Pith. sign in

REVIEW 4 major objections 5 minor 1 cited by

DoGFlow derives dense LiDAR scene flow from radar Doppler measurements, without any manual labels.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

Radar Doppler velocities, clustered under rigidity assumptions, can be propagated to LiDAR as pseudo scene flow labels, outperforming self-supervised baselines on TruckScenes and improving label efficiency.

T0 review reviewed 2026-08-05 challenge →

load-bearing objection Real novelty and believable gains, but the Doppler-to-full-3D recovery is under-specified and needs a conditioning analysis before the full-3D claim lands. the 4 major comments →

arxiv 2508.18506 v1 pith:UIW67J5U submitted 2025-08-25 cs.CV

DoGFlow: Self-Supervised LiDAR Scene Flow via Cross-Modal Doppler Guidance

Ajinkya Khoche , Qingwen Zhang , Yixi Cai , Sina Sharif Mansouri , Patric Jensfelt This is my paper
classification cs.CV
keywords scene flowLiDAR4D radarDoppler velocityself-supervised learningpseudo-labelscross-modal label transferadverse weather
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The reading

The paper sets out to show that LiDAR scene flow can be learned without expensive human annotation if 4D radar Doppler measurements are used as the teacher. Because each radar point yields only a one-dimensional radial speed, DoGFlow clusters dynamic radar returns and solves for each cluster's full 3D velocity under a rigidity assumption; it then transfers those velocities to associated LiDAR clusters, using range-adaptive association and forward-projection to resolve ambiguous candidates. On the MAN TruckScenes dataset, the resulting pseudo-labels beat existing self-supervised methods, particularly beyond 35 meters and in snow, and push a LiDAR-only scene flow network to more than 90 percent of fully supervised accuracy with only 10 percent ground truth. The same pipeline can run at runtime for direct flow estimation or offline to generate training labels.

Core claim

The paper's central claim is that the missing motion information in LiDAR can be supplied by radar Doppler through a two-stage cross-modal transfer. First, dynamic radar points are separated by an ego-motion-compensated Doppler threshold and clustered with connected components on a spatial-plus-Doppler graph. For each cluster, the one-dimensional radial constraints from all its points form a linear system whose solution is the cluster's full 3D velocity, solved by least squares with physical bounds. Second, these velocities are propagated to LiDAR: ground-removed LiDAR is clustered by intensity, dynamic labels are assigned by majority vote over associated radar clusters, and when several can

What carries the argument

The load-bearing identity is the Doppler observation model v_j^comp = u_j^T R_ego^{Si} v_k, which converts each radar measurement into one scalar constraint on a three-dimensional cluster velocity. Sufficiently diverse line-of-sight directions inside a rigid cluster make the resulting linear system determine the cluster's full velocity; connected-components clustering on the spatial-plus-Doppler graph provides the rigid groups, and the ambiguity-resolved propagation step carries those velocities across to LiDAR points.

Load-bearing premise

The method assumes that the Doppler readings within each cluster are enough to determine all three components of the cluster's velocity; for small or distant objects that do not span many viewing angles, the estimate depends on the solver's physical bounds rather than on the measurements.

What would settle it

Select radar clusters whose points span a narrow range of line-of-sight directions at long range (e.g., a distant truck seen nearly head-on) and compare their estimated full velocity against ground truth. If DoGFlow still produces a full 3D velocity close to truth, the recovery is doing more than the Doppler constraints can support; if the velocity is pinned to the least-squares bound, the mechanism is confirmed to be rank-deficient exactly there. A simpler aggregate check is to plot cluster-level velocity error against the angular diversity of the cluster's radar points.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • LiDAR-only scene flow networks can be trained to near-supervised quality with no manual labels: on TruckScenes, an SSF backbone pretrained on DoGFlow pseudo-labels reaches about 90 percent of fully supervised performance with 10 percent ground truth.
  • Doppler-based dynamic classification is itself a reusable signal: swapping it into the SeFlow pipeline improves both near-range and far-range dynamic EPE and IoU.
  • The label-transfer pipeline is training-free and can be used either as a direct runtime estimator or as an offline pseudo-label generator, decoupling the quality of the teacher from the speed of the student.
  • Because radar is relatively unaffected by rain and snow, the method keeps dynamic IoU high in adverse weather where LiDAR-only geometric self-supervision degrades sharply.
  • The framework is modular: any LiDAR scene flow backbone can consume the pseudo-labels, so the label-efficiency result is not tied to a single network architecture.

Where Pith is reading between the lines

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

  • It follows, though the paper does not test it, that the same pseudo-label recipe should transfer to other 4D-radar-equipped urban datasets, with calibration quality and radar density as the main limiting factors.
  • A natural stress test beyond the paper: as radar-LiDAR extrinsic calibration drifts, the range-adaptive association threshold should cause label quality to degrade smoothly; quantifying that sensitivity would show whether online extrinsic correction is needed.
  • Articulated objects such as pedestrians violate the per-cluster rigidity assumption; splitting clusters into locally rigid parts would extend the method but would also create more candidate velocities to disambiguate at propagation time.
  • The dynamic/static decision comes directly from Doppler thresholding, so the method could contribute to tracking and detection even in regimes where the exact flow vector is noisy.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. DoGFlow proposes a self-supervised LiDAR scene flow method that generates pseudo-labels from 4D radar Doppler measurements and transfers them to LiDAR. In the first stage, dynamic radar points are thresholded by ego-compensated radial velocity, clustered via Connected Components Labeling in spatial/Doppler space, and each cluster's full 3D velocity is estimated by solving a linear system from per-point radial constraints. In the second stage, LiDAR points are denoised, clustered with HDBSCAN, associated to radar clusters, and labeled as dynamic by majority voting; ambiguous multi-velocity cases are resolved by forward projection and Chamfer distance. The authors evaluate on MAN TruckScenes, reporting improvements over self-supervised baselines in range-wise dynamic EPE and dynamic IoU, favorable pseudo-labeling results when training an SSF backbone, label-efficiency gains, and robust behavior in adverse weather.

Significance. The core idea is original and physically motivated: using Doppler as an independent cross-modal cue for LiDAR scene flow is a promising alternative to purely geometric self-supervision. The paper also addresses a practically important regime—long-range and adverse-weather LiDAR—where chamfer-based methods typically fail. A strength is that the label source (radar Doppler) is physically independent of the box-based ground truth used for evaluation, so the reported improvements are not trivially explained by label leakage. Another strength is the pseudo-labeling evaluation, which shows that the generated labels can train a LiDAR backbone and reduce annotation cost. However, the central claim of recovering full 3D velocities from Doppler constraints needs more rigorous support, and the empirical claims would be stronger with error bars and threshold sensitivity analysis. If these issues are addressed, the work would be a solid contribution to self-supervised scene flow.

major comments (4)
  1. [IV-A, Eq. (6)] The central velocity-recovery step treats each radar point as one scalar constraint on a 3D cluster velocity. Full recovery is only possible if the rows u_j^T R_ego^{S_i} within a cluster span R^3. For compact objects at long range—the regime where DoGFlow claims its main advantage—these rows are nearly parallel, making the system rank-deficient or ill-conditioned. The text states the system 'may be over or under-constrained' and defers to 'physical velocity bounds [39]', but the bound values, the fraction of rank-deficient clusters, and the condition numbers are never reported. If many clusters are unobservable, the propagated LiDAR flow is determined by the unspecified bounds, not by Doppler evidence. Please report observability statistics and ablations with and without the bounds.
  2. [Tables I–IV and Fig. 5] All quantitative results are single-run point estimates without error bars. For example, Table I shows DoGFlow's close-range dynamic EPE (0.6892) is worse than NSFP (0.6152) and FastNSF (0.6019), while its far-range advantage is 0.7013 vs 0.8829; without variance or significance testing it is unclear which differences are reliable. Figure 5's headline claim of 'over 90% of fully supervised performance with 10% GT' rests on a comparison of 0.4541 vs 0.4119. Please report multiple seeds with standard deviations and, if feasible, significance tests for the main comparisons.
  3. [V-D and Algorithm 1] The method depends on at least seven manually set thresholds: δ_dyn, δ_spatial, δ_velocity, δ_intensity, δ_neighbor, and δ_adaptive^min/max. No sensitivity analysis is provided. Because the pseudo-label pipeline is not learned, these thresholds directly determine cluster boundaries and, consequently, the quality of the propagated flow. Without a sensitivity study, it is difficult to judge whether the reported gains are robust or tuned to the validation set. Please add a sensitivity sweep over the most influential thresholds, especially δ_spatial, δ_velocity, and δ_dyn.
  4. [Abstract and Table I] The abstract and introduction claim that DoGFlow computes motion pseudo-labels 'in real-time' and enables real-time inference. However, Table I reports 2291 ± 465 ms per frame for direct scene flow estimation, which is far above the 100 ms frame interval at 10 Hz. Only the feedforward model trained on pseudo-labels runs at 25 ± 12 ms. The real-time claim for the proposed method itself is not supported. Please revise the wording or report a version of the pipeline that meets real-time constraints.
minor comments (5)
  1. [Throughout] There are several typos: 'sligtly' in Section VI-A, 'leverging' in Section II-A, 'DogFlow' in the Figure 4 caption, and an extra 'and' in Section IV-A.
  2. [IV-A] The notation v_j^comp is used both as a scalar (Eq. (3)) and as a vector (v_j^comp = v_j^comp u_j). Please use distinct symbols to avoid ambiguity.
  3. [V-D / VI-A] The Doppler threshold δ_dyn = 0.05 m/s is much smaller than the dynamic GT threshold of 0.05 m per frame (equivalent to 0.5 m/s at 10 Hz). The relation between these thresholds should be discussed, since a very low Doppler threshold may classify slow or static-but-noisy returns as dynamic.
  4. [V-A] The annotation schedule (bounding boxes every fifth frame) and the evaluation protocol (which frames are used for scene flow metrics) should be clarified. It is not clear whether metrics are computed only on annotated frames or on all frames.
  5. [Repository] The code link is only promised 'upon acceptance.' Since reproducibility is a key strength of pseudo-labeling papers, please provide an anonymous repository or more detailed implementation details, especially for the least-squares solver and bound values.

Circularity Check

0 steps flagged

No significant circularity: the Doppler-derived pseudo-label chain is an independent physical measurement, and the self-citations are architectural or baseline choices, not load-bearing circularity.

full rationale

The derivation chain is: raw 4D radar Doppler radial velocities are ego-compensated in Eq. (3), dynamic radar points are thresholded in Eq. (4), clustered using spatial and Doppler proximity in Eq. (5), and each cluster's 3D velocity is recovered by solving the linear system in Eq. (6) under a rigidity assumption. This is an inversion of an independent sensor signal (Doppler), not a fit to the ground-truth flow used for evaluation. The cross-modal propagation in Algorithm 1 transfers those cluster velocities to LiDAR by nearest-neighbor association and Chamfer-based disambiguation; the resulting pseudo-labels are then used to train an SSF backbone. Nothing in this chain defines the predicted scene flow in terms of the box-derived ground-truth flow or in terms of the reported metrics, and the MAN TruckScenes ground truth is external to the method. The Section IV-A caveat that the linear system 'may be over or under-constrained' is a genuine observability/conditioning limitation, but it is not a circular step: even if physical velocity bounds [39] dominate for some compact clusters, the pipeline is not assuming the answer it claims to produce. The self-references (SSF [7], SeFlow [13], DUFOMap [29]) are used as backbone, baseline, and mapping module; they do not supply the Doppler-to-LiDAR label transfer mechanism, so they are not load-bearing circularity. The label-efficiency experiment is an empirical comparison of pseudo-label pretraining plus varying GT fractions, not a statistical tautology. Therefore no circular step can be exhibited, and the circularity score is 0.

Axiom & Free-Parameter Ledger

7 free parameters · 6 axioms · 0 invented entities

No new physical entities, forces, or conserved quantities are introduced. The pseudo-labels are derived quantities, not invented entities. The main burden is carried by the hand-set thresholds and the rigidity and calibration assumptions.

free parameters (7)
  • delta_dyn = 0.05 m/s
    Ego-compensated Doppler magnitude above this threshold marks a radar point dynamic in Eq. (4); filters slow movers and multipath noise.
  • delta_spatial = 3.0 m
    Spatial proximity threshold for CCL graph edges in Eq. (5).
  • delta_velocity = 1.5 m/s
    Doppler-velocity proximity threshold for CCL graph edges in Eq. (5).
  • delta_intensity = 0.008
    Splits LiDAR points into high and low intensity subsets; low-intensity points are treated as potential weather noise.
  • delta_neighbor = 0.5 m
    Distance threshold for reassigning filtered low-intensity LiDAR points to their nearest cluster.
  • delta_adaptive range = 0.1 m to 5.0 m, linear interpolation
    Range-adaptive radar-to-LiDAR association threshold; the interpolation formula is not specified.
  • velocity bounds in LS solver = not reported
    Bound-constrained least-squares solver [39] is used for underdetermined clusters; the physical bounds on v_k are never stated.
axioms (6)
  • domain assumption Radar points with compensated Doppler magnitude above delta_dyn are moving objects.
    Eq. (4) uses a single fixed threshold to separate dynamic from static; it fails for slow objects, blindspots, and multipath, as acknowledged in Section VI-A and the conclusion.
  • domain assumption Each radar CCL cluster is a single rigidly moving object.
    Section IV-A solves a per-cluster linear system A v_k = b under this rigidity prior; articulated or merged objects violate it.
  • domain assumption Ego motion F_ego is known exactly from GPS or localization.
    Problem statement in Section III-A says F_ego is 'assumed known through GPS or a localization pipeline'.
  • domain assumption Radar-LiDAR extrinsic calibration is accurate and static.
    The conclusion explicitly states the method assumes this; sensor drift or slight misalignment would introduce systematic velocity errors.
  • domain assumption High-intensity LiDAR points are reliable and low-intensity points are noise unless radar-supported.
    Section IV-B filters by intensity before HDBSCAN; snow and rain can also produce high-intensity returns, making this assumption imperfect.
  • domain assumption Nearest-neighbor radar association within a range-adaptive threshold is a valid correspondence.
    Algorithm 1, lines 4 and 5, associates every LiDAR point to exactly one radar point; this can merge or miss objects when sensors are misaligned, sparse, or occluded.

reviewed 2026-08-05 · how reviews work

0 comments
Cite this review

Pith. "Pith review of DoGFlow: Self-Supervised LiDAR Scene Flow via Cross-Modal Doppler Guidance." pith.science (2026). https://pith.science/paper/UIW67J5U

@misc{pith2026250818506,
  author       = {Pith},
  title        = {Pith review of: DoGFlow: Self-Supervised LiDAR Scene Flow via Cross-Modal Doppler Guidance},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/UIW67J5U}},
  note         = {Machine review of arXiv:2508.18506}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Accurate 3D scene flow estimation is critical for autonomous systems to navigate dynamic environments safely, but creating the necessary large-scale, manually annotated datasets remains a significant bottleneck for developing robust perception models. Current self-supervised methods struggle to match the performance of fully supervised approaches, especially in challenging long-range and adverse weather scenarios, while supervised methods are not scalable due to their reliance on expensive human labeling. We introduce DoGFlow, a novel self-supervised framework that recovers full 3D object motions for LiDAR scene flow estimation without requiring any manual ground truth annotations. This paper presents our cross-modal label transfer approach, where DoGFlow computes motion pseudo-labels in real-time directly from 4D radar Doppler measurements and transfers them to the LiDAR domain using dynamic-aware association and ambiguity-resolved propagation. On the challenging MAN TruckScenes dataset, DoGFlow substantially outperforms existing self-supervised methods and improves label efficiency by enabling LiDAR backbones to achieve over 90% of fully supervised performance with only 10% of the ground truth data. For more details, please visit https://ajinkyakhoche.github.io/DogFlow/

Figures

Figures reproduced from arXiv: 2508.18506 by Ajinkya Khoche, Patric Jensfelt, Qingwen Zhang, Sina Sharif Mansouri, Yixi Cai.

Figure 1
Figure 1. Figure 1: Qualitative scene flow comparison on a real-world [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Radar Doppler anomalies for a rigid object observed by [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Overview of the DoGFlow pipeline. Radar points (larger circles) are first classified as dynamic using ego-motion [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Range-wise breakdown of scene flow performance. [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Label efficiency evaluation comparing SSF trained [PITH_FULL_IMAGE:figures/full_fig_p007_5.png] view at source ↗

discussion (0)

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

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. CaRLi-V: Camera-RADAR-LiDAR Point-Wise 3D Velocity Estimation

    cs.RO 2025-11 unverdicted novelty 7.0

    CaRLi-V fuses RADAR velocity cube, camera optical flow, and LiDAR ranges in a closed-form solution to produce dense point-wise 3D velocity estimates that outperform scene flow methods on a custom dataset.

Reference graph

Works this paper leans on

42 extracted references · 40 canonical work pages · cited by 1 Pith paper · 3 internal anchors

  1. [1]

    Union: Unsupervised 3d object detection using object appearance-based pseudo-classes,

    T. Lentsch, H. Caesar, and D. Gavrila, “Union: Unsupervised 3d object detection using object appearance-based pseudo-classes,” Advances in Neural Information Processing Systems , vol. 37, pp. 22 028–22 046, 2024

    work page 2024

  2. [2]

    Ratrack: moving object detection and tracking with 4d radar point cloud,

    Z. Pan, F. Ding, H. Zhong, and C. X. Lu, “Ratrack: moving object detection and tracking with 4d radar point cloud,” in 2024 IEEE International Conference on Robotics and Automation (ICRA) . IEEE, 2024, pp. 4480–4487

    work page 2024

  3. [3]

    SemanticFlow: A Self-Supervised Framework for Joint Scene Flow Prediction and Instance Segmentation in Dynamic Environments

    Y . Chen, M. Zhang, Q. Hao, and G. Zhou, “Semanticflow: A self- supervised framework for joint scene flow prediction and instance seg- mentation in dynamic environments,” arXiv preprint arXiv:2503.14837, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  4. [4]

    Ofmpnet: Deep end-to-end model for occu- pancy and flow prediction in urban environment,

    Y . Murhij and D. Yudin, “Ofmpnet: Deep end-to-end model for occu- pancy and flow prediction in urban environment,” Neurocomputing, vol. 586, p. 127649, 2024

    work page 2024

  5. [5]

    Scalable scene flow from point clouds in the real world,

    P. Jund, C. Sweeney et al. , “Scalable scene flow from point clouds in the real world,” IEEE Robotics and Automation Letters , vol. 7, no. 2, pp. 1589–1596, 2021

    work page 2021

  6. [6]

    DeFlow: Decoder of scene flow network in autonomous driving,

    Q. Zhang, Y . Yang et al. , “DeFlow: Decoder of scene flow network in autonomous driving,” in 2024 IEEE International Conference on Robotics and Automation (ICRA) , 2024, pp. 2105–2111

    work page 2024

  7. [7]

    Ssf: Sparse long-range scene flow for autonomous driving,

    A. Khoche, Q. Zhang et al. , “Ssf: Sparse long-range scene flow for autonomous driving,” arXiv preprint arXiv:2501.17821 , 2025

    Pith/arXiv arXiv 2025

  8. [8]

    Flow4d: Leveraging 4d voxel network for lidar scene flow estimation,

    J. Kim, J. Woo et al., “Flow4d: Leveraging 4d voxel network for lidar scene flow estimation,” IEEE Robotics and Automation Letters , 2025

    work page 2025

  9. [9]

    Towards long-range 3d object detection for autonomous vehicles,

    A. Khoche, L. P. S ´anchez et al., “Towards long-range 3d object detection for autonomous vehicles,” in 2024 IEEE Intelligent Vehicles Symposium (IV). IEEE, 2024, pp. 2206–2212

    work page 2024

  10. [10]

    ZeroFlow: Scalable Scene Flow via Distillation

    K. Vedder, N. Peri et al., “Zeroflow: Scalable scene flow via distillation,” arXiv preprint arXiv:2305.10424 , 2023

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  11. [11]

    Neural eulerian scene flow fields,

    ——, “Neural eulerian scene flow fields,” in The Thirteenth International Conference on Learning Representations , 2025

    work page 2025

  12. [12]

    Neural scene flow prior,

    X. Li, J. Kaesemodel Pontes, and S. Lucey, “Neural scene flow prior,” Advances in Neural Information Processing Systems , vol. 34, pp. 7838– 7851, 2021

    work page 2021

  13. [13]

    SeFlow: A self-supervised scene flow method in autonomous driving,

    Q. Zhang, Y . Yang et al., “SeFlow: A self-supervised scene flow method in autonomous driving,” in European Conference on Computer Vision (ECCV). Springer, 2024, p. 353–369

    work page 2024

  14. [14]

    Just go with the flow: Self-supervised scene flow estimation,

    H. Mittal, B. Okorn, and D. Held, “Just go with the flow: Self-supervised scene flow estimation,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , 2020, pp. 11 177–11 185

    work page 2020

  15. [15]

    HiMo: High-speed objects motion com- pensation in point cloud,

    Q. Zhang, A. Khoche et al. , “HiMo: High-speed objects motion com- pensation in point cloud,” arXiv preprint arXiv:2503.00803 , 2025

    work page arXiv 2025

  16. [16]

    Self-supervised pillar motion learning for autonomous driving,

    C. Luo, X. Yang, and A. Yuille, “Self-supervised pillar motion learning for autonomous driving,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2021, pp. 3183–3192

    work page 2021

  17. [17]

    3-d scene flow estimation on pseudo-lidar: Bridging the gap on estimating point motion,

    C. Jiang, G. Wang, Y . Miao, and H. Wang, “3-d scene flow estimation on pseudo-lidar: Bridging the gap on estimating point motion,” IEEE Transactions on Industrial Informatics , vol. 19, no. 6, pp. 7346–7354, 2022

    work page 2022

  18. [18]

    Floxels: Fast Unsupervised Voxel Based Scene Flow Estimation

    D. T. Hoffmann, S. H. Raza et al. , “Floxels: Fast unsupervised voxel based scene flow estimation,” arXiv preprint arXiv:2503.04718 , 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  19. [19]

    Survey on lidar perception in adverse weather conditions,

    M. Dreissig, D. Scheuble, F. Piewak, and J. Boedecker, “Survey on lidar perception in adverse weather conditions,” 2023 IEEE Intelligent Vehicles Symposium (IV), pp. 1–8, 2023. [Online]. Available: https://api.semanticscholar.org/CorpusID:258108386

    work page 2023

  20. [20]

    A scalable and accurate de-snowing algorithm for lidar point clouds in winter,

    W. Wang, X. You et al., “A scalable and accurate de-snowing algorithm for lidar point clouds in winter,” Remote Sensing, vol. 14, no. 6, p. 1468, 2022

    work page 2022

  21. [21]

    Lisnownet: Real- time snow removal for lidar point clouds,

    M.-Y . Yu, R. Vasudevan, and M. Johnson-Roberson, “Lisnownet: Real- time snow removal for lidar point clouds,” in 2022 IEEE/RSJ Interna- tional Conference on Intelligent Robots and Systems (IROS) . IEEE, 2022, pp. 6820–6826

    work page 2022

  22. [22]

    Self-supervised scene flow estimation with 4-d automotive radar,

    F. Ding, Z. Pan et al., “Self-supervised scene flow estimation with 4-d automotive radar,” IEEE Robotics and Automation Letters , vol. 7, no. 3, pp. 8233–8240, 2022

    work page 2022

  23. [23]

    nuscenes: A multimodal dataset for autonomous driving,

    H. Caesar, V . Bankiti et al. , “nuscenes: A multimodal dataset for autonomous driving,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , 2020, pp. 11 621–11 631

    work page 2020

  24. [24]

    Argoverse 2: Next generation datasets for self- driving perception and forecasting,

    B. Wilson, W. Qi et al., “Argoverse 2: Next generation datasets for self- driving perception and forecasting,” in Thirty-fifth Conference on Neural Information Processing Systems Datasets and Benchmarks Track

    work page

  25. [25]

    Man truckscenes: A multimodal dataset for autonomous trucking in diverse conditions,

    F. Fent, F. Kuttenreich et al., “Man truckscenes: A multimodal dataset for autonomous trucking in diverse conditions,” Advances in Neural Information Processing Systems , vol. 37, pp. 62 062–62 082, 2024

    work page 2024

  26. [26]

    Icp-flow: Lidar scene flow estimation with icp,

    Y . Lin and H. Caesar, “Icp-flow: Lidar scene flow estimation with icp,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024, pp. 15 501–15 511

    work page 2024

  27. [27]

    Fast neural scene flow,

    X. Li, J. Zheng et al. , “Fast neural scene flow,” in Proceedings of the IEEE/CVF International Conference on Computer Vision , 2023, pp. 9878–9890

    work page 2023

  28. [28]

    Rigidflow: Self-supervised scene flow learning on point clouds by local rigidity prior,

    R. Li, C. Zhang et al., “Rigidflow: Self-supervised scene flow learning on point clouds by local rigidity prior,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2022, pp. 16 959–16 968

    work page 2022

  29. [29]

    Dufomap: Efficient dynamic awareness mapping,

    D. Duberg, Q. Zhang, M. Jia, and P. Jensfelt, “Dufomap: Efficient dynamic awareness mapping,” IEEE Robotics and Automation Letters , 2024

    work page 2024

  30. [30]

    Emernerf: Emergent spatial-temporal scene decomposition via self-supervision,

    J. Yang, B. Ivanovic et al., “Emernerf: Emergent spatial-temporal scene decomposition via self-supervision,” arXiv preprint arXiv:2311.02077 , 2023

    Pith/arXiv arXiv 2023

  31. [31]

    Full-velocity radar returns by radar-camera fusion,

    Y . Long, D. Morris et al., “Full-velocity radar returns by radar-camera fusion,” in Proceedings of the IEEE/CVF International Conference on Computer Vision, 2021, pp. 16 198–16 207

    work page 2021

  32. [32]

    Hidden gems: 4d radar scene flow learning using cross-modal supervision,

    F. Ding, A. Palffy, D. M. Gavrila, and C. X. Lu, “Hidden gems: 4d radar scene flow learning using cross-modal supervision,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023, pp. 9340–9349

    work page 2023

  33. [33]

    Kiss-icp: In defense of point-to-point icp–simple, accurate, and robust registration if done the right way,

    I. Vizzo, T. Guadagnino et al. , “Kiss-icp: In defense of point-to-point icp–simple, accurate, and robust registration if done the right way,”IEEE Robotics and Automation Letters , vol. 8, no. 2, pp. 1029–1036, 2023

    work page 2023

  34. [34]

    Cfar-based interference mitigation for fmcw automotive radar systems,

    J. Wang, “Cfar-based interference mitigation for fmcw automotive radar systems,” IEEE Transactions on Intelligent Transportation Systems , vol. 23, no. 8, pp. 12 229–12 238, 2021

    work page 2021

  35. [35]

    Mcrb on doa estimation for automotive mimo radar in the presence of multipath,

    M. Levy-Israel, I. Bilik, and J. Tabrikian, “Mcrb on doa estimation for automotive mimo radar in the presence of multipath,” IEEE Transactions on Aerospace and Electronic Systems , vol. 59, no. 5, pp. 4831–4843, 2023

    work page 2023

  36. [36]

    Fast rule-based clutter detection in automotive radar data,

    J. Kopp, D. Kellner, A. Piroli, and K. Dietmayer, “Fast rule-based clutter detection in automotive radar data,” in 2021 IEEE International Intelligent Transportation Systems Conference (ITSC). IEEE, 2021, pp. 3010–3017

    work page 2021

  37. [37]

    hdbscan: Hierarchical density based clustering

    L. McInnes, J. Healy, S. Astels et al. , “hdbscan: Hierarchical density based clustering.” J. Open Source Softw. , vol. 2, no. 11, p. 205, 2017

    work page 2017

  38. [38]

    Fully sparse 3d object detection,

    L. Fan, F. Wang, N. Wang, and Z.-X. Zhang, “Fully sparse 3d object detection,” Advances in Neural Information Processing Systems, vol. 35, pp. 351–363, 2022

    work page 2022

  39. [39]

    A subspace, interior, and con- jugate gradient method for large-scale bound-constrained minimization problems,

    M. A. Branch, T. F. Coleman, and Y . Li, “A subspace, interior, and con- jugate gradient method for large-scale bound-constrained minimization problems,” SIAM Journal on Scientific Computing , vol. 21, no. 1, pp. 1–23, 1999

    work page 1999

  40. [40]

    Groundgrid: Lidar point cloud ground segmentation and terrain estimation,

    N. Steinke, D. Goehring, and R. Rojas, “Groundgrid: Lidar point cloud ground segmentation and terrain estimation,” IEEE Robotics and Automation Letters, vol. 9, no. 1, pp. 420–426, 2023

    work page 2023

  41. [41]

    The pascal visual object classes challenge: A retrospective,

    M. Everingham, S. A. Eslami et al. , “The pascal visual object classes challenge: A retrospective,” International journal of computer vision , vol. 111, pp. 98–136, 2015

    work page 2015

  42. [42]

    Re-evaluating lidar scene flow,

    N. Chodosh, D. Ramanan, and S. Lucey, “Re-evaluating lidar scene flow,” in Proceedings of the IEEE/CVF Winter Conference on Applica- tions of Computer Vision (WACV) , January 2024, pp. 6005–6015

    work page 2024

This paper was first reviewed by deepseek-v4-flash on August 5, 2026.