pith. sign in

arxiv: 2511.16988 · v2 · submitted 2025-11-21 · 💻 cs.GR

PhysMorph-GS: Render-Guided Volumetric Morphing with Differentiable Physics

Chang-Yong Song , David Hyde This is my paper

Pith reviewed 2026-05-17 20:54 UTC · model grok-4.3

classification 💻 cs.GR
keywords volumetric morphingdifferentiable physics3D Gaussian splattingmaterial point methoddeformation gradientrender-guided simulationphased plasticity
0
0 comments X

The pith

By routing visual supervision through the deformation gradient rather than particle positions, render-guided morphing keeps physics trajectories intact while capturing fine details.

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

The paper establishes a framework for target-driven volumetric shape morphing in differentiable physics simulations. Direct coupling of image losses to particle positions tends to destabilize elastic dynamics, while physics-only mass matching misses fine geometric detail. PhysMorph-GS solves this by injecting visual supervision from differentiable 3D Gaussian splatting through the deformation gradient F. Render gradients thus provide control-space guidance, and trajectories stay governed by physics. Phased Chamfer-guided plasticity further delays rest-state migration until coarse structure forms, and rendering uses a surface-focused particle subset for efficiency.

Core claim

The central claim is that the main challenge in render-guided differentiable morphing is not simply adding stronger image losses but injecting visual guidance compatibly with elastic simulation. This is achieved by supervising through the deformation gradient F in a coupled MPM and 3D Gaussian splatting setup, with phased plasticity, resulting in different sources being driven toward a shared target-determined attractor by plasticity-driven rest-state migration.

What carries the argument

The deformation gradient F serving as the conduit for render gradients from 3D Gaussian splatting into MPM simulation, alongside phased Chamfer-guided plasticity.

If this is right

  • Silhouette error reduces by 25.8%, 10.8%, and 49.9% on representative examples relative to a physics-only baseline.
  • Largest gains occur on models with thin features.
  • Plasticity-driven rest-state migration distinguishes physics-based morphing from interpolation between registered shape pairs.
  • Surface-focused particle subset for rendering improves efficiency and concentrates gradients.

Where Pith is reading between the lines

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

  • This routing principle could be tested in other particle or mesh-based simulators to see if visual fidelity improves without losing physical plausibility.
  • The method implies that compatible gradient injection may be more important than loss strength in differentiable physics applications.
  • Extensions to dynamic scene reconstruction from video could use similar deformation gradient supervision to recover plausible motions for thin structures.

Load-bearing premise

Routing render gradients through the deformation gradient F remains compatible with elastic simulation dynamics and does not introduce instability or require extensive tuning of the phased plasticity schedule.

What would settle it

Measuring silhouette error and checking for simulation instability when applying the method to additional thin-featured models would test if the error reductions hold or if dynamics break.

Figures

Figures reproduced from arXiv: 2511.16988 by Chang-Yong Song, David Hyde.

Figure 1
Figure 1. Figure 1: Pipeline overview. (1) Deformation control: Differentiable MPM evolves anchor particles (blue spheres) under log-based grid mass loss Lmass to maintain material conservation. (2) Surface-aware upsampling: Subdivision upsampling spawns child particles (cyan) in proportion to local deformation magnitude | det(F) − 1|, upsampling from sparse anchors to a dense set of render particles. A multi￾scale F-field th… view at source ↗
Figure 2
Figure 2. Figure 2: Subdivision upsampling process (left to right). (a) Original shape: Initial anchors (8,918 particles) represent the coarse geometry. (b) Parent selection: 99 parents are chosen based on high local deformation, visualized by color. (c) Parent–child relationships: For clarity, we show only the top 20 high-deformation parents (red) and their associated children (orange), highlighting how subdivision concentra… view at source ↗
Figure 3
Figure 3. Figure 3: Particle distribution after subdivision (left: x, mid￾dle: y, right: z). The adaptive scheme generates a spatially con￾tinuous and dense particle field. By concentrating resolution in high-deformation regions, the method ensures smooth coverage without clustering or sparsity artifacts (see [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 5
Figure 5. Figure 5: Rendering loss evolution. Top: Lα localizes errors along shape boundaries (t = 0, 10, 40). Bottom: depth hits visu￾alize Lshrink activation for interior pruning. The combined super￾vision ensures clean silhouettes and removes internal artifacts as morphing progresses [PITH_FULL_IMAGE:figures/full_fig_p006_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Impact of Material Stiffness on Optimization Be￾havior. Optimization results under varying constitutive param￾eters. (Left) Hard Material (E ≈ 1.4 × 105 , ν ≈ 0.20): High stiffness restricts deformation, resulting in low anisotropy (mean ≈ 1.02) and rigid-body motion. (Center) Medium Mate￾rial (E ≈ 2.8 × 104 , ν ≈ 0.40): Intermediate stiffness shows transitional behavior with moderate stretching. (Right) S… view at source ↗
Figure 7
Figure 7. Figure 7: High resolution results with Physics-Guided Shape Morphing. Snapshots from source sphere into targets. PhysMorph-GS successfully guides the physical simulation to match the target geometry. Starting from a coarse MPM initializa￾tion, the pipeline progressively optimizes deformation gradients and positions. Note how the system captures fine geometric de￾tails (ears, tail) while maintaining physical connecti… view at source ↗
Figure 8
Figure 8. Figure 8: Training convergence. Physics loss (main objective) decreases 96.5%, while depth loss (geometric guidance) reduces 75.4%. Alpha loss functions as a boundary regularizer, maintain￾ing low values (∼0.06) to preserve silhouette sharpness without dominating optimization [PITH_FULL_IMAGE:figures/full_fig_p008_8.png] view at source ↗
read the original abstract

Differentiable particle-based simulation can produce physically plausible motion, but target-driven volumetric shape morphing remains underconstrained: physics-only mass matching captures coarse global structure yet struggles with fine geometric detail, while naive image-space coupling destabilizes elastic dynamics. We present PhysMorph-GS, a render-guided morphing framework that couples material point method simulation with differentiable 3D Gaussian splatting. The key idea is to inject visual supervision through the deformation gradient $\mathbf{F}$ rather than particle positions, so render gradients act as control-space guidance while trajectories remain governed by physics. We further introduce phased Chamfer-guided plasticity that delays rest-state migration until coarse structure has formed; in practice, rendering is evaluated on a surface-focused particle subset for efficiency and gradient concentration. Relative to a physics-only baseline, our method reduces silhouette error by 25.8\%, 10.8\%, and 49.9\% on representative examples, with the largest gains on models with thin features. These results suggest that the main challenge in render-guided differentiable morphing is not simply adding stronger image losses, but injecting visual guidance in a way that remains compatible with elastic simulation. We further observe that plasticity-driven rest-state migration drives different sources toward a shared target-determined attractor, distinguishing physics-based morphing from interpolation between registered shape pairs.

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 / 3 minor

Summary. The paper claims to introduce PhysMorph-GS, a framework coupling material point method (MPM) simulation with differentiable 3D Gaussian splatting for target-driven volumetric morphing. The key mechanism routes visual supervision through the deformation gradient F (rather than particle positions) so that render gradients provide control-space guidance while trajectories remain physics-governed. It further introduces phased Chamfer-guided plasticity to delay rest-state migration until coarse structure forms and evaluates rendering on a surface-focused particle subset. Relative to a physics-only baseline, the method reports silhouette error reductions of 25.8%, 10.8%, and 49.9% on representative examples, with largest gains on thin-featured models. The work concludes that the main challenge is injecting visual guidance compatibly with elastic simulation rather than simply strengthening image losses.

Significance. If the central mechanism holds, the paper offers a concrete approach to making render losses compatible with elastic MPM dynamics, which could benefit animation, shape optimization, and physics-based rendering pipelines. The explicit coupling via F and the observation that plasticity drives convergence to a target-determined attractor distinguish this from pure geometric interpolation. The reported quantitative improvements on thin features are a positive signal, though the absence of isolating ablations and stability diagnostics limits immediate generalizability.

major comments (2)
  1. [Abstract] Abstract: the reported silhouette error reductions (25.8%, 10.8%, 49.9%) are obtained with the full method that includes both F-supervision and the phased Chamfer-guided plasticity schedule. No ablation is shown that applies F-routing versus direct position supervision while holding the plasticity phasing fixed; this is load-bearing for the claim that routing through F specifically ensures compatibility with elastic dynamics.
  2. [Abstract] Abstract: no quantitative stability diagnostics (e.g., minimum det(F), strain-energy bounds, or divergence counts) are provided for the thin-feature cases that exhibit the largest gain (49.9%). The introduction of phased plasticity as a practical necessity leaves open whether the F-route alone keeps the MPM constitutive model inside its stable regime (det(F) > 0, bounded energy) on those examples.
minor comments (3)
  1. The abstract refers to 'representative examples' without naming the specific models, datasets, or mesh resolutions used.
  2. No error bars, number of runs, or statistical details accompany the percentage error reductions.
  3. The precise criteria for selecting the surface-focused particle subset and the exact schedule parameters for phased plasticity are not specified, hindering reproducibility.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the thoughtful and detailed report. The comments highlight important aspects of our claims regarding the role of deformation-gradient routing and the need for supporting diagnostics. We address each major comment below and commit to revisions that strengthen the manuscript without altering its core contributions.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the reported silhouette error reductions (25.8%, 10.8%, 49.9%) are obtained with the full method that includes both F-supervision and the phased Chamfer-guided plasticity schedule. No ablation is shown that applies F-routing versus direct position supervision while holding the plasticity phasing fixed; this is load-bearing for the claim that routing through F specifically ensures compatibility with elastic dynamics.

    Authors: We agree that the reported gains reflect the combined system and that an isolating ablation is necessary to substantiate the specific benefit of routing supervision through F. In the revised manuscript we will add a controlled ablation that applies F-routing versus direct particle-position supervision while holding the phased Chamfer-guided plasticity schedule fixed. This addition will directly address the load-bearing claim about compatibility with elastic MPM dynamics. revision: yes

  2. Referee: [Abstract] Abstract: no quantitative stability diagnostics (e.g., minimum det(F), strain-energy bounds, or divergence counts) are provided for the thin-feature cases that exhibit the largest gain (49.9%). The introduction of phased plasticity as a practical necessity leaves open whether the F-route alone keeps the MPM constitutive model inside its stable regime (det(F) > 0, bounded energy) on those examples.

    Authors: We acknowledge that explicit stability metrics were not reported. Although successful convergence on the thin-feature examples indicates that the simulation remained stable, we will augment the revision with quantitative diagnostics (minimum det(F), strain-energy bounds, and divergence counts) specifically for those cases. These metrics will clarify that the F-route maintains the constitutive model within its stable regime even when plasticity phasing is held constant. revision: yes

Circularity Check

0 steps flagged

No significant circularity; method introduces explicit new coupling without reducing claims to inputs by construction

full rationale

The paper's core proposal routes render gradients through the deformation gradient F as an explicit new supervisory mechanism while adding phased Chamfer-guided plasticity as a practical schedule to maintain stability. No equations, predictions, or first-principles derivations in the abstract or described framework reduce a claimed result to a fitted parameter, self-citation chain, or renamed input by construction. The reported error reductions are presented as empirical outcomes of the introduced coupling rather than tautological re-derivations, and the derivation remains self-contained against external simulation and rendering benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review; no explicit free parameters, axioms, or invented entities are quantified. The phased Chamfer-guided plasticity and surface-focused subset are introduced as new algorithmic choices whose precise parameterization is not stated.

pith-pipeline@v0.9.0 · 5532 in / 1158 out tokens · 34307 ms · 2026-05-17T20:54:42.015128+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

50 extracted references · 50 canonical work pages

  1. [1]

    Adaptively sampled particle fluids

    Bart Adams, Mark Pauly, Richard Keiser, and Leonidas J Guibas. Adaptively sampled particle fluids. InACM SIG- GRAPH, page 48, 2007. 1

    work page 2007

  2. [2]

    As-rigid- as-possible shape interpolation

    Marc Alexa, Daniel Cohen-Or, and David Levin. As-rigid- as-possible shape interpolation. InSeminal Graphics Papers: Pushing the Boundaries, Volume 2, pages 165–172. 2023. 2

    work page 2023

  3. [3]

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

    Jonathan T Barron, Ben Mildenhall, Matthew Tancik, Peter Hedman, Ricardo Martin-Brualla, and Pratul P Srinivasan. Mip-nerf: A multiscale representation for anti-aliasing neu- ral radiance fields. InInt. Conf. Comput. Vis., pages 5855– 5864, 2021. 2

    work page 2021

  4. [4]

    Projective dynamics: Fusing constraint projections for fast simulation

    Sofien Bouaziz, Sebastian Martin, Tiantian Liu, Ladislav Ka- van, and Mark Pauly. Projective dynamics: Fusing constraint projections for fast simulation. InSeminal Graphics Papers: Pushing the Boundaries, Volume 2, pages 787–797. 2023. 2

    work page 2023

  5. [5]

    Cover and P

    T. Cover and P. Hart. Nearest neighbor pattern classifica- tion.IEEE Transactions on Information Theory, 13(1):21– 27, 1967. 4

    work page 1967

  6. [6]

    A differentiable physics engine for deep learning in robotics.Frontiers in Neurorobotics, 13:6, 2019

    Jonas Degrave, Michiel Hermans, Joni Dambre, and Francis Wyffels. A differentiable physics engine for deep learning in robotics.Frontiers in Neurorobotics, 13:6, 2019. 2

    work page 2019

  7. [7]

    Diffpd: Dif- ferentiable projective dynamics.ACM Trans

    Tao Du, Kui Wu, Pingchuan Ma, Sebastien Wah, Andrew Spielberg, Daniela Rus, and Wojciech Matusik. Diffpd: Dif- ferentiable projective dynamics.ACM Trans. Graph., 41(2): 1–21, 2021. 2

    work page 2021

  8. [8]

    Target-driven smoke an- imation.ACM Trans

    Raanan Fattal and Dani Lischinski. Target-driven smoke an- imation.ACM Trans. Graph., 23(3):441–448, 2004. 2

    work page 2004

  9. [9]

    Dynamic 3d gaussian fields for urban areas

    Tobias Fischer et al. Dynamic 3d gaussian fields for urban areas. InInt. Conf. Comput. Vis., 2024. 2

    work page 2024

  10. [10]

    Superpoint gaussian splatting for real-time high-fidelity dynamic scene reconstruction

    Diwen Gao, Jingdong Xu, Ziwei Qiao, et al. Superpoint gaussian splatting for real-time high-fidelity dynamic scene reconstruction. InInt. Conf. Mach. Learn., 2024. 2

    work page 2024

  11. [11]

    A moving least squares material point method with displacement discontinu- ity and two-way rigid body coupling.ACM Trans

    Yuanming Hu, Yu Fang, Ziheng Ge, Ziyin Qu, Yixin Zhu, Andre Pradhana, and Chenfanfu Jiang. A moving least squares material point method with displacement discontinu- ity and two-way rigid body coupling.ACM Trans. Graph., 37(4):150, 2018. 2, 9

    work page 2018

  12. [12]

    Difftaichi: Differentiable programming for physical simulation.arXiv preprint arXiv:1910.00935, 2019

    Yuanming Hu, Luke Anderson, Tzu-Mao Li, Qi Sun, Nathan Carr, Jonathan Ragan-Kelley, and Fr´edo Durand. Difftaichi: Differentiable programming for physical simulation.arXiv preprint arXiv:1910.00935, 2019. 1, 2

    work page arXiv 1910

  13. [13]

    Tenenbaum, William T

    Yuanming Hu, Jiancheng Liu, Andrew Spielberg, Joshua B. Tenenbaum, William T. Freeman, Jiajun Wu, Daniela Rus, and Wojciech Matusik. Chainqueen: A real-time differen- tiable physical simulator for soft robotics. InIEEE Int. Conf. Robot. Autom., pages 6265–6271, 2019. 2

    work page 2019

  14. [14]

    The material point method for simulating continuum materials

    Chenfanfu Jiang, Craig Schroeder, Joseph Teran, Alexey Stomakhin, and Andrew Selle. The material point method for simulating continuum materials. InACM SIGGRAPH 2016 Courses, pages 1–52. 2016. 1, 2, 9

    work page 2016

  15. [15]

    Billion- scale similarity search with GPUs.IEEE Transactions on Big Data, 7(3):535–547, 2019

    Jeff Johnson, Matthijs Douze, and Herv ´e J ´egou. Billion- scale similarity search with GPUs.IEEE Transactions on Big Data, 7(3):535–547, 2019. 4

    work page 2019

  16. [16]

    Splatam: Splat track & map 3d gaussians for dense rgb-d slam

    Nikhil Keetha, Jay Karhade, Krishna Murthy Jatavallabhula, Gengshan Yang, Sebastian Scherer, Deva Ramanan, and Jonathon Luiten. Splatam: Splat track & map 3d gaussians for dense rgb-d slam. InIEEE Conf. Comput. Vis. Pattern Recog., pages 21357–21366, 2024. 2

    work page 2024

  17. [17]

    3d gaussian splatting for real-time radiance field rendering.ACM Trans

    Bernhard Kerbl, Georgios Kopanas, Thomas Leimk ¨uhler, and George Drettakis. 3d gaussian splatting for real-time radiance field rendering.ACM Trans. Graph., 42(4):139:1– 139:14, 2023. 1, 2

    work page 2023

  18. [18]

    Kingma and Jimmy Ba

    Diederik P. Kingma and Jimmy Ba. Adam: A method for stochastic optimization, 2017. 9

    work page 2017

  19. [19]

    Modular primitives for high-performance differentiable rendering.ACM Trans

    Samuli Laine, Janne Hellsten, Tero Karras, Yeongho Seol, Jaakko Lehtinen, and Timo Aila. Modular primitives for high-performance differentiable rendering.ACM Trans. Graph., 39(6):1–14, 2020. 2

    work page 2020

  20. [20]

    Fully explicit dynamic gaussian splat- ting

    Junoh Lee, Changyeon Won, Hyunjun Jung, Inhwan Bae, and Hae-Gon Jeon. Fully explicit dynamic gaussian splat- ting. InAdv. Neural Inform. Process. Syst., 2024. 2

    work page 2024

  21. [21]

    Pac- nerf: Physics augmented continuum neural radiance fields for geometry-agnostic system identification

    Xuan Li, Yi-Ling Qiao, Peter Yichen Chen, et al. Pac- nerf: Physics augmented continuum neural radiance fields for geometry-agnostic system identification. InInt. Conf. Learn. Represent., 2023. 2, 8

    work page 2023

  22. [22]

    Diffcloth: Differentiable cloth simulation with dry frictional contact.ACM Trans

    Yifei Li, Tao Du, Kui Wu, Jie Xu, and Wojciech Matusik. Diffcloth: Differentiable cloth simulation with dry frictional contact.ACM Trans. Graph., 42(1):2:1–2:20, 2022. 2

    work page 2022

  23. [23]

    Omniphysgs: 3d con- stitutive gaussians for general physics-based dynamics generation.arXiv preprint arXiv:2501.18982,

    Yuchen Lin, Chenguo Lin, Jianjin Xu, and Yadong Mu. Omniphysgs: 3d constitutive gaussians for gen- eral physics-based dynamics generation.arXiv preprint arXiv:2501.18982, 2025. 1, 2, 3, 8 13

    work page arXiv 2025

  24. [24]

    Marching cubes: A high resolution 3d surface construction algorithm

    William E Lorensen and Harvey E Cline. Marching cubes: A high resolution 3d surface construction algorithm. InSem- inal graphics: pioneering efforts that shaped the field, pages 347–353. 1998. 1, 2

    work page 1998

  25. [25]

    Space-time sculpting of liq- uid animation

    Pierre-Luc Manteaux, Ulysse Vimont, Chris Wojtan, Damien Rohmer, and Marie-Paule Cani. Space-time sculpting of liq- uid animation. InInt. Conf. on Motion in Games (MIG), pages 61–71, 2016. 2

    work page 2016

  26. [26]

    Fluid control using the adjoint method.ACM Trans

    Antoine McNamara, Adrien Treuille, Zoran Popovi´c, and Jos Stam. Fluid control using the adjoint method.ACM Trans. Graph., 23(3):449–456, 2004. 2

    work page 2004

  27. [27]

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

    Ben Mildenhall, Pratul P Srinivasan, Matthew Tancik, Jonathan T Barron, Ravi Ramamoorthi, and Ren Ng. Nerf: Representing scenes as neural radiance fields for view syn- thesis. InEur. Conf. Comput. Vis., 2020. 2

    work page 2020

  28. [28]

    Instant neural graphics primitives with a multires- olution hash encoding.ACM Trans

    Thomas M ¨uller, Alex Evans, Christoph Schied, and Alexan- der Keller. Instant neural graphics primitives with a multires- olution hash encoding.ACM Trans. Graph., 41(4):102:1– 102:15, 2022. 2

    work page 2022

  29. [29]

    Vdb: High-resolution sparse volumes with dynamic topology.ACM Trans

    Ken Museth. Vdb: High-resolution sparse volumes with dynamic topology.ACM Trans. Graph., 32(3):27:1–27:22,

    work page

  30. [30]

    Ohtake, A

    Y . Ohtake, A. Belyaev, and H.P. Seidel. A multi-scale ap- proach to 3d scattered data interpolation with compactly sup- ported basis functions. InShape Modeling Int. (SMI), pages 153–161, 2003. 1

    work page 2003

  31. [31]

    Efficient solver for spacetime control of smoke.ACM Trans

    Zherong Pan and Dinesh Manocha. Efficient solver for spacetime control of smoke.ACM Trans. Graph., 36(5): 162:1–162:13, 2017. 2

    work page 2017

  32. [32]

    Hypernerf: A higher-dimensional representation for topologically varying neural radiance fields,

    Keunhong Park, Utkarsh Sinha, Peter Hedman, Jonathan T Barron, Sofien Bouaziz, Dan B Goldman, Ricardo Martin- Brualla, and Steven M Seitz. Hypernerf: A higher- dimensional representation for topologically varying neural radiance fields.arXiv preprint arXiv:2106.13228, 2021. 2

    work page arXiv 2021

  33. [33]

    D-nerf: Neural radiance fields for dynamic scenes

    Albert Pumarola, Enric Corona, Gerard Pons-Moll, and Francesc Moreno-Noguer. D-nerf: Neural radiance fields for dynamic scenes. InIEEE Conf. Comput. Vis. Pattern Recog.,

    work page

  34. [34]

    A material point method for snow simulation.ACM Trans

    Alexey Stomakhin, Craig Schroeder, Lawrence Chai, Joseph Teran, and Andrew Selle. A material point method for snow simulation.ACM Trans. Graph., 32(4):1–10, 2013. 1, 2

    work page 2013

  35. [35]

    A particle method for history-dependent materials.Comput

    Deborah Sulsky, Zhen Chen, and Howard L Schreyer. A particle method for history-dependent materials.Comput. Methods Appl. Mech. Eng., 118(1-2):179–196, 1994. 1

    work page 1994

  36. [36]

    Keyframe control of smoke simulations.ACM Trans

    Adrien Treuille, Antoine McNamara, Zoran Popovi´c, and Jos Stam. Keyframe control of smoke simulations.ACM Trans. Graph., 22(3):716–723, 2003. 2

    work page 2003

  37. [37]

    Shape transformation us- ing variational implicit functions

    Greg Turk and James F O’brien. Shape transformation us- ing variational implicit functions. InACM SIGGRAPH 2005 Courses, pages 13–es. 2005. 2

    work page 2005

  38. [38]

    Neus: Learning neural implicit surfaces by volume rendering for multi-view reconstruction

    Peng Wang, Lingjie Liu, Yuan Liu, Christian Theobalt, Taku Komura, and Wenping Wang. Neus: Learning neural implicit surfaces by volume rendering for multi-view reconstruction. InAdv. Neural Inform. Process. Syst., 2021. 2

    work page 2021

  39. [39]

    4d gaussian splatting for real-time dynamic scene rendering,

    Guanjun Wu, Taoran Yi, Jiemin Fang, Lingxi Xie, Xiaopeng Zhang, Wei Wei, Wenyu Liu, Qi Tian, and Xinggang Wang. 4d gaussian splatting for real-time dynamic scene rendering,

    work page

  40. [40]

    S- volsdf: Sparse multi-view stereo regularization of neural im- plicit surfaces

    Haoyu Wu, Alexandros Graikos, and Dimitris Samaras. S- volsdf: Sparse multi-view stereo regularization of neural im- plicit surfaces. InInt. Conf. Comput. Vis., pages 3556–3568,

    work page

  41. [41]

    Pbr-nerf: Inverse rendering with physics-based neural fields

    Sean Wu, Shamik Basu, Tim Broedermann, Luc Van Gool, and Christos Sakaridis. Pbr-nerf: Inverse rendering with physics-based neural fields. InIEEE Conf. Comput. Vis. Pat- tern Recog., pages 10974–10984, 2025. 2

    work page 2025

  42. [42]

    Physgaussian: Physics- integrated 3d gaussians for generative dynamics

    Tianyi Xie, Zeshun Zong, Yuxing Qiu, Xuan Li, Yutao Feng, Yin Yang, and Chenfanfu Jiang. Physgaussian: Physics- integrated 3d gaussians for generative dynamics. InIEEE Conf. Comput. Vis. Pattern Recog., pages 4389–4398, 2024. 1, 2, 8

    work page 2024

  43. [43]

    A differentiable material point method framework for shape morphing.IEEE Trans

    Michael Xu, Chang-Yong Song, David Levin, and David Hyde. A differentiable material point method framework for shape morphing.IEEE Trans. Vis. Comput. Graph., 31(10): 9140–9153, 2025. 1, 2, 3, 9

    work page 2025

  44. [44]

    Deformable 3d gaussians for high- fidelity monocular dynamic scene reconstruction, 2023

    Ziyi Yang, Xinyu Gao, Wen Zhou, Shaohui Jiao, Yuqing Zhang, and Xiaogang Jin. Deformable 3d gaussians for high- fidelity monocular dynamic scene reconstruction, 2023. 2

    work page 2023

  45. [45]

    Differentiable surface splatting for point-based geometry processing.ACM Trans

    Wang Yifan, Felice Serena, Shihao Wu, Cengiz ¨Oztireli, and Olga Sorkine-Hornung. Differentiable surface splatting for point-based geometry processing.ACM Trans. Graph., 38 (6):1–14, 2019. 2

    work page 2019

  46. [46]

    Gradient surgery for multi-task learning

    Tianhe Yu, Saurabh Kumar, Abhishek Gupta, Sergey Levine, Karol Hausman, and Chelsea Finn. Gradient surgery for multi-task learning. InAdv. Neural Inform. Process. Syst.,

    work page

  47. [47]

    Progressive point cloud upsam- pling via differentiable rendering.IEEE Trans

    Pingping Zhang, Xu Wang, Lin Ma, Shiqi Wang, Sam Kwong, and Jianmin Jiang. Progressive point cloud upsam- pling via differentiable rendering.IEEE Trans. Circuits Syst. Video Technol., 31(12):4673–4685, 2021. 1

    work page 2021

  48. [48]

    Feng, Changxi Zheng, Noah Snavely, Jiajun Wu, and William T

    Tianyuan Zhang, Hong-Xing Yu, Rundi Wu, Brandon Y . Feng, Changxi Zheng, Noah Snavely, Jiajun Wu, and William T. Freeman. Physdreamer: Physics-based interac- tion with 3d objects via video generation, 2024. 3, 8

    work page 2024

  49. [49]

    Gaussian- painter: Painting point cloud into 3d gaussians with normal guidance

    Jingqiu Zhou, Lue Fan, Xuesong Chen, et al. Gaussian- painter: Painting point cloud into 3d gaussians with normal guidance. InAAAI Conf. Artif. Intell., 2025. 3

    work page 2025

  50. [50]

    Real-time rendering of procedural multiscale materials

    Tobias Zirr and Anton S Kaplanyan. Real-time rendering of procedural multiscale materials. InACM SIGGRAPH Symp. Interact. 3D Graph. Games, pages 139–148, 2016. 1 14

    work page 2016