Pith. sign in

REVIEW 4 major objections 7 minor 60 references

EventTracer: Fast Path Tracing-based Event Stream Rendering

T0 review · 4 major / 7 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read EventTracer: realistic event streams at 1,000 FPS from a cheap noisy renderer plus a bipolar spiking denoiser.

desk verdict Clever pipeline, speed real, but the fidelity comparison confounds temporal resolution with sensor-model realism. read the letter →

arxiv 2508.18071 v2 pith:XRBNGTCB submitted 2025-08-25 cs.CV

Zhenyang Li , Xiaoyang Bai , Jinfan Lu , Pengfei Shen , Edmund Y. Lam , Yifan Peng This is my paper
classification cs.CV
keywords eventcamerasimulationpathtracingspikingneuralnetworksbipolarLIFevent-to-videoreconstructionevent-baseddepthestimationreal2simevaluationsyntheticdata
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

The paper claims that realistic event-camera data can be synthesized from dynamic 3D scenes without the expensive noise-free rendering that limits existing simulators to 100–300 FPS. It renders with only 64 samples per pixel—one thirty-second of the usual 2,048—and hands the resulting noisy, high-frame-rate video to a lightweight spiking network whose bipolar integrate-and-fire neuron converts pixel illuminance changes into positive and negative event pulses. The network is trained against V2E-generated ground truth with a bidirectional earth-mover-distance loss, and it reproduces two sensor behaviors that frame-difference simulators miss: per-pixel internal-state bias and post-fire saturation. If the claim holds, EventTracer produces 1,000-FPS event streams at 1.12 minutes per second of 360p video, making large-scale synthetic event-RGB training data cheap, with depth and other annotations nearly free from the path-tracing backbone. The authors' Real2Sim evaluation—pretrained event-to-video and event-to-depth models scoring higher on their synthetic events than on V2E's or V2CE's—supports the realism claim.

What carries the argument

The BiLIF (bipolar leaky integrate-and-fire) unit—a spiking neuron with symmetric thresholds ±Vth that integrates log-illuminance differences over time, emits +1/0/−1 spikes, and resets after firing—is the object that carries the argument. It reproduces internal-state bias (randomly initialized membrane voltages make identical brightness changes fire at different times pixel by pixel) and the saturation effect (post-reset trailing pulses after large changes), both of which frame-difference event simulators miss. Training combines a bidirectional 1D earth-mover loss, computed on cumulative sums in both directions and separated by polarity, with a spike-count regularizer; the whole pixel-wise

What would settle it

Project or display the same dynamic 3D scene as both an EventTracer stream and a V2E stream, capture each with a real event camera, and compare the captured event statistics (inter-spike-interval distributions, event-rate histograms, and reconstruction error of a fixed E2V model against known ground-truth frames). If EventTracer's captured events are not measurably closer to a real camera's response to the same scene than V2E's—or if the Real2Sim advantage vanishes when both are tested against matched real captures—the central realism claim is refuted.

Watch

Extended reading notes

Core claim

EventTracer's central claim is that the bottleneck in event simulation is not rendering cost but the demand for noiseless frames: existing pipelines render clean RGB at an equivalent 100–300 FPS, far below the ≥1,000 FPS of real event cameras. The paper shows that path tracing at 64 SPP (1/32 the usual cost) followed by a small temporal convolutional network ending in a bipolar LIF spike layer yields streams that pretrained event-to-video and event-to-depth models handle better than streams from V2E or V2CE. The spiking unit carries the physical modeling: symmetric thresholds emit both polarities, the leaky membrane stores each pixel's internal state across the clip (internal-state bias), an

Load-bearing premise

EventTracer's realism is anchored to V2E: the network is trained to imitate V2E's output, so the pipeline can only be as faithful to a real camera as V2E's sensor model, and the claimed superiority over V2E rests on the assumption that low-SPP render noise plus learned denoising adds realistic sensor noise rather than artifacts.

Editorial extensions

If this is right

  • Event-RGB training data at true event-camera temporal resolution becomes cheap enough for large-scale synthesis: 1.12 minutes of compute per second of 360p video at 1,000 FPS, beating the full V2E pipeline's 1.85 minutes and V2CE's effective cost when run at 60-FPS input.
  • Because the backbone is full path tracing, per-pixel depth, surface normals, segmentation, and optical flow come at almost no extra cost, extending synthetic event datasets beyond RGB-event pairs.
  • The trained EvSNet amortizes its 12+ hour training cost across arbitrary new scenes, and since the pipeline is differentiable it can serve as a rendering engine for optimization as well as dataset generation.
  • Real2Sim scores on both event-to-video and event-to-depth show a smaller sim-to-real gap than V2E or V2CE, supporting the use of synthetic events for pretraining perception stacks in robotics and autonomous driving.

Reading between the lines

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

  • The recipe—cheap stochastic renderer plus a learned, physics-constrained pulse model—is a general pattern: the same division of labor could simulate other sensing artifacts, such as rolling-shutter skew, fixed-pattern noise, or photon-shot-noise-limited sensors, where the noise is itself informative.
  • Because EvSNet imitates V2E-generated ground truth, EventTracer's realism ceiling is set by V2E's sensor model; retraining the same architecture on real event captures or on a more detailed sensor model should lift the ceiling without changing the pipeline.
  • The paper acknowledges its spike-count loss only loosely constrains total event counts, which it says hurts event-to-video reconstruction; adding explicit count-regularized training is a direct, paper-suggested path to improvement that would also make the E2V results tighter.
  • The paper's own Sim2Real check shows the 90-second ETScenes corpus is too small and static to train production models; the next decisive test is whether a much larger ETScenes-style corpus with driving-scale dynamics yields models that transfer to real event data.
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 / 7 minor

Summary. The paper proposes EventTracer, a pipeline for synthesizing event streams from dynamic 3D scenes. It combines low-SPP Monte Carlo path tracing with a lightweight event spiking network (EvSNet) that converts noisy rendered illuminance sequences into event spikes. EvSNet uses a bipolar leaky integrate-and-fire (BiLIF) unit and is trained with a bidirectional EMD loss plus an event-count loss. The authors report generation at 1,000 FPS with a speed of 1.12 minutes per second of 360p video, and construct the ETScenes dataset. They evaluate realism through Real2Sim tests on event-to-video reconstruction (EVREAL) and event-to-depth estimation (e2depth), comparing against V2E and V2CE. The paper claims that EventTracer produces events with greater similarity to real-world data than these baselines.

Significance. If the central claim held, EventTracer would be a practically valuable tool for generating large-scale, high-temporal-resolution event-RGB datasets with auxiliary annotations, at lower cost than existing physics-based simulators. The pipeline design is reasonable: low-SPP rendering plus a learned denoiser is a sensible way to accelerate path tracing, and the integration with a spiking network is novel. The paper reports clear runtime numbers and a reproducible training setup. However, the fidelity claim is not established by the reported experiments because the comparison is confounded by temporal resolution and the training target is V2E itself. The paper's own Sim2Real section reports poor transfer, further undermining the abstract's claim of narrowing the sim-to-real gap. The contribution is promising but requires a controlled evaluation before it can be accepted as presented.

major comments (4)
  1. [Sec. 4.2–4.4] The Real2Sim comparison is confounded by temporal resolution. EventTracer generates events at 1,000 FPS, while the V2E and V2CE baselines are fed RGB videos at 60 FPS and 15 FPS respectively (Sec. 4.2: 'from RGB videos whose frame rate does not exceed 60 FPS for a fair comparison'). For both E2V reconstruction and E2D estimation, models generally perform better with denser event streams, so the improved scores in Tables 1 and 2 could simply reflect the higher event rate rather than a more realistic sensor model. The label 'for a fair comparison' is misleading. To support the claim of greater fidelity, the authors should compare at matched temporal resolution—for example, by temporally downsampling EventTracer's 1,000 FPS output to 60 FPS, or by running V2E on 1,000 FPS high-SPP rendered frames—and show that the advantage persists.
  2. [Sec. 3.3] EvSNet is trained to match V2E-generated events: 'we render high-quality, Fevt FPS videos with high-SPP path tracing and feed them into the V2E toolbox to generate ground truth events' (Sec. 3.3). Therefore EventTracer is, by construction, a learned approximation of V2E at a higher temporal sampling rate. It cannot genuinely surpass V2E's sensor-model realism unless the network injects additional validated noise characteristics, but no evidence is provided that the learned deviations from V2E improve realism. The Real2Sim superiority, if it survives the temporal-resolution control, still conflates the denoiser's effect with the higher input rate. An experiment training EvSNet on real-captured events, or comparing against V2E with equal input frame rate, is needed to separate these factors.
  3. [Sec. 5 (Bells & whistles: Sim2Real)] The paper's own Sim2Real section states that models trained on ETScenes 'lead to low-quality results on datasets such as UZH DA VIS and MVSEC' and attributes this to a 'domain gap' and insufficient scale. This directly contradicts the abstract's claim that EventTracer 'narrow[s] the sim-to-real gap.' The authors place this limitation after the conclusion under an informal heading, which obscures its weight. The abstract and introduction should be revised to reflect the actual evidence, and the Sim2Real results should be moved into the main evaluation if they are to be considered part of the validation.
  4. [Sec. 4.3] The evaluation protocol does not control for event count or rate. The energy-of-Laplacian (LPA) metric favors sharp reconstructions, and higher event rates naturally yield sharper E2V outputs. While SSIM and LPIPS are also reported, all three metrics are computed on reconstructions from event streams of vastly different densities. To establish realism, the authors should either match the event rate or report results as a function of event rate. Additionally, the use of histogram equalization on both ground truth and reconstruction may hide intensity-level errors that are relevant to event fidelity.
minor comments (7)
  1. [Abstract] Typo: 'establishes its as a promising tool' should be 'establishes it as a promising tool'.
  2. [Sec. 1] Introduction: 'event sensory have been widely integrated' should be 'event sensors have been widely integrated'; 'levitates its prospect' is awkward and unclear.
  3. [Fig. 1 caption] The caption for Figure 1 appears to be from a different manuscript: it discusses 'Deformation-based 4DGS' and 'FAD strategy', none of which are described in this paper. Please replace with a caption that describes the actual pipeline and qualitative comparison.
  4. [Sec. 4.4] Table 2 lists five metrics, but the text says 'achieves better scores in all three metrics.' The number should be corrected.
  5. [Sec. 4.5] Grammar: 'Were 60 FPS inputs are used for V2CE' should be 'If 60 FPS inputs are used for V2CE'.
  6. [Sec. 5] The heading 'Bells & whistles: Sim2Real' is informal and confusingly placed after the conclusion. This section should be integrated into the validation section, with a descriptive heading such as 'Sim2Real Evaluation.'
  7. [Table 1] Table 1's format with slash-separated triplets is difficult to read. Please split SSIM, LPIPS, and LPA into separate columns or clearly dedicated sub-tables.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity; EvSNet is trained on V2E output, but the central Real2Sim claim is evaluated with external real-world-trained models, so it is not a circular derivation.

full rationale

The paper's core derivation is not circular. EventTracer's EvSNet is trained to reproduce V2E-generated events (Sec. 3.3: 'we render high-quality, Fevt FPS videos with high-SPP path tracing and feed them into the V2E toolbox [9] to generate ground truth events E'), so EventTracer's sensor-noise realism is inherited from V2E in a narrow sense. However, the claimed 'greater similarity to real-world event data than other event simulators' is supported by Real2Sim evaluations using pretrained models (EVREAL for event-to-video, e2depth for event-to-depth) that were trained on real or mixed real-synthetic datasets, not by the training loss itself. The comparison is confounded: EventTracer operates at 1,000 FPS while V2E and V2CE are fed 60 FPS and 15 FPS RGB inputs (Sec. 4.2), and the ETScenes evaluation set includes the same Bistro scenes used for training EvSNet. These are experimental validity limitations, not circular reductions: no equation in the paper defines EventTracer's superiority over V2E in terms of its training inputs, and no load-bearing claim rests solely on an author self-citation. The author-overlapping references (e.g., [23], [59]) are used only as related work or future-work motivation, not to justify the core simulation result. Thus the circularity score is 0, with the noted confounds belonging to correctness risk rather than circularity.

Assumptions & free parameters 6 free parameters · 4 assumptions · 1 invented entities

The pipeline relies on a learned network trained on V2E output, several unreported hyperparameters, and the assumption that Real2Sim evaluation reflects real-world fidelity. No code or dataset is released.

free parameters (6)
  • EvSNet network weights = trained on 7,936 frames from Bistro scenes
    Learned mapping from low-SPP renders to V2E-like events; central to the pipeline.
  • Contrast threshold rho (Eq. 1)
    Predefined threshold in the log-luminance transform; value not specified in the paper.
  • BiLIF decay tau and threshold Vth (Eqs. 2-3)
    Hyperparameters controlling spiking dynamics; values not reported.
  • Loss weight lambda (Eq. 11)
    Weight balancing EMD and count losses; value not reported.
  • SPP = 64 = 64
    Chosen to trade rendering speed for noise; N=64 reduces runtime to 1/32 of 2048 SPP.
  • Receptive field width W = 43 = 43
    Determined by kernel size 7 and depth M=3; fixed by architecture design.
assumptions (4)
  • domain assumption V2E produces realistic event ground truth
    Sec. 3.3 uses V2E-generated events as training targets; if V2E is not faithful to real sensors, EvSNet inherits that bias.
  • domain assumption Low-SPP path tracing plus learned denoising preserves event-relevant information
    The pipeline assumes 64 SPP renders contain enough signal to generate accurate events despite heavy noise.
  • domain assumption Real2Sim scores with pretrained models measure similarity to real data
    Sec. 4.2 uses the Real2Sim protocol; assumes downstream performance on simulated data is a valid proxy for data fidelity.
  • standard math Standard math results (EMD formula, surrogate gradients)
    Used for the loss function and gradient-based training of the spiking network.
invented entities (1)
  • BiLIF neuron (bipolar leaky integrate-and-fire)
    purpose: Emit positive and negative event spikes with saturation effect, mimicking event camera behavior
    Described in Sec. 3.3; no external evidence that it matches real sensor dynamics beyond the paper's own claims.

how reviews work

0 comments
Cite this review

Pith. "Pith review of EventTracer: Fast Path Tracing-based Event Stream Rendering." pith.science (2026). https://pith.science/paper/XRBNGTCB

@misc{pith2026250818071,
  author       = {Pith},
  title        = {Pith review of: EventTracer: Fast Path Tracing-based Event Stream Rendering},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/XRBNGTCB}},
  note         = {Machine review of arXiv:2508.18071}
}
read the original abstract

Simulating event streams from 3D scenes has become a common practice in event-based vision research, as it meets the demand for large-scale, high temporal frequency data without setting up expensive hardware devices or undertaking extensive data collections. Yet existing methods in this direction typically work with noiseless RGB frames that are costly to render, and therefore they can only achieve a temporal resolution equivalent to 100-300 FPS, far lower than that of real-world event data. In this work, we propose EventTracer, a path tracing-based rendering pipeline that simulates high-fidelity event sequences from complex 3D scenes in an efficient and physics-aware manner. Specifically, we speed up the rendering process via low sample-per-pixel (SPP) path tracing, and train a lightweight event spiking network to denoise the resulting RGB videos into realistic event sequences. To capture the physical properties of event streams, the network is equipped with a bipolar leaky integrate-and-fired (BiLIF) spiking unit and trained with a bidirectional earth mover distance (EMD) loss. Our EventTracer pipeline runs at a speed of about 4 minutes per second of 720p video, and it inherits the merit of accurate spatiotemporal modeling from its path tracing backbone. We show in two downstream tasks that EventTracer captures better scene details and demonstrates a greater similarity to real-world event data than other event simulators, which establishes it as a promising tool for creating large-scale event-RGB datasets at a low cost, narrowing the sim-to-real gap in event-based vision, and boosting various application scenarios such as robotics, autonomous driving, and VRAR.

Figures

Figures reproduced from arXiv: 2508.18071 by the authors.

Figure 1
Figure 1. Left: Deformation-based 4DGS encounters difficulties in reconstructing scenes and rendering novel views under challenging conditions, such as significant motion and other complex dynamics. Our EvSNet exhibits com￾mendable performance on given scenes (e.g., “cut-roasted-beef”). Middle: We compare the ground-truth optical flow, the deformation network of baseline (4DGS), and the velocity field rendered by our method. … view at source ↗
Figure 2
Figure 2. Left: (top) Each pixel is initialized with a randomly sampled internal state; (bottom) Identical brightness changes are applied to all pixels, causing them to fire events at different timestamps. Right: Internal State Bias: Two event pixels with different initial states (S1 and S2) produce drastically different event pulses (blue and red arrows) even when they undergo identical illuminance changes; Saturation Bias: … view at source ↗
Figure 3
Figure 3. Pipeline of our EventTracer. Bottom left: Our task is to directly render event streams from 3D scenes. Top-left: In each scene, we define a camera trajectory and run the path tracer at varying sample-per-pixel (SPP) to obtain low-SPP (noisy, high FPS) and high-SPP (noiseless, low FPS) RGB videos. Top-right: Our EvSNet network takes noisy pixel illuminances as input and outputs discrete event spikes via the BiLIF uni… view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Comparison of events obtained using various rendering approaches–EventTracer, V2E, and V2CE–on three [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: Histograms of event intensity generated by three investigated simulators. We take all frames in the [PITH_FULL_IMAGE:figures/full_fig_p009_5.png]
Figure 6
Figure 6. Figure 6: Real2Sim validation results on E2V. Each column contains the ground truth grayscale frame as well as pairs of reconstructed image and event frame for each event simulator. When evaluated on our ETScenes dataset, the pretrained FireNet [55] model is able to reconstruct …
Figure 7
Figure 7. Figure 7: Real2Sim validation results on E2D. Each column contains the ground truth depth map as well as pairs of estimated depth and event frame for each event simulator. Note that ETScenes yields the best results of three, where not only detailed structures (e.g. desks, stairs…

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

60 extracted references · 48 canonical work pages

  1. [1]

    Event-based Vision: A Survey

    Guillermo Gallego, Tobi Delbruck, Garrick Orchard, Chiara Bartolozzi, Brian Taba, Andrea Censi, Stefan Leutenegger, Andrew Davison, Joerg Conradt, Kostas Daniilidis, and Davide Scaramuzza. Event-based Vision: A Survey. IEEE Transactions on Pattern Analysis and Machine Intelligence, 44(1):154–180, January 2022

    work page 2022

  2. [2]

    Event based, near eye gaze tracking beyond 10,000 hz

    Anastasios N Angelopoulos, Julien NP Martel, Amit PS Kohli, Jorg Conradt, and Gordon Wetzstein. Event based, near eye gaze tracking beyond 10,000 hz. arXiv preprint arXiv:2004.03577, 2020

    arXiv 2004

  3. [3]

    Taming event cameras with bio-inspired architecture and algorithm: a case for drone obstacle avoidance

    Jingao Xu, Danyang Li, Zheng Yang, Yishujie Zhao, Hao Cao, Yunhao Liu, and Longfei Shangguan. Taming event cameras with bio-inspired architecture and algorithm: a case for drone obstacle avoidance. In Proceedings of the 29th annual international conference on mobile computing and networking, pages 1–16, 2023

    work page 2023

  4. [4]

    Low-latency automotive vision with event cameras

    Daniel Gehrig and Davide Scaramuzza. Low-latency automotive vision with event cameras. Nature, 629(8014):1034–1040, 2024

    work page 2024

  5. [5]

    Deep learning for event-based vision: A comprehensive survey and benchmarks

    Xu Zheng, Yexin Liu, Yunfan Lu, Tongyan Hua, Tianbo Pan, Weiming Zhang, Dacheng Tao, and Lin Wang. Deep learning for event-based vision: A comprehensive survey and benchmarks. arXiv preprint arXiv:2302.08890, 2023

    arXiv 2023

  6. [6]

    A 240× 180 130 db 3 µs latency global shutter spatiotemporal vision sensor

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

    2014

  7. [7]

    Visevent: Reliable object tracking via collaboration of frame and event flows

    Xiao Wang, Jianing Li, Lin Zhu, Zhipeng Zhang, Zhe Chen, Xin Li, Yaowei Wang, Yonghong Tian, and Feng Wu. Visevent: Reliable object tracking via collaboration of frame and event flows. IEEE Transactions on Cybernetics, 54(3):1997–2010, 2023

    work page 1997

  8. [8]

    Even- taid: Benchmarking event-aided image/video enhancement algorithms with real-captured hybrid dataset

    Peiqi Duan, Boyu Li, Yixin Yang, Hanyue Lou, Minggui Teng, Xinyu Zhou, Yi Ma, and Boxin Shi. Even- taid: Benchmarking event-aided image/video enhancement algorithms with real-captured hybrid dataset. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2025

    work page 2025

Show all 60 references
  1. [9]

    V2e: From Video Frames to Realistic DVS Events

    Yuhuang Hu, Shih-Chii Liu, and Tobi Delbruck. V2e: From Video Frames to Realistic DVS Events. 2021

    2021

  2. [10]

    V2CE: Video to Continuous Events Simulator, April 2024

    Zhongyang Zhang, Shuyang Cui, Kaidong Chai, Haowen Yu, Subhasis Dasgupta, Upal Mahbub, and Tauhidur Rahman. V2CE: Video to Continuous Events Simulator, April 2024

    2024

  3. [11]

    ESIM: An Open Event Camera Simulator

    Henri Rebecq, Daniel Gehrig, and Davide Scaramuzza. ESIM: An Open Event Camera Simulator. In Proceedings of The 2nd Conference on Robot Learning, pages 969–982. PMLR, October 2018. 12 EventTracer: Fast Path Tracing-based Event Stream Rendering

    2018

  4. [12]

    Physical-Based Event Camera Simulator

    Haiqian Han, Jiacheng Lyu, Jianing Li, Henglu Wei, Cheng Li, Yajing Wei, Shu Chen, and Xiangyang Ji. Physical-Based Event Camera Simulator. December 2024

    2024

  5. [13]

    BlinkFlow: A Dataset to Push the Limits of Event-based Optical Flow Estimation, October 2024

    Yijin Li, Zhaoyang Huang, Shuo Chen, Xiaoyu Shi, Hongsheng Li, Hujun Bao, Zhaopeng Cui, and Guofeng Zhang. BlinkFlow: A Dataset to Push the Limits of Event-based Optical Flow Estimation, October 2024

    2024

  6. [14]

    The Falcor rendering framework, 8 2022

    Simon Kallweit, Petrik Clarberg, Craig Kolb, Tom’aš Davidoviˇc, Kai-Hwa Yao, Theresa Foley, Yong He, Lifan Wu, Lucy Chen, Tomas Akenine-Möller, Chris Wyman, Cyril Crassin, and Nir Benty. The Falcor rendering framework, 8 2022. https://github.com/NVIDIAGameWorks/Falcor

    2022

  7. [15]

    Nvidia tensorrt, 2025

    NVIDIA. Nvidia tensorrt, 2025

    2025

  8. [16]

    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. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 3943–3952, 2023

    2023

  9. [17]

    Learning monocular dense depth from events

    Javier Hidalgo-Carrió, Daniel Gehrig, and Davide Scaramuzza. Learning monocular dense depth from events. In 2020 International Conference on 3D Vision (3DV), pages 534–542. IEEE, 2020

    2020

  10. [18]

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

    Patrick Lichtsteiner, Christoph Posch, and Tobi Delbruck. A 128 × 128 120 db 15 µs latency asynchronous temporal contrast vision sensor. IEEE journal of solid-state circuits, 43(2):566–576, 2008

    2008

  11. [19]

    Event-based 3d motion flow estimation using 4d spatio temporal subspaces properties

    Sio-Hoi Ieng, João Carneiro, and Ryad B Benosman. Event-based 3d motion flow estimation using 4d spatio temporal subspaces properties. Frontiers in Neuroscience, 10:596, 2017

    2017

  12. [20]

    Learning to reconstruct high speed and high dynamic range videos from events

    Yunhao Zou, Yinqiang Zheng, Tsuyoshi Takatani, and Ying Fu. Learning to reconstruct high speed and high dynamic range videos from events. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 2024–2033, 2021

    2024

  13. [21]

    Eventps: Real-time photometric stereo using an event camera

    Bohan Yu, Jieji Ren, Jin Han, Feishi Wang, Jinxiu Liang, and Boxin Shi. Eventps: Real-time photometric stereo using an event camera. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 9602–9611, 2024

    2024

  14. [22]

    All-in-Focus Imaging from Event Focal Stack

    Hanyue Lou, Minggui Teng, Yixin Yang, and Boxin Shi. All-in-Focus Imaging from Event Focal Stack. In 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 17366–17375, Vancouver, BC, Canada, June 2023. IEEE

    2023

  15. [23]

    Em- bodied neuromorphic synergy for lighting-robust machine vision to see in extreme bright.Nature Communications, 15(1):10781, 2024

    Shijie Lin, Guangze Zheng, Ziwei Wang, Ruihua Han, Wanli Xing, Zeqing Zhang, Yifan Peng, and Jia Pan. Em- bodied neuromorphic synergy for lighting-robust machine vision to see in extreme bright.Nature Communications, 15(1):10781, 2024

    2024

  16. [24]

    Codedevents: optimal point-spread-function engineering for 3d-tracking with event cameras

    Sachin Shah, Matthew A Chan, Haoming Cai, Jingxi Chen, Sakshum Kulshrestha, Chahat Deep Singh, Yiannis Aloimonos, and Christopher A Metzler. Codedevents: optimal point-spread-function engineering for 3d-tracking with event cameras. In Proceedings of the IEEE/CVF conference on ...

    2024

  17. [25]

    Ced: Color event camera dataset

    Cedric Scheerlinck, Henri Rebecq, Timo Stoffregen, Nick Barnes, Robert Mahony, and Davide Scaramuzza. Ced: Color event camera dataset. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, pages 0–0, 2019

    2019

  18. [26]

    LED: A Large-scale Real-world Paired Dataset for Event Camera Denoising

    Yuxing Duan. LED: A Large-scale Real-world Paired Dataset for Event Camera Denoising. In 2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 25637–25647, Seattle, W A, USA, June

    2024

  19. [27]

    The Multi Vehicle Stereo Event Camera Dataset: An Event Camera Dataset for 3D Perception

    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, July 2018

  20. [28]

    Ev-imo: Motion segmentation dataset and learning pipeline for event cameras

    Anton Mitrokhin, Chengxi Ye, Cornelia Fermüller, Yiannis Aloimonos, and Tobi Delbruck. Ev-imo: Motion segmentation dataset and learning pipeline for event cameras. In 2019 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pages 6105–6112. IEEE, 2019

    2019

  21. [29]

    CIFAR10-DVS: An Event-Stream Dataset for Object Classification

    Hongmin Li, Hanchao Liu, Xiangyang Ji, Guoqi Li, and Luping Shi. CIFAR10-DVS: An Event-Stream Dataset for Object Classification. Frontiers in Neuroscience, 11:309, May 2017

    2017

  22. [30]

    Super slomo: High quality estimation of multiple intermediate frames for video interpolation

    Huaizu Jiang, Deqing Sun, Varun Jampani, Ming-Hsuan Yang, Erik Learned-Miller, and Jan Kautz. Super slomo: High quality estimation of multiple intermediate frames for video interpolation. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 9...

    2018

  23. [31]

    Time structure of the activity in neural network models

    Wulfram Gerstner. Time structure of the activity in neural network models. Physical review E, 51(1):738, 1995

    1995

  24. [32]

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

    Wolfgang Maass. Networks of spiking neurons: the third generation of neural network models. Neural networks, 10(9):1659–1671, 1997. 13 EventTracer: Fast Path Tracing-based Event Stream Rendering

    1997

  25. [33]

    Deep learning in spiking neural networks

    Amirhossein Tavanaei, Masoud Ghodrati, Saeed Reza Kheradpisheh, Timothée Masquelier, and Anthony Maida. Deep learning in spiking neural networks. Neural networks, 111:47–63, 2019

    2019

  26. [34]

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

    Paul A Merolla, John V Arthur, Rodrigo Alvarez-Icaza, Andrew S Cassidy, Jun Sawada, Filipp Akopyan, Bryan L Jackson, Nabil Imam, Chen Guo, Yutaka Nakamura, et al. A million spiking-neuron integrated circuit with a scalable communication network and interface. Science, 345(6197...

    2014

  27. [35]

    Towards artificial general intelligence with hybrid tianjic chip architecture

    Jing Pei, Lei Deng, Sen Song, Mingguo Zhao, Youhui Zhang, Shuang Wu, Guanrui Wang, Zhe Zou, Zhenzhi Wu, Wei He, et al. Towards artificial general intelligence with hybrid tianjic chip architecture. Nature, 572(7767):106– 111, 2019

    2019

  28. [36]

    Cambridge University Press, 2014

    Wulfram Gerstner, Werner M Kistler, Richard Naud, and Liam Paninski.Neuronal dynamics: From single neurons to networks and models of cognition. Cambridge University Press, 2014

    2014

  29. [37]

    Spiking deep networks with lif neurons

    Eric Hunsberger and Chris Eliasmith. Spiking deep networks with lif neurons. arXiv preprint arXiv:1510.08829, 2015

    2015 arXiv

  30. [38]

    Optimal conversion of conventional artificial neural networks to spiking neural networks

    Shikuang Deng and Shi Gu. Optimal conversion of conventional artificial neural networks to spiking neural networks. arXiv preprint arXiv:2103.00476, 2021

    2021 arXiv

  31. [39]

    Spatio-temporal backpropagation for training high- performance spiking neural networks

    Yujie Wu, Lei Deng, Guoqi Li, Jun Zhu, and Luping Shi. Spatio-temporal backpropagation for training high- performance spiking neural networks. Frontiers in neuroscience, 12:331, 2018

    2018

  32. [40]

    Surrogate gradient learning in spiking neural networks: Bringing the power of gradient-based optimization to spiking neural networks

    Emre O Neftci, Hesham Mostafa, and Friedemann Zenke. Surrogate gradient learning in spiking neural networks: Bringing the power of gradient-based optimization to spiking neural networks. IEEE Signal Processing Magazine, 36(6):51–63, 2019

    2019

  33. [41]

    Opportunities for neuromorphic computing algorithms and applications

    Catherine D Schuman, Shruti R Kulkarni, Maryam Parsa, J Parker Mitchell, Prasanna Date, and Bill Kay. Opportunities for neuromorphic computing algorithms and applications. Nature Computational Science, 2(1):10– 19, 2022

    2022

  34. [42]

    Exploring neuromorphic computing based on spiking neural networks: Algorithms to hardware

    Nitin Rathi, Indranil Chakraborty, Adarsh Kosta, Abhronil Sengupta, Aayush Ankit, Priyadarshini Panda, and Kaushik Roy. Exploring neuromorphic computing based on spiking neural networks: Algorithms to hardware. ACM Computing Surveys, 55(12):1–49, 2023

    2023

  35. [43]

    Hybrid snn-ann: Energy-efficient classification and object detection for event-based vision

    Alexander Kugele, Thomas Pfeil, Michael Pfeiffer, and Elisabetta Chicca. Hybrid snn-ann: Energy-efficient classification and object detection for event-based vision. In DAGM German Conference on Pattern Recognition, pages 297–312. Springer, 2021

    2021

  36. [44]

    A hybrid snn-ann network for event-based object detection with spatial and temporal attention

    Soikat Hasan Ahmed, Jan Finkbeiner, and Emre Neftci. A hybrid snn-ann network for event-based object detection with spatial and temporal attention. arXiv preprint arXiv:2403.10173, 2024

    2024 arXiv

  37. [45]

    Event-based video reconstruction via potential-assisted spiking neural network

    Lin Zhu, Xiao Wang, Yi Chang, Jianing Li, Tiejun Huang, and Yonghong Tian. Event-based video reconstruction via potential-assisted spiking neural network. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 3594–3604, 2022

    2022

  38. [46]

    Temporal- wise attention spiking neural networks for event streams classification

    Man Yao, Huanhuan Gao, Guangshe Zhao, Dingheng Wang, Yihan Lin, Zhaoxu Yang, and Guoqi Li. Temporal- wise attention spiking neural networks for event streams classification. In Proceedings of the IEEE/CVF interna- tional conference on computer vision, pages 10221–10230, 2021

    2021

  39. [47]

    Event-based action recognition using motion information and spiking neural networks

    Qianhui Liu, Dong Xing, Huajin Tang, De Ma, and Gang Pan. Event-based action recognition using motion information and spiking neural networks. In IJCAI, pages 1743–1749, 2021

    2021

  40. [48]

    Self-supervised learning of event-based optical flow with spiking neural networks

    Jesse Hagenaars, Federico Paredes-Vallés, and Guido De Croon. Self-supervised learning of event-based optical flow with spiking neural networks. Advances in Neural Information Processing Systems, 34:7167–7179, 2021

    2021

  41. [49]

    Fine-tuning surrogate gradient learning for optimal hardware performance in spiking neural networks

    Ilkin Aliyev and Tosiron Adegbija. Fine-tuning surrogate gradient learning for optimal hardware performance in spiking neural networks. In 2024 Design, Automation & Test in Europe Conference & Exhibition (DATE), pages 1–2. IEEE, 2024

    2024

  42. [50]

    Optimal transport: old and new, volume 338

    Cédric Villani et al. Optimal transport: old and new, volume 338. Springer, 2008

    2008

  43. [51]

    Amazon lumberyard bistro, open research content archive (orca), July 2017

    Amazon Lumberyard. Amazon lumberyard bistro, open research content archive (orca), July 2017

    2017

  44. [52]

    Spikingjelly: An open-source machine learning infrastructure platform for spike-based intelligence

    Wei Fang, Yanqi Chen, Jianhao Ding, Zhaofei Yu, Timothée Masquelier, Ding Chen, Liwei Huang, Huihui Zhou, Guoqi Li, and Yonghong Tian. Spikingjelly: An open-source machine learning infrastructure platform for spike-based intelligence. Science Advances, 9(40):eadi1480, 2023

    2023

  45. [53]

    Rendering resources, 2016

    Benedikt Bitterli. Rendering resources, 2016. https://benedikt-bitterli.me/resources/

    2016

  46. [54]

    The multivehicle stereo event camera dataset: An event camera dataset for 3d perception

    Alex Zihao Zhu, Dinesh Thakur, Tolga Özaslan, Bernd Pfrommer, Vijay Kumar, and Kostas Daniilidis. The multivehicle stereo event camera dataset: An event camera dataset for 3d perception. IEEE Robotics and Automation Letters, 3(3):2032–2039, 2018. 14 EventTracer: Fast Path Trac...

  47. [55]

    Fast image reconstruction with an event camera

    Cedric Scheerlinck, Henri Rebecq, Daniel Gehrig, Nick Barnes, Robert Mahony, and Davide Scaramuzza. Fast image reconstruction with an event camera. In Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision, pages 156–163, 2020

    2020

  48. [56]

    The unreasonable effectiveness of deep features as a perceptual metric

    Richard Zhang, Phillip Isola, Alexei A Efros, Eli Shechtman, and Oliver Wang. The unreasonable effectiveness of deep features as a perceptual metric. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 586–595, 2018

    2018

  49. [57]

    Focusing techniques

    Murali Subbarao, Tae-Sun Choi, and Arman Nikzad. Focusing techniques. Optical Engineering, 32(11):2824– 2836, 1993

    1993

  50. [58]

    Curriculum learning for ab initio deep learned refractive optics

    Xinge Yang, Qiang Fu, and Wolfgang Heidrich. Curriculum learning for ab initio deep learned refractive optics. Nature communications, 15(1):6572, 2024

    2024

  51. [59]

    Learned multi-aperture color-coded optics for snapshot hyperspectral imaging

    Zheng Shi, Xiong Dun, Haoyu Wei, Siyu Dong, Zhanshan Wang, Xinbin Cheng, Felix Heide, and Yifan Peng. Learned multi-aperture color-coded optics for snapshot hyperspectral imaging. ACM Transactions on Graphics (TOG), 43(6):1–11, 2024

    2024

  52. [60]

    The event-camera dataset and simulator: Event-based data for pose estimation, visual odometry, and slam

    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. 15

    2017

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.