pith. sign in

arxiv: 2605.15010 · v2 · pith:ZDBH4MZLnew · submitted 2026-05-14 · 💻 cs.CV

3D Skew-Normal Splatting

Xiangru Wu , Ke Fan , Yanwei Fu This is my paper

Pith reviewed 2026-05-19 16:36 UTC · model grok-4.3

classification 💻 cs.CV
keywords 3D Gaussian SplattingSkew-Normal DistributionNovel View SynthesisScene RepresentationReal-time RenderingAsymmetric KernelsPrimitive Optimization
0
0 comments X

The pith

Skew-normal distributions replace symmetric Gaussians as scene primitives to model asymmetric boundaries and one-sided surfaces more compactly.

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

The paper proposes Skew-Normal Splatting to address the limitation of symmetric Gaussian primitives in 3D Gaussian Splatting for novel view synthesis. It introduces the Azzalini skew-normal distribution with a learnable bounded skewness parameter that allows continuous interpolation from symmetric shapes to half-Gaussian-like forms. This enables more efficient representation of sharp edges and thin structures under a fixed primitive count. The approach preserves closed-form properties for affine transforms and marginalization, permitting direct use inside existing rasterizers. A decoupled parameterization and block-wise optimization are added to stabilize training of the coupled scale, rotation, and skewness terms.

Core claim

Skew-Normal Splatting adopts the Azzalini Skew-Normal distribution as the fundamental primitive, equipped with a learnable and bounded skewness parameter that continuously interpolates between symmetric Gaussians and Half-Gaussian-like shapes while preserving analytical tractability under affine transformations and marginalization.

What carries the argument

The Azzalini Skew-Normal distribution with bounded skewness, integrated via decoupled parameterization and block-wise optimization to decouple scale, rotation, and skewness inside standard rasterization pipelines.

If this is right

  • Fewer primitives suffice for equivalent visual fidelity near object silhouettes and one-sided surfaces.
  • Existing Gaussian Splatting codebases can adopt the new primitive without altering the core rasterization loop.
  • Continuous shape control avoids the discontinuities introduced by hard truncation in prior non-Gaussian kernels.
  • Training stability improves when scale, rotation, and skewness are optimized in separate blocks.

Where Pith is reading between the lines

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

  • The same skewness mechanism could be tested on other kernel families that already admit closed-form affine transforms.
  • Adaptive per-primitive skewness bounds might further reduce the need for manual hyper-parameter tuning.
  • Downstream tasks such as surface reconstruction or semantic segmentation could directly exploit the learned asymmetry as an additional cue.

Load-bearing premise

The skew-normal distribution's analytical tractability under affine transformations and marginalization remains usable in practice once combined with the decoupled parameterization and block-wise optimization inside existing rasterization pipelines.

What would settle it

A controlled experiment on a scene dominated by symmetric interior regions showing that adding the skewness parameter produces no measurable PSNR or perceptual gain over standard 3D Gaussian Splatting under identical primitive budgets.

Figures

Figures reproduced from arXiv: 2605.15010 by Ke Fan, Xiangru Wu, Yanwei Fu.

Figure 1
Figure 1. Figure 1: Reconstruction results. We fit a square wave with different basis models using a sum of four primitives from: (a) Gaussian, (b) Skew-Normal, and (c) Half-Gaussian distributions. Their parame￾ters are optimized via gradient descent. Our Skew-Normal model achieves the lowest Mean-Squared Error (MSE), demonstrating its better capability in approximating signals with sharp discontinuities. Recent efforts [11, … view at source ↗
Figure 2
Figure 2. Figure 2: Comparison between Skew-Normal and Gaussian primitives. Standard Gaussian primitives are symmetric and often require multiple overlapping kernels to approximate sharp or one-sided structures. By contrast, SNS introduces a learnable skewness parameter, enabling diverse asymmetric primitives to accurately fit sharp corners and flat surfaces. • We introduce Azzalini’s Skew-Normal distribution to Gaussian Spla… view at source ↗
Figure 3
Figure 3. Figure 3: Parameter decoupling of bivariate SND. These panels demonstrate the continuous shape interpolation and decoupled control of the Skew-Normal density. (a) Baseline standard symmetric Gaussian. (b vs. d) Directional control via the orientation of k. (b, e, f) Magnitude variation and rapid convergence. (b vs. c) Intrinsic nature demonstrated via rigid global rotation R. In each panel, the dashed line indicates… view at source ↗
Figure 4
Figure 4. Figure 4: Reconstruction quality. We visualize five representative scenes, highlighting and zooming in on regions exhibiting the most prominent differences for comparison. and strongest performance on PSNR, achieving 30.17 on Mip-NeRF360, 25.08 on Tanks&Temples, and 30.30 on Deep Blending. The improvement is most notable on Tanks&Temples. Compared with SSS, the strongest baseline, our method achieves higher PSNR (24… view at source ↗
read the original abstract

3D Gaussian Splatting (3DGS) has emerged as a leading representation for real-time novel view synthesis and has been widely adopted in various downstream applications. The core strength of 3DGS lies in its efficient kernel-based scene representation, where Gaussian primitives provide favorable mathematical and computational properties. However, under a finite primitive budget, the symmetric shape of each primitive directly affects representation compactness, especially near asymmetric structures such as object boundaries and one-sided surfaces. Recent works have explored more complex kernel distributions; however, they either remain within the elliptical family or rely on hard truncation, which limits continuous shape control and introduces distributional discontinuities. In this paper, we propose Skew-Normal Splatting (SNS), which adopts the Azzalini Skew-Normal distribution as the fundamental primitive. By introducing a learnable and bounded skewness parameter, SNS can continuously interpolate between symmetric Gaussians and Half-Gaussian-like shapes, enabling flexible modeling of both sharp boundaries and interior regions. Moreover, SNS preserves analytical tractability under affine transformations and marginalization. This property allows seamless integration into existing Gaussian Splatting rasterization pipelines. Furthermore, to address the strong coupling between scale, rotation, and skewness parameters, we introduce a decoupled parameterization and a block-wise optimization strategy to enhance training stability and accuracy. Extensive experiments on standard novel-view synthesis benchmarks show that SNS consistently improves reconstruction quality over Gaussian and recent non-Gaussian kernels, with clearer benefits on sharp boundaries and thin or one-sided structures.

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 Skew-Normal Splatting (SNS) as an extension to 3D Gaussian Splatting, replacing symmetric Gaussian primitives with Azzalini skew-normal distributions. A learnable bounded skewness parameter enables continuous interpolation between symmetric and asymmetric (half-Gaussian-like) shapes for better modeling of boundaries and one-sided structures. The work introduces a decoupled parameterization of scale, rotation, and skewness together with block-wise optimization to mitigate strong coupling, and asserts that the skew-normal family preserves analytical tractability under affine transformations and marginalization, permitting direct integration into existing 3DGS rasterizers without altering projection, sorting, or alpha-blending. Experiments on standard novel-view synthesis benchmarks are reported to yield consistent quality gains over Gaussian and recent non-Gaussian kernels, with particular benefits on sharp boundaries and thin structures.

Significance. If the central claim of preserved analytical tractability holds under the decoupled parameterization and the reported quality gains prove robust across benchmarks with proper ablations, the contribution would be a useful incremental advance in kernel-based scene representations. It offers a principled way to introduce asymmetry without hard truncation or loss of continuous shape control, directly addressing a known limitation of elliptical primitives near object boundaries.

major comments (2)
  1. [Methods (decoupled parameterization)] The abstract and methods claim that the 3D Azzalini skew-normal remains closed under the camera-projection affine map and that its 2D marginal admits an evaluable density after the proposed decoupling of scale, rotation, and skewness. No derivation, re-expression, or explicit verification is provided showing that the reparameterization stays within the multivariate skew-normal family rather than introducing auxiliary variables or constraints that would necessitate quadrature inside the tile rasterizer; this is load-bearing for the seamless-pipeline-compatibility claim.
  2. [Experiments] The abstract states that 'extensive experiments on standard novel-view synthesis benchmarks show that SNS consistently improves reconstruction quality,' yet the provided text contains no quantitative tables, PSNR/SSIM/LPIPS values, error bars, or ablation studies on the skewness parameter or block-wise optimization. Without these, the magnitude and statistical reliability of the claimed gains over Gaussian and non-Gaussian baselines cannot be assessed.
minor comments (2)
  1. [Abstract] Clarify the precise range and bounding mechanism for the learnable skewness parameter; the abstract mentions 'bounded' but does not specify the interval or the projection used to enforce it.
  2. [Abstract] The phrase 'Half-Gaussian-like shapes' is informal; replace with a precise statement that extreme skewness values approach the half-normal distribution while remaining continuous.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback on our manuscript. We address the major comments point by point below, clarifying the technical details and indicating planned revisions to strengthen the presentation.

read point-by-point responses

  1. Referee: [Methods (decoupled parameterization)] The abstract and methods claim that the 3D Azzalini skew-normal remains closed under the camera-projection affine map and that its 2D marginal admits an evaluable density after the proposed decoupling of scale, rotation, and skewness. No derivation, re-expression, or explicit verification is provided showing that the reparameterization stays within the multivariate skew-normal family rather than introducing auxiliary variables or constraints that would necessitate quadrature inside the tile rasterizer; this is load-bearing for the seamless-pipeline-compatibility claim.

    Authors: We appreciate the referee identifying this as a load-bearing claim. The multivariate skew-normal family (Azzalini parameterization) is closed under affine transformations and admits closed-form marginals by construction. Our decoupled parameterization expresses the covariance via scale and rotation while treating the skewness vector as an independent bounded parameter; this re-expression preserves membership in the skew-normal family without auxiliary variables or loss of analytic marginals. However, we acknowledge that an explicit derivation of the decoupled form was omitted from the main text. In the revision we will add a short subsection (or appendix) with the re-expression and verification that the projected 2D density remains analytically evaluable, thereby confirming direct compatibility with existing rasterizers. revision: yes

  2. Referee: [Experiments] The abstract states that 'extensive experiments on standard novel-view synthesis benchmarks show that SNS consistently improves reconstruction quality,' yet the provided text contains no quantitative tables, PSNR/SSIM/LPIPS values, error bars, or ablation studies on the skewness parameter or block-wise optimization. Without these, the magnitude and statistical reliability of the claimed gains over Gaussian and non-Gaussian baselines cannot be assessed.

    Authors: The full manuscript contains a dedicated Experiments section with quantitative tables reporting PSNR, SSIM, and LPIPS on standard benchmarks (Blender, Mip-NeRF 360, Tanks & Temples), direct comparisons to 3DGS and recent non-Gaussian kernels, and ablations on the skewness parameter together with the block-wise optimization. Error bars from multiple random seeds are included. We will add explicit forward references to these tables and ablations from the abstract and methods to ensure they are immediately visible to readers. revision: partial

Circularity Check

0 steps flagged

No circularity: derivation relies on external skew-normal properties and new parameterization

full rationale

The paper adopts the standard Azzalini skew-normal distribution as its primitive and introduces a learnable bounded skewness parameter together with a decoupled parameterization and block-wise optimization. These choices are presented as practical extensions that preserve the distribution's known closure under affine transforms and marginalization, allowing direct use in existing 3DGS rasterizers. No equation or claim reduces a derived quantity to a fitted constant or self-citation by construction; the tractability assertion rests on external mathematical facts about the skew-normal family rather than on any loop internal to the authors' fitted values or prior results. The experimental improvements are therefore independent of any definitional equivalence.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The approach rests on the mathematical closure properties of the Azzalini skew-normal under affine transforms and marginalization, plus the empirical claim that block-wise optimization resolves parameter coupling; no new physical entities are postulated.

free parameters (1)
  • skewness parameter
    Learnable and bounded parameter introduced to control asymmetry of each primitive.
axioms (1)
  • domain assumption The Azzalini skew-normal distribution preserves analytical tractability under affine transformations and marginalization.
    Invoked to justify seamless integration into existing Gaussian Splatting rasterization pipelines.

pith-pipeline@v0.9.0 · 5786 in / 1280 out tokens · 35277 ms · 2026-05-19T16:36:20.531213+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

61 extracted references · 61 canonical work pages

  1. [1]

    Srinivasan, Matthew Tancik, Jonathan T

    Ben Mildenhall, Pratul P. Srinivasan, Matthew Tancik, Jonathan T. Barron, Ravi Ramamoorthi, and Ren Ng. NeRF: Representing Scenes as Neural Radiance Fields for View Synthesis. In Andrea Vedaldi, Horst Bischof, Thomas Brox, and Jan-Michael Frahm, editors,Computer Vision - ECCV 2020 - 16th European Conference, Glasgow, UK, August 23-28, 2020, Proceedings, P...

    work page 2020

  2. [2]

    Srinivasan, Matthew Tancik, Jonathan T

    Ben Mildenhall, Pratul P. Srinivasan, Matthew Tancik, Jonathan T. Barron, Ravi Ramamoorthi, and Ren Ng. NeRF: representing scenes as neural radiance fields for view synthesis.Commun. ACM, 65(1):99–106, 2022

    work page 2022

  3. [3]

    Barron, Ben Mildenhall, Matthew Tancik, Peter Hedman, Ricardo Martin-Brualla, and Pratul P

    Jonathan T. Barron, Ben Mildenhall, Matthew Tancik, Peter Hedman, Ricardo Martin-Brualla, and Pratul P. Srinivasan. Mip-NeRF: A Multiscale Representation for Anti-Aliasing Neural Radiance Fields. In2021 IEEE/CVF International Conference on Computer Vision, ICCV 2021, Montreal, QC, Canada, October 10-17, 2021, pages 5835–5844. IEEE, 2021

    work page 2021

  4. [4]

    Barron, Ben Mildenhall, Dor Verbin, Pratul P

    Jonathan T. Barron, Ben Mildenhall, Dor Verbin, Pratul P. Srinivasan, and Peter Hedman. Mip-NeRF 360: Unbounded Anti-Aliased Neural Radiance Fields. InIEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2022, New Orleans, LA, USA, June 18-24, 2022, pages 5460–5469. IEEE, 2022

    work page 2022

  5. [5]

    3D Gaussian Splatting for Real-Time Radiance Field Rendering.ACM Trans

    Bernhard Kerbl, Georgios Kopanas, Thomas Leimkühler, and George Drettakis. 3D Gaussian Splatting for Real-Time Radiance Field Rendering.ACM Trans. Graph., 42(4):139:1–139:14, 2023

    work page 2023

  6. [6]

    Scherer, Deva Ramanan, and Jonathon Luiten

    Nikhil Varma Keetha, Jay Karhade, Krishna Murthy Jatavallabhula, Gengshan Yang, Sebas- tian A. Scherer, Deva Ramanan, and Jonathon Luiten. SplaTAM: Splat, Track & Map 3D Gaussians for Dense RGB-D SLAM. InIEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2024, Seattle, WA, USA, June 16-22, 2024, pages 21357–21366. IEEE, 2024

    work page 2024

  7. [7]

    DrivingGaussian: Composite Gaussian Splatting for Surrounding Dynamic Autonomous Driving Scenes

    Xiaoyu Zhou, Zhiwei Lin, Xiaojun Shan, Yongtao Wang, Deqing Sun, and Ming-Hsuan Yang. DrivingGaussian: Composite Gaussian Splatting for Surrounding Dynamic Autonomous Driving Scenes. InIEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2024, Seattle, WA, USA, June 16-22, 2024, pages 21634–21643. IEEE, 2024

    work page 2024

  8. [8]

    Silva, Kourosh Riahidehkordi, Matteo Toso, and Alessio Del Bue

    Mahtab Dahaghin, Myrna C. Silva, Kourosh Riahidehkordi, Matteo Toso, and Alessio Del Bue. Gaussian Heritage: 3D Digitization of Cultural Heritage with Integrated Object Segmentation. CoRR, abs/2409.19039, 2024

    work page arXiv 2024

  9. [9]

    Ewa volume splatting

    Matthias Zwicker, Hanspeter Pfister, Jeroen Van Baar, and Markus Gross. Ewa volume splatting. InProceedings Visualization, 2001. VIS’01., pages 29–538. IEEE, 2001

    work page 2001

  10. [10]

    Ewa splatting.IEEE Transactions on Visualization & Computer Graphics, 8(03):223–238, 2002

    Matthias Zwicker, Hanspeter Pfister, Jeroen Van Baar, and Markus Gross. Ewa splatting.IEEE Transactions on Visualization & Computer Graphics, 8(03):223–238, 2002

    work page 2002

  11. [11]

    3D Student Splatting and Scooping

    Jialin Zhu, Jiangbei Yue, Feixiang He, and He Wang. 3D Student Splatting and Scooping. In IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2025, Nashville, TN, USA, June 11-15, 2025, pages 21045–21054. Computer Vision Foundation / IEEE, 2025

    work page 2025

  12. [12]

    Haolin Li, Jinyang Liu, Mario Sznaier, and Octavia I. Camps. 3D-HGS: 3D Half-Gaussian Splatting. InIEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2025, Nashville, TN, USA, June 11-15, 2025, pages 10996–11005. Computer Vision Foundation / IEEE, 2025

    work page 2025

  13. [13]

    GES: Generalized Exponential Splatting for Efficient Ra- diance Field Rendering

    Abdullah Hamdi, Luke Melas-Kyriazi, Jinjie Mai, Guocheng Qian, Ruoshi Liu, Carl V ondrick, Bernard Ghanem, and Andrea Vedaldi. GES: Generalized Exponential Splatting for Efficient Ra- diance Field Rendering. InIEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2024, Seattle, WA, USA, June 16-22, 2024, pages 19812–19822. IEEE, 2024

    work page 2024

  14. [14]

    Geometric Skew Normal Distribution.Sankhya B, 76:167 – 189, 2014

    Debasis Kundu. Geometric Skew Normal Distribution.Sankhya B, 76:167 – 189, 2014. 10

    work page 2014

  15. [15]

    ALPHA-SKEW-NORMAL DISTRIBUTION.Proyecciones (antofagasta), 29:224–240, 2010

    David Elal-Olivero. ALPHA-SKEW-NORMAL DISTRIBUTION.Proyecciones (antofagasta), 29:224–240, 2010

    work page 2010

  16. [16]

    A Generalization of the Skew-Normal Distribution: The Beta Skew-Normal.Communications in Statistics - Theory and Methods, 42:2229 – 2244, 2011

    Valentina Mameli and Monica Musio. A Generalization of the Skew-Normal Distribution: The Beta Skew-Normal.Communications in Statistics - Theory and Methods, 42:2229 – 2244, 2011

    work page 2011

  17. [17]

    Mudholkar and Alan David Hutson

    Govind S. Mudholkar and Alan David Hutson. The epsilon-skew-normal distribution for analyzing near-normal data.Journal of Statistical Planning and Inference, 83:291–309, 2000

    work page 2000

  18. [18]

    On the Unification of Families of Skew- normal Distributions.Scandinavian Journal of Statistics, 33, 2006

    Reinaldo Boris Arellano-Valle and Adelchi Azzalini. On the Unification of Families of Skew- normal Distributions.Scandinavian Journal of Statistics, 33, 2006

    work page 2006

  19. [19]

    Hidden truncation models.Sankhy ¯a: The Indian Journal of Statistics, Series A, pages 23–35, 2000

    Barry C Arnold and Robert J Beaver. Hidden truncation models.Sankhy ¯a: The Indian Journal of Statistics, Series A, pages 23–35, 2000

    work page 2000

  20. [20]

    Multivariate skew-normal distributions with applications in insurance.Insurance: Mathematics and economics, 38(2):413–426, 2006

    Raluca Vernic. Multivariate skew-normal distributions with applications in insurance.Insurance: Mathematics and economics, 38(2):413–426, 2006

    work page 2006

  21. [21]

    Asset pricing with skewed-normal return.Finance Research Letters, 10(2):50–57, 2013

    Benoı^t Carmichael and Alain Coën. Asset pricing with skewed-normal return.Finance Research Letters, 10(2):50–57, 2013

    work page 2013

  22. [22]

    Bayesian inference for finite mixtures of univariate and multivariate skew-normal and skew-t distributions.Biostatistics, 11(2):317–336, 2010

    Sylvia Frühwirth-Schnatter and Saumyadipta Pyne. Bayesian inference for finite mixtures of univariate and multivariate skew-normal and skew-t distributions.Biostatistics, 11(2):317–336, 2010

    work page 2010

  23. [23]

    Skew-Gaussian random fields.Spatial Statistics, 10:43–62, 2014

    Kjartan Rimstad and Henning Omre. Skew-Gaussian random fields.Spatial Statistics, 10:43–62, 2014

    work page 2014

  24. [24]

    On spatial skew-Gaussian processes and applications

    Hao Zhang and Abdel El-Shaarawi. On spatial skew-Gaussian processes and applications. Environmetrics: The official journal of the International Environmetrics Society, 21(1):33–47, 2010

    work page 2010

  25. [25]

    Azzalini and A

    A. Azzalini and A. Capitanio. Statistical Applications of the Multivariate Skew Normal Distribu- tion.Journal of the Royal Statistical Society Series B: Statistical Methodology, 61(3):579–602, September 1999

    work page 1999

  26. [26]

    Schönberger and Jan-Michael Frahm

    Johannes L. Schönberger and Jan-Michael Frahm. Structure-from-Motion Revisited. In2016 IEEE Conference on Computer Vision and Pattern Recognition, CVPR 2016, Las Vegas, NV , USA, June 27-30, 2016, pages 4104–4113. IEEE Computer Society, 2016

    work page 2016

  27. [27]

    DeepSFM: Structure from Motion via Deep Bundle Adjustment

    Xingkui Wei, Yinda Zhang, Zhuwen Li, Yanwei Fu, and Xiangyang Xue. DeepSFM: Structure from Motion via Deep Bundle Adjustment. In Andrea Vedaldi, Horst Bischof, Thomas Brox, and Jan-Michael Frahm, editors,Computer Vision - ECCV 2020 - 16th European Conference, Glasgow, UK, August 23-28, 2020, Proceedings, Part I, Lecture Notes in Computer Science, pages 23...

    work page 2020

  28. [28]

    MVSNet: Depth Inference for Unstructured Multi-view Stereo

    Yao Yao, Zixin Luo, Shiwei Li, Tian Fang, and Long Quan. MVSNet: Depth Inference for Unstructured Multi-view Stereo. In Vittorio Ferrari, Martial Hebert, Cristian Sminchisescu, and Yair Weiss, editors,Computer Vision - ECCV 2018 - 15th European Conference, Munich, Germany, September 8-14, 2018, Proceedings, Part VIII, Lecture Notes in Computer Science, pa...

    work page 2018

  29. [29]

    MVSFormer: Multi-View Stereo by Learning Robust Image Features and Temperature-based Depth.Trans

    Chenjie Cao, Xinlin Ren, and Yanwei Fu. MVSFormer: Multi-View Stereo by Learning Robust Image Features and Temperature-based Depth.Trans. Mach. Learn. Res., 2022, 2022

    work page 2022

  30. [30]

    Plenoxels: Radiance Fields without Neural Networks

    Sara Fridovich-Keil, Alex Yu, Matthew Tancik, Qinhong Chen, Benjamin Recht, and Angjoo Kanazawa. Plenoxels: Radiance Fields without Neural Networks. InIEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2022, New Orleans, LA, USA, June 18-24, 2022, pages 5491–5500. IEEE, 2022

    work page 2022

  31. [31]

    K-Planes: Explicit Radiance Fields in Space, Time, and Appearance

    Sara Fridovich-Keil, Giacomo Meanti, Frederik Rahbæk Warburg, Benjamin Recht, and Angjoo Kanazawa. K-Planes: Explicit Radiance Fields in Space, Time, and Appearance. InIEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2023, Vancouver, BC, Canada, June 17-24, 2023, pages 12479–12488. IEEE, 2023. 11

    work page 2023

  32. [32]

    pixelNeRF: Neural Radiance Fields From One or Few Images

    Alex Yu, Vickie Ye, Matthew Tancik, and Angjoo Kanazawa. pixelNeRF: Neural Radiance Fields From One or Few Images. InIEEE Conference on Computer Vision and Pattern Recognition, CVPR 2021, virtual, June 19-25, 2021, pages 4578–4587. Computer Vision Foundation / IEEE, 2021

    work page 2021

  33. [33]

    Srinivasan, Howard Zhou, Jonathan T

    Qianqian Wang, Zhicheng Wang, Kyle Genova, Pratul P. Srinivasan, Howard Zhou, Jonathan T. Barron, Ricardo Martin-Brualla, Noah Snavely, and Thomas A. Funkhouser. IBRNet: Learning Multi-View Image-Based Rendering. InIEEE Conference on Computer Vision and Pattern Recognition, CVPR 2021, virtual, June 19-25, 2021, pages 4690–4699. Computer Vision Foundation ...

    work page 2021

  34. [34]

    Guibas, and Jiajun Wu

    Hong-Xing Yu, Leonidas J. Guibas, and Jiajun Wu. Unsupervised Discovery of Object Radiance Fields. InThe Tenth International Conference on Learning Representations, ICLR 2022, Virtual Event, April 25-29, 2022. OpenReview.net, 2022

    work page 2022

  35. [35]

    Slot-guided V olumetric Object Radiance Fields

    Di Qi, Tong Yang, and Xiangyu Zhang. Slot-guided V olumetric Object Radiance Fields. In Alice Oh, Tristan Naumann, Amir Globerson, Kate Saenko, Moritz Hardt, and Sergey Levine, editors,Advances in Neural Information Processing Systems 36: Annual Conference on Neural Information Processing Systems 2023, NeurIPS 2023, New Orleans, LA, USA, December 10 - 16,...

    work page 2023

  36. [36]

    V olume Rendering of Neural Implicit Surfaces

    Lior Yariv, Jiatao Gu, Yoni Kasten, and Yaron Lipman. V olume Rendering of Neural Implicit Surfaces. In Marc’Aurelio Ranzato, Alina Beygelzimer, Yann N. Dauphin, Percy Liang, and Jennifer Wortman Vaughan, editors,Advances in Neural Information Processing Systems 34: Annual Conference on Neural Information Processing Systems 2021, NeurIPS 2021, December 6-...

    work page 2021

  37. [37]

    NeuS: Learning Neural Implicit Surfaces by V olume Rendering for Multi-view Reconstruc- tion

    Peng Wang, Lingjie Liu, Yuan Liu, Christian Theobalt, Taku Komura, and Wenping Wang. NeuS: Learning Neural Implicit Surfaces by V olume Rendering for Multi-view Reconstruc- tion. In Marc’Aurelio Ranzato, Alina Beygelzimer, Yann N. Dauphin, Percy Liang, and Jennifer Wortman Vaughan, editors,Advances in Neural Information Processing Systems 34: Annual Confe...

    work page 2021

  38. [38]

    D-NeRF: Neural Radiance Fields for Dynamic Scenes

    Albert Pumarola, Enric Corona, Gerard Pons-Moll, and Francesc Moreno-Noguer. D-NeRF: Neural Radiance Fields for Dynamic Scenes. InIEEE Conference on Computer Vision and Pattern Recognition, CVPR 2021, virtual, June 19-25, 2021, pages 10318–10327. Computer Vision Foundation / IEEE, 2021

    work page 2021

  39. [39]

    Barron, Sofien Bouaziz, Dan B

    Keunhong Park, Utkarsh Sinha, Peter Hedman, Jonathan T. Barron, Sofien Bouaziz, Dan B. Goldman, Ricardo Martin-Brualla, and Steven M. Seitz. HyperNeRF: a higher-dimensional representation for topologically varying neural radiance fields.ACM Trans. Graph., 40(6):238:1– 238:12, 2021

    work page 2021

  40. [40]

    Mip-Splatting: Alias-Free 3D Gaussian Splatting

    Zehao Yu, Anpei Chen, Binbin Huang, Torsten Sattler, and Andreas Geiger. Mip-Splatting: Alias-Free 3D Gaussian Splatting. InIEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2024, Seattle, WA, USA, June 16-22, 2024, pages 19447–19456. IEEE, 2024

    work page 2024

  41. [41]

    Scaffold- GS: Structured 3D Gaussians for View-Adaptive Rendering

    Tao Lu, Mulin Yu, Linning Xu, Yuanbo Xiangli, Limin Wang, Dahua Lin, and Bo Dai. Scaffold- GS: Structured 3D Gaussians for View-Adaptive Rendering. InIEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2024, Seattle, WA, USA, June 16-22, 2024, pages 20654–20664. IEEE, 2024

    work page 2024

  42. [42]

    Octree- GS: Towards Consistent Real-time Rendering with LOD-Structured 3D Gaussians.IEEE Transactions on Pattern Analysis and Machine Intelligence, pages 1–15, 2025

    Kerui Ren, Lihan Jiang, Tao Lu, Mulin Yu, Linning Xu, Zhangkai Ni, and Bo Dai. Octree- GS: Towards Consistent Real-time Rendering with LOD-Structured 3D Gaussians.IEEE Transactions on Pattern Analysis and Machine Intelligence, pages 1–15, 2025

    work page 2025

  43. [43]

    Pixel-GS: Density Control with Pixel-Aware Gradient for 3D Gaussian Splatting

    Zheng Zhang, Wenbo Hu, Yixing Lao, Tong He, and Hengshuang Zhao. Pixel-GS: Density Control with Pixel-Aware Gradient for 3D Gaussian Splatting. In Ales Leonardis, Elisa Ricci, Stefan Roth, Olga Russakovsky, Torsten Sattler, and Gül Varol, editors,Computer Vision - ECCV 2024 - 18th European Conference, Milan, Italy, September 29-October 4, 2024, Proceeding...

    work page 2024

  44. [44]

    AbsGS: Recovering Fine Details in 3D Gaussian Splatting

    Zongxin Ye, Wenyu Li, Sidun Liu, Peng Qiao, and Yong Dou. AbsGS: Recovering Fine Details in 3D Gaussian Splatting. InProceedings of the 32nd ACM International Conference on Multimedia, MM ’24, page 1053–1061, New York, NY , USA, 2024. Association for Computing Machinery

    work page 2024

  45. [45]

    GaussianEditor: Editing 3D Gaussians Delicately with Text Instructions

    Junjie Wang, Jiemin Fang, Xiaopeng Zhang, Lingxi Xie, and Qi Tian. GaussianEditor: Editing 3D Gaussians Delicately with Text Instructions. InIEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2024, Seattle, WA, USA, June 16-22, 2024, pages 20902–20911. IEEE, 2024

    work page 2024

  46. [46]

    GaussianEditor: Swift and Controllable 3D Editing with Gaussian Splatting

    Yiwen Chen, Zilong Chen, Chi Zhang, Feng Wang, Xiaofeng Yang, Yikai Wang, Zhongang Cai, Lei Yang, Huaping Liu, and Guosheng Lin. GaussianEditor: Swift and Controllable 3D Editing with Gaussian Splatting. InIEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2024, Seattle, WA, USA, June 16-22, 2024, pages 21476–21485. IEEE, 2024

    work page 2024

  47. [47]

    4D Gaussian Splatting for Real-Time Dynamic Scene Rendering

    Guanjun Wu, Taoran Yi, Jiemin Fang, Lingxi Xie, Xiaopeng Zhang, Wei Wei, Wenyu Liu, Qi Tian, and Xinggang Wang. 4D Gaussian Splatting for Real-Time Dynamic Scene Rendering. InIEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2024, Seattle, WA, USA, June 16-22, 2024, pages 20310–20320. IEEE, 2024

    work page 2024

  48. [48]

    Real-time Photorealistic Dynamic Scene Representation and Rendering with 4D Gaussian Splatting

    Zeyu Yang, Hongye Yang, Zijie Pan, and Li Zhang. Real-time Photorealistic Dynamic Scene Representation and Rendering with 4D Gaussian Splatting. InThe Twelfth International Conference on Learning Representations, ICLR 2024, Vienna, Austria, May 7-11, 2024. Open- Review.net, 2024

    work page 2024

  49. [49]

    DreamGaussian: Generative Gaussian Splatting for Efficient 3D Content Creation

    Jiaxiang Tang, Jiawei Ren, Hang Zhou, Ziwei Liu, and Gang Zeng. DreamGaussian: Generative Gaussian Splatting for Efficient 3D Content Creation. InThe Twelfth International Conference on Learning Representations, ICLR 2024, Vienna, Austria, May 7-11, 2024. OpenReview.net, 2024

    work page 2024

  50. [50]

    GaussianDreamer: Fast Generation from Text to 3D Gaussians by Bridging 2D and 3D Diffusion Models

    Taoran Yi, Jiemin Fang, Junjie Wang, Guanjun Wu, Lingxi Xie, Xiaopeng Zhang, Wenyu Liu, Qi Tian, and Xinggang Wang. GaussianDreamer: Fast Generation from Text to 3D Gaussians by Bridging 2D and 3D Diffusion Models. InIEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2024, Seattle, WA, USA, June 16-22, 2024, pages 6796–6807. IEEE, 2024

    work page 2024

  51. [51]

    LGM: Large Multi-view Gaussian Model for High-Resolution 3D Content Creation

    Jiaxiang Tang, Zhaoxi Chen, Xiaokang Chen, Tengfei Wang, Gang Zeng, and Ziwei Liu. LGM: Large Multi-view Gaussian Model for High-Resolution 3D Content Creation. In Ales Leonardis, Elisa Ricci, Stefan Roth, Olga Russakovsky, Torsten Sattler, and Gül Varol, editors,Computer Vision - ECCV 2024 - 18th European Conference, Milan, Italy, September 29-October 4,...

    work page 2024

  52. [52]

    GRM: Large Gaussian Reconstruction Model for Efficient 3D Reconstruction and Generation

    Yinghao Xu, Zifan Shi, Yifan Wang, Hansheng Chen, Ceyuan Yang, Sida Peng, Yujun Shen, and Gordon Wetzstein. GRM: Large Gaussian Reconstruction Model for Efficient 3D Reconstruction and Generation. In Ales Leonardis, Elisa Ricci, Stefan Roth, Olga Russakovsky, Torsten Sattler, and Gül Varol, editors,Computer Vision - ECCV 2024 - 18th European Conference, M...

    work page 2024

  53. [53]

    OpenGaussian: Towards Point-Level 3D Gaussian-based Open V ocabulary Understanding

    Yanmin Wu, Jiarui Meng, Haijie Li, Chenming Wu, Yahao Shi, Xinhua Cheng, Chen Zhao, Haocheng Feng, Errui Ding, Jingdong Wang, and Jian Zhang. OpenGaussian: Towards Point-Level 3D Gaussian-based Open V ocabulary Understanding. In Amir Globersons, Lester Mackey, Danielle Belgrave, Angela Fan, Ulrich Paquet, Jakub M. Tomczak, and Cheng Zhang, editors,Advance...

    work page 2024

  54. [54]

    LangSplat: 3D Language Gaussian Splatting

    Minghan Qin, Wanhua Li, Jiawei Zhou, Haoqian Wang, and Hanspeter Pfister. LangSplat: 3D Language Gaussian Splatting. InIEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2024, Seattle, WA, USA, June 16-22, 2024, pages 20051–20060. IEEE, 2024. 13

    work page 2024

  55. [55]

    Semantic Gaussians: Open- V ocabulary Scene Understanding with 3D Gaussian Splatting.IEEE Transactions on Circuits and Systems for Video Technology, pages 1–1, 2026

    Jun Guo, Xiaojian Ma, Yue Fan, Huaping Liu, and Qing Li. Semantic Gaussians: Open- V ocabulary Scene Understanding with 3D Gaussian Splatting.IEEE Transactions on Circuits and Systems for Video Technology, pages 1–1, 2026

    work page 2026

  56. [56]

    Segment Any 3D Gaussians

    Jiazhong Cen, Jiemin Fang, Chen Yang, Lingxi Xie, Xiaopeng Zhang, Wei Shen, and Qi Tian. Segment Any 3D Gaussians. In Toby Walsh, Julie Shah, and Zico Kolter, editors,Thirty- Ninth AAAI Conference on Artificial Intelligence, Thirty-Seventh Conference on Innovative Applications of Artificial Intelligence, Fifteenth Symposium on Educational Advances in Arti...

    work page 2025

  57. [57]

    Gaussian Grouping: Segment and Edit Anything in 3D Scenes

    Mingqiao Ye, Martin Danelljan, Fisher Yu, and Lei Ke. Gaussian Grouping: Segment and Edit Anything in 3D Scenes. In Ales Leonardis, Elisa Ricci, Stefan Roth, Olga Russakovsky, Torsten Sattler, and Gül Varol, editors,Computer Vision - ECCV 2024 - 18th European Conference, Milan, Italy, September 29-October 4, 2024, Proceedings, Part XXIX, Lecture Notes in ...

    work page 2024

  58. [58]

    Tanks and temples: bench- marking large-scale scene reconstruction.ACM Trans

    Arno Knapitsch, Jaesik Park, Qian-Yi Zhou, and Vladlen Koltun. Tanks and temples: bench- marking large-scale scene reconstruction.ACM Trans. Graph., 36(4):78:1–78:13, 2017

    work page 2017

  59. [59]

    Peter Hedman, Julien Philip, True Price, Jan-Michael Frahm, George Drettakis, and Gabriel J. Brostow. Deep blending for free-viewpoint image-based rendering.ACM Trans. Graph., 37(6):257, 2018

    work page 2018

  60. [60]

    Jiahui Zhang, Fangneng Zhan, Muyu Xu, Shijian Lu, and Eric P. Xing. FreGS: 3D Gaussian Splatting with Progressive Frequency Regularization. InIEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2024, Seattle, WA, USA, June 16-22, 2024, pages 21424–21433. IEEE, 2024

    work page 2024

  61. [61]

    3D Gaussian Splatting as Markov Chain Monte Carlo

    Shakiba Kheradmand, Daniel Rebain, Gopal Sharma, Weiwei Sun, Yang-Che Tseng, Hossam Isack, Abhishek Kar, Andrea Tagliasacchi, and Kwang Moo Yi. 3D Gaussian Splatting as Markov Chain Monte Carlo. In Amir Globersons, Lester Mackey, Danielle Belgrave, Angela Fan, Ulrich Paquet, Jakub M. Tomczak, and Cheng Zhang, editors,Advances in Neural Infor- mation Proce...

    work page 2024