arxiv: 2605.07254 · v1 · submitted 2026-05-08 · 💻 cs.CV · cs.GR

Recognition: no theorem link

High-Fidelity Surface Splatting-Based 3D Reconstruction from Multi-View Images

Nandhana Sunil , Abhirami R Iyer , Avirup Mandal

Authors on Pith no claims yet

Pith reviewed 2026-05-11 01:34 UTC · model grok-4.3

classification 💻 cs.CV cs.GR

keywords 3D reconstructionsurface splattingimplicit moving least squarespolynomial kernelmulti-view imageshigh-frequency geometryLaplacian regularizationmesh extraction

0 comments

The pith

A compact polynomial kernel with local support in implicit moving least squares preserves high-frequency details better than exponential kernels for 3D surface reconstruction from multi-view images.

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

The paper seeks to advance direct mesh reconstruction from sets of photos by fixing a limitation in implicit moving least squares. Exponential kernels used in prior work tend to smooth away sharp edges and fine textures. The authors replace them with a compact polynomial kernel that has built-in local support, which gives explicit control over the frequencies retained in the surface. They further add stochastic Laplacian regularization during optimization to protect small-scale features without causing instability. If the approach works, it would let practitioners obtain accurate 3D meshes and high-quality renders in one joint training pass rather than relying on separate post-processing stages.

Core claim

The paper claims that replacing exponential kernels with a compact polynomial kernel of local support in the implicit moving least squares formulation, together with stochastic Laplacian regularization, yields improved geometric fidelity and sharper appearance from multi-view images. This formulation supports end-to-end conversion of point clouds into signed distance and texture fields, removing the need for post-processing mesh extraction used by methods such as 3D Gaussian Splatting.

What carries the argument

Compact polynomial kernel with local support inside the implicit moving least squares surface splatting model, augmented by stochastic Laplacian filtering.

If this is right

Direct end-to-end optimization of geometry and appearance becomes feasible without separate mesh extraction steps.
High-frequency geometric details are retained more reliably from sparse input views.
Rendering quality improves through sharper textures and reduced smoothing artifacts.
The method remains stable under stochastic regularization, supporting consistent training across scenes.
Surface splatting can be applied to practical multi-view capture pipelines with fewer post-processing requirements.

Where Pith is reading between the lines

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

The local support property may allow the same kernel to be used in real-time incremental reconstruction settings where only nearby points are updated.
Polynomial degree could be treated as a tunable hyperparameter to trade off smoothness against detail in different capture densities.
The approach might reduce the number of input views needed for acceptable fidelity, lowering the cost of 3D scanning sessions.
Integration with existing point-cloud pipelines could become straightforward because the kernel operates directly on local neighborhoods.

Load-bearing premise

The compact polynomial kernel with local support and Laplacian regularization preserves high-frequency structure better than exponential kernels without introducing new artifacts or optimization instability.

What would settle it

A controlled experiment on the DTU multi-view benchmark in which the new kernel produces higher Chamfer distance or lower PSNR than an otherwise identical exponential-kernel baseline would falsify the claimed advantage.

Figures

Figures reproduced from arXiv: 2605.07254 by Abhirami R Iyer, Avirup Mandal, Nandhana Sunil.

**Figure 1.** Figure 1: Visualization of the proposed kernel γours under varying parameters ki and mi , compared against the exponential kernel γimls used in IMLS Splatting Yang et al. [2025]. The 1-D profiles (top) demonstrate that the proposed kernel drops to exactly zero at the compact support boundary miki (dotted line), unlike the exponential kernel, which extends infinitely and requires arbitrary truncation. The 2-D heatmap… view at source ↗

**Figure 2.** Figure 2: A qualitative comparison of the generated meshes. Meshes generated using IMLS Splat [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗

**Figure 3.** Figure 3: Qualitative rendering comparison with IMLS Splatting Yang et al. [2025] method. Images [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗

**Figure 4.** Figure 4: Reconstructed Mesh comparison on the Mic scene (Top Row). Mesh reconstructions without (left) and with (right) SP and Laplacian, showing that the baseline exhibits floating artifacts and geometric inconsistencies, particularly in thin structures, while the full model produces smoother and more coherent surfaces. Rendered images comparison on the Ficus scene (Bottom Row). (Left to Right). Ground truth image… view at source ↗

**Figure 5.** Figure 5: Reconstructed Mesh comparison on the Ficus scene (Top Row). (Left) Reconstruction without stochastic preconditioning (SP) and Laplacian regularization exhibits floating artifacts and structural inconsistencies. (Right) Incorporating SP and Laplacian produces smoother and more coherent surfaces. Rendered images comparison on the Ficus scene (Bottom Row). (Left to Right). Ground truth, reconstruction without… view at source ↗

**Figure 6.** Figure 6: Reconstructed Mesh comparison on the Lego scene (Top Row). (Left) Reconstruction without stochastic preconditioning (SP) and Laplacian regularization exhibits floating artifacts and structural inconsistencies. (Right) Incorporating SP and Laplacian produces smoother and more coherent surfaces. Rendered images comparison on the Lego scene (Bottom Row). (Left to Right) ground truth, reconstruction without SP… view at source ↗

**Figure 7.** Figure 7: Reconstructed Mesh comparison on the Lego scene (Top Row) (View from a different angle). (Left) Reconstruction without stochastic preconditioning (SP) and Laplacian regularization exhibits floating artifacts and structural inconsistencies. (Right) Incorporating SP and Laplacian produces smoother and more coherent surfaces. Rendered images comparison on the Lego scene (Bottom Row). (Left to Right). Ground t… view at source ↗

**Figure 8.** Figure 8: Reconstructed Mesh comparison on the scan24 scene (Top Row). (Left to Right). Ground truth, reconstruction without stochastic preconditioning (SP), and Laplacian regularization, and reconstruction with SP and Laplacian. Reconstruction without stochastic preconditioning and Laplacian regularization exhibits floating artifacts and structural inconsistencies. Reconstruction with SP and Laplacian mitigates the… view at source ↗

**Figure 9.** Figure 9: Reconstructed Mesh comparison on the scan122 scene. (Left to Right). Ground truth, reconstruction without stochastic preconditioning (SP), and Laplacian regularization, reconstruction with stochastic preconditioning (SP), and Laplacian regularization. Reconstruction without stochastic preconditioning and Laplacian regularization exhibits floating artifacts and structural inconsistencies. Reconstruction wit… view at source ↗

read the original abstract

Multi-view mesh reconstruction remains a core challenge in computer graphics and vision, especially for recovering high-frequency geometry from sparse observations. Recent methods such as 3D Gaussian Splatting (3DGS) and Neural Radiance Fields (NeRF) rely on post-processing for mesh extraction, thereby limiting joint optimization of geometry and appearance. Implicit Moving Least Squares (IMLS) instead enables direct conversion of point clouds into signed distance and texture fields, supporting end-to-end reconstruction and rendering. However, existing IMLS formulations use exponential kernels that struggle with high-frequency detail. We introduce a compact polynomial kernel with local support and greater flexibility, allowing better control over frequency content and improved geometric fidelity. To further enhance fine details, we incorporate stochastic regularization with Laplacian filtering. Together, these improve the preservation of high-frequency structure while maintaining stable optimization. Experiments show state-of-the-art performance in both surface reconstruction and rendering, yielding more accurate geometry and sharper visuals from multi-view data.

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.

Desk Editor's Note private letter to a colleague

The paper introduces a polynomial kernel for IMLS plus Laplacian regularization for sharper multi-view reconstruction, but needs ablations to confirm the kernel drives the gains.

read the letter

The main takeaway is that the authors replace the exponential kernel in Implicit Moving Least Squares with a compact polynomial kernel that has local support, and they add stochastic Laplacian regularization to preserve more high-frequency geometry details during multi-view reconstruction. This approach is new in its specific kernel choice and the way it combines with regularization for better frequency control. The paper does well by focusing on direct mesh optimization from points without needing extra post-processing, which is a practical advantage over methods like 3D Gaussian Splatting or NeRF that extract meshes separately. It identifies a clear problem with over-smoothing in existing IMLS and offers a targeted fix that could lead to sharper surfaces and better visuals if the results check out. The soft spots center on proving the kernel's contribution. The abstract says the polynomial kernel and regularization together improve high-frequency preservation, yet there are no details on ablations that test the kernel in isolation or on metrics that measure frequency content directly, such as edge sharpness or spectral analysis. Without those, the gains might stem from other implementation choices rather than the kernel itself. The stress-test concern about missing validation for the kernel's role is accurate based on the provided information. This paper targets researchers in computer vision and graphics who deal with 3D reconstruction from multiple images. Someone working on improving implicit surface representations would get useful ideas from the method, assuming the full experiments include solid baselines and quantitative support. It deserves peer review because the problem is relevant and the proposal is grounded, though the authors should be asked to add the missing ablations for stronger claims.

Referee Report

2 major / 1 minor

Summary. The paper introduces a compact polynomial kernel with local support for Implicit Moving Least Squares (IMLS) surface reconstruction from multi-view images, paired with stochastic Laplacian regularization. It claims this combination better preserves high-frequency geometric details than prior exponential-kernel IMLS formulations, enabling direct end-to-end optimization of geometry and appearance and yielding state-of-the-art results in both surface reconstruction accuracy and rendering quality.

Significance. If the claims are substantiated, the work could meaningfully advance multi-view 3D reconstruction by providing a direct mesh-extraction pathway that avoids post-processing artifacts common in NeRF and 3D Gaussian Splatting pipelines, while offering explicit control over frequency content. This would be particularly relevant for applications needing high-fidelity surfaces from sparse views.

major comments (2)

[Abstract] Abstract: The assertion of 'state-of-the-art performance in both surface reconstruction and rendering' is presented without any quantitative metrics, error bars, baseline comparisons, datasets, or experimental protocol. This absence prevents assessment of whether the claimed improvements in geometry accuracy and visual sharpness are supported by evidence.
[Method/Experiments] Method/Experiments: The central attribution that the compact polynomial kernel (with local support) plus Laplacian regularization 'together improve' high-frequency preservation is not isolated. No kernel-swap ablation (holding regularization fixed) or frequency-domain metrics (e.g., power-spectrum error or edge-sharpness histograms) are referenced, leaving open the possibility that gains arise from regularization, data terms, or optimizer choices rather than the kernel itself. This is load-bearing for the main technical contribution.

minor comments (1)

[Abstract] Abstract: Consider adding one sentence on the specific multi-view datasets or scenes used to ground the SOTA claim.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their thoughtful comments and the opportunity to clarify and strengthen our manuscript. We address each major comment below and outline the revisions we will make.

read point-by-point responses

Referee: [Abstract] Abstract: The assertion of 'state-of-the-art performance in both surface reconstruction and rendering' is presented without any quantitative metrics, error bars, baseline comparisons, datasets, or experimental protocol. This absence prevents assessment of whether the claimed improvements in geometry accuracy and visual sharpness are supported by evidence.

Authors: We agree that the abstract would benefit from including specific quantitative evidence to support the state-of-the-art claims. In the revised manuscript, we will update the abstract to include key performance metrics (e.g., average Chamfer distance on DTU dataset and PSNR for rendering) and mention the main baselines and datasets used. This will provide readers with a clearer indication of the improvements while keeping the abstract concise. revision: yes
Referee: [Method/Experiments] Method/Experiments: The central attribution that the compact polynomial kernel (with local support) plus Laplacian regularization 'together improve' high-frequency preservation is not isolated. No kernel-swap ablation (holding regularization fixed) or frequency-domain metrics (e.g., power-spectrum error or edge-sharpness histograms) are referenced, leaving open the possibility that gains arise from regularization, data terms, or optimizer choices rather than the kernel itself. This is load-bearing for the main technical contribution.

Authors: We appreciate this point on isolating the contributions. Our current experiments compare the full method against prior exponential-kernel IMLS and other SOTA approaches, demonstrating superior high-frequency detail preservation. However, to more rigorously attribute the gains to the polynomial kernel, we will add a kernel-swap ablation study where we replace our polynomial kernel with the exponential one while keeping the Laplacian regularization and other components fixed. Additionally, we will include frequency-domain analysis, such as power spectrum error plots, to quantify the high-frequency preservation. These additions will be incorporated in the revised version. revision: yes

Circularity Check

0 steps flagged

No significant circularity; novel kernel and regularization introduced without self-referential definitions or fitted inputs renamed as predictions.

full rationale

The paper proposes a compact polynomial kernel with local support plus stochastic Laplacian regularization as improvements over exponential-kernel IMLS. No equations appear in the provided abstract or description that define any quantity in terms of itself, treat a fitted parameter as an independent prediction, or rely on a uniqueness theorem imported from the authors' prior work. The central claims rest on end-to-end experimental superiority rather than any derivation that reduces by construction to the inputs. This is the most common honest outcome for a methods paper that introduces a new ansatz and validates it externally via benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract supplies no equations, parameters, or assumptions; free_parameters, axioms, and invented_entities cannot be populated.

pith-pipeline@v0.9.0 · 5472 in / 1041 out tokens · 38238 ms · 2026-05-11T01:34:00.080346+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

73 extracted references · 73 canonical work pages

[1]

2022 , isbn =

Kim, Theodore and Eberle, David , title =. 2022 , isbn =. doi:10.1145/3532720.3535628 , booktitle =

work page doi:10.1145/3532720.3535628 2022
[2]

and Li, Jie and Narain, Rahul , journal=

Overby, Matthew and Brown, George E. and Li, Jie and Narain, Rahul , journal=. ADMM Projective Dynamics: Fast Simulation of Hyperelastic Models with Dynamic Constraints , year=

work page
[3]

Foundations and Trends® in Machine Learning , author =

Foundations and Trends in Machine Learning , title =. 2011 , volume =. doi:10.1561/2200000016 , issn =

work page doi:10.1561/2200000016 2011
[4]

2019 , issue_date =

Xian, Zangyueyang and Tong, Xin and Liu, Tiantian , title =. 2019 , issue_date =. doi:10.1145/3355089.3356486 , journal =

work page doi:10.1145/3355089.3356486 2019
[5]

2025 , issue_date =

Lu, Jia-Ming and Yuan, Tailing and Mo, Zhe-Han and Hu, Shi-Min , title =. 2025 , issue_date =. doi:10.1145/3763327 , journal =

work page doi:10.1145/3763327 2025
[6]

Capture and modeling of non-linear heterogeneous soft tissue , year =

Bickel, Bernd and B\". Capture and modeling of non-linear heterogeneous soft tissue , year =. doi:10.1145/1531326.1531395 , journal =

work page doi:10.1145/1531326.1531395
[7]

and Ramamoorthi, Ravi , title =

Wang, Huamin and O'Brien, James F. and Ramamoorthi, Ravi , title =. 2011 , issue_date =. doi:10.1145/2010324.1964966 , journal =

work page doi:10.1145/2010324.1964966 2011
[8]

Computer Graphics Forum , volume =

Miguel, E. and Bradley, D. and Thomaszewski, B. and Bickel, B. and Matusik, W. and Otaduy, M. A. and Marschner, S. , title =. Computer Graphics Forum , volume =. doi:https://doi.org/10.1111/j.1467-8659.2012.03031.x , year =

work page doi:10.1111/j.1467-8659.2012.03031.x 2012
[9]

, title =

Miguel, Eder and Miraut, David and Otaduy, Miguel A. , title =. Computer Graphics Forum , volume =. doi:https://doi.org/10.1111/cgf.12840 , year =

work page doi:10.1111/cgf.12840
[10]

Nonlinear material design using principal stretches , year =

Xu, Hongyi and Sin, Funshing and Zhu, Yufeng and Barbi. Nonlinear material design using principal stretches , year =. doi:10.1145/2766917 , journal =

work page doi:10.1145/2766917
[11]

: Integer-grid maps for reliable quad meshing

R\'. Embedded thin shells for wrinkle simulation , year =. doi:10.1145/2461912.2462018 , journal =

work page doi:10.1145/2461912.2462018
[12]

, title =

Li, Pengbo and Kry, Paul G. , title =. 2014 , isbn =. doi:10.1145/2668064.2668089 , booktitle =

work page doi:10.1145/2668064.2668089 2014
[13]

and Pai, Dinesh K

Li, Duo and Sueda, Shinjiro and Neog, Debanga R. and Pai, Dinesh K. , title =. 2013 , issue_date =. doi:10.1145/2461912.2462008 , journal =

work page doi:10.1145/2461912.2462008 2013
[14]

Strain based dynamics , year =

M\". Strain based dynamics , year =. Proceedings of the ACM SIGGRAPH/Eurographics Symposium on Computer Animation , pages =

work page
[15]

1987 , isbn =

Terzopoulos, Demetri and Platt, John and Barr, Alan and Fleischer, Kurt , title =. 1987 , isbn =. doi:10.1145/37401.37427 , booktitle =

work page doi:10.1145/37401.37427 1987
[16]

1998 , isbn =

Baraff, David and Witkin, Andrew , title =. 1998 , isbn =. doi:10.1145/280814.280821 , booktitle =

work page doi:10.1145/280814.280821 1998
[17]

and Yang, Weiwei and Tong, Y

Kharevych, L. and Yang, Weiwei and Tong, Y. and Kanso, E. and Marsden, J. E. and Schr\". Geometric, variational integrators for computer animation , year =. Proceedings of the 2006 ACM SIGGRAPH/Eurographics Symposium on Computer Animation , pages =

work page 2006
[18]

2011 , isbn =

Martin, Sebastian and Thomaszewski, Bernhard and Grinspun, Eitan and Gross, Markus , title =. 2011 , isbn =. doi:10.1145/1964921.1964967 , booktitle =

work page doi:10.1145/1964921.1964967 2011
[19]

Gast, T. F. and Schroeder, C. , title =. Proceedings of the ACM SIGGRAPH/Eurographics Symposium on Computer Animation , pages =. 2015 , publisher =

work page 2015
[20]

and O'Brien, James F

Liu, Tiantian and Bargteil, Adam W. and O'Brien, James F. and Kavan, Ladislav , title =. 2013 , issue_date =. doi:10.1145/2508363.2508406 , journal =

work page doi:10.1145/2508363.2508406 2013
[21]

2014 , issue_date =

Bouaziz, Sofien and Martin, Sebastian and Liu, Tiantian and Kavan, Ladislav and Pauly, Mark , title =. 2014 , issue_date =. doi:10.1145/2601097.2601116 , journal =

work page doi:10.1145/2601097.2601116 2014
[22]

Computer Graphics Forum , volume =

Bouaziz, Sofien and Deuss, Mario and Schwartzburg, Yuliy and Weise, Thibaut and Pauly, Mark , title =. Computer Graphics Forum , volume =. doi:https://doi.org/10.1111/j.1467-8659.2012.03171.x , year =

work page doi:10.1111/j.1467-8659.2012.03171.x 2012
[23]

2016 , isbn =

Weiler, Marcel and Koschier, Dan and Bender, Jan , title =. 2016 , isbn =. doi:10.1145/2994258.2994282 , booktitle =

work page doi:10.1145/2994258.2994282 2016
[24]

2016 , issue_date =

Fratarcangeli, Marco and Tibaldo, Valentina and Pellacini, Fabio , title =. 2016 , issue_date =. doi:10.1145/2980179.2982437 , journal =

work page doi:10.1145/2980179.2982437 2016
[25]

XPBD: position-based simulation of compliant constrained dynamics , year =

Macklin, Miles and M\". XPBD: position-based simulation of compliant constrained dynamics , year =. doi:10.1145/2994258.2994272 , booktitle =

work page doi:10.1145/2994258.2994272
[26]

Position based dynamics , year =

M\". Position based dynamics , year =. doi:10.1016/j.jvcir.2007.01.005 , journal =

work page doi:10.1016/j.jvcir.2007.01.005 2007
[27]

and Teschner, Matthias and Macklin, Miles , title =

Bender, Jan and Müller, Matthias and Otaduy, Miguel A. and Teschner, Matthias and Macklin, Miles , title =. Computer Graphics Forum , volume =. doi:https://doi.org/10.1111/cgf.12346 , year =

work page doi:10.1111/cgf.12346
[28]

2017 , issue_date =

Liu, Tiantian and Bouaziz, Sofien and Kavan, Ladislav , title =. 2017 , issue_date =. doi:10.1145/3072959.2990496 , journal =

work page doi:10.1145/3072959.2990496 2017
[29]

, title =

Nocedal, Jorge and Wright, Stephen J. , title =. 2006 , isbn =

work page 2006
[30]

SIAM Journal on Control and Optimization , volume =

Tseng, Paul , title =. SIAM Journal on Control and Optimization , volume =. 1991 , doi =

work page 1991
[31]

2008 , url=

An Efficient Primal-Dual Hybrid Gradient Algorithm For Total Variation Image Restoration , author=. 2008 , url=

work page 2008
[32]

2015 , issue_date =

Wang, Huamin , title =. 2015 , issue_date =. doi:10.1145/2816795.2818063 , journal =

work page doi:10.1145/2816795.2818063 2015
[33]

and Cano, Alfonso and Giannakis, Georgios B

Forero, Pedro A. and Cano, Alfonso and Giannakis, Georgios B. , journal=. Distributed Clustering Using Wireless Sensor Networks , year=

work page
[34]

doi:10.1088/0266-5611/28/11/115010 , year =

Wen, Zaiwen and Yang, Chao and Liu, Xin and Marchesini, Stefano , title =. doi:10.1088/0266-5611/28/11/115010 , year =

work page doi:10.1088/0266-5611/28/11/115010
[35]

and Févotte, Cédric , booktitle=

Sun, Dennis L. and Févotte, Cédric , booktitle=. Alternating direction method of multipliers for non-negative matrix factorization with the beta-divergence , year=

work page
[36]

Convergence analysis of alternating direction method of multipliers for a family of nonconvex problems , year=

Hong, Mingyi and Luo, Zhi-Quan and Razaviyayn, Meisam , booktitle=. Convergence analysis of alternating direction method of multipliers for a family of nonconvex problems , year=

work page
[37]

SIAM Journal on Optimization , volume =

Li, Guoyin and Pong, Ting Kei , title =. SIAM Journal on Optimization , volume =. 2015 , doi =

work page 2015
[38]

IEEE Transactions on Control of Network Systems , title=

Magn\'. IEEE Transactions on Control of Network Systems , title=. 2016 , volume=

work page 2016
[39]

Mathematical Programming , year =

Goldfarb, Donald and Ma, Shiqian and Scheinberg, Katya , title =. Mathematical Programming , year =. doi:10.1007/s10107-012-0530-2 , url =

work page doi:10.1007/s10107-012-0530-2
[40]

SIAM Journal on Imaging Sciences , volume =

Goldstein, Tom and O'Donoghue, Brendan and Setzer, Simon and Baraniuk, Richard , title =. SIAM Journal on Imaging Sciences , volume =. 2014 , doi =

work page 2014
[41]

2015 , isbn =

Kadkhodaie, Mojtaba and Christakopoulou, Konstantina and Sanjabi, Maziar and Banerjee, Arindam , title =. 2015 , isbn =. doi:10.1145/2783258.2783400 , booktitle =

work page doi:10.1145/2783258.2783400 2015
[42]

2014 , issue_date =

Gregson, James and Ihrke, Ivo and Thuerey, Nils and Heidrich, Wolfgang , title =. 2014 , issue_date =. doi:10.1145/2601097.2601147 , journal =

work page doi:10.1145/2601097.2601147 2014
[43]

, title =

Narain, Rahul and Samii, Armin and O'Brien, James F. , title =. ACM Trans. Graph. , month = nov, articleno =. 2012 , issue_date =

work page 2012
[44]

and Bulbul, Abdullah and Ward, Gregory J

Narain, Rahul and Albert, Rachel A. and Bulbul, Abdullah and Ward, Gregory J. and Banks, Martin S. and O'Brien, James F. , title =. ACM Trans. Graph. , month = jul, articleno =. 2015 , issue_date =

work page 2015
[45]

ACM Trans

Smith, Breannan and Goes, Fernando De and Kim, Theodore , title =. ACM Trans. Graph. , month = mar, articleno =. 2018 , issue_date =

work page 2018
[46]

ACM Trans

Smith, Breannan and Goes, Fernando De and Kim, Theodore , title =. ACM Trans. Graph. , month = feb, articleno =. 2019 , issue_date =

work page 2019
[47]

Bianco and Sharon Gannot and Peter Gerstoft , publisher =

Efren Fernandez-Grande and Michael J. Bianco and Sharon Gannot and Peter Gerstoft , publisher =. DTU three-channel room impulse response dataset for direction of arrival estimation 2020 , year =

work page 2020
[48]

2020 , eprint=

NeRF: Representing Scenes as Neural Radiance Fields for View Synthesis , author=. 2020 , eprint=

work page 2020
[49]

Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV) , year=

NeuManifold: Neural Watertight Manifold Reconstruction with Efficient and High-Quality Rendering Support , author=. Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV) , year=

work page
[50]

2014 IEEE Conference on Computer Vision and Pattern Recognition , pages=

Large scale multi-view stereopsis evaluation , author=. 2014 IEEE Conference on Computer Vision and Pattern Recognition , pages=. 2014 , organization=

work page 2014
[51]

and Voulgaris, Petros and Stipanović, Dušan , booktitle=

Li, Yichuan and Freris, Nikolaos M. and Voulgaris, Petros and Stipanović, Dušan , booktitle=. DN-ADMM: Distributed Newton ADMM for Multi-agent Optimization , year=

work page
[52]

and Roosta, Fred and Mahoney, Michael W

Fang, Chih-Hao and Kylasa, Sudhir B. and Roosta, Fred and Mahoney, Michael W. and Grama, Ananth , booktitle=. Newton-ADMM: A Distributed GPU-Accelerated Optimizer for Multiclass Classification Problems , year=

work page
[53]

1981 , isbn=

Practical optimization , author=. 1981 , isbn=

work page 1981
[54]

ACM Trans

Yang, Kaizhi and Dai, Liu and Liu, Isabella and Zhang, Xiaoshuai and Sun, Xiaoyan and Chen, Xuejin and Xu, Zexiang and Su, Hao , title =. ACM Trans. Graph. , month = jul, articleno =. 2025 , issue_date =

work page 2025
[55]

ACM Trans

Kolluri, Ravikrishna , title =. ACM Trans. Algorithms , month = may, articleno =. 2008 , issue_date =

work page 2008
[56]

Proceedings of the Twenty-Fifth Annual ACM-SIAM Symposium on Discrete Algorithms , pages =

Cheng, Siu-Wing and Chiu, Man-Kwun , title =. Proceedings of the Twenty-Fifth Annual ACM-SIAM Symposium on Discrete Algorithms , pages =. 2014 , isbn =

work page 2014
[57]

ACM Trans

Ling, Selena and Nimier-David, Merlin and Jacobson, Alec and Sharp, Nicholas , title =. ACM Trans. Graph. , month = jul, articleno =. 2025 , issue_date =

work page 2025
[58]

Neural-IMLS: Learning Implicit Moving Least-Squares for Surface Reconstruction from Unoriented Point clouds , journal =

Zixiong Wang and Pengfei Wang and Qiujie Dong and Junjie Gao and Shuang. Neural-IMLS: Learning Implicit Moving Least-Squares for Surface Reconstruction from Unoriented Point clouds , journal =. 2021 , eprinttype =. 2109.04398 , timestamp =

work page arXiv 2021
[59]

Deep Implicit Moving Least-Squares Functions for 3D Reconstruction , journal =

Shi. Deep Implicit Moving Least-Squares Functions for 3D Reconstruction , journal =. 2021 , eprinttype =. 2103.12266 , timestamp =

work page arXiv 2021
[60]

Advances in Computational Mathematics , year =

Holger Wendland , title =. Advances in Computational Mathematics , year =

work page
[61]

Thirty-Fifth Conference on Neural Information Processing Systems , year=

Volume rendering of neural implicit surfaces , author=. Thirty-Fifth Conference on Neural Information Processing Systems , year=

work page
[62]

2023 , eprint=

Neuralangelo: High-Fidelity Neural Surface Reconstruction , author=. 2023 , eprint=

work page 2023
[63]

2023 , eprint=

3D Gaussian Splatting for Real-Time Radiance Field Rendering , author=. 2023 , eprint=

work page 2023
[64]

2023 , eprint=

SuGaR: Surface-Aligned Gaussian Splatting for Efficient 3D Mesh Reconstruction and High-Quality Mesh Rendering , author=. 2023 , eprint=

work page 2023
[65]

2024 , eprint=

Gaussian Opacity Fields: Efficient Adaptive Surface Reconstruction in Unbounded Scenes , author=. 2024 , eprint=

work page 2024
[66]

uller, Thomas and Fidler, Sanja , title =

Munkberg, Jacob and Hasselgren, Jon and Shen, Tianchang and Gao, Jun and Chen, Wenzheng and Evans, Alex and M\"uller, Thomas and Fidler, Sanja , title = ". Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , month =. 2022 , pages =

work page 2022
[67]

Flexible isosurface extraction for gradient-based mesh optimization.ACM Trans

Shen, Tianchang and Munkberg, Jacob and Hasselgren, Jon and Yin, Kangxue and Wang, Zian and Chen, Wenzheng and Gojcic, Zan and Fidler, Sanja and Sharp, Nicholas and Gao, Jun , title =. 2023 , issue_date =. doi:10.1145/3592430 , journal =

work page doi:10.1145/3592430 2023
[68]

Structure-from-Motion Revisited , booktitle=

Sch\". Structure-from-Motion Revisited , booktitle=

work page
[69]

European Conference on Computer Vision (ECCV) , year=

Pan, Linfei and Barath, Daniel and Pollefeys, Marc and Sch\". European Conference on Computer Vision (ECCV) , year=

work page
[70]

Designing efficient sorting algorithms for manycore GPUs , year=

Satish, Nadathur and Harris, Mark and Garland, Michael , booktitle=. Designing efficient sorting algorithms for manycore GPUs , year=

work page
[71]

ACM Transactions on Graphics , year =

Modular Primitives for High-Performance Differentiable Rendering , author =. ACM Transactions on Graphics , year =

work page
[72]

Proceedings of the 35th International Conference on Neural Information Processing Systems , articleno =

Wang, Peng and Liu, Lingjie and Liu, Yuan and Theobalt, Christian and Komura, Taku and Wang, Wenping , title =. Proceedings of the 35th International Conference on Neural Information Processing Systems , articleno =. 2021 , isbn =

work page 2021
[73]

SIGGRAPH 2024 Conference Papers , year =

2D Gaussian Splatting for Geometrically Accurate Radiance Fields , author=. SIGGRAPH 2024 Conference Papers , year =

work page 2024