Pith. sign in

REVIEW 4 major objections 5 minor 2 cited by

DeblurSplat claims to be the first SfM-free deblurring 3D Gaussian Splatting method, replacing COLMAP with DUSt3R point-cloud initialization and adding event-camera latent-image supervision to achieve state-of-the-art novel-view quality on

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 ↗

T0 review · deepseek-v4-flash

2026-08-04 15:39 UTC pith:GCVVYNGM

load-bearing objection DeblurSplat is a solid engineering combination of DUSt3R and event-based deblurring for 3D-GS with strong reported results, but the SOTA claim is undercut by a missing nearest-baseline comparison and a real-world evaluation that gives every method the proposed initialization. the 4 major comments →

arxiv 2509.18898 v2 pith:GCVVYNGM submitted 2025-09-23 cs.CV

DeblurSplat: SfM-free 3D Gaussian Splatting with Event Camera for Robust Deblurring

Pengteng Li , Yunfan Lu , Pinhao Song , Weiyu Guo , Huizai Yao , F. Richard Yu , Hui Xiong This is my paper
classification cs.CV
keywords 3D Gaussian Splattingmotion deblurringevent cameraSfM-free reconstructionDUSt3Rnovel view synthesisEDI decouplingbundle adjustment
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 paper proposes DeblurSplat, the first Structure-from-Motion-free framework for deblurring 3D Gaussian Splatting. It argues that motion-blurred images can be reconstructed reliably by replacing SfM initialization with a pretrained dense stereo module that predicts point clouds and confidence scores directly, and by using event streams to recover latent sharp frames that supervise optimization. The central claim is that this combination avoids cumulative pose errors from blur and yields state-of-the-art novel-view quality and rendering speed. A sympathetic reader would care because real-world captures under low light or long exposure routinely suffer motion blur, and existing SfM-based deblurring pipelines degrade or fail in those settings.

Core claim

DeblurSplat replaces the fragile SfM initialization stage of deblurring 3D-GS with DUSt3R, a pretrained dense stereo network that outputs, from blurred images directly, a set of 3D points and per-point confidence scores without estimating camera poses as an intermediate. The paper introduces Confidence Balanced Sampling to keep low-confidence but spatially informative points, and Progressive Alignment: each blurred image is decoupled via the Event-based Double Integral (EDI) into u latent sharp frames, while the camera trajectory during exposure is parameterized as a linear interpolation between two endpoint SE(3) corrections. The Gaussians render these latent frames; a grayscale alignment l

What carries the argument

The central mechanism is the pairing of DUSt3R's dense-stereo point clouds (with confidence scores) with event-decoupled latent sharp images. DUSt3R provides blur-robust initialization without pose estimation; Confidence Balanced Sampling converts its confidence map into Gaussian primitives; EDI decouples each blurred frame into u latent sharp images; Progressive Alignment optimizes two endpoint SE(3) pose corrections per view (Eq. 5) and renders the u latent frames, imposing a grayscale alignment loss (Eq. 6) against the EDI-decoded frames and a blur-formation loss (Eq. 7) averaging the renders against the blurred frame. The event stream is the fine-grained supervision that resolves the amb

Load-bearing premise

The load-bearing premise is that each blurred image is exactly the average of u latent sharp frames related by an exponential event response with a constant threshold (Theta = 0.27) and a camera trajectory that is a linear SE(3) interpolation between two endpoint corrections; if real exposure motion is non-uniform or the EDI model is inaccurate, the latent sharp images used for supervision are systematically biased.

What would settle it

Record a scene with a robotic arm executing a deliberately non-uniform trajectory during a long exposure while an event camera logs the true motion, and compare DeblurSplat's reconstructed latent frames against ground-truth high-speed sharp frames: if the recovered latent frames deviate exactly where the trajectory departs from linear SE(3) interpolation, the blur model, not the optimization, is the limiting factor.

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

If this is right

  • If the method works as claimed, motion-blurred captures can be reconstructed without running COLMAP, removing a common failure point and cutting initialization time by roughly 3 times.
  • The event-guided Progressive Alignment provides a template for using continuous event streams to correct latent poses, not just deblur pixels.
  • The confidence-balanced sampling principle suggests that discarding low-confidence points from learned stereo is suboptimal; balanced sampling preserves geometry.
  • The method claims real-time rendering (above 210 FPS) after training, so deblurring 3D reconstruction could be used interactively.
  • Consistent gains on increasingly blurry real scenes suggest the approach degrades gracefully as blur worsens.

Where Pith is reading between the lines

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

  • An editor's testable extension: the same DUSt3R-plus-event pipeline might extend to rolling-shutter distortion or other spatially varying degradations, since the event stream carries continuous motion information (the paper itself suggests rolling-shutter as future work).
  • The EDI decoupling assumes a fixed contrast threshold (Theta = 0.27); making Theta learnable per scene could improve robustness on real sensors where the threshold drifts, and it is a direct experiment to run.
  • If DUSt3R's confidence is unreliable on blurry inputs (the paper's own limitation), an alternative is to fuse multiple stereo hypotheses or to refine confidence during Gaussian training; this could be tested on synthetic scenes with known ground-truth geometry.
  • The grayscale alignment loss implicitly assumes latent images share edge structure with rendered ones; adding a color-aware or perceptual term might help on scenes with strong color fringes, though it could also introduce new failure modes.

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

4 major / 5 minor

Summary. DeblurSplat proposes a deblurring 3D Gaussian Splatting framework that replaces COLMAP initialization with DUSt3R dense-stereo point clouds and adds an event-based Progressive Alignment module. Blurred images are decoupled into latent sharp frames via EDI; rendered latent frames are aligned in grayscale with event-derived latent images, while a blurred reconstruction is formed by averaging latent renders. The camera trajectory is parameterized by SE(3) interpolation and optimized jointly with the Gaussians. Experiments on synthetic (normal/hard) and real-world datasets report consistent gains over B-NeRF, B-3DGS, UFP-GS, EFN-GS, E2GS, BAD-NeRF, BAD-GS, and EDNeRF.

Significance. If the empirical claims hold, the paper makes a useful contribution: dense-stereo initialization is a plausible remedy for COLMAP failure under blur, and event-based latent-image alignment gives a strong deblurring signal. The paper also introduces new challenging benchmarks (7 Blender hard scenes and 5 real scenes with multiple blur levels) and includes useful ablations of sampling strategy, interval count, event-loss weight, and pose accuracy. However, the headline SOTA claim is currently conditional on a missing closest competitor and on a real-world protocol in which all baselines are initialized by the proposed DUSt3R pipeline. The contribution is therefore promising but needs stronger empirical grounding before publication.

major comments (4)
  1. [Abstract and Section III-B / Appendix A-B] The claim that the method does not calculate camera poses is contradicted by the paper itself. Appendix A-B derives focal length, pairwise RANSAC+PnP poses, and globally aligned poses via Eq. (14); Eq. (5) parameterizes and optimizes latent poses; Table XII reports ATE. This is not SfM-free in the usual sense. Please revise the wording and clearly state that DUSt3R provides an initial pose estimate that is subsequently refined.
  2. [Section IV-C and Section II-A] EVA-Gaussians [12] is described as event-stream-assisted 3D Gaussian splatting from blurry images but is absent from the baseline list and all tables. Because it is the closest prior work, the SOTA claim in Tables I-II is not yet established. Add EVA-Gaussians under its native COLMAP initialization and under the DUSt3R-initialized protocol; report per-scene metrics. If it matches or exceeds DeblurSplat under either initialization, the SOTA claim is not supported.
  3. [Section IV-B/IV-D and Appendix D] The real-world protocol weakens the comparison. Appendix D states COLMAP fails and 'we initialize these scenes by our method,' so every baseline is initialized from the proposed DUSt3R point cloud (1k random subsample) without re-tuning. In the hard synthetic scenes, by contrast, baselines receive sharp-image COLMAP initialization. This asymmetry makes it hard to attribute the +3.36 dB real and +2.74 dB hard gains to the proposed modules rather than to the initialization handout. Provide COLMAP-initialized baselines where possible, or explicitly report all methods under a common initialization and add an initialization-source sensitivity study.
  4. [Section III-B, Eqs. (2)-(5)] The event-guided latent images are obtained by solving Eq. (4) from the same blurred input and events under EDI assumptions, so L_e is not supervision from independent sharp ground truth. Any error in the uniform-blur model, constant threshold Θ=0.27, bin count u, or the linear SE(3) interpolation in Eq. (5) is systematically reinforced by the alignment loss. Please add sensitivity analyses over Θ and u, and test with non-linear/interpolated ground-truth trajectories on synthetic scenes (where high-speed sharp frames are available) to quantify the bias.
minor comments (5)
  1. [Section IV-A] Notation is inconsistent: Section III-A uses M intervals and L points, while Section IV-A says 'm=40 and Q_m=125'; also 'internals' should be 'intervals'.
  2. [Appendix A] The exposure-time definition t_exp = t_start - t_end has the wrong sign; it should be t_end - t_start.
  3. [References] References [49] and [50] are duplicate entries for the same paper.
  4. [Figure 2] The caption mentions a proportion β that is never defined in the main text.
  5. [Section IV-D] EDNeRF is cited as [29] in Section IV-C but as [30] in Section IV-D; please resolve the reference numbering.

Circularity Check

0 steps flagged

No significant circularity: the derivation is self-contained; event-guided supervision is a model-based regularizer, not an identity, and the main limitations are experimental-design concerns rather than circular reasoning.

full rationale

The paper's central derivation is not circular. The event-guided supervision in Eq. 6 aligns rendered latent images to EDI-decoded latent images (Eq. 4). Although the EDI latent images are a deterministic re-encoding of the blurred input plus events, the rendered images come from independent 3DGS parameters and camera poses; the loss is a soft, grayscale regularizer (with λ_e = 5e-3) on top of the blur-reconstruction loss Eq. 7. The final evaluation is against held-out sharp ground-truth views (Tables I, II, V-VIII), not against the EDI-decoded images, so the target is external. No equation reduces to an identity: the rendered latent images are not algebraically forced to equal the EDI-decoded latent images, and the optimization could fail to match them. The DUSt3R initialization is an external pretrained model with stated assumptions, and the pose refinement is a standard bundle adjustment; no 'uniqueness' theorem or load-bearing self-citation is invoked. The two self-citations (Refs. [18] and [31] include co-authors) appear only in related-work lists and do not support the central claim. The paper's real limitations—the EDI motion-blur model (Eqs. 2-5) and the Appendix D admission that real-world baselines are initialized by the proposed method—are correctness/experimental-design concerns, not circularity. In particular, Appendix D states 'we initialize these scenes by our method and simulate the sparse condition by random sampling 1k points on point clouds,' which weakens the comparison but does not make the derivation equivalent to its inputs. Similarly, the omission of EVA-Gaussians [12] from the baseline table is an evaluation gap, not a circular step. The derivation chain from blur inputs, DUSt3R point clouds, and event streams to the final renders is not self-referential in the sense prohibited by the circularity review.

Axiom & Free-Parameter Ledger

6 free parameters · 5 axioms · 0 invented entities

The central pipeline rests on two borrowed modules (DUSt3R and EDI) plus a pose-trajectory model; the main free parameters are the sampling and loss weights, several of which are tuned on the same datasets used for evaluation. No new physical entities are introduced.

free parameters (6)
  • M (number of confidence intervals) = 40
    Tuned in Table XI; M=40 balances fine-grained semantic extraction vs. over-weighting low-confidence noise.
  • L (sampled point count) = 5,000
    Tuned in Table X; larger counts (10k to 100k) degrade optimization performance.
  • λe (event loss weight) = 5e-3 (default)
    Swept in Table IX; 5e-2 causes performance collapse and values above 5e-1 alter the trade-off.
  • λb (blur reconstruction loss weight) = 0.2
    Set in Implementation Details; standard 3D-GS reconstruction trade-off.
  • u (number of event bins) = 10
    Set 'following previous works' in Sec. IV-A; controls EDI discretization of the exposure interval.
  • Θ (event response threshold) = 0.27
    Adopted from prior works [8, 3, 41] as stated in Sec. IV-A; material to the EDI latent-image computation.
axioms (5)
  • domain assumption DUSt3R, pretrained on sharp images, produces reliable point clouds and confidence maps for motion-blurred inputs.
    Core motivation of the SfM-free claim; Appendix G admits the confidence accuracy under ambiguous conditions is unclear because DUSt3R was trained on sharp images.
  • domain assumption Motion blur is a time-weighted average of latent sharp frames and the event-camera response is an exponential with constant threshold Θ (the EDI model).
    Eqs. (2)-(4) in Sec. III-B inherit EDI [16]; if this assumption is violated, the latent supervision signal is biased.
  • domain assumption The camera trajectory during an exposure is a smooth linear interpolation in SE(3) between two endpoint corrections ΔT1 and ΔTu.
    Eq. (5) in Sec. III-B parameterizes the latent poses this way; non-uniform motion during an exposure would break the trajectory model.
  • domain assumption Confidence maps from DUSt3R can be partitioned into M intervals and sampled proportionally to confidence without losing essential spatial coverage.
    Sec. III-A sampling design; based on the empirical study in Fig. 2, which is itself limited to a few scenes.
  • standard math The 3D Gaussian Splatting renderer and its differentiable rasterization are a valid forward model.
    Appendix A.3; standard background from [1].

pith-pipeline@v1.3.0-alltime-deepseek · 20929 in / 12795 out tokens · 124083 ms · 2026-08-04T15:39:35.357691+00:00 · methodology

0 comments
read the original abstract

In this paper, we propose the first Structure-from-Motion (SfM)-free deblurring 3D Gaussian Splatting method via event camera, dubbed DeblurSplat. We address the motion-deblurring problem in two ways. First, we leverage the pretrained capability of the dense stereo module (DUSt3R) to directly obtain accurate initial point clouds from blurred images. Without calculating camera poses as an intermediate result, we avoid the cumulative errors transfer from inaccurate camera poses to the initial point clouds' positions. Second, we introduce the event stream into the deblur pipeline for its high sensitivity to dynamic change. By decoding the latent sharp images from the event stream and blurred images, we can provide a fine-grained supervision signal for scene reconstruction optimization. Extensive experiments across a range of scenes demonstrate that DeblurSplat not only excels in generating high-fidelity novel views but also achieves significant rendering efficiency compared to the SOTAs in deblur 3D-GS.

Figures

Figures reproduced from arXiv: 2509.18898 by F. Richard Yu, Hui Xiong, Huizai Yao, Pengteng Li, Pinhao Song, Weiyu Guo, Yunfan Lu.

Figure 1
Figure 1. Figure 1: Illustration of differences between our work and previous works for [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 4
Figure 4. Figure 4: Recently, SfM-free novel-view synthesis approaches [17], [5], [18], [19], [20] have emerged, showing potential to circumvent SfM pitfalls by directly estimating scene geometry in an end-to-end manner. In particular, DUSt3R [5] employs a dense stereo module pretrained on large-scale datasets, demonstrating greater robustness in producing point clouds arXiv:2509.18898v1 [cs.CV] 23 Sep 2025 [PITH_FULL_IMAGE:… view at source ↗
Figure 2
Figure 2. Figure 2: Empirical study of initial point clouds generated from blur images by DUSt3R. Here, “Freq.” denotes frequency, “Dist.” denotes distribution, “Conf.” [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Illustration of our proposed DeblurSplat. We first input the blur images into the dense stereo module (DUSt3R [5]) for getting estimated point [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Quantitative comparisons of time consuming (min.) of initialization [PITH_FULL_IMAGE:figures/full_fig_p005_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Qualitative comparison on the synthetic and real dataset. It shows that our method achieves better performance in rendering novel views compared [PITH_FULL_IMAGE:figures/full_fig_p006_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Analysis of efficiency of Confidence Balanced Sampling (CBS) in [PITH_FULL_IMAGE:figures/full_fig_p007_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Illustration of our crafted hard scene datasets. For each scene, the first line is the rendered mode and the second line is the wire frame mode with [PITH_FULL_IMAGE:figures/full_fig_p010_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Illustration of our crafted camera trajectory for hard scenes. Compared to normal scene datasets [6]. The average movement distance of the camera [PITH_FULL_IMAGE:figures/full_fig_p011_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Illustration of our equipment for creating real datasets. [PITH_FULL_IMAGE:figures/full_fig_p011_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Different blur degrees for each scene in the real-world dataset captured using the DVS346 camera. To capture videos with varying blur levels, we [PITH_FULL_IMAGE:figures/full_fig_p012_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Qualitative comparison on the remaining synthetic dataset on novel views synthesis. It shows that our method achieves better performance in rendering [PITH_FULL_IMAGE:figures/full_fig_p017_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: Qualitative comparison on the normal scenes dataset on deblur views rendering. It shows that our method achieves better performance in deblur [PITH_FULL_IMAGE:figures/full_fig_p018_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: Qualitative comparison on the hard scenes dataset on deblur views rendering. It shows that our method achieves better performance in deblur [PITH_FULL_IMAGE:figures/full_fig_p019_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: Qualitative comparison on the real-world scenes on deblur views rendering. [PITH_FULL_IMAGE:figures/full_fig_p020_14.png] view at source ↗

discussion (0)

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

Forward citations

Cited by 2 Pith papers

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

  1. JADE-GS: Joint Allocation of Deblurring Evidence for Event-Assisted 3D Gaussian Splatting

    cs.CV 2026-07 conditional novelty 7.0

    JADE-GS trains a small per-pixel router to combine analytical event-inversion and learned frame–event restoration as extra supervision for 3D Gaussian Splatting, improving perceptual and fidelity metrics on event-debl...

  2. JADE-GS: Joint Allocation of Deblurring Evidence for Event-Assisted 3D Gaussian Splatting

    cs.CV 2026-07 conditional novelty 6.0

    Fusing a physics-based event deblurring prior with a learned refiner through an adaptive gate, and letting 3D Gaussian Splatting renders regularize the deblurrer in a bidirectional loop, yields state-of-the-art percep...

Reference graph

Works this paper leans on

51 extracted references · 15 linked inside Pith · cited by 1 Pith paper

  1. [1]

    3d gaussian splatting for real-time radiance field rendering

    B. Kerbl, G. Kopanas, T. Leimk ¨uhler, and G. Drettakis, “3d gaussian splatting for real-time radiance field rendering.”ACM TOG, vol. 42, no. 4, pp. 139–1, 2023

    2023

  2. [2]

    Nerf: Representing scenes as neural radiance fields for view synthesis,

    B. Mildenhall, P. P. Srinivasan, M. Tancik, J. T. Barron, R. Ramamoorthi, and R. Ng, “Nerf: Representing scenes as neural radiance fields for view synthesis,”Communications of the ACM, vol. 65, no. 1, pp. 99–106, 2021

    2021

  3. [3]

    Bad-gaussians: Bundle adjusted deblur gaussian splatting,

    L. Zhao, P. Wang, and P. Liu, “Bad-gaussians: Bundle adjusted deblur gaussian splatting,” inECCV. Springer, 2025, pp. 233–250

    2025

  4. [4]

    Structure-from-motion revisited,

    J. L. Sch ¨onberger and J.-M. Frahm, “Structure-from-motion revisited,” inCVPR, 2016

    2016

  5. [5]

    Dust3r: Geometric 3d vision made easy,

    S. Wang, V . Leroy, Y . Cabon, B. Chidlovskii, and J. Revaud, “Dust3r: Geometric 3d vision made easy,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024, pp. 20 697–20 709

    2024

  6. [6]

    Deblur-nerf: Neural radiance fields from blurry images,

    L. Ma, X. Li, J. Liao, Q. Zhang, X. Wang, J. Wang, and P. V . Sander, “Deblur-nerf: Neural radiance fields from blurry images,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2022, pp. 12 861–12 870

    2022

  7. [7]

    Dp-nerf: Deblurred neural radiance field with physical scene priors,

    D. Lee, M. Lee, C. Shin, and S. Lee, “Dp-nerf: Deblurred neural radiance field with physical scene priors,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023, pp. 12 386–12 396

    2023

  8. [8]

    Bad-nerf: Bundle adjusted deblur neural radiance fields,

    P. Wang, L. Zhao, R. Ma, and P. Liu, “Bad-nerf: Bundle adjusted deblur neural radiance fields,” inCVPR, 2023, pp. 4170–4179

    2023

  9. [9]

    Event- based vision: A survey,

    G. Gallego, T. Delbr ¨uck, G. Orchard, C. Bartolozzi, B. Taba, A. Censi, S. Leutenegger, A. J. Davison, J. Conradt, K. Daniilidiset al., “Event- based vision: A survey,”IEEE transactions on pattern analysis and machine intelligence, vol. 44, no. 1, pp. 154–180, 2020

    2020

  10. [10]

    E2gs: Event enhanced gaussian splatting,

    H. Deguchi, M. Masuda, T. Nakabayashi, and H. Saito, “E2gs: Event enhanced gaussian splatting,” inICIP. IEEE, 2024, pp. 1676–1682

    2024

  11. [11]

    E3nerf: Efficient event- enhanced neural radiance fields from blurry images,

    Y . Qi, J. Li, Y . Zhao, Y . Zhang, and L. Zhu, “E3nerf: Efficient event- enhanced neural radiance fields from blurry images,”arXiv:2408.01840, 2024

    arXiv 2024

  12. [12]

    Evagaus- sians: Event stream assisted gaussian splatting from blurry images,

    W. Yu, C. Feng, J. Tang, X. Jia, L. Yuan, and Y . Tian, “Evagaus- sians: Event stream assisted gaussian splatting from blurry images,” arXiv:2405.20224, 2024

    Pith/arXiv arXiv 2024

  13. [13]

    Eventsplat: 3d gaus- sian splatting from moving event cameras for real-time rendering,

    T. Yura, A. Mirzaei, and I. Gilitschenski, “Eventsplat: 3d gaus- sian splatting from moving event cameras for real-time rendering,” arXiv:2412.07293, 2024

    Pith/arXiv arXiv 2024

  14. [14]

    Besplat–gaussian splatting from a single blurry image and event stream,

    G. R. Matta, R. Trisha, and K. Mitra, “Besplat–gaussian splatting from a single blurry image and event stream,”arXiv:2412.19370, 2024

    Pith/arXiv arXiv 2024

  15. [15]

    Elite-evgs: Learning event-based 3d gaussian splatting by distilling event-to-video priors,

    Z. Zhang, K. Chen, and L. Wang, “Elite-evgs: Learning event-based 3d gaussian splatting by distilling event-to-video priors,”arXiv:2409.13392, 2024

    Pith/arXiv arXiv 2024

  16. [16]

    Bringing a blurry frame alive at high frame-rate with an event camera,

    L. Pan, C. Scheerlinck, X. Yu, R. Hartley, M. Liu, and Y . Dai, “Bringing a blurry frame alive at high frame-rate with an event camera,” inCVPR, 2019, pp. 6820–6829

    2019

  17. [17]

    Light3r- sfm: Towards feed-forward structure-from-motion,

    S. Elflein, Q. Zhou, S. Agostinho, and L. Leal-Taix ´e, “Light3r- sfm: Towards feed-forward structure-from-motion,”arXiv preprint arXiv:2501.14914, 2025

    Pith/arXiv arXiv 2025

  18. [18]

    Align3r: Aligned monocular depth estimation for dynamic videos,

    J. Lu, T. Huang, P. Li, Z. Dou, C. Lin, Z. Cui, Z. Dong, S.-K. Yeung, W. Wang, and Y . Liu, “Align3r: Aligned monocular depth estimation for dynamic videos,”arXiv preprint arXiv:2412.03079, 2024

    Pith/arXiv arXiv 2024

  19. [19]

    Freesplatter: Pose-free gaussian splatting for sparse-view 3d reconstruction,

    J. Xu, S. Gao, and Y . Shan, “Freesplatter: Pose-free gaussian splatting for sparse-view 3d reconstruction,”arXiv preprint arXiv:2412.09573, 2024

    Pith/arXiv arXiv 2024

  20. [20]

    Flare: Feed-forward geometry, appearance and camera estimation from uncalibrated sparse views,

    S. Zhang, J. Wang, Y . Xu, N. Xue, C. Rupprecht, X. Zhou, Y . Shen, and G. Wetzstein, “Flare: Feed-forward geometry, appearance and camera estimation from uncalibrated sparse views,”arXiv preprint arXiv:2502.12138, 2025

    arXiv 2025

  21. [21]

    Eventnerf: Neural radiance fields from a single colour event camera,

    V . Rudnev, M. Elgharib, C. Theobalt, and V . Golyanik, “Eventnerf: Neural radiance fields from a single colour event camera,” inCVPR, 2023, pp. 4992–5002

    2023

  22. [22]

    E-nerf: Neural radiance fields from a moving event camera,

    S. Klenk, L. Koestler, D. Scaramuzza, and D. Cremers, “E-nerf: Neural radiance fields from a moving event camera,”IEEE RAL, vol. 8, no. 3, pp. 1587–1594, 2023

    2023

  23. [23]

    Event3dgs: Event-based 3d gaussian splatting for high- speed robot egomotion,

    T. Xiong, J. Wu, B. He, C. Fermuller, Y . Aloimonos, H. Huang, and C. Metzler, “Event3dgs: Event-based 3d gaussian splatting for high- speed robot egomotion,” in8th Annual Conference on Robot Learning, 2024

    2024

  24. [24]

    Event-3dgs: Event-based 3d re- construction using 3d gaussian splatting,

    H. Han, J. Li, H. Wei, and X. Ji, “Event-3dgs: Event-based 3d re- construction using 3d gaussian splatting,” inThe Thirty-eighth Annual Conference on Neural Information Processing Systems, 2023

    2023

  25. [25]

    Robust e-nerf: Nerf from sparse & noisy events under non-uniform motion,

    W. F. Low and G. H. Lee, “Robust e-nerf: Nerf from sparse & noisy events under non-uniform motion,” inICCV, 2023, pp. 18 335–18 346

    2023

  26. [26]

    Deblur e-nerf: Nerf from motion-blurred events under high-speed or low-light conditions,

    ——, “Deblur e-nerf: Nerf from motion-blurred events under high-speed or low-light conditions,” inECCV. Springer, 2025, pp. 192–209

    2025

  27. [27]

    Sweepevgs: Event- based 3d gaussian splatting for macro and micro radiance field rendering from a single sweep,

    J. Wu, S. Zhu, C. Wang, B. Shi, and E. Y . Lam, “Sweepevgs: Event- based 3d gaussian splatting for macro and micro radiance field rendering from a single sweep,”arXiv:2412.11579, 2024

    arXiv 2024

  28. [28]

    E2nerf: Event enhanced neural radiance fields from blurry images,

    Y . Qi, L. Zhu, Y . Zhang, and J. Li, “E2nerf: Event enhanced neural radiance fields from blurry images,” inICCV, 2023, pp. 13 254–13 264

    2023

  29. [29]

    Ev-gs: Event-based gaussian splatting for efficient and accurate radiance field rendering,

    J. Wu, S. Zhu, C. Wang, and E. Y . Lam, “Ev-gs: Event-based gaussian splatting for efficient and accurate radiance field rendering,” inMLSP. IEEE, 2024, pp. 1–6

    2024

  30. [30]

    Mitigating motion blur in neural radiance fields with events and frames,

    M. Cannici and D. Scaramuzza, “Mitigating motion blur in neural radiance fields with events and frames,” inCVPR, 2024, pp. 9286–9296

    2024

  31. [31]

    Liftimage3d: Lifting any single image to 3d gaussians with video generation priors,

    Y . Chen, C. Yang, J. Fang, X. Zhang, L. Xie, W. Shen, W. Dai, H. Xiong, and Q. Tian, “Liftimage3d: Lifting any single image to 3d gaussians with video generation priors,”arXiv preprint arXiv:2412.09597, 2024

    Pith/arXiv arXiv 2024

  32. [32]

    Fast3r: Towards 3d reconstruction of 1000+ images in one forward pass,

    J. Yang, A. Sax, K. J. Liang, M. Henaff, H. Tang, A. Cao, J. Chai, F. Meier, and M. Feiszli, “Fast3r: Towards 3d reconstruction of 1000+ images in one forward pass,”arXiv preprint arXiv:2501.13928, 2025

    Pith/arXiv arXiv 2025

  33. [33]

    Grounding image matching in 3d with mast3r,

    V . Leroy, Y . Cabon, and J. Revaud, “Grounding image matching in 3d with mast3r,” inEuropean Conference on Computer Vision. Springer, 2024, pp. 71–91

    2024

  34. [34]

    Pf3plat: Pose-free feed-forward 3d gaussian splatting,

    S. Hong, J. Jung, H. Shin, J. Han, J. Yang, C. Luo, and S. Kim, “Pf3plat: Pose-free feed-forward 3d gaussian splatting,”arXiv preprint arXiv:2410.22128, 2024

    Pith/arXiv arXiv 2024

  35. [35]

    Splatt3r: Zero- shot gaussian splatting from uncalibrated image pairs,

    B. Smart, C. Zheng, I. Laina, and V . A. Prisacariu, “Splatt3r: Zero- shot gaussian splatting from uncalibrated image pairs,”arXiv preprint arXiv:2408.13912, 2024

    Pith/arXiv arXiv 2024

  36. [36]

    No pose, no problem: Surprisingly simple 3d gaussian splats from sparse unposed images,

    B. Ye, S. Liu, H. Xu, X. Li, M. Pollefeys, M.-H. Yang, and S. Peng, “No pose, no problem: Surprisingly simple 3d gaussian splats from sparse unposed images,”arXiv preprint arXiv:2410.24207, 2024

    Pith/arXiv arXiv 2024

  37. [37]

    Spars3r: Semantic prior align- ment and regularization for sparse 3d reconstruction,

    Y . Tang, Y . Guo, D. Li, and C. Peng, “Spars3r: Semantic prior align- ment and regularization for sparse 3d reconstruction,”arXiv preprint arXiv:2411.12592, 2024

    Pith/arXiv arXiv 2024

  38. [38]

    Instantsplat: Unbounded sparse-view pose-free gaussian splatting in 40 seconds,

    Z. Fan, W. Cong, K. Wen, K. Wang, J. Zhang, X. Ding, D. Xu, B. Ivanovic, M. Pavone, G. Pavlakoset al., “Instantsplat: Unbounded sparse-view pose-free gaussian splatting in 40 seconds,”arXiv preprint arXiv:2403.20309, vol. 2, no. 3, p. 4, 2024

    Pith/arXiv arXiv 2024

  39. [39]

    Pointnet++: Deep hierarchical feature learning on point sets in a metric space,

    C. R. Qi, L. Yi, H. Su, and L. J. Guibas, “Pointnet++: Deep hierarchical feature learning on point sets in a metric space,”Advances in neural information processing systems, vol. 30, 2017

    2017

  40. [40]

    High speed and high dynamic range video with an event camera,

    H. Rebecq, R. Ranftl, V . Koltun, and D. Scaramuzza, “High speed and high dynamic range video with an event camera,”IEEE transactions on pattern analysis and machine intelligence, vol. 43, no. 6, pp. 1964–1980, 2019

    1964

  41. [41]

    Inceventgs: Pose-free gaussian splatting from a single event camera,

    J. Huang, C. Dong, and P. Liu, “Inceventgs: Pose-free gaussian splatting from a single event camera,”arXiv:2410.08107, 2024

    Pith/arXiv arXiv 2024

  42. [42]

    Self-supervised non- uniform kernel estimation with flow-based motion prior for blind image deblurring,

    Z. Fang, F. Wu, W. Dong, X. Li, J. Wu, and G. Shi, “Self-supervised non- uniform kernel estimation with flow-based motion prior for blind image deblurring,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2023, pp. 18 105–18 114

    2023

  43. [43]

    Event-based fusion for motion deblurring with cross- modal attention,

    L. Sun, C. Sakaridis, J. Liang, Q. Jiang, K. Yang, P. Sun, Y . Ye, K. Wang, and L. V . Gool, “Event-based fusion for motion deblurring with cross- modal attention,” inEuropean conference on computer vision. Springer, 2022, pp. 412–428

    2022

  44. [44]

    Image quality assessment: from error visibility to structural similarity,

    Z. Wang, A. C. Bovik, H. R. Sheikh, and E. P. Simoncelli, “Image quality assessment: from error visibility to structural similarity,”IEEE transactions on image processing, vol. 13, no. 4, pp. 600–612, 2004

    2004

  45. [45]

    The unreasonable effectiveness of deep features as a perceptual metric,

    R. Zhang, P. Isola, A. A. Efros, E. Shechtman, and O. Wang, “The unreasonable effectiveness of deep features as a perceptual metric,” in Proceedings of the IEEE conference on computer vision and pattern recognition, 2018, pp. 586–595

    2018

  46. [46]

    The weiszfeld algorithm: proof, amendments, and exten- sions,

    F. Plastria, “The weiszfeld algorithm: proof, amendments, and exten- sions,”Foundations of location analysis, pp. 357–389, 2011

    2011

  47. [47]

    Random sample consensus: a paradigm for model fitting with applications to image analysis and automated car- tography,

    M. FISCHLER AND, “Random sample consensus: a paradigm for model fitting with applications to image analysis and automated car- tography,”Commun. ACM, vol. 24, no. 6, pp. 381–395, 1981

    1981

  48. [48]

    Ep n p: An accurate o (n) solution to the p n p problem,

    V . Lepetit, F. Moreno-Noguer, and P. Fua, “Ep n p: An accurate o (n) solution to the p n p problem,”International journal of computer vision, vol. 81, pp. 155–166, 2009

    2009

  49. [49]

    Video to events: Recycling video datasets for event cameras,

    D. Gehrig, M. Gehrig, J. Hidalgo-Carri ´o, and D. Scaramuzza, “Video to events: Recycling video datasets for event cameras,” inIEEE Conf. Comput. Vis. Pattern Recog. (CVPR), June 2020

    2020

  50. [50]

    Video to events: Recycling video datasets for event cameras,

    D. Gehrig and D. Scaramuzza, “Video to events: Recycling video datasets for event cameras,” inIEEE Conf. Comput. Vis. Pattern Recog. (CVPR), June 2020. 9 APPENDIXA PRELIMINARIES A. Event generation model Unlike frame-based cameras that record the brightness of each pixel at a fixed frame rate [9], the event camera asyn- chronously generates an evente(x, y...

    arXiv 2020

  51. [51]

    In contrast, our method significantly outperforms these baselines by rendering high-fidelity deblur views and novel views

    Our findings indicate that while previous state-of-the- art methods perform well in normal scenes, they struggle in 16 challenging cases like hard scenes, particularly when handling complex textures and large-scale camera motion, often lead- ing to severe blurring. In contrast, our method significantly outperforms these baselines by rendering high-fidelit...