Pith. sign in

REVIEW 5 major objections 6 minor 33 references

Multimodal Fusion SLAM with Fourier Attention

T0 review · 5 major / 6 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read FMF-SLAM claims that a SLAM front-end with Fourier-based attention and RGB-depth fusion can localize reliably in dark, dim, and fast-moving scenes, reporting zero tracking failures and an average trajectory error of 1.07 cm on the…

desk verdict A useful RGB-D SLAM system, but the headline robustness claim compares RGB-only DROID-SLAM against RGB-D FMF-SLAM, so the advantage is uncontrolled. read the letter →

arxiv 2506.18204 v2 pith:T3L2GV3K submitted 2025-06-22 cs.CV cs.AI

Youjie Zhou , Guofeng Mei , Yiming Wang , Yi Wan , Fabio Poiesi This is my paper · ORCID
classification cs.CVcs.AI
keywords visualSLAMRGB-DfusionFourierattentionfasttransformknowledgedistillationopticalflowGNSS-RTKreal-timerobotics
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

FMF-SLAM is a learning-based visual SLAM system designed to keep localizing reliably when the RGB image is noisy, dim, or dark. It claims that replacing the dot-product attention inside a DROID-SLAM-style front-end with attention computed via the fast Fourier transform, and fusing RGB with depth features through multi-scale knowledge distillation, gives both higher robustness and lower compute. On TUM and the challenging TartanAir splits it reports zero outliers across the 20 evaluated sequences and an average absolute trajectory error of 1.07 cm, against 6.02 cm for DROID-SLAM. The same front-end runs on a security robot at about 100 ms per frame, fusing GNSS-RTK positions through global bundle adjustment. If these results hold, frequency-domain attention is sufficient for dense optical flow in SLAM, which makes multimodal fusion practical on real robots.

What carries the argument

The load-bearing component is the Fourier-based multimodal fusion encoder (FMF encoder), a two-branch RGB/depth feature extractor that replaces the front-end of DROID-SLAM. Its defining operation is Eqs. (2)--(5): for each branch, self-attention is computed as $\mathrm{IFFT}\big(\mathrm{FFT}(Q_i)\,\overline{\mathrm{FFT}(K_i)}\big)$, a fast correlation in the frequency domain with no softmax and no learned attention weights, after which cross-attention exchanges queries and keys between branches. The multi-scale knowledge distillation loss then pulls the two branches into agreement across three granularities. This assembly is what carries the efficiency claim (77.76 ms per frame on TartanAir, 100.85 ms on the robot) and the robustness claim (zero tracking failures in the challenging test set).

What would settle it

Train the identical two-branch encoder, loss schedule, and DROID-SLAM back-end, but replace Fourier attention with standard dot-product self- and cross-attention (with softmax). If that variant matches or exceeds FMF-SLAM's ACC1px of 82.04% and AEPE2D of 1.51 on TartanAir, or achieves comparable ATE on the challenging test split, then the Fourier operation is not the source of the reported accuracy and the efficiency claim needs a new benchmark. A second check is to replace the conjugate product in Eq. (2) with a plain FFT product and see whether flow quality collapses.

Watch

Extended reading notes

Core claim

The paper's central discovery is that a SLAM front-end can compute self-attention and cross-modal attention entirely in the frequency domain, as element-wise products of FFTs rather than matrix dot products, and still extract the correlation information needed for dense optical flow. With RGB and depth branches, the self-attention map is $A_i = \mathrm{IFFT}\big(\mathrm{FFT}(Q_i)\,\overline{\mathrm{FFT}(K_i)}\big)$, normalized and applied to values as $V_i = \mathcal{N}(A_i)\,V_i$, with a $1\times1$ convolution residual; cross-attention then mixes the two modalities bidirectionally. A multi-scale distillation loss $L_k = \alpha L_{L2} + \beta L_s + \delta L_c$ aligns the RGB and depth feature maps in $L2$, spatial, and channel terms. On TUM and on the TartanAir fast-moving, dimmed, and dark splits, FMF-SLAM reports zero outliers across the 20 evaluated sequences and an average ATE of 1.07 cm, versus one outlier and 6.02 cm for DROID-SLAM. The authors present this as evidence that an efficient Fourier-based multimodal encoder can replace expensive dot-product attention inside a learning-based SLAM front-end without sacrificing accuracy.

Load-bearing premise

The system's gains rest on the assumption that swapping the usual learned attention weights for a fixed frequency-domain multiplication costs no information needed for accurate optical flow, and the experiments never compare the two attention schemes directly.

Editorial extensions

If this is right

  • A learning-based SLAM front-end that consumes depth together with RGB can remain accurate where RGB-only systems fail: across the 20 tested sequences in Table I the FMF encoder produced zero tracking outliers.
  • Fourier-domain attention is an efficiency lever worth taking seriously: the full pipeline runs at 77.76 ms per frame on TartanAir and 100.85 ms per frame on the robot, within range of real-time deployment.
  • Fusing GNSS-RTK through global bundle adjustment keeps large outdoor loops tight, with accumulation errors of 0.09 m over 147.45 m and 0.26 m over 391.82 m.
  • On real indoor data under light-changing and dark conditions, the depth branch supplies information the RGB stream lacks, giving average ATE of 11.03 cm and 16.84 cm where RGB-only baselines diverge or fail.

Reading between the lines

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

  • The paper does not compare its Fourier attention against standard dot-product attention under identical training, so the accuracy gains from the attention block could come from the depth branch and distillation rather than from the Fourier operation itself; a direct head-to-head would settle this.
  • Because the Fourier attention has no learned weights, it behaves as a fixed frequency-domain correlation; if that is the true source of the gains, similar FFT-based correlation layers could replace other costly matching modules in SLAM front-ends beyond attention.
  • The same two-branch encoder could transfer to other modality pairs, such as stereo or thermal-plus-RGB, wherever a second depth-like signal is available; this is a testable extension the paper does not pursue.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 6 minor

Summary. The paper proposes FMF-SLAM, a learning-based RGB-D SLAM system built on DROID-SLAM, with three claimed contributions: a Fourier-based self- and cross-attention mechanism for multimodal feature extraction, a multi-scale knowledge distillation loss between RGB and depth branches, and integration with GNSS-RTK and global bundle adjustment for deployment on a security robot. The method is evaluated on TUM, TartanAir (with held-out challenging sequences), and the authors' own Realtest-GT and Realtest-Visual datasets. The central claim is that FMF-SLAM achieves state-of-the-art localization robustness and accuracy under fast motion, dim lighting, and darkness, with zero tracking failures on the selected challenging sequences and an average ATE of 1.07 cm versus 6.02 cm for DROID-SLAM, while running at about 100 ms per frame on a robot.

Significance. If the central claims hold, the paper would be a useful demonstration that an efficient Fourier-domain mixing operation can replace dot-product attention in a learning-based SLAM front-end and that RGB-D multimodal fusion improves robustness in visually degraded scenes. The paper has notable strengths: it reports results on external benchmarks (TUM, TartanAir), uses a held-out TartanAir test split disjoint from training, provides a quantitative ablation (Table IV), reports parameter counts and inference times, and states that code and datasets will be released. However, the headline comparison is not controlled: FMF-SLAM uses RGB-D input while DROID-SLAM is evaluated with RGB only, although DROID-SLAM natively supports RGB-D input. The ablation does not include a standard softmax-attention baseline, so the specific benefit of the Fourier attention operation is not isolated. These issues are load-bearing for the claims of state-of-the-art robustness and of Fourier attention as the cause of the gains.

major comments (5)
  1. [Tables I and II] The principal comparison is uncontrolled with respect to input modality. DROID-SLAM is listed as using RGB input only while FMF-SLAM uses RGB-D, and DROID-SLAM natively supports RGB-D cameras (Teed and Deng, NeurIPS 2021). Without a DROID-SLAM RGB-D baseline evaluated on the same TartanAir and Realtest-GT sequences, the zero-outlier result and the average ATE of 1.07 cm versus 6.02 cm cannot be attributed to the proposed Fourier attention or distillation; the additional depth stream alone could explain the difference. The TUM rows, where FMF-SLAM is slightly worse than RGB-only DROID-SLAM (fr1/desk 1.73 vs. 1.63, fr3/office 1.28 vs. 1.18), reinforce this concern. Please add the missing RGB-D baseline or substantially rephrase the claims so that they do not imply an architectural advantage over DROID-SLAM.
  2. [Section III-A, Eqs. (2)-(6)] The Fourier attention mechanism is not fully specified and is not demonstrated to act as attention. As written, Eq. (2) is an element-wise product in the Fourier domain followed by inverse FFT, which is a circular correlation/convolution operation; there is no softmax and no learned attention weight. The text says "FFT(·) denotes the conjugate transpose operation" but the symbols in Eqs. (2) and (5) do not make clear which operand is conjugated or along which axis the FFT is applied (spatial dimensions, channel dimension, or both). This makes the method irreproducible as stated. In addition, Table IV Exp. 3 vs. Exp. 4 only compares Fourier attention against no attention; a standard dot-product/softmax attention or a linear attention baseline is needed to support the claim that the Fourier-domain mixing preserves the correlation information needed for optical flow in SLAM.
  3. [Section III-A, Eq. (5)] Only the cross-attention output for the RGB branch (V_r) is defined. The claim that cross-attention is applied bidirectionally ("RGB ↔ depth") requires an analogous equation for the depth branch, V_d, which is never given. Without this equation, the bidirectional exchange and the exact inputs to the distillation losses are not fully specified. Please provide the missing expression and clarify how the two cross-attention branches are combined in the loss.
  4. [Tables I, II, and IV] All quantitative results are reported as single runs with no error bars, standard deviations, or number of random seeds. Several reported differences are small relative to plausible run-to-run variation (e.g., fr1/xyz 0.96 vs. 1.04, OfficeP05 0.22 vs. 0.21, EndofworldP06 0.45 vs. 0.58), so the claimed improvements are not statistically supported. Please report mean ± std over multiple runs or otherwise justify that the evaluation procedure is deterministic and the differences are significant.
  5. [Section IV-D] The real-time and deployment claims need more context. "100.85 ms per frame" is stated without specifying the sensor frame rate, the hardware pipeline, the input resolution, or whether the time includes depth estimation. The TartanAir figure of 77.76 ms per frame is similarly reported without a runtime comparison to DROID-SLAM on the same machine. Moreover, the outdoor robot experiments use ZoeDepth-estimated depth rather than sensor depth, so they do not directly validate the RGB-D front-end. Please provide a runtime comparison table and state the estimated-depth limitation explicitly.
minor comments (6)
  1. [References] Reference [9] attributes "Attention is all you need" to Mayer et al. at NIPS; the correct authors are Vaswani et al., NeurIPS 2017. Reference [15] gives the TartanAir year as 2012, but the dataset was published in 2020. Please correct these citations.
  2. [Section II-A] There is a typo, "direct direct", in the first paragraph; please fix it.
  3. [Eqs. (7)-(9)] The distillation losses are hard to parse because of ambiguous norm notation and unclear summation ranges; for example, in Eq. (8) it is not clear whether the 1/HW average is applied to each feature map before the norm, and in Eq. (9) the argument of the norm mixes tensor and scalar quantities. Please rewrite these equations with explicit indexing.
  4. [Figure 1] The block diagram is nearly unreadable in the PDF: the fonts are tiny and the connections are dense. Please provide a higher-resolution figure with larger labels.
  5. [Contributions, page 2] The claim "the first learning-based multimodal SLAM system" is too strong; DROID-SLAM already supports RGB-D input, and there are other learning-based RGB-D SLAM systems. Please qualify this as "the first to our knowledge" and cite the relevant prior work.
  6. [Section IV-E] The definitions of ACC1px, AEPE2D, Rot0.1°, and Tra0.01m cite reference [7], which is the authors' own paper. Please define these metrics directly in the text or cite the original sources.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation found: FMF-SLAM's claims are empirical and externally benchmarked; the main caveat is an uncontrolled RGB versus RGB-D baseline comparison, not circularity.

full rationale

FMF-SLAM is an empirical architecture paper whose claimed improvements are measured on external benchmarks (TUM, held-out TartanAir sequences) and on a laser-tracker ground-truth dataset, rather than derived from a closed-form model whose output is equivalent to its input. The Fourier attention equations (2)-(6) define a mechanism; they are not predictions fitted from the evaluation data. The knowledge-distillation loss (7)-(10) is a training objective whose effect is tested through ablations that perturb the model, so the reported gains are not forced by construction. The only self-citation, [7], is used for motivation and for the ACC1px/AEPE2D metric definitions in the ablation; these are conventional evaluation quantities and are not load-bearing for the central claim, and no uniqueness theorem or forced choice is imported from prior work. The most serious limitation is not circularity: Table I compares RGB-only DROID-SLAM against RGB-D FMF-SLAM without a depth-enabled DROID-SLAM baseline, so the headline ATE gap may be explained by the additional depth modality. That is a controlled-comparison confound, not a derivation that reduces to its own input. Therefore no circular step is exhibited and the score is 0.

Assumptions & free parameters 2 free parameters · 5 assumptions · 0 invented entities

The central result depends on hand-tuned distillation loss weights, standard FFT math, assumptions about depth availability and synthetic-to-real transfer, and an ad hoc Fourier-attention-as-attention assumption. The ledger is small, but the unstated loss weights and the missing standard-attention baseline are the main gaps.

free parameters (2)
  • alpha, beta, delta (distillation loss weights) = not reported
    Eq. (10) defines Lk = alpha * LL2 + beta * Ls + delta * Lc, but the paper never reports the values used in training. These hand-chosen weights directly influence the learned features and therefore the final SLAM accuracy.
  • gamma (loss weighting base) = not reported
    Used in Eq. (11) and (12) to weight pose and optical flow losses across iterations. The value is not stated, making the training loss function incompletely specified.
assumptions (5)
  • standard math FFT and IFFT are exactly invertible and conjugate products in the frequency domain implement correlation.
    Invoked in Eq. (2)-(5) for the proposed Fourier attention. This is standard signal processing background, but the paper does not specify the exact axis and conjugation conventions.
  • domain assumption Depth maps are available and sufficiently accurate at test time.
    Indoor tests use RealSense L515 depth; outdoor deployment uses ZoeDepth-predicted depth (Sec. IV-D). If the depth channel is unreliable, the reported multimodal gains and outdoor accumulation errors may not replicate.
  • domain assumption Synthetic TartanAir training transfers to real indoor and outdoor environments.
    The model is trained only on TartanAir (Sec. IV-A) and then evaluated on TUM, real indoor sequences, and an outdoor robot. No domain adaptation step is described.
  • ad hoc to paper Fourier-domain element-wise multiplication without softmax acts as attention for optical flow features.
    Eq. (2)-(5) replace dot-product attention with FFT products, but no theoretical argument or baseline comparison to standard attention is provided (Sec. IV-E).
  • ad hoc to paper Multi-scale knowledge distillation between RGB and depth branches improves fusion.
    The distillation losses in Eq. (7)-(10) are designed specifically for this architecture. The ablation supports the design, but no alternative distillation choices are tested.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Multimodal Fusion SLAM with Fourier Attention." pith.science (2026). https://pith.science/paper/T3L2GV3K

@misc{pith2026250618204,
  author       = {Pith},
  title        = {Pith review of: Multimodal Fusion SLAM with Fourier Attention},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/T3L2GV3K}},
  note         = {Machine review of arXiv:2506.18204}
}
read the original abstract

Visual SLAM is particularly challenging in environments affected by noise, varying lighting conditions, and darkness. Learning-based optical flow algorithms can leverage multiple modalities to address these challenges, but traditional optical flow-based visual SLAM approaches often require significant computational resources.To overcome this limitation, we propose FMF-SLAM, an efficient multimodal fusion SLAM method that utilizes fast Fourier transform (FFT) to enhance the algorithm efficiency. Specifically, we introduce a novel Fourier-based self-attention and cross-attention mechanism to extract features from RGB and depth signals. We further enhance the interaction of multimodal features by incorporating multi-scale knowledge distillation across modalities. We also demonstrate the practical feasibility of FMF-SLAM in real-world scenarios with real time performance by integrating it with a security robot by fusing with a global positioning module GNSS-RTK and global Bundle Adjustment. Our approach is validated using video sequences from TUM, TartanAir, and our real-world datasets, showcasing state-of-the-art performance under noisy, varying lighting, and dark conditions.Our code and datasets are available at https://github.com/youjie-zhou/FMF-SLAM.git.

Figures

Figures reproduced from arXiv: 2506.18204 by the authors.

Figure 1
Figure 1. Block diagram of FMF-SLAM. (a) Encoder: Inputs are [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. a) Front view of the laser tracker. b) Front view of our security robot. [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Relationship between the number of successfully tracked sequences [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (4 more)
Figure 5
Figure 5. Figure 5: Comparison between the groundtruth and estimated trajectories for [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 6
Figure 6. Figure 6: Relationship between successfully tracked sequences (Tab. II) in the [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]
Figure 8
Figure 8. Figure 8: 3D reconstruction of our FMF-SLAM around garden and tall building [PITH_FULL_IMAGE:figures/full_fig_p007_8.png]
Figure 7
Figure 7. Figure 7: 3D reconstruction of our FMF-SLAM under standard, light-changing, [PITH_FULL_IMAGE:figures/full_fig_p007_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

33 extracted references · 30 canonical work pages

  1. [1]

    Past, present, and future of simultaneous localization and mapping: Toward the robust-perception age,

    C. Cadena and et al, “Past, present, and future of simultaneous localization and mapping: Toward the robust-perception age,” T-RO, 2016

    work page 2016

  2. [2]

    Gnss-stereo-inertial slam for arable farming,

    J. Cremona, J. Civera, E. Kofman, and T. Pire, “Gnss-stereo-inertial slam for arable farming,” Journal of Field Robotics , 2023

    work page 2023

  3. [3]

    Real-time indoor localization with visual slam for in-building emergency response,

    P. Tseng, J. Lin, and et al, “Real-time indoor localization with visual slam for in-building emergency response,” Automation in Construction , 2022

    work page 2022

  4. [4]

    Present and future of slam in extreme environments: The darpa subt challenge,

    K. Ebadi, L. Bernreiter, and et al, “Present and future of slam in extreme environments: The darpa subt challenge,” T-RO, 2023

    work page 2023

  5. [5]

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

    R. Mur-Artal and J. Tard ´os, “Orb-slam2: An open-source slam system for monocular, stereo, and rgb-d cameras,” T-RO, vol. 33, no. 5, pp. 1255– 1262, 2017

    work page 2017

  6. [6]

    Direct sparse odometry,

    J. Engel, V . Koltun, and D. Cremers, “Direct sparse odometry,” TPAMI, vol. 40, no. 3, pp. 611–625, 2017

    work page 2017

  7. [7]

    Attentive multimodal fusion for optical and scene flow,

    Y . Zhou, G. Mei, Y . Wang, Y . Wan, and F. Poiesi, “Attentive multimodal fusion for optical and scene flow,” RA-L, 2023

    work page 2023

  8. [8]

    A review on the attention mechanism of deep learning,

    Z. Niu, G. Zhong, and H. Yu, “A review on the attention mechanism of deep learning,” Neurocomputing, 2021

    work page 2021

Show all 33 references
  1. [9]

    Attention is all you need,

    N. Mayer, E. Ilg, P. Hausser, and P. Fischer, “Attention is all you need,” in NIPS, Los Angeles, US, December 2017

    2017

  2. [10]

    DROID-SLAM: Deep Visual SLAM for Monoc- ular, Stereo, and RGB-D Cameras,

    Z. Teed and J. Deng, “DROID-SLAM: Deep Visual SLAM for Monoc- ular, Stereo, and RGB-D Cameras,” in NeurIPS, December 2021

    2021

  3. [11]

    Guest editors introduction to the top 10 algorithms,

    J. Dongarra and F. Sullivan, “Guest editors introduction to the top 10 algorithms,” CiSE, 2000

    2000

  4. [12]

    Alleviating Catas- trophic Forgetting of Incremental Object Detection via Within-Class and Between-Class Knowledge Distillation,

    M. Kang, J. Zhang, J. Zhang, and X. Wang, “Alleviating Catas- trophic Forgetting of Incremental Object Detection via Within-Class and Between-Class Knowledge Distillation,” in ICCV, Paris, October 2023

    2023

  5. [13]

    3d mapping database-aided gnss rtk and its assessments in urban canyons,

    H. NG and L. HSU, “3d mapping database-aided gnss rtk and its assessments in urban canyons,” IEEE Transactions on Aerospace and Electronic Systems, 2021

    2021

  6. [14]

    A bench mark for the evaluation of RGB-D SLAM systems,

    J. Sturm, N. Engelhard, and et al, “A bench mark for the evaluation of RGB-D SLAM systems,” in IROS, Vilamoura, Algarve, October 2012

    2012

  7. [15]

    TartanAir: A dataset to push the limits of visual slam,

    W. Wang, D. Zhu, X. Wang, and et al, “TartanAir: A dataset to push the limits of visual slam,” in IROS, October 2012

    2012

  8. [16]

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

    C. Campos, R. Elvira, and J. Rodr ´ıguez, “Orb-slam3: An accurate open- source library for visual, visual–inertial, and multimap slam,” T-RO, vol. 37, no. 6, pp. 1874–1890, 2021

    2021

  9. [17]

    SURF: Speeded up robust features,

    H. Bay, T. Tuytelaars, and L. V . Gool, “SURF: Speeded up robust features,” in ECCV, Graz, Austria, May 2006

    2006

  10. [18]

    ORB: An efficient alternative to SIFT or SURF,

    E. Rublee, V . Rabaud, and K. Konolige, “ORB: An efficient alternative to SIFT or SURF,” in ICCV, Barcelona,Spain, November 2011

    2011

  11. [19]

    Superpoint: Self- supervised interest point detection and description,

    D. DeTone, T. Malisiewicz, and A. Rabinovich, “Superpoint: Self- supervised interest point detection and description,” in CVPR, Salt Lake City, USA, June 2018

    2018

  12. [20]

    Superglue: Learning feature matching with graph neural networks,

    P. Sarlin, D. DeTone, T. Malisiewicz, and et al, “Superglue: Learning feature matching with graph neural networks,” in CVPR, June 2020

    2020

  13. [21]

    Deepvo: Towards end-to-end visual odometry with deep recurrent convolutional neural networks,

    S. Wang, R. Clark, H. Wen, and et al, “Deepvo: Towards end-to-end visual odometry with deep recurrent convolutional neural networks,” in ICRA, Marina Bay Sands, Singapore, May 2020

    2020

  14. [22]

    Deepv2d: Video to depth with differentiable structure from motion,

    Z. Teed and J. Deng, “Deepv2d: Video to depth with differentiable structure from motion,” arXiv preprint arXiv:1812.04605 , 2018

    2018 arXiv

  15. [23]

    Deepfactors: Real-time proba- bilistic dense monocular slam,

    J. Czarnowski, T. Laidlow, and et al, “Deepfactors: Real-time proba- bilistic dense monocular slam,” RA-L, vol. 5, no. 2, pp. 721–728, 2020

    2020

  16. [24]

    Tartanvo: A generalizable learning- based vo,

    W. Wang, Y . Hu, and S. Scherer, “Tartanvo: A generalizable learning- based vo,” in CoRL, London, England, November 2021

    2021

  17. [25]

    Deep virtual stereo odometry: Leveraging deep depth prediction for monocular direct sparse odometry,

    N. Yang, R. Wang, J. Stuckler, and et al, “Deep virtual stereo odometry: Leveraging deep depth prediction for monocular direct sparse odometry,” in ECCV, Munich, Germany, September 2018

    2018

  18. [26]

    D3VO: Deep depth, deep pose and deep uncertainty for monocular visual odometry,

    N. Yang, L. Stumberg, R. Wang, and et al, “D3VO: Deep depth, deep pose and deep uncertainty for monocular visual odometry,” in CVPR, June 2020

    2020

  19. [27]

    Deep patch visual odometry,

    Z. Teed, L. Lipson, and J. Deng, “Deep patch visual odometry,” in NeurIPS, 2024

    2024

  20. [28]

    Fast-fnet: Accelerating trans- former encoder models via efficient fourier layers,

    N. Sevim, E. Ozyedek, and et al, “Fast-fnet: Accelerating trans- former encoder models via efficient fourier layers,” arXiv preprint arXiv:2209.12816, 2022

    2022 arXiv

  21. [29]

    DistillBEV: Boosting Multi-Camera 3D Object Detection with Cross-Modal Knowledge Distil- lation,

    Z. Wang, D. Li, C. Luo, C. Xie, and X. Yang, “DistillBEV: Boosting Multi-Camera 3D Object Detection with Cross-Modal Knowledge Distil- lation,” in ICCV, Paris, FR, October 2023

    2023

  22. [30]

    RESLAM: A real-time robust edge-based SLAM system,

    F. Schenk and F. Fraundorfer, “RESLAM: A real-time robust edge-based SLAM system,” in ICRA, Montreal, Canada, May 2019

    2019

  23. [31]

    Measurement accuracy and practical assessment of the lidar camera Intel RealSense L515,

    A. Breitbarth, C. Hake, and G. Notni, “Measurement accuracy and practical assessment of the lidar camera Intel RealSense L515,” in OMSII XII, June 2021

    2021

  24. [32]

    Dynamic performance evaluation of 6D laser tracker sensor,

    K. Lau, Y . Yang, and et al, “Dynamic performance evaluation of 6D laser tracker sensor,” in PerMIS, New York, USA, September 2010

    2010

  25. [33]

    Zoedepth: Zero-shot transfer by combining relative and metric depth,

    S. Bhat, R. Birkl, and e. a. D. Wofk and, “Zoedepth: Zero-shot transfer by combining relative and metric depth,” arXiv preprint arXiv:2302.12288., 2023

    2023 arXiv

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.