pith. sign in

arxiv: 2607.01698 · v1 · pith:7CMBTTJ7new · submitted 2026-07-02 · 💻 cs.CV

Structure-Aware Gaussian Splatting for Large-Scale Scene Reconstruction

Weiyi Xue , Fan Lu , Chi Zhang , Tianhang Wang , Sanqing Qu , Zehan Zheng , Boyuan Zheng , Junqiao Zhao
show 1 more author
Guang Chen
This is my paper

Pith reviewed 2026-07-03 16:54 UTC · model grok-4.3

classification 💻 cs.CV
keywords 3D Gaussian Splattinglarge-scale scene reconstructionfrequency schedulingsignal structure recoverynovel view synthesisGaussian densificationSphere-Constrained Gaussians
0
0 comments X

The pith

Deriving average sampling frequency from 3D representations lets a scheduler align image resolution and Gaussian densification for large-scale scenes.

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

Large-scale scenes in 3D Gaussian Splatting often start with sparse points in under-observed areas. High-frequency image supervision on these low-frequency starts causes uncontrolled densification and extra primitives that hurt both speed and output quality. The paper reframes reconstruction as signal structure recovery and introduces the SIG scheduler, which calculates average sampling frequency and bandwidth to adjust training image resolution and densification on the fly. It adds Sphere-Constrained Gaussians that use point-cloud spatial priors to keep optimization geometry-aware. The result is training that stays frequency-consistent without floaters.

Core claim

By calculating the average sampling frequency and bandwidth of 3D representations, the SIG scheduler synchronizes image supervision with the convergence of Gaussian frequencies and regulates both resolution and densification accordingly. Sphere-Constrained Gaussians further enforce spatial priors from the initial point cloud to limit optimization drift. This combination produces frequency-consistent, geometry-aware, and floater-free training that improves efficiency and rendering quality over prior hardcoded scheduling approaches.

What carries the argument

SIG scheduler that derives average sampling frequency and bandwidth of 3D representations to regulate image resolution and Gaussian densification in sync with scene frequency convergence.

If this is right

  • Training produces fewer redundant primitives in sparsely observed areas.
  • Rendering quality improves while training time decreases in large scenes.
  • The process avoids both under-densification of high-frequency details and over-densification artifacts.

Where Pith is reading between the lines

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

  • The same frequency-matching logic could be tested on other point-based or primitive-based reconstruction pipelines.
  • If the bandwidth estimate can be updated incrementally, the scheduler might adapt to streaming or online capture setups.
  • Applying the sphere constraint idea to other initialization priors could reduce optimization drift in different scene types.

Load-bearing premise

The derived average sampling frequency and bandwidth of 3D representations give a reliable general signal for adjusting image resolution and Gaussian densification without missing details or creating new artifacts.

What would settle it

In a large-scale test scene with known sparse regions, run the method and check whether floaters or redundant primitives still appear at rates comparable to or higher than existing hardcoded schedulers.

Figures

Figures reproduced from arXiv: 2607.01698 by Boyuan Zheng, Chi Zhang, Fan Lu, Guang Chen, Junqiao Zhao, Sanqing Qu, Tianhang Wang, Weiyi Xue, Zehan Zheng.

Figure 1
Figure 1. Figure 1: Floaters and redundancy. Prior methods that directly supervise with high-frequency images lead to redundancy and cannot exploit more primitives to capture high-frequency details. Delving deeper into the issue, given that large-scale scenes inevitably require more image supervision and Gaussians, the degradation can be further attributed to the imbalance between the rendering resolution and the densificatio… view at source ↗
Figure 2
Figure 2. Figure 2: Overview. We define Gaussian frequency based on the opacity field, represented as a weighted sum of Gaussians. Using a frequency-matching module, we synchronize image supervision with Gaussian frequencies (SIG), and optimize with Sphere-Constrained Gaussians to incorporate geometric priors. 3.1 PRELIMINARIES 3DGS utilizes Gaussian primitives {Gi(x) = exp − 1 2 (x − pi) ⊤Σ −1 i (x − pi)  } N i=1 to represe… view at source ↗
Figure 3
Figure 3. Figure 3: Effectiveness Validation: (a) Average Scene Bandwidth during Training under Different Image Resolutions and w3dB Estimation Methods. (b) Our approximation of the power spectrum. 5 [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Qualitative Results. The top two rows illustrate sparse observations, whereas the bottom two rep￾resent regular regions. For regions corresponding to sparse viewpoints, our approach effectively minimizes the generation of redundant and erroneous Gaussian components. For general regions, our method recovers high-frequency details while maintaining intact geometric structures. 4.2 COMPARISON WITH OTHER METHO… view at source ↗
Figure 5
Figure 5. Figure 5: Ablation. CityGS-L: lower densification threshold of CityGS for higher-frequency fitting [PITH_FULL_IMAGE:figures/full_fig_p009_5.png] view at source ↗
read the original abstract

3D Gaussian Splatting has demonstrated remarkable potential in novel view synthesis. In contrast to small-scale scenes, large-scale scenes inevitably contain sparsely observed regions with excessively sparse initial points. In this case, supervising Gaussians initialized from low-frequency sparse points with high-frequency images often induces uncontrolled densification and redundant primitives, degrading both efficiency and quality. Intuitively, this issue can be mitigated with scheduling strategies, which can be categorized into two paradigms: modulating target signal frequency via densification and modulating sampling frequency via image resolution. However, previous scheduling strategies are primarily hardcoded, failing to perceive the convergence behavior of scene frequency. To address this, we reframe the scene reconstruction problem from the perspective of signal structure recovery and propose SIG, a novel scheduler that synchronizes image supervision with Gaussian frequencies. Specifically, we derive the average sampling frequency and bandwidth of 3D representations, and then regulate the training image resolution and the Gaussian densification process based on scene frequency convergence. Furthermore, we introduce Sphere-Constrained Gaussians, which leverage the spatial prior of initialized point clouds to control Gaussian optimization. Our framework enables frequency-consistent, geometry-aware, and floater-free training, achieving state-of-the-art performance by a substantial margin in both efficiency and rendering quality in large-scale scenes. The code is available at: https://github.com/weiyixue999/Signal_Structure_Aware_Gaussian

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 / 1 minor

Summary. The manuscript proposes SIG, a novel scheduler for 3D Gaussian Splatting in large-scale scenes that reframes reconstruction as signal structure recovery. It derives the average sampling frequency and bandwidth of 3D representations to dynamically regulate training image resolution and Gaussian densification according to scene frequency convergence. The work also introduces Sphere-Constrained Gaussians that exploit spatial priors from initialized point clouds to control optimization, aiming for frequency-consistent, geometry-aware, and floater-free training with claimed SOTA gains in efficiency and rendering quality for large scenes. Code is released.

Significance. If the central claims hold, the signal-processing perspective on frequency scheduling could meaningfully advance large-scale novel view synthesis by reducing redundant primitives and uncontrolled densification in sparsely observed regions. The code release supports reproducibility, and the geometry-aware constraint is a concrete addition that may generalize beyond the specific scheduler.

major comments (2)
  1. [Abstract] Abstract (SIG scheduler paragraph): the claim that the derived average sampling frequency and bandwidth provide a reliable, general signal for dynamically regulating image resolution and Gaussian densification is load-bearing for the SOTA assertion, yet the manuscript supplies no equations, derivation steps, or validation that this avoids missing high-frequency details or introducing new artifacts.
  2. [Abstract] Abstract: the assertion of 'state-of-the-art performance by a substantial margin' in both efficiency and rendering quality for large-scale scenes requires quantitative backing (tables, ablations, error analysis) that is absent from the provided description, leaving the central empirical claim unverified.
minor comments (1)
  1. [Abstract] The abstract could briefly indicate the datasets or scene scales used to support the efficiency and quality claims.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the careful review and constructive feedback on the abstract. We address each major comment below and will revise the manuscript accordingly to strengthen the presentation of our technical claims and empirical results.

read point-by-point responses

  1. Referee: [Abstract] Abstract (SIG scheduler paragraph): the claim that the derived average sampling frequency and bandwidth provide a reliable, general signal for dynamically regulating image resolution and Gaussian densification is load-bearing for the SOTA assertion, yet the manuscript supplies no equations, derivation steps, or validation that this avoids missing high-frequency details or introducing new artifacts.

    Authors: The full manuscript derives the average sampling frequency and bandwidth in Section 3.2 using signal-processing analysis of 3D Gaussian representations, with explicit equations and steps for computing scene frequency convergence. These are used to regulate image resolution and densification. Validation appears in Section 4.3 ablations, which quantify that the scheduler preserves high-frequency details (via PSNR/SSIM metrics) without introducing artifacts (via visual inspection and floater counts). We will revise the abstract to briefly reference these equations and validation results. revision: yes

  2. Referee: [Abstract] Abstract: the assertion of 'state-of-the-art performance by a substantial margin' in both efficiency and rendering quality for large-scale scenes requires quantitative backing (tables, ablations, error analysis) that is absent from the provided description, leaving the central empirical claim unverified.

    Authors: Quantitative backing is provided in the manuscript via Tables 1-3 (comparisons to prior methods on large-scale datasets), Section 4.1 (efficiency metrics such as training time and primitive count), Section 4.2 (ablations on the scheduler), and Section 4.3 (error analysis and rendering quality). We will revise the abstract to include specific quantitative margins (e.g., relative improvements in PSNR and FPS) to make the claim self-contained while preserving brevity. revision: yes

Circularity Check

0 steps flagged

No significant circularity in derivation chain

full rationale

The paper reframes reconstruction as signal structure recovery and derives average sampling frequency/bandwidth of 3D representations to drive the SIG scheduler for image resolution and densification. This step is presented as an application of signal processing to the problem rather than a self-definition, fitted input renamed as prediction, or reduction to self-citation. Sphere-Constrained Gaussians are introduced separately using spatial priors from point clouds. No load-bearing claim in the abstract or described method reduces by construction to its own inputs; the central SOTA claims follow from the derived scheduler plus the new Gaussian constraint. This matches the default expectation for non-circular papers.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 1 invented entities

The central claim rests on the ability to derive average sampling frequency and bandwidth from 3D representations and on the assumption that frequency convergence can be used to control training without side effects. Sphere-Constrained Gaussians are a new postulated variant whose behavior is controlled by an external spatial prior.

axioms (2)
  • domain assumption Scene reconstruction can be reframed as signal structure recovery
    Used to motivate the entire SIG scheduler approach.
  • domain assumption Average sampling frequency and bandwidth of 3D representations can be derived and used to regulate training
    Core step for deciding image resolution and densification timing.
invented entities (1)
  • Sphere-Constrained Gaussians no independent evidence
    purpose: Leverage spatial prior of initialized point clouds to control Gaussian optimization and reduce floaters
    New variant introduced to enforce geometry awareness during optimization.

pith-pipeline@v0.9.1-grok · 5799 in / 1393 out tokens · 38137 ms · 2026-07-03T16:54:40.673389+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

77 extracted references · 10 canonical work pages · 5 internal anchors

  1. [1]

    Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

    Point-nerf: Point-based neural radiance fields , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  2. [2]

    ACM Transactions on Graphics (ToG) , volume=

    Tanks and temples: Benchmarking large-scale scene reconstruction , author=. ACM Transactions on Graphics (ToG) , volume=. 2017 , publisher=

    2017

  3. [3]

    arXiv preprint arXiv:2511.04283 , year=

    FastGS: Training 3D Gaussian Splatting in 100 Seconds , author=. arXiv preprint arXiv:2511.04283 , year=

    work page arXiv

  4. [4]

    Proceedings of the Computer Vision and Pattern Recognition Conference , pages=

    Flashgs: Efficient 3d gaussian splatting for large-scale and high-resolution rendering , author=. Proceedings of the Computer Vision and Pattern Recognition Conference , pages=

  5. [5]

    Advances in neural information processing systems , volume=

    Pytorch: An imperative style, high-performance deep learning library , author=. Advances in neural information processing systems , volume=

  6. [6]

    Adam: A Method for Stochastic Optimization

    A method for stochastic optimization , author=. arXiv preprint arXiv:1412.6980 , volume=

    work page internal anchor Pith review Pith/arXiv arXiv

  7. [7]

    Proceedings of the Computer Vision and Pattern Recognition Conference , pages=

    DashGaussian: Optimizing 3D Gaussian Splatting in 200 Seconds , author=. Proceedings of the Computer Vision and Pattern Recognition Conference , pages=

  8. [8]

    SIGGRAPH Asia 2024 Conference Papers , pages=

    Taming 3dgs: High-quality radiance fields with limited resources , author=. SIGGRAPH Asia 2024 Conference Papers , pages=

    2024

  9. [9]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    Surfelgan: Synthesizing realistic sensor data for autonomous driving , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  10. [10]

    2023 , booktitle=

    K-Planes: Explicit Radiance Fields in Space, Time, and Appearance , author=. 2023 , booktitle=

    2023

  11. [11]

    NeRF: Neural Radiance Field in 3D Vision: A Comprehensive Review (Updated Post-Gaussian Splatting)

    Nerf: Neural radiance field in 3d vision, a comprehensive review , author=. arXiv preprint arXiv:2210.00379 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  12. [12]

    Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

    Mip-nerf 360: Unbounded anti-aliased neural radiance fields , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  13. [13]

    ACM SIGGRAPH 2024 conference papers , pages=

    2d gaussian splatting for geometrically accurate radiance fields , author=. ACM SIGGRAPH 2024 conference papers , pages=

    2024

  14. [14]

    A Survey on 3D Gaussian Splatting

    A survey on 3d gaussian splatting , author=. arXiv preprint arXiv:2401.03890 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  15. [15]

    IEEE Transactions on Visualization and Computer Graphics , year=

    Can I Get There? Negotiated User-to-User Teleportations in Social VR , author=. IEEE Transactions on Visualization and Computer Graphics , year=

  16. [16]

    2019 IEEE Conference on Virtual Reality and 3D User Interfaces (VR) , pages=

    Novel view synthesis with multiple 360 images for large-scale 6-dof virtual reality system , author=. 2019 IEEE Conference on Virtual Reality and 3D User Interfaces (VR) , pages=. 2019 , organization=

    2019

  17. [17]

    Communications of the ACM , volume=

    Nerf: Representing scenes as neural radiance fields for view synthesis , author=. Communications of the ACM , volume=. 2021 , publisher=

    2021

  18. [18]

    European Conference on Computer Vision , pages=

    Citygaussian: Real-time high-quality large-scale scene rendering with gaussians , author=. European Conference on Computer Vision , pages=. 2024 , organization=

    2024

  19. [19]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    Vastgaussian: Vast 3d gaussians for large scene reconstruction , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  20. [20]

    IEEE Transactions on Geoscience and Remote Sensing , volume=

    Remote sensing novel view synthesis with implicit multiplane representations , author=. IEEE Transactions on Geoscience and Remote Sensing , volume=. 2022 , publisher=

    2022

  21. [21]

    The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences , volume=

    Mega-NeRF++: An Improved Scalable NeRFs for High-resolution Photogrammetric Images , author=. The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences , volume=. 2024 , publisher=

    2024

  22. [22]

    Advances in Neural Information Processing Systems , volume=

    DOGS: Distributed-Oriented Gaussian Splatting for Large-Scale 3D Reconstruction Via Gaussian Consensus , author=. Advances in Neural Information Processing Systems , volume=

  23. [23]

    , author=

    3d gaussian splatting for real-time radiance field rendering. , author=. ACM Trans. Graph. , volume=

  24. [24]

    ACM Transactions on Graphics (TOG) , volume=

    A hierarchical 3d gaussian representation for real-time rendering of very large datasets , author=. ACM Transactions on Graphics (TOG) , volume=. 2024 , publisher=

    2024

  25. [25]

    Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

    Nerf in the wild: Neural radiance fields for unconstrained photo collections , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  26. [26]

    Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

    Mega-nerf: Scalable construction of large-scale nerfs for virtual fly-throughs , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  27. [27]

    IEEE Transactions on Pattern Analysis and Machine Intelligence , volume=

    Neural radiance fields from sparse rgb-d images for high-quality view synthesis , author=. IEEE Transactions on Pattern Analysis and Machine Intelligence , volume=. 2022 , publisher=

    2022

  28. [28]

    The Eleventh International Conference on Learning Representations , year=

    Switch-nerf: Learning scene decomposition with mixture of experts for large-scale neural radiance fields , author=. The Eleventh International Conference on Learning Representations , year=

  29. [29]

    and Srinivasan, Pratul P

    Verbin, Dor and Hedman, Peter and Mildenhall, Ben and Zickler, Todd and Barron, Jonathan T. and Srinivasan, Pratul P. , journal=. Ref-NeRF: Structured View-Dependent Appearance for Neural Radiance Fields , year=

  30. [30]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    Grid-guided neural radiance fields for large urban scenes , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  31. [31]

    Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

    Structure-from-motion revisited , author=. Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

  32. [32]

    Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

    Block-nerf: Scalable large scene neural view synthesis , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  33. [33]

    GaussNav: Gaussian Splatting for Visual Navigation , year=

    Lei, Xiaohan and Wang, Min and Zhou, Wengang and Li, Houqiang , journal=. GaussNav: Gaussian Splatting for Visual Navigation , year=

  34. [34]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    Scaffold-gs: Structured 3d gaussians for view-adaptive rendering , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  35. [35]

    arXiv preprint arXiv:2403.17898 (2024)

    Octree-gs: Towards consistent real-time rendering with lod-structured 3d gaussians , author=. arXiv preprint arXiv:2403.17898 , year=

    work page arXiv

  36. [36]

    Seminal Graphics Papers: Pushing the Boundaries, Volume 2 , pages=

    Light field rendering , author=. Seminal Graphics Papers: Pushing the Boundaries, Volume 2 , pages=

  37. [37]

    Proceedings of the 23rd annual conference on Computer graphics and interactive techniques , pages=

    View morphing , author=. Proceedings of the 23rd annual conference on Computer graphics and interactive techniques , pages=

  38. [38]

    Visibility-Aware Point-Based Multi-View Stereo Network , year=

    Chen, Rui and Han, Songfang and Xu, Jing and Su, Hao , journal=. Visibility-Aware Point-Based Multi-View Stereo Network , year=

  39. [39]

    Stereo Matching Using Multi-Level Cost Volume and Multi-Scale Feature Constancy , year=

    Liang, Zhengfa and Guo, Yulan and Feng, Yiliu and Chen, Wei and Qiao, Linbo and Zhou, Li and Zhang, Jianfeng and Liu, Hengzhu , journal=. Stereo Matching Using Multi-Level Cost Volume and Multi-Scale Feature Constancy , year=

  40. [40]

    S-NeRF++: Autonomous Driving Simulation via Neural Reconstruction and Generation , year=

    Chen, Yurui and Zhang, Junge and Xie, Ziyang and Li, Wenye and Zhang, Feihu and Lu, Jiachen and Zhang, Li , journal=. S-NeRF++: Autonomous Driving Simulation via Neural Reconstruction and Generation , year=

  41. [41]

    Stereoscopic Displays and Virtual Reality Systems II , volume=

    Head-tracked stereoscopic display using image warping , author=. Stereoscopic Displays and Virtual Reality Systems II , volume=. 1995 , organization=

    1995

  42. [42]

    A Hybrid Multiview Stereo Algorithm for Modeling Urban Scenes , year=

    Lafarge, Florent and Keriven, Renaud and Brédif, Mathieu and Vu, Hoang-Hiep , journal=. A Hybrid Multiview Stereo Algorithm for Modeling Urban Scenes , year=

  43. [43]

    Multi-Scale Geometric Consistency Guided and Planar Prior Assisted Multi-View Stereo , year=

    Xu, Qingshan and Kong, Weihang and Tao, Wenbing and Pollefeys, Marc , journal=. Multi-Scale Geometric Consistency Guided and Planar Prior Assisted Multi-View Stereo , year=

  44. [44]

    IEEE Transactions on Pattern Analysis and Machine Intelligence , volume=

    Cascaded and Generalizable Neural Radiance Fields for Fast View Synthesis , author=. IEEE Transactions on Pattern Analysis and Machine Intelligence , volume=. 2023 , publisher=

    2023

  45. [45]

    arXiv preprint arXiv:2412.04887 , year=

    Momentum-gs: Momentum gaussian self-distillation for high-quality large scene reconstruction , author=. arXiv preprint arXiv:2412.04887 , year=

    work page arXiv

  46. [46]

    Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

    Deepstereo: Learning to predict new views from the world's imagery , author=. Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

  47. [47]

    Advances in Neural Information Processing Systems , volume=

    GeoNLF: Geometry guided Pose-Free Neural LiDAR Fields , author=. Advances in Neural Information Processing Systems , volume=

  48. [48]

    Proceedings of the IEEE/CVF international conference on computer vision , pages=

    Barf: Bundle-adjusting neural radiance fields , author=. Proceedings of the IEEE/CVF international conference on computer vision , pages=

  49. [49]

    European Conference on Computer Vision , pages=

    Revising densification in gaussian splatting , author=. European Conference on Computer Vision , pages=. 2024 , organization=

    2024

  50. [50]

    arXiv preprint arXiv:2504.09048 , year=

    BlockGaussian: Efficient Large-Scale Scene Novel View Synthesis via Adaptive Block-Based Gaussian Splatting , author=. arXiv preprint arXiv:2504.09048 , year=

    work page arXiv

  51. [51]

    Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

    Mip-splatting: Alias-free 3d gaussian splatting , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  52. [52]

    Stereo Magnification: Learning View Synthesis using Multiplane Images

    Stereo magnification: Learning view synthesis using multiplane images , author=. arXiv preprint arXiv:1805.09817 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  53. [53]

    Real-Time Globally Consistent Dense 3D Reconstruction With Online Texturing , year=

    Han, Lei and Gu, Siyuan and Zhong, Dawei and Quan, Shuxue and Fang, Lu , journal=. Real-Time Globally Consistent Dense 3D Reconstruction With Online Texturing , year=

  54. [54]

    Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

    Plenoxels: Radiance fields without neural networks , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  55. [55]

    Proceedings of the IEEE/CVF international conference on computer vision , pages=

    Mip-nerf: A multiscale representation for anti-aliasing neural radiance fields , author=. Proceedings of the IEEE/CVF international conference on computer vision , pages=

  56. [56]

    , author=

    Patchmatch stereo-stereo matching with slanted support windows. , author=. Bmvc , volume=

  57. [57]

    Proceedings of the IEEE international conference on computer vision , pages=

    Massively parallel multiview stereopsis by surface normal diffusion , author=. Proceedings of the IEEE international conference on computer vision , pages=

  58. [58]

    and Frahm, Jan-Michael , booktitle=

    Schönberger, Johannes L. and Frahm, Jan-Michael , booktitle=. Structure-from-Motion Revisited , year=

  59. [59]

    ACM transactions on graphics (TOG) , volume=

    Instant neural graphics primitives with a multiresolution hash encoding , author=. ACM transactions on graphics (TOG) , volume=. 2022 , publisher=

    2022

  60. [60]

    European conference on computer vision , pages=

    Tensorf: Tensorial radiance fields , author=. European conference on computer vision , pages=. 2022 , organization=

    2022

  61. [61]

    Proceedings of the IEEE/CVF international conference on computer vision , pages=

    Mvsnerf: Fast generalizable radiance field reconstruction from multi-view stereo , author=. Proceedings of the IEEE/CVF international conference on computer vision , pages=

  62. [62]

    Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

    Ibrnet: Learning multi-view image-based rendering , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  63. [63]

    Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

    The unreasonable effectiveness of deep features as a perceptual metric , author=. Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

  64. [64]

    European Conference on Computer Vision , pages=

    Street gaussians: Modeling dynamic urban scenes with gaussian splatting , author=. European Conference on Computer Vision , pages=. 2024 , organization=

    2024

  65. [65]

    ECCV , pages =

    Capturing, Reconstructing, and Simulating: the UrbanScene3D Dataset , author=. ECCV , pages =

  66. [66]

    2024 , publisher=

    SparseGS: Real-time 360° sparse view synthesis using Gaussian splatting , author=. 2024 , publisher=

    2024

  67. [67]

    Advances in neural information processing systems , volume=

    Lightgaussian: Unbounded 3d gaussian compression with 15x reduction and 200+ fps , author=. Advances in neural information processing systems , volume=

  68. [68]

    arXiv preprint arXiv:2311.18159 , volume=

    Compact3d: Compressing gaussian splat radiance field models with vector quantization , author=. arXiv preprint arXiv:2311.18159 , volume=

    work page arXiv

  69. [69]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    Gaussian splatting slam , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  70. [70]

    Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

    Matrixcity: A large-scale city dataset for city-scale neural rendering and beyond , author=. Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

  71. [71]

    Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

    Hugs: Human gaussian splats , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  72. [72]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    Splatam: Splat track & map 3d gaussians for dense rgb-d slam , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  73. [73]

    Depth Anything V2

    Depth Anything V2 , author=. arXiv:2406.09414 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  74. [74]

    Proceedings of the seventh IEEE international conference on computer vision , volume=

    Manhattan world: Compass direction from a single image by bayesian inference , author=. Proceedings of the seventh IEEE international conference on computer vision , volume=. 1999 , organization=

    1999

  75. [75]

    Communications of the ACM , volume=

    Building rome in a day , author=. Communications of the ACM , volume=. 2011 , publisher=

    2011

  76. [76]

    Proceedings of the 28th annual conference on Computer graphics and interactive techniques , pages=

    An efficient representation for irradiance environment maps , author=. Proceedings of the 28th annual conference on Computer graphics and interactive techniques , pages=

  77. [77]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    Depth-regularized optimization for 3d gaussian splatting in few-shot images , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=