pith. sign in

arxiv: 2606.26636 · v1 · pith:YDWRZVQAnew · submitted 2026-06-25 · 💻 cs.CV · cs.LG· eess.IV

FracEvent: Event-Camera Simulation via Fractional-Relaxation Pixel Dynamics

Langyi Chen , Chuanzhi Xu , Haoxian Zhou , Pengfei Ye , Ziyu Luo , Haodong Chen , Qiang Qu , Xiaoming Chen
show 1 more author
Weidong Cai
This is my paper

Pith reviewed 2026-06-26 05:55 UTC · model grok-4.3

classification 💻 cs.CV cs.LGeess.IV
keywords event camera simulationfractional relaxationpixel dynamicsevent generationtemporal structuredownstream transferimage reconstructionoptical flow
0
0 comments X

The pith

FracEvent models event-camera pixels with fractional-relaxation voltage dynamics to produce more accurate event timing than contrast-threshold simulators.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

Collecting large-scale real event camera data is difficult due to hardware and annotation needs, so simulation is key for advancing event-based vision. Standard simulators simplify pixel responses with contrast thresholds, which often distorts event timing and hurts transfer to other tasks. FracEvent instead drives a stack of relaxation modes from log-intensity trajectories to build a continuous voltage state, emits events at threshold crossings, and keeps the modes active after each update so residual effects shape future timing. This retained memory links earlier intensity changes to later event decisions in a way prior models omit. On multiple datasets the resulting streams show better temporal structure and stronger performance when used for image reconstruction and optical flow estimation.

Core claim

FracEvent models the pixel lifecycle with fractional-relaxation voltage dynamics: a compact stack of relaxation modes is driven from a log-intensity trajectory, their responses are combined into a voltage state, ON/OFF events are emitted at threshold crossings on the continuous trajectory, and the reference is updated while the underlying modes retain memory to affect later timing.

What carries the argument

a compact stack of relaxation modes combined into voltage whose retained memory after each event influences subsequent threshold crossings

If this is right

  • Simulated event streams exhibit improved temporal structure matching real data more closely than baselines.
  • Downstream models for image reconstruction achieve stronger results when trained on FracEvent outputs.
  • Downstream models for optical flow estimation achieve stronger results when trained on FracEvent outputs.
  • The method outperforms competing simulator baselines across multiple evaluation datasets.

Where Pith is reading between the lines

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

  • Retained memory modes could support simulation of longer-term adaptation behaviors observed in physical sensors.
  • The same retained-state approach might extend to modeling other asynchronous sensors that report threshold crossings.
  • Adding spatial coupling among neighboring pixel modes could address limitations in scenes with strong local contrast changes.

Load-bearing premise

A stack of fractional relaxation modes driven from log-intensity trajectories and updated only at event times sufficiently captures the dominant temporal dynamics of real event-camera pixels.

What would settle it

Direct comparison of inter-event time distributions or voltage response shapes produced by FracEvent against recordings from a physical event camera on the same controlled log-intensity stimulus sequence.

Figures

Figures reproduced from arXiv: 2606.26636 by Chuanzhi Xu, Haodong Chen, Haoxian Zhou, Langyi Chen, Pengfei Ye, Qiang Qu, Weidong Cai, Xiaoming Chen, Ziyu Luo.

Figure 1
Figure 1. Figure 1: Overview. Given input APS frames, FracEvent converts the resulting log-brightness trajectory into an event stream through three sensor-side mechanisms: fractional voltage memory, continuous-time crossing localization, and retained-memory updates. The generated event stream is evaluated through event-stream comparison and downstream event-based tasks. Abstract Event cameras asynchronously report brightness … view at source ↗
Figure 2
Figure 2. Figure 2: Equivalent pixel-dynamics view of FracEvent. The log-intensity trajectory drives a finite fractional-memory mode stack whose weighted sum forms the sensor-side event-generation voltage state. Events are emitted by continuous-time ON/OFF threshold crossing relative to a reference state. After an event, only the reference is updated while the memory modes are preserved. [3, 20, 26]. Fast modes track immediat… view at source ↗
Figure 3
Figure 3. Figure 3: Qualitative event-stream comparison. Each row shows a spatiotemporal event plot for one 50 ms DAVIS240C win￾dow, with columns comparing real events and generated events from the same interval. Red and blue denote ON and OFF events. log10 ∆t with the 1-Wasserstein distance [33]. • Polarity measures ON/OFF balance as |PrS(p = +1) − PrR(p = +1)|. • Time surface measures whether recent events occur at similar … view at source ↗
Figure 5
Figure 5. Figure 5: Qualitative image reconstruction comparison on DAVIS240C. The figure shows three DAVIS240C examples and compares image reconstruction results from models trained with event streams synthesized by different baselines [PITH_FULL_IMAGE:figures/full_fig_p007_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Optical-flow predictions on MVSEC. The figure shows representative outdoor and indoor evaluations. Columns compare ground truth with predictions from models trained on real events or event streams synthesized by different baselines [PITH_FULL_IMAGE:figures/full_fig_p008_6.png] view at source ↗
read the original abstract

Event cameras asynchronously report brightness changes with microsecond-level temporal resolution, but real event data remain difficult to collect at scale because specialized sensors, careful synchronization, and task-specific annotations are required. Event-camera simulation is therefore important to event-based vision tasks. Most practical simulators build on contrast-threshold event generation, some with additional filtering, stochastic noise, or hand-tuned sensor parameters. While effective, such formulations often simplify the temporal structure produced by the lifecycle of each pixel, which can distort event timing and weaken downstream transfer. We introduce FracEvent, an event simulator that models this pixel-level lifecycle with fractional-relaxation voltage dynamics. Given a log-intensity trajectory, FracEvent drives a compact stack of relaxation modes, combines their responses into a voltage state, emits ON/OFF events by localizing threshold crossings on the continuous voltage trajectory, and updates the reference while retaining the underlying memory modes. This retained state links residual voltage response to later event timing. We evaluate FracEvent through event-stream comparison and downstream transfer on image reconstruction and optical flow estimation. Across multiple datasets, FracEvent improves the temporal structure of generated events and achieves stronger downstream-transfer results than competing simulator baselines, showing its practical value for event-camera simulation.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 0 minor

Summary. The paper introduces FracEvent, an event-camera simulator that models each pixel's voltage lifecycle via a compact stack of fractional-relaxation modes driven from log-intensity trajectories. Events are emitted by localizing threshold crossings on the continuous voltage state; after each event the reference is updated while the underlying memory modes are retained to influence subsequent timing. The method is evaluated via event-stream statistics and downstream transfer on image reconstruction and optical flow tasks, with the claim that it produces more realistic temporal event structure and stronger transfer performance than prior simulators across multiple datasets.

Significance. If the fractional-relaxation formulation demonstrably improves fidelity to real pixel dynamics, the simulator could narrow the sim-to-real gap for event-based vision, enabling more reliable synthetic data for training and evaluation. The explicit retention of memory modes after reference updates is a concrete modeling choice that could preserve temporal dependencies not captured by simpler threshold or filtered models.

major comments (3)
  1. [Abstract] Abstract: the central claim that FracEvent 'improves the temporal structure of generated events and achieves stronger downstream-transfer results' is stated without any quantitative metrics (e.g., event-rate histograms, timing-error distributions, or transfer-task accuracies with error bars), rendering the magnitude and reliability of the reported gains impossible to assess.
  2. [Pixel lifecycle modeling paragraph] Pixel-lifecycle-modeling paragraph (and associated method description): the assumption that driving a stack of fractional relaxation modes from log-intensity trajectories and updating only at event times sufficiently reproduces dominant real-pixel voltage dynamics is load-bearing for the entire contribution, yet no direct comparison to measured pixel responses, continuous-time simulations, or non-fractional baselines is supplied to test this assumption.
  3. [Evaluation section] Evaluation section: no ablation results are reported on the number of relaxation modes in the stack or on the choice of fractional orders, both of which are free parameters introduced by the method and directly affect the claimed temporal-structure improvements.

Simulated Author's Rebuttal

3 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 claim that FracEvent 'improves the temporal structure of generated events and achieves stronger downstream-transfer results' is stated without any quantitative metrics (e.g., event-rate histograms, timing-error distributions, or transfer-task accuracies with error bars), rendering the magnitude and reliability of the reported gains impossible to assess.

    Authors: We agree that the abstract would be strengthened by quantitative support. The revised abstract will include key metrics such as timing-error reductions and downstream accuracy gains with error bars drawn from the evaluation results. revision: yes

  2. Referee: [Pixel lifecycle modeling paragraph] Pixel-lifecycle-modeling paragraph (and associated method description): the assumption that driving a stack of fractional relaxation modes from log-intensity trajectories and updating only at event times sufficiently reproduces dominant real-pixel voltage dynamics is load-bearing for the entire contribution, yet no direct comparison to measured pixel responses, continuous-time simulations, or non-fractional baselines is supplied to test this assumption.

    Authors: The current validation uses event-stream statistics and transfer performance as indirect evidence. We will add direct comparisons against non-fractional baselines and continuous-time simulations. Measured real-pixel responses are not available in the public datasets used, so that specific comparison cannot be supplied. revision: partial

  3. Referee: [Evaluation section] Evaluation section: no ablation results are reported on the number of relaxation modes in the stack or on the choice of fractional orders, both of which are free parameters introduced by the method and directly affect the claimed temporal-structure improvements.

    Authors: We agree that ablations on these parameters are necessary. The revised evaluation will report results for varying numbers of modes and fractional orders, quantifying their effect on temporal structure and transfer performance. revision: yes

Circularity Check

0 steps flagged

No significant circularity; modeling choice is independent

full rationale

The paper presents FracEvent as an explicit modeling decision: a stack of fractional-relaxation modes driven from log-intensity trajectories, with events emitted at continuous threshold crossings and memory modes retained after reference updates. This is introduced as a new ansatz for pixel lifecycle dynamics rather than derived from prior equations or fitted parameters within the paper. No self-citations are invoked as load-bearing uniqueness theorems, no parameters are fitted to a subset and then relabeled as predictions, and the reported gains in event timing and downstream transfer are framed as empirical outcomes on external datasets. The derivation chain therefore remains self-contained against the stated modeling assumptions.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Only the abstract is available; no explicit free parameters, axioms, or invented entities are stated. The modeling choice of fractional relaxation itself functions as an unverified domain assumption.

pith-pipeline@v0.9.1-grok · 5780 in / 1000 out tokens · 18050 ms · 2026-06-26T05:55:26.904335+00:00 · methodology

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

17 extracted references

  1. [1]

    A 240 x 180 130 db 3 us latency global shutter spatiotemporal vision sensor.IEEE Journal of Solid-State Circuits, 49(10):2333–2341, 2014

    Christian Brandli, Raphael Berner, Minhao Yang, Shih-Chii Liu, and Tobi Delbruck. A 240 x 180 130 db 3 us latency global shutter spatiotemporal vision sensor.IEEE Journal of Solid-State Circuits, 49(10):2333–2341, 2014

    2014

  2. [2]

    EVREAL: Towards a comprehensive benchmark and analysis suite for event-based video reconstruction

    Burak Ercan, Onur Eker, Aykut Erdem, and Erkut Erdem. EVREAL: Towards a comprehensive benchmark and analysis suite for event-based video reconstruction. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, pages 3943–3952. IEEE, 2023

    2023

  3. [3]

    Video to events: Recycling video datasets for event cameras

    Daniel Gehrig, Mathias Gehrig, Javier Hidalgo-Carrio, and Davide Scaramuzza. Video to events: Recycling video datasets for event cameras. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 3586–3595, 2020

    2020

  4. [4]

    v2e: From video frames to realistic DVS events

    Yuhuang Hu, Shih-Chii Liu, and Tobi Delbruck. v2e: From video frames to realistic DVS events. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, pages 1312–1321, 2021

    2021

  5. [5]

    Shi, and Ryad B

    Xavier Lagorce, Garrick Orchard, Francesco Galluppi, Bertram E. Shi, and Ryad B. Benosman. HOTS: A hierarchy of event-based time-surfaces for pattern recognition.IEEE Transactions on Pattern Analysis and Machine Intelligence, 39(7):1346–1359, 2017

    2017

  6. [6]

    DVS-V oltmeter: Stochastic process-based event simulator for dynamic vision sensors

    Songnan Lin, Ye Ma, Zhenhua Guo, and Bihan Wen. DVS-V oltmeter: Stochastic process-based event simulator for dynamic vision sensors. InComputer Vision – ECCV 2022, pages 578–593. Springer Nature Switzerland, 2022

    2022

  7. [7]

    The event-camera dataset and simulator: Event-based data for pose estimation, visual odometry, and SLAM.The International Journal of Robotics Research, 36(2):142–149, 2017

    Elias Mueggler, Henri Rebecq, Guillermo Gallego, Tobi Delbruck, and Davide Scaramuzza. The event-camera dataset and simulator: Event-based data for pose estimation, visual odometry, and SLAM.The International Journal of Robotics Research, 36(2):142–149, 2017

    2017

  8. [8]

    Deep multi-scale convolutional neural network for dynamic scene deblurring

    Seungjun Nah, Tae Hyun Kim, and Kyoung Mu Lee. Deep multi-scale convolutional neural network for dynamic scene deblurring. InProceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages 3883–3891, 2017

    2017

  9. [9]

    ESIM: An open event camera simulator

    Henri Rebecq, Daniel Gehrig, and Davide Scaramuzza. ESIM: An open event camera simulator. InProceedings of The 2nd Confer- ence on Robot Learning, pages 969–982. PMLR, 2018

    2018

  10. [10]

    Events-to-video: Bringing modern computer vision to event cameras

    Henri Rebecq, Rene Ranftl, Vladlen Koltun, and Davide Scaramuzza. Events-to-video: Bringing modern computer vision to event cameras. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 3857–3866, 2019

    2019

  11. [11]

    HATS: Histograms of averaged time surfaces for robust event-based object classification

    Amos Sironi, Manuele Brambilla, Nicolas Bourdis, Xavier Lagorce, and Ryad Benosman. HATS: Histograms of averaged time surfaces for robust event-based object classification. InProceedings of the IEEE Conference on Computer Vision and Pattern Recog- nition, pages 1731–1740, 2018

    2018

  12. [12]

    Reducing the sim-to-real gap for event cameras

    Timo Stoffregen, Cedric Scheerlinck, Davide Scaramuzza, Tom Drummond, Nick Barnes, Lindsay Kleeman, and Robert Mahony. Reducing the sim-to-real gap for event cameras. InEuropean Conference on Computer Vision, pages 534–549. Springer, 2020

    2020

  13. [13]

    Springer, Berlin, Heidelberg, 2009

    Cedric Villani.Optimal Transport: Old and New. Springer, Berlin, Heidelberg, 2009

    2009

  14. [14]

    Bovik, Hamid R

    Zhou Wang, Alan C. Bovik, Hamid R. Sheikh, and Eero P. Simoncelli. Image quality assessment: From error visibility to structural similarity.IEEE Transactions on Image Processing, 13(4):600–612, 2004

    2004

  15. [15]

    Efros, Eli Shechtman, and Oliver Wang

    Richard Zhang, Phillip Isola, Alexei A. Efros, Eli Shechtman, and Oliver Wang. The unreasonable effectiveness of deep features as a perceptual metric. InProceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages 586–595, 2018

    2018

  16. [16]

    The multi vehicle stereo event camera dataset: An event camera dataset for 3D perception.IEEE Robotics and Automation Letters, 3(3):2032–2039, 2018

    Alex Zihao Zhu, Dinesh Thakur, Tolga Ozaslan, Bernd Pfrommer, Vijay Kumar, and Kostas Daniilidis. The multi vehicle stereo event camera dataset: An event camera dataset for 3D perception.IEEE Robotics and Automation Letters, 3(3):2032–2039, 2018

    2032

  17. [17]

    EV-FlowNet: Self-supervised optical flow estimation for event-based cameras

    Alex Zihao Zhu, Liangzhe Yuan, Kenneth Chaney, and Kostas Daniilidis. EV-FlowNet: Self-supervised optical flow estimation for event-based cameras. InRobotics: Science and Systems, 2018. 23

    2018