Pith. sign in

REVIEW 5 major objections 7 minor 23 references

Enhanced 3D Object Detection via Diverse Feature Representations of 4D Radar Tensor

T0 review · 5 major / 7 minor · reviewed 2026-08-08 · deepseek-v4-flash

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

Pith's one-line read This paper claims that multi-teacher distillation from diverse 4D radar preprocessings lets a sparse-input student detector beat the same-input baseline by 7.3 AP3D and approach 90x denser inputs.

desk verdict A real distillation gain on sparse radar input is buried under an unverified multi-teacher claim and two corrupted ablation tables. read the letter →

arxiv 2502.06114 v3 pith:HL5LPSJC submitted 2025-02-10 cs.CV

Seung-Hyun Song , Dong-Hee Paek , Minh-Quan Dao , Ezio Malis , Seung-Hyun Kong This is my paper · ORCID
classification cs.CV
keywords 4Dradar3Dobjectdetectionknowledgedistillationmulti-teacherpre-processingBEVfeaturedensificationsparsepointcloud
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

The paper tries to show that a lightweight 3D object detector can get most of the benefit of dense 4D radar data while reading only an extremely sparse point cloud. It trains several teacher detectors, each on a different preprocessed view of the same 4D radar tensor, fuses their bird's-eye-view features, and distills that fused knowledge into a student that consumes a 99.9th-percentile point cloud of about 0.1 MB per frame. On the K-Radar sedan class the student reaches 44.16 AP3D and 52.59 APBEV, against 36.84 and 43.10 for the same-backbone RTNH baseline fed identical sparse input, and it comes close to teachers that use roughly 90 times more data. If the claim holds, dense radar pre-processing knowledge can be transferred into a sparse, fast student without paying the memory and compute cost at inference.

What carries the argument

The carrying mechanism is multi-teacher knowledge distillation with feature densification. Three teacher models, each trained on a different 4D radar preprocessing (polar fixed percentile at 80th and 90th thresholds, Cartesian interpolation-based percentile filtering, and CA-CFAR), produce BEV features that are aligned by per-teacher alignment blocks, fused by a CBAM-based attention block with a 1x1 convolution, and then used as the regression target for the student. The student's sparse BEV feature passes through a dual-pass encoder–decoder Densify Module, extended from Sparse2Dense, whose output is combined with the original feature by residual connections; the masked MSE loss in Eq. (2) forces the densified feature to match the fused teacher feature on ground-truth object regions. At inference only the student and its densified feature path remain.

What would settle it

Ablate the Densify Module by feeding the raw student BEV feature into the masked MSE loss while keeping everything else fixed; the paper's mechanism predicts a clear drop in sedan AP3D. If AP3D stays near 44.16, then feature densification is not the load-bearing part of the claimed transfer.

Watch

Extended reading notes

Core claim

The central claim is that complementary signal content hidden in the raw 4D radar tensor—spread across different preprocessing assumptions—can be captured by a set of teacher backbones and recombined into a single feature target for a sparse-input student. The paper reports that the student, using the same RTNH backbone and the same 99.9th-percentile input as the baseline, improves sedan AP3D by 7.3 points and APBEV by 9.5 points, and improves bus/truck AP3D by 6.1 points, with a per-frame input of 0.1 MB; this places it near the RTNH90 teacher that consumes 9 MB per frame. The authors attribute the gain to the combination of a multi-teacher aggregation module and a Densify Module that reconstructs dense student BEV features from sparse evidence, supervised by a ground-truth-masked MSE distillation loss.

Load-bearing premise

The claim rests on the assumption that the Densify Module can turn the student's almost empty BEV features, derived from a point cloud with almost no object-interior points, into features whose object-region similarity to the teachers is meaningful enough for a masked MSE loss to carry real information.

Editorial extensions

If this is right

  • A 0.1 MB-per-frame sparse radar input can deliver detection accuracy close to a 9 MB-per-frame input when trained under multi-teacher distillation, cutting per-frame input data by roughly 90 times.
  • The student alone runs at inference, so the runtime cost stays that of one RTNH backbone while absorbing knowledge from several richer representations.
  • Teacher choice matters: combining the three most diverse preprocessings does not always beat a combination of two percentile teachers and the interpolation teacher, so complementarity, not just diversity, drives the gain.
  • Improvements transfer across object classes: sedan and bus/truck both show higher AP3D and APBEV than the same-input baseline.

Reading between the lines

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

  • A testable extension the paper does not run: mask out the object-region Gaussian from the distillation loss and keep detection loss; if AP3D stays near 44, the distillation target is not the source of the gain.
  • The same recipe could apply to other sensors whose raw data admit multiple valid preprocessings—for example, LiDAR intensity thresholds or camera ISP variants—suggesting a general sparse-student distillation pattern.
  • Because the distillation loss requires ground-truth boxes to build the mask, the method as stated is limited to labeled data; applying it to teacher-ensemble pseudo-labels would test whether the densified features transfer beyond supervised regimes.
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

5 major / 7 minor

Summary. The paper proposes 4DR-MR, a multi-teacher knowledge distillation framework for 3D object detection from 4D radar data. Multiple teacher models are trained on point clouds produced by different radar pre-processing techniques (fixed percentile in polar and Cartesian domains, CA-CFAR), their BEV features are fused by an aggregation module, and the fused representation is distilled into a lightweight student that operates on a very sparse 99.9th-percentile point cloud. Experiments on the K-Radar dataset report 44.16 AP3D and 52.59 APBEV for the sedan class, a large improvement over the RTNH baseline on the same sparse input, while using about 90x less input data than denser preprocessings. The paper also compares against prior 4D radar detectors and provides ablations of the aggregation and densify modules.

Significance. If the claims are correct, the work is a useful contribution: it shows that a sparse-input student can absorb knowledge from multiple dense radar representations through multi-teacher distillation, which addresses a real efficiency-accuracy trade-off in 4D radar perception. The controlled comparison against RTNH on the same sparse input is a strength, and the paper is transparent that teacher choice matters. However, the evidence as presented does not currently establish the central claim: the only fully specified multi-teacher configuration slightly underperforms its best single teacher, the headline multi-teacher result uses an undefined teacher, and the fusion ablation duplicates baseline numbers. The Densify module, which is load-bearing for the mechanism, is not validated with feature-level analysis. Therefore the contribution is currently not substantiated, though the underlying idea remains plausible.

major comments (5)
  1. [Section IV-B2, Table II] The only multi-teacher set whose members are all defined in Section IV-A, namely (1),(2),(3), achieves 44.34 AP3D and 47.21 APBEV, which is lower than the best single-teacher RTNH80 (44.52 AP3D, 47.79 APBEV) and also lower than single-teacher RTNH (44.48 AP3D, 47.52 APBEV). Since the paper's core claim is that diverse multi-teacher distillation improves over single-teacher distillation, this fully specified configuration directly contradicts the claim. The authors need to identify a multi-teacher set that beats all single teachers, or revise the claim to acknowledge that teacher diversity alone is not beneficial and that an unspecified configuration drives the gain.
  2. [Table II, row (1),(1-1),(2)] The teacher label (1-1) is never defined in Section IV-A or Table I. The headline result of 52.59 APBEV is obtained with this undefined teacher, so the gain cannot be attributed to the proposed multi-teacher framework as described. The authors must specify what preprocessing (1-1) corresponds to, why it is absent from the single-teacher ablation, and report its single-teacher performance so that the contribution of teacher diversity can be assessed.
  3. [Table IV and Table III] The fusion ablation reports exactly 36.84 AP3D / 43.10 APBEV for 'Learnable query-based' and 37.58 AP3D / 46.02 APBEV for 'CBAM', which are the same numbers as the RTNH and Radar PillarNet rows in Table III. This is not plausible for independently measured fusion strategies, and it invalidates the conclusion that the proposed Aggregation Module is necessary. The authors must rerun these ablations and report the actual measured values, or the aggregation module's contribution remains unverified.
  4. [Section III-D and Section IV-C2] The Densify Module is the mechanism claimed to bridge the density mismatch between sparse student features and dense teacher features, but the paper provides no quantitative feature-similarity analysis and no ablation that removes the Densify Module. The final AP numbers and the qualitative Figure 5 do not establish that the student is learning from teacher features rather than memorizing teacher priors or simply benefiting from the stronger detection loss. A feature-level comparison or a densify-module ablation is needed to support the stated mechanism.
  5. [Section IV-A2 and Section IV-B] The experimental setup promises: 'Inference speed is reported in frames per second (FPS) to assess the runtime efficiency of our framework.' No FPS or runtime result appears anywhere in the paper. The efficiency claim is based only on input data size (0.1 MB vs. 9 MB per frame), not on actual computational cost. The authors should either report measured inference speed and compare it with dense-input baselines, or remove the promise and temper the efficiency claims.
minor comments (7)
  1. [Algorithm 1, line 10] Line 10 contains 'eg' where 'ele' is intended in the elevation-to-z conversion: 'z = eg · sin(ele)' should read 'z = ele · sin(ele)' or similar.
  2. [All experimental tables] Results are reported without error bars or multiple seeds. Given the small AP differences in Table II (e.g., 44.34 vs. 44.52), variance estimates are essential to know whether the differences are meaningful.
  3. [General] No code release or trained-model release is mentioned. Providing the code and evaluated checkpoints would substantially improve reproducibility, especially given the undefined teacher configuration and the duplicated ablation numbers.
  4. [References] References [13] and [21] appear to be the same paper ('3-D object detection for multiframe 4-D automotive millimeter-wave radar point cloud') with overlapping author lists; please merge them or disambiguate.
  5. [Section III-E, Eq. (1)] The loss weights alpha and beta are fixed to 1 without sensitivity analysis; a small study or justification of these choices would strengthen the paper.
  6. [Section IV-B1, Bus or Truck results] The text reports a 3.42% improvement in APBEV for Bus or Truck, while Table I shows 29.02 vs. 25.60, which is a 3.42 percentage-point difference; please state the metric explicitly as percentage points or relative improvement.
  7. [Figure 5] The qualitative feature-map comparison would benefit from higher-resolution crops and a clear annotation of which parts correspond to the claimed object-relevant regions, so that readers can verify the visual enhancement claimed in the text.

Circularity Check

1 steps flagged · score 3.0 of 10

The central detection result is measured on held-out data and is not circular; however, the aggregation-module ablation contains a construction-level reduction, as its two alternative 'fusion' rows duplicate the RTNH and Radar PillarNet baseline rows of Table III.

  1. renaming known result [Section IV-C.1, Table IV (Ablation Study, Aggregation Module)]
    "Fusion Method AP3D APBEV / Learnable query-based 36.84 43.10 / CBAM 37.58 46.02 / Aggregation module 44.16 52.59 — compared with Table III: RTNH[7] 36.84 43.10; Radar PillarNet[22] 37.58 46.02."

    The two non-proposed fusion rows are numerically identical, to two decimals, to the RTNH and Radar PillarNet rows already reported in Table III. The paper introduces these rows as tested alternative fusion strategies, but the equality shows they are not measured fusion variants; they are the previously reported baseline networks under new labels. Consequently, the claimed superiority of the Aggregation Module in this ablation is produced by relabeling known numbers rather than by an experimental comparison, which is a renaming-of-known-result reduction.

full rationale

Score 3 because the core detection comparison (Table I) is not circular: the student is evaluated on the held-out K-Radar test split, the distillation loss masks with ground truth only during training, and the student at inference is a standalone network. However, one supporting ablation in Table IV is a construction-level reduction: the 'Learnable query-based' and 'CBAM' rows have AP values identical to the RTNH and Radar PillarNet rows in Table III, so the Aggregation Module's advantage in that comparison is produced by renaming known baseline numbers rather than by measuring alternative fusion strategies. I did not classify the undefined teacher label '(1-1)' in Table II as circularity, but it is a missing-definition problem that undercuts the multi-teacher claim: Section IV-A defines only (1) fixed percentile with 80/90 thresholds, (2) RTNH interpolation, and (3) CA-CFAR, and the fully specified combination '(1),(2),(3)' scores 44.34/47.21, below single-teacher RTNH80 at 44.52/47.79. Those issues affect evidentiary weight, not circularity.

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

No invented physical entities are introduced; the only new objects are neural modules (alignment block, fusion block, densify module), which are not hypothesized physical entities. The free parameters are training and preprocessing hyperparameters, not fitted physical constants.

free parameters (5)
  • Loss weights alpha and beta = 1
    Hand-selected in Eq. (1) to balance detection and distillation losses; no sensitivity analysis is provided.
  • Student input percentile = 99.9
    The student operates on the 99.9th percentile point cloud; this sparsity level is the claimed operating point and is chosen rather than derived.
  • Teacher preprocessing thresholds = 80th and 90th percentile, RTNH-style interpolation, CA-CFAR TLP
    Teacher representations come from hand-picked preprocessing configurations, several inherited from cited works; the choice of teacher set affects student performance in Table II.
  • CBAM reduction ratio r = 8
    Set in Appendix A for the channel and spatial attention; no ablation tests other reduction ratios.
  • Number of teachers N = 3
    Set to 3 because of hardware limits (Section IV.A.1), so the effect of adding more or different teachers is untested.
assumptions (5)
  • domain assumption K-Radar provides correct ground-truth 3D boxes and representative weather and road conditions for training and evaluation.
    The entire evaluation rests on this public dataset and its annotations; no independent data is used.
  • domain assumption The RTNH backbone and its AP at 0.3 IoU protocol are a valid baseline and evaluation metric.
    Used as teacher and student backbone and as the baseline; the paper adopts RTNH settings without questioning their biases.
  • ad hoc to paper Different radar preprocessing techniques produce complementary, not merely redundant, feature representations.
    This is the core premise of the method (Section III-C); Table II gives only partial support because a single teacher is best on AP3D.
  • domain assumption Feature-level MSE distillation with ground-truth masking is a meaningful training signal for radar BEV features.
    Eq. (2) assumes that matching student features to masked teacher features improves detection; the paper does not compare against alternative distillation losses.
  • domain assumption The Sparse2Dense densification approach transfers from LiDAR point cloud detection to 4D radar sparse features.
    The Densify Module extends reference [24] without an independent justification that radar sparsity behaves like LiDAR sparsity.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Enhanced 3D Object Detection via Diverse Feature Representations of 4D Radar Tensor." pith.science (2026). https://pith.science/paper/HL5LPSJC

@misc{pith2026250206114,
  author       = {Pith},
  title        = {Pith review of: Enhanced 3D Object Detection via Diverse Feature Representations of 4D Radar Tensor},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/HL5LPSJC}},
  note         = {Machine review of arXiv:2502.06114}
}
read the original abstract

Recent advances in automotive four-dimensional (4D) Radar have enabled access to raw 4D Radar Tensor (4DRT), offering richer spatial and Doppler information than conventional point clouds. While most existing methods rely on heavily pre-processed, sparse Radar data, recent attempts to leverage raw 4DRT face high computational costs and limited scalability. To address these limitations, we propose a novel three-dimensional (3D) object detection framework that maximizes the utility of 4DRT while preserving efficiency. Our method introduces a multi-teacher knowledge distillation (KD), where multiple teacher models are trained on point clouds derived from diverse 4DRT pre-processing techniques, each capturing complementary signal characteristics. These teacher representations are fused via a dedicated aggregation module and distilled into a lightweight student model that operates solely on a sparse Radar input. Experimental results on the K-Radar dataset demonstrate that our framework achieves improvements of 7.3% in AP_3D and 9.5% in AP_BEV over the baseline RTNH model when using extremely sparse inputs. Furthermore, it attains comparable performance to denser-input baselines while significantly reducing the input data size by about 90 times, confirming the scalability and efficiency of our approach.

Figures

Figures reproduced from arXiv: 2502.06114 by the authors.

Figure 1
Figure 1. Comparison of 4D Radar point clouds based on differ [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. 4DR-MR: Overall architecture of the proposed 3D object detection framework utilizing multi-representations of 4D Radar. Features from multiple teacher networks, each using different Radar pre-processing, are fused and distilled into a compact student model for efficient inference [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Radar processing pipeline: Microwave signals are [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Aggregation module: The module consists of a Representation Alignment Block that refines each teacher feature and [PITH_FULL_IMAGE:figures/full_fig_p004_4.png]
Figure 5
Figure 5. Figure 5: Qualitative comparison of BEV feature maps and [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

23 extracted references · 15 canonical work pages

  1. [1]

    Are we ready for autonomous driving? the kitti vision benchmark suite,

    A. Geiger, P. Lenz, and R. Urtasun, “Are we ready for autonomous driving? the kitti vision benchmark suite,” in 2012 IEEE conference on computer vision and pattern recognition . IEEE, 2012, pp. 3354–3361

    2012

  2. [2]

    nuscenes: A multimodal dataset for autonomous driving,

    H. Caesar, V . Bankiti, A. H. Lang, S. V ora, V . E. Liong, Q. Xu, A. Krishnan, Y . Pan, G. Baldan, and O. Beijbom, “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

    2020

  3. [3]

    Robustness of object detectors in degrading weather conditions,

    M. J. Mirza, C. Buerkle, J. Jarquin, M. Opitz, F. Oboril, K.-U. Scholl, and H. Bischof, “Robustness of object detectors in degrading weather conditions,” in 2021 IEEE International Intelligent Transportation Sys- tems Conference (ITSC) . IEEE, 2021, pp. 2719–2724

    work page 2021

  4. [4]

    A benchmark for lidar sensors in fog: Is detection breaking down?

    M. Bijelic, T. Gruber, and W. Ritter, “A benchmark for lidar sensors in fog: Is detection breaking down?” in 2018 IEEE intelligent vehicles symposium (IV) . IEEE, 2018, pp. 760–767

    work page 2018

  5. [5]

    Multi- class road user detection with 3+ 1d radar in the view-of-delft dataset,

    A. Palffy, E. Pool, S. Baratam, J. F. Kooij, and D. M. Gavrila, “Multi- class road user detection with 3+ 1d radar in the view-of-delft dataset,” IEEE Robotics and Automation Letters , vol. 7, no. 2, pp. 4961–4968, 2022

    work page 2022

  6. [6]

    Tj4dradset: A 4d radar dataset for autonomous driving,

    L. Zheng, Z. Ma, X. Zhu, B. Tan, S. Li, K. Long, W. Sun, S. Chen, L. Zhang, M. Wan et al., “Tj4dradset: A 4d radar dataset for autonomous driving,” in 2022 IEEE 25th International Conference on Intelligent Transportation Systems (ITSC) . IEEE, 2022, pp. 493–498

    work page 2022

  7. [7]

    K-radar: 4d radar object detection for autonomous driving in various weather conditions,

    D.-H. Paek, S.-H. Kong, and K. T. Wijaya, “K-radar: 4d radar object detection for autonomous driving in various weather conditions,” Ad- vances in Neural Information Processing Systems , vol. 35, pp. 3819– 3829, 2022. 7 Preprint submitted to arXiv

    work page 2022

  8. [8]

    Rpfa-net: A 4d radar pillar feature attention network for 3d object de- tection,

    B. Xu, X. Zhang, L. Wang, X. Hu, Z. Li, S. Pan, J. Li, and Y . Deng, “Rpfa-net: A 4d radar pillar feature attention network for 3d object de- tection,” in 2021 IEEE International Intelligent Transportation Systems Conference (ITSC). IEEE, 2021, pp. 3061–3066

    work page 2021

Show all 23 references
  1. [9]

    Pointpillars: Fast encoders for object detection from point clouds,

    A. H. Lang, S. V ora, H. Caesar, L. Zhou, J. Yang, and O. Beijbom, “Pointpillars: Fast encoders for object detection from point clouds,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2019, pp. 12 697–12 705

    2019

  2. [10]

    Smurf: Spatial multi-representation fusion for 3d object detection with 4d imaging radar,

    J. Liu, Q. Zhao, W. Xiong, T. Huang, Q.-L. Han, and B. Zhu, “Smurf: Spatial multi-representation fusion for 3d object detection with 4d imaging radar,” IEEE Transactions on Intelligent V ehicles , 2023

    2023

  3. [11]

    Mvfan: Multi-view feature assisted network for 4d radar object detection,

    Q. Yan and Y . Wang, “Mvfan: Multi-view feature assisted network for 4d radar object detection,” in International Conference on Neural Information Processing. Springer, 2023, pp. 493–511

    2023

  4. [12]

    Smiformer: Learning spatial feature representation for 3d object detection from 4d imaging radar via multi-view interactive transformers,

    W. Shi, Z. Zhu, K. Zhang, H. Chen, Z. Yu, and Y . Zhu, “Smiformer: Learning spatial feature representation for 3d object detection from 4d imaging radar via multi-view interactive transformers,” Sensors, vol. 23, no. 23, p. 9429, 2023

    2023

  5. [14]

    Raw high-definition radar for multi-task learning,

    J. Rebut, A. Ouaknine, W. Malik, and P. P ´erez, “Raw high-definition radar for multi-task learning,” in Proceedings of the IEEE/CVF Confer- ence on Computer Vision and Pattern Recognition , 2022, pp. 17 021– 17 030

    2022

  6. [15]

    Adcnet: Learning from raw radar data via distillation,

    B. Yang, I. Khatri, M. Happold, and C. Chen, “Adcnet: Learning from raw radar data via distillation,” arXiv preprint arXiv:2303.11420 , 2023

    2023 arXiv

  7. [16]

    T-fftradnet: Object detection with swin vision transformers from raw adc radar signals,

    J. Giroux, M. Bouchard, and R. Laganiere, “T-fftradnet: Object detection with swin vision transformers from raw adc radar signals,” in Proceed- ings of the IEEE/CVF International Conference on Computer Vision , 2023, pp. 4030–4039

    2023

  8. [17]

    Dpft: Dual perspective fusion trans- former for camera-radar-based object detection,

    F. Fent, A. Palffy, and H. Caesar, “Dpft: Dual perspective fusion trans- former for camera-radar-based object detection,” IEEE TRANSACTIONS ON INTELLIGENT VEHICLES , 2024

    2024

  9. [18]

    Rtnh+: Enhanced 4d radar object detection network using two-level preprocessing and vertical encoding,

    S.-H. Kong, D.-H. Paek, and S. Lee, “Rtnh+: Enhanced 4d radar object detection network using two-level preprocessing and vertical encoding,” IEEE Transactions on Intelligent V ehicles , 2024

    2024

  10. [19]

    Attention is all you need,

    A. Vaswani, “Attention is all you need,” Advances in Neural Information Processing Systems, 2017

    2017

  11. [20]

    Enhanced k-radar: Optimal density reduction to improve detection performance and accessibility of 4d radar tensor-based object detection,

    D.-H. Paek, S.-H. Kong, and K. T. Wijaya, “Enhanced k-radar: Optimal density reduction to improve detection performance and accessibility of 4d radar tensor-based object detection,” in 2023 IEEE Intelligent V ehicles Symposium (IV). IEEE, 2023, pp. 1–6

    2023

  12. [21]

    3-d object detection for multiframe 4-d automotive millimeter-wave radar point cloud,

    B. Tan, Z. Ma, X. Zhu, S. Li, L. Zheng, S. Chen, L. Huang, and J. Bai, “3-d object detection for multiframe 4-d automotive millimeter-wave radar point cloud,” IEEE Sensors Journal , vol. 23, no. 11, pp. 11 125– 11 138, 2022

    2022

  13. [22]

    Rcfusion: Fusing 4-d radar and camera with bird’s-eye view features for 3-d object detection,

    L. Zheng, S. Li, B. Tan, L. Yang, S. Chen, L. Huang, J. Bai, X. Zhu, and Z. Ma, “Rcfusion: Fusing 4-d radar and camera with bird’s-eye view features for 3-d object detection,” IEEE Transactions on Instrumentation and Measurement, vol. 72, pp. 1–14, 2023

    2023

  14. [23]

    Deformable {detr}: Deformable transformers for end-to-end object detection,

    X. Zhu, W. Su, L. Lu, B. Li, X. Wang, and J. Dai, “Deformable {detr}: Deformable transformers for end-to-end object detection,” in International Conference on Learning Representations , 2021

    2021

  15. [24]

    Sparse2dense: Learning to densify 3d features for 3d object detection,

    T. Wang, X. Hu, Z. Liu, and C.-W. Fu, “Sparse2dense: Learning to densify 3d features for 3d object detection,” Advances in Neural Information Processing Systems , vol. 35, pp. 38 533–38 545, 2022. 8

    2022

Pith tools

Reviewed August 8, 2026 · model on record in the stance chip above.