Pith. sign in

REVIEW 2 major objections 69 references

Spiking neural networks detect 3D objects in LiDAR bird's eye view at 92 AP while using 3.33 times less synaptic energy than CNNs.

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 ↗

T0 review · grok-4.3

2026-06-30 11:37 UTC pith:WGVX2VB4

load-bearing objection SNN for LiDAR BEV detection gets decent numbers with learned encodings, but repeating the same frame across timesteps undercuts claims about real event-driven streaming benefits. the 2 major comments →

arxiv 2605.25293 v1 pith:WGVX2VB4 submitted 2026-05-24 cs.CV cs.AIcs.RO

Neuromorphic LiDAR-based Bird's Eye View Object Detection using Energy-efficient Spiking Neural Networks

Sambit Mohapatra , Senthil Yogamani , Heinrich Gotzig , Patrick Mader This is my paper
classification cs.CV cs.AIcs.RO
keywords spiking neural networksbird's eye viewLiDAR object detectionneuromorphic perceptionautonomous drivingsurrogate gradientKITTI benchmarkenergy efficiency
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 paper shows that an end-to-end spiking encoder-decoder network can perform accurate bird's eye view object detection on LiDAR point clouds. Two variants are trained with surrogate gradient descent: one that reads membrane potentials at the output for peak accuracy and one that stays fully binary for direct neuromorphic hardware use. Learned spike encodings from data outperform fixed schemes such as Poisson or latency coding when the same BEV frame is repeated across time steps. A block-wise energy breakdown finds a 3.33-fold drop in synaptic operations relative to a matched convolutional network. These results establish that spiking networks can meet both accuracy and power targets for neuromorphic perception in driving.

Core claim

An end-to-end spiking encoder-decoder network trained by surrogate gradient backpropagation achieves 92.05/87.04/86.51 AP at IoU=0.5 on the KITTI benchmark while delivering a 3.33× reduction in synaptic operation energy compared with an equivalent CNN; a membrane-potential readout variant reaches the quoted accuracy and a fully binary spike-train variant supports direct neuromorphic deployment.

What carries the argument

Surrogate-gradient-trained spiking encoder-decoder that converts BEV LiDAR features into spike trains and decodes them to 3D bounding boxes, with learned input encoding that replaces hand-crafted spike schemes.

Load-bearing premise

Repeating the identical bird's eye view frame across multiple time steps supplies enough temporal structure for the spiking network to use its event-driven dynamics even though the benchmark contains no real sequential data.

What would settle it

Running the identical trained spiking network on actual sequential LiDAR frames recorded at 10 Hz and measuring whether accuracy or the energy advantage collapses relative to the repeated-frame results.

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

If this is right

  • Neuromorphic chips can host perception stacks that stay under strict power budgets while retaining detection performance close to conventional CNNs.
  • Learned spike encodings remove the need for manual design of input representations when moving from static images to event-based sensors.
  • The fully binary variant can be mapped directly to event-driven hardware without additional conversion steps.
  • Block-level energy accounting shows that most savings come from sparsity in the encoder and decoder layers.

Where Pith is reading between the lines

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

  • The same architecture could be tested on streaming data from event cameras or radar to check whether the energy gain persists when real temporal structure is present.
  • Replacing the repeated-frame proxy with actual multi-frame fusion might further improve the moderate and hard AP numbers.
  • The 3.33× energy figure assumes loop-based simulation; direct measurement on a physical neuromorphic processor would give the true deployment cost.

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

2 major / 0 minor

Summary. The paper proposes an end-to-end spiking encoder-decoder network for bird's eye view (BEV) object detection from LiDAR point clouds, trained via surrogate gradient backpropagation. It evaluates two variants on KITTI—a membrane-potential version achieving 92.05/87.04/86.51 AP at IoU=0.5 (Easy/Moderate/Hard) and a fully binary spiking version—along with four input spike-encoding strategies (learned representations outperforming Poisson, latency, and z-axis schemes). A block-wise energy analysis reports a 3.33× reduction in synaptic-operation energy versus an equivalent CNN under conservative loop-based operation, using repeated identical BEV frames across timesteps as a proxy for temporal streaming since KITTI supplies no sequential data.

Significance. If the accuracy and energy results hold under the reported conditions, the work would provide concrete evidence that SNNs can reach competitive detection performance on a standard 3D perception benchmark while offering substantial energy savings, supporting their viability for neuromorphic hardware in autonomous-driving perception pipelines.

major comments (2)
  1. [Abstract] Abstract: The central viability claim for 'neuromorphic perception in autonomous driving' rests on the SNN exploiting event-driven temporal dynamics, yet the setup feeds the identical static BEV map at every timestep because 'sequential frames are unavailable.' This means all temporal evolution arises from internal recurrence rather than sensor-driven input changes, weakening the extrapolation to real streaming LiDAR scenarios where point clouds evolve continuously.
  2. [Abstract] Abstract: The reported 3.33× synaptic-operation energy reduction is obtained from a 'block-wise energy analysis' under 'conservative loop-based operation,' but the manuscript provides no explicit description of the CNN baseline architecture, the precise energy model (e.g., per-synapse cost, accumulation vs. multiply-add), or how loop-based versus event-driven counts are tallied; without these, the factor cannot be independently verified or compared to other SNN energy studies.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback. We address each major comment below and indicate planned revisions.

read point-by-point responses

  1. Referee: [Abstract] Abstract: The central viability claim for 'neuromorphic perception in autonomous driving' rests on the SNN exploiting event-driven temporal dynamics, yet the setup feeds the identical static BEV map at every timestep because 'sequential frames are unavailable.' This means all temporal evolution arises from internal recurrence rather than sensor-driven input changes, weakening the extrapolation to real streaming LiDAR scenarios where point clouds evolve continuously.

    Authors: We acknowledge that repeating the identical BEV frame is a proxy necessitated by the lack of sequential data in KITTI. This still permits the SNN to demonstrate internal temporal dynamics via recurrence. We will revise the abstract to explicitly note the proxy nature of the setup and add a limitations paragraph discussing extrapolation to real streaming LiDAR. revision: partial

  2. Referee: [Abstract] Abstract: The reported 3.33× synaptic-operation energy reduction is obtained from a 'block-wise energy analysis' under 'conservative loop-based operation,' but the manuscript provides no explicit description of the CNN baseline architecture, the precise energy model (e.g., per-synapse cost, accumulation vs. multiply-add), or how loop-based versus event-driven counts are tallied; without these, the factor cannot be independently verified or compared to other SNN energy studies.

    Authors: The referee is correct that the energy model and baseline details are insufficiently specified. We will expand the methods and experimental sections to fully describe the CNN architecture, per-operation energy costs, and the exact counting procedure for synaptic operations under both loop-based and event-driven regimes. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical benchmark results with independent evaluation

full rationale

The paper reports direct empirical outcomes from training an SNN encoder-decoder on KITTI BEV inputs and measuring AP at fixed IoU thresholds plus a block-wise synaptic energy count. These quantities are computed from model outputs on held-out data and from explicit operation tallies; they do not reduce by construction to any fitted parameter or self-referential definition. The repeated-frame proxy is stated explicitly as a methodological choice required by the benchmark, not derived from prior results or self-citations. No equations, uniqueness theorems, or ansatzes are invoked that would create the enumerated circular patterns. The central claims therefore rest on external test-set performance rather than on any internal reduction to the inputs.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

Abstract provides no explicit free parameters, axioms, or invented entities beyond standard SNN training assumptions; ledger is therefore minimal.

axioms (1)
  • domain assumption Surrogate gradient backpropagation enables effective training of deep spiking networks
    Invoked to train the proposed encoder-decoder architecture.

pith-pipeline@v0.9.1-grok · 5829 in / 1275 out tokens · 52874 ms · 2026-06-30T11:37:38.311295+00:00 · methodology

0 comments
read the original abstract

Autonomous driving perception demands accurate and efficient processing of three-dimensional sensor data under strict power constraints. Traditional convolutional neural networks achieve strong detection accuracy but are computationally intensive, limiting their suitability for deployment on resource-constrained neuromorphic platforms. Spiking neural networks offer a compelling alternative through event-driven sparse computation, yet their application to complex real-world perception tasks such as three-dimensional object detection remains limited. In this work, we propose an end-to-end spiking encoder-decoder network for object detection in bird's eye view representations of LiDAR point clouds, trained using surrogate gradient backpropagation. We train two variants: a membrane potential variant that reads continuous neuron state at the output stage for maximum accuracy, achieving $92.05$/$87.04$/$86.51$ AP at $\mathrm{IoU}\!=\!0.5$ (Easy/Moderate/Hard), and, a fully binary spiking variant that operates exclusively on spike trains at every layer for direct neuromorphic deployment. We evaluate four input spike encoding strategies and demonstrate that allowing the network to learn spike representations directly from data outperforms hand-crafted Poisson, latency, and z-axis encoding schemes on the KITTI benchmark, where sequential frames are unavailable and the BEV input is presented repeatedly across timesteps as a proxy for temporal streaming. A block-wise energy analysis demonstrates a $3.33\times$ reduction in synaptic operation energy over an equivalent CNN under conservative loop-based operation. Together, these results demonstrate the viability of spiking neural networks for accurate and energy-efficient neuromorphic perception in autonomous driving.

Figures

Figures reproduced from arXiv: 2605.25293 by Heinrich Gotzig, Patrick Mader, Sambit Mohapatra, Senthil Yogamani.

Figure 1
Figure 1. Figure 1: Network architecture (top): Downsampling Blocks (DB) extract features and reduce spatial resolution while Upsampling Blocks (UB) aggregate multi-scale features and restore spatial resolution. The convolutional layers act as synaptic connections and feature extractors between the Leaky Integrate-and-Fire (LIF) neuron layers. where we < wf reflect that the early window is noisier and should contribute less t… view at source ↗
Figure 2
Figure 2. Figure 2: Pipeline of the proposed Population-Coded Box Regression Loss Lbox: Binary spike trains S (t) are tempo￾rally integrated into a continuous rate map Rˆ box (Eq. 19), then spatially pooled over a k×k neighborhood to form the population consensus Rˆ pop (Eq. 20). The ℓ1 regression loss (Eq. 22) is computed solely at object centers defined by the keypoint mask Y∗ kp. 3) Rotation Classification Loss: Rotation a… view at source ↗
Figure 3
Figure 3. Figure 3: Qualitative comparison of 3D object detection on the KITTI validation set across three diverse scenes: Green boxes denote network predictions. The left column demonstrates the continuous vmem readout, while the right column demonstrates the fully binary spike readout. Red circles highlight characteristic failure modes induced by purely binary spike inference, including false negatives on distant, sparse ob… view at source ↗
Figure 4
Figure 4. Figure 4: Spike activity over time for 5 time steps and spike rate for the keypoint detection head: Each plane shows the binary spike pattern at a single timestep (t = 0 through t = 4), followed by the temporal mean firing rate (avg) and the bird’s-eye-view LiDAR input (BEV). Cyan spots on the BEV highlight spatial regions with elevated average spike activity. 0 27 54 81 108 135 162 189 Epoch 0.00 0.05 0.10 0.15 0.2… view at source ↗
Figure 5
Figure 5. Figure 5: Mean firing rate of the proposed SNN across training epochs: The network learns progressively sparser spike representations. The decoder dominates SNN energy at 78.37%, driven by the large MAC counts of the transposed convolution upsampling blocks operating at progressively higher spatial resolutions. UB3 4 alone accounts for 28.64% of total SNN energy, yet still achieves 2.78× over its CNN equivalent. The… view at source ↗

discussion (0)

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

Reference graph

Works this paper leans on

69 extracted references · 69 canonical work pages · 1 internal anchor

  1. [1]

    Joseph and A

    L. Joseph and A. K. Mondal,Autonomous driving and advanced driver- assistance systems (ADAS): applications, development, legal issues, and testing. CRC Press, 2021

    work page 2021

  2. [2]

    Near-field depth estimation using monocular fisheye camera: A semi-supervised learning approach using sparse lidar data,

    V . R. Kumar, S. Milz, C. Witt, M. Simon, K. Amende, J. Petzold, S. Yogamani, and T. Pech, “Near-field depth estimation using monocular fisheye camera: A semi-supervised learning approach using sparse lidar data,” inCVPR Workshop, vol. 7, 2018, p. 2. 14

    work page 2018

  3. [3]

    Efficient neuromorphic signal processing with Loihi 2,

    G. Orchard, E. P. Frady, D. B. Rubin, S. Sanborn, S. B. Shrestha, F. T. Sommer, and M. Davies, “Efficient neuromorphic signal processing with Loihi 2,” inIEEE Workshop on Signal Processing Systems (SiPS), 2021, pp. 254–259

    work page 2021

  4. [4]

    A million spiking-neuron integrated circuit with a scalable communication network and interface,

    P. A. Merolla, J. V . Arthur, R. Alvarez-Icaza, A. S. Cassidy, J. Sawada, F. Akopyan, B. L. Jackson, N. Imam, C. Guo, Y . Nakamuraet al., “A million spiking-neuron integrated circuit with a scalable communication network and interface,”Science, vol. 345, no. 6197, pp. 668–673, 2014

    work page 2014

  5. [5]

    The SpiNNaker project,

    S. B. Furber, F. Galluppi, S. Temple, and L. A. Plana, “The SpiNNaker project,” vol. 102, no. 5, 2014, pp. 652–665

    work page 2014

  6. [6]

    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),” inIEEE International Solid-State Circuits Conference (ISSCC), 2014, pp. 10–14

    work page 2014

  7. [7]

    LiDAR for autonomous driving: The principles, challenges, and trends for automotive LiDAR and perception systems,

    Y . Li and J. Ibanez-Guzman, “LiDAR for autonomous driving: The principles, challenges, and trends for automotive LiDAR and perception systems,”IEEE Signal Processing Magazine, vol. 37, no. 4, pp. 50–61, 2020

    work page 2020

  8. [8]

    Event- based vision: A survey,

    G. Gallego, T. Delbr ¨uck, G. Orchard, C. Bartolozzi, B. Taba, A. Censi, S. Leutenegger, A. J. Davison, J. Conradt, K. Daniilidiset al., “Event- based vision: A survey,”IEEE transactions on pattern analysis and machine intelligence, vol. 44, no. 1, pp. 154–180, 2020

    work page 2020

  9. [9]

    Bevcar: Camera-radar fusion for bev map and object segmentation,

    J. Schramm, N. V ¨odisch, K. Petek, B. R. Kiran, S. Yogamani, W. Bur- gard, and A. Valada, “Bevcar: Camera-radar fusion for bev map and object segmentation,” in2024 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). IEEE, 2024, pp. 1435–1442

    work page 2024

  10. [10]

    Vision meets robotics: The kitti dataset,

    A. Geiger, P. Lenz, C. Stiller, and R. Urtasun, “Vision meets robotics: The kitti dataset,”The international journal of robotics research, vol. 32, no. 11, pp. 1231–1237, 2013

    work page 2013

  11. [11]

    Training spiking neural networks using lessons from deep learning,

    J. K. Eshraghian, M. Ward, E. O. Neftci, X. Wang, G. Lenz, G. Dwivedi, M. Bennamoun, D. S. Jeong, and W. D. Lu, “Training spiking neural networks using lessons from deep learning,”Proceedings of the IEEE, vol. 111, no. 9, pp. 1016–1054, 2023

    work page 2023

  12. [12]

    V ote3deep: Fast object detection in 3d point clouds using efficient convolutional neural networks,

    M. Engelcke, D. Rao, D. Z. Wang, C. H. Tong, and I. Posner, “V ote3deep: Fast object detection in 3d point clouds using efficient convolutional neural networks,” in2017 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2017, pp. 1355–1361

    work page 2017

  13. [13]

    V oxelnet: End-to-end learning for point cloud based 3d object detection,

    Y . Zhou and O. Tuzel, “V oxelnet: End-to-end learning for point cloud based 3d object detection,” inProceedings of the IEEE conference on computer vision and pattern recognition, 2018, pp. 4490–4499

    work page 2018

  14. [14]

    Second: Sparsely embedded convolutional detection,

    Y . Yan, Y . Mao, and B. Li, “Second: Sparsely embedded convolutional detection,”Sensors, vol. 18, no. 10, p. 3337, 2018

    work page 2018

  15. [15]

    Pointrcnn: 3d object proposal generation and detection from point cloud,

    S. Shi, X. Wang, and H. Li, “Pointrcnn: 3d object proposal generation and detection from point cloud,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2019, pp. 770– 779

    work page 2019

  16. [16]

    Pointnet++: Deep hierarchical feature learning on point sets in a metric space,

    C. R. Qi, L. Yi, H. Su, and L. J. Guibas, “Pointnet++: Deep hierarchical feature learning on point sets in a metric space,”Advances in neural information processing systems, vol. 30, 2017

    work page 2017

  17. [17]

    Pv-rcnn: Point-voxel feature set abstraction for 3d object detection,

    S. Shi, C. Guo, L. Jiang, Z. Wang, J. Shi, X. Wang, and H. Li, “Pv-rcnn: Point-voxel feature set abstraction for 3d object detection,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2020, pp. 10 529–10 538

    work page 2020

  18. [18]

    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,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2019, pp. 12 697–12 705

    work page 2019

  19. [19]

    Pointnet: Deep learning on point sets for 3d classification and segmentation,

    C. R. Qi, H. Su, K. Mo, and L. J. Guibas, “Pointnet: Deep learning on point sets for 3d classification and segmentation,” inProceedings of the IEEE conference on computer vision and pattern recognition, 2017, pp. 652–660

    work page 2017

  20. [20]

    PIXOR: Real-time 3D object detection from point clouds,

    B. Yang, W. Luo, and R. Urtasun, “PIXOR: Real-time 3D object detection from point clouds,” inProceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2018, pp. 7652– 7660

    work page 2018

  21. [21]

    Bevdet- net: Bird’s eye view lidar point cloud based real-time 3d object detection for autonomous driving,

    S. Mohapatra, S. Yogamani, H. Gotzig, S. Milz, and P. Mader, “Bevdet- net: Bird’s eye view lidar point cloud based real-time 3d object detection for autonomous driving,” in2021 IEEE International Intelligent Trans- portation Systems Conference (ITSC). IEEE, 2021, pp. 2809–2815

    work page 2021

  22. [22]

    Lidar-based intensity-aware outdoor 3d object detection,

    A. Y . Naich and J. R. Carri ´on, “Lidar-based intensity-aware outdoor 3d object detection,”Sensors, vol. 24, no. 9, p. 2942, 2024

    work page 2024

  23. [23]

    Lift: Lightweight, fpga-tailored 3d object detection based on lidar data,

    K. Lis, T. Kryjak, and M. Gorgo ´n, “Lift: Lightweight, fpga-tailored 3d object detection based on lidar data,” inInternational Workshop on Design and Architectures for Signal and Image Processing. Springer, 2025, pp. 28–40

    work page 2025

  24. [24]

    Spiking deep convolutional neural networks for energy-efficient object recognition,

    Y . Cao, Y . Chen, and D. Khosla, “Spiking deep convolutional neural networks for energy-efficient object recognition,”International Journal of Computer Vision, vol. 113, no. 1, pp. 54–66, 2015

    work page 2015

  25. [25]

    Fast- classifying, high-accuracy spiking deep networks through weight and threshold balancing,

    P. U. Diehl, D. Neil, J. Binas, M. Cook, S.-C. Liu, and M. Pfeiffer, “Fast- classifying, high-accuracy spiking deep networks through weight and threshold balancing,” in2015 International joint conference on neural networks (IJCNN). ieee, 2015, pp. 1–8

    work page 2015

  26. [26]

    Con- version of continuous-valued deep networks to efficient event-driven networks for image classification,

    B. Rueckauer, I.-A. Lungu, Y . Hu, M. Pfeiffer, and S.-C. Liu, “Con- version of continuous-valued deep networks to efficient event-driven networks for image classification,”Frontiers in neuroscience, vol. 11, p. 682, 2017

    work page 2017

  27. [27]

    Imagenet: A large-scale hierarchical image database,

    J. Deng, W. Dong, R. Socher, L.-J. Li, K. Li, and L. Fei-Fei, “Imagenet: A large-scale hierarchical image database,” in2009 IEEE conference on computer vision and pattern recognition. Ieee, 2009, pp. 248–255

    work page 2009

  28. [28]

    Learning multiple layers of features from tiny images,

    A. Krizhevsky, G. Hintonet al., “Learning multiple layers of features from tiny images,” University of Toronto, Tech. Rep., 2009

    work page 2009

  29. [29]

    Spiking-yolo: spiking neural network for energy-efficient object detection,

    S. Kim, S. Park, B. Na, and S. Yoon, “Spiking-yolo: spiking neural network for energy-efficient object detection,” inProceedings of the AAAI conference on artificial intelligence, vol. 34, no. 07, 2020, pp. 11 270–11 277

    work page 2020

  30. [30]

    Challenges in de- signing datasets and validation for autonomous driving,

    M. Uric ´ar, D. Hurych, P. Krizek, and S. Yogamani, “Challenges in de- signing datasets and validation for autonomous driving,” inProceedings of the International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications (VISAPP), 2019

    work page 2019

  31. [31]

    Spiking neural network for ultralow-latency and high-accurate object detection,

    J. Qu, Z. Gao, T. Zhang, Y . Lu, H. Tang, and H. Qiao, “Spiking neural network for ultralow-latency and high-accurate object detection,”IEEE Transactions on Neural Networks and Learning Systems, vol. 36, no. 3, pp. 4934–4946, 2024

    work page 2024

  32. [32]

    Spikili: A spiking simulation of lidar based real-time object detection for autonomous driving,

    S. Mohapatra, T. Mesquida, M. Hodaei, S. Yogamani, H. Gotzig, and P. M¨ader, “Spikili: A spiking simulation of lidar based real-time object detection for autonomous driving,” in2022 IEEE Intelligent Vehicles Symposium (IV). IEEE, 2022, pp. 1118–1125

    work page 2022

  33. [33]

    Spiking pointcnn: An efficient converted spiking neural network under a flexible framework,

    Y . Tao and Q. Wu, “Spiking pointcnn: An efficient converted spiking neural network under a flexible framework,”Electronics, vol. 13, no. 18, p. 3626, 2024

    work page 2024

  34. [34]

    Pointcnn: Con- volution on x-transformed points,

    Y . Li, R. Bu, M. Sun, W. Wu, X. Di, and B. Chen, “Pointcnn: Con- volution on x-transformed points,” inAdvances in Neural Information Processing Systems (NeurIPS), vol. 31, 2018, pp. 820–830

    work page 2018

  35. [35]

    Efficient converted spiking neural network for 3d and 2d classification,

    S. Lan, M. Zhang, Q. Wuet al., “Efficient converted spiking neural network for 3d and 2d classification,” inProceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), 2023, pp. 10 356– 10 365

    work page 2023

  36. [37]

    Spatio-temporal backpropa- gation for training high-performance spiking neural networks,

    Y . Wu, L. Deng, G. Li, J. Zhu, and L. Shi, “Spatio-temporal backpropa- gation for training high-performance spiking neural networks,”Frontiers in neuroscience, vol. 12, p. 331, 2018

    work page 2018

  37. [38]

    Brain-inspired spiking neural networks for energy-efficient object detection,

    Z. Li, M. Yao, X. Qiuet al., “Brain-inspired spiking neural networks for energy-efficient object detection,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2025

    work page 2025

  38. [39]

    Object detection with spiking neural networks on automotive event data,

    L. Cordone, B. Miramond, and P. Thierion, “Object detection with spiking neural networks on automotive event data,” inInternational Joint Conference on Neural Networks (IJCNN), 2022, pp. 1–8

    work page 2022

  39. [40]

    Autonomous driv- ing with spiking neural networks,

    R.-J. Zhu, Z. Wang, L. Gilpin, and J. K. Eshraghian, “Autonomous driv- ing with spiking neural networks,” inAdvances in Neural Information Processing Systems (NeurIPS), 2024

    work page 2024

  40. [41]

    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,” inAdvances in Neural Information Processing Systems (NeurIPS), vol. 36, 2023

    work page 2023

  41. [42]

    Point-to-spike residual learning for energy-efficient 3D point cloud classification,

    Q. Wu, Q. Zhang, C. Tan, Y . Zhou, and C. Sun, “Point-to-spike residual learning for energy-efficient 3D point cloud classification,” in Proceedings of the AAAI Conference on Artificial Intelligence, vol. 38, 2024, pp. 6092–6099

    work page 2024

  42. [43]

    Spiking PointNet: Spiking neural networks for point clouds,

    H. Ren, Z. Zhao, A. Lombardini, G. Tang, and P. Li, “Spiking PointNet: Spiking neural networks for point clouds,” inAdvances in Neural Information Processing Systems (NeurIPS), 2023

    work page 2023

  43. [44]

    Deep SCNN- based real-time object detection for self-driving vehicles using LiDAR temporal data,

    S. Lian, J. Luo, Z. Zhao, S. Li, S. Yu, and L. Deng, “Deep SCNN- based real-time object detection for self-driving vehicles using LiDAR temporal data,”IEEE Access, vol. 8, pp. 76 903–76 912, 2020

    work page 2020

  44. [45]

    Networks of spiking neurons: the third generation of neural network models,

    W. Maass, “Networks of spiking neurons: the third generation of neural network models,”Neural networks, vol. 10, no. 9, pp. 1659–1671, 1997

    work page 1997

  45. [46]

    The impulses produced by sensory nerve endings,

    E. D. Adrian and Y . Zotterman, “The impulses produced by sensory nerve endings,”The Journal of physiology, vol. 61, no. 4, pp. 465–483, 1926

    work page 1926

  46. [47]

    Rapid visual processing using spike asyn- chrony,

    S. Thorpe and J. Gautrais, “Rapid visual processing using spike asyn- chrony,”Advances in neural information processing systems, vol. 9, 1996. 15

    work page 1996

  47. [48]

    Deep directly- trained spiking neural networks for object detection,

    Q. Su, Y . Chou, Y . Hu, J. Li, S. Mei, Z. Zhang, and G. Li, “Deep directly- trained spiking neural networks for object detection,” inProceedings of the IEEE/CVF International Conference on Computer Vision, 2023, pp. 6555–6565

    work page 2023

  48. [49]

    Information processing with population codes,

    A. Pouget, P. Dayan, and R. Zemel, “Information processing with population codes,”Nature Reviews Neuroscience, vol. 1, pp. 125–132, 2000

    work page 2000

  49. [50]

    Lidar-bevmtn: Real-time lidar bird’s-eye view multi-task perception network for autonomous driving,

    S. Mohapatra, S. Yogamani, V . R. Kumar, S. Milz, H. Gotzig, and P. M ¨ader, “Lidar-bevmtn: Real-time lidar bird’s-eye view multi-task perception network for autonomous driving,”IEEE transactions on intelligent transportation systems, vol. 26, no. 2, pp. 1547–1561, 2025

    work page 2025

  50. [51]

    LiMoSeg: Real- time Bird’s Eye View based LiDAR Motion Segmentation,

    S. Mohapatra, M. Hodaei, S. Yogamani, S. Milzet al., “LiMoSeg: Real- time Bird’s Eye View based LiDAR Motion Segmentation,” inProceed- ings of the 17th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications (VISIGRAPP 2022), 2022

    work page 2022

  51. [52]

    Birdnet+: End-to- end 3d object detection in lidar bird’s eye view,

    A. Barrera, C. Guindel, J. Beltr ´an, and F. Garc ´ıa, “Birdnet+: End-to- end 3d object detection in lidar bird’s eye view,” in2020 IEEE 23rd International Conference on Intelligent Transportation Systems (ITSC). IEEE, 2020, pp. 1–6

    work page 2020

  52. [53]

    U-net: Convolutional networks for biomedical image segmentation,

    O. Ronneberger, P. Fischer, and T. Brox, “U-net: Convolutional networks for biomedical image segmentation,” inInternational Conference on Medical image computing and computer-assisted intervention. Springer, 2015, pp. 234–241

    work page 2015

  53. [54]

    Focal loss for dense object detection,

    T.-Y . Lin, P. Goyal, R. Girshick, K. He, and P. Doll ´ar, “Focal loss for dense object detection,” inProceedings of the IEEE International Conference on Computer Vision (ICCV), 2017, pp. 2980–2988

    work page 2017

  54. [55]

    Cornernet: Detecting objects as paired key- points,

    H. Law and J. Deng, “Cornernet: Detecting objects as paired key- points,” inProceedings of the European Conference on Computer Vision (ECCV), 2018, pp. 734–750

    work page 2018

  55. [56]

    V-Net: Fully convolutional neural networks for volumetric medical image segmentation,

    F. Milletari, N. Navab, and S.-A. Ahmadi, “V-Net: Fully convolutional neural networks for volumetric medical image segmentation,” inInter- national Conference on 3D Vision (3DV), 2016, pp. 565–571

    work page 2016

  56. [57]

    Generalised dice overlap as a deep learning loss function for highly unbalanced segmentations,

    C. H. Sudre, W. Li, T. Vercauteren, S. Ourselin, and M. J. Cardoso, “Generalised dice overlap as a deep learning loss function for highly unbalanced segmentations,” inDeep Learning in Medical Image Analysis and Multimodal Learning for Clinical Decision Support (DLMIA), 2017, pp. 240–248

    work page 2017

  57. [58]

    Information processing with population codes,

    A. Pouget, P. Dayan, and R. Zemel, “Information processing with population codes,”Nature Reviews Neuroscience, vol. 1, no. 2, pp. 125– 132, 2000

    work page 2000

  58. [59]

    Surrogate gradient learning in spiking neural networks: Bringing the power of gradient-based opti- mization to spiking neural networks,

    E. O. Neftci, H. Mostafa, and F. Zenke, “Surrogate gradient learning in spiking neural networks: Bringing the power of gradient-based opti- mization to spiking neural networks,”IEEE Signal Processing Magazine, vol. 36, no. 6, pp. 51–63, 2019

    work page 2019

  59. [60]

    Training deep spiking neural networks using backpropagation,

    J. H. Lee, T. Delbruck, and M. Pfeiffer, “Training deep spiking neural networks using backpropagation,” inFrontiers in Neuroscience, vol. 10, 2016, p. 508

    work page 2016

  60. [61]

    Rethinking the inception architecture for computer vision,

    C. Szegedy, V . Vanhoucke, S. Ioffe, J. Shlens, and Z. Wojna, “Rethinking the inception architecture for computer vision,” inProceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2016, pp. 2818–2826

    work page 2016

  61. [62]

    Spikeclouds: Streaming spike-based processing of lidar for fast and efficient object detection,

    M. Neumeier, N. Fasfous, B. Li, and A. von Arnim, “Spikeclouds: Streaming spike-based processing of lidar for fast and efficient object detection,”IEEE Robotics and Automation Letters, 2025

    work page 2025

  62. [63]

    Monocular 3D object detection for autonomous driving,

    X. Chen, K. Kundu, Z. Zhang, H. Ma, S. Fidler, and R. Urtasun, “Monocular 3D object detection for autonomous driving,” inPro- ceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2016, pp. 2147–2156

    work page 2016

  63. [64]

    Adam: A Method for Stochastic Optimization

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

    work page internal anchor Pith review Pith/arXiv arXiv 2014

  64. [65]

    SGDR: Stochastic gradient descent with warm restarts,

    I. Loshchilov and F. Hutter, “SGDR: Stochastic gradient descent with warm restarts,” inInternational Conference on Learning Representations (ICLR), 2017

    work page 2017

  65. [66]

    nuScenes: A multimodal dataset for autonomous driving,

    H. Caesar, V . Bankiti, A. H. Lang, S. V ora, V . E. Liong, Q. Xu, A. Kr- ishnan, Y . Pan, G. Baldan, and O. Beijbom, “nuScenes: A multimodal dataset for autonomous driving,” inProceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2020, pp. 11 621– 11 631

    work page 2020

  66. [67]

    A survey of encoding techniques for signal processing in spiking neural networks,

    D. Auge, J. Hille, E. Mueller, and A. Knoll, “A survey of encoding techniques for signal processing in spiking neural networks,”Neural Processing Letters, vol. 53, pp. 4395–4429, 2021

    work page 2021

  67. [68]

    Towards spike-based machine intelligence with neuromorphic computing,

    K. Roy, A. Jaiswal, and P. Panda, “Towards spike-based machine intelligence with neuromorphic computing,”Nature, vol. 575, pp. 607– 617, 2019

    work page 2019

  68. [69]

    A wafer-scale neuromorphic hardware system for large-scale neural modeling,

    J. Schemmel, D. Br ¨uderle, A. Gr¨ubl, M. Hock, K. Meier, and S. Millner, “A wafer-scale neuromorphic hardware system for large-scale neural modeling,” inProceedings of the IEEE International Symposium on Circuits and Systems (ISCAS), 2010, pp. 1947–1950

    work page 2010

  69. [70]

    SpikiLi: A spiking based LiDAR point cloud object detection model for autonomous driving,

    X. Li, P. Bhatt, R. Li, W. Zhang, and U. Bhatt, “SpikiLi: A spiking based LiDAR point cloud object detection model for autonomous driving,” inIEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2023

    work page 2023