pith. sign in

arxiv: 2411.18966 · v2 · submitted 2024-11-28 · 💻 cs.CV · cs.GR· cs.MM

SVGS: Enhancing Gaussian Splatting Using Primitives with Spatially Varying Colors

Rui Xu , Wenyue Chen , Jiepeng Wang , Yuan Liu , Peng Wang , Cheng Lin , Shiqing Xin , Xin Li
show 2 more authors
Wenping Wang Taku Komura
This is my paper

Pith reviewed 2026-05-23 17:16 UTC · model grok-4.3

classification 💻 cs.CV cs.GRcs.MM
keywords Gaussian splattingnovel view synthesis2D Gaussian surfelsspatially varying colorsneural rendering3D reconstructionview synthesis
0
0 comments X

The pith

Spatially varying colors and opacity in single 2D Gaussian surfels improve novel view synthesis over standard single-color 3D Gaussians.

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

The paper replaces the fixed color and opacity of each Gaussian primitive with functions that vary those values across the surface of a 2D Gaussian surfel. This change lets one primitive encode more appearance detail, which matters for scenes that combine rich textures with simple geometry because fewer primitives can then achieve the same fidelity. The authors implement the variation with bilinear interpolation, movable kernels, and tiny neural networks, then train and render using the 2D surfels. Quantitative results on multiple datasets show all three variants beat the baseline, and movable kernels give the largest gain in novel-view quality while geometry quality stays high.

Core claim

Equipping 2D Gaussian surfels with spatially varying color and opacity functions, realized through bilinear interpolation, movable kernels, or tiny neural networks, yields a more compact scene representation that outperforms standard single-color 3D Gaussian Splatting on novel-view synthesis metrics across several datasets while preserving high-quality geometric reconstruction, especially for real-world scenes that pair complex textures with relatively simple geometry.

What carries the argument

Spatially varying functions (bilinear interpolation, movable kernels, or tiny neural networks) that assign per-point color and opacity inside each 2D Gaussian surfel primitive.

If this is right

  • Real-world scenes with detailed textures and simple shapes require fewer primitives for equivalent visual quality.
  • Movable kernels deliver the strongest novel-view synthesis gains among the three tested functions.
  • Geometric reconstruction accuracy remains comparable to the baseline despite the added appearance flexibility.
  • The approach applies directly to existing Gaussian Splatting pipelines with only local changes to the primitive definition.

Where Pith is reading between the lines

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

  • The same spatially varying idea could be tested on other explicit primitives such as points or meshes to check whether the compactness benefit generalizes.
  • An adaptive scheduler that chooses the variation function per surfel based on local texture complexity might further reduce total primitive count.
  • Integration with dynamic or time-varying scenes would test whether the extra per-surfels parameters remain tractable under motion.

Load-bearing premise

That 2D surfels carrying internal color and opacity variation can capture textured appearance more compactly than many fixed-color 3D Gaussians without creating new rendering artifacts or excessive compute.

What would settle it

A controlled test on a scene whose geometry is highly non-planar where the 2D surfel model produces visible artifacts or lower PSNR than the single-color 3D Gaussian baseline.

Figures

Figures reproduced from arXiv: 2411.18966 by Cheng Lin, Jiepeng Wang, Peng Wang, Rui Xu, Shiqing Xin, Taku Komura, Wenping Wang, Wenyue Chen, Xin Li, Yuan Liu.

Figure 1
Figure 1. Figure 1: SuperGaussians gives each Gaussian the ability to vary spatially. Compared with 2DGS [ [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: 3DGS [16] uses Gaussian ellipsoids to express scenes, and a learnable color is defined on each ellipsoid. 2DGS [11] uses Gaussian surfels to express scenes, and a learnable color is defined on each Gaussian surfel. Our SuperGaussians uses spatially vary￾ing Gaussian surfels to express scenes, and the color and opacity changes with the spatial position on each surfel. 3. Method 3.1. Spatially Varying Gaussi… view at source ↗
Figure 4
Figure 4. Figure 4: The parameter amounts of the three proposed spatial [PITH_FULL_IMAGE:figures/full_fig_p004_4.png] view at source ↗
Figure 6
Figure 6. Figure 6: Visualization comparison with 2DGS [11] on the Syn￾thetic Blender [22] dataset. The blue zoom-in windows show the error map against the ground truth image. the surface reconstruction, use Chamfer Distance (CD) to measure the accuracy of geometry on the DTU [13] dataset. Comparison on Different Spatially Varying Functions. First, we present the comparison between 2DGS [11] and three of our spatially varying… view at source ↗
Figure 5
Figure 5. Figure 5: Visual comparison between three different spatially [PITH_FULL_IMAGE:figures/full_fig_p005_5.png] view at source ↗
Figure 7
Figure 7. Figure 7: Visual comparison with 2DGS [11] and 3DGS [16] on both the Mip-NeRF360 [3] dataset and the Tanks&Temples [17] dataset shows that SuperGaussians can reconstruct details better due to stronger expressiveness. terpolation also achieves good results in some of the scenes. The tiny neural network performs well when the number of Gaussian primitive is limited, demonstrating its strong rep￾resentation ability. Ho… view at source ↗
Figure 8
Figure 8. Figure 8: Visual comparison between three different spatially [PITH_FULL_IMAGE:figures/full_fig_p007_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Here we demonstrate the capabilities of three different [PITH_FULL_IMAGE:figures/full_fig_p011_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: More geometric reconstruction results on the DTU [ [PITH_FULL_IMAGE:figures/full_fig_p013_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Here we show more novel view results of SuperGaussians on the Mip-NeRF360 [ [PITH_FULL_IMAGE:figures/full_fig_p014_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: Here we show more novel view results of SuperGaussians on the first part of Synthetic Blender [ [PITH_FULL_IMAGE:figures/full_fig_p015_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: Here we show more novel view results of SuperGaussians on the second part of Synthetic Blender [ [PITH_FULL_IMAGE:figures/full_fig_p016_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: Here we show more rendering results of SuperGaussians on the first part of DTU [ [PITH_FULL_IMAGE:figures/full_fig_p017_14.png] view at source ↗
Figure 15
Figure 15. Figure 15: Here we show more rendering results of SuperGaussians on the second part of DTU [ [PITH_FULL_IMAGE:figures/full_fig_p018_15.png] view at source ↗
read the original abstract

Gaussian Splatting demonstrates impressive results in multi-view reconstruction based on Gaussian explicit representations. However, the current Gaussian primitives only have a single view-dependent color and an opacity to represent the appearance and geometry of the scene, resulting in a non-compact representation. In this paper, we introduce a new method called SVGS (Spatially Varying Gaussian Splatting) that utilizes spatially varying colors and opacity in a single Gaussian primitive to improve its representation ability. We have implemented bilinear interpolation, movable kernels, and tiny neural networks as spatially varying functions. SVGS employs 2D Gaussian surfels as primitives, which significantly enhances novel-view synthesis while maintaining high-quality geometric reconstruction. This approach is particularly effective in practical applications, as scenes combining complex textures with relatively simple geometry occur frequently in real-world environments. Quantitative and qualitative experimental results demonstrate that all three functions outperform the baseline, with the best movable kernels achieving superior novel view synthesis performance on multiple datasets, highlighting the strong potential of spatially varying functions. Project page: https://ruixu.me/html/SuperGaussians/index.html

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 / 0 minor

Summary. The paper introduces SVGS, which augments Gaussian Splatting by replacing single-color 3D Gaussians with 2D Gaussian surfels that carry spatially varying color and opacity via one of three functions (bilinear interpolation, movable kernels, or tiny neural networks). It claims that the approach yields superior novel-view synthesis on multiple datasets while preserving high-quality geometric reconstruction and enabling more compact scene representations for scenes that combine complex textures with relatively simple geometry.

Significance. If the quantitative gains hold under controlled conditions and the net parameter count is demonstrably lower than standard 3DGS at matched quality, the method would address a recognized limitation of explicit radiance fields on textured scenes. The empirical comparison of the three spatially varying functions and the shift to 2D surfels constitute the core technical contribution.

major comments (2)
  1. [Abstract] Abstract: the central motivation that SVGS yields representations 'more compactly' than single-color 3D Gaussians is load-bearing, yet the provided text contains no quantitative comparison of total primitive counts, parameter budgets, or storage sizes versus the 3DGS baseline at equivalent PSNR; each spatially varying function adds per-primitive overhead, so the net compactness claim cannot be evaluated without these data.
  2. [Abstract] Abstract and experimental sections: the assertion of quantitative and qualitative outperformance lacks any reference to baselines, error bars, dataset splits, or ablation controls on primitive count; without these, it is impossible to confirm that the reported gains are robust or that post-hoc selection has been ruled out.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments, which highlight important aspects of clarity and rigor in presenting our claims. We address each major comment below and will make the corresponding revisions to strengthen the manuscript.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the central motivation that SVGS yields representations 'more compactly' than single-color 3D Gaussians is load-bearing, yet the provided text contains no quantitative comparison of total primitive counts, parameter budgets, or storage sizes versus the 3DGS baseline at equivalent PSNR; each spatially varying function adds per-primitive overhead, so the net compactness claim cannot be evaluated without these data.

    Authors: We agree that the abstract does not contain explicit quantitative comparisons of primitive counts, parameter budgets, or storage sizes. In the revised manuscript we will add a concise quantitative summary (e.g., a short table or sentence) reporting the reduction in total primitives and overall storage relative to 3DGS at matched PSNR on the evaluated datasets. This will allow readers to assess the net compactness after accounting for the per-primitive overhead of the spatially varying functions. revision: yes

  2. Referee: [Abstract] Abstract and experimental sections: the assertion of quantitative and qualitative outperformance lacks any reference to baselines, error bars, dataset splits, or ablation controls on primitive count; without these, it is impossible to confirm that the reported gains are robust or that post-hoc selection has been ruled out.

    Authors: The experimental section already reports direct comparisons against the 3DGS baseline (and other methods) using standard metrics on established datasets. Nevertheless, we acknowledge that error bars, explicit dataset splits, and primitive-count ablations are not sufficiently highlighted. In the revision we will add error bars from repeated runs, state the train/test splits used, and include an ablation varying primitive count while holding other factors fixed, thereby addressing concerns about robustness and post-hoc selection. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical extension validated externally

full rationale

The paper introduces SVGS as an empirical method extending 3D Gaussian Splatting with 2D surfels and three spatially varying color/opacity functions (bilinear, movable kernels, tiny NNs). Claims of improved novel-view synthesis and compactness rest on quantitative results against baselines on multiple external datasets, not on any derivation, equation, or self-citation that reduces outputs to inputs by construction. No load-bearing step equates a 'prediction' to a fitted parameter or renames a known result; the central performance advantage is presented as an experimental outcome.

Axiom & Free-Parameter Ledger

2 free parameters · 1 axioms · 0 invented entities

The approach rests on the domain assumption that 2D surfels suffice for the targeted scene class and on fitted parameters inside the tiny neural networks and kernel placements; no new physical entities are postulated.

free parameters (2)
  • weights of tiny neural networks
    Weights are learned from data to realize the spatially varying function.
  • movable kernel parameters
    Kernel positions or sizes are optimized per primitive.
axioms (1)
  • domain assumption 2D Gaussian surfels suffice to represent scenes with complex textures and simple geometry
    Abstract states the method is particularly effective for such scenes.

pith-pipeline@v0.9.0 · 5750 in / 1210 out tokens · 46630 ms · 2026-05-23T17:16:59.053729+00:00 · methodology

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. 3D Skew Gaussian Splatting with Any Camera Trajectory Visualization Engine

    cs.CV 2026-05 unverdicted novelty 6.0

    3D Skew Gaussian Splatting extends standard 3D Gaussian Splatting with skew primitives, enhanced opacity, depth-aware densification, and a re-derived CUDA pipeline for a free-camera visualization engine.

  2. FACT-GS: Frequency-Aligned Complexity-Aware Texture Reparameterization for 2D Gaussian Splatting

    cs.CV 2025-11 unverdicted novelty 6.0

    FACT-GS allocates higher texture sampling density to high-frequency areas in 2D Gaussian Splatting through a learnable deformation field, recovering sharper details at the same parameter budget.

Reference graph

Works this paper leans on

39 extracted references · 39 canonical work pages · cited by 2 Pith papers · 1 internal anchor

  1. [1]

    Mip-nerf: A multiscale representation for anti-aliasing neu- ral radiance fields

    Jonathan T Barron, Ben Mildenhall, Matthew Tancik, Peter Hedman, Ricardo Martin-Brualla, and Pratul P Srinivasan. Mip-nerf: A multiscale representation for anti-aliasing neu- ral radiance fields. In Proceedings of the IEEE/CVF Inter- national Conference on Computer Vision, pages 5855–5864,

    work page

  2. [2]

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

    Jonathan T Barron, Ben Mildenhall, Dor Verbin, Pratul P Srinivasan, and Peter Hedman. Mip-nerf 360: Unbounded anti-aliased neural radiance fields. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 5470–5479, 2022. 2

    work page 2022

  3. [3]

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

    Jonathan T Barron, Ben Mildenhall, Dor Verbin, Pratul P Srinivasan, and Peter Hedman. Mip-nerf 360: Unbounded anti-aliased neural radiance fields. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 5470–5479, 2022. 2, 5, 6, 7, 11, 12, 14

    work page 2022

  4. [4]

    Zip-nerf: Anti-aliased grid-based neural radiance fields

    Jonathan T Barron, Ben Mildenhall, Dor Verbin, Pratul P Srinivasan, and Peter Hedman. Zip-nerf: Anti-aliased grid-based neural radiance fields. In Proceedings of the IEEE/CVF International Conference on Computer Vision , pages 19697–19705, 2023. 2

    work page 2023

  5. [5]

    On the mathematical properties of the structural similarity index

    Dominique Brunet, Edward R Vrscay, and Zhou Wang. On the mathematical properties of the structural similarity index. IEEE Transactions on Image Processing , 21(4):1488–1499,

    work page

  6. [6]

    pixelsplat: 3d gaussian splats from image pairs for scalable generalizable 3d reconstruction

    David Charatan, Sizhe Lester Li, Andrea Tagliasacchi, and Vincent Sitzmann. pixelsplat: 3d gaussian splats from image pairs for scalable generalizable 3d reconstruction. In Pro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 19457–19467, 2024. 2

    work page 2024

  7. [7]

    High-quality surface re- construction using gaussian surfels

    Pinxuan Dai, Jiamin Xu, Wenxiang Xie, Xinguo Liu, Huamin Wang, and Weiwei Xu. High-quality surface re- construction using gaussian surfels. In SIGGRAPH 2024 Conference Papers. Association for Computing Machinery,

    work page 2024

  8. [8]

    Accurate, dense, and robust multiview stereopsis

    Yasutaka Furukawa and Jean Ponce. Accurate, dense, and robust multiview stereopsis. IEEE Transactions on Pattern Analysis and Machine Intelligence , 32(8):1362–1376, 2010. 2

    work page 2010

  9. [9]

    Sugar: Surface- aligned gaussian splatting for efficient 3d mesh reconstruc- tion and high-quality mesh rendering

    Antoine Gu ´edon and Vincent Lepetit. Sugar: Surface- aligned gaussian splatting for efficient 3d mesh reconstruc- tion and high-quality mesh rendering. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 5354–5363, 2024. 2, 12

    work page 2024

  10. [10]

    Ges: Generalized exponential splatting for efficient radiance 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 radiance field rendering. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 19812–19822, 2024. 2

    work page 2024

  11. [11]

    2d gaussian splatting for geometrically ac- curate radiance fields

    Binbin Huang, Zehao Yu, Anpei Chen, Andreas Geiger, and Shenghua Gao. 2d gaussian splatting for geometrically ac- curate radiance fields. In ACM SIGGRAPH 2024 Conference Papers, pages 1–11, 2024. 1, 2, 3, 4, 5, 6, 7, 8, 11, 12, 13

    work page 2024

  12. [12]

    Nerf-texture: Texture synthesis with neural radi- ance fields

    Yi-Hua Huang, Yan-Pei Cao, Yu-Kun Lai, Ying Shan, and Lin Gao. Nerf-texture: Texture synthesis with neural radi- ance fields. In ACM SIGGRAPH 2023 Conference Proceed- ings, pages 1–10, 2023. 2

    work page 2023

  13. [13]

    Large scale multi-view stereopsis eval- uation

    Rasmus Jensen, Anders Dahl, George V ogiatzis, Engin Tola, and Henrik Aanæs. Large scale multi-view stereopsis eval- uation. In Proceedings of the IEEE conference on computer vision and pattern recognition , pages 406–413, 2014. 2, 5, 7, 11, 12, 13, 17, 18

    work page 2014

  14. [14]

    Neggs: Negative gaussian splatting

    Artur Kasymov, Bartosz Czekaj, Marcin Mazur, and Prze- mysław Spurek. Neggs: Negative gaussian splatting. arXiv preprint arXiv:2405.18163, 2024. 3

    work page arXiv 2024

  15. [15]

    Poisson surface reconstruction

    Michael Kazhdan, Matthew Bolitho, and Hugues Hoppe. Poisson surface reconstruction. In Proceedings of the fourth Eurographics symposium on Geometry processing, 2006. 2

    work page 2006

  16. [16]

    3d gaussian splatting for real-time radiance field rendering

    Bernhard Kerbl, Georgios Kopanas, Thomas Leimk ¨uhler, and George Drettakis. 3d gaussian splatting for real-time radiance field rendering. ACM Trans. Graph., 42(4):139–1,

    work page

  17. [17]

    1, 2, 3, 4, 5, 6, 7, 8, 11, 12

    work page

  18. [18]

    Tanks and temples: Benchmarking large-scale scene reconstruction

    Arno Knapitsch, Jaesik Park, Qian-Yi Zhou, and Vladlen Koltun. Tanks and temples: Benchmarking large-scale scene reconstruction. ACM Transactions on Graphics (ToG) , 36 (4):1–13, 2017. 2, 5, 6, 11

    work page 2017

  19. [19]

    3d-hgs: 3d half-gaussian splatting

    Haolin Li, Jinyang Liu, Mario Sznaier, and Octavia Camps. 3d-hgs: 3d half-gaussian splatting. arXiv preprint arXiv:2406.02720, 2024. 2, 3

    work page arXiv 2024

  20. [20]

    Sur- face reconstruction from point clouds without normals by parametrizing the gauss formula

    Siyou Lin, Dong Xiao, Zuoqiang Shi, and Bin Wang. Sur- face reconstruction from point clouds without normals by parametrizing the gauss formula. ACM Transactions on Graphics, 42(2):1–19, 2022. 2

    work page 2022

  21. [21]

    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. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 20654–20664, 2024. 2

    work page 2024

  22. [22]

    P-mvsnet: Learning patch-wise matching confidence aggregation for multi-view stereo

    Keyang Luo, Tao Guan, Lili Ju, Haipeng Huang, and Yawei Luo. P-mvsnet: Learning patch-wise matching confidence aggregation for multi-view stereo. In Proceedings of the IEEE/CVF International Conference on Computer Vision , pages 10452–10461, 2019. 2

    work page 2019

  23. [23]

    Nerf: Representing scenes as neural radiance fields for view syn- thesis

    Ben Mildenhall, Pratul P Srinivasan, Matthew Tancik, Jonathan T Barron, Ravi Ramamoorthi, and Ren Ng. Nerf: Representing scenes as neural radiance fields for view syn- thesis. Communications of the ACM, 65(1):99–106, 2021. 1, 2, 5, 7, 8, 11, 12, 15, 16

    work page 2021

  24. [24]

    Polyfit: Polygonal surface reconstruction from point clouds

    Liangliang Nan and Peter Wonka. Polyfit: Polygonal surface reconstruction from point clouds. InProceedings of the IEEE International Conference on Computer Vision , pages 2353– 2361, 2017. 2

    work page 2017

  25. [25]

    Openmvs: Open multi-view stereo reconstruc- tion library

    OpenMVS. Openmvs: Open multi-view stereo reconstruc- tion library. 2

    work page

  26. [26]

    Structure- from-motion revisited

    Johannes L Schonberger and Jan-Michael Frahm. Structure- from-motion revisited. In Proceedings of the IEEE con- ference on computer vision and pattern recognition , pages 4104–4113, 2016. 2

    work page 2016

  27. [27]

    Pixelwise view selection for un- structured multi-view stereo

    Johannes Lutz Sch ¨onberger, Enliang Zheng, Marc Pollefeys, and Jan-Michael Frahm. Pixelwise view selection for un- structured multi-view stereo. In European Conference on Computer Vision (ECCV), 2016. 2 9

    work page 2016

  28. [28]

    A comparison and evalua- tion of multi-view stereo reconstruction algorithms

    Steven M Seitz, Brian Curless, James Diebel, Daniel Scharstein, and Richard Szeliski. A comparison and evalua- tion of multi-view stereo reconstruction algorithms. In 2006 IEEE computer society conference on computer vision and pattern recognition (CVPR’06), pages 519–528. IEEE, 2006. 2

    work page 2006

  29. [29]

    NeuS: Learning Neural Implicit Surfaces by Volume Rendering for Multi-view Reconstruction

    Peng Wang, Lingjie Liu, Yuan Liu, Christian Theobalt, Taku Komura, and Wenping Wang. Neus: Learning neural implicit surfaces by volume rendering for multi-view reconstruction. arXiv preprint arXiv:2106.10689, 2021. 2, 11

    work page internal anchor Pith review Pith/arXiv arXiv 2021

  30. [30]

    Rfeps: Re- constructing feature-line equipped polygonal surface

    Rui Xu, Zixiong Wang, Zhiyang Dou, Chen Zong, Shiqing Xin, Mingyan Jiang, Tao Ju, and Changhe Tu. Rfeps: Re- constructing feature-line equipped polygonal surface. ACM Transactions on Graphics (TOG), 41(6):1–15, 2022. 2

    work page 2022

  31. [31]

    Globally consistent normal orientation for point clouds by regularizing the winding-number field.ACM Transactions on Graphics (TOG), 42(4):1–15, 2023

    Rui Xu, Zhiyang Dou, Ningna Wang, Shiqing Xin, Shuang- min Chen, Mingyan Jiang, Xiaohu Guo, Wenping Wang, and Changhe Tu. Globally consistent normal orientation for point clouds by regularizing the winding-number field.ACM Transactions on Graphics (TOG), 42(4):1–15, 2023. 2

    work page 2023

  32. [32]

    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 Proceedings of the European conference on computer vi- sion (ECCV), pages 767–783, 2018. 2

    work page 2018

  33. [33]

    Recurrent mvsnet for high-resolution multi-view stereo depth inference

    Yao Yao, Zixin Luo, Shiwei Li, Tianwei Shen, Tian Fang, and Long Quan. Recurrent mvsnet for high-resolution multi-view stereo depth inference. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 5525–5534, 2019. 2

    work page 2019

  34. [34]

    V ol- ume rendering of neural implicit surfaces

    Lior Yariv, Jiatao Gu, Yoni Kasten, and Yaron Lipman. V ol- ume rendering of neural implicit surfaces. Advances in Neu- ral Information Processing Systems , 34:4805–4815, 2021. 11

    work page 2021

  35. [35]

    Fast-mvsnet: Sparse-to- dense multi-view stereo with learned propagation and gauss- newton refinement

    Zehao Yu and Shenghua Gao. Fast-mvsnet: Sparse-to- dense multi-view stereo with learned propagation and gauss- newton refinement. In Proceedings of the IEEE/CVF con- ference on computer vision and pattern recognition , pages 1949–1958, 2020. 2

    work page 1949

  36. [36]

    Mip-splatting: Alias-free 3d gaussian splat- ting

    Zehao Yu, Anpei Chen, Binbin Huang, Torsten Sattler, and Andreas Geiger. Mip-splatting: Alias-free 3d gaussian splat- ting. In Proceedings of the IEEE/CVF Conference on Com- puter Vision and Pattern Recognition , pages 19447–19456,

    work page

  37. [37]

    Vis-mvsnet: Visibility-aware multi-view stereo net- work

    Jingyang Zhang, Shiwei Li, Zixin Luo, Tian Fang, and Yao Yao. Vis-mvsnet: Visibility-aware multi-view stereo net- work. International Journal of Computer Vision , 131(1): 199–214, 2023. 2

    work page 2023

  38. [38]

    The unreasonable effectiveness of deep features as a perceptual metric

    Richard Zhang, Phillip Isola, Alexei A Efros, Eli Shechtman, and Oliver Wang. The unreasonable effectiveness of deep features as a perceptual metric. In CVPR, 2018. 5

    work page 2018

  39. [39]

    images 4

    Matthias Zwicker, Hanspeter Pfister, Jeroen Van Baar, and Markus Gross. Ewa volume splatting. In Proceedings Visu- alization, 2001. VIS’01., pages 29–538. IEEE, 2001. 3 10 A. Implementation Details Following 2DGS [11] and 3DGS [16], we tested the Syn- thetic Blender dataset [22] and Tanks&Temples [17] at their native resolution. We tested the DTU [13] dat...

    work page arXiv 2001