pith. sign in

arxiv: 2605.20185 · v2 · pith:DV7UFES4new · submitted 2026-05-19 · 💻 cs.GR · cs.CV

PiG-Avatar: Hierarchical Neural-Field-Guided Gaussian Avatars

Julian Kaltheuner , Jan Spindler , Sina Kitz , Patrick Stotko , Reinhard Klein This is my paper

Pith reviewed 2026-05-21 07:22 UTC · model grok-4.3

classification 💻 cs.GR cs.CV
keywords Gaussian avatarsneural fieldsvolumetric representationbarycentric transportself-organizationhierarchical reconstructionreal-time renderingclothing geometry
0
0 comments X

The pith

Avatars are modeled as Gaussians anchored in a volumetric neural-field space instead of on a body template surface.

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

The paper tries to establish that representing avatars as Gaussians in a continuous volumetric canonical space governed by a neural field, while using the body model only for kinematic transport, decouples geometry from template constraints and lets complex layered clothing emerge naturally. A sympathetic reader would care because prior surface-based methods limit off-body and non-rigid elements like loose clothing by entangling representation with deformation. The approach maintains motion coherence through 3D barycentric anchor transport that permits free deviation from the template while producing stable correspondences. Dual-level optimization with Sobolev-preconditioned field updates and KNN-based anchor preconditioning drives an emergent self-organization where anchors concentrate in high-curvature and high-variation regions. This single setup also enables hierarchical detail levels with shared supervision across scales.

Core claim

PiG-Avatar represents the avatar as Gaussians anchored in a volumetric canonical space governed by a continuous neural field, using the parametric body model solely for kinematic transport via 3D barycentric coordinates. This decouples the representation from template topology so anchors can deviate freely from the surface. Dual-level spatially coherent optimization combines Sobolev-preconditioned neural-field updates with KNN-based preconditioning of canonical anchors, which induces emergent self-organization of anchor density toward regions of high curvature, appearance variation, and non-coherent motion. The result is high-fidelity capture of complex clothing geometry and layered surfaces

What carries the argument

Neural-field-guided Gaussian anchors in volumetric canonical space, with dual-level spatially coherent optimization that combines Sobolev-preconditioned field updates and KNN-based anchor preconditioning to induce self-organization of density while using barycentric transport for kinematic coherence.

Load-bearing premise

The premise that Sobolev-preconditioned neural-field updates plus KNN-based preconditioning will reliably cause anchors to self-organize toward high-detail regions and preserve coherence even when anchors move freely away from the template surface.

What would settle it

After training on a sequence with loose clothing and large non-rigid motion, visualize final anchor positions; if density does not increase markedly in folds, wrinkles, and joint areas relative to smooth regions, or if rendering quality collapses under forced uniform anchor spacing, the self-organization mechanism would be shown not to work as claimed.

Figures

Figures reproduced from arXiv: 2605.20185 by Jan Spindler, Julian Kaltheuner, Patrick Stotko, Reinhard Klein, Sina Kitz.

Figure 1
Figure 1. Figure 1: We present PiG-Avatar, a Gaussian avatar method that decouples representation and deformation: the parametric model provides only kinematic transport, while a canonical neural field over learnable anchors is independent of template topology. Anchors self-organize in canonical space and, combined with time-conditioned neural features, produce temporally consistent posed splats, enabling complex, layered clo… view at source ↗
Figure 2
Figure 2. Figure 2: Existing methods rely on UV maps, surface/triangle binding, or skele [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Overview of PiG-Avatar. We learn a canonical, anchor-based Gaussian representation guided by a shared multi-resolution latent field. Conditioned on [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Emergent anchor density from our spatially coherent optimization. [PITH_FULL_IMAGE:figures/full_fig_p005_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Position and opacity remain stable across target LODs, whereas scale [PITH_FULL_IMAGE:figures/full_fig_p006_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Illustration of anchor transport through the deforming proxy mesh. [PITH_FULL_IMAGE:figures/full_fig_p007_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Qualitative comparison on DNA for novel-pose synthesis (top, 0165) [PITH_FULL_IMAGE:figures/full_fig_p008_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: LOD comparison of our shared hierarchical representation, showing [PITH_FULL_IMAGE:figures/full_fig_p009_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Qualitative structural ablations on DNA. Compared to our full model, [PITH_FULL_IMAGE:figures/full_fig_p009_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Robustness to noisy SMPL-X parameters. Left: ground-truth image. [PITH_FULL_IMAGE:figures/full_fig_p010_10.png] view at source ↗
read the original abstract

Existing Gaussian avatar methods typically parameterize geometry on a body-template surface, which entangles the avatar's representation space with the template's deformation space and limits the capture of layered, off-body, and non-rigid clothing geometry. We present PiG-Avatar, which addresses this limitation by using the parametric body model solely for kinematic transport, while representing the avatar as Gaussians anchored in a volumetric canonical space governed by a continuous neural field. This decouples representation from template topology, avoiding the geometric constraints of surface-based parameterizations. Kinematic coherence is maintained through 3D barycentric anchor transport, which guides motion without constraining geometry and allows anchors to deviate freely from the template surface, yielding dense, stable temporal surface correspondences by construction. To make this unconstrained formulation tractable, we introduce dual-level spatially coherent optimization, combining Sobolev-preconditioned neural-field updates with a novel KNN-based preconditioning of canonical anchor geometry. Together, these mechanisms induce an emergent self-organization of anchor density: anchors migrate toward regions of high curvature, appearance variation, and non-coherent motion without explicit heuristics. As a result, complex clothing geometry and layered surfaces emerge as natural, high-fidelity outputs. This single representation further supports hierarchical reconstruction across multiple levels of detail, with coarse-level supervision propagating to finer levels through the shared field and coupled anchor graph. On established benchmarks featuring subjects with complex clothing and challenging non-rigid motion, PiG-Avatar achieves state-of-the-art rendering quality, generalizes robustly to imperfect body model initialization, and renders in real time across all detail levels.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper introduces PiG-Avatar, a Gaussian avatar method that decouples geometry from the parametric body template by anchoring Gaussians in a volumetric canonical space governed by a continuous neural field. Kinematic coherence is maintained via 3D barycentric anchor transport that permits free deviation from the template surface, while dual-level spatially coherent optimization (Sobolev-preconditioned neural-field updates combined with KNN-based preconditioning of canonical anchors) induces emergent self-organization of anchor density toward high-curvature and non-coherent regions. The single representation supports hierarchical reconstruction across detail levels and is reported to achieve SOTA rendering quality, robustness to imperfect body-model initialization, and real-time performance on benchmarks with complex clothing and challenging non-rigid motion.

Significance. If the central claims hold, the work would be significant for neural avatar and Gaussian-splatting research by removing surface-topology constraints that limit capture of layered or off-body clothing. The combination of barycentric transport with preconditioned optimization for emergent density self-organization and hierarchical propagation offers a potentially generalizable route to more flexible, real-time avatars. Explicit credit is due for the parameter-free kinematic transport mechanism and the attempt at self-organizing anchor placement without explicit heuristics.

major comments (2)
  1. [Method (dual-level optimization) and Experiments (temporal stability)] The central claim that 3D barycentric transport plus dual-level preconditioning (Sobolev neural-field updates and KNN anchor preconditioning) reliably prevents kinematic drift when anchors deviate freely from the template surface is load-bearing for the decoupling argument. The manuscript provides no quantitative temporal-coherence metrics (e.g., anchor trajectory variance or correspondence error over long sequences) or failure-case analysis showing the regime in which the preconditioners cease to dominate free degrees of freedom.
  2. [Experiments and Ablations] Table reporting quantitative results on complex-clothing benchmarks: the SOTA claim is presented without ablations that isolate the contribution of KNN-based preconditioning versus Sobolev preconditioning alone. Without these controls it is unclear whether the observed self-organization of anchor density is emergent or an artifact of the specific optimization schedule.
minor comments (2)
  1. [Preliminaries / Method] Notation for the volumetric canonical space and the precise form of the neural field (e.g., input coordinates, output features) should be stated explicitly in the methods section to allow reproduction.
  2. [Figures] Figures illustrating anchor migration would benefit from side-by-side density histograms or curvature-correlation plots to substantiate the emergent self-organization claim.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback on our manuscript. The comments highlight key areas where additional evidence would strengthen the claims regarding kinematic stability and component contributions. We address each point below and will revise the manuscript accordingly to include the requested analyses.

read point-by-point responses

  1. Referee: [Method (dual-level optimization) and Experiments (temporal stability)] The central claim that 3D barycentric transport plus dual-level preconditioning (Sobolev neural-field updates and KNN anchor preconditioning) reliably prevents kinematic drift when anchors deviate freely from the template surface is load-bearing for the decoupling argument. The manuscript provides no quantitative temporal-coherence metrics (e.g., anchor trajectory variance or correspondence error over long sequences) or failure-case analysis showing the regime in which the preconditioners cease to dominate free degrees of freedom.

    Authors: We agree that explicit quantitative temporal-coherence metrics are not reported in the current manuscript. The 3D barycentric transport is intended to enforce coherence by construction through dense correspondences, and the dual-level preconditioning is designed to stabilize deviations. However, to directly address the concern, we will add quantitative evaluations in the revision, including anchor trajectory variance, long-sequence correspondence error, and analysis of failure regimes where preconditioners may be insufficient. revision: yes

  2. Referee: [Experiments and Ablations] Table reporting quantitative results on complex-clothing benchmarks: the SOTA claim is presented without ablations that isolate the contribution of KNN-based preconditioning versus Sobolev preconditioning alone. Without these controls it is unclear whether the observed self-organization of anchor density is emergent or an artifact of the specific optimization schedule.

    Authors: We acknowledge the value of isolating the individual contributions of KNN-based and Sobolev preconditioning to clarify whether self-organization is truly emergent. The current results demonstrate the combined effect, but we will incorporate new ablation experiments in the revised manuscript that disable each preconditioner separately while measuring impacts on anchor density distribution, self-organization behavior, and final rendering quality. revision: yes

Circularity Check

0 steps flagged

No significant circularity; derivation introduces independent mechanisms

full rationale

The paper's core derivation relies on newly proposed components: a volumetric canonical space governed by a continuous neural field, 3D barycentric anchor transport for kinematic coherence without surface constraints, and dual-level spatially coherent optimization (Sobolev-preconditioned neural-field updates plus KNN-based preconditioning). These are presented as addressing limitations of prior surface-based Gaussian avatar methods without reducing by the paper's own equations or self-citations to quantities already fitted or defined in the inputs. The abstract explicitly frames outcomes like emergent anchor density self-organization and stable temporal correspondences as results of the unconstrained formulation plus preconditioners, rather than tautological redefinitions or fitted-input predictions. No load-bearing step is shown to collapse into prior work via self-citation chains or ansatz smuggling. The central claims retain independent content from the described architecture and optimization strategy.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 2 invented entities

Review is limited to the abstract, so the ledger is necessarily incomplete; the paper introduces several new technical constructs whose precise parameter counts and supporting assumptions cannot be audited without the full text.

axioms (1)
  • domain assumption The parametric body model can be used solely for kinematic transport without imposing geometric constraints on the avatar representation.
    Invoked when the abstract states that the body model is used only for transport while geometry lives in an unconstrained volumetric space.
invented entities (2)
  • volumetric canonical space governed by a continuous neural field no independent evidence
    purpose: To host Gaussian anchors independently of the body-template surface topology.
    Core representational choice that enables decoupling from surface-based constraints.
  • dual-level spatially coherent optimization with Sobolev and KNN preconditioning no independent evidence
    purpose: To render the unconstrained formulation tractable and induce emergent anchor self-organization.
    Novel optimization technique introduced to stabilize training without explicit density heuristics.

pith-pipeline@v0.9.0 · 5831 in / 1500 out tokens · 51783 ms · 2026-05-21T07:22:29.333917+00:00 · methodology

Review history (2 revisions) →

discussion (0)

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

Lean theorems connected to this paper

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

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

Reference graph

Works this paper leans on

135 extracted references · 135 canonical work pages

  1. [1]

    Kaltheuner, Julian and Dr

    work page

  2. [2]

    Li, Yang and Takehara, Hikari and Taketomi, Takafumi and Zheng, Bo and Nie

    work page

  3. [3]

    2008 , publisher=

    Vlasic, Daniel and Baran, Ilya and Matusik, Wojciech and Popovi. 2008 , publisher=

    work page 2008

  4. [4]

    Bogo, Federica and Romero, Javier and Pons-Moll, Gerard and Black, Michael J , booktitle=CVPR, year=

    work page

  5. [5]

    Kaltheuner, Julian and Oebel, Alexander and Dr

    work page

  6. [6]

    Yao, Yuxin and Ren, Siyu and Hou, Junhui and Deng, Zhi and Zhang, Juyong and Wang, Wenping , booktitle=ECCV, year=

    work page

  7. [7]

    Cao, Wei and Luo, Chang and Zhang, Biao and Nie

    work page

  8. [8]

    Lei, Jiahui and Daniilidis, Kostas , booktitle=CVPR, year=

    work page

  9. [9]

    Tancik, Matthew and Srinivasan, Pratul and Mildenhall, Ben and Fridovich-Keil, Sara and Raghavan, Nithin and Singhal, Utkarsh and Ramamoorthi, Ravi and Barron, Jonathan and Ng, Ren , journal=NeurIPS, volume=

    work page

  10. [10]

    Kingma, Diederik P and Ba, Jimmy , booktitle=ICLR, year=

    work page

  11. [11]

    2013 , publisher=

    Kazhdan, Michael and Hoppe, Hugues , journal=TOG, volume=. 2013 , publisher=

    work page 2013

  12. [12]

    2021 , organization=

    Zhang, Jiayi Eris and Jacobson, Alec and Alexa, Marc , journal=CGF, volume=. 2021 , organization=

    work page 2021

  13. [13]

    2021 , publisher=

    Nicolet, Baptiste and Jacobson, Alec and Jakob, Wenzel , journal=TOG, volume=. 2021 , publisher=

    work page 2021

  14. [14]

    Bozic, Aljaz and Palafox, Pablo and Zollhofer, Michael and Thies, Justus and Dai, Angela and Nie

    work page

  15. [15]

    Bozic, Aljaz and Palafox, Pablo and Zollh

    work page

  16. [16]

    2015 , publisher=

    Loper, Matthew and Mahmood, Naureen and Romero, Javier and Pons-Moll, Gerard and Black, Michael J , journal=TOG, volume=. 2015 , publisher=

    work page 2015

  17. [17]

    Li, Tianye and Bolkart, Timo and Black, Michael J and Li, Hao and Romero, Javier , journal=TOG, volume=

    work page

  18. [18]

    2017 , publisher=

    Romero, Javier and Tzionas, Dimitrios and Black, Michael J , journal=TOG, volume=. 2017 , publisher=

    work page 2017

  19. [19]

    Pavlakos, Georgios and Choutas, Vasileios and Ghorbani, Nima and Bolkart, Timo and Osman, Ahmed AA and Tzionas, Dimitrios and Black, Michael J , booktitle=CVPR, year=

    work page

  20. [20]

    Wang, Lizhen and Chen, Zhiyuan and Yu, Tao and Ma, Chenguang and Li, Liang and Liu, Yebin , booktitle=CVPR, year=

    work page

  21. [21]

    2005 , publisher=

    Anguelov, Dragomir and Srinivasan, Praveen and Koller, Daphne and Thrun, Sebastian and Rodgers, Jim and Davis, James , journal=TOG, volume=. 2005 , publisher=

    work page 2005

  22. [22]

    Xu, Hongyi and Bazavan, Eduard Gabriel and Zanfir, Andrei and Freeman, William T and Sukthankar, Rahul and Sminchisescu, Cristian , booktitle=CVPR, year=

    work page

  23. [23]

    Li, Yang and Harada, Tatsuya , journal=NeurIPS, volume=

    work page

  24. [24]

    2007 , publisher=

    Embedded deformation for shape manipulation , author=. 2007 , publisher=

    work page 2007

  25. [25]

    Huang, Jiahui and Gojcic, Zan and Atzmon, Matan and Litany, Or and Fidler, Sanja and Williams, Francis , booktitle=CVPR, year=

    work page

  26. [26]

    Peng, Songyou and Jiang, Chiyu and Liao, Yiyi and Niemeyer, Michael and Pollefeys, Marc and Geiger, Andreas , journal=NeurIPS, volume=

    work page

  27. [27]

    2020 , publisher=

    Hanocka, Rana and Metzer, Gal and Giryes, Raja and Cohen-Or, Daniel , journal=TOG, volume=. 2020 , publisher=

    work page 2020

  28. [28]

    Williams, Francis and Schneider, Teseo and Silva, Claudio and Zorin, Denis and Bruna, Joan and Panozzo, Daniele , booktitle=CVPR, year=

    work page

  29. [29]

    Niemeyer, Michael and Mescheder, Lars and Oechsle, Michael and Geiger, Andreas , booktitle=ICCV, year=

    work page

  30. [30]

    Yenamandra, Tarun and Tewari, Ayush and Bernard, Florian and Seidel, Hans-Peter and Elgharib, Mohamed and Cremers, Daniel and Theobalt, Christian , booktitle=CVPR, year=

    work page

  31. [31]

    Bednarik, Jan and Kim, Vladimir G and Chaudhuri, Siddhartha and Parashar, Shaifali and Salzmann, Mathieu and Fua, Pascal and Aigerman, Noam , booktitle=ICCV, year=

    work page

  32. [32]

    Jiang, Boyan and Zhang, Yinda and Wei, Xingkui and Xue, Xiangyang and Fu, Yanwei , booktitle=CVPR, year=

    work page

  33. [33]

    Chen, Ricky TQ and Rubanova, Yulia and Bettencourt, Jesse and Duvenaud, David K , journal=NeurIPS, volume=

    work page

  34. [34]

    Huang, Jingwei and Jiang, Chiyu Max and Leng, Baiqiang and Wang, Bin and Guibas, Leonidas , journal=

    work page

  35. [35]

    Li, Zhengqi and Niklaus, Simon and Snavely, Noah and Wang, Oliver , booktitle=CVPR, year=

    work page

  36. [36]

    Tang, Jiapeng and Xu, Dan and Jia, Kui and Zhang, Lei , booktitle=CVPR, year=

    work page

  37. [37]

    Pumarola, Albert and Corona, Enric and Pons-Moll, Gerard and Moreno-Noguer, Francesc , booktitle=CVPR, year=

    work page

  38. [38]

    Park, Keunhong and Sinha, Utkarsh and Barron, Jonathan T and Bouaziz, Sofien and Goldman, Dan B and Seitz, Steven M and Martin-Brualla, Ricardo , booktitle=ICCV, year=

    work page

  39. [39]

    Newcombe, Richard A and Fox, Dieter and Seitz, Steven M , booktitle=CVPR, year=

    work page

  40. [40]

    Bhatnagar, Bharat Lal and Tiwari, Garvita and Theobalt, Christian and Pons-Moll, Gerard , booktitle=ICCV, year=

    work page

  41. [41]

    Bogo, Federica and Kanazawa, Angjoo and Lassner, Christoph and Gehler, Peter and Romero, Javier and Black, Michael J , booktitle=ECCV, year=

    work page

  42. [42]

    Sanyal, Soubhik and Bolkart, Timo and Feng, Haiwen and Black, Michael J , booktitle=CVPR, year=

    work page

  43. [43]

    2021 , publisher=

    Liu, Lingjie and Habermann, Marc and Rudnev, Viktor and Sarkar, Kripasindhu and Gu, Jiatao and Theobalt, Christian , journal=TOG, volume=. 2021 , publisher=

    work page 2021

  44. [44]

    Alldieck, Thiemo and Magnor, Marcus and Bhatnagar, Bharat Lal and Theobalt, Christian and Pons-Moll, Gerard , booktitle=CVPR, year=

    work page

  45. [45]

    Tiwari, Garvita and Bhatnagar, Bharat Lal and Tung, Tony and Pons-Moll, Gerard , booktitle=ECCV, year=

    work page

  46. [46]

    Efficient Non-linear Optimization via Multi-scale Gradient Filtering , author=

    work page

  47. [47]

    2004 , publisher=

    Renka, Robert J , journal=CAGD, volume=. 2004 , publisher=

    work page 2004

  48. [48]

    1995 , publisher=

    Renka, Robert J and Neuberger, JW , journal=SIAM, volume=. 1995 , publisher=

    work page 1995

  49. [49]

    Slavcheva, Miroslava and Baust, Maximilian and Ilic, Slobodan , booktitle=CVPR, year=

    work page

  50. [50]

    Jung, Yucheol and Kim, Hyomin and Yoon, Hyejeong and Lee, Seungyong , booktitle=CGF, year=

    work page

  51. [51]

    Chang, Wesley and Yang, Xuanda and Belhe, Yash and Ramamoorthi, Ravi and Li, Tzu-Mao , booktitle=SIGGRAPH_ASIA_CONF, year=

    work page

  52. [52]

    1985 , publisher=

    Neuberger, JW , journal=. 1985 , publisher=

    work page 1985

  53. [53]

    Eckstein, Ilya and Pons, J-P and Tong, Yiying and Kuo, C-CJ and Desbrun, Mathieu , booktitle=SGP, year=

    work page

  54. [54]

    2013 , publisher=

    Krishnan, Dilip and Fattal, Raanan and Szeliski, Richard , journal=TOG, volume=. 2013 , publisher=

    work page 2013

  55. [55]

    Claici, Sebastian and Bessmeltsev, Mikhail and Schaefer, Scott and Solomon, Justin , booktitle=CGF, volume=

    work page

  56. [56]

    Chen, Jie , booktitle=ICLR, year=

    work page

  57. [57]

    Rudikov, Alexander and Fanaskov, Vladimir and Muravleva, Ekaterina and Laevsky, Yuri M and Oseledets, Ivan , booktitle=ICML, year=

    work page

  58. [58]

    Li, Yichen and Chen, Peter Yichen and Du, Tao and Matusik, Wojciech , booktitle=ICML, year=

    work page

  59. [59]

    Trifonov, Vladislav and Rudikov, Alexander and Iliev, Oleg and Laevsky, Yuri M and Oseledets, Ivan and Muravleva, Ekaterina , journal=

    work page

  60. [60]

    2005 , publisher=

    Aksoylu, Burak and Khodakovsky, Andrei and Schr. 2005 , publisher=

    work page 2005

  61. [61]

    2016 , publisher=

    Kovalsky, Shahar Z and Galun, Meirav and Lipman, Yaron , journal=TOG, volume=. 2016 , publisher=

    work page 2016

  62. [62]

    2019 , organization=

    Wu, Zhenchao and Li, Kun and Lai, Yu-Kun and Yang, Jingyu , booktitle=. 2019 , organization=

    work page 2019

  63. [63]

    2019 , publisher=

    Yang, Jingyu and Guo, Daoliang and Li, Kun and Wu, Zhenchao and Lai, Yu-Kun , journal=TIP, volume=. 2019 , publisher=

    work page 2019

  64. [64]

    Dynamic Neural Surfaces for Elastic 4D Shape Representation and Analysis , author=

    work page

  65. [65]

    4DTAM: Non-Rigid Tracking and Mapping via Dynamic Surface Gaussians , author=

    work page

  66. [66]

    WildGS-SLAM: Monocular Gaussian Splatting SLAM in Dynamic Environments , author=

    work page

  67. [67]

    DynaGSLAM: Real-Time Gaussian-Splatting SLAM for Online Rendering, Tracking, Motion Predictions of Moving Objects in Dynamic Scenes , author=

    work page

  68. [68]

    3D Gaussian splatting for real-time radiance field rendering , author=

    work page

  69. [69]

    Mihajlovic, Marko and Zhang, Siwei and Li, Gen and Zhao, Kaifeng and Muller, Lea and Tang, Siyu , booktitle=ICCV, year=

    work page

  70. [70]

    Park, Jinhyung and Romero, Javier and Saito, Shunsuke and Prada, Fabian and Shiratori, Takaaki and Xu, Yichen and Bogo, Federica and Yu, Shoou-I and Kitani, Kris and Khirodkar, Rawal , booktitle=ICCV, year=

    work page

  71. [71]

    2022 , publisher=

    M. 2022 , publisher=

    work page 2022

  72. [72]

    Chen, Anpei and Xu, Zexiang and Geiger, Andreas and Yu, Jingyi and Su, Hao , booktitle=ECCV, year=

    work page

  73. [73]

    Fridovich-Keil, Sara and Meanti, Giacomo and Warburg, Frederik Rahb

    work page

  74. [74]

    Cao, Ang and Johnson, Justin , booktitle=CVPR, year=

    work page

  75. [75]

    Zhu, Jiaxuan and Tang, Hao , journal=

    work page

  76. [76]

    Bae, Jeongmin and Kim, Seoha and Yun, Youngsik and Lee, Hahyun and Bang, Gun and Uh, Youngjung , booktitle=ECCV, year=

    work page

  77. [77]

    Chen, Honghu and Peng, Bo and Tao, Yunfan and Zhang, Juyong , booktitle=CVPR, year=

    work page

  78. [78]

    Oh, Jong Kwon and Lyu, Hanbaek and Son, Hwijae , journal=

    work page

  79. [79]

    2025 , publisher=

    Cho, Namkyeong and Ryu, Junseung and Hwang, Hyung Ju , journal=. 2025 , publisher=

    work page 2025

  80. [80]

    2024 , publisher=

    Kilicsoy, AOM and Liedmann, J and Valdebenito, MA and Barthold, F-J and Faes, MGR , journal=. 2024 , publisher=

    work page 2024

Showing first 80 references.