arxiv: 2604.02696 · v1 · submitted 2026-04-03 · 💻 cs.CV · cs.RO

Recognition: 2 theorem links

· Lean Theorem

VBGS-SLAM: Variational Bayesian Gaussian Splatting Simultaneous Localization and Mapping

Jie Xu, Wei Ren, Yanyu Zhang, Yuhan Zhu

Authors on Pith no claims yet

Pith reviewed 2026-05-13 20:53 UTC · model grok-4.3

classification 💻 cs.CV cs.RO

keywords 3D Gaussian SplattingSLAMVariational InferenceUncertainty EstimationPose TrackingScene ReconstructionNovel View Synthesis

0 comments

The pith

VBGS-SLAM uses variational Bayesian inference to maintain explicit uncertainty over both camera poses and 3D Gaussian scene parameters.

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

The paper presents VBGS-SLAM, which reformulates 3D Gaussian Splatting simultaneous localization and mapping as a generative probabilistic problem. It applies variational inference to couple the refinement of the splat map with camera pose tracking. Using conjugate properties of multivariate Gaussians, the approach obtains efficient closed-form updates for the posteriors. This explicit uncertainty representation helps mitigate drift and boosts robustness in challenging scenarios while keeping the original efficiency and rendering quality. Experiments confirm better performance in tracking and novel view synthesis on synthetic and real scenes.

Core claim

By modeling 3DGS SLAM in a generative probabilistic form and leveraging conjugate Gaussian properties for variational inference, the method enables closed-form updates that explicitly maintain posterior uncertainty over both camera poses and scene parameters, thereby reducing drift and improving robustness without compromising efficiency or rendering quality.

What carries the argument

Variational inference on a joint generative model of poses and 3D Gaussian splats, using multivariate Gaussian conjugacy to derive closed-form posterior updates.

If this is right

Uncertainty-aware optimization reduces drift in long sequence tracking.
Robustness improves in challenging conditions such as low texture or fast motion.
Efficiency and high-quality novel view synthesis are preserved from standard 3DGS.
Explicit posterior uncertainty is available for both poses and scene parameters.

Where Pith is reading between the lines

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

Uncertainty estimates could inform selective updates or loop closure decisions in extended mapping systems.
The closed-form structure may extend to other mixture-based scene models in probabilistic SLAM.
Pose uncertainty could enhance fusion with inertial or other sensor data.
Validation against datasets providing ground-truth uncertainty would further test the robustness gains.

Load-bearing premise

The joint distribution over camera poses and Gaussian scene parameters admits a generative model where variational inference yields sufficiently accurate closed-form updates without introducing large errors in the uncertainty estimates.

What would settle it

Observing on ground-truth datasets that the estimated uncertainties fail to correlate with actual pose errors, or that the system does not outperform standard 3DGS SLAM in drift-heavy sequences.

Figures

Figures reproduced from arXiv: 2604.02696 by Jie Xu, Wei Ren, Yanyu Zhang, Yuhan Zhu.

**Figure 2.** Figure 2: Qualitative Comparison of Rendering image to ground truth. The top row displays results on AR-TABLE dataset, [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗

read the original abstract

3D Gaussian Splatting (3DGS) has shown promising results for 3D scene modeling using mixtures of Gaussians, yet its existing simultaneous localization and mapping (SLAM) variants typically rely on direct, deterministic pose optimization against the splat map, making them sensitive to initialization and susceptible to catastrophic forgetting as map evolves. We propose Variational Bayesian Gaussian Splatting SLAM (VBGS-SLAM), a novel framework that couples the splat map refinement and camera pose tracking in a generative probabilistic form. By leveraging conjugate properties of multivariate Gaussians and variational inference, our method admits efficient closed-form updates and explicitly maintains posterior uncertainty over both poses and scene parameters. This uncertainty-aware method mitigates drift and enhances robustness in challenging conditions, while preserving the efficiency and rendering quality of existing 3DGS. Our experiments demonstrate superior tracking performance and robustness in long sequence prediction, alongside efficient, high-quality novel view synthesis across diverse synthetic and real-world scenes.

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.

Desk Editor's Note private letter to a colleague

VBGS-SLAM adds a variational Bayesian layer to 3DGS-SLAM for explicit pose and map uncertainty, but the closed-form conjugacy claim is likely undercut by the non-linear rendering steps.

read the letter

The paper's main move is to recast 3D Gaussian Splatting SLAM as a joint generative model over poses and scene parameters, then apply variational inference with conjugate Gaussian updates so that posterior uncertainty is carried explicitly and updated in closed form. This is new relative to the deterministic pose optimization used in earlier 3DGS-SLAM systems, and it directly targets drift and initialization sensitivity in long sequences. The abstract positions the method as keeping the original rendering speed and quality while gaining robustness, which is a reasonable goal for robotics and AR use cases. Experiments are claimed to show better tracking on both synthetic and real data, so the practical payoff is at least asserted. The soft spot is the conjugacy assumption itself. Standard 3DGS rendering applies perspective projection, covariance rotation and scaling, and alpha-blended volume rendering; none of these operations leave the joint distribution inside the conjugate multivariate Gaussian family without additional linearization or mean-field steps. If those steps are present, the uncertainty estimates are no longer guaranteed to be accurate, which undercuts the drift-mitigation story. The abstract gives no derivations or error analysis, so it is impossible to tell how large the approximation error actually is. This work is for people already building or extending 3DGS-SLAM pipelines who want a probabilistic variant. A reader focused on uncertainty-aware mapping would find the framing useful even if the math needs tightening. It is worth sending to peer review so the update equations and any approximation details can be checked against the claimed performance gains.

Referee Report

3 major / 3 minor

Summary. The paper proposes VBGS-SLAM, a variational Bayesian framework for 3D Gaussian Splatting SLAM. It formulates joint inference over camera poses and scene Gaussian parameters as a generative probabilistic model, leveraging conjugate multivariate Gaussian properties and variational inference to obtain efficient closed-form posterior updates. The method explicitly tracks uncertainty in both poses and the map to mitigate drift and improve robustness in challenging conditions, while claiming to preserve the rendering speed and quality of standard 3DGS. Experiments on synthetic and real-world sequences report superior tracking accuracy and long-sequence stability compared to prior deterministic 3DGS-SLAM baselines.

Significance. If the claimed closed-form variational updates hold with accurate uncertainty quantification, the work would provide a principled Bayesian treatment of 3DGS-SLAM that addresses a clear limitation of existing deterministic optimization approaches. The ability to maintain and exploit posterior uncertainty for drift reduction could influence future robust mapping systems, particularly in long-term or low-texture scenarios, while retaining the efficiency advantages of Gaussian splatting.

major comments (3)

[§3.2] §3.2, Eq. (7)–(9): The derivation asserts that the variational updates remain exactly closed-form due to conjugacy of the multivariate Gaussian likelihood and priors, yet the rendering process (perspective projection of covariances, alpha-blending, and volume integration) is non-linear in pose and depth; no explicit linearization, mean-field factorization, or error bound is provided to justify that conjugacy is preserved, which directly undercuts the central claim of accurate posterior uncertainty.
[§4.3] §4.3, Table 2: The reported ATE reductions (e.g., 12–18% on long sequences) are presented without ablation on the uncertainty propagation term or comparison against a deterministic baseline with identical optimization schedule; it is therefore unclear whether the gains stem from the variational formulation or from other implementation details.
[§3.4] §3.4: The pose posterior update is described as maintaining a full covariance, but the projection Jacobian used in the update is not shown; without this or a sensitivity analysis, the claimed robustness to initialization errors cannot be verified as arising from the uncertainty-aware mechanism.

minor comments (3)

[Figure 3] Figure 3: The uncertainty visualization ellipses are not accompanied by a quantitative calibration metric (e.g., expected calibration error), making it difficult to assess whether the reported posteriors are well-calibrated.
[Related Work] Related Work: The discussion of prior Bayesian SLAM methods (e.g., those using factor graphs or particle filters) is brief and does not explicitly contrast the computational scaling of the proposed closed-form updates against those approaches.
[§5.1] §5.1: Minor notation inconsistency—σ_p is used both for pose noise and for a scene parameter variance in the same paragraph.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the detailed and constructive feedback. We agree that the points raised require clarification and additional experiments to strengthen the manuscript. Below we address each major comment, and the revised version will incorporate the suggested changes.

read point-by-point responses

Referee: [§3.2] §3.2, Eq. (7)–(9): The derivation asserts that the variational updates remain exactly closed-form due to conjugacy of the multivariate Gaussian likelihood and priors, yet the rendering process (perspective projection of covariances, alpha-blending, and volume integration) is non-linear in pose and depth; no explicit linearization, mean-field factorization, or error bound is provided to justify that conjugacy is preserved, which directly undercuts the central claim of accurate posterior uncertainty.

Authors: We acknowledge that the non-linearity of the full rendering pipeline must be addressed explicitly. In the current derivation we linearize the rendering function (including projection and alpha-blending) via a first-order Taylor expansion around the current pose estimate; this renders the effective likelihood Gaussian in the pose parameters and thereby preserves conjugacy for the variational updates. We will add the explicit linearization step, the resulting Jacobian, and a short error-bound discussion (based on the second-order remainder term) to the revised §3.2. revision: yes
Referee: [§4.3] §4.3, Table 2: The reported ATE reductions (e.g., 12–18% on long sequences) are presented without ablation on the uncertainty propagation term or comparison against a deterministic baseline with identical optimization schedule; it is therefore unclear whether the gains stem from the variational formulation or from other implementation details.

Authors: We agree that isolating the contribution of the uncertainty propagation is essential. In the revised manuscript we will add an ablation study that (i) disables uncertainty propagation by replacing posterior covariances with point estimates and (ii) compares against a deterministic 3DGS-SLAM baseline that uses exactly the same optimization schedule, learning rates, and hyperparameters. The new results will be reported alongside the existing Table 2. revision: yes
Referee: [§3.4] §3.4: The pose posterior update is described as maintaining a full covariance, but the projection Jacobian used in the update is not shown; without this or a sensitivity analysis, the claimed robustness to initialization errors cannot be verified as arising from the uncertainty-aware mechanism.

Authors: We will include the explicit form of the projection Jacobian (derived via the chain rule through the Gaussian projection and splatting operations) in the revised §3.4. In addition, we will add a sensitivity analysis that reports tracking accuracy under controlled initialization noise levels, directly comparing the full-covariance variational update against a point-estimate baseline to demonstrate the robustness benefit. revision: yes

Circularity Check

0 steps flagged

No significant circularity; new generative formulation with standard conjugacy

full rationale

The paper presents VBGS-SLAM as a novel coupling of 3DGS with variational Bayesian inference that exploits conjugate Gaussian properties for closed-form updates. No equations, fitted parameters, or self-citations are shown that reduce the claimed uncertainty maintenance or drift mitigation to quantities defined by the method's own outputs. The derivation relies on external mathematical facts about multivariate Gaussian conjugacy and variational inference rather than re-expressing prior fitted results or importing uniqueness from self-citations. The framework is therefore self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

Review performed on abstract only; full model specification, parameter counts, and assumption details are unavailable.

axioms (1)

domain assumption Conjugate properties of multivariate Gaussians allow closed-form variational updates for the joint pose-and-map posterior in this SLAM setting
Abstract states that these properties admit efficient closed-form updates; no further justification supplied.

pith-pipeline@v0.9.0 · 5468 in / 1293 out tokens · 37905 ms · 2026-05-13T20:53:16.815337+00:00 · methodology

discussion (0)

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

IndisputableMonolith/Cost/FunctionalEquation.lean washburn_uniqueness_aczel unclear
By leveraging conjugate properties of multivariate Gaussians and variational inference, our method admits efficient closed-form updates... q(z,μs,Σs,μc,Σc,π,Tt) ... NIW ... Dirichlet
IndisputableMonolith/Foundation/AlexanderDuality.lean alexander_duality_circle_linking unclear
linearizing the spatial likelihood around the current pose estimate μξ,t using first-order Taylor expansion on the tangent space (Eqs. 11-15)

Reference graph

Works this paper leans on

42 extracted references · 42 canonical work pages · 2 internal anchors

[1]

Real-time 3d reconstruction at scale using voxel hashing,

M. Nießner, M. Zollh ¨ofer, S. Izadi, and M. Stamminger, “Real-time 3d reconstruction at scale using voxel hashing,”ACM Transactions on Graphics (ToG), vol. 32, no. 6, pp. 1–11, 2013

work page 2013
[2]

Intelligent hotel ros-based service robot,

Y . Zhang, X. Wang, X. Wu, W. Zhang, M. Jiang, and M. Al- Khassaweneh, “Intelligent hotel ros-based service robot,” in2019 IEEE International Conference on Electro Information Technology (EIT), 2019, pp. 399–403

work page 2019
[3]

Bundle- fusion: Real-time globally consistent 3d reconstruction using on- the-fly surface reintegration,

A. Dai, M. Nießner, M. Zollh ¨ofer, S. Izadi, and C. Theobalt, “Bundle- fusion: Real-time globally consistent 3d reconstruction using on- the-fly surface reintegration,”ACM Transactions on Graphics (ToG), vol. 36, no. 4, p. 1, 2017

work page 2017
[4]

Kinectfusion: Real-time dense surface mapping and tracking,

R. A. Newcombe, S. Izadi, O. Hilliges, D. Molyneaux, D. Kim, A. J. Davison, P. Kohi, J. Shotton, S. Hodges, and A. Fitzgibbon, “Kinectfusion: Real-time dense surface mapping and tracking,” in 2011 10th IEEE international symposium on mixed and augmented reality. Ieee, 2011, pp. 127–136

work page 2011
[5]

Plk-calib: Single-shot and target- less lidar-camera extrinsic calibration using plcker lines,

Y . Zhang, J. Xu, and W. Ren, “Plk-calib: Single-shot and target- less lidar-camera extrinsic calibration using plcker lines,” in2025 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2025, pp. 16 091–16 097

work page 2025
[6]

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,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2022, pp. 12 786–12 796

work page 2022
[7]

Occupancy networks: Learning 3d reconstruction in function space,

L. Mescheder, M. Oechsle, M. Niemeyer, S. Nowozin, and A. Geiger, “Occupancy networks: Learning 3d reconstruction in function space,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2019, pp. 4460–4470

work page 2019
[8]

Di-fusion: Online implicit 3d reconstruction with deep priors,

J. Huang, S.-S. Huang, H. Song, and S.-M. Hu, “Di-fusion: Online implicit 3d reconstruction with deep priors,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2021, pp. 8932–8941

work page 2021
[9]

Deepsdf: Learning continuous signed distance functions for shape representation,

J. J. Park, P. Florence, J. Straub, R. Newcombe, and S. Lovegrove, “Deepsdf: Learning continuous signed distance functions for shape representation,” inProceedings of the IEEE/CVF conference on com- puter vision and pattern recognition, 2019, pp. 165–174

work page 2019
[10]

Nerf- vio: Map-based visual-inertial odometry with initialization leveraging neural radiance fields,

Y . Zhang, D. Wang, J. Xu, M. Liu, P. Zhu, and W. Ren, “Nerf- vio: Map-based visual-inertial odometry with initialization leveraging neural radiance fields,” in2025 IEEE 21st International Conference on Automation Science and Engineering (CASE), 2025, pp. 3506–3511

work page 2025
[11]

V ox- fusion: Dense tracking and mapping with voxel-based neural implicit representation,

X. Yang, H. Li, H. Zhai, Y . Ming, Y . Liu, and G. Zhang, “V ox- fusion: Dense tracking and mapping with voxel-based neural implicit representation,” in2022 IEEE International Symposium on Mixed and Augmented Reality (ISMAR). IEEE, 2022, pp. 499–507

work page 2022
[12]

Gaussian splatting slam,

H. Matsuki, R. Murai, P. H. Kelly, and A. J. Davison, “Gaussian splatting slam,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024, pp. 18 039–18 048

work page 2024
[13]

Splatam: Splat track & map 3d gaussians for dense rgb-d slam,

N. Keetha, J. Karhade, K. M. Jatavallabhula, G. Yang, S. Scherer, D. Ramanan, and J. Luiten, “Splatam: Splat track & map 3d gaussians for dense rgb-d slam,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024, pp. 21 357–21 366

work page 2024
[14]

Catastrophic forgetting in connectionist networks,

R. M. French, “Catastrophic forgetting in connectionist networks,” Trends in cognitive sciences, vol. 3, no. 4, pp. 128–135, 1999

work page 1999
[15]

Variational bayes gaussian splatting,

T. Van de Maele, O. Catal, A. Tschantz, C. L. Buckley, and T. Verbelen, “Variational bayes gaussian splatting,”arXiv preprint arXiv:2410.03592, 2024

work page arXiv 2024
[16]

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

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

work page 2015
[17]

Orb-slam2: An open-source slam system for monocular, stereo, and rgb-d cameras,

R. Mur-Artal and J. D. Tards, “Orb-slam2: An open-source slam system for monocular, stereo, and rgb-d cameras,”IEEE Transactions on Robotics, vol. 33, no. 5, pp. 1255–1262, 2017

work page 2017
[18]

Lsd-slam: Large-scale di- rect monocular slam,

J. Engel, T. Sch ¨ops, and D. Cremers, “Lsd-slam: Large-scale di- rect monocular slam,” inEuropean conference on computer vision. Springer, 2014, pp. 834–849

work page 2014
[19]

Visual navigation using sparse optical flow and time-to-transit,

C. Boretti, P. Bich, Y . Zhang, and J. Baillieul, “Visual navigation using sparse optical flow and time-to-transit,” in2022 International Conference on Robotics and Automation (ICRA), 2022, pp. 9397– 9403

work page 2022
[20]

Pl-cvio: Point-line cooperative visual- inertial odometry,

Y . Zhang, P. Zhu, and W. Ren, “Pl-cvio: Point-line cooperative visual- inertial odometry,” in2023 IEEE Conference on Control Technology and Applications (CCTA), 2023, pp. 859–865

work page 2023
[21]

Mast3r-slam: Real- time dense slam with 3d reconstruction priors,

R. Murai, E. Dexheimer, and A. J. Davison, “Mast3r-slam: Real- time dense slam with 3d reconstruction priors,” inProceedings of the Computer Vision and Pattern Recognition Conference, 2025, pp. 16 695–16 705

work page 2025
[22]

A multi-state constraint kalman filter for vision-aided inertial navigation,

A. I. Mourikis and S. I. Roumeliotis, “A multi-state constraint kalman filter for vision-aided inertial navigation,” inProceedings 2007 IEEE International Conference on Robotics and Automation, 2007, pp. 3565–3572

work page 2007
[23]

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

B. Mildenhall, P. P. Srinivasan, M. Tancik, J. T. Barron, R. Ramamoor- thi, and R. Ng, “Nerf: Representing scenes as neural radiance fields for view synthesis,”Communications of the ACM, vol. 65, no. 1, pp. 99–106, 2021

work page 2021
[24]

Nerf-vins: A real-time neural radiance field map- based visual-inertial navigation system,

S. Katragadda, W. Lee, Y . Peng, P. Geneva, C. Chen, C. Guo, M. Li, and G. Huang, “Nerf-vins: A real-time neural radiance field map- based visual-inertial navigation system,” in2024 IEEE International Conference on Robotics and Automation (ICRA), 2024, pp. 10 230– 10 237

work page 2024
[25]

3d gaussian splatting for real-time radiance field rendering

B. Kerbl, G. Kopanas, T. Leimk ¨uhler, and G. Drettakis, “3d gaussian splatting for real-time radiance field rendering.”ACM Trans. Graph., vol. 42, no. 4, pp. 139–1, 2023

work page 2023
[26]

Dgs-slam: Gaussian splatting slam in dynamic environment,

M. Kong, J. Lee, S. Lee, and E. Kim, “Dgs-slam: Gaussian splatting slam in dynamic environment,”arXiv preprint arXiv:2411.10722, 2024

work page arXiv 2024
[27]

Dy- nagslam: Real-time gaussian-splatting slam for online rendering, track- ing, motion predictions of moving objects in dynamic scenes,

R. B. Li, M. Shaghaghi, K. Suzuki, X. Liu, V . Moparthi, B. Du, W. Curtis, M. Renschler, K. M. B. Lee, N. Atanasovet al., “Dy- nagslam: Real-time gaussian-splatting slam for online rendering, track- ing, motion predictions of moving objects in dynamic scenes,”arXiv preprint arXiv:2503.11979, 2025

work page arXiv 2025
[28]

Large- scale gaussian splatting slam,

Z. Xin, C. Wu, P. Huang, Y . Zhang, Y . Mao, and G. Huang, “Large- scale gaussian splatting slam,”arXiv preprint arXiv:2505.09915, 2025

work page arXiv 2025
[29]

Gs- livo: Real-time lidar, inertial, and visual multi-sensor fused odometry with gaussian mapping,

S. Hong, C. Zheng, Y . Shen, C. Li, F. Zhang, T. Qin, and S. Shen, “Gs- livo: Real-time lidar, inertial, and visual multi-sensor fused odometry with gaussian mapping,”arXiv preprint arXiv:2501.08672, 2025

work page arXiv 2025
[30]

Adam: A Method for Stochastic Optimization

D. P. Kingma and J. Ba, “Adam: A method for stochastic optimiza- tion,”arXiv preprint arXiv:1412.6980, 2014

work page internal anchor Pith review Pith/arXiv arXiv 2014
[31]

Y . L. Tong,The multivariate normal distribution. Springer Science & Business Media, 2012

work page 2012
[32]

Bayesian inference for categorical data analysis,

A. Agresti and D. B. Hitchcock, “Bayesian inference for categorical data analysis,”Statistical Methods and Applications, vol. 14, no. 3, pp. 297–330, 2005

work page 2005
[33]

The wishart and inverse wishart distributions,

S. W. Nydick, “The wishart and inverse wishart distributions,”Elec- tronic Journal of Statistics, vol. 6, no. 1-19, 2012

work page 2012
[34]

Dirichlet and related distri- butions: Theory, methods and applications,

K. W. Ng, G.-L. Tian, and M.-L. Tang, “Dirichlet and related distri- butions: Theory, methods and applications,” 2011

work page 2011
[35]

Kullback-leibler divergence,

J. M. Joyce, “Kullback-leibler divergence,” inInternational encyclo- pedia of statistical science. Springer, 2011, pp. 720–722

work page 2011
[36]

Variational infer- ence: A review for statisticians,

D. M. Blei, A. Kucukelbir, and J. D. McAuliffe, “Variational infer- ence: A review for statisticians,”Journal of the American statistical Association, vol. 112, no. 518, pp. 859–877, 2017

work page 2017
[37]

An introduction to variational methods for graphical models,

M. I. Jordan, Z. Ghahramani, T. S. Jaakkola, and L. K. Saul, “An introduction to variational methods for graphical models,”Machine learning, vol. 37, no. 2, pp. 183–233, 1999

work page 1999
[38]

The Replica Dataset: A Digital Replica of Indoor Spaces

J. Straub, T. Whelan, L. Ma, Y . Chen, E. Wijmans, S. Green, J. J. Engel, R. Mur-Artal, C. Ren, S. Verma, A. Clarkson, M. Yan, B. Budge, Y . Yan, X. Pan, J. Yon, Y . Zou, K. Leon, N. Carter, J. Briales, T. Gillingham, E. Mueggler, L. Pesqueira, M. Savva, D. Batra, H. M. Strasdat, R. D. Nardi, M. Goesele, S. Lovegrove, and R. Newcombe, “The Replica dataset...

work page internal anchor Pith review Pith/arXiv arXiv 1906
[39]

A benchmark for the evaluation of rgb-d slam systems,

J. Sturm, N. Engelhard, F. Endres, W. Burgard, and D. Cremers, “A benchmark for the evaluation of rgb-d slam systems,” inProc. of the International Conference on Intelligent Robot Systems (IROS), Oct. 2012

work page 2012
[40]

Monocular visual-inertial odometry with planar regularities,

C. Chen, P. Geneva, Y . Peng, W. Lee, and G. Huang, “Monocular visual-inertial odometry with planar regularities,” inProc. of the IEEE International Conference on Robotics and Automation, London, UK, 2023

work page 2023
[41]

A tutorial on quantitative trajectory evaluation for visual(-inertial) odometry,

Z. Zhang and D. Scaramuzza, “A tutorial on quantitative trajectory evaluation for visual(-inertial) odometry,” in2018 IEEE/RSJ Interna- tional Conference on Intelligent Robots and Systems (IROS), 2018, pp. 7244–7251

work page 2018
[42]

Point- slam: Dense neural point cloud-based slam,

E. Sandstr ¨om, Y . Li, L. Van Gool, and M. R. Oswald, “Point- slam: Dense neural point cloud-based slam,” inProceedings of the IEEE/CVF International Conference on Computer Vision, 2023, pp. 18 433–18 444

work page 2023