Event-camera simulation

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

An event camera simulator that keeps pixel-level voltage memory and localizes ON/OFF threshold crossings inside frame intervals.

Langyi Chen¹, Chuanzhi Xu¹, Haoxian Zhou¹, Pengfei Ye², Ziyu Luo³, Haodong Chen¹, Qiang Qu¹, Xiaoming Chen³, Weidong Cai¹

¹The University of Sydney · ²Massachusetts Institute of Technology · ³Beijing Technology and Business University

Code Paper arXiv Bib

Overview of the FracEvent simulator and evaluation workflow — FracEvent converts APS frames into simulated event streams through fractional voltage memory, continuous-time crossing localization, and retained-memory reference updates.

Abstract

Simulating events by modeling the pixel lifecycle.

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.

FracEvent models this pixel-level lifecycle with fractional-relaxation voltage dynamics. Given a log-intensity trajectory, it 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.

The retained state links residual voltage response to later event timing. We evaluate FracEvent through direct event-stream comparison and downstream transfer on image reconstruction and optical flow estimation.

Method

Fractional memory, continuous crossing, retained state.

Pixel event lifecycle

Each pixel keeps a finite stack of relaxation modes instead of treating frame differences as isolated contrast increments.

Sub-frame timing

Events are localized on the voltage trajectory inside each frame interval, avoiding frame-boundary-only emission.

Residual memory

Event references update after threshold crossings, while the relaxation modes are carried into the next interval.

Input Log-intensity trajectory

Frames provide the continuous interval endpoints used by the simulator.

Memory Fractional modes

Weighted relaxation channels approximate broad temporal memory.

Events Threshold roots

Active pixels are solved for ON/OFF crossing times within the interval.

Carry Reference update

References advance by emitted levels while mode state remains available.

Detailed FracEvent method diagram — FracEvent combines fractional-relaxation pixel dynamics, threshold crossing, and retained memory state.

Results

Visualization and evaluation results.

Visualization

DAVIS240C APS frames on the left, accumulated FracEvent events on the right.

MVSEC APS frames on the left, accumulated FracEvent events on the right.

Visual comparison examples for image reconstruction.

Visual comparison examples for optical flow estimation.

Eval result

Event-stream comparison

On matched DAVIS240C windows, FracEvent achieves the lowest IEI distance and polarity error among the compared simulators.

Image reconstruction

With the fixed E2VID-style training protocol, FracEvent gives the lowest DAVIS240C MSE (0.030) and LPIPS (0.460); ESIM has the highest DAVIS240C SSIM.

Optical flow

For EV-FlowNet-style training, FracEvent reaches 2.68 mean AEE, the lowest among simulated training sources and closest to training with real MVSEC events at 2.42.

Citation

Cite FracEvent.

@misc{chen2026fracevent,
      title={FracEvent: Event-Camera Simulation via Fractional-Relaxation Pixel Dynamics},
      author={Langyi Chen and Chuanzhi Xu and Haoxian Zhou and Pengfei Ye and Ziyu Luo and Haodong Chen and Qiang Qu and Xiaoming Chen and Weidong Cai},
      year={2026},
      eprint={2606.26636},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2606.26636},
}