Pith. sign in

REVIEW 4 major objections 6 minor 28 references

EDmamba: Rethinking Efficient Event Denoising with Spatiotemporal Decoupled SSMs

T0 review · 4 major / 6 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read Spatial and temporal event noise can be suppressed independently, and a decoupled state-space network does it faster and more accurately than joint 4D models.

desk verdict Decoupled spatial/temporal SSM design is genuinely new and the efficiency numbers are striking, but the SOTA claim rests on baseline numbers that were partly transcribed, not re-run. read the letter →

arxiv 2505.05391 v3 pith:SYK4UH6U submitted 2025-05-08 cs.CV

Ciyu Ruan , Zihang Gong , Ruishan Guo , Jingao Xu , Xinlei Chen This is my paper · ORCID
classification cs.CV
keywords eventcameradenoisingstatespacemodelsMambaspatiotemporaldecoupledarchitecturebackgroundactivitynoisehotpixelstreamprocessingreal-timeefficiency
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 tries to establish that event-camera noise separates into spatially structured artifacts (like hot pixels and leakage) and temporally inconsistent firings (background activity), so denoising can be done by two independent lightweight state-space branches instead of one joint 4D model. If true, the practical payoff is large: the network is small enough (88.98K parameters, 2.27 GFLOPs) and fast enough (100K events in 68.5ms) to run in real time, while matching or beating prior Transformer and CNN denoisers in accuracy. The paper reports state-of-the-art AUC on four public benchmarks, outscoring the strongest prior model by about 2.1 percentage points. A sympathetic reader takes this as evidence that architectural specialization to noise physics can beat simply scaling a joint spatiotemporal backbone.

What carries the argument

The central object is EDmamba, a U-Net-style encoder-decoder whose core is a pair of decoupled Mamba branches. The Spatial-SSM flattens the spatial domain with space-filling curves (Z-order and Hilbert) to preserve local geometry while processing sequences with bidirectional state updates, so fixed-pattern noise is suppressed through learned spatial smoothing. The Temporal-SSM sorts events by normalized timestamp and runs a bidirectional Mamba block to capture forward and backward motion patterns, exploiting the causal consistency of real motion versus the acausal randomness of temporal noise. A coarse feature extraction module separates each event's representation into geometry and polarity streams before the branches, and the two branches interact through shared spatial-temporal state-space blocks rather than entangled attention.

What would settle it

Re-run the Table 1 comparison from scratch: train every learning-based baseline and EDmamba on the same ED24 split, evaluate each on DND21 and DVSCLEAN with multiple seeds, and report the spread; if the best baseline's mean AUC overlaps EDmamba's within one standard deviation, the claimed accuracy lead is not established.

Watch

Extended reading notes

Core claim

The paper claims that spatial and temporal event noise originate from different physical mechanisms and can therefore be suppressed independently, and that acting on this separation improves both accuracy and efficiency. Its proposed network, EDmamba, processes raw events as a 4D event cloud $(x, y, t, p)$ using a coarse feature extractor that encodes geometry and polarity separately, then routes the features through two decoupled Mamba branches: a Spatial-SSM that learns location-conditioned filters to silence persistent spatial artifacts, and a Temporal-SSM that models causal motion continuity to eliminate bursty background events. The paper reports that this decoupled design reaches higher AUC than prior models on DND21 and most DVSCLEAN sequences, while using far fewer parameters and running 36 times faster than a Transformer-based denoiser.

Load-bearing premise

The state-of-the-art claim rests on the assumption that every comparison number in Table 1 came from identically run, fair evaluations, and the paper gives no code, no variance estimates, and no statement about whether non-learned baselines were re-run, so a reader cannot yet check whether the reported 1-3% margins exceed run-to-run noise.

Editorial extensions

If this is right

  • If the central claim is correct, event denoising no longer needs heavy spatiotemporal attention: a small decoupled state-space network can reach or beat prior accuracy while processing 100K events in 68ms, making denoising practical for real-time UAV and autonomous-driving pipelines.
  • The decoupling principle appears to generalize across noise levels, since the reported gains hold for shot noise rates of 1-10 Hz/pixel on DND21 and for both 50% and 100% noise injection on DVSCLEAN.
  • Ablations removing either branch degrade accuracy, with the Temporal-SSM contributing more in motion-heavy scenes, so interpreting temporal modeling as motion-continuity enforcement is a direct corollary of the paper's experiments.
  • The reported MESR gains on unlabeled real-world datasets imply the denoising behavior transfers beyond synthetic training data to daylight and nighttime conditions without requiring labels.
  • The Joint-SSM control variant, which shares one Mamba block for both spatial and temporal sequences, needs more parameters and runs slower while achieving lower accuracy, supporting the paper's claim that noise-specific specialization is the source of the gain.

Reading between the lines

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

  • The same decoupling insight likely transfers to other event-stream tasks where spatial artifacts and temporal jitter have distinct causes, such as event-based tracking, SLAM preprocessing, or high-speed obstacle avoidance, where removing background activity before downstream perception could improve robustness.
  • Because the spatial and temporal branches rely on linear state-space scans rather than quadratic attention, a testable extension is deploying the model on low-power neuromorphic or embedded hardware to check whether the 36x speed advantage over Transformers persists at the edge.
  • A natural experimental follow-up would vary the relative strength of spatial versus temporal noise in a controlled synthetic dataset and measure whether the two branches degrade independently; the paper's current ablations only remove whole branches, not manipulate noise types separately.
  • The paper's logic suggests that adding a third specialized branch for polarity-flip noise, rather than folding polarity into the coarse features, could yield further gains, though this goes beyond what the authors tested.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 6 minor

Summary. The manuscript proposes EDmamba, an event-denoising network that represents raw events as 4D point clouds, extracts coarse geometric and polarity features, and then routes them through two decoupled state-space branches: a Spatial-SSM for location-dependent artifacts such as hot pixels, and a Temporal-SSM for temporally inconsistent background activity. These branches are assembled in a U-Net-style encoder-decoder with shared Spatial-Temporal State Space Blocks. The paper reports 88.98K parameters, 2.27 GFLOPs, and 0.0685 s per 100K events on a single GPU, and claims state-of-the-art AUC on DND21 and DVSCLEAN plus strong MESR on E-MLB and DND21. Ablations on DND21 isolate the contributions of the geometric features, polarity features, and the two SSM branches, including a Joint-SSM control variant.

Significance. The core design idea is attractive and practically relevant: separating spatial and temporal denoising is a natural response to the different physical origins of hot-pixel artifacts and background-activity noise, and the reported parameter and latency figures are unusually low. The paper also includes a self-designed Joint-SSM control, which is a useful methodological step, and it evaluates on both labeled and unlabeled real-world benchmarks. If the comparisons against prior methods are made fully fair, reproducible, and statistically grounded, the result would be a meaningful advance for real-time event denoising. At present, however, the headline state-of-the-art claim is not yet supported because the baseline tables mix re-run and transcribed results, several reported differences are small, and no variance information is provided.

major comments (4)
  1. [Section 4.2, Table 2] The claim that EDmamba establishes new state-of-the-art accuracy on four public benchmarks is not supported by Table 2: on E-MLB Daylight ND1, GET reaches 1.051 versus 0.976 for EDmamba; on Daylight ND16, TS reaches 0.980 versus 0.985; and on DND21 MESR, EventZoom reaches 1.059 versus 1.057. The authors should either qualify the state-of-the-art claim to specific benchmarks and metrics, or explain how these non-best values are compatible with the abstract and Section 1 statements.
  2. [Section 4.1, Section 4.2, Tables 1-2, Figure 1] The Figure 1 caption states that all methods were evaluated under identical settings, but Table 2 footnotes say that IETS, GET, and EventZoom results are derived from E-MLB [25] and Table 3 says Pre-Mamba efficiency is derived from Pre-Mamba [12]. The paper must state explicitly which baselines were re-run by the authors and which were transcribed from prior papers. For re-run baselines, the authors should report the training split, noise-injection protocol, event-sampling scheme, evaluation script, and hardware configuration. Without this information, the 1.15-3.01% relative AUC improvements over EDformer in Table 1 and the small MESR margins in Table 2 could plausibly be caused by protocol differences rather than by the proposed method.
  3. [Section 4.3, Table 3] The efficiency comparison is not fully coherent. The 'Rel. Speed' column uses TS as the 1.0x reference for filtering methods but EDnCNN as the 1.0x reference for learning-based methods, so the relative speed values cannot be compared across the two groups. In addition, Pre-Mamba's inference time is 'derived' rather than measured, and no details are given on whether the baseline runtimes were obtained with the same operator-level implementation, batch size, and memory configuration. The 36x speedup claim over EDformer should be supported with measured runtimes under a common benchmarking protocol, or explicitly labeled as an estimate.
  4. [Section 4.4, Table 4] Ablations are reported for a single noise level (5 Hz/pixel) and without error bars or multiple seeds. Given that several differences are only 0.22-0.36 AUC points (e.g., w/o Geometry Feat. on Hotel-bar), the robustness of the ablation conclusions is uncertain. The same lack of variance information applies to the headline comparisons in Tables 1 and 2, where the reported margins are often smaller than typical run-to-run variation for event-denoising benchmarks.
minor comments (6)
  1. [Table 1] Table 1 contains a formatting error: several DVSCLEAN entries are concatenated (e.g., '0.78270.93940.76790.93010.7615'), making the table difficult to read.
  2. [Section 4, Implementation Details] The learning rate is described as '8 x 10^-5 per sample'; please clarify whether this is a per-sample or per-batch learning rate and how it interacts with the batch size of 128.
  3. [Section 3.2, Event Sampling] The text says one event is randomly sampled from each non-empty voxel during training; please state what sampling is used at inference time, since random sampling would introduce run-to-run variation in the reported metrics.
  4. [Figure 1] Figure 1 has a legend with 'Not mentioned' appearing twice, which leaves the marker encoding of parameter counts unclear; a cleaner legend would improve readability.
  5. [Abstract and Section 1] The abstract uses '2.1 percentage points' while Section 1 says '2.08% improvement to denoising accuracy'; these should be reconciled and the relative versus absolute distinction made explicit.
  6. [General] The paper does not state whether code or trained models will be released; providing them would substantially improve reproducibility, especially for baseline numbers that are currently derived from the literature.

Circularity Check

1 steps flagged · score 2.0 of 10

No load-bearing circularity; central accuracy claim is externally benchmarked, with only a peripheral self-cited efficiency baseline.

  1. other [Table 3 footnote and Figure 1 caption; Section 4.3]
    "All methods were evaluated under identical settings. ... †: The result is derived from Pre-Mamba [12]."

    The efficiency comparison uses a speed number transcribed from the authors' own prior work, so the stated 'identical settings' guarantee does not actually hold for that baseline row. This is a minor self-citation, not a construction-level circularity: it is not used to define the loss, metric, architecture, or any fitted parameter, and the accuracy state-of-the-art claim is measured on external public benchmarks against published methods. The step is flagged as a transparency caveat, not as a reduction of the central claim to its own input.

full rationale

EDmamba's central accuracy claim is evaluated on external public benchmarks (DND21, DVSCLEAN, E-MLB) using AUC and MESR against prior published methods; no equation in the method is fitted to produce a prediction, and the cross-entropy objective is independent of the evaluation metric. The decoupled S-SSM/T-SSM design is an architectural hypothesis tested by ablations, not an output forced by its own definition. The only same-author citation is Pre-Mamba [12], used as an efficiency baseline whose result is derived from the authors' prior paper rather than re-run; this introduces a comparability caveat for the relative-speed claim but does not support the accuracy claim. Table 2 also transcribes some MESR baseline values from E-MLB [25] because official code is unavailable, which is a reproducibility/protocol concern rather than circularity. Overall, the paper is self-contained against external benchmarks, and no load-bearing step reduces by construction to a fitted input or a self-citation chain.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

The paper's central quantitative results rest on hand-set hyperparameters, on a physical noise decomposition that is motivational rather than derived, and on transfer from the ED24 training set to public test sets. No new physical entities are postulated.

free parameters (4)
  • Temporal voxel resolution v = 0.1
    Grid resolution of normalized timestamp z used for voxelization; set by hand in Implementation Details, Section 4. It controls sampling and the sequences fed to the two SSM branches.
  • Event segment sample size N = 10240
    Fixed number of events per segment used in training; affects both accuracy and the reported throughput (Section 4).
  • U-Net depth and channel schedule = encoder [2,4], decoder [2], channels 8 to 16
    Architecture topology chosen by hand; it fixes the 88.98K parameter count and 2.27 GFLOPs (Section 4, Fig. 2).
  • Optimizer schedule = AdamW lr 8e-5, weight decay 5e-2, 50 epochs, batch 128
    Standard hyperparameters, no sensitivity study reported (Section 4).
assumptions (5)
  • domain assumption Spatial and temporal noise are physically decoupled and can be suppressed independently (Eqs. 1-3).
    Motivates the two-branch design; the paper does not measure the actual correlation of noise components in real sensors (Section 3.1).
  • domain assumption Space-filling curve scans preserve enough spatial locality for an SSM to learn location-conditioned filters.
    Core to S-SSM; no analysis of distortion for sparse event clouds (Section 3.3).
  • standard math Mamba blocks provide the claimed linear-time sequence modeling on event sequences.
    Inherited from Gu and Dao [11]; treated as a black box (Section 3.3-3.4).
  • domain assumption Training on ED24 transfers to DND21, DVSCLEAN, and E-MLB without significant domain shift.
    The test sets are used for reporting only; no domain-gap analysis is given (Section 4).
  • domain assumption AUC on DND21/DVSCLEAN and MESR on E-MLB are valid measures of denoising quality.
    Metrics are taken from prior work; assumes ground-truth labels are noise-free (Section 4).

how reviews work

0 comments
Cite this review

Pith. "Pith review of EDmamba: Rethinking Efficient Event Denoising with Spatiotemporal Decoupled SSMs." pith.science (2026). https://pith.science/paper/SYK4UH6U

@misc{pith2026250505391,
  author       = {Pith},
  title        = {Pith review of: EDmamba: Rethinking Efficient Event Denoising with Spatiotemporal Decoupled SSMs},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/SYK4UH6U}},
  note         = {Machine review of arXiv:2505.05391}
}
read the original abstract

Event cameras provide micro-second latency and broad dynamic range, yet their raw streams are marred by spatial artifacts (e.g., hot pixels) and temporally inconsistent background activity. Existing methods jointly process the entire 4D event volume (x, y, p, t), forcing heavy spatio-temporal attention that inflates parameters, FLOPs, and latency. We introduce EDmamba, a compact event-denoising framework that embraces the key insight that spatial and temporal noise arise from different physical mechanisms and can therefore be suppressed independently. A polarity- and geometry-aware encoder first extracts coarse cues, which are then routed to two lightweight state-space branches: a Spatial-SSM that learns location-conditioned filters to silence persistent artifacts, and a Temporal-SSM that models causal signal dynamics to eliminate bursty background events. This decoupled design distills the network to only 88.9K parameters and 2.27GFLOPs, enabling real-time throughput of 100K events in 68ms on a single GPU, 36x faster than recent Transformer baselines. Despite its economy, EDmamba establishes new state-of-the-art accuracy on four public benchmarks, outscoring the strongest prior model by 2.1 percentage points.

Figures

Figures reproduced from arXiv: 2505.05391 by the authors.

Figure 1
Figure 1. Performance vs. efficiency on event denoising with [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Overview of the EDmamba architecture. Raw events are grouped into 4D event clouds capturing spatial, temporal, and [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. (a) Visualization of signal events and spatiotempo [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Visual comparison on the E-MLB [25] dataset under daytime (top two rows) and nighttime (bottom two rows) conditions. EDmamba effectively suppresses background noise while preserving fine motion and structural details. In contrast, baseline methods either leave residual…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

28 extracted references · 25 canonical work pages

  1. [16]

    Shasha Guo and Tobi Delbruck. 2022. Low cost and latency event camera back- ground activity denoising.IEEE Transactions on Pattern Analysis and Machine Intelligence45, 1 (2022), 785–795

    work page 2022

  2. [25]

    Saizhe Ding, Jinze Chen, Yang Wang, Yu Kang, Weiguo Song, Jie Cheng, and Yang Cao. 2023. E-MLB: Multilevel benchmark for event-based camera denoising. IEEE Transactions on Multimedia26 (2023), 65–76

    work page 2023

  3. [12]

    Ciyu Ruan, Ruishan Guo, Zihang Gong, Jingao Xu, Wenhan Yang, and Xinlei Chen. 2025. PRE-Mamba: A 4D State Space Model for Ultra-High-Frequent Event Camera Deraining.arXiv preprint arXiv:2505.05307(2025)

    work page Pith review arXiv 2025

  4. [1]

    Rui Jiang, Xiaozheng Mou, Shunshun Shi, Yueyin Zhou, Qinyi Wang, Meng Dong, and Shoushun Chen. 2020. Object tracking on event cameras with offline–online learning.CAAI Transactions on Intelligence Technology5, 3 (2020), 165–171

    work page 2020

  5. [2]

    Daniel Gehrig and Davide Scaramuzza. 2024. Low-latency automotive vision with event cameras.Nature629, 8014 (2024), 1034–1040

    work page 2024

  6. [3]

    Davide Falanga, Kevin Kleber, and Davide Scaramuzza. 2020. Dynamic obstacle avoidance for quadrotors with event cameras.Science Robotics5, 40 (2020), eaaz9712

    work page 2020

  7. [4]

    Guillermo Gallego, Tobi Delbrück, Garrick Orchard, Chiara Bartolozzi, Brian Taba, Andrea Censi, Stefan Leutenegger, Andrew J Davison, Jörg Conradt, Kostas Daniilidis, et al. 2020. Event-based vision: A survey.IEEE transactions on pattern analysis and machine intelligence44, 1 (2020), 154–180

    2020

  8. [5]

    Yang Feng, Hengyi Lv, Hailong Liu, Yisa Zhang, Yuyao Xiao, and Chengshan Han. 2020. Event density based denoising method for dynamic vision sensor. Ruan and Gong, et al. Applied Sciences10, 6 (2020), 2024

    work page 2020

Show all 28 references
  1. [6]

    R Wes Baldwin, Mohammed Almatrafi, Jason R Kaufman, Vijayan Asari, and Keigo Hirakawa. 2019. Inceptive event time-surfaces for object classification using neuromorphic cameras. InImage Analysis and Recognition: 16th International Conference, ICIAR 2019, Waterloo, ON, Canada, A...

    2019

  2. [7]

    Yanxiang Wang, Bowen Du, Yiran Shen, Kai Wu, Guangrong Zhao, Jianguo Sun, and Hongkai Wen. 2019. EV-gait: Event-based robust gait recognition using dynamic vision sensors. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition. 6358–6367

    2019

  3. [8]

    R Baldwin, Mohammed Almatrafi, Vijayan Asari, and Keigo Hirakawa. 2020. Event probability mask (epm) and event denoising convolutional neural network (edncnn) for neuromorphic cameras. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 1701–1710

    2020

  4. [9]

    Huachen Fang, Jinjian Wu, Leida Li, Junhui Hou, Weisheng Dong, and Guang- ming Shi. 2022. AEDNet: Asynchronous event denoising with Spatial-Temporal correlation among irregular data. InProceedings of the 30th ACM International Conference on Multimedia. 1427–1435

    2022

  5. [10]

    Bin Jiang, Bo Xiong, Bohan Qu, M Salman Asif, You Zhou, and Zhan Ma. 2024. EDformer: Transformer-Based Event Denoising Across Varied Noise Levels. In European Conference on Computer Vision. Springer, 200–216

    2024

  6. [11]

    Albert Gu and Tri Dao. 2023. Mamba: Linear-time sequence modeling with selective state spaces.arXiv preprint arXiv:2312.00752(2023)

    2023 arXiv

  7. [13]

    Tobi Delbruck et al. 2008. Frame-free dynamic digital vision. InProceedings of Intl. Symp. on Secure-Life Electronics, Advanced Electronics for Quality Life and Society, Vol. 1. Citeseer, 21–26

    2008

  8. [14]

    Hongjie Liu, Christian Brandli, Chenghan Li, Shih-Chii Liu, and Tobi Delbruck

  9. [15]

    2018.𝑂(𝑁) O (N)-space spatiotemporal filter for reducing noise in neuromorphic vision sensors.IEEE Transactions on Emerging Topics in Computing9, 1 (2018), 15–23

    Alireza Khodamoradi and Ryan Kastner. 2018.𝑂(𝑁) O (N)-space spatiotemporal filter for reducing noise in neuromorphic vision sensors.IEEE Transactions on Emerging Topics in Computing9, 1 (2018), 15–23

    2018

  10. [17]

    Sio-Hoi Ieng, Christoph Posch, and Ryad Benosman. 2014. Asynchronous neuro- morphic event-driven image filtering.Proc. IEEE102, 10 (2014), 1485–1499

    2014

  11. [18]

    Daniel Czech and Garrick Orchard. 2016. Evaluating noise filtering for event- based asynchronous change detection image sensors. In2016 6th IEEE Interna- tional Conference on Biomedical Robotics and Biomechatronics (BioRob). IEEE, 19–24

    2016

  12. [19]

    Jinjian Wu, Chuanwei Ma, Xiaojie Yu, and Guangming Shi. 2020. Denoising of event-based sensors with spatial-temporal correlation. InICASSP 2020-2020 IEEE international conference on acoustics, speech and signal processing (ICASSP). IEEE, 4437–4441

    2020

  13. [20]

    Peiqi Duan, Zihao W Wang, Boxin Shi, Oliver Cossairt, Tiejun Huang, and Agge- los K Katsaggelos. 2021. Guided event filtering: Synergy between intensity images and neuromorphic events for high performance imaging.IEEE Transactions on Pattern Analysis and Machine Intelligence44...

    2021

  14. [21]

    Xavier Lagorce, Garrick Orchard, Francesco Galluppi, Bertram E Shi, and Ryad B Benosman. 2016. Hots: a hierarchy of event-based time-surfaces for pattern recognition.IEEE transactions on pattern analysis and machine intelligence39, 7 (2016), 1346–1359

    2016

  15. [22]

    Xuemei Xie, Jiang Du, Guangming Shi, Jianxiu Yang, Wan Liu, and Wang Li. 2018. DVS image noise removal using K-SVD method. InNinth International Conference on Graphic and Image Processing (ICGIP 2017), Vol. 10615. SPIE, 1099–1107

    2018

  16. [23]

    Peiqi Duan, Zihao W Wang, Xinyu Zhou, Yi Ma, and Boxin Shi. 2021. EventZoom: Learning to denoise and super resolve neuromorphic events. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition. 12824–12833

    2021

  17. [24]

    Yusra Alkendi, Rana Azzam, Abdulla Ayyad, Sajid Javed, Lakmal Seneviratne, and Yahya Zweiri. 2022. Neuromorphic camera denoising using graph neural network-driven transformers.IEEE Transactions on Neural Networks and Learning Systems35, 3 (2022), 4110–4124

    2022

  18. [26]

    Olaf Ronneberger, Philipp Fischer, and Thomas Brox. 2015. U-net: Convolutional networks for biomedical image segmentation. InInternational Conference on Medical image computing and computer-assisted intervention. Springer, 234–241

    2015

  19. [27]

    Yang Feng, Hengyi Lv, Hailong Liu, Yisa Zhang, Yuyao Xiao, and Chengshan Han. 2020. Event Density Based Denoising Method for Dynamic Vision Sensor. Applied Sciences(2020)

    2020

  20. [2015]

    In2015 IEEE International Symposium on Circuits and Systems (ISCAS)

    Design of a spatiotemporal correlation filter for event-based sensors. In2015 IEEE International Symposium on Circuits and Systems (ISCAS). IEEE, 722–725

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.