Pith. sign in

REVIEW 5 major objections 4 minor 36 references

SpikingRTNH: Spiking Neural Network for 4D Radar Object Detection

T0 review · 5 major / 4 minor · reviewed 2026-08-09 · deepseek-v4-flash

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

Pith's one-line read SpikingRTNH claims to be the first spiking neural network for 4D Radar 3D object detection, achieving detection accuracy comparable to the ANN baseline while cutting estimated per-frame energy by 78%.

desk verdict First SNN for 4D radar detection, but the headline 78% energy saving rests on a factor-of-ten arithmetic error and a cost model that omits memory traffic, so the energy claims are not established. read the letter →

arxiv 2502.00074 v1 pith:ZGPJPK4C submitted 2025-01-31 cs.CV cs.AIcs.NE

Dong-Hee Paek , Seung-Hyun Kong This is my paper · ORCID
classification cs.CVcs.AIcs.NE
keywords spikingneuralnetwork4Dradar3DobjectdetectionLIFneuronenergyefficiencyautonomousdrivingtop-downinference
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 aims to establish that spiking neural networks (SNNs) can replace conventional artificial neural networks for 4D Radar 3D object detection in autonomous vehicles, cutting energy use without sacrificing accuracy. It converts the RTNH radar detector into SpikingRTNH by swapping every ReLU activation for a leaky integrate-and-fire (LIF) spiking neuron, and adds a biological top-down inference (BTI) schedule that feeds successively sparser point clouds across inference time steps. On the K-Radar test set, SpikingRTNH with BTI reaches 51.1% AP 3D and 57.0% AP BEV, closely matching the ANN baseline of 50.7% and 56.5%, while the paper estimates per-frame energy drops from 7.16 J to 1.58 J, a 78% reduction. A single-time-step version uses 92% less energy at a small accuracy cost. If correct, this makes spike-based radar perception an energy-efficient option for continuous on-vehicle sensing.

What carries the argument

The central machinery is the leaky integrate-and-fire (LIF) neuron, a spiking unit that accumulates incoming currents into a membrane potential and fires an output spike when the potential crosses a threshold, leaking its potential over time. Replacing every ReLU activation in RTNH with LIF neurons converts each multiply-accumulate (MAC) into an accumulate (AC) operation, which the paper prices at 0.9 pJ versus 4.6 pJ per MAC in 45-nm CMOS. Because the spike function's Heaviside step is non-differentiable, training uses a scaled tanh surrogate gradient and a single simulation time step, while inference runs multiple time steps. The second mechanism is biological top-down inference (BTI), which at each time step keeps the top r% of points by radar power, feeding the network a sparser, lower-noise point cloud; the final detection comes from the last time step's feature maps.

What would settle it

Recompute the energy comparison using the paper's own unit costs and a deployment model that includes memory traffic: 156G MACs at 4.6 pJ equals 0.72 J, not 7.16 J, and if the SNN's real measured energy on event-driven hardware is not roughly one-fifth of the ANN's total energy, the 78% claim is refuted.

Watch

Extended reading notes

Core claim

SpikingRTNH is the first spiking neural network for 3D object detection from 4D Radar data. The central discovery is that a direct ReLU-to-LIF conversion of the RTNH architecture, trained with a surrogate-gradient approximation of the spike derivative and run at inference over multiple time steps with progressively sparser point-cloud inputs, produces detection accuracy comparable to the original ANN while drastically lowering the estimated energy cost per frame. With BTI (r = 80% density retention, T = 3 time steps), the network reports 51.1% AP 3D and 57.0% AP BEV on the K-Radar test set versus 50.7% and 56.5% for RTNH, and the energy estimate falls from 7.16 J/frame to 1.58 J/frame, a 78% reduction; without BTI (T = 1) it falls to 0.551 J/frame, a 92% reduction, at 48.1% AP 3D. The paper interprets these numbers as demonstrating that spike-based computing is viable for dense 4D Radar perception in autonomous driving, including under adverse weather.

Load-bearing premise

The paper's 78% energy saving rests on the assumption that every operation in the spiking network costs an idealized 0.9 pJ accumulation and every operation in the ANN costs 4.6 pJ, with no additional energy for memory access, data movement, or control logic in either network.

Editorial extensions

If this is right

  • A radar perception network can run on spike-based hardware with roughly one-fifth the per-frame energy of the ANN version, making continuous 4D Radar processing more feasible for onboard deployment.
  • The energy-accuracy trade-off is controlled by the number of inference time steps T and the density retention ratio r: T = 1 trades 2.6 points of AP 3D for a 92% energy cut, and BTI at T = 3 restores the accuracy for a 78% cut.
  • BTI's density schedule improves over single-step SNN inference by 3.0% AP 3D, indicating that temporally processing progressively sparser radar clouds is itself a source of accuracy gain.
  • The same ReLU-to-LIF conversion recipe transfers from the camera and LiDAR SNN literature to 4D Radar tensor inputs, broadening the input modalities for which spike-based detection is reported.

Reading between the lines

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

  • The conversion recipe and density-schedule idea could plausibly transfer to other point-cloud detectors beyond RTNH, such as pillar- or voxel-based networks on LiDAR, although the paper only demonstrates it for one radar architecture.
  • On true event-driven neuromorphic hardware, the real energy saving might be larger than the paper's 78% estimate, because silent neurons in sparse lower-density clouds consume near-zero dynamic energy; the paper's ideal-CMOS accounting does not count such event-driven effects.
  • A direct test of BTI's mechanism would sweep T beyond 3 and r below 50%: if accuracy keeps rising with T, the benefit is temporal integration; if it saturates or drops, the benefit is specifically noise pruning at the 80% retention level.
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 / 4 minor

Summary. The manuscript proposes SpikingRTNH, the first spiking neural network (SNN) for 3D object detection from 4D Radar data, obtained by replacing ReLU activations in the RTNH architecture with leaky integrate-and-fire (LIF) neurons. It also introduces biological top-down inference (BTI), which sequentially processes successively lower-density point clouds at successive time steps. On the K-Radar test set, SpikingRTNH with BTI is reported to achieve 51.1% AP3D and 57.0% APBEV, comparable to RTNH (50.7% and 56.5%), while reducing inference energy by 78% (1.58 J/frame vs 7.16 J/frame). The paper claims this demonstrates the viability of SNNs for energy-efficient 4D Radar perception.

Significance. If the energy-efficiency claim were established, the paper would make a meaningful contribution by showing that SNNs can match ANN performance on a challenging 4D Radar detection task while consuming substantially less energy. The detection results themselves are plausible and the BTI idea of progressive density reduction is simple but interesting. However, the central quantitative claim is not currently supported: the reported energy numbers contain a factor-of-ten arithmetic error, the operation counts are not reproducible from the released materials, and the cost model omits memory traffic and assumes idealized neuromorphic hardware. The paper also appears to select the BTI hyperparameters on the test set. These issues are fixable but require substantial revision; the paper does not currently establish its headline result.

major comments (5)
  1. [IV-B, Table I] The energy values in Table I and Section IV-B are a factor of 10 too high. For RTNH, 156G MACs at 4.6 pJ/MAC gives 0.718 J, not 7.16 J. For SpikingRTNH with BTI, 7.43G MACs at 4.6 pJ plus 137G ACs at 0.9 pJ gives 0.157 J, not 1.58 J. For the T=1 row, 2.48G MACs plus 48.6G ACs gives 0.055 J, not 0.551 J. The relative reductions (78% and 92%) are unchanged by this correction, but the absolute numbers are wrong, and this arithmetic error calls into question the care with which the energy analysis was performed.
  2. [IV-A, IV-B] The paper states that the thop library was used to measure computational requirements, but thop counts dense MAC/FLOP operations and cannot produce spike-dependent AC counts that depend on the timing and sparsity of spikes. The reported AC counts (48.6G and 137G) are therefore not obtainable from the cited tool. Moreover, the link provided for 'All codes are available' points to the K-Radar dataset repository, not to a SpikingRTNH implementation, so the reported MAC/AC counts are not reproducible from the manuscript or the released code.
  3. [IV-B, Energy Analysis] The energy comparison assigns 4.6 pJ per MAC and 0.9 pJ per AC using ideal 45nm CMOS estimates from Horowitz, but this model omits memory access, data movement, and control logic. For 3D convolutional layers operating on dense radar tensors, memory traffic typically dominates energy, and a single SRAM read can cost tens of pJ, far exceeding the per-operation costs used here. In addition, the actual PyTorch/GPU implementation executes MAC operations in the SNN simulation, so the claimed savings would only be realized on dedicated neuromorphic hardware, which is not described or analyzed. The 78% reduction is thus an idealized upper bound, not a property of any deployed or simulated system.
  4. [IV-C, Table II] The ablation study in Table II appears to be performed on the K-Radar test set, and the BTI hyperparameters r=80% and T=3 are selected based on these test-set results. Tuning on the test set can inflate the reported performance (51.1% AP3D, 57.0% APBEV). A separate validation split should be used for hyperparameter selection, with the test set reserved for a single final evaluation.
  5. [III-B, Section IV-C] The BTI procedure feeds different point clouds (with lower density at each time step) into the same network at successive time steps, and the final detection result is taken from the last time step's features. This is not temporal integration of a single input in the usual SNN sense; the LIF membrane potential is carried across different inputs, which is not formally justified. The interaction between the SNN temporal dynamics and the changing input is unclear, and the paper does not provide an ablation separating the effect of multi-step processing from the effect of density reduction.
minor comments (4)
  1. [Abstract, Footnote] The statement 'All codes are available at https://github.com/kaist-avelab/k-radar' is misleading, as the link points to the K-Radar dataset repository rather than to the proposed SpikingRTNH implementation; please either provide the actual code or state that it is not yet released.
  2. [III-A, Eq. (7)] The decay factor lambda is set to 0.25 in the text with a citation, but it is unclear whether this value is used in all experiments and whether it is optimized or fixed; please clarify.
  3. [II-A] The sentence 'SNNs activate neurons only when spike events occur' is an oversimplification, since the LIF membrane potential is updated at every time step and the reported energy model charges 0.9 pJ per AC regardless of sparsity; consider wording it more precisely.
  4. [IV-C, Table II] The ablation table would be easier to interpret if it included a row for RTNH or a column indicating the relative energy reduction for each configuration, since the absolute MAC and AC counts alone do not make the efficiency trade-off transparent.

Circularity Check

0 steps flagged · score 1.0 of 10

No circular derivation; the 78% energy claim is an external cost-model calculation, not a self-referential fit, though it contains a factor-of-ten arithmetic error.

full rationale

The paper's central derivation chain is not circular. SpikingRTNH is obtained from RTNH by replacing ReLU activations with LIF neurons; detection performance is measured on K-Radar and compared directly with the ANN baseline. BTI is a heuristic sequential subsampling scheme (Eqs. 10-11) whose hyperparameters (r=80%, T=3) are selected by ablation; no equation in the paper defines a prediction in terms of the target quantity. The energy comparison is an accounting model: MAC and AC counts are multiplied by external 45nm CMOS unit costs from Horowitz [21]. The relative 78% figure follows from the ratio of these costs and is not forced by a fitted parameter. However, the absolute energies in Table I are wrong by a factor of ten (156G x 4.6pJ = 0.72J, not 7.16J; 7.43G x 4.6pJ + 137G x 0.9pJ = 0.16J, not 1.58J), and the analysis omits memory-traffic and data-movement energy, which could narrow the estimated gap; these are correctness/reproducibility concerns, not circularity. The reliance on the authors' own RTNH baseline, K-Radar dataset, and prior self-citations is normal and not load-bearing in a circular sense. The main non-circular risks are test-set hyperparameter selection (if the Table II ablations were run on the test set) and the unavailability of SpikingRTNH code (the link points to the K-Radar dataset repository), but neither constitutes a derivation that reduces to its inputs.

Assumptions & free parameters 2 free parameters · 4 assumptions · 0 invented entities

The central claims rest on standard SNN training assumptions, a theoretical energy cost model, and a hypothesized benefit of top-power point selection. No new physical entities are introduced. The main hidden costs are the test-set tuning of BTI hyperparameters and the unverified transfer of 45nm CMOS per-op costs to a real implementation.

free parameters (2)
  • r (density retention ratio) = 80%
    BTI keeps the top 80% of points by power between consecutive time steps. The value was chosen as the best configuration in the ablation table on the K-Radar test set, so the reported 51.1 AP3D is a best-of-grid result.
  • T (number of inference time steps) = 3
    The network is trained with T=1 but evaluated with T=3 in BTI. T was also selected from the test-set ablation table, creating both a tuning and a train/inference mismatch.
assumptions (4)
  • domain assumption The discrete LIF update with surrogate-gradient approximation (Eqs. 7-9) is a valid differentiable proxy for spiking dynamics.
    Adopted from Wu et al. [16] and Ren et al. [34]; the approximation error is not analyzed and the entire training procedure depends on it.
  • domain assumption Per-operation energy constants from 45nm CMOS (4.6 pJ/MAC, 0.9 pJ/AC) transfer directly to the deployed network.
    Used in Section IV-B to convert op counts into J/frame. No hardware measurement, memory energy, data movement, or GPU implementation overhead is included, so the absolute energy claim depends on this assumption.
  • domain assumption Lower-density 4DRPC formed by keeping top-power points is less noisy and more representative of targets.
    This underlies BTI's expected benefit but is supported only by the K-Radar ablations, not by an independent noise model or by a baseline that isolates the effect.
  • domain assumption A network trained at T=1 can be evaluated at T=3 and still produce meaningful feature maps.
    The paper trains with a single time step to avoid gradient issues but runs multi-step inference in BTI; this generalization is assumed rather than validated by a dedicated experiment.

how reviews work

0 comments
Cite this review

Pith. "Pith review of SpikingRTNH: Spiking Neural Network for 4D Radar Object Detection." pith.science (2026). https://pith.science/paper/ZGPJPK4C

@misc{pith2026250200074,
  author       = {Pith},
  title        = {Pith review of: SpikingRTNH: Spiking Neural Network for 4D Radar Object Detection},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ZGPJPK4C}},
  note         = {Machine review of arXiv:2502.00074}
}
read the original abstract

Recently, 4D Radar has emerged as a crucial sensor for 3D object detection in autonomous vehicles, offering both stable perception in adverse weather and high-density point clouds for object shape recognition. However, processing such high-density data demands substantial computational resources and energy consumption. We propose SpikingRTNH, the first spiking neural network (SNN) for 3D object detection using 4D Radar data. By replacing conventional ReLU activation functions with leaky integrate-and-fire (LIF) spiking neurons, SpikingRTNH achieves significant energy efficiency gains. Furthermore, inspired by human cognitive processes, we introduce biological top-down inference (BTI), which processes point clouds sequentially from higher to lower densities. This approach effectively utilizes points with lower noise and higher importance for detection. Experiments on K-Radar dataset demonstrate that SpikingRTNH with BTI significantly reduces energy consumption by 78% while achieving comparable detection performance to its ANN counterpart (51.1% AP 3D, 57.0% AP BEV). These results establish the viability of SNNs for energy-efficient 4D Radar-based object detection in autonomous driving systems. All codes are available at https://github.com/kaist-avelab/k-radar.

Figures

Figures reproduced from arXiv: 2502.00074 by the authors.

Figure 1
Figure 1. 4D Radar system and data representation: (a) Overview of the 4D [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Network architecture comparison between RTNH and SpikingRTNH. Top: RTNH processes a single high-density 4D Radar point cloud using [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Qualitative comparison of 3D object detection results across different weather conditions (Normal, Sleet, and Heavy snow). The first column [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

36 extracted references · 23 canonical work pages

  1. [1]

    4d mmwave radar for autonomous driving perception: A comprehensive survey,

    L. Fan, J. Wang, Y . Chang, Y . Li, Y . Wang, and D. Cao, “4d mmwave radar for autonomous driving perception: A comprehensive survey,” IEEE Transactions on Intelligent Vehicles , vol. 9, no. 4, pp. 4606– 4620, 2024

    work page 2024

  2. [2]

    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,” Advances in Neural Information Processing Systems , vol. 35, pp. 3819–3829, 2022

    2022

  3. [3]

    Automotive radars: A review of signal processing techniques,

    S. M. Patole, M. Torlak, D. Wang, and M. Ali, “Automotive radars: A review of signal processing techniques,” IEEE Signal Processing Magazine, vol. 34, no. 2, pp. 22–35, 2017

    2017

  4. [4]

    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 Vehicles Symposium (IV) . IEEE, 2023, pp. 1–6

    2023

  5. [5]

    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 Vehicles, pp. 1–14, 2024

    work page 2024

  6. [6]

    Dual radar: A multi-modal dataset with dual 4d radar for autononous driving,

    X. Zhang, L. Wang, J. Chen, C. Fang, L. Yang, Z. Song, G. Yang, Y . Wang, X. Zhang, and J. Li, “Dual radar: A multi-modal dataset with dual 4d radar for autononous driving,” arXiv preprint arXiv:2310.07602, 2023

    arXiv 2023

  7. [7]

    Introduction to 4D radar: Hardware, MIMO, signal processing, dataset, and AI,

    D.-H. Paek and J. Guan, “Introduction to 4D radar: Hardware, MIMO, signal processing, dataset, and AI,” in 2024 IEEE Intelligent Vehicles Symposium , June 2024, tutorial. [Online]. Available: https://www.ieee-iv-4dradar.org/

    work page 2024

  8. [8]

    Spiking neural networks: A survey,

    J. D. Nunes, M. Carvalho, D. Carneiro, and J. S. Cardoso, “Spiking neural networks: A survey,” IEEE Access, vol. 10, pp. 60 738–60 764, 2022

    work page 2022

Show all 36 references
  1. [9]

    Direct learning-based deep spiking neural networks: a review,

    Y . Guo, X. Huang, and Z. Ma, “Direct learning-based deep spiking neural networks: a review,” Frontiers in Neuroscience , vol. 17, p. 1209795, 2023

    2023

  2. [10]

    Deep learning in spiking neural networks,

    A. Tavanaei, M. Ghodrati, S. R. Kheradpisheh, T. Masquelier, and A. Maida, “Deep learning in spiking neural networks,” Neural net- works, vol. 111, pp. 47–63, 2019

    2019

  3. [11]

    Are snns really more energy-efficient than anns? an in-depth hardware-aware study,

    M. Dampfhoffer, T. Mesquida, A. Valentian, and L. Anghel, “Are snns really more energy-efficient than anns? an in-depth hardware-aware study,” IEEE Transactions on Emerging Topics in Computational Intelligence, vol. 7, no. 3, pp. 731–741, 2023

    2023

  4. [12]

    Autonomous driving with spiking neural networks,

    R.-J. Zhu, Z. Wang, L. H. Gilpin, and J. Eshraghian, “Autonomous driving with spiking neural networks,” in The Thirty-eighth Annual Conference on Neural Information Processing Systems , 2024

    2024

  5. [13]

    St-p3: End- to-end vision-based autonomous driving via spatial-temporal feature learning,

    S. Hu, L. Chen, P. Wu, H. Li, J. Yan, and D. Tao, “St-p3: End- to-end vision-based autonomous driving via spatial-temporal feature learning,” in European Conference on Computer Vision . Springer, 2022, pp. 533–549

    2022

  6. [14]

    Lanesnns: Spiking neural networks for lane detection on the loihi neuromorphic processor,

    A. Viale, A. Marchisio, M. Martina, G. Masera, and M. Shafique, “Lanesnns: Spiking neural networks for lane detection on the loihi neuromorphic processor,” in 2022 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , 2022, pp. 79–86

    2022

  7. [15]

    Toward neuromorphic perception: Spike-driven lane segmentation for autonomous driving using lidar sensor,

    G. Zhuang, Z. Bing, K. Huang, and A. Knoll, “Toward neuromorphic perception: Spike-driven lane segmentation for autonomous driving using lidar sensor,” in 2023 IEEE 26th International Conference on Intelligent Transportation Systems (ITSC) , 2023, pp. 2448–2453

    2023

  8. [16]

    Direct training for spiking neural networks: Faster, larger, better,

    Y . Wu, L. Deng, G. Li, J. Zhu, Y . Xie, and L. Shi, “Direct training for spiking neural networks: Faster, larger, better,” in Proceedings of the AAAI conference on artificial intelligence , vol. 33, no. 01, 2019, pp. 1311–1318

    2019

  9. [17]

    Top-down influences on visual processing,

    C. D. Gilbert and W. Li, “Top-down influences on visual processing,” Nature Reviews Neuroscience, vol. 14, no. 5, pp. 350–363, 2013

    2013

  10. [18]

    Infant visual attention and object recognition,

    G. D. Reynolds, “Infant visual attention and object recognition,” Behavioral Brain Research, vol. 285, pp. 34–43, May 15 2015

    2015

  11. [19]

    Lapicque’s introduction of the integrate-and-fire model neuron (1907),

    L. F. Abbott, “Lapicque’s introduction of the integrate-and-fire model neuron (1907),” Brain Research Bulletin, vol. 50, no. 5-6, pp. 303–304, 1999

    1907

  12. [20]

    Unsupervised learning of digit recognition using spike-timing-dependent plasticity,

    P. U. Diehl and M. Cook, “Unsupervised learning of digit recognition using spike-timing-dependent plasticity,” Frontiers in Computational Neuroscience, vol. 9, 2015

    2015

  13. [21]

    1.1 computing’s energy problem (and what we can do about it),

    M. Horowitz, “1.1 computing’s energy problem (and what we can do about it),” in 2014 IEEE International Solid-State Circuits Conference Digest of Technical Papers (ISSCC) , 2014, pp. 10–14

    2014

  14. [22]

    Vision meets robotics: The kitti dataset,

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

    2013

  15. [23]

    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

  16. [24]

    Automotive radar dataset for deep learning based 3d object detection,

    M. Meyer and G. Kuschk, “Automotive radar dataset for deep learning based 3d object detection,” in 2019 16th European Radar Conference (EuRAD), 2019, pp. 129–132

    2019

  17. [25]

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

    A. Palffy, E. Pool, S. Baratam, J. F. P. 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

    2022

  18. [26]

    Tj4dradset: A 4d radar dataset for au- tonomous 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 au- tonomous driving,” in 2022 IEEE 25th International Conference on Intelligent Transportation Systems (ITSC). IEEE, 2022, pp. 493–498

    2022

  19. [27]

    Rpfa-net: a 4d radar pillar feature attention network for 3d object detection,

    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 detection,” 09 2021, pp. 3061–3066

    2021

  20. [28]

    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

  21. [29]

    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, 2023

    2023

  22. [30]

    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

  23. [31]

    Very deep convolutional networks for large-scale image recognition,

    K. Simonyan and A. Zisserman, “Very deep convolutional networks for large-scale image recognition,” arXiv preprint arXiv:1409.1556 , 2014

    2014 arXiv

  24. [32]

    Feature pyramid networks for object detection,

    T.-Y . Lin, P. Doll´ar, R. Girshick, K. He, B. Hariharan, and S. Belongie, “Feature pyramid networks for object detection,” in Proceedings of the IEEE conference on computer vision and pattern recognition , 2017, pp. 2117–2125

    2017

  25. [33]

    You only look once: Unified, real-time object detection,

    J. Redmon, “You only look once: Unified, real-time object detection,” in Proceedings of the IEEE conference on computer vision and pattern recognition, 2016

    2016

  26. [34]

    Spik- ing pointnet: Spiking neural networks for point clouds,

    D. Ren, Z. Ma, Y . Chen, W. Peng, X. Liu, Y . Zhang, and Y . Guo, “Spik- ing pointnet: Spiking neural networks for point clouds,” Advances in Neural Information Processing Systems , vol. 36, 2024

    2024

  27. [35]

    Orienting of attention,

    M. I. Posner, “Orienting of attention,” Quarterly Journal of Experi- mental Psychology, vol. 32, no. 1, pp. 3–25, 1980

    1980

  28. [36]

    pytorch-opcounter: Count operations in pytorch models,

    Lyken17, “pytorch-opcounter: Count operations in pytorch models,” https://github.com/Lyken17/pytorch-OpCounter, 2019

    2019

Pith tools

Reviewed August 9, 2026 · model on record in the stance chip above.