Pith. sign in

REVIEW 4 major objections 5 minor 2 cited by

Event-based vision for egomotion estimation using precise event timing

T0 review · 4 major / 5 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read A shallow, unlearned spiking neural network estimates a vehicle's yaw rate directly from event-camera data, reporting lower error than published learning-based methods.

desk verdict A credible on-chip TDE demonstration undermined by an uncalibrated readout: the SOTA accuracy numbers are not interpretable without a scale factor. read the letter →

arxiv 2501.11554 v1 pith:UOCTBGLI submitted 2025-01-20 cs.CV cs.ARcs.RO

Hugh Greatorex , Michele Mastella , Madison Cotteret , Ole Richter , Elisabetta Chicca This is my paper · ORCID
classification cs.CVcs.ARcs.RO
keywords event-basedvisionegomotionestimationspikingneuralnetworkstimedifferenceencoderneuromorphichardwareopticalflowMVSECdatasetyawrate
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

This paper claims that a shallow, unlearned spiking neural network can estimate a vehicle's egomotion—specifically its yaw rate—directly from the asynchronous event stream of an event camera, without ever turning events into frames. The network is built from time difference encoders (TDEs), which convert the precise timing between events at two pixels into bursts of spikes whose rate encodes local optical flow. On the MVSEC driving dataset, the simulated full-field network reports average relative rotational errors of 0.00065, 0.00086, 0.00053, and 0.00067 rad on outdoor day1, night1, night2, and night3, and an on-chip version with a single TDE circuit reports 0.00014 rad on outdoor day1. The paper argues these numbers are at least an order of magnitude better than existing event-based egomotion methods, while the emulated on-chip network draws an estimated 1.8 nW. If correct, this shows that precise event timing alone, with no learning and no frame-based preprocessing, can carry accurate self-motion estimation.

What carries the argument

The central object is the Time Difference Encoder (TDE), a two-input spiking unit in which an event on the facilitatory input starts an exponentially decaying trace and an event on the trigger input samples that trace; the resulting current, and therefore the number and timing of output spikes, falls exponentially with the inter-event time difference, $I_{\mathrm{TDE}} \propto e^{-\Delta t / \tau_{\mathrm{FAC}}}$. When the two inputs are connected to pixels separated in space, the TDE responds preferentially to motion in that direction, functioning as an elementary motion detector. The full network is a layer of such TDEs split equally between two opposing orientations; a leaky integrator accumulates left-right spikes positively and right-left spikes negatively, and the difference signal, normalised to its sample maximum, is taken as the estimate of yaw rate. This fixed, parameter-free temporal-correlation computation carries the entire argument: no learning, no frame aggregation, and no trainable weights enter the pipeline.

What would settle it

Take the network's normalised output from one MVSEC sample, fit a single scale factor that maps it to the ground-truth yaw rate in radians per second, and apply that same fixed scale to the other three samples before recomputing the ARRE; if the accuracy collapses by orders of magnitude, the reported numbers reflect per-sample normalisation rather than a fixed egomotion estimate. A simpler laboratory test is to rotate the camera at a known constant yaw rate and check whether the network output amplitude stays constant as scene texture and event rate vary.

Watch

Extended reading notes

Core claim

The central discovery the paper proposes is that the time difference encoder—a CMOS circuit that emits a burst of spikes whose frequency falls exponentially with the time difference between a facilitatory and a trigger input event—acts as a direction-selective motion detector when its two inputs are connected to pixels separated in space. A population of such units, half preferring left-right motion and half right-left, yields an integrated differential activity that tracks the yaw rate of a moving vehicle. The paper demonstrates this in two regimes: a 200-TDE network emulated serially on a single fabricated circuit, and a 178,880-TDE network simulated over the full visual field. In both regimes the network output is normalised to its absolute maximum over the sample, and the resulting shape matches the ground-truth yaw rate; the reported ARRE values are 0.00014 rad on-chip and sub-milliradian in simulation, with the simulated network's accumulated heading drifting less than the IMU-derived heading in three of four samples. The claim is not that the network learns this—it does not; the claim is that a fixed, hand-wired temporal-correlation computation is sufficient for high-accuracy egomotion readout.

Load-bearing premise

The headline accuracy figures assume that the normalised, unitless TDE-activity signal can be compared directly with ground-truth yaw rate in radians per second, but the paper never states or derives the scale factor that converts spike activity into angular velocity.

Editorial extensions

If this is right

  • A hardware implementation of the full 178,880-TDE network is projected to run at roughly 200 µW, making the approach feasible for power-constrained platforms such as micro-drones, edge devices, and VR headsets.
  • Because the network requires no training, it can be deployed on new scenes or domains without re-tuning, in contrast to learned event-frame CNNs.
  • The event-based readout can augment or recalibrate IMU-based navigation, since the simulated network's accumulated heading drifts less than the IMU in three of the four MVSEC samples.
  • The reported accuracy gain over prior event-based egomotion methods—ARRE of 0.00053–0.00086 rad versus 0.00139–0.121 rad in the cited comparisons—suggests that preserving precise event timing, rather than aggregating events into frames, is the decisive design choice.

Reading between the lines

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

  • The paper never states a scale factor that converts the normalised, unitless TDE activity into radians per second; applying one fixed scale learned from a single sample to the other samples would test whether the reported accuracy is a genuine prediction or a per-sample shape fit.
  • By orienting TDE pairs vertically rather than horizontally, the same mechanism could be extended from yaw to pitch and roll, and with four orientations it could in principle estimate full 3-DOF angular velocity.
  • Because TDEs respond to event timing rather than event count, the network should in principle be robust to independently moving objects, which the discussion gestures at but does not test; a scene containing a walking pedestrian would provide a direct test.
  • The 200 µW scaling projection assumes the readout aggregation and spike-routing circuitry are also implemented on-chip; the paper measured only the TDE core, so actual system power depends on the aggregation hardware.
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 / 5 minor

Summary. The paper proposes a fully event-based pipeline for egomotion estimation built on a Time Difference Encoder (TDE) circuit. It reports silicon measurements of the TDE on the cognigr1 chip, an on-chip emulation of a 200-TDE network on the MVSEC outdoor_day1 sequence, and larger JAX simulations with 178,880 TDEs on four MVSEC sequences. The headline results are ARRE values of 0.00014 rad on-chip and 0.00053--0.00086 rad in simulation, which the authors claim are an order of magnitude better than published event-based egomotion methods.

Significance. If the accuracy claim were supportable, the paper would be notable: a shallow, unlearned, event-timing network outperforming learned frame-conversion baselines on MVSEC, together with a low-power mixed-signal implementation. The silicon measurements and the circuit--model match in Fig. 2 are credible and useful. However, the central quantitative comparison is undermined by the absence of a physical calibration from the unitless, per-sample-normalised network activity to yaw rate in rad/s. As written, the headline ARRE values and the comparison in Table I are not interpretable, so the claimed significance is not established.

major comments (4)
  1. [Eq. 6 / Results, On-chip egomotion network emulation] The network output is never converted to radians per second. Eq. 6 defines A(t) as a leaky integral of a spike-count difference in arbitrary units, and the Results state that the signal is normalised to its maximum absolute value over the duration of the sample. Figures 6 and 7 nevertheless plot the output on axes labelled in rad/sec and report ARRE values. Supplementary S2 defines ARRE on predicted Euler-angle vectors using a matrix logarithm; it is not defined for a unitless, per-sample-normalised scalar waveform. If the normalisation constant, or any additional scale, is taken from the ground-truth maximum, the reported ARRE values become partly a fit rather than a prediction. A causal calibration from known pixel geometry and camera intrinsics, or an explicit scale-calibration experiment, is needed before ARRE can be reported in radians.
  2. [Results, Scaled-up simulations / Table I] The comparison in Table I is not apples to apples. The cited baselines estimate egomotion as full pose or rotation, whereas this work estimates a single yaw-rate correlate from a population of TDEs. Because Supplementary S2 defines ARRE on rotation matrices, the values in Table I cannot be compared unless the same metric is computed identically for all methods on the same output representation. The claim of an improvement of at least one order of magnitude over previous works is therefore not supported by the evidence presented.
  3. [Methods, Egomotion network; Results, On-chip egomotion network emulation] The per-sample normalisation is non-causal. The normalisation constant is determined from the absolute maximum activity over the whole test sample, so the outputs plotted in Figs. 6 and 7 could not be produced online in a streaming or low-latency system. Either the evaluation should be repeated with a causal normalisation or a fixed calibration, or the real-time and low-latency claims should be qualified accordingly.
  4. [Results, On-chip network power consumption; Methods IV.1] The on-chip experiment is a sequential emulation of 200 TDEs on a single circuit, not a concurrent 200-TDE array. The 1.8 nW figure is an extrapolation that explicitly excludes the off-chip aggregation readout. This is acceptable as a scaling estimate, but the abstract and discussion should be worded so that readers do not infer that a parallel 200-TDE network was measured on the chip.
minor comments (5)
  1. [References] Reference [13] is cited for the MVSEC dataset, but the bibliography entry is the EV-FlowNet paper by Zhu et al.; the MVSEC dataset citation should be corrected.
  2. [Figs. 6 and 7] The accumulated-heading panels are labelled in rad but are obtained by integrating a normalised unitless signal; the axis units and the scaling used to produce the headings should be clarified.
  3. [Methods, Egomotion network] The random placement of TDE units in the two boxes is said to be identical, but no seed or code is provided; specifying the seed would improve reproducibility.
  4. [Supplementary S2] The ARRE definition should state explicitly whether it is applied to instantaneous yaw rate or to the integrated heading trajectory, and with what scaling; the current formula refers only to Euler-angle vectors and rotation matrices.
  5. [Eqs. (1)--(2) and Eq. (6)] The symbol I_TDE is used for the circuit current in Eqs. (1)--(2) and the integrated activity in Eq. (6); distinct notation for these quantities would avoid confusion.

Circularity Check

1 steps flagged · score 6.0 of 10

The headline ARRE values are not an independent physical prediction: Eq. 6 defines only a unitless, per-sample-maximum-normalized spike activity, and no conversion to radians is supplied before computing ARRE against MVSEC ground truth.

  1. fitted input called prediction [Results 'On-chip egomotion network emulation', Methods 'Egomotion network' Eq. (6), Fig. 6, Supplementary S2]
    "The estimated egomotion signal was generated by integrating the TDE spiking activity and computing the difference between the right-left and left-right oriented TDEs. ... The resulting signal was then normalised to the maximum absolute value over the duration of the sample. ... The egomotion signal representing the yaw rate of the vehicle (ψ˙) was measured with an Average Relative Rotational Error (ARRE) (Supplementary S2) of 0.000 14 rad on-chip."

    Eq. (6) defines A(t) as a leaky integral of spike counts; after the stated sample-maximum normalization, A(t) is dimensionless. Supplementary S2 defines ARRE on predicted Euler angles P in radians, so converting A(t) to ψ˙(t) in rad/s requires a scale factor. The paper supplies no such factor. The normalization constant is computed over the very sample on which ARRE is reported, so if the missing scale is taken from the ground-truth maximum, or chosen by least-squares matching, the reported 0.00014–0.00086 rad values quantify a per-sample fit rather than an independent prediction. Even without a ground-truth-derived scale, ARRE is undefined for a unitless waveform, so the Table I state-of-the-art comparison is not a physical prediction.

full rationale

The core signal-processing chain is otherwise not circular: the TDE parameters are fixed and not fitted to the egomotion target, the circuit is measured against the simulation model, and the network is unlearned. The circularity is localized to the evaluation scale that carries the headline quantitative claim. Eq. (6) and the normalization sentence produce a unitless activity, yet the paper reports ARRE in radians and plots the signal on a rad/sec axis. Because ARRE is defined for predicted Euler angles, a rad/s calibration must be inserted at some point; the only calibration-related quantity stated in the paper is the per-sample maximum normalization, which is derived from the same data used for evaluation. If that normalization is matched to the ground-truth amplitude, the state-of-the-art accuracy is partly a fit by construction, not a genuinely predicted physical yaw rate.

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

The central claim rests on a small set of hand-set model parameters and, critically, on an unspecified per-sample scale that converts spike activity into physical yaw rate. The TDE circuit itself is taken from prior work (refs 28 and 31), so no new physical entity is introduced.

free parameters (4)
  • per-sample output normalisation scale = not stated; maximum absolute activity over the sample
    The egomotion signal is divided by the maximum absolute value over the duration of the sample (Results and Methods Eq. 6). This scale is needed to convert unitless spike activity into rad/s for ARRE; without it the reported sub-milliradian errors are unexplained.
  • leaky integrator time constant tau_A = 0.75 s (Brian2/on-chip), 10 ms (JAX)
    Eq. 6 uses tau_A to smooth TDE activity; the two simulation setups use 75 times different values with no sensitivity analysis, so the smoothing level is a hand-chosen free parameter that affects the output signal.
  • TDE dynamics parameters = tau_FAC = tau_TRG = tau_m = 20 ms, u_theta = 50, w_FAC = w_TRG = 1
    Chosen to give high spiking dynamic range over a 50 ms input interval (Results). They are not fitted to the egomotion target, but the yaw estimate depends on them.
  • network geometry choices = stride 2, 20x20 boxes, 100 TDEs per box, 178,880 TDEs in simulation
    Box locations, stride, and TDE density are selected by hand; the on-chip result uses a different geometry from the scaled-up simulation, and no sensitivity analysis is provided.
assumptions (4)
  • domain assumption A static scene imaged by a moving event camera produces optical flow whose horizontal component is dominated by yaw rotation, so the difference of left-right versus right-left TDE activity estimates yaw rate.
    Invoked in Results and Methods when TDE orientations are chosen to be left-right or right-left and their activity difference is summed. Independently moving objects or depth-dependent translation flow would corrupt the estimate.
  • domain assumption The TDE circuit transfer function ITDE proportional to exp(-Delta t / tau_FAC) holds on the fabricated chip under the chosen subthreshold biases.
    Stated in Eqs. 1-2 and Fig. 2; the circuit measurements support it in one configuration, but the egomotion results assume it holds for all TDE instances and event rates.
  • ad hoc to paper The ARRE metric in Supplementary S2 is a valid error measure when applied to unitless, per-sample-normalised network outputs.
    The predicted signal is normalised by its sample maximum, yet ARRE is defined on Euler angles in physical units; the paper never states how the normalised signal is converted to radians, so this assumption is introduced without support.
  • domain assumption Brian2 and JAX simulations faithfully reproduce the analog circuit response.
    The paper shows one comparison in Fig. 2 and claims high similarity, but the scaled-up simulations use the abstract TDE model, not the measured circuit, and use different tau_A values.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Event-based vision for egomotion estimation using precise event timing." pith.science (2026). https://pith.science/paper/UOCTBGLI

@misc{pith2026250111554,
  author       = {Pith},
  title        = {Pith review of: Event-based vision for egomotion estimation using precise event timing},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/UOCTBGLI}},
  note         = {Machine review of arXiv:2501.11554}
}
read the original abstract

Egomotion estimation is crucial for applications such as autonomous navigation and robotics, where accurate and real-time motion tracking is required. However, traditional methods relying on inertial sensors are highly sensitive to external conditions, and suffer from drifts leading to large inaccuracies over long distances. Vision-based methods, particularly those utilising event-based vision sensors, provide an efficient alternative by capturing data only when changes are perceived in the scene. This approach minimises power consumption while delivering high-speed, low-latency feedback. In this work, we propose a fully event-based pipeline for egomotion estimation that processes the event stream directly within the event-based domain. This method eliminates the need for frame-based intermediaries, allowing for low-latency and energy-efficient motion estimation. We construct a shallow spiking neural network using a synaptic gating mechanism to convert precise event timing into bursts of spikes. These spikes encode local optical flow velocities, and the network provides an event-based readout of egomotion. We evaluate the network's performance on a dedicated chip, demonstrating strong potential for low-latency, low-power motion estimation. Additionally, simulations of larger networks show that the system achieves state-of-the-art accuracy in egomotion estimation tasks with event-based cameras, making it a promising solution for real-time, power-constrained robotics applications.

Figures

Figures reproduced from arXiv: 2501.11554 by the authors.

Figure 1
Figure 1. Photograph of the cognigr1 chip, fabricated in 180 nm technology along with the schematic of the TDE synapse. a) The relevant structures on the die are indicated. The total size of the TDE circuit is 19 µm×56 µm including guard rings and is biased by an on-chip DAC. b) Schematic of the state-of-the-art CMOS TDE synapse [28], with facilitatory and trigger blocks labeled. widely used, are inadequate for capturing the … view at source ↗
Figure 2
Figure 2. Silicon measurements of the TDE circuit. a) The time (∆t) between FAC and TRG input events was increased systematically. Each plot shows a ∆t increment of 20 ms from 0 ms to 80 ms. The membrane potential, Vmem of the neuron integrates the input current from the TDE synapse and outputs a spike when the threshold is reached. The encoding of the temporal distance between the input events is exhibited by the dynamics of… view at source ↗
Figure 3
Figure 3. The TDE can be applied to event-based vision tasks by “connecting” the FAC and TRG inputs to specific (x, y) pixels. In this way the TDE becomes receptive to a particular direction of motion. In this case the TDE is sensitive to left-right motion, indicated by the arrow representing the TDE connectivity. We define the stride as the pixel-wise separation of FAC and TRG inputs, in this example it is 1. On-chip egomoti… view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: An illustration of how the FAC and TRG connections are oriented for each TDE unit in relation to the (x, y) event data from the event camera. b) A single frame (generated by integrating activity over an arbitrary time step) of the MVSEC sample outdoor day1. This event …
Figure 5
Figure 5. Figure 5: Measurements from the TDE circuit on the cognigr1 chip, implementing the egomotion estimation task. For this task the first 90 seconds of events from the outdoor day1 MVSEC sample were used. a) The event data from the event-camera mounted on the driving car. The two bo…
Figure 6
Figure 6. Figure 6: The egomotion estimation of the vehicle using the congigr1 chip. a) The egomotion network activity compared with the IMU measurements from the vehicle and the recorded ground truth. The estimated motion is the angular velocity of the car, ψ˙ , as it turns, otherwise kn…
Figure 7
Figure 7. Figure 7: The simulated scaled-up egomotion network sampling from the entire visual field. The network estimation is compared to the IMU recording and ground truth. The ARRE compared to the ground truth is reported for each driving sequence and the accumulated heading direction,…

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 2 Pith papers

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

  1. All Eyes, no IMU: Learning Flight Attitude from Vision Alone

    cs.RO 2025-07 conditional novelty 7.0 of 10

    A quadrotor was flown with closed-loop attitude and rate control driven purely by event-camera vision through a recurrent CNN, without using an IMU in the inner loop.

  2. A scalable event-driven spatiotemporal feature extraction circuit

    eess.SP 2025-01 conditional novelty 6.0 of 10

    A redesigned subthreshold CMOS time-difference encoder circuit achieves 61% lower coefficient of variation in transmitted charge across Monte Carlo mismatch simulations, supporting scalable event-based vision arrays.

Reference graph

Works this paper leans on

62 extracted references · 36 canonical work pages · cited by 2 Pith papers

  1. [1]

    Mobile robot positioning: Sensors and techniques

    J. Borenstein, H. R. Everett, L. Feng, and D. Wehe, “Mobile robot positioning: Sensors and techniques”, Journal of Robotic Systems, vol. 14, no. 4, pp. 231–249, 1997. DOI: 10.1002/ (SICI)1097-4563(199704)14:4⟨231::AID-ROB2⟩3.0.CO;2-R

    work page 1997

  2. [2]

    Visual Odometry [Tuto- rial]

    D. Scaramuzza and F. Fraundorfer, “Visual Odometry [Tuto- rial]”, IEEE Robotics & Automation Magazine , vol. 18, no. 4, pp. 80–92, 2011. DOI: 10.1109/MRA.2011.943233

    arXiv 2011

  3. [3]

    Continuous-Time Visual-Inertial Odometry for Event Cam- eras

    E. Mueggler, G. Gallego, H. Rebecq, and D. Scaramuzza, “Continuous-Time Visual-Inertial Odometry for Event Cam- eras”, IEEE Transactions on Robotics, vol. 34, no. 6, pp. 1425– 1440, 2018. DOI: 10.1109/TRO.2018.2858287

    arXiv 2018

  4. [4]

    Visual odometry,

    D. Nister, O. Naroditsky, and J. Bergen, “Visual odometry,” in Proceedings of the 2004 IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2004. CVPR 2004., vol. 1, 2004, pp. I–I. DOI: 10 . 1109 / CVPR . 2004 . 1315094

    work page 2004

  5. [5]

    Visual odometry for ground vehicle applications

    D. Nist ´er, O. Naroditsky, and J. Bergen, “Visual odometry for ground vehicle applications”, Journal of Field Robotics , vol. 23, no. 1, pp. 3–20, 2006. DOI: 10.1002/rob.20103

    work page doi:10.1002/rob.20103 2006

  6. [6]

    Low-latency automotive vision with event cameras

    D. Gehrig and D. Scaramuzza, “Low-latency automotive vision with event cameras”, Nature, vol. 629, no. 8014, pp. 1034– 1040, 2024. DOI: 10.1038/s41586-024-07409-w

    doi:10.1038/s41586-024-07409-w 2024

  7. [7]

    The Silicon Retina

    M. Mahowald, “The Silicon Retina”, An Analog VLSI System for Stereoscopic Vision, pp. 4–65, 1994. DOI: 10.1007/978-1- 4615-2724-4 2

    doi:10.1007/978-1- 1994

  8. [8]

    A silicon model of early visual processing

    C. A. Mead and M. A. Mahowald, “A silicon model of early visual processing”, Neural Networks, vol. 1, no. 1, pp. 91–97,

    work page

Show all 62 references
  1. [9]

    A 128 × 128 120 dB 15 µs latency asynchronous temporal contrast vision sensor

    P. Lichtsteiner, C. Posch, and T. Delbruck, “A 128 × 128 120 dB 15 µs latency asynchronous temporal contrast vision sensor”, IEEE Journal of Solid-State Circuits , vol. 43, no. 2, pp. 566–576, 2008. DOI: 10.1109/JSSC.2007.914337

    2008

  2. [10]

    A QVGA 143 dB dynamic range frame-free PWM image sensor with lossless pixel-level video compression and time-domain CDS

    C. Posch, D. Matolin, and R. Wohlgenannt, “A QVGA 143 dB dynamic range frame-free PWM image sensor with lossless pixel-level video compression and time-domain CDS”, IEEE Journal of Solid-State Circuits , vol. 46, no. 1, pp. 259–275,

  3. [11]

    Event-Based Neuromorphic Vision for Autonomous Driving: A Paradigm Shift for Bio-Inspired Visual Sensing and Perception

    G. Chen, H. Cao, J. Conradt, H. Tang, F. Rohrbein, and A. Knoll, “Event-Based Neuromorphic Vision for Autonomous Driving: A Paradigm Shift for Bio-Inspired Visual Sensing and Perception”, IEEE Signal Processing Magazine , vol. 37, no. 4, pp. 34–49, 2020. DOI: 10.1109/MSP.2020.2985815

    2020

  4. [12]

    Champion-level drone racing us- ing deep reinforcement learning

    E. Kaufmann, L. Bauersfeld, A. Loquercio, M. M ¨uller, V . Koltun, and D. Scaramuzza, “Champion-level drone racing us- ing deep reinforcement learning”, Nature, vol. 620, no. 7976, pp. 982–987, 2023. DOI: 10.1038/s41586-023-06419-4

    2023 doi

  5. [13]

    EV- FlowNet: Self-Supervised Optical Flow Estimation for Event- based Cameras

    A. Z. Zhu, L. Yuan, K. Chaney, and K. Daniilidis, “EV- FlowNet: Self-Supervised Optical Flow Estimation for Event- based Cameras”, Robotics: Science and Systems , 2018. DOI: 10.15607/RSS.2018.XIV .062

    2018 doi

  6. [14]

    Robust Event-Based Vision Model Estimation by Dispersion Minimisation

    U. M. Nunes and Y . Demiris, “Robust Event-Based Vision Model Estimation by Dispersion Minimisation”, IEEE Trans- actions on Pattern Analysis and Machine Intelligence , vol. 44, no. 12, pp. 9561–9573, 2022. DOI: 10 . 1109 / TPAMI . 2021 . 3130049

    2022

  7. [15]

    Spike-FlowNet: Event-Based Optical Flow Estima- tion with Energy-Efficient Hybrid Neural Networks

    C. Lee, A. K. Kosta, A. Z. Zhu, K. Chaney, K. Daniilidis, and K. Roy, “Spike-FlowNet: Event-Based Optical Flow Estima- tion with Energy-Efficient Hybrid Neural Networks”, Lecture Notes in Computer Science , vol. 12374 LNCS, pp. 366–382,

  8. [16]

    Optical flow estimation from event-based cam- eras and spiking neural networks

    J. Cuadrado, U. Ranc ¸on, B. R. Cottereau, F. Barranco, and T. Masquelier, “Optical flow estimation from event-based cam- eras and spiking neural networks”, Frontiers in Neuroscience, vol. 17, p. 1 160 034, 2023. DOI: 10 . 3389 / FNINS . 2023 . 1160034/BIBTEX

    2023

  9. [17]

    Unsu- pervised Learning of Depth and Ego-Motion From Video,

    T. Zhou, M. Brown, N. Snavely, and D. G. Lowe, “Unsu- pervised Learning of Depth and Ego-Motion From Video,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2017, pp. 1851–1858

    2017

  10. [18]

    Low-latency event-based visual odometry

    A. Censi and D. Scaramuzza, “Low-latency event-based visual odometry”, Proceedings - IEEE International Conference on Robotics and Automation , pp. 703–710, 2014. DOI: 10.1109/ ICRA.2014.6906931

    2014

  11. [19]

    Event-driven Vision and Control for UA Vs on a Neuromorphic Chip

    A. Vitale, A. Renner, C. Nauer, D. Scaramuzza, and Y . Sandamirskaya, “Event-driven Vision and Control for UA Vs on a Neuromorphic Chip”, Proceedings - IEEE International Conference on Robotics and Automation , vol. 2021-May, pp. 103–109, 2021. DOI: 10.1109/ICRA48506.2021.9560881

    2021

  12. [20]

    Event-Based Moving Object Detection and Track- ing

    A. Mitrokhin, C. Fermuller, C. Parameshwara, and Y . Aloi- monos, “Event-Based Moving Object Detection and Track- ing”, IEEE International Conference on Intelligent Robots and Systems, pp. 6895–6902, 2018. DOI: 10 . 1109 / IROS . 2018 . 8593805

    2018

  13. [21]

    EV-IMO: Motion Segmentation Dataset and Learning Pipeline for Event Cameras,

    A. Mitrokhin, C. Ye, C. Ferm ¨uller, Y . Aloimonos, and T. Del- bruck, “EV-IMO: Motion Segmentation Dataset and Learning Pipeline for Event Cameras,” in 2019 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , 2019, pp. 6105–6112. DOI: 10.1109/IROS4089...

    2019

  14. [22]

    Event-based, 6-DOF Camera Tracking from Photometric Depth Maps

    G. Gallego, J. E. A. Lund, E. Mueggler, H. Rebecq, T. Delbruck, and D. Scaramuzza, “Event-based, 6-DOF Camera Tracking from Photometric Depth Maps”, IEEE Transactions on Pattern Analysis and Machine Intelligence , vol. 40, no. 10, pp. 2402–2412, 2018. DOI: 10.1109/TPAMI.2017.2769655

    2018

  15. [23]

    Accurate Angular Velocity Estimation With an Event Camera

    G. Gallego and D. Scaramuzza, “Accurate Angular Velocity Estimation With an Event Camera”, IEEE Robotics and Au- tomation Letters , vol. 2, no. 2, pp. 632–639, 2017. DOI: 10. 1109/LRA.2016.2647639

    2017

  16. [24]

    Visual odometry with neuromorphic resonator networks

    A. Renner, L. Supic, A. Danielescu, G. Indiveri, E. P. Frady, F. T. Sommer, and Y . Sandamirskaya, “Visual odometry with neuromorphic resonator networks”, Nature Machine Intelli- gence, vol. 6, no. 6, pp. 653–663, 2024. DOI: 10.1038/s42256- 024-00846-2

    2024 doi

  17. [25]

    Event-Based Visual Flow

    R. Benosman, C. Clercq, X. Lagorce, S.-H. Ieng, and C. Bartolozzi, “Event-Based Visual Flow”, IEEE Transactions on Neural Networks and Learning Systems , vol. 25, no. 2, pp. 407–417, 2014. DOI: 10.1109/TNNLS.2013.2273537

    2014

  18. [26]

    Spiking Optical Flow for Event-Based Sensors Using IBM’s TrueNorth Neurosynaptic System

    G. Haessig, A. Cassidy, R. Alvarez, R. Benosman, and G. Or- chard, “Spiking Optical Flow for Event-Based Sensors Using IBM’s TrueNorth Neurosynaptic System”, IEEE Transactions on Biomedical Circuits and Systems , vol. 12, no. 4, pp. 860– 870, 2018. DOI: 10.1109/TBCAS.2018.2834558

    2018

  19. [27]

    A Low Power, Fully Event-Based Gesture Recognition System,

    A. Amir, B. Taba, D. Berg, T. Melano, J. McKinstry, C. Di Nolfo, T. Nayak, A. Andreopoulos, et al., “A Low Power, Fully Event-Based Gesture Recognition System,” in Proceed- ings of the IEEE Conference on Computer Vision and Pattern Recognition, 2017, pp. 7243–7252. 9

    2017

  20. [28]

    A scalable event-driven spatiotemporal feature extraction circuit

    H. Greatorex, M. Mastella, O. Richter, M. Cotteret, W. S. Gir˜ao, E. Janotte, and E. Chicca, “A scalable event-driven spatiotemporal feature extraction circuit”, 2025. DOI: 10 . 48550/arXiv.2501.10155

    2025 doi arXiv

  21. [29]

    EvGNN: An Event- driven Graph Neural Network Accelerator for Edge Vision

    Y . Yang, A. Kneip, and C. Frenkel, “EvGNN: An Event- driven Graph Neural Network Accelerator for Edge Vision”, IEEE Transactions on Circuits and Systems for Artificial Intelligence, pp. 1–14, 2024. DOI: 10.1109/TCASAI.2024. 3520905

    2024 doi

  22. [30]

    Spike-based dynamic computing with asynchronous sensing-computing neuromorphic chip

    M. Yao, O. Richter, G. Zhao, N. Qiao, Y . Xing, D. Wang, T. Hu, W. Fang, et al., “Spike-based dynamic computing with asynchronous sensing-computing neuromorphic chip”, Nature Communications, vol. 15, no. 1, p. 4464, 2024. DOI: 10.1038/ s41467-024-47811-6

    2024

  23. [31]

    Spiking Elementary Motion Detector in Neuromorphic Systems

    M. B. Milde, O. J. N. Bertrand, H. Ramachandran, M. Egel- haaf, and E. Chicca, “Spiking Elementary Motion Detector in Neuromorphic Systems”, Neural Computation, vol. 30, no. 9, pp. 2384–2417, 2018. DOI: 10.1162/neco a 01112

    2018 doi

  24. [32]

    Finding the gap: Neuromorphic motion-vision in dense environments

    T. Schoepe, E. Janotte, M. B. Milde, O. J. N. Bertrand, M. Egelhaaf, and E. Chicca, “Finding the gap: Neuromorphic motion-vision in dense environments”, Nature Communica- tions, vol. 15, no. 1, p. 817, 2024. DOI: 10 . 1038 / s41467 - 024-45063-y

    2024

  25. [33]

    Seeing Things in Motion: Models, Circuits, and Mechanisms

    A. Borst and T. Euler, “Seeing Things in Motion: Models, Circuits, and Mechanisms”, Neuron, vol. 71, no. 6, pp. 974– 994, 2011. DOI: 10.1016/J.NEURON.2011.08.031

    2011 doi

  26. [34]

    Autokorrelations-Auswertung als Funktion- sprinzip des Zentralnervensystems: (bei der optischen Bewe- gungswahrnehmung eines Insektes)

    W. Reichardt, “Autokorrelations-Auswertung als Funktion- sprinzip des Zentralnervensystems: (bei der optischen Bewe- gungswahrnehmung eines Insektes)”, Zeitschrift f ¨ur Natur- forschung B, vol. 12, no. 7, pp. 448–457, 1957. DOI: 10.1515/ znb-1957-0707

    1957

  27. [35]

    Event-Based Eccentric Motion Detection Exploiting Time Difference Encoding

    G. D’Angelo, E. Janotte, T. Schoepe, J. O’Keeffe, M. B. Milde, E. Chicca, and C. Bartolozzi, “Event-Based Eccentric Motion Detection Exploiting Time Difference Encoding”, Frontiers in Neuroscience, vol. 14, 2020. DOI: 10.3389/fnins.2020.00451

    2020

  28. [36]

    Closed-loop sound source localization in neuromor- phic systems

    T. Schoepe, D. Gutierrez-Galan, J. P. Dominguez-Morales, H. Greatorex, A. Jimenez-Fernandez, A. Linares-Barranco, and E. Chicca, “Closed-loop sound source localization in neuromor- phic systems”, Neuromorphic Computing and Engineering , vol. 3, no. 2, 2023. DOI: 10.1088/2634-4...

    2023 doi

  29. [37]

    Odour Localization in Neu- romorphic Systems,

    T. Schoepe, D. Drix, F. M. Sch ¨uffny, R. Miko, S. Sutton, E. Chicca, and M. Schmuker, “Odour Localization in Neu- romorphic Systems,” in 2024 IEEE International Symposium on Circuits and Systems (ISCAS), 2024, pp. 1–5. DOI: 10.1109/ ISCAS58744.2024.10558186

    2024

  30. [38]

    Low- Latency Monocular Depth Estimation Using Event Timing on Neuromorphic Hardware,

    S. Chiavazza, S. M. Meyer, and Y . Sandamirskaya, “Low- Latency Monocular Depth Estimation Using Event Timing on Neuromorphic Hardware,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 2023, pp. 4071–4080

    2023

  31. [39]

    Fully neuromorphic vision and control for autonomous drone flight

    F. Paredes-Vall ´es, J. J. Hagenaars, J. Dupeyroux, S. Stroobants, Y . Xu, and G. C. H. E. de Croon, “Fully neuromorphic vision and control for autonomous drone flight”, Science Robotics , vol. 9, no. 90, eadi0591, 2024. DOI: 10 . 1126 / scirobotics . adi0591

    2024

  32. [40]

    A current-mode conductance-based silicon neuron for address-event neuromorphic systems,

    P. Livi and G. Indiveri, “A current-mode conductance-based silicon neuron for address-event neuromorphic systems,” in 2009 IEEE International Symposium on Circuits and Systems , 2009, pp. 2898–2901. DOI: 10.1109/ISCAS.2009.5118408

    2009

  33. [41]

    Neu- romorphic Electronic Circuits for Building Autonomous Cog- nitive Systems

    E. Chicca, F. Stefanini, C. Bartolozzi, and G. Indiveri, “Neu- romorphic Electronic Circuits for Building Autonomous Cog- nitive Systems”, Proceedings of the IEEE , vol. 102, no. 9, pp. 1367–1388, 2014. DOI: 10.1109/JPROC.2014.2313954

    2014

  34. [42]

    Efficient Neuromor- phic Signal Processing with Loihi 2,

    G. Orchard, E. P. Frady, D. B. D. Rubin, S. Sanborn, S. B. Shrestha, F. T. Sommer, and M. Davies, “Efficient Neuromor- phic Signal Processing with Loihi 2,” in 2021 IEEE Workshop on Signal Processing Systems (SiPS), Coimbra, Portugal, 2021, pp. 254–259. DOI: 10.1109/SiPS52927....

    2021

  35. [43]

    Learning sensorimotor control with neuromorphic sen- sors: Toward hyperdimensional active perception

    A. Mitrokhin, P. Sutor, C. Ferm ¨uller, and Y . Aloimonos, “Learning sensorimotor control with neuromorphic sen- sors: Toward hyperdimensional active perception”, Science Robotics, vol. 4, no. 30, 2019. DOI: 10 . 1126 / scirobotics . aaw6736

    2019

  36. [44]

    Unsupervised learning of dense optical flow, depth and egomotion with event-based sensors

    C. Ye, A. Mitrokhin, C. Fermuller, J. A. Yorke, and Y . Aloi- monos, “Unsupervised learning of dense optical flow, depth and egomotion with event-based sensors”, IEEE International Conference on Intelligent Robots and Systems, pp. 5831–5838,

  37. [45]

    Unsu- pervised Event-Based Learning of Optical Flow, Depth, and Egomotion,

    A. Z. Zhu, L. Yuan, K. Chaney, and K. Daniilidis, “Unsu- pervised Event-Based Learning of Optical Flow, Depth, and Egomotion,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2019, pp. 989–997

    2019

  38. [46]

    E- RAFT: Dense Optical Flow from Event Cameras

    M. Gehrig, M. Millhausler, D. Gehrig, and D. Scaramuzza, “E- RAFT: Dense Optical Flow from Event Cameras”, Proceed- ings - 2021 International Conference on 3D Vision, 3DV 2021, pp. 197–206, 2021. DOI: 10.1109/3DV53792.2021.00030

    2021

  39. [47]

    Taming Contrast Maximization for Learning Sequential, Low-latency, Event-based Optical Flow,

    F. Paredes-Vall ´es, K. Y . W. Scheper, C. De Wagter, and G. C. H. E. de Croon, “Taming Contrast Maximization for Learning Sequential, Low-latency, Event-based Optical Flow,” in Proceedings of the IEEE/CVF International Conference on Computer Vision, 2023, pp. 9695–9705

    2023

  40. [48]

    DOI: 10.1109/IROS45743.2020.9341224

    2020

  41. [49]

    A 64-mW DNN-Based Visual Nav- igation Engine for Autonomous Nano-Drones

    D. Palossi, A. Loquercio, F. Conti, E. Flamand, D. Scara- muzza, and L. Benini, “A 64-mW DNN-Based Visual Nav- igation Engine for Autonomous Nano-Drones”, IEEE Internet of Things Journal , vol. 6, no. 5, pp. 8357–8371, 2019. DOI: 10.1109/JIOT.2019.2917066

    2019

  42. [50]

    TDE-3: An improved prior for optical flow com- putation in spiking neural networks

    M. Yedutenko, F. Paredes-Valles, L. Khacef, and G. C. H. E. De Croon, “TDE-3: An improved prior for optical flow com- putation in spiking neural networks”, 2024. DOI: 10.48550/ arXiv.2402.11662

    2024 doi

  43. [51]

    A Mixed-Signal Near-Sensor Con- volutional Imager SoC with Charge-Based 4b-Weighted 5- to-84-TOPS/W MAC Operations for Feature Extraction and Region-of-Interest Detection,

    M. Lefebvre and D. Bol, “A Mixed-Signal Near-Sensor Con- volutional Imager SoC with Charge-Based 4b-Weighted 5- to-84-TOPS/W MAC Operations for Feature Extraction and Region-of-Interest Detection,” in 2024 IEEE Custom Inte- grated Circuits Conference (CICC) , 2024, pp. 1–2. DO...

    2024

  44. [52]

    Secrets of Event- Based Optical Flow, Depth and Ego-Motion Estimation by Contrast Maximization

    S. Shiba, Y . Klose, Y . Aoki, and G. Gallego, “Secrets of Event- Based Optical Flow, Depth and Ego-Motion Estimation by Contrast Maximization”, IEEE Transactions on Pattern Anal- ysis and Machine Intelligence, vol. 46, no. 12, pp. 7742–7759,

  45. [53]

    Brian 2, an intuitive and efficient neural simulator

    M. Stimberg, R. Brette, and D. F. Goodman, “Brian 2, an intuitive and efficient neural simulator”, eLife, vol. 8, e47314,

  46. [57]

    Synaptic Dynamics in Analog VLSI

    C. Bartolozzi and G. Indiveri, “Synaptic Dynamics in Analog VLSI”, Neural Computation, vol. 19, no. 10, pp. 2581–2603,

  47. [61]

    Synaptic Dynamics in Analog VLSI

    C. Bartolozzi and G. Indiveri, “Synaptic Dynamics in Analog VLSI”, Neural Computation , vol. 19, no. 10, pp. 2581–2603, 2007. DOI: 10.1162/neco.2007.19.10.2581. 3

    2007 doi

  48. [63]

    Robust Spiking Attractor Networks with a Hard Winner-Take-All Neuron Circuit,

    M. Cotteret, O. Richter, M. Mastella, H. Greatorex, E. Janotte, W. S. Gir ˜ao, M. Ziegler, and E. Chicca, “Robust Spiking Attractor Networks with a Hard Winner-Take-All Neuron Circuit,” in 2023 IEEE International Symposium on Circuits and Systems (ISCAS) , 2023, pp. 1–5. DOI: ...

    2023

  49. [64]

    Synaptic Normalisation for On-Chip Learning in Analog CMOS Spiking Neural Networks,

    M. Mastella, H. Greatorex, M. Cotteret, E. Janotte, W. Soares Girao, O. Richter, and E. Chicca, “Synaptic Normalisation for On-Chip Learning in Analog CMOS Spiking Neural Networks,” in Proceedings of the 2023 International Conference on Neuromorphic Systems, ser. ICONS ’23, Ne...

    2023

  50. [1988]

    DOI: 10.1016/0893-6080(88)90024-X

    doi

  51. [2011]

    DOI: 10.1109/JSSC.2010.2085952

    2010

  52. [2019]

    10 SUPPLEMENTARY MATERIAL S1

    DOI: 10.7554/eLife.47314. 10 SUPPLEMENTARY MATERIAL S1. TDE SYNAPSE CIRCUIT ANALYSIS The following section refers to Fig. 1b and the currents, capacitances and voltages shown on the schematic. This approach to characterise the behaviour of the TDE circuit is derived from simil...

    doi

  53. [2020]

    DOI: 10.1007/978-3-030-58526-6 22/COVER

    doi

  54. [2024]

    DOI: 10.1109/TPAMI.2024.3396116

    2024

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.