Pith. sign in

REVIEW 4 major objections 4 minor 32 references

Vibe2Spike: Batteryless Wireless Tags for Vibration Sensing with Event Cameras and Spiking Networks

T0 review · 4 major / 4 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read Vibe2Spike claims that a sub-$1 tag made of a piezo disc, a Zener diode, and an LED can convert device vibrations into light pulses that an event camera and an evolved spiking neural network classify with 94.9% average F1 accuracy.

desk verdict The 94.9% claim is a best-of-999 validation score, not a tested accuracy; the hardware idea is neat but the evaluation lacks a proper test set and error bars. read the letter →

arxiv 2508.11640 v1 pith:NN53LE7I submitted 2025-07-31 eess.SP cs.AIcs.HCcs.LG

Danny Scott , William LaForest , Hritom Das , Ioannis Polykretis , Catherine D. Schuman , Charles Rizzo , James Plank , Sai Swaminathan This is my paper
classification eess.SPcs.AIcs.HCcs.LG
keywords spikingneuralnetworksvisiblelightcommunicationbattery-freesensingeventcamerasvibrationpiezoelectricenergyharvestingevolutionaryoptimizationedgeAI
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

Vibe2Spike proposes an end-to-end, battery-free vibration sensing pipeline: a sub-$1 tag made of a piezoelectric disc, a Zener diode, and an LED harvests mechanical vibration and emits sparse pulses of visible light whose timing pattern carries the device's identity. An event camera records those pulses, and a spiking neural network evolved by the EONS framework classifies which device is vibrating. The paper reports 94.9% average classification fitness across five device classes, although the evaluation section describes only four handheld devices, leaving the class count inconsistent. The result matters because, if it holds, dense smart-environment sensing could run on tags with no batteries, radios, or microcontrollers, communicating over light instead of RF.

What carries the argument

At the core is the Vibe2Spike tag: a piezoelectric disc in parallel with a Zener diode and an LED, housed in a 3D-printed case with a dual-spring mechanism that mechanically amplifies vibration so the piezo's weak current can drive the LED. The Zener clamps voltage and rectifies the AC signal, so the LED fires only on forward-biased surges, producing sparse optical spikes whose timing encodes the vibration pattern. On the receiver side, an event camera reports only pixel-level brightness changes, generating asynchronous ON/OFF events; a temporal binning pipeline aggregates these events into time windows, and the Argyle-4 scheme encodes the binned events for input to a spiking neural network. The SNN uses an encoder-decoder architecture optimized by EONS, an evolutionary framework that searches network topology and parameters, with classification accuracy as the fitness metric. The paper shows that bin size is a key tuning knob: 250 ms windows give 81.16% accuracy, 2.5 s windows give 94.88%, and 5 s windows give 93.52%.

What would settle it

Mount the same tag design on the same device multiple times with different mounting torques and orientations, record the event streams, and check whether the per-class temporal signatures cluster consistently; if class separation collapses across mounts or tag instances, the 94.9% result is a property of the lab setup, not the tag. A second test: run the SNN at inference on the full pixel field without the best-pixel selection heuristic, and compare accuracy against the reported number.

Watch

Extended reading notes

Core claim

On the paper's own terms, the discovery is that a passive tag containing only a piezo disc, a Zener diode, and an LED can convert ambient mechanical vibration into a detectable optical signature, and that this signature is informative enough for an evolved spiking neural network to tell different vibrating tools apart. The tag requires no battery, no RF transmitter, and no microcontroller; mechanical amplification through a dual-spring mount boosts the piezo's output so that vibration-driven current can flash the LED. An event camera at roughly 30 cm captures the sparse light spikes as asynchronous ON/OFF events, which are grouped into temporal bins and encoded for the SNN. Across four tested devices, the best configuration (2.5 second bins) achieves 94.9% F1 accuracy, with shorter bins trading accuracy for lower latency. The paper frames this as the first integration of battery-free sensing, visible light communication, and neuromorphic processing into one deployable system.

Load-bearing premise

The system's central bet is that the tag's vibration-to-light conversion produces a consistent, repeatable temporal signature for each device class, and that this signature survives changes in mounting, vibration strength, and tag instance; the current evaluation samples only one tag, four devices, a fixed camera, and five trials per device.

Editorial extensions

If this is right

  • Battery-free, radio-free tags could give everyday tools and appliances a low-cost way to announce their identity and activity to a nearby event camera, removing battery maintenance from dense sensing deployments.
  • Visible light communication confines the signal to line-of-sight, so the sensing channel is resistant to RF congestion and can be privacy-controlled simply by covering the LED.
  • The measured latency-accuracy trade-off gives system designers a direct choice: a 250 ms window for fast alerts at 81% accuracy, or a 2.5 s window for near-95% reliability when decisions can wait.
  • Because the tag is just three passive components plus a mount, the marginal cost of adding another sensed device is close to the cost of the tag itself, which the paper puts below one dollar.

Reading between the lines

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

  • If the single-pixel selection heuristic used during training is not actually matched at inference, the stated accuracy may depend on a cleaner signal than the full camera field provides; a direct test of pixel-parallel inference would settle this.
  • The discrepancy between the abstract's five device classes and the four devices in the evaluation section suggests either an unstated fifth class (such as a no-vibration baseline) or an over-count; the reported 94.9% should be read as applying to the four shown classes unless clarified.
  • The paper evaluates one tag instance, one camera distance, and five trials per device; a stress test with multiple tag instances, varied mounting angles, and weaker vibration levels would show whether the learned temporal signatures are device-specific or setup-specific.
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

4 major / 4 minor

Summary. The paper introduces Vibe2Spike, a battery-free vibration-sensing tag built from a piezoelectric disc, a Zener diode, and an LED, whose light pulses are captured by an event camera and classified by SNNs evolved with the EONS framework. The central claim is that this pipeline achieves 94.9% average classification fitness across five device classes, with temporal binning as a key latency-accuracy trade-off. The evaluation is based on four handheld devices, five trials per device, a single tag instance, and a fixed camera distance of about 30 cm.

Significance. If the evaluation were statistically sound, the hardware contribution would be notable: a sub-dollar, batteryless, RF-free tag that converts mechanical vibration into optical spikes readable by an event camera, combined with an evolved SNN classifier, is a genuinely interesting integration. The system builds on the authors' prior NeuroCamTags and EONS work, and the paper is strongest when it describes the physical tag and the event-camera pipeline. However, the headline accuracy does not currently support the deployability claims because it is a validation score selected from 999 EONS runs with no independent test set, and the paper contains a four-versus-five device-class inconsistency. The reported result is plausible but not yet an unbiased estimate of real-world performance.

major comments (4)
  1. [Section IV-C, Table I] The 94.9% figure is the validation accuracy of the best model selected from a population of 999 randomly initialized SNNs, with no held-out test set described anywhere in Sections IV or V. Selecting the maximum validation score over 999 models and reporting that same validation score introduces winner's-curse bias, especially with only five trials per device. The manuscript must either introduce a proper train/validation/test split (or nested cross-validation) and report the test accuracy, or report the mean and standard deviation over multiple evolutionary runs. Without this, the abstract's central quantitative claim cannot be interpreted as an unbiased estimate.
  2. [Section IV-A vs. abstract and Fig. 6] The paper says 'four common devices' in Section IV-A and shows four tools in Fig. 4, while the abstract, the conclusion, and the Fig. 6 caption refer to 'five device classes' or 'five classes.' This is not a cosmetic issue: the confusion matrix in Fig. 6 is described as containing five classes, but the device list and the per-class averaging are ambiguous. The authors should specify whether the fifth class is a background or negative class, or correct the counts throughout; the current inconsistency makes the reported per-class performance impossible to interpret.
  3. [Section V, pixel-selection heuristic] The claim that the trained SNN 'generalizes across pixels' is asserted without any supporting experiment. The training-time heuristic of selecting the pixel with the highest cumulative event count is explicitly described as a way to improve label fidelity, but no evidence is given that a network trained on one pixel transfers to other pixels or to a full-field, pixel-parallel pipeline. Since this generalization is invoked to justify the inference-time deployment claim, the authors should provide an experiment that trains on one pixel and evaluates on held-out pixels, or alternatively restrict the deployment claim to the single-pixel mode actually validated.
  4. [Section IV-A, Section IV-C] The evaluation scale is very small: one tag instance, four devices, five trials per device, and a fixed camera at approximately 30 cm. There are no error bars, no confidence intervals, and no comparison baseline, so it is unclear whether the reported F1 scores are stable or whether a simple baseline such as event-count histograms would achieve similar performance. The 'deployable system' claim in the conclusion needs either additional data across multiple tag instances, mountings, and camera geometries, or a more cautious statement that these results are a proof-of-concept from a single lab setup.
minor comments (4)
  1. [Section V] The sentence beginning 'With an average F1 accuracy of 94.9' is incomplete and should be finished or merged with the following paragraph. Also, the Table I title 'F1-ACCURACYVALIDATIONSCORES' is missing spaces, and 'F1 accuracy' conflates F1 score with accuracy; the paper should report both metrics explicitly if both are intended.
  2. [Section IV-A] The text says 'five trials per device, yielding a total of 25 minutes of labeled event data,' but with four devices this would be 20 trials unless some trials are longer than others; please clarify the per-trial duration and the total recording time.
  3. [References] References [4] and [29] are the same EONS paper and should be consolidated into a single citation to avoid duplication.
  4. [Section I] The introduction contains a typo where 'we proposeVibe2Spike' lacks a space between 'propose' and 'Vibe2Spike'; the manuscript should also be checked for similar missing-space artifacts from LaTeX source.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the 94.9% accuracy is a measured validation outcome, not a derived constant, and self-citations are context, not load-bearing.

full rationale

The paper's central claim is an empirical measurement: Vibe2Spike tags convert vibration into light pulses, an event camera captures those pulses, and an evolved spiking neural network classifies the resulting temporal bin features. The 94.9% figure in Table I is a reported validation score from an experiment, not a parameter fitted into a theory and then renamed as a prediction. No equation in the paper defines the classification accuracy in terms of the tag's circuit, the camera settings, or the SNN's inputs in a way that would make the result true by construction. The self-citations to NeuroCamTags [27] and EONS [4, 29] are used as prior tools and context: NeuroCamTags informs the claim that the LED remains detectable at oblique angles, and EONS is the evolutionary framework used to optimize the SNNs. Neither citation is invoked to prove the accuracy result itself, and neither is a uniqueness theorem that forbids alternative explanations. The heuristic of selecting the pixel with the highest cumulative event count during training and the assertion that the SNN 'generalizes across pixels' are unverified, but those are external-validity or experimental-soundness concerns, not circular reasoning. Additionally, selecting the best validation model among 999 evolved networks and reporting that validation score is a model-selection bias concern, not a case where the prediction reduces to its inputs. Because the evaluation is a direct measurement against real device classes and the reported accuracy is not an algebraic consequence of any fitted quantity, the derivation chain is not circular.

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

The ledger is empty of invented entities because the paper introduces no new physical objects beyond a hardware tag made of off-the-shelf parts. The free parameters listed are the user-chosen and evolution-fitted quantities that the 94.9% result depends on.

free parameters (4)
  • Temporal bin size (latency window) = 250 ms, 500 ms, 2,500 ms, 5,000 ms
    Chosen by hand as a swept design parameter; the 2,500 ms bin is selected because it gives the best reported F1, so the headline number is conditional on this choice.
  • EONS evolutionary hyperparameters = population size 999; other parameters unspecified
    The SNN architectures and weights are evolved with classification accuracy as fitness; the reported model is the best of the population, and the unspecified hyperparameters affect reproducibility.
  • Training-time pixel selection heuristic = pixel with highest cumulative event count
    Used to isolate the LED signal during training; the paper asserts generalization to other pixels without a supporting experiment.
  • Argyle-4 spike encoding parameters
    The encoder settings are described by name but not specified in the visible text, so the feature representation is under-determined.
assumptions (4)
  • domain assumption The piezoelectric disc generates enough electrical energy to drive the LED only with mechanical amplification, and the dual-spring mechanism provides that amplification.
    Relies on the physical characterization of the CEB-44D06 piezo and spring assembly; no measured electrical characterization is included in the visible text.
  • domain assumption An event camera at roughly 30 cm with up to 85 degrees of angular offset reliably captures the LED's optical spikes.
    The paper cites prior NeuroCamTags work for this tolerance rather than measuring it in this setup.
  • domain assumption The vibration signatures of the selected devices are stable and separable across trials.
    Only five trials per device are collected; the paper itself notes performance depends on vibration strength and consistency, which may vary across contexts.
  • ad hoc to paper The SNN trained on the single highest-event pixel generalizes to other pixels at inference.
    Stated in Section V without an experiment; it is a training-time heuristic treated as harmless at inference.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Vibe2Spike: Batteryless Wireless Tags for Vibration Sensing with Event Cameras and Spiking Networks." pith.science (2026). https://pith.science/paper/NN53LE7I

@misc{pith2026250811640,
  author       = {Pith},
  title        = {Pith review of: Vibe2Spike: Batteryless Wireless Tags for Vibration Sensing with Event Cameras and Spiking Networks},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/NN53LE7I}},
  note         = {Machine review of arXiv:2508.11640}
}
read the original abstract

The deployment of dense, low-cost sensors is critical for realizing ubiquitous smart environments. However, existing sensing solutions struggle with the energy, scalability, and reliability trade-offs imposed by battery maintenance, wireless transmission overhead, and data processing complexity. In this work, we present Vibe2Spike, a novel battery-free, wireless sensing framework that enables vibration-based activity recognition using visible light communication (VLC) and spiking neural networks (SNNs). Our system uses ultra-low-cost tags composed only of a piezoelectric disc, a Zener diode, and an LED, which harvest vibration energy and emit sparse visible light spikes without requiring batteries or RF radios. These optical spikes are captured by event cameras and classified using optimized SNN models evolved via the EONS framework. We evaluate Vibe2Spike across five device classes, achieving 94.9\% average classification fitness while analyzing the latency-accuracy trade-offs of different temporal binning strategies. Vibe2Spike demonstrates a scalable, and energy-efficient approach for enabling intelligent environments in a batteryless manner.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

32 extracted references · 25 canonical work pages

  1. [1]

    Coexistence evaluation of densely deployed ble-based body area networks,

    Q. D. La, D. Nguyen-Nam, M. V . Ngo, and T. Q. S. Quek, “Coexistence evaluation of densely deployed ble-based body area networks,” inGLOBECOM 2017 - 2017 IEEE Global Communications Conference. IEEE Press, 2017, p. 1–6. [Online]. Available: https://doi.org/10.1109/GLOCOM.2017.8253947

    work page arXiv 2017

  2. [2]

    High-rate wifi broadcasting in crowded scenarios via lightweight coordination of multiple access points,

    H. Qiu, K. Psounis, G. Caire, K. M. Chugg, and K. Wang, “High-rate wifi broadcasting in crowded scenarios via lightweight coordination of multiple access points,” inProceedings of the 17th ACM International Symposium on Mobile Ad Hoc Networking and Computing, ser. MobiHoc ’16. New York, NY , USA: Association for Computing Machinery, 2016, p. 301–310. [Onl...

    work page arXiv 2016

  3. [4]

    Evolutionary optimization for neuromorphic systems,

    C. D. Schuman, J. P. Mitchell, R. M. Patton, T. E. Potok, and J. S. Plank, “Evolutionary optimization for neuromorphic systems,” inProceedings of the 2020 Annual Neuro-Inspired Computational Elements Workshop, 2020, pp. 1–9

    work page 2020

  4. [5]

    Design of an rfid-based battery-free programmable sensing platform,

    A. P. Sample, D. J. Yeager, P. S. Powledge, A. V . Mamishev, and J. R. Smith, “Design of an rfid-based battery-free programmable sensing platform,”IEEE transactions on instrumentation and measurement, vol. 57, no. 11, pp. 2608–2615, 2008

    2008

  5. [6]

    of Massachusetts Amherst

    U. of Massachusetts Amherst. (2023) Unichmoo. Uni- versity of Massachusetts Amherst. [Online]. Available: https://github.com/spqr/umichmoo

    work page 2023

  6. [7]

    Hybrid analog- digital backscatter platform for high data rate, battery-free sensing,

    V . Talla, M. Buettner, D. Wetherall, and J. R. Smith, “Hybrid analog- digital backscatter platform for high data rate, battery-free sensing,” in2013 IEEE Topical Conference on Wireless Sensors and Sensor Networks (WiSNet), IEEE. Unknown: IEEE, 2013, pp. 1–3

    work page 2013

  7. [8]

    Tip-tap: battery-free discrete 2d fingertip input,

    K. Katsuragawa, J. Wang, Z. Shan, N. Ouyang, O. Abari, and D. V ogel, “Tip-tap: battery-free discrete 2d fingertip input,” inProceedings of the 32nd Annual ACM Symposium on User Interface Software and Technology. New York, NY , USA: ACM, 2019, pp. 1045–1057

    work page 2019

  8. [9]

    RF Bandaid: A Fully-Analog and Passive Wireless Interface for Wearable Sensors,

    V . Ranganathan, S. Gupta, J. Lester, J. R. Smith, and D. Tan, “RF Bandaid: A Fully-Analog and Passive Wireless Interface for Wearable Sensors,”Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies, vol. 2, no. 2, pp. 1–21, Jul. 2018. [Online]. Available: https://dl.acm.org/doi/10.1145/3214282

    work page doi:10.1145/3214282 2018

Show all 32 references
  1. [10]

    Livetag: Sensing human-object interaction through passive chipless wi-fi tags,

    C. Gao, Y . Li, and X. Zhang, “Livetag: Sensing human-object interaction through passive chipless wi-fi tags,”GetMobile: Mobile Computing and Communications, vol. 22, no. 3, pp. 32–35, 2019

    2019

  2. [11]

    LoRa Backscatter: Enabling The Vision of Ubiquitous Connectivity,

    V . Talla, M. Hessar, B. Kellogg, A. Najafi, J. R. Smith, and S. Gollakota, “LoRa Backscatter: Enabling The Vision of Ubiquitous Connectivity,” Proc. ACM Interact. Mob. Wearable Ubiquitous Technol., vol. 1, no. 3, pp. 1–24, Sep. 2017

    2017

  3. [12]

    Pible: battery-free mote for perpetual indoor BLE applications,

    F. Fraternali, B. Balaji, Y . Agarwal, L. Benini, and R. Gupta, “Pible: battery-free mote for perpetual indoor BLE applications,” inProceedings of the 5th Conference on Systems for Built Environments, ser. BuildSys ’18. New York, NY , USA: Association for Computing Machinery, ...

    2018

  4. [13]

    luXbeacon—A Batteryless Beacon for Green IoT: Design, Modeling, and Field Tests,

    K. E. Jeon, J. She, J. Xue, S.-H. Kim, and S. Park, “luXbeacon—A Batteryless Beacon for Green IoT: Design, Modeling, and Field Tests,” IEEE Internet of Things Journal, vol. 6, no. 3, pp. 5001–5012, Jun. 2019, conference Name: IEEE Internet of Things Journal

    2019

  5. [14]

    LED-to-LED visible light communication networks,

    S. Schmid, G. Corbellini, S. Mangold, and T. R. Gross, “LED-to-LED visible light communication networks,” inProceedings of the fourteenth ACM international symposium on Mobile ad hoc networking and computing. Bangalore India: ACM, Jul. 2013, pp. 1–10. [Online]. Available: http...

    2013

  6. [15]

    Retro-VLC: Enabling Battery-free Duplex Visible Light Communication for Mobile and IoT Applications,

    J. Li, A. Liu, G. Shen, L. Li, C. Sun, and F. Zhao, “Retro-VLC: Enabling Battery-free Duplex Visible Light Communication for Mobile and IoT Applications,” inProceedings of the 16th International Workshop on Mobile Computing Systems and Applications. Santa Fe New Mexico USA: AC...

    2015

  7. [16]

    PassiveVLC: Enabling Practical Visible Light Backscatter Communication for Battery-free IoT Applications,

    X. Xu, Y . Shen, J. Yang, C. Xu, G. Shen, G. Chen, and Y . Ni, “PassiveVLC: Enabling Practical Visible Light Backscatter Communication for Battery-free IoT Applications,” inProceedings of the 23rd Annual International Conference on Mobile Computing and Networking. Snowbird Uta...

    2017

  8. [17]

    Turboboosting Visible Light Backscatter Communication,

    Y . Wu, P. Wang, K. Xu, L. Feng, and C. Xu, “Turboboosting Visible Light Backscatter Communication,” inProceedings of the Annual conference of the ACM Special Interest Group on Data Communication on the applications, technologies, architectures, and protocols for computer comm...

    2020

  9. [18]

    Low- Latency Visible Light Backscatter Networking with RetroMUMIMO,

    K. Xu, C. Gong, B. Liang, Y . Wu, B. Di, L. Song, and C. Xu, “Low- Latency Visible Light Backscatter Networking with RetroMUMIMO,” inProceedings of the 20th ACM Conference on Embedded Networked Sensor Systems. Boston Massachusetts: ACM, Nov. 2022, pp. 448–461. [Online]. Availa...

    2022

  10. [19]

    Lightanchors: Appropriating point lights for spatially-anchored augmented reality interfaces,

    K. Ahuja, S. Pareddy, R. Xiao, M. Goel, and C. Harrison, “Lightanchors: Appropriating point lights for spatially-anchored augmented reality interfaces,” inProceedings of the 32nd Annual ACM Symposium on User Interface Software and Technology, ser. UIST ’19. New York, NY , USA:...

    2019

  11. [20]

    Every Byte Matters: Traffic Analysis of Bluetooth Wearable Devices,

    L. Barman, A. Dumur, A. Pyrgelis, and J.-P. Hubaux, “Every Byte Matters: Traffic Analysis of Bluetooth Wearable Devices,”Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies, vol. 5, no. 2, pp. 1–45, Jun. 2021. [Online]. Available: https://dl.acm...

    2021 doi

  12. [21]

    Powering for privacy: improving user trust in smart speaker microphones with intentional powering and perceptible assurance,

    Y . Do, N. Arora, A. Mirzazadeh, I. Moon, E. Xu, Z. Zhang, G. D. Abowd, and S. Das, “Powering for privacy: improving user trust in smart speaker microphones with intentional powering and perceptible assurance,” in32nd USENIX Security Symposium (USENIX Security 23). Anaheim, CA...

    2023

  13. [22]

    A neuromorphic multi-scale approach for real-time heart rate and state detection,

    C. De Luca, M. Tincani, G. Indiveri, and E. Donati, “A neuromorphic multi-scale approach for real-time heart rate and state detection,”npj Unconventional Computing, vol. 2, no. 1, p. 6, 2025

    2025

  14. [23]

    Adaptively pruned spiking neural networks for energy-efficient intracortical neural decoding,

    F. Rivelli, M. Popov, C. S. Kouzinopoulos, and G. Tang, “Adaptively pruned spiking neural networks for energy-efficient intracortical neural decoding,”arXiv preprint arXiv:2504.11568, 2025

    2025 arXiv

  15. [24]

    Lightweight event- based optical flow estimation via iterative deblurring,

    Y . Wu, F. Paredes-Vall ´es, and G. C. De Croon, “Lightweight event- based optical flow estimation via iterative deblurring,” in2024 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2024, pp. 14 708–14 715

    2024

  16. [25]

    Neuromor- phic split computing with wake-up radios: Architecture and design via digital twinning,

    J. Chen, S. Park, P. Popovski, H. V . Poor, and O. Simeone, “Neuromor- phic split computing with wake-up radios: Architecture and design via digital twinning,”IEEE Transactions on Signal Processing, 2024

    2024

  17. [26]

    Recent Event Camera Innovations: A Survey,

    B. Chakravarthi, A. A. Verma, K. Daniilidis, C. Fermuller, and Y . Yang, “Recent Event Camera Innovations: A Survey,” Aug. 2024, arXiv:2408.13627 [cs]. [Online]. Available: http://arxiv.org/abs/2408.13627

    2024 arXiv

  18. [27]

    NeuroCamTags: Long-Range, Battery-free, Wireless Sensing with Neuromorphic Cameras,

    D. Scott, M. Bringle, I. Fahad, G. Morales, A. Zahid, and S. Swaminathan, “NeuroCamTags: Long-Range, Battery-free, Wireless Sensing with Neuromorphic Cameras,”Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies, vol. 8, no. 3, Sep. 2024. [Online...

    2024 doi

  19. [28]

    The TENNLab exploratory neuromorphic computing framework,

    J. S. Plank, C. D. Schuman, G. Bruer, M. E. Dean, and G. S. Rose, “The TENNLab exploratory neuromorphic computing framework,”IEEE Letters of the Computer Society, vol. 1, no. 2, pp. 17–20, July-Dec 2018. [Online]. Available: https://doi.ieeecomputersociety.org/10.1109/LOCS.201...

    2018

  20. [29]

    Evolutionary Optimization for Neuromorphic Systems,

    C. D. Schuman, J. P. Mitchell, R. M. Patton, T. E. Potok, and J. S. Plank, “Evolutionary Optimization for Neuromorphic Systems,” inProceedings of the 2020 Annual Neuro-Inspired Computational Elements Workshop, ser. NICE ’20. New York, NY , USA: Association for Computing Machin...

    2020

  21. [30]

    Non-traditional input encoding schemes for spiking neuromorphic systems,

    C. D. Schuman, J. S. Plank, G. Bruer, and J. Anantharaj, “Non-traditional input encoding schemes for spiking neuromorphic systems,” inIJCNN: The International Joint Conference on Neural Networks, Budapest, 2019, pp. 1–10

    2019

  22. [31]

    The cart- pole application as a benchmark for neuromorphic computing,

    J. S. Plank, C. P. Rizzo, C. A. White, and C. D. Schuman, “The cart- pole application as a benchmark for neuromorphic computing,”Journal of Low Power Electronics and Applications, vol. 15, no. 1, pp. 1–27,

  23. [32]

    The RISP neuroprocessor – open source support for embedded neu- romorphic computing,

    J. S. Plank, K. E. M. Dent, B. Gullett, C. P. Rizzo, and C. D. Schuman, “The RISP neuroprocessor – open source support for embedded neu- romorphic computing,” inIEEE International Conference on Rebooting Computing (ICRC), San Diego, December 2024

    2024

  24. [2025]

    Available: https://www.mdpi.com/2079-9268/15/1/5

    [Online]. Available: https://www.mdpi.com/2079-9268/15/1/5

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.