how this affects the reported reconstruction gains.
minor comments (5)
  1. [Figures 3 and 4] The PSNR values printed in the figure captions and images do not match the values in Tables I and II; for example, Fig. 3 shows several PSNR entries that differ from the corresponding table entries. The authors should align the figure annotations with the tables.
  2. [References] References [32] and [33] are duplicates of [23] and [24] (NoPe-NeRF and LocalRF). This should be fixed with a single citation per work in the bibliography.
  3. [Table V] The 'Length [m]' column entries such as '25.46×8' appear to be dimensions or area rather than a trajectory length, and the column header should be clarified or split into 'Size [m]' and 'Traversed Length [m]'.
  4. [Section III-C] The phrase 'SALAD descriptor [30], a foundation-model-based visual descriptor from DINO v2 [31]' is slightly inaccurate: SALAD is a global descriptor aggregation method trained on features from DINOv2, not itself a foundation model. The wording should be adjusted.
  5. [Section VI] The paper promises code release at a GitHub URL, but the repository is not currently available. The authors should either provide an anonymous link during review or state explicitly that the code will be released upon acceptance.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation found; the reported pose and reconstruction results are benchmarked against external references, though the SOTA comparison is weakened by the absence of the direct predecessor PLGSLAM as a baseline.

full rationale

The paper's derivation chain is self-contained: pose tracking uses a differentiable dense bundle adjustment layer (Eq. 8) with RAFT optical flow, mapping is supervised by photometric loss, normalized monocular depth loss (Eq. 10), and optical flow loss (Eq. 12), and loop closure uses the external SALAD/DINOv2 descriptor. None of these training objectives is later reported as a prediction metric; evaluation metrics (PSNR/SSIM/LPIPS, ATE/RPE) are computed against ground-truth images, COLMAP references, or standard benchmarks. The self-citation of PLGSLAM [28] in Section II-C describes an architectural starting point, not a result used to prove the current claims, so it is not load-bearing circularity. The absence of PLGSLAM from Tables I-VII and the unit inconsistencies in Tables III-IV/VII are correctness and comparison weaknesses, but they do not make any predicted quantity equivalent to a fitted input by construction. Therefore no circular step meeting the required evidentiary standard is present.

Assumptions & free parameters 6 free parameters · 6 assumptions · 0 invented entities

The numbers the system depends on are mostly thresholds and pretrained models. There are six hand-set parameters, the most consequential being the submap trigger thresholds and the loop closure threshold; no sensitivity analysis is provided. The main external assumptions are the monocular depth prior, the SALAD descriptor, and the DROID-SLAM flow and BA backbone. No new physical entities are introduced.

free parameters (6)
  • submap creation distance threshold delta_dist = 2.0 m
    Trigger in Eq. (5); controls submap size, memory, and overlapping volumes; hand-set in Section IV-A.
  • submap visibility threshold delta_vis = 0.6
    Second trigger in Eq. (5); determines when a new submap is created; hand-set.
  • keyframe flow threshold tau_flow = 2.5 px
    New keyframe creation in Section III-C; affects graph density and tracking.
  • loop closure similarity threshold tau_co = 0.75
    SALAD descriptor matching threshold; controls recall and precision of loop detections in Section IV-A.
  • distillation weights omega_g and omega_c = not reported
    Eq. (13) balances geometry and color consistency; values are not stated in the paper.
  • local sliding window size N_local = 15
    Local BA window size, hand-set in Section IV-A.
assumptions (6)
  • standard math NeRF volume rendering and differentiable ray sampling provide valid geometry and color supervision.
    Equations (6) and (9) follow the standard NeRF formulation; this is accepted background.
  • domain assumption ViT-based monocular depth network supplies dependable scale/shift-invariant depth priors for all tested scenes.
    Invoked in Section III-A; Eq. (10) normalizes depth, but if the prior is wrong in low-texture or outdoor regions the mapping objective is biased.
  • domain assumption SALAD/DINOv2 descriptors remain discriminative under drastic viewpoint and appearance changes.
    Loop closure relies on descriptor similarity with threshold 0.75; no precision or recall verification is reported (Section III-C and Section IV-A).
  • domain assumption Dense bundle adjustment from DROID-SLAM has accurate optical flow and depth initialization in city-scale scenes.
    Tracking backbone is taken from DROID-SLAM; its error model and flow predictions are treated as reliable (Section III-C).
  • domain assumption Freezing a submap's parameters prevents catastrophic forgetting without causing inconsistency at boundaries.
    Progressive submap strategy freezes earlier submaps while adding new ones (Section III-B); boundary consistency is then repaired by distillation.
  • domain assumption The scene is predominantly static.
    Acknowledged limitation in the Conclusion; moving objects would corrupt depth and photometric losses.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Multi-Submap Implicit Neural SLAM with Local-to-Global Loop Closure for Large-Scale Scene Reconstruction." pith.science (2026). https://pith.science/paper/2NPO7FMW

@misc{pith2026260809146,
  author       = {Pith},
  title        = {Pith review of: Multi-Submap Implicit Neural SLAM with Local-to-Global Loop Closure for Large-Scale Scene Reconstruction},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2NPO7FMW}},
  note         = {Machine review of arXiv:2608.09146}
}
read the original abstract

Neural Radiance Fields (NeRF)-based SLAM has demonstrated impressive results in small-scale scene reconstruction, yet scaling these methods to extensive, complex environments remains challenging due to catastrophic forgetting and accumulated trajectory drift. This paper presents a robust, large-scale neural SLAM system featuring a multi-submap architecture and a dual-tier loop closure mechanism. Specifically, we propose a progressive mapping strategy that dynamically allocates neural submaps to maintain high-fidelity representations without memory explosion. For robust pose estimation, an optical-flow-based tracking module is integrated to handle aggressive motions. To address global consistency, we introduce a local-to-global loop closure framework leveraging the foundation model for high-performance global descriptor extraction, significantly enhancing relocalization accuracy under varying viewpoints. Furthermore, an inter-submap online distillation algorithm is designed during back-end optimization to enforce geometric and appearance consistency across overlapping submap boundaries. To validate the system, we developed a customized handheld mechatronic platform and conducted extensive evaluations on both public benchmarks and our large-scale indoor-outdoor datasets. Experimental results, including direct deployment on an onboard computing unit, demonstrate that our approach outperforms state-of-the-art neural SLAM methods in reconstruction quality and localization robustness, providing a scalable solution for real-world robotic perception and digital twinning. We will release the code publicly on \href{https://github.com/dtc111111/MSN-SLAM}{https://github.com/dtc111111/MSN-SLAM} .

Figures

Figures reproduced from arXiv: 2608.09146 by the authors.

Figure 1
Figure 1. Overview of the proposed system pipeline. The framework consists of three core components: (a) Progressive multi-submap representation, which [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Overview of the proposed handheld device: (a) CAD model showing [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Visual quality comparison on Tanks and Temples sequences. Compared to SOTA methods including NoPe-NeRF [32] and LocalRF [33], our method [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Visual quality comparison on the city-scale KITTI dataset. Compared to SOTA methods including NoPe-NeRF and LocalRF, our method provides [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]
Figure 5
Figure 5. Figure 5: Reconstruction results on the Static Hikes dataset [33] . Our Progressive Multi-Submap strategy enables the efficient management of memory [PITH_FULL_IMAGE:figures/full_fig_p009_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

36 extracted references · 20 canonical work pages

  1. [1]

    What is the best 3d scene representation for robotics? from geometric to foundation models,

    T. Deng, Y . Pan, S. Yuan, D. Li, C. Wang, M. Li, L. Chen, L. Xie, D. Wang, J. Wang, J. Civera, H. Wang, and W. Chen, “What is the best 3d scene representation for robotics? from geometric to foundation models,”arXiv preprint arXiv:2512.03422, 2025

    arXiv 2025

  2. [2]

    Mpoc-slam: An rgb-d slam system with motion probability and object category in high dynamic environments,

    S. Wu, X. Zhang, S. Zhang, Z. Song, R. Wang, and J. Yuan, “Mpoc-slam: An rgb-d slam system with motion probability and object category in high dynamic environments,”IEEE/ASME Transactions on Mechatron- ics, vol. 30, no. 2, pp. 1061–1071, 2025

    work page 2025

  3. [3]

    Pmo-slam: Enhancing incremental optimization with probabilistic metrics for object slam,

    X. Lin, S. Ji, Z. Liao, L. He, X. Zhou, and H. Zhang, “Pmo-slam: Enhancing incremental optimization with probabilistic metrics for object slam,”IEEE/ASME Transactions on Mechatronics, pp. 1–11, 2025

    work page 2025

  4. [4]

    Orb-slam: A versatile and accurate monocular slam system,

    R. Mur-Artal, J. M. M. Montiel, and J. D. Tard ´os, “Orb-slam: A versatile and accurate monocular slam system,”IEEE Transactions on Robotics, vol. 31, no. 5, pp. 1147–1163, 2015

    2015

  5. [5]

    Vins-mono: A robust and versatile monocular visual-inertial state estimator,

    T. Qin, P. Li, and S. Shen, “Vins-mono: A robust and versatile monocular visual-inertial state estimator,”IEEE Transactions on Robotics, vol. 34, no. 4, pp. 1004–1020, 2018

    2018

  6. [6]

    Dtam: Dense tracking and mapping in real-time,

    R. A. Newcombe, S. J. Lovegrove, and A. J. Davison, “Dtam: Dense tracking and mapping in real-time,” in2011 international conference on computer vision. IEEE, 2011, pp. 2320–2327

    2011

  7. [7]

    Kinectfusion: real-time 3d reconstruction and interaction using a moving depth cam- era,

    S. Izadi, D. Kim, O. Hilliges, D. Molyneaux, R. Newcombe, P. Kohli, J. Shotton, S. Hodges, D. Freeman, A. Davisonet al., “Kinectfusion: real-time 3d reconstruction and interaction using a moving depth cam- era,” inProceedings of the 24th annual ACM symposium on User interface software and technology, 2011, pp. 559–568

    2011

  8. [8]

    imap: Implicit mapping and positioning in real-time,

    E. Sucar, S. Liu, J. Ortiz, and A. J. Davison, “imap: Implicit mapping and positioning in real-time,” inICCV, October 2021, pp. 6229–6238

    work page 2021

Show all 36 references
  1. [9]

    Nice-slam: Neural implicit scalable encoding for slam,

    Z. Zhu, S. Peng, V . Larsson, W. Xu, H. Bao, Z. Cui, M. R. Oswald, and M. Pollefeys, “Nice-slam: Neural implicit scalable encoding for slam,” inCVPR, June 2022, pp. 12 786–12 796

    2022

  2. [10]

    Co-slam: Joint coordinate and sparse parametric encodings for neural real-time slam,

    H. Wang, J. Wang, and L. Agapito, “Co-slam: Joint coordinate and sparse parametric encodings for neural real-time slam,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023, pp. 13 293–13 302

    2023

  3. [11]

    Eslam: Efficient dense slam system based on hybrid representation of signed distance fields,

    M. M. Johari, C. Carta, and F. Fleuret, “Eslam: Efficient dense slam system based on hybrid representation of signed distance fields,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023, pp. 17 408–17 419

    2023

  4. [12]

    Orb-slam3: An accurate open-source library for visual, visual–inertial, and multimap slam,

    C. Campos, R. Elvira, J. J. G. Rodr ´ıguez, J. M. M. Montiel, and J. D. Tard ´os, “Orb-slam3: An accurate open-source library for visual, visual–inertial, and multimap slam,”IEEE Transactions on Robotics, vol. 37, no. 6, pp. 1874–1890, 2021

    2021

  5. [13]

    Long-term visual simultaneous localization and mapping: Using a bayesian persistence filter-based global map prediction,

    T. Deng, H. Xie, J. Wang, and W. Chen, “Long-term visual simultaneous localization and mapping: Using a bayesian persistence filter-based global map prediction,”IEEE Robotics & Automation Magazine, vol. 30, no. 1, pp. 36–49, 2023

    2023

  6. [14]

    Robust incremental long- term visual topological localization in changing environments,

    H. Xie, T. Deng, J. Wang, and W. Chen, “Robust incremental long- term visual topological localization in changing environments,”IEEE Transactions on Instrumentation and Measurement, vol. 72, pp. 1–14, 2022

    2022

  7. [15]

    Elasticfusion: Real-time dense slam and light source estimation,

    T. Whelan, R. F. Salas-Moreno, B. Glocker, A. J. Davison, and S. Leutenegger, “Elasticfusion: Real-time dense slam and light source estimation,”The International Journal of Robotics Research, vol. 35, no. 14, pp. 1697–1716, 2016

    2016

  8. [16]

    Codeslam — learning a compact, optimisable representation for dense visual slam,

    M. Bloesch, J. Czarnowski, R. Clark, S. Leutenegger, and A. J. Davison, “Codeslam — learning a compact, optimisable representation for dense visual slam,” inProceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), June 2018

    2018

  9. [17]

    Droid-slam: Deep visual slam for monocular, stereo, and rgb-d cameras,

    Z. Teed and J. Deng, “Droid-slam: Deep visual slam for monocular, stereo, and rgb-d cameras,”Advances in neural information processing systems, vol. 34, pp. 16 558–16 569, 2021

    2021

  10. [18]

    Nerf: Representing scenes as neural radiance fields for view synthesis,

    B. Mildenhall, P. P. Srinivasan, M. Tancik, J. T. Barron, R. Ramamoorthi, and R. Ng, “Nerf: Representing scenes as neural radiance fields for view synthesis,” inECCV, 2020

    2020

  11. [19]

    Instant neural graphics primitives with a multiresolution hash encoding,

    T. M ¨uller, A. Evans, C. Schied, and A. Keller, “Instant neural graphics primitives with a multiresolution hash encoding,”ACM transactions on graphics (TOG), vol. 41, no. 4, pp. 1–15, 2022

    2022

  12. [20]

    Direct voxel grid optimization: Super- fast convergence for radiance fields reconstruction,

    C. Sun, M. Sun, and H.-T. Chen, “Direct voxel grid optimization: Super- fast convergence for radiance fields reconstruction,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2022, pp. 5459–5469

    2022

  13. [21]

    Tensorf: Tensorial radiance fields,

    A. Chen, Z. Xu, A. Geiger, J. Yu, and H. Su, “Tensorf: Tensorial radiance fields,” inEuropean conference on computer vision. Springer, 2022, pp. 333–350

    2022

  14. [22]

    Barf: Bundle-adjusting neural radiance fields,

    C.-H. Lin, W.-C. Ma, A. Torralba, and S. Lucey, “Barf: Bundle-adjusting neural radiance fields,” inProceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), October 2021, pp. 5741–5751

    2021

  15. [23]

    Nope-nerf: Optimising neural radiance field with no pose prior,

    W. Bian, Z. Wang, K. Li, J.-W. Bian, and V . A. Prisacariu, “Nope-nerf: Optimising neural radiance field with no pose prior,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2023, pp. 4160–4169

    2023

  16. [25]

    Prosgnerf: Progressive dynamic neural scene graph with frequency modulated foundation model in urban scenes,

    T. Deng, Y . Wang, Y . Liu, C. Su, J. Wang, H. Wang, D. Wang, S.-Y . Lo, and W. Chen, “Prosgnerf: Progressive dynamic neural scene graph with frequency modulated foundation model in urban scenes,”International Journal of Computer Vision, vol. 134, no. 8, p. 365, 2026

    2026

  17. [26]

    Incremental joint learning of depth, pose and implicit scene representation on monocular camera in large-scale scenes,

    T. Deng, N. Wang, C. Wang, S. Yuan, J. Wang, H. Wang, D. Wang, and W. Chen, “Incremental joint learning of depth, pose and implicit scene representation on monocular camera in large-scale scenes,”IEEE Transactions on Automation Science and Engineering, pp. 1–1, 2025

    2025

  18. [27]

    Go-slam: Global optimization for consistent 3d instant reconstruction,

    Y . Zhang, F. Tosi, S. Mattoccia, and M. Poggi, “Go-slam: Global optimization for consistent 3d instant reconstruction,” inProceedings of the IEEE/CVF International Conference on Computer Vision, 2023, pp. 3727–3737

    2023

  19. [28]

    Plgslam: Progressive neural scene represenation with local to global bundle adjustment,

    T. Deng, G. Shen, T. Qin, J. Wang, W. Zhao, J. Wang, D. Wang, and W. Chen, “Plgslam: Progressive neural scene represenation with local to global bundle adjustment,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), June 2024, pp. 19 657–19 666

    2024

  20. [29]

    Raft: Recurrent all-pairs field transforms for op- tical flow,

    Z. Teed and J. Deng, “Raft: Recurrent all-pairs field transforms for op- tical flow,” inComputer Vision–ECCV 2020: 16th European Conference, Glasgow, UK, August 23–28, 2020, Proceedings, Part II 16. Springer, 2020, pp. 402–419

    2020

  21. [30]

    Optimal transport aggregation for visual place recognition,

    S. Izquierdo and J. Civera, “Optimal transport aggregation for visual place recognition,” inProceedings of the ieee/cvf conference on com- puter vision and pattern recognition, 2024, pp. 17 658–17 668

    2024

  22. [31]

    Dinov2: Learning robust visual features without supervision,

    M. Oquab, T. Darcet, T. Moutakanni, H. V o, M. Szafraniec, V . Khalidov, P. Fernandez, D. Haziza, F. Massa, A. El-Noubyet al., “Dinov2: Learning robust visual features without supervision,”arXiv preprint arXiv:2304.07193, 2023

    2023 arXiv

  23. [32]

    Nope-nerf: Optimising neural radiance field with no pose prior,

    W. Bian, Z. Wang, K. Li, J.-W. Bian, and V . A. Prisacariu, “Nope-nerf: Optimising neural radiance field with no pose prior,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023, pp. 4160–4169

    2023

  24. [33]

    Progressively optimized local radiance fields for robust view synthesis,

    A. Meuleman, Y .-L. Liu, C. Gao, J.-B. Huang, C. Kim, M. H. Kim, and J. Kopf, “Progressively optimized local radiance fields for robust view synthesis,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023, pp. 16 539–16 548

    2023

  25. [34]

    Flow-nerf: Joint learning of geometry, poses, and dense flow within unified neural representations,

    X. Zheng and D. Xu, “Flow-nerf: Joint learning of geometry, poses, and dense flow within unified neural representations,” inProceedings JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2021 11 of the Computer Vision and Pattern Recognition Conference, 2025, pp. 993–1002

    2021

  26. [35]

    Vision meets robotics: The kitti dataset,

    A. Geiger, P. Lenz, C. Stiller, and R. Urtasun, “Vision meets robotics: The kitti dataset,”International Journal of Robotics Research (IJRR), 2013

    2013

  27. [36]

    Fast-livo2: Fast, direct lidar–inertial–visual odometry,

    C. Zheng, W. Xu, Z. Zou, T. Hua, C. Yuan, D. He, B. Zhou, Z. Liu, J. Lin, F. Zhuet al., “Fast-livo2: Fast, direct lidar–inertial–visual odometry,”IEEE Transactions on Robotics, vol. 41, pp. 326–346, 2024

    2024

  28. [37]

    Tanks and temples: Benchmarking large-scale scene reconstruction,

    A. Knapitsch, J. Park, Q.-Y . Zhou, and V . Koltun, “Tanks and temples: Benchmarking large-scale scene reconstruction,”ACM Transactions on Graphics (ToG), vol. 36, no. 4, pp. 1–13, 2017

    2017

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.