pith. sign in

arxiv: 2605.18334 · v1 · pith:H6FLL6QUnew · submitted 2026-05-18 · 💻 cs.CV · cs.GR

3D Skew Gaussian Splatting with Any Camera Trajectory Visualization Engine

Beizhen Zhao , Yifan Zhou , Gaochao Song , Ziran Yin , Hao Wang This is my paper

Pith reviewed 2026-05-20 11:20 UTC · model grok-4.3

classification 💻 cs.CV cs.GR
keywords 3D Gaussian SplattingSkew GaussianReal-time RenderingView SynthesisScene RepresentationInteractive VisualizationCUDA RasterizationCamera Trajectory
0
0 comments X

The pith

Skew Gaussian primitives replace symmetric ones to cut blurriness and redundancy in real-time 3D scene rendering.

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

The paper sets out to show that the symmetric Gaussian kernels used in standard 3D Gaussian Splatting create visual artifacts at shape and color boundaries, producing both blurriness and unnecessary extra primitives that complicate spatial data exploration. It introduces skew Gaussian primitives that keep the fast rasterization of the original method but add built-in asymmetry to model those boundaries more accurately. The work pairs this change with an improved opacity model and a depth-aware way to add or remove primitives, then rebuilds the underlying CUDA code so the same engine can handle both symmetric and skew forms under any camera path. If correct, the result is higher visual fidelity and tighter scene descriptions that still run at interactive speeds for free exploration of detailed environments.

Core claim

The central claim is that extending the 3D Gaussian primitive to a general skew Gaussian form supplies intrinsic asymmetric modeling power while retaining the efficient rasterization behavior of the symmetric case. This is combined with an enhanced opacity representation and a depth-aware densification strategy, all re-implemented in a CUDA pipeline that supports arbitrary camera trajectories inside a decoupled interactive visualization engine.

What carries the argument

The generalized skew Gaussian primitive, which extends the standard symmetric Gaussian to handle asymmetric shape and color distributions while keeping the same fast rasterization pipeline.

If this is right

  • Shape and color discontinuities are captured with less blurring and fewer redundant primitives.
  • Scene representations become structurally more compact while preserving visual quality in intricate regions.
  • Complex transparency is handled more effectively through the enhanced opacity model.
  • The same rasterizer supports both symmetric and skew forms under any camera trajectory without code changes.
  • Real-time frame rates remain available for fluid interactive visual exploration of the rendered scenes.

Where Pith is reading between the lines

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

  • The same primitive change could be tested on dynamic or time-varying scenes to see whether the densification strategy scales.
  • Integration with existing 3D reconstruction pipelines might reduce memory footprints when exporting compact models for downstream analysis.
  • The asymmetric modeling may help in applications that require precise boundary localization, such as measuring object dimensions from rendered views.

Load-bearing premise

The skew Gaussian can be rasterized at the same speed and efficiency as the original symmetric Gaussian despite its added flexibility.

What would settle it

Running the method on a benchmark scene with fine geometric or color edges and finding either no measurable drop in blurriness or primitive count or a drop in frame rate below real-time interactive levels would falsify the central performance claim.

Figures

Figures reproduced from arXiv: 2605.18334 by Beizhen Zhao, Gaochao Song, Hao Wang, Yifan Zhou, Ziran Yin.

Figure 1
Figure 1. Figure 1: Comparison of Skew Normal and Normal Kernels in Fitting a Square Wave. (a): Distribution shapes with different skew parameters in the Skew [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: The architectural pipeline of our 3D Skew Gaussian Splatting (3DSGS) framework. We fundamentally extend standard symmetric 3D Gaussians into 3D [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: The left sub-figure shows how the enhanced opacity parameter affects [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Comparison Results. Visual differences are highlighted with red insets for better clarity. Our approach consistently outperforms other models on Mip-NeRF 360 [35] dataset, demonstrating clear advantages in challenging scenarios such as thin geometries and fine-scale details. Best viewed in color [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Interactive Exploration and View Synthesis Quality. Top row: The unconstrained free camera trajectory and the 3DSGS point cloud proxy visualized within our Blender frontend. Bottom row: The corresponding novel views rendered by our CUDA rasterizer at the selected poses. Data Splitting Protocol: To ensure a fair and consistent eval￾uation, we strictly adhere to the standard protocol established in prior vol… view at source ↗
Figure 6
Figure 6. Figure 6: Comparison Results. Visual differences are highlighted with red insets for better clarity. Our approach consistently outperforms other models on Mip-NeRF 360 [35] dataset, demonstrating advantages in challenging scenarios. Best viewed in color. plementary Materials. Baselines. Given the extensive volume of research related to 3DGS, we select the original 3DGS [6] alongside the most recent advancements that… view at source ↗
Figure 7
Figure 7. Figure 7: This figure illustrates the fitting error analysis between different rendering models and the ground truth. (a) represents the ground truth, while subfigures [PITH_FULL_IMAGE:figures/full_fig_p010_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Ablation Results. We design the ablation study about the performance of our methods with 3DGS under different optimization iteration settings across all three datasets. Our method can outperform 3DGS consistently. TABLE III ABLATION EVALUATION ON THE MIP-NERF 360 [35] DATASET. Method PSNR↑ SSIM↑ LPIPS↓ w/o opacity regularization 29.53 0.881 0.159 w/o densification optimization 29.47 0.878 0.166 Ours 29.78 … view at source ↗
Figure 9
Figure 9. Figure 9: Comparison Results. Visual differences are highlighted with red insets for better clarity. Our approach consistently outperforms other models on Tanks & Temples [36] and Deep Blending [37] dataset, demonstrating advantages in challenging scenarios such as thin geometries and fine-scale details. Best viewed in color [PITH_FULL_IMAGE:figures/full_fig_p016_9.png] view at source ↗
read the original abstract

While 3D Gaussian Splatting (3DGS) has revolutionized real-time photorealistic view synthesis, its fundamental reliance on symmetric Gaussian distributions introduces visual artifacts that hinder accurate spatial data exploration. Specifically, symmetric kernels struggle to capture shape and color discontinuities , which cause blurriness and primitive redundancy that mislead human perception during visual analysis. To address these visualization barriers, we introduce 3D Skew Gaussian Splatting (3DSGS), a novel framework that significantly enhances the structural fidelity and compactness of explicit scene representations. Our key insight lies in extending the standard primitive to a general Skew Gaussian counterpart. This generalized primitive inherits the highly efficient rasterization properties of standard Gaussians while gaining intrinsic asymmetric modeling capabilities. We couple this with an enhanced opacity representation to better handle complex transparency, alongside a depth-aware densification strategy that intelligently manages primitive allocation. Furthermore, to make these advancements actionable for real-world visual analytics, we re-derive the CUDA rasterization pipeline to universally support both symmetric and skew Gaussians, integrating it into a decoupled, free-camera interactive visualization engine. Extensive experiments demonstrate that 3DSGS achieves superior rendering quality and structural compactness, particularly in regions with intricate details, while maintaining the real-time frame rates necessary for fluid interactive exploration. Supplementary derivations and visual results are available at \textbf{\textit{https://3d-skew-gs.github.io/}}.

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

1 major / 1 minor

Summary. The manuscript introduces 3D Skew Gaussian Splatting (3DSGS) as an extension of 3D Gaussian Splatting to mitigate artifacts from symmetric kernels, such as blurriness at discontinuities. The core contribution is a generalized skew Gaussian primitive that purportedly retains efficient rasterization while enabling asymmetric modeling, paired with enhanced opacity handling, depth-aware densification, and a re-derived CUDA rasterization pipeline integrated into a decoupled visualization engine supporting arbitrary camera trajectories. The authors assert through experiments that the method yields superior rendering quality and structural compactness in intricate regions while preserving real-time frame rates.

Significance. If the efficiency and quality claims hold, the work would advance explicit scene representations in computer vision by improving fidelity for asymmetric structures without sacrificing interactivity, with direct utility for visual analytics and free-camera exploration. The re-derivation of the rasterization pipeline to handle both symmetric and skew primitives represents a concrete technical step forward if it avoids hidden per-pixel costs.

major comments (1)
  1. [Abstract] Abstract: The central assertion that the skew Gaussian primitive 'inherits the highly efficient rasterization properties of standard Gaussians' while adding asymmetry is load-bearing for the real-time performance guarantee, yet the abstract (and by extension the manuscript) provides no derivation, complexity analysis, or pseudocode showing that the projected density remains evaluable in closed form or O(1) per primitive; standard 3DGS relies on the analytic 2D Gaussian projection for tile sorting and alpha compositing, and skew extensions typically require numerical methods that risk violating this.
minor comments (1)
  1. [Abstract] The supplementary derivations URL is referenced but should include a brief summary of the key projection equations in the main text to allow readers to assess the rasterization claim without external access.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for their careful reading and constructive feedback on our work. The single major comment raises a valid point about clarity in the abstract and supporting details for the efficiency claims. We address this directly below and are happy to revise the manuscript accordingly.

read point-by-point responses

  1. Referee: [Abstract] Abstract: The central assertion that the skew Gaussian primitive 'inherits the highly efficient rasterization properties of standard Gaussians' while adding asymmetry is load-bearing for the real-time performance guarantee, yet the abstract (and by extension the manuscript) provides no derivation, complexity analysis, or pseudocode showing that the projected density remains evaluable in closed form or O(1) per primitive; standard 3DGS relies on the analytic 2D Gaussian projection for tile sorting and alpha compositing, and skew extensions typically require numerical methods that risk violating this.

    Authors: We agree that the abstract would benefit from greater precision on this point. The full manuscript (Section 3.2 and Appendix A) contains the re-derivation of the 2D projection for skew Gaussians, showing that the density remains evaluable in closed form via a modified covariance and skew-adjusted mean after perspective projection; this preserves the same O(1) per-primitive cost for tile sorting and alpha blending as standard 3DGS. The CUDA kernel implements this analytically without numerical quadrature or iterative solvers. We will revise the abstract to explicitly reference the closed-form projection and add a short complexity statement. Pseudocode for the rasterizer will be moved from the supplement into the main text for visibility. revision: yes

Circularity Check

0 steps flagged

No circularity: derivation is a direct extension with re-derived pipeline

full rationale

The paper introduces 3DSGS by extending the standard 3D Gaussian primitive with skew parameters for asymmetry, coupled with an enhanced opacity model and depth-aware densification. It explicitly states that the CUDA rasterization pipeline is re-derived to support both symmetric and skew Gaussians, presented as an independent implementation step rather than a reduction to prior fitted values or self-referential definitions. No equations or claims reduce by construction to inputs (e.g., no fitted parameters renamed as predictions, no load-bearing self-citations, and no ansatz smuggled via prior work). The central claims of improved fidelity and real-time performance rest on the new primitive and pipeline re-derivation, which are self-contained against external 3DGS benchmarks without circular reduction.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 1 invented entities

The central claim rests on the unverified assumption that skew Gaussians can be rasterized as efficiently as symmetric ones; no free parameters or additional axioms are detailed in the abstract.

invented entities (1)
  • Skew Gaussian primitive no independent evidence
    purpose: To capture shape and color discontinuities with asymmetric modeling
    Presented as a generalized extension of standard Gaussians but without independent evidence or external validation in the abstract.

pith-pipeline@v0.9.0 · 5786 in / 1356 out tokens · 46740 ms · 2026-05-20T11:20:42.420009+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

44 extracted references · 44 canonical work pages · 1 internal anchor

  1. [1]

    Time- multiplexed neural holography: a flexible framework for holographic near-eye displays with fast heavily-quantized spatial light modulators,

    S. Choi, M. Gopakumar, Y . Peng, J. Kim, O’Tooleet al., “Time- multiplexed neural holography: a flexible framework for holographic near-eye displays with fast heavily-quantized spatial light modulators,” inACM SIGGRAPH 2022 Conference Proceedings, 2022, pp. 1–9

    work page 2022

  2. [2]

    Bundlesdf: Neural 6-dof tracking and 3d reconstruction of unknown objects,

    B. Wen, J. Tremblay, V . Blukis, S. Tyree, T. M ¨uller, A. Evans, D. Fox et al., “Bundlesdf: Neural 6-dof tracking and 3d reconstruction of unknown objects,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023, pp. 606–617

    work page 2023

  3. [3]

    Sparseocc: Rethink- ing sparse latent representation for vision-based semantic occupancy prediction,

    P. Tang, Z. Wang, G. Wang, J. Zheng, X. Renet al., “Sparseocc: Rethink- ing sparse latent representation for vision-based semantic occupancy prediction,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024, pp. 15 035–15 044

    work page 2024

  4. [4]

    Look at the sky: Sky-aware efficient 3d gaussian splatting in the wild,

    Y . Wang, J. Wang, R. Gao, Y . Qu, W. Duan, S. Yang, and Y . Qi, “Look at the sky: Sky-aware efficient 3d gaussian splatting in the wild,”IEEE Transactions on Visualization and Computer Graphics, 2025

    work page 2025

  5. [5]

    Splatloc: 3d gaussian splatting-based visual localization for augmented reality,

    H. Zhai, X. Zhang, B. Zhao, H. Li, Y . He, Z. Cuiet al., “Splatloc: 3d gaussian splatting-based visual localization for augmented reality,”IEEE Transactions on Visualization and Computer Graphics, 2025

    work page 2025

  6. [6]

    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 Trans. Graph., vol. 42, no. 4, pp. 139–1, 2023

    work page 2023

  7. [7]

    3d gaussian splatting as markov chain monte carlo,

    S. Kheradmand, D. Rebain, G. Sharma, W. Sun, Y .-C. Tseng, H. Isack, A. Kar, A. Tagliasacchi, and K. M. Yi, “3d gaussian splatting as markov chain monte carlo,”Advances in Neural Information Processing Systems, vol. 37, pp. 80 965–80 986, 2024

    work page 2024

  8. [8]

    Wildgaussians: 3d gaussian splatting in the wild,

    J. Kulhanek, S. Peng, Z. Kukelova, M. Pollefeys, and T. Sattler, “Wildgaussians: 3d gaussian splatting in the wild,”arXiv preprint arXiv:2407.08447, 2024

    work page arXiv 2024

  9. [9]

    arXiv preprint arXiv:2405.20693

    R. Zha, T. J. Lin, Y . Cai, J. Cao, Y . Zhang, and H. Li, “R 2-gaussian: Rectifying radiative gaussian splatting for tomographic reconstruction,” arXiv preprint arXiv:2405.20693, 2024

    work page arXiv 2024

  10. [10]

    Gaus- sianshader: 3d gaussian splatting with shading functions for reflective surfaces,

    Y . Jiang, J. Tu, Y . Liu, X. Gao, X. Long, W. Wang, and Y . Ma, “Gaus- sianshader: 3d gaussian splatting with shading functions for reflective surfaces,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024, pp. 5322–5332

    work page 2024

  11. [11]

    3d gaussian splatting with deferred reflection,

    K. Ye, Q. Hou, and K. Zhou, “3d gaussian splatting with deferred reflection,” inACM SIGGRAPH 2024 Conference Papers, 2024, pp. 1– 10

    work page 2024

  12. [12]

    Spec-gaussian: Anisotropic view-dependent appearance for 3d gaussian splatting,

    Z. Yang, X. Gao, Y .-T. Sun, Y . Huang, X. Lyu, W. Zhou, S. Jiao, X. Qi, and X. Jin, “Spec-gaussian: Anisotropic view-dependent appearance for 3d gaussian splatting,”Advances in Neural Information Processing Systems, vol. 37, pp. 61 192–61 216, 2024

    work page 2024

  13. [13]

    Human hair reconstruction with strand-aligned 3d gaussians,

    E. Zakharov, V . Sklyarova, M. Black, G. Nam, J. Thies, and O. Hilliges, “Human hair reconstruction with strand-aligned 3d gaussians,” inEuro- pean Conference on Computer Vision. Springer, 2024, pp. 409–425

    work page 2024

  14. [14]

    Beyond gaussians: Fast and high-fidelity 3d splatting with linear kernels,

    H. Chen, R. Chen, Q. Qu, Z. Wang, T. Liu, X. Chen, and Y . Y . Chung, “Beyond gaussians: Fast and high-fidelity 3d splatting with linear kernels,”arXiv preprint arXiv:2411.12440, 2024

    work page arXiv 2024

  15. [15]

    Revising densification in gaussian splatting,

    S. Rota Bul `o, L. Porzi, and P. Kontschieder, “Revising densification in gaussian splatting,” inEuropean Conference on Computer Vision. Springer, 2024, pp. 347–362

    work page 2024

  16. [16]

    SVGS: Enhancing Gaussian Splatting Using Primitives with Spatially Varying Colors

    R. Xu, W. Chen, J. Wang, Y . Liu, P. Wang, L. Gao, S. Xin, T. Ko- mura, X. Li, and W. Wang, “Supergaussians: Enhancing gaussian splatting using primitives with spatially varying colors,”arXiv preprint arXiv:2411.18966, 2024

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  17. [17]

    Effective rank analysis and regularization for enhanced 3d gaussian splatting,

    J. Hyung, S. Hong, S. Hwang, J. Lee, J. Choo, and J.-H. Kim, “Effective rank analysis and regularization for enhanced 3d gaussian splatting,” arXiv preprint arXiv:2406.11672, 2024

    work page arXiv 2024

  18. [18]

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

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

    work page 2021

  19. [19]

    Mip-nerf: A multiscale representation for anti- aliasing neural radiance fields,

    J. T. Barron, B. Mildenhall, M. Tancik, P. Hedman, R. Martin-Brualla, and P. P. Srinivasan, “Mip-nerf: A multiscale representation for anti- aliasing neural radiance fields,” inProceedings of the IEEE/CVF inter- national conference on computer vision, 2021, pp. 5855–5864

    work page 2021

  20. [20]

    Instant neural graphics primitives with a multiresolution hash encoding,

    T. M ¨uller, A. Evans, C. Schied, and A. Keller, “Instant neural graphics primitives with a multiresolution hash encoding,”ACM transactions on graphics (TOG), vol. 41, no. 4, pp. 1–15, 2022

    work page 2022

  21. [21]

    ADOP: Approximate differentiable one-pixel point rendering,

    D. R ¨uckert, L. Franke, and M. Stamminger, “ADOP: Approximate differentiable one-pixel point rendering,”ACM Transactions on Graphics (TOG), vol. 41, no. 4, pp. 1–14, 2022

    work page 2022

  22. [22]

    Deformable 3d gaussians for high-fidelity monocular dynamic scene reconstruction,

    Z. Yang, X. Gao, W. Zhou, S. Jiao, Y . Zhang, and X. Jin, “Deformable 3d gaussians for high-fidelity monocular dynamic scene reconstruction,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2024, pp. 20 331–20 341

    work page 2024

  23. [23]

    Luciddreamer: Domain-free gen- eration of 3d gaussian splatting scenes

    J. Chung, S. Lee, H. Nam, J. Lee, and K. M. Lee, “Luciddreamer: Domain-free generation of 3d gaussian splatting scenes,”arXiv preprint arXiv:2311.13384, 2023

    work page arXiv 2023

  24. [24]

    Wild-gs: Real-time novel view synthesis from unconstrained photo collections,

    J. Xu, Y . Mei, and V . Patel, “Wild-gs: Real-time novel view synthesis from unconstrained photo collections,”Advances in Neural Information Processing Systems, vol. 37, pp. 103 334–103 355, 2024

    work page 2024

  25. [25]

    Dreamscene360: Unconstrained text-to- 3d scene generation with panoramic gaussian splatting,

    S. Zhou, Z. Fan, D. Xu, H. Chang, P. Chari, T. Bharadwaj, S. You, Z. Wang, and A. Kadambi, “Dreamscene360: Unconstrained text-to- 3d scene generation with panoramic gaussian splatting,” inEuropean Conference on Computer Vision. Springer, 2024, pp. 324–342

    work page 2024

  26. [26]

    3dgs-det: Empower 3d gaussian splatting with boundary guidance and box-focused sampling for 3d object detection,

    Y . Cao, Y . Jv, and D. Xu, “3dgs-det: Empower 3d gaussian splatting with boundary guidance and box-focused sampling for 3d object detection,” arXiv preprint arXiv:2410.01647, 2024

    work page arXiv 2024

  27. [27]

    F-3dgs: Factorized coordinates and representations for 3d gaussian splatting,

    X. Sun, J. C. Lee, D. Rho, J. H. Ko, U. Ali, and E. Park, “F-3dgs: Factorized coordinates and representations for 3d gaussian splatting,” in Proceedings of the 32nd ACM International Conference on Multimedia, 2024, pp. 7957–7965

    work page 2024

  28. [28]

    Compact 3d scene representation via self-organizing gaussian grids,

    W. Morgenstern, F. Barthel, A. Hilsmann, and P. Eisert, “Compact 3d scene representation via self-organizing gaussian grids,” inEuropean Conference on Computer Vision. Springer, 2024, pp. 18–34

    work page 2024

  29. [29]

    3d student splatting and scooping,

    J. Zhu, J. Yue, F. He, and H. Wang, “3d student splatting and scooping,” arXiv preprint arXiv:2503.10148, 2025

    work page arXiv 2025

  30. [30]

    2D Gaussian Splat- ting for geometrically accurate radiance fields,

    B. Huang, Z. Yu, A. Chen, A. Geiger, and S. Gao, “2D Gaussian Splat- ting for geometrically accurate radiance fields,” inACM SIGGRAPH 2024 Conference Papers, 2024, pp. 1–11

    work page 2024

  31. [31]

    3d-hgs: 3d half-gaussian splatting

    H. Li, J. Liu, M. Sznaier, and O. Camps, “3d-hgs: 3d half-gaussian splatting,”arXiv preprint arXiv:2406.02720, 2024

    work page arXiv 2024

  32. [32]

    Deformable radial kernel splatting,

    Y .-H. Huang, M.-X. Lin, Y .-T. Sun, Z. Yang, X. Lyu, Y .-P. Cao, and X. Qi, “Deformable radial kernel splatting,”arXiv preprint arXiv:2412.11752, 2024

    work page arXiv 2024

  33. [33]

    Mip-splatting: Alias-free 3d gaussian splatting,

    Z. Yu, A. Chen, B. Huang, T. Sattler, and A. Geiger, “Mip-splatting: Alias-free 3d gaussian splatting,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024, pp. 19 447–19 456

    work page 2024

  34. [34]

    3d convex splatting: Radiance field rendering with 3d smooth convexes,

    J. Held, R. Vandeghen, A. Hamdi, A. Deliege, A. Cioppa, S. Giancola, A. Vedaldi, B. Ghanem, and M. Van Droogenbroeck, “3d convex splatting: Radiance field rendering with 3d smooth convexes,”arXiv preprint arXiv:2411.14974, 2024

    work page arXiv 2024

  35. [35]

    Mip-nerf 360: Unbounded anti-aliased neural radiance fields,

    J. T. Barron, B. Mildenhall, D. Verbin, P. P. Srinivasan, and P. Hedman, “Mip-nerf 360: Unbounded anti-aliased neural radiance fields,” inPro- ceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2022, pp. 5470–5479

    work page 2022

  36. [36]

    Tanks and temples: Benchmarking large-scale scene reconstruction,

    A. Knapitsch, J. Park, Q.-Y . Zhou, and V . Koltun, “Tanks and temples: Benchmarking large-scale scene reconstruction,”ACM Transactions on Graphics, vol. 36, no. 4, 2017

    work page 2017

  37. [37]

    Deep blending for free-viewpoint image-based rendering,

    P. Hedman, J. Philip, T. Price, J.-M. Frahm, G. Drettakis, and G. Bros- tow, “Deep blending for free-viewpoint image-based rendering,”ACM Transactions on Graphics (ToG), vol. 37, no. 6, pp. 1–15, 2018

    work page 2018

  38. [38]

    Ges: Generalized exponential splatting for efficient radiance field rendering,

    A. Hamdi, L. Melas-Kyriazi, J. Mai, G. Qian, R. Liu, C. V ondrick, B. Ghanem, and A. Vedaldi, “Ges: Generalized exponential splatting for efficient radiance field rendering,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024, pp. 19 812–19 822

    work page 2024

  39. [39]

    2d gaussian splatting for geometrically accurate radiance fields,

    B. Huang, Z. Yu, A. Chen, A. Geiger, and S. Gao, “2d gaussian splatting for geometrically accurate radiance fields,” inSIGGRAPH 2024 Conference Papers. Association for Computing Machinery, 2024

    work page 2024

  40. [40]

    Fregs: 3d gaussian splatting with progressive frequency regularization,

    J. Zhang, F. Zhan, M. Xu, S. Lu, and E. Xing, “Fregs: 3d gaussian splatting with progressive frequency regularization,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024, pp. 21 424–21 433

    work page 2024

  41. [41]

    Scaffold-gs: Structured 3d gaussians for view-adaptive rendering,

    T. Lu, M. Yu, L. Xu, Y . Xiangli, L. Wang, D. Lin, and B. Dai, “Scaffold-gs: Structured 3d gaussians for view-adaptive rendering,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024, pp. 20 654–20 664

    work page 2024

  42. [42]

    Mip-splatting: Alias-free 3d gaussian splatting,

    Z. Yu, A. Chen, B. Huang, T. Sattler, and A. Geiger, “Mip-splatting: Alias-free 3d gaussian splatting,” inProceedings of the IEEE/CVF JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2021 13 Conference on Computer Vision and Pattern Recognition (CVPR), June 2024, pp. 19 447–19 456

    work page 2021

  43. [43]

    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

    work page 2004

  44. [44]

    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. APPENDIX This work adheres to the IEEE Code of Ethics. In this study, no human subjects or animal experimentation was invo...

    work page 2018