Pith. sign in

REVIEW 3 major objections 5 minor 1 cited by

E-4DGS claims the first event-only dynamic 3D Gaussian framework, reconstructing view-consistent 4D scenes from multi-view event streams without any RGB input.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

E-4DGS is a deformable 3D Gaussian Splatting method that reconstructs dynamic scenes directly from multi-view event camera streams, outperforming event-to-image baseline approaches.

T0 review reviewed 2026-08-05 challenge →

load-bearing objection Solid benchmark and event-only dynamic GS pipeline, but all quantitative results assume known poses and real-data evidence is qualitative, so the headline claim overshoots. the 3 major comments →

arxiv 2508.09912 v2 pith:2PPMNCQB submitted 2025-08-13 cs.CV

E-4DGS: High-Fidelity Dynamic Reconstruction from the Multi-view Event Cameras

Chaoran Feng , Zhenyu Tang , Wangbo Yu , Yatian Pang , Yian Zhao , Jianbin Zhao , Li Yuan , Yonghong Tian This is my paper
classification cs.CV
keywords Event-driven 4D reconstruction3D Gaussian SplattingNovel view synthesisEvent camerasDynamic scene reconstructionMulti-view event streamsHigh-speed egomotion
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

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 tries to establish that dynamic scenes can be reconstructed directly from multi-view event streams, bypassing the standard step of converting events into intensity frames. It proposes E-4DGS, a deformable 3D Gaussian representation supervised purely by event accumulation slices, supported by event-aware initialization, adaptive slicing of event windows, and an intensity-based pruning step. On a new synthetic benchmark with six moving event cameras per scene, the event-only version reaches 25.38 dB average PSNR versus 19.88 dB for E2VID+D3DGS and 21.66 dB for Deblur4DGS. The authors argue that operating on raw events preserves temporal sharpness and multi-view consistency better than two-stage event-to-video pipelines, and that adding a few blurry RGB frames pushes quality higher still.

Core claim

E-4DGS adapts Deformable3DGS so that the Gaussian field is optimized against event streams rather than rendered frames. Canonical Gaussians are deformed by a tiny MLP conditioned on time, and the supervision signal is the difference in predicted log-intensity between two timestamps, compared with the observed event count map through a learnable contrast threshold. To make this work, the method initializes Gaussians without SfM (a fixed cube or a radial sphere), randomly varies the length of the event accumulation window and injects noise into event-free pixels to avoid event neutralization, and prunes Gaussians whose alpha-weighted importance is below a threshold to suppress floaters. On eig

What carries the argument

The central object is a deformable 3D Gaussian field (canonical Gaussians plus a time-conditioned deformation MLP), supervised by event accumulation slices rather than intensity frames. The key identity is the event-count model: the predicted number of events between two timestamps is proportional to the change in rendered log-intensity divided by a learnable contrast threshold C-hat, and this is compared with the observed event count via a Huber-like loss. Supporting mechanisms are event-adaptive slicing (random window lengths plus noise injection), intensity importance pruning (removing Gaussians with low transmittance-weighted alpha), and adaptive contrast-threshold optimization.

Load-bearing premise

The method assumes that the exact camera positions and orientations at every moment are already known (interpolated linearly and via spherical linear interpolation), so if the moving cameras' trajectories are even slightly wrong, the entire multi-view consistency argument and the reported reconstruction quality rest on a pose oracle the method itself never estimates.

What would settle it

Take the eight synthetic scenes and perturb the provided camera trajectories with realistic odometry noise (for example, a few centimeters of position error and a few degrees of rotation error), then rerun E-4DGS and the baselines; if the event-only PSNR drops below that of the two-stage baseline, the claimed advantage depends on oracle poses rather than on the event-driven approach itself.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • Event-only dynamic reconstruction can reach competitive quality without RGB frames, SfM, or dense depth, at least in simulated multi-view settings.
  • Direct event supervision beats two-stage event-to-video conversion, implying that intermediate intensity reconstruction loses temporal precision that raw events retain.
  • Multi-view event coverage resolves some of the object-versus-camera motion ambiguity, yielding fewer floating artifacts and better 3D consistency than monocular event pipelines.
  • The optional fusion of a small number of blurry RGB frames further improves fidelity, suggesting a practical hybrid mode for real rigs.
  • The released synthetic benchmark, with six moving cameras and varying motion difficulty, provides a standardized testbed for future event-based 4D reconstruction methods.

Where Pith is reading between the lines

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

  • If the result transfers to real multi-view event setups, event-only capture could enable high-speed, low-light 3D reconstruction without the synchronization and exposure constraints of RGB cameras.
  • A natural next step the paper does not explore is jointly estimating camera trajectories while optimizing the Gaussian field; the method currently assumes accurate extrinsics, so a pose-free variant would broaden its applicability.
  • The learnable contrast threshold is a single scalar; real event sensors exhibit spatial and temporal threshold variation, so replacing it with a spatially varying or per-pixel threshold could be a testable improvement.
  • If the benchmarks were generated with a more realistic event simulator that includes sensor noise, the margin over E2VID+D3DGS might change; an extension would be to evaluate on a real multi-view event rig.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The paper presents E-4DGS, an end-to-end method for dynamic novel-view synthesis from multi-view event streams, built on a deformable 3D Gaussian Splatting representation. The main contributions are an event-driven radial-sphere Gaussian initialization, an event-adaptive slicing strategy for event-window supervision, an intensity-importance pruning scheme to remove floaters, and a learnable event contrast threshold. Optional blurry RGB frames can be fused through a weighted event/RGB loss. The method is evaluated on a new synthetic benchmark of eight Blender scenes rendered from six moving viewpoints at 3000 fps and processed with the v2e simulator, plus qualitative experiments on DSEC driving sequences. The reported event-only averages are PSNR 25.38 / SSIM 0.896 / LPIPS 0.134, compared with 19.88 / 0.728 / 0.397 for E2VID+D3DGS and 21.66 / 0.797 / 0.291 for Deblur4DGS; event-RGB fusion reaches 25.62 / 0.903 / 0.129. The paper claims to be the first event-only framework enabling view-consistent 3D Gaussian reconstruction in dynamic scenes.

Significance. If the results hold, the paper provides a useful step toward exploiting event cameras for high-speed dynamic reconstruction, and the planned release of the synthetic multi-view event benchmark is valuable to the community. The paper has clear strengths: the per-scene breakdown and component ablations are reported, the authors openly state where the method underperforms or where the data are insufficient, and the proposed components are simple and plausible. However, the significance is conditional. All quantitative evidence is synthetic, with ground-truth camera trajectories that the method does not estimate, and the event-only evaluation uses an inference-time color correction fitted to ground-truth log-intensity. These issues directly affect the strength of the headline claims and need to be addressed before the paper can be accepted as published.

major comments (3)
  1. [§5.1.1, §5.2.2]
  2. [§4.4, Eqs. (11)–(12) and Eq. (14)]
  3. [§5.2.1, Table 1, Appendix A.1]
minor comments (5)
  1. [§5.3, Table 2]
  2. [Figure 2 and §4]
  3. [§4.4, Eq. (10)]
  4. [References, Related Work]
  5. [Appendix A.1, Data Limitations]

Circularity Check

1 steps flagged

Quantitative NVS result is partly self-referential: final renderings are color-corrected by an OLS fit to ground-truth log-intensity at inference, so the reported numbers are not a pure event-only prediction of absolute appearance.

specific steps
  1. fitted input called prediction [Section 4.4, Eq. (14); Section 5.1.1 Implementation Details]
    "there is a scale ambiguity in the reconstructed color balance and illumination of the scene, when only the event contrast threshold is known. Thus, it's necessary to correct and align the color value for every color channel ... using the correction formula as follow: L̂(u_k,t_k) ≜ g_c·L̂(u_k,t_k)+Δ_c, (14) where, g_c and Δ_c are the color correction parameters, and derived via ordinary least squares [37] with the ground-truth log-intensity L(u_k,t_k) ... color correction is applied only at inference time in all scenes."

    The images evaluated in Tables 1, 3, and 5 are not raw model outputs but the outputs of Eq. (14), with g_c and Δ_c fit by ordinary least squares to the ground-truth log-intensity image. If this fit is performed on the same target frames used for PSNR/SSIM/LPIPS, then the evaluated `prediction' has its per-channel gain and offset constructed from the target itself, so the absolute intensity/color component of the reported metrics is partly fitted rather than predicted. The spatial structure still comes from the event-driven reconstruction, so this is a partial, not total, circularity.

full rationale

The core event-supervision pipeline is not circular: the loss in Eqs. (9)-(12) compares rendered intensity differences to observed event counts, and the learnable threshold C-hat is a single scalar calibration parameter that cannot fabricate the spatial event pattern; it is a gauge parameter, not a target-derived prediction. The paper's self-citations (AE-NeRF, GS2E, EvaGaussians, NeuralGS) appear only in related-work or background contexts and are not load-bearing. The most significant weakness is external validity, not circularity: the quantitative benchmark is synthetic with known camera trajectories, and the authors explicitly state that the real DSEC data 'does not support accurate quantitative NVS evaluation' (Sec. 5.2.2). That is an unvalidated pose/domain assumption, not a self-referential derivation. The one self-referential element is the inference-time OLS color correction, which uses ground-truth log-intensity to align the rendered color and thus partially contaminates the absolute-intensity component of the reported quantitative results. Hence a moderate score of 4.

Axiom & Free-Parameter Ledger

5 free parameters · 3 axioms · 0 invented entities

The paper introduces no new physical entities or forces; the ledger's burden is entirely carried by hand-set hyperparameters, a synthetic event simulator used as ground truth, and an assumed pose oracle. The largest unstated cost is the pose oracle: the entire quantitative evaluation assumes perfect camera trajectories, while the real data where trajectories come from VIO is only used qualitatively.

free parameters (5)
  • Initialization sphere radius and point count = r = 10.0, 200,000 points
    Chosen by hand for the synthetic center-focus scenes, not derived from the data or from an optimization; the paper reports it is a trade-off against E2VID+COLMAP.
  • Event count range per window [N_min, N_max] = [5e3, 1e4] object scenes, [1e5, 1e6] indoor scenes
    The adaptive slicing ranges are set per scene category by hand and affect supervision frequency and detail, with no sensitivity analysis.
  • Learnable contrast threshold C_hat = Initialized 0.15 synthetic, 0.2 real; jointly optimized
    A free parameter whose initialization is based on expected event camera settings and co-optimized with the model, absorbing the factual contrast threshold discrepancy.
  • Gaussian noise sigma (event jitter) = 0.02
    Hand-chosen noise added to event-void pixels during accumulation.
  • Importance pruning threshold = 0.015
    Fixed pruning threshold adopted from LightGaussians, not analyzed in the ablation.
axioms (3)
  • domain assumption The v2e event simulator generates sufficient realistic events from Blender renderings at 3000 FPS.
    The entire synthetic benchmark assumes v2e is a faithful event camera model; the paper cites the simulator's limitations for extreme lighting, which are acknowledged in Appendix A.1.
  • domain assumption Accurate camera intrinsics and high-frequency extrinsics are available, with linear/SLERP interpolation.
    Stated as a Training Assumption in Section 5.1.1; no camera pose estimation is performed, so all reported results depend on this oracle.
  • domain assumption Using D3DGS as the dynamic representation is adequate for arbitrary multi-view dynamic scenes.
    The paper replaces D3DGS supervision but inherits its assumptions about deformable scenes, such as no topology change and limited deformation.

reviewed 2026-08-05 · how reviews work

0 comments
Cite this review

Pith. "Pith review of E-4DGS: High-Fidelity Dynamic Reconstruction from the Multi-view Event Cameras." pith.science (2026). https://pith.science/paper/2PPMNCQB

@misc{pith2026250809912,
  author       = {Pith},
  title        = {Pith review of: E-4DGS: High-Fidelity Dynamic Reconstruction from the Multi-view Event Cameras},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2PPMNCQB}},
  note         = {Machine review of arXiv:2508.09912}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Novel view synthesis and 4D reconstruction techniques predominantly rely on RGB cameras, thereby inheriting inherent limitations such as the dependence on adequate lighting, susceptibility to motion blur, and a limited dynamic range. Event cameras, offering advantages of low power, high temporal resolution and high dynamic range, have brought a new perspective to addressing the scene reconstruction challenges in high-speed motion and

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Does it matter which Gaussians you pick in 4D Gaussian streaming?

    cs.CV 2026-03 conditional novelty 4.0

    A reinforcement-learned plug-in sampler can match or beat IGS@8192 quality on N3DV and MeetingRoom using as few as 256 anchors while reducing per-frame time.

Reference graph

Works this paper leans on

101 extracted references · 54 canonical work pages · cited by 1 Pith paper · 14 internal anchors

  1. [1]

    Anish Bhattacharya, Ratnesh Madaan, Fernando Cladera, Sai Vemprala, Roge- rio Bonatti, Kostas Daniilidis, Ashish Kapoor, Vijay Kumar, Nikolai Matni, and Jayesh K Gupta. 2024. Evdnerf: Reconstructing event data with dynamic neural radiance fields. In Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision. 5846–5855

    2024

  2. [2]

    Marco Cannici and Davide Scaramuzza. 2024. Mitigating Motion Blur in Neural Radiance Fields with Events and Frames. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

    2024

  3. [3]

    Ang Cao and Justin Johnson. 2023. HexPlane: A Fast Representation for Dynamic Scenes. In Computer Vision and Pattern Recognition (CVPR) . https://caoang327. github.io/HexPlane/

    2023

  4. [4]

    Kang Chen, Jiyuan Zhang, Zecheng Hao, Yajing Zheng, Tiejun Huang, and Zhaofei Yu. 2024. USP-Gaussian: Unifying Spike-based Image Reconstruction, Pose Correction and Gaussian Splatting. arXiv preprint arXiv:2411.10504 (2024)

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  5. [5]

    Runnan Chen, Youquan Liu, Lingdong Kong, Xinge Zhu, Yuexin Ma, Yikang Li, Yuenan Hou, Yu Qiao, and Wenping Wang. 2023. Clip2scene: Towards label- efficient 3d scene understanding by clip. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition . 7020–7030

    2023

  6. [6]

    Blender Online Community. 2018. Blender - a 3D modelling and rendering package. Blender Foundation, Stichting Blender Foundation, Amsterdam. http://www. blender.org

    2018

  7. [7]

    Matt Deitke, Ruoshi Liu, Matthew Wallingford, Huong Ngo, Oscar Michel, Aditya Kusupati, Alan Fan, Christian Laforte, Vikram Voleti, Samir Yitzhak Gadre, et al

  8. [8]

    Yuanxing Duan, Fangyin Wei, Qiyu Dai, Yuhang He, Wenzheng Chen, and Bao- quan Chen. 2024. 4d-rotor gaussian splatting: towards efficient novel view synthesis for dynamic scenes. In ACM SIGGRAPH 2024 Conference Papers . 1–11

    2024

  9. [9]

    Burak Ercan, Onur Eker, Canberk Saglam, Aykut Erdem, and Erkut Erdem. 2024. Hypere2vid: Improving event-based video reconstruction via hypernetworks. IEEE Transactions on Image Processing (2024)

    2024

  10. [10]

    Zhiwen Fan, Wenyan Cong, Kairun Wen, Kevin Wang, Jian Zhang, Xinghao Ding, Danfei Xu, Boris Ivanovic, Marco Pavone, Georgios Pavlakos, Zhangyang Wang, and Yue Wang. 2024. InstantSplat: Unbounded Sparse-view Pose-free Gaussian Splatting in 40 Seconds. arXiv:2403.20309 [cs.CV]

    Pith/arXiv arXiv 2024

  11. [11]

    Zhiwen Fan, Kevin Wang, Kairun Wen, Zehao Zhu, Dejia Xu, and Zhangyang Wang. 2023. Lightgaussian: Unbounded 3d gaussian compression with 15x reduction and 200+ fps. arXiv preprint arXiv:2311.17245 (2023)

    Pith/arXiv arXiv 2023

  12. [12]

    Chaoran Feng, Wangbo Yu, Xinhua Cheng, Zhenyu Tang, Junwu Zhang, Li Yuan, and Yonghong Tian. 2025. AE-NeRF: Augmenting Event-Based Neural Radiance Fields for Non-ideal Conditions and Larger Scene.arXiv preprint arXiv:2501.02807 (2025)

    Pith/arXiv arXiv 2025

  13. [13]

    Guillermo Gallego, Tobi Delbruck, Garrick Orchard, Chiara Bartolozzi, Brian Taba, Andrea Censi, Stefan Leutenegger, Andrew J Davison, Jorg Conradt, Kostas Daniilidis, et al. 2020. Event-based vision: A survey. IEEE Trans. Pattern Analysis and Machine Intelligence (PAMI) (2020)

    2020

  14. [14]

    Guillermo Gallego, Henri Rebecq, and Davide Scaramuzza. 2018. A unifying contrast maximization framework for event cameras, with applications to mo- tion, depth, and optical flow estimation. In Proceedings of the IEEE conference on computer vision and pattern recognition . 3867–3876

    2018

  15. [15]

    Mathias Gehrig, Willem Aarents, Daniel Gehrig, and Davide Scaramuzza. 2021. DSEC: A Stereo Event Camera Dataset for Driving Scenarios. IEEE Robotics and Automation Letters (2021). doi:10.1109/LRA.2021.3068942

    work page arXiv 2021

  16. [16]

    Gaurvi Goyal, Franco Di Pietro, Nicolo Carissimi, Arren Glover, and Chiara Bartolozzi. 2023. Moveenet: online high-frequency human pose estimation with an event camera. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 4024–4033

    2023

  17. [17]

    Shuang Guo and Guillermo Gallego. 2024. CMax-SLAM: Event-based rotational- motion bundle adjustment and SLAM system using contrast maximization. IEEE Transactions on Robotics (2024)

    2024

  18. [18]

    Yijia Guo, Liwen Hu, Yuanxi Bai, Jiawei Yao, Lei Ma, and Tiejun Huang. 2024. Spikegs: Reconstruct 3d scene via fast-moving bio-inspired sensors.arXiv preprint arXiv:2407.03771 (2024)

    Pith/arXiv arXiv 2024

  19. [19]

    Haiqian Han, Jianing Li, Henglu Wei, and Xiangyang Ji. 2024. Event-3DGS: Event-based 3D Reconstruction Using 3D Gaussian Splatting. Advances in Neural Information Processing Systems 37 (2024), 128139–128159

    2024

  20. [20]

    Haiqian Han, Jiacheng Lyu, Jianing Li, Henglu Wei, Cheng Li, Yajing Wei, Shu Chen, and Xiangyang Ji. 2024. Physical-Based Event Camera Simulator. In European Conference on Computer Vision . Springer, 19–35

    2024

  21. [21]

    Bing He, Yunuo Chen, Guo Lu, Qi Wang, Qunshan Gu, Rong Xie, Li Song, and Wenjun Zhang. 2024. S4D: Streaming 4D Real-World Reconstruction with Gaus- sians and 3D Control Points. arXiv:2408.13036 [cs.CV]

    work page arXiv 2024

  22. [22]

    Weihua He, Kaichao You, Zhendong Qiao, Xu Jia, Ziyang Zhang, Wenhui Wang, Huchuan Lu, Yaoyuan Wang, and Jianxing Liao. 2022. Timereplayer: Unlocking the potential of event cameras for video interpolation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition . 17804–17813

    2022

  23. [23]

    Javier Hidalgo-Carrió, Guillermo Gallego, and Davide Scaramuzza. 2022. Event- aided direct sparse odometry. In Proceedings of the IEEE/CVF Conference on Com- puter Vision and Pattern Recognition . 5781–5790

    2022

  24. [24]

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

    2021

  25. [25]

    Jian Huang, Chengrui Dong, and Peidong Liu. 2024. IncEventGS: Pose-Free Gaussian Splatting from a Single Event Camera. arXiv preprint arXiv:2410.08107 (2024)

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  26. [26]

    Inwoo Hwang, Junho Kim, and Young Min Kim. 2023. Ev-nerf: Event based neural radiance field. In Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision. 837–847

    2023

  27. [27]

    Bernhard Kerbl, Georgios Kopanas, Thomas Leimkuhler, and George Drettakis

  28. [28]

    Simon Klenk, Jason Chui, Nikolaus Demmel, and Daniel Cremers. 2021. TUM- VIE: The TUM stereo visual-inertial event dataset. In 2021 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) . IEEE, 8601–8608

    work page 2021

  29. [29]

    Simon Klenk, Lukas Koestler, Davide Scaramuzza, and Daniel Cremers. 2023. E-nerf: Neural radiance fields from a moving event camera. IEEE Robotics and Automation Letters 8, 3 (2023), 1587–1594

    work page 2023

  30. [30]

    Seungjun Lee and Gim Hee Lee. 2025. DiET-GS: Diffusion Prior and Event Stream- Assisted Motion Deblurring 3D Gaussian Splatting. arXiv:2503.24210 [cs.CV] https://arxiv.org/abs/2503.24210

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  31. [31]

    Hao Li, Curise Jia, Peng Jin, Zesen Cheng, Kehan Li, Jialu Sui, Chang Liu, and Li Yuan. 2023. Freestyleret: Retrieving images from style-diversified queries. arXiv preprint arXiv:2312.02428 (2023)

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  32. [32]

    Hao Li, Da Long, Li Yuan, Yu Wang, Yonghong Tian, Xinchang Wang, and Fanyang Mo. 2025. Decoupled peak property learning for efficient and inter- pretable electronic circular dichroism spectrum prediction.Nature Computational Science (2025), 1–11

    work page 2025

  33. [33]

    Yuchen Li*, Chaoran Feng*, Zhenyu Tang, Kaiyuan Deng, Wangbo Yu, Yonghong Tian, and Li Yuan. 2025. GS2E: Gaussian Splatting is an Effective Data Generator for Event Stream Generation. arXiv preprint arXiv:2505.15287 (2025)

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  34. [34]

    Jinwei Lin. 2024. Dynamic NeRF: A Review. arXiv preprint arXiv:2405.08609 (2024)

    Pith/arXiv arXiv 2024

  35. [35]

    Songnan Lin, Ye Ma, Zhenhua Guo, and Bihan Wen. 2022. Dvs-voltmeter: Sto- chastic process-based event simulator for dynamic vision sensors. In European Conference on Computer Vision . Springer, 578–593

    work page 2022

  36. [36]

    Youtian Lin, Zuozhuo Dai, Siyu Zhu, and Yao Yao. 2024. Gaussian-flow: 4d reconstruction with dynamic 3d gaussian particle. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition . 21136–21145

    work page 2024

  37. [37]

    Weng Fei Low and Gim Hee Lee. 2023. Robust e-nerf: Nerf from sparse & noisy events under non-uniform motion. In Proceedings of the IEEE/CVF International Conference on Computer Vision

    work page 2023

  38. [38]

    Jiahao Lu, Jiacheng Deng, Ruijie Zhu, Yanzhe Liang, Wenfei Yang, Xu Zhou, and Tianzhu Zhang. 2025. Dn-4dgs: Denoised deformable network with temporal- spatial aggregation for dynamic scene rendering. Advances in Neural Information Processing Systems 37 (2025), 84114–84138

    work page 2025

  39. [39]

    Qi Ma, Danda Pani Paudel, Ajad Chhatkuli, and Luc Van Gool. 2023. Deformable neural radiance fields using rgb and event cameras. InProceedings of the IEEE/CVF International Conference on Computer Vision . 3590–3600

    work page 2023

  40. [40]

    Branislav Micusik and Tomavs Pajdla. [n. d.]. Structure from motion with wide circular field of view cameras. ([n. d.])

    work page

  41. [41]

    Local Light Field Fusion: Practical View Synthesis with Prescriptive Sampling Guidelines

    Ben Mildenhall, Pratul P. Srinivasan, Rodrigo Ortiz-Cayon, Nima Khademi Kalantari, Ravi Ramamoorthi, Ren Ng, and Abhishek Kar. 2019. Local Light Field Fusion: Practical View Synthesis with Prescriptive Sampling Guidelines. arXiv:1905.00889 [cs.CV] https://arxiv.org/abs/1905.00889

    work page internal anchor Pith review Pith/arXiv arXiv 2019

  42. [42]

    Ben Mildenhall, Pratul P Srinivasan, Matthew Tancik, Jonathan T Barron, Ravi Ramamoorthi, and Ren Ng. 2020. NeRF: Representing Scenes as Neural Radiance Fields for View Synthesis. In European Conference on Computer Vision (ECCV) . https://www.matthewtancik.com/nerf

    work page 2020

  43. [43]

    Liyuan Pan, Cedric Scheerlinck, Xin Yu, Richard Hartley, Miaomiao Liu, and Yuchao Dai. 2019. Bringing a blurry frame alive at high frame-rate with an event camera. In Computer Vision and Pattern Recognition (CVPR)

    work page 2019

  44. [44]

    Yatian Pang, Peng Jin, Shuo Yang, Bin Lin, Bin Zhu, Zhenyu Tang, Liuhan Chen, Francis EH Tay, Ser-Nam Lim, Harry Yang, et al. 2024. Next patch prediction for autoregressive visual generation. arXiv preprint arXiv:2412.15321 (2024)

    Pith/arXiv arXiv 2024

  45. [45]

    Yatian Pang, Bin Zhu, Bin Lin, Mingzhe Zheng, Francis EH Tay, Ser-Nam Lim, Harry Yang, and Li Yuan. 2024. DreamDance: Animating Human Images by E-4DGS: High-Fidelity Dynamic Reconstruction from the Multi-view Event Cameras MM ’25, October 27–31, 2025, Dublin, Ireland. Enriching 3D Geometry Cues from 2D Poses. arXiv preprint arXiv:2412.00397 (2024)

    Pith/arXiv arXiv 2024

  46. [46]

    Keunhong Park, Utkarsh Sinha, Jonathan T Barron, Sofien Bouaziz, Dan B Gold- man, Steven M Seitz, and Ricardo Martin-Brualla. 2021. Nerfies: Deformable neural radiance fields. In Proceedings of the IEEE/CVF International Conference on Computer Vision. 5865–5874

    work page 2021

  47. [47]

    Keunhong Park, Utkarsh Sinha, Peter Hedman, Jonathan T Barron, Sofien Bouaziz, Dan B Goldman, Ricardo Martin-Brualla, and Steven M Seitz. 2021. Hypernerf: A higher-dimensional representation for topologically varying neural radiance fields. arXiv preprint arXiv:2106.13228 (2021)

    Pith/arXiv arXiv 2021

  48. [48]

    Shihan Peng, Hanyu Zhou, Hao Dong, Zhiwei Shi, Haoyue Liu, Yuxing Duan, Yi Chang, and Luxin Yan. 2024. CoSEC: A coaxial stereo event camera dataset for autonomous driving. arXiv preprint arXiv:2408.08500 (2024)

    Pith/arXiv arXiv 2024

  49. [49]

    Albert Pumarola, Enric Corona, Gerard Pons-Moll, and Francesc Moreno-Noguer

  50. [50]

    Yunshan Qi, Lin Zhu, Yu Zhang, and Jia Li. 2023. E2NeRF: Event Enhanced Neural Radiance Fields from Blurry Images. In International Conference on Computer Vision (ICCV)

    work page 2023

  51. [51]

    Maxime Raafat and contributors. 2024. BlenderNeRF: Easy NeRF synthetic dataset creation within Blender. https://github.com/maximeraafat/BlenderNeRF. Accessed: 2025-04-07

    work page 2024

  52. [52]

    Henri Rebecq, René Ranftl, Vladlen Koltun, and Davide Scaramuzza. 2019. High Speed and High Dynamic Range Video with an Event Camera.IEEE Trans. Pattern Anal. Mach. Intell. (T-PAMI) (2019)

    work page 2019

  53. [53]

    Viktor Rudnev, Mohamed Elgharib, Christian Theobalt, and Vladislav Golyanik

    work page

  54. [54]

    Viktor Rudnev, Gereon Fox, Mohamed Elgharib, Christian Theobalt, and Vladislav Golyanik. 2024. Dynamic EventNeRF: Reconstructing General Dynamic Scenes from Multi-view Event Cameras. arXiv preprint arXiv:2412.06770 (2024)

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  55. [55]

    Arman Savran, Raffaele Tavarone, Bertrand Higy, Leonardo Badino, and Chiara Bartolozzi. 2018. Energy and computation efficient audio-visual voice activity detection driven by event-cameras. In 2018 13th IEEE International Conference on Automatic Face & Gesture Recognition (FG 2018) . IEEE, 333–340

    work page 2018

  56. [56]

    Johannes L Schonberger and Jan-Michael Frahm. 2016. Structure-from-motion Revisited. In Computer Vision and Pattern Recognition (CVPR)

    work page 2016

  57. [57]

    In Computer Vision and Pattern Recognition (CVPR)

    EventNeRF: Neural Radiance Fields from a Single Colour Event Camera. In Computer Vision and Pattern Recognition (CVPR)

    work page

  58. [58]

    Zihang Shao, Xuanye Fang, Yaxin Li, Chaoran Feng, Jiangrong Shen, and Qi Xu

    work page

  59. [59]

    Noah Snavely, Steven M Seitz, and Richard Szeliski. 2006. Photo tourism: explor- ing photo collections in 3D. In ACM siggraph 2006 papers. 835–846

    work page 2006

  60. [60]

    Ganchao Tan, Yang Wang, Han Han, Yang Cao, Feng Wu, and Zheng-Jun Zha

    work page

  61. [61]

    Jiwei Shan, Zeyu Cai, Cheng-Tai Hsieh, Shing Shin Cheng, and Hesheng Wang

    work page

  62. [62]

    Zhenyu Tang, Chaoran Feng, Xinhua Cheng, Wangbo Yu, Junwu Zhang, Yuan Liu, Xiaoxiao Long, Wenping Wang, and Li Yuan. 2025. NeuralGS: Bridging Neural Fields and 3D Gaussian Splatting for Compact 3D Representations. arXiv preprint arXiv:2503.23162 (2025)

    Pith/arXiv arXiv 2025

  63. [63]

    Zhenyu Tang, Junwu Zhang, Xinhua Cheng, Wangbo Yu, Chaoran Feng, Ya- tian Pang, Bin Lin, and Li Yuan. 2024. Cycle3D: High-quality and Consistent Image-to-3D Generation via Generation-Reconstruction Cycle. arXiv preprint arXiv:2407.19548 (2024)

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  64. [64]

    Advances in Neural Information Processing Systems 36 (2023), 32117–32128

    EICIL: joint excitatory inhibitory cycle iteration learning for deep spiking neural networks. Advances in Neural Information Processing Systems 36 (2023), 32117–32128

    work page 2023

  65. [65]

    Diwen Wan, Yuxiang Wang, Ruijie Lu, and Gang Zeng. 2024. Template-free Articulated Gaussian Splatting for Real-time Reposable Dynamic View Synthesis. arXiv preprint arXiv:2412.05570 (2024)

    Pith/arXiv arXiv 2024

  66. [66]

    Haoyang Wang, Ruishan Guo, Pengtao Ma, Ciyu Ruan, Xinyu Luo, Wenhua Ding, Tianyang Zhong, Jingao Xu, Yunhao Liu, and Xinlei Chen. 2025. Towards Mobile Sensing with Event Cameras on High-mobility Resource-constrained Devices: A Survey. arXiv preprint arXiv:2503.22943 (2025)

    arXiv 2025

  67. [67]

    Jiaxu Wang, Junhao He, Ziyi Zhang, Mingyuan Sun, Jingkai Sun, and Renjing Xu. 2024. EvGGS: A Collaborative Learning Framework for Event-based Gener- alizable Gaussian Splatting. arXiv preprint arXiv:2405.14959 (2024)

    Pith/arXiv arXiv 2024

  68. [68]

    Wei Zhi Tang, Daniel Rebain, Kostantinos G Derpanis, and Kwang Moo Yi. 2024. LSE-NeRF: Learning Sensor Modeling Errors for Deblured Neural Radiance Fields with RGB-Event Stereo. arXiv preprint arXiv:2409.06104 (2024)

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  69. [69]

    Jiahao Wu, Rui Peng, Zhiyan Wang, Lu Xiao, Luyang Tang, Jinbo Yan, Kaiqiang Xiong, and Ronggang Wang. 2025. Swift4D: Adaptive divide-and-conquer Gauss- ian Splatting for compact and efficient reconstruction of dynamic scene. arXiv preprint arXiv:2503.12307 (2025)

    Pith/arXiv arXiv 2025

  70. [70]

    Jingqian Wu, Shuo Zhu, Chutian Wang, and Edmund Y Lam. 2024. Ev-GS: Event- based gaussian splatting for efficient and accurate radiance field rendering. In 2024 IEEE 34th International Workshop on Machine Learning for Signal Processing (MLSP). IEEE, 1–6

    work page 2024

  71. [71]

    Stepan Tulyakov, Daniel Gehrig, Stamatios Georgoulis, Julius Erbach, Mathias Gehrig, Yuanyou Li, and Davide Scaramuzza. 2021. Time lens: Event-based video frame interpolation. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 16155–16164

    work page 2021

  72. [72]

    Renlong Wu, Zhilu Zhang, Mingyang Chen, Xiaopeng Fan, Zifei Yan, and Wang- meng Zuo. 2024. Deblur4DGS: 4D Gaussian Splatting from Blurry Monocular Video. arXiv preprint arXiv:2412.06424 (2024)

    arXiv 2024

  73. [73]

    Ziyi Wu, Mathias Gehrig, Qing Lyu, Xudong Liu, and Igor Gilitschenski. 2024. Leod: Label-efficient object detection for event cameras. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition . 16933–16943

    work page 2024

  74. [74]

    Tianyi Xiong, Jiayi Wu, Botao He, Cornelia Fermuller, Yiannis Aloimonos, Heng Huang, and Christopher A Metzler. 2024. Event3DGS: Event-based 3D Gaussian Splatting for Fast Egomotion. arXiv preprint arXiv:2406.02972 (2024)

    Pith/arXiv arXiv 2024

  75. [75]

    Qianqian Wang, Vickie Ye, Hang Gao, Jake Austin, Zhengqi Li, and Angjoo Kanazawa. 2024. Shape of motion: 4d reconstruction from a single video. arXiv preprint arXiv:2407.13764 (2024)

    arXiv 2024

  76. [76]

    Wenhao Xu, Wenming Weng, Yueyi Zhang, Ruikang Xu, and Zhiwei Xiong. 2024. Event-boosted Deformable 3D Gaussians for Fast Dynamic Scene Reconstruction. arXiv preprint arXiv:2411.16180 (2024)

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  77. [77]

    Le Xue, Mingfei Gao, Chen Xing, Roberto Martín-Martín, Jiajun Wu, Caiming Xiong, Ran Xu, Juan Carlos Niebles, and Silvio Savarese. 2023. Ulip: Learning a unified representation of language, images, and point clouds for 3d understand- ing. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 1179–1189

    work page 2023

  78. [78]

    Jingqian Wu, Shuo Zhu, Chutian Wang, Boxin Shi, and Edmund Y Lam. 2024. SweepEvGS: Event-Based 3D Gaussian Splatting for Macro and Micro Radiance Field Rendering from a Single Sweep. arXiv preprint arXiv:2412.11579 (2024)

    work page arXiv 2024

  79. [79]

    Jinbo Yan, Rui Peng, Luyang Tang, and Ronggang Wang. 2024. 4D Gaussian Splatting with Scale-aware Residual Field and Adaptive Optimization for Real- time rendering of temporally complex dynamic scenes. In Proceedings of the 32nd ACM International Conference on Multimedia . 7871–7880

    work page 2024

  80. [80]

    Zhiwen Yan, Chen Li, and Gim Hee Lee. 2023. Nerf-ds: Neural radiance fields for dynamic specular objects. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition . 8285–8295

    work page 2023

Showing first 80 references.

This paper was first reviewed by deepseek-v4-flash on August 5, 2026.