pith. sign in

arxiv: 2604.12803 · v1 · submitted 2026-04-14 · 💻 cs.CV · cs.LG

Generative Anonymization in Event Streams

Adam T. M\"uller , Mihai Kocsis , Nicolaj C. Stache This is my paper

Pith reviewed 2026-05-10 16:28 UTC · model grok-4.3

classification 💻 cs.CV cs.LG
keywords generative anonymizationevent streamsneuromorphic visionprivacy preservationevent-to-video reconstructiondata utilityanonymization frameworksynchronized dataset
0
0 comments X

The pith

A pipeline converts event streams to intensity images, generates fake identities with pretrained models, and converts back to preserve task utility while blocking identity recovery.

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

Neuromorphic sensors produce sparse event streams that can be turned into recognizable video, creating privacy risks in public spaces. Standard ways to hide identities by masking or scrambling destroy the timing and structure needed for detection and tracking. The paper shows that routing the events through an intensity-image stage lets existing generative models replace real people with invented ones before the data returns to event form. Experiments indicate the resulting streams still support accurate perception models yet resist identity extraction. A new robotic-captured dataset of paired event and RGB sequences is released to support this evaluation.

Core claim

The authors claim that bridging the modality gap by first projecting asynchronous events into an intermediate intensity representation, then applying pretrained spatial generative models to synthesize non-existent identities, and finally re-encoding the result back into the neuromorphic domain produces anonymized event streams that reliably prevent identity recovery from event-to-video reconstructions while preserving the spatio-temporal structure required by downstream vision tasks.

What carries the argument

The three-stage pipeline that projects events into intensity images, leverages pretrained generative models for identity synthesis, and re-encodes the output into the event domain.

If this is right

  • Public deployment of neuromorphic cameras becomes feasible because event data can be released without exposing personal identities.
  • Downstream models for object detection, tracking, and segmentation continue to operate at near-original performance on the processed streams.
  • The introduced synchronized event-RGB dataset provides a repeatable benchmark for measuring both privacy protection and task utility in future work.
  • The same intermediate-representation strategy could be applied to other asynchronous sensor data where reconstruction risks privacy.

Where Pith is reading between the lines

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

  • Extending the method to video-rate generation might allow anonymization on live camera feeds rather than recorded streams.
  • If the intensity stage can be replaced by a lighter model, the pipeline could run on edge devices with limited compute.
  • The approach suggests that privacy protection need not be a separate post-processing step but can be embedded in the data-generation process itself.

Load-bearing premise

Projecting events to intensity and back after generative replacement keeps the exact timing and spatial layout that downstream perception models need, without adding artifacts that lower their accuracy.

What would settle it

A controlled test in which a standard object detector or tracker shows a large drop in accuracy on the anonymized streams compared with the original streams would indicate that structural integrity was not preserved.

Figures

Figures reproduced from arXiv: 2604.12803 by Adam T. M\"uller, Mihai Kocsis, Nicolaj C. Stache.

Figure 1
Figure 1. Figure 1: Architectural overview of the generative anonymization pipeline. The framework translates raw asynchronous event data into continuous grayscale frames to detect and swap faces using established generative models. The anonymized identity is subsequently projected back into the event space via a V2E conversion, preserving the underlying spatiotemporal structure. according to the retention probability: P(ei ∈… view at source ↗
Figure 2
Figure 2. Figure 2: Qualitative examples of source and synthetic identi￾ties. Comparison of three subjects (rows). Columns from left-to￾right: Source event streams, intermediate E2V representations, the anonymized generative output (Anon), V2E projection into a new event stream, and the final downstream E2V validation. row three). However, the model occasionally struggles to translate subtle visual micro-expressions, particul… view at source ↗
Figure 3
Figure 3. Figure 3: V2E discretization and density artifacts. Viewed tilted from the top-down position, where more recent events are closer to the frontal cross-section. The reverse projection step relies on stan￾dard V2E conversion, which leads to discretization in the event￾space. Notably only in the parts of the event stream where infor￾mation has been replaced (facial region). we do not yet directly alter the raw event st… view at source ↗
read the original abstract

Neuromorphic vision sensors offer low latency and high dynamic range, but their deployment in public spaces raises severe data protection concerns. Recent Event-to-Video (E2V) models can reconstruct high-fidelity intensity images from sparse event streams, inadvertently exposing human identities. Current obfuscation methods, such as masking or scrambling, corrupt the spatio-temporal structure, severely degrading data utility for downstream perception tasks. In this paper, to the best of our knowledge, we present the first generative anonymization framework for event streams to resolve this utility-privacy trade-off. By bridging the modality gap between asynchronous events and standard spatial generative models, our pipeline projects events into an intermediate intensity representation, leverages pretrained models to synthesize realistic, non-existent identities, and re-encodes the features back into the neuromorphic domain. Experiments demonstrate that our method reliably prevents identity recovery from E2V reconstructions while preserving the structural data integrity required for downstream vision tasks. Finally, to facilitate rigorous evaluation, we introduce a novel, synchronized real-world event and RGB dataset captured via precise robotic trajectories, providing a robust benchmark for future research in privacy-preserving neuromorphic vision.

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

2 major / 2 minor

Summary. The paper proposes the first generative anonymization framework for event streams captured by neuromorphic vision sensors. The pipeline converts asynchronous events to an intermediate intensity representation, applies pretrained spatial generative models to synthesize non-existent identities, and re-encodes the result back into the event domain. The central claim is that this approach prevents identity recovery from subsequent event-to-video reconstructions while preserving the spatio-temporal structure needed for downstream perception tasks. The work also introduces a new synchronized real-world event and RGB dataset captured along precise robotic trajectories to support evaluation.

Significance. If the round-trip preservation of event timing, polarity, and density holds, the method would offer a practical way to address privacy risks in public neuromorphic deployments without the severe utility degradation caused by masking or scrambling. The new dataset with robotic synchronization could become a useful benchmark for privacy-preserving event-based vision. The reliance on existing pretrained models lowers the barrier to adoption.

major comments (2)
  1. [§3] §3 (Pipeline): The re-encoding step after intensity-based generative synthesis is load-bearing for the utility-preservation claim. The manuscript must quantify how well microsecond timestamps, polarity, and event sparsity are retained; any temporal smoothing or density mismatch introduced by the generative model would break downstream neuromorphic algorithms even if the intensity images appear realistic.
  2. [§5] §5 (Experiments): The abstract states that experiments demonstrate reliable prevention of identity recovery and preservation of structural integrity, yet no quantitative metrics, baselines, or protocols for measuring identity leakage (e.g., face recognition accuracy on E2V outputs) are referenced. Without these details the central trade-off claim cannot be evaluated.
minor comments (2)
  1. [Abstract] The repeated phrase 'to the best of our knowledge' in the abstract and introduction is redundant; a single, well-supported novelty statement suffices.
  2. [§3] Notation for the event-to-intensity projection and the subsequent re-encoding operator should be defined once in a dedicated subsection rather than inline.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the thoughtful and constructive review. We address each major comment below and have revised the manuscript to strengthen the presentation of the pipeline and experimental evaluation.

read point-by-point responses

  1. Referee: [§3] §3 (Pipeline): The re-encoding step after intensity-based generative synthesis is load-bearing for the utility-preservation claim. The manuscript must quantify how well microsecond timestamps, polarity, and event sparsity are retained; any temporal smoothing or density mismatch introduced by the generative model would break downstream neuromorphic algorithms even if the intensity images appear realistic.

    Authors: We agree that explicit quantification of the re-encoding step is necessary to support the utility-preservation claim. The original manuscript described the re-encoding procedure at a high level but did not report numerical metrics on timestamp fidelity, polarity retention, or density preservation. In the revised version we have expanded §3 with a dedicated analysis subsection that measures (i) mean absolute timestamp deviation, (ii) polarity match rate, and (iii) event-density ratio between input and output streams. These metrics are computed on the new robotic dataset and show that the re-encoding step introduces only negligible temporal smoothing while preserving polarity and sparsity within acceptable bounds for downstream neuromorphic algorithms. revision: yes

  2. Referee: [§5] §5 (Experiments): The abstract states that experiments demonstrate reliable prevention of identity recovery and preservation of structural integrity, yet no quantitative metrics, baselines, or protocols for measuring identity leakage (e.g., face recognition accuracy on E2V outputs) are referenced. Without these details the central trade-off claim cannot be evaluated.

    Authors: We acknowledge that the experimental section lacked explicit quantitative protocols and numbers for identity leakage. The revised manuscript now includes a dedicated evaluation subsection in §5 that reports face-recognition accuracy (using two standard models) on E2V reconstructions from both original and anonymized event streams. We also describe the exact evaluation protocol, including the train/test split on the new dataset and the baseline comparison against masking and scrambling methods. The added results show a substantial drop in identity recovery while downstream task performance remains comparable to the original streams. revision: yes

Circularity Check

0 steps flagged

No significant circularity in derivation chain

full rationale

The paper describes a pipeline that converts event streams to intensity frames via existing E2V reconstruction, applies off-the-shelf pretrained generative models for identity synthesis, and re-encodes the output back to the event domain. No load-bearing equations, fitted parameters, or self-citations are shown that would make the claimed preservation of spatio-temporal structure (polarity, timestamps, density) equivalent to the input data by construction. The utility-privacy trade-off resolution is presented as an empirical outcome of the modular pipeline rather than a tautological redefinition or renamed known result. The framework therefore remains self-contained against external benchmarks and pretrained components.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

Abstract-only review; pipeline assumes standard pretrained generative models exist and that event-to-intensity projection is invertible enough for downstream tasks. No explicit free parameters, axioms, or invented entities are stated.

axioms (2)
  • domain assumption Pretrained spatial generative models can produce realistic non-identities when applied to event-derived intensity images.
    Central to the anonymization step; no justification or reference provided in abstract.
  • domain assumption Re-encoding anonymized intensity images back to events preserves the spatio-temporal statistics needed for perception.
    Required for the utility claim; not demonstrated in abstract.

pith-pipeline@v0.9.0 · 5498 in / 1307 out tokens · 54428 ms · 2026-05-10T16:28:21.951732+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

47 extracted references · 47 canonical work pages

  1. [1]

    E2PRIV: Privacy- Preserving Event-to-Video Reconstruction with Face Anonymization

    Mira Adra and Jean-Luc Dugelay. E2PRIV: Privacy- Preserving Event-to-Video Reconstruction with Face Anonymization. In2025 13th International Workshop on Biometrics and Forensics (IWBF), pages 1–6, 2025. 2

    work page 2025

  2. [2]

    Event-driven Re-Id: A New Bench- mark and Method Towards Privacy-Preserving Person Re- Identification

    Shafiq Ahmad, Gianluca Scarpellini, Pietro Morerio, and Alessio Del Bue. Event-driven Re-Id: A New Bench- mark and Method Towards Privacy-Preserving Person Re- Identification. In2022 IEEE/CVF Winter Conference on Ap- plications of Computer Vision Workshops (WACVW), pages 459–468, 2022. ISSN: 2690-621X. 1, 2

    work page 2022

  3. [3]

    Person Re-Identification without Identification via Event anonymization

    Shafiq Ahmad, Pietro Morerio, and Alessio Del Bue. Person Re-Identification without Identification via Event anonymization. InProceedings of the IEEE/CVF Interna- tional Conference on Computer Vision, pages 11132–11141,

    work page

  4. [4]

    Event Anonymization: Privacy-Preserving Person Re- Identification and Pose Estimation in Event-Based Vision

    Shafiq Ahmad, Pietro Morerio, and Alessio Del Bue. Event Anonymization: Privacy-Preserving Person Re- Identification and Pose Estimation in Event-Based Vision. IEEE Access, 12:66964–66980, 2024. 1, 2, 8

    work page 2024

  5. [5]

    H. G. Barrow, J. M. Tenenbaum, R. C. Bolles, and H. C. Wolf. Parametric Correspondence and Chamfer Matching: Two New Techniques for Image Matching. 1977. Number: TN153. 5

    work page 1977

  6. [6]

    Understanding Human Reactions Looking at Facial Mi- croexpressions With an Event Camera.IEEE Transactions on Industrial Informatics, 18(12):9112–9121, 2022

    Federico Becattini, Federico Palai, and Alberto Del Bimbo. Understanding Human Reactions Looking at Facial Mi- croexpressions With an Event Camera.IEEE Transactions on Industrial Informatics, 18(12):9112–9121, 2022. 1

    work page 2022

  7. [7]

    AnonyNoise: Anonymizing Event Data with Smart Noise to Outsmart Re-Identification and Preserve Privacy

    Katharina Bendig, René Schuster, Nicole Thiemer, Karen Joisten, and Didier Stricker. AnonyNoise: Anonymizing Event Data with Smart Noise to Outsmart Re-Identification and Preserve Privacy. InProceedings of the Winter Confer- ence on Applications of Computer Vision, pages 3159–3161,

    work page

  8. [8]

    Neuromorphic Event-Based Facial Expression Recognition

    Lorenzo Berlincioni, Luca Cultrera, Chiara Albisani, Lisa Cresti, Andrea Leonardo, Sara Picchioni, Federico Becattini, and Alberto Del Bimbo. Neuromorphic Event-Based Facial Expression Recognition. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 4109–4119, 2023. 2, 6

    work page 2023

  9. [9]

    Faces in Event Streams (FES): An Annotated Face Dataset for Event Cameras.Sen- sors, 24(5), 2024

    Ulzhan Bissarinova, Tomiris Rakhimzhanova, Daulet Ken- zhebalin, and Huseyin Atakan Varol. Faces in Event Streams (FES): An Annotated Face Dataset for Event Cameras.Sen- sors, 24(5), 2024. 1, 5, 7, 8

    work page 2024

  10. [10]

    Sliced and Radon Wasserstein Barycenters of Mea- sures.Journal of Mathematical Imaging and Vision, 51(1): 22–45, 2015

    Nicolas Bonneel, Julien Rabin, Gabriel Peyré, and Hanspeter Pfister. Sliced and Radon Wasserstein Barycenters of Mea- sures.Journal of Mathematical Imaging and Vision, 51(1): 22–45, 2015. 6

    work page 2015

  11. [11]

    How Far Are We From Solving the 2D & 3D Face Alignment Problem? (And a Dataset of 230,000 3D Facial Landmarks)

    Adrian Bulat and Georgios Tzimiropoulos. How Far Are We From Solving the 2D & 3D Face Alignment Problem? (And a Dataset of 230,000 3D Facial Landmarks). InProceedings of the IEEE International Conference on Computer Vision, pages 1021–1030, 2017. 5

    work page 2017

  12. [12]

    Burt and Edward H

    Peter J. Burt and Edward H. Adelson. A multiresolution spline with application to image mosaics.ACM Trans. Graph., 2(4):217–236, 1983. 3

    work page 1983

  13. [13]

    Event-Based Neuromorphic Vi- sion for Autonomous Driving: A Paradigm Shift for Bio- Inspired Visual Sensing and Perception.IEEE Signal Pro- cessing Magazine, 37(4):34–49, 2020

    Guang Chen, Hu Cao, Jorg Conradt, Huajin Tang, Florian Rohrbein, and Alois Knoll. Event-Based Neuromorphic Vi- sion for Autonomous Driving: A Paradigm Shift for Bio- Inspired Visual Sensing and Perception.IEEE Signal Pro- cessing Magazine, 37(4):34–49, 2020. 1

    work page 2020

  14. [14]

    SimSwap: An efficient framework for high fidelity face swapping

    Renwang Chen, Xuanhong Chen, Bingbing Ni, and Yanhao Ge. SimSwap: An efficient framework for high fidelity face swapping. InProceedings of the 28th ACM International Conference on Multimedia, pages 2003–2011. Association for Computing Machinery, 2020. 3, 6

    work page 2003

  15. [15]

    ArcFace: Additive Angular Mar- gin Loss for Deep Face Recognition.IEEE Transactions on Pattern Analysis and Machine Intelligence, 44(10):5962– 5979, 2022

    Jiankang Deng, Jia Guo, Jing Yang, Niannan Xue, Irene Kot- sia, and Stefanos Zafeiriou. ArcFace: Additive Angular Mar- gin Loss for Deep Face Recognition.IEEE Transactions on Pattern Analysis and Machine Intelligence, 44(10):5962– 5979, 2022. 3, 5

    work page 2022

  16. [16]

    Acceler- ating the Super-Resolution Convolutional Neural Network

    Chao Dong, Chen Change Loy, and Xiaoou Tang. Acceler- ating the Super-Resolution Convolutional Neural Network. InComputer Vision – ECCV 2016, pages 391–407, Cham,

    work page 2016

  17. [17]

    Springer International Publishing. 6

    work page

  18. [18]

    Event Encryption for Neuro- morphic Vision Sensors: Framework, Algorithm, and Evalu- ation.Sensors, 21(13), 2021

    Bowen Du, Weiqi Li, Zeju Wang, Manxin Xu, Tianchen Gao, Jiajie Li, and Hongkai Wen. Event Encryption for Neuro- morphic Vision Sensors: Framework, Algorithm, and Evalu- ation.Sensors, 21(13), 2021. 2

    work page 2021

  19. [19]

    Now You See Me, Now You Don’t: A Unified Framework for Ex- pression Consistent Anonymization in Talking Head Videos

    Anil Egin, Andrea Tangherloni, and Antitza Dantcheva. Now You See Me, Now You Don’t: A Unified Framework for Ex- pression Consistent Anonymization in Talking Head Videos. InProceedings of the IEEE/CVF International Conference on Computer Vision, pages 5925–5934, 2025. 5

    work page 2025

  20. [20]

    EVREAL: Towards a Comprehensive Benchmark and Anal- ysis Suite for Event-Based Video Reconstruction

    Burak Ercan, Onur Eker, Aykut Erdem, and Erkut Erdem. EVREAL: Towards a Comprehensive Benchmark and Anal- ysis Suite for Event-Based Video Reconstruction. InPro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 3943–3952, 2023. 1, 2, 6

    work page 2023

  21. [21]

    Regulation (eu) 2024/1689 of the european parliament and of the council of 13 june 2024 laying down harmonised rules on artificial intelligence

    European Union. Regulation (eu) 2024/1689 of the european parliament and of the council of 13 june 2024 laying down harmonised rules on artificial intelligence. Official Journal of the European Union, L 2024/1689, 2024. Accessed: 2026- 01-19. 1

    work page 2024

  22. [22]

    Haoqiang Fan, Hao Su, and Leonidas J. Guibas. A Point Set Generation Network for 3D Object Reconstruction From a Single Image. InProceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages 605–613,

    work page

  23. [23]

    Event-Based Video Reconstruction With Deep Spatial- Frequency Unfolding Network.IEEE Transactions on Image Processing, 34:1779–1794, 2025

    Chengjie Ge, Xueyang Fu, Kunyu Wang, and Zheng-Jun Zha. Event-Based Video Reconstruction With Deep Spatial- Frequency Unfolding Network.IEEE Transactions on Image Processing, 34:1779–1794, 2025. 2

    work page 2025

  24. [24]

    Dense Continuous-Time Optical Flow From Event Cameras

    Mathias Gehrig, Manasi Muglikar, and Davide Scaramuzza. Dense Continuous-Time Optical Flow From Event Cameras. IEEE Transactions on Pattern Analysis and Machine Intelli- gence, 46(7):4736–4746, 2024. 2

    work page 2024

  25. [25]

    Probabilistic Online Event Downsampling

    Andreu Girbau-Xalabarder, Jun Nagata, and Shinichi Sumiyoshi. Probabilistic Online Event Downsampling. In Proceedings of the IEEE/CVF Conference on Computer Vi- sion and Pattern Recognition, pages 4896–4904, 2025. 3

    work page 2025

  26. [26]

    v2e: From video frames to realistic DVS events

    Yuhuang Hu, Shih-Chii Liu, and Tobi Delbruck. v2e: From video frames to realistic DVS events. In2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW). IEEE, 2021. 1, 2, 3, 6

    work page 2021

  27. [27]

    DeepPrivacy: A Generative Adversarial Network for Face Anonymization

    Håkon Hukkelås, Rudolf Mester, and Frank Lindseth. DeepPrivacy: A Generative Adversarial Network for Face Anonymization. InAdvances in Visual Computing, pages 565–578, Cham, 2019. Springer International Publishing. 2

    work page 2019

  28. [28]

    Observational evaluation of event cameras performance in optical space surveillance

    Krzysztof Kami ´nski, Gregory Cohen, Tobi Delbruck, Michał ˙Zołnowski, and Marcin G˛ edek. Observational evaluation of event cameras performance in optical space surveillance. In 1st NEO and Debris Detection Conference, 2019. 1

    work page 2019

  29. [29]

    LDFA: Latent Diffusion Face Anonymiza- tion for Self-Driving Applications

    Marvin Klemp, Kevin Rösch, Royden Wagner, Jannik Quehl, and Martin Lauer. LDFA: Latent Diffusion Face Anonymiza- tion for Self-Driving Applications. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 3199–3205, 2023. 1, 2

    work page 2023

  30. [30]

    Grand Challenge of 106-Point Fa- cial Landmark Localization

    Yinglu Liu, Hao Shen, Yue Si, Xiaobo Wang, Xiangyu Zhu, Hailin Shi, Zhibin Hong, Hanqi Guo, Ziyuan Guo, Yanqin Chen, Bi Li, Teng Xi, Jun Yu, Haonian Xie, Guochen Xie, Mengyan Li, Qing Lu, Zengfu Wang, Shenqi Lai, Zhenhua Chai, and Xiaoming Wei. Grand Challenge of 106-Point Fa- cial Landmark Localization. In2019 IEEE International Conference on Multimedia ...

    work page 2019

  31. [31]

    CIA- GAN: Conditional Identity Anonymization Generative Ad- versarial Networks

    Maxim Maximov, Ismail Elezi, and Laura Leal-Taixe. CIA- GAN: Conditional Identity Anonymization Generative Ad- versarial Networks. InProceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition, pages 5447–5456, 2020. 2

    work page 2020

  32. [32]

    Qiang Qu, Yiran Shen, Xiaoming Chen, Yuk Ying Chung, and Tongliang Liu. E2HQV: High-Quality Video Generation from Event Camera via Theory-Inspired Model-Aided Deep Learning.Proceedings of the AAAI Conference on Artificial Intelligence, 38(5):4632–4640, 2024. 1, 2

    work page 2024

  33. [33]

    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. 2

    work page 2019

  34. [34]

    You Only Look Once: Unified, Real-Time Object Detection

    Joseph Redmon, Santosh Divvala, Ross Girshick, and Ali Farhadi. You Only Look Once: Unified, Real-Time Object Detection. InProceedings of the IEEE Conference on Com- puter Vision and Pattern Recognition, pages 779–788, 2016. 5, 7

    work page 2016

  35. [35]

    High-Resolution Image Synthesis With Latent Diffusion Models

    Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Björn Ommer. High-Resolution Image Synthesis With Latent Diffusion Models. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 10684–10695, 2022. 3

    work page 2022

  36. [36]

    Yossi Rubner, Carlo Tomasi, and Leonidas J. Guibas. The Earth Mover’s Distance as a Metric for Image Retrieval. International Journal of Computer Vision, 40(2):99–121,

    work page

  37. [37]

    Nataniel Ruiz, Eunji Chong, and James M. Rehg. Fine- Grained Head Pose Estimation Without Keypoints. InPro- ceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops, pages 2074–2083, 2018. 5

    work page 2074

  38. [38]

    Fast Im- age Reconstruction with an Event Camera

    Cedric Scheerlinck, Henri Rebecq, Daniel Gehrig, Nick Barnes, Robert Mahony, and Davide Scaramuzza. Fast Im- age Reconstruction with an Event Camera. InProceedings of the IEEE/CVF Winter Conference on Applications of Com- puter Vision, pages 156–163, 2020. 2

    work page 2020

  39. [39]

    EventNet: Asynchronous Recursive Event Processing

    Yusuke Sekikawa, Kosuke Hara, and Hideo Saito. EventNet: Asynchronous Recursive Event Processing. InProceedings of the IEEE/CVF Conference on Computer Vision and Pat- tern Recognition, pages 3887–3896, 2019. 5

    work page 2019

  40. [40]

    A Comprehensive Review of YOLO Architectures in Computer Vision: From YOLOv1 to YOLOv8 and YOLO-NAS.Machine Learning and Knowl- edge Extraction, 5(4):1680–1716, 2023

    Juan Terven, Diana-Margarita Córdova-Esparza, and Julio- Alejandro Romero-González. A Comprehensive Review of YOLO Architectures in Computer Vision: From YOLOv1 to YOLOv8 and YOLO-NAS.Machine Learning and Knowl- edge Extraction, 5(4):1680–1716, 2023. 5, 7, 8

    work page 2023

  41. [41]

    Event- Based Video Reconstruction Using Transformer

    Wenming Weng, Yueyi Zhang, and Zhiwei Xiong. Event- Based Video Reconstruction Using Transformer. InProceed- ings of the IEEE/CVF International Conference on Com- puter Vision, pages 2563–2572, 2021. 2

    work page 2021

  42. [42]

    G²Face: High-Fidelity Reversible Face Anonymization via Generative and Geometric Priors.IEEE Transactions on In- formation Forensics and Security, 19:8773–8785, 2024

    Haoxin Yang, Xuemiao Xu, Cheng Xu, Huaidong Zhang, Jing Qin, Yi Wang, Pheng-Ann Heng, and Shengfeng He. G²Face: High-Fidelity Reversible Face Anonymization via Generative and Geometric Priors.IEEE Transactions on In- formation Forensics and Security, 19:8773–8785, 2024. 3

    work page 2024

  43. [43]

    DreamID: High-Fidelity and Fast diffusion-based Face Swapping via Triplet ID Group Learning, 2025

    Fulong Ye, Miao Hua, Pengze Zhang, Xinghui Li, Qichao Sun, Songtao Zhao, Qian He, and Xinglong Wu. DreamID: High-Fidelity and Fast diffusion-based Face Swapping via Triplet ID Group Learning, 2025. arXiv:2504.14509 [cs]. 5, 8

    work page arXiv 2025

  44. [44]

    V2CE: Video to Continuous Events Simulator

    Zhongyang Zhang, Shuyang Cui, Kaidong Chai, Haowen Yu, Subhasis Dasgupta, Upal Mahbub, and Tauhidur Rah- man. V2CE: Video to Continuous Events Simulator. In2024 IEEE International Conference on Robotics and Automation (ICRA), pages 12455–12461, 2024. 8

    work page 2024

  45. [45]

    Deep Event-based Ob- ject Detection in Autonomous Driving: A Survey, 2024

    Bingquan Zhou and Jie Jiang. Deep Event-based Ob- ject Detection in Autonomous Driving: A Survey, 2024. arXiv:2405.03995 [cs]. 1

    work page arXiv 2024

  46. [46]

    EventHDR: From Event to High-Speed HDR Videos and Beyond.IEEE Transactions on Pattern Analysis and Machine Intelligence, 47(1):32–50, 2025

    Yunhao Zou, Ying Fu, Tsuyoshi Takatani, and Yinqiang Zheng. EventHDR: From Event to High-Speed HDR Videos and Beyond.IEEE Transactions on Pattern Analysis and Machine Intelligence, 47(1):32–50, 2025. 2

    work page 2025

  47. [47]

    Context-Aware Full Body Anonymization using Text-to-Image Diffusion Models, 2024

    Pascal Zwick, Kevin Roesch, Marvin Klemp, and Oliver Bringmann. Context-Aware Full Body Anonymization using Text-to-Image Diffusion Models, 2024. arXiv:2410.08551 [cs]. 1, 2

    work page arXiv 2024