CoChat
Sign Up
Cited by 372Computer Graphics and Visualization Techniques3D Shape Modeling and AnalysisModel Reduction and Neural Networks

Deep Fluids: A Generative Network for Parameterized Fluid Simulations

Open access
Kim, Byungsoo, Azevedo, Vinicius C., Thuerey, Nils, Kim, Theodore, Gross, Markus, Solenthaler, Barbara·2019·Computer Graphics Forum
DOI 10.1111/cgf.13619
Open this paper in CoChat — free accountOpen PDF

Abstract

This paper presents a novel generative model to synthesize fluid simulations from a set of reduced parameters. A convolutional neural network is trained on a collection of discrete, parameterizable fluid simulation velocity fields. Due to the capability of deep learning architectures to learn representative features of the data, our generative model is able to accurately approximate the training data set, while providing plausible interpolated in‐betweens. The proposed generative model is optimized for fluids by a novel loss function that guarantees divergence‐free velocity fields at all times. In addition, we demonstrate that we can handle complex parameterizations in reduced spaces, and advance simulations in time by integrating in the latent space with a second network. Our method models a wide variety of fluid behaviors, thus enabling applications such as fast construction of simulations, interpolation of fluids with different parameters, time re‐sampling, latent space simulations, and compression of fluid simulation data. Reconstructed velocity fields are generated up to 700× faster than re‐simulating the data with the underlying CPU solver, while achieving compression rates of up to 1300×.

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.1111/cgf.13619

Cites

3 of 3 references verified0 citation errors · 0 not checkable

Statistics

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

  • Existing attempts to integrate physics into video generation either add physics losses directly at the pixel level, which is noisy and computationally heavy, or tightly couple a generative model to a specific simulator, limiting flexibility and domain coverage.

    “Existing attempts either (i) add physics losses directly at the pixel level, which is noisy and computationally heavy [51,53,54,57], or (ii) tightly couple a generative model to a specific simulator, limiting flexibility and domain coverage.”

    From PhyStruct-Video: Structured Physics State Spaces for Physically Consistent Video Generation (2025) · cites this paper as [53]

    Section 3.1 (pp. 3–4) describes a physics-motivated loss for reconstructed fluid velocity fields and notes that minimizing L1 distance alone can produce noise. It does not describe pixel-level physics losses for video generation or say that such losses are computationally heavy. The citation is attached to the pixel-level-loss clause, not the subsequent simulator-coupling clause. (LLM, conf=0.75)

    Experimental — accuracy not yet measured

  • For real videos, PhyStruct-Video relies on approximate geometry, materials and forces inferred from vision models and heuristics, which can introduce discrepancies between true and modeled dynamics.

    “For real videos, we rely on approximate geometry, materials and forces inferred from vision models and heuristics [38,39,40,51,53,54,56].This can introduce discrepancies between the true and modeled dynamics.”

    From PhyStruct-Video: Structured Physics State Spaces for Physically Consistent Video Generation (2025) · cites this paper as [53]

    The Introduction (p. 2) and Conclusion (p. 11) describe generating fluid simulations from reduced parameters. The cited paper does not address the first sentence’s claim that geometry, materials, and forces for real videos are inferred using vision models and heuristics; searches for those terms found no supporting passage. The following sentence about discrepancies is the citing authors’ own conclusion. (LLM, conf=0.74)

    Experimental — accuracy not yet measured

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Cites

Flagged first, then by citations

Research works printed but not linked: 48

  1. [ATW15] ANDO R., THUEREY N., WOJTAN C.: A stream function solver for liquid simulations. ACM Transactions on Graphics 34 , 4 (2015), 53:1-53:9. 9
  2. [BSM17] BERTHELOT D., SCHUMM T., METZ L.: BEGAN: Boundary Equilibrium Generative Adversarial Networks. arXiv:1703.10717 . 4 · arXiv
  3. [CT17b] CHU M., THUEREY N.: Data-driven synthesis of smoke flows with CNN-based feature descriptors. ACMTransactions on Graphics 36 , 4 (2017), 1-14. 2
  4. [DLF12] DE WITT T., LESSIG C., FIUME E.: Fluid simulation using Laplacian eigenfunctions. ACM Trans. Graph. 31 , 1 (2012), 1-11. 2
  5. [FGP17] FARIMANI A. B., GOMES J., PANDE V. S.: Deep Learning the Physics of Transport Phenomena. arXiv:1709.02432 . 2 · arXiv
  6. [GDG ∗ 17] GUERIN E., DIGNE J., GALIN E., PEYTAVIE A., WOLF C., BENES B., MARTINEZ B.: Interactive Example-Based Terrain Authoring with Conditional Generative Adversarial Networks. ACM Transactions on Graphics (proceedings of Siggraph Asia 2017) 36 , 6 (2017). 1
  7. [GLI16] GUO X., LI W., IORIO F.: Convolutional Neural Networks for Steady Flow Approximation. In Proc. of Knowledge Discovery and Data Mining (2016), pp. 481-490. 2
  8. [GN07] GUPTA M., NARASIMHAN S. G.: Legendre fluids: a unified framework for analytic reduced space modeling and rendering of participating media. In Symp. on Computer Anim. (2007), pp. 17-25. 2
  9. [GPAM ∗ 14] GOODFELLOW I., POUGET-ABADIE J., MIRZA M., XU B., WARDE-FARLEY D., OZAIR S., COURVILLE A., BENGIO Y.: Generative Adversarial Nets. In Advances in Neural Information Processing Systems 27 . 2014, pp. 2672-2680. 2
  10. [JAFF16] JOHNSON J., ALAHI A., FEI-FEI L.: Perceptual losses for real-time style transfer and super-resolution. In European Conference on Computer Vision (2016), Springer, pp. 694-711. 3
  11. [JSK16] JONES A. D., SEN P., KIM T.: Compressing Fluid Subspaces. In Symposium on Computer Animation (2016), pp. 77-84. 1, 2, 9
  12. [KA15] KINGMA D., ADAM J. B.: A method for stochastic optimization. In International Conference on Learning Representations (ICLR) (2015), vol. 5. 5
  13. [KALL17] KARRAS T., AILA T., LAINE S., LEHTINEN J.: Progressive Growing of GANs for Improved Quality, Stability, and Variation. arXiv:1710.10196 . 1 · arXiv
  14. [KD13] KIM T., DELANEY J.: Subspace fluid re-simulation. ACMTransactions on Graphics 32 , 4 (2013), 1. 1, 2, 9
  15. [Kim08] KIM T.: Hardware-aware Analysis and Optimization of Stable Fluids. In Interactive 3D Graphics and Games (2008), pp. 99-106. 9
  16. [KMM ∗ 17] KALLWEIT S., MÜLLER T., MCWILLIAMS B., GROSS M., NOVÁK J.: Deep scattering: Rendering atmospheric clouds with radiance-predicting neural networks. ACM Trans Graph 36 , 6 (2017). 1
  17. [KWKT15] KULKARNI T. D., WHITNEY W. F., KOHLI P., TENENBAUM J.: Deep convolutional inverse graphics network. In Advances in neural information processing systems (2015), pp. 2539-2547. 5
  18. [LH16] LOSHCHILOV I., HUTTER F.: SGDR: Stochastic Gradient Descent with Warm Restarts. arXiv:1608.03983 . 6 · arXiv
  19. [LI06] LINDSTROM P., ISENBURG M.: Fast and efficient compression of floating-point data. IEEE Transactions on Visualization and Computer Graphics 12 , 5 (2006), 1245-1250. 9
  20. [LJF16] LU W., JIN N., FEDKIW R.: Two-way Coupling of Fluids to Reduced Deformable Bodies. In SCA (2016), pp. 67-76. 2
  21. [LJS ∗ 15] LADICKÝ L., JEONG S., SOLENTHALER B., POLLEFEYS M., GROSS M.: Data-driven fluid simulations using regression forests. ACM Transactions on Graphics 34 , 6 (2015), 1-9. 1, 2
  22. [LKT16] LING J., KURZAWSKI A., TEMPLETON J.: Reynolds averaged turbulence modelling using deep neural networks with embedded invariance. Journal of Fluid Mechanics 807 (2016), 155-166. 3
  23. [LMH ∗ 15] LIU B., MASON G., HODGSON J., TONG Y., DESBRUN M.: Model-reduced variational fluid simulation. ACM Transactions on Graphics 34 , 6 (2015), 244. 2
  24. [LRS ∗ 18] LIU G., REDA F. A., SHIH K. J., WANG T., TAO A., CATANZARO B.: Image inpainting for irregular holes using partial convolutions. arXiv:1804.07723 . 11 · arXiv
  25. [LTH ∗ 17] LEDIG C., THEIS L., HUSZÁR F., CABALLERO J., CUNNINGHAM A., ACOSTA A., AITKEN A. P., TEJANI A., TOTZ J., WANG Z., ET AL.: Photo-realistic single image super-resolution using a generative adversarial network. In CVPR (2017), vol. 2, p. 4. 3
  26. [Lum67] LUMLEY J. L.: The Structure of Inhomogeneous Turbulent Flows. In Atmospheric turbulence and radio propagation . Nauka, 1967, pp. 166-178. 2
  27. [MCL15] MATHIEU M., COUPRIE C., LECUN Y.: Deep multi-scale video prediction beyond mean square error. arXiv:1511.05440 . 4 · arXiv
  28. [MCP ∗ 09] MULLEN P., CRANE K., PAVLOV D., TONG Y., DESBRUN M.: Energy-preserving integrators for fluid animation. ACM Transactions on Graphics 28 , 3 (2009), 1. 9
  29. [MDB17] MOSSER L., DUBRULE O., BLUNT M. J.: Reconstruction of three-dimensional porous media using generative adversarial neural networks. Physical Review E 96 , 4 (2017), 043309. 3
  30. [MHN13] MAAS A., HANNUN A., NG A.: Rectifier nonlinearities improve neural network acoustic models. In ICML (2013), vol. 30. 4
  31. [MTP ∗ 18] MA P., TIAN Y., PAN Z., REN B., MANOCHA D.: Fluid directed rigid body control using deep reinforcement learning. ACM Transactions on Graphics (SIGGRAPH Proceedings) (2018). 2
  32. [PBT17] PRANTL L., BONEV B., THUEREY N.: Pre-computed Liquid Spaces with Generative Neural Networks and Optical Flow. arXiv: 1704.07854 . 2, 11 · arXiv
  33. [PdON18] PAGANINI M., DE OLIVEIRA L., NACHMAN B.: Accelerating science with generative adversarial networks: an application to 3d particle showers in multilayer calorimeters. Physical review letters 120 , 4 (2018), 042003. 3
  34. [RLM ∗ 17] RAVANBAKHSH S., LANUSSE F., MANDELBAUM R., SCHNEIDER J. G., POCZOS B.: Enabling dark energy science with deep generative models of galaxy images. In AAAI (2017), pp. 1488-1494. 3
  35. [Smi17] SMITH L. N.: Cyclical learning rates for training neural networks. In Applications of Computer Vision (2017), pp. 464-472. 6
  36. [SSW ∗ 13] STANTON M., SHENG Y., WICKE M., PERAZZI F., YUEN A., NARASIMHAN S., TREUILLE A.: Non-polynomial galerkin projection on deforming meshes. ACM Trans Graph 32 , 4 (2013), 86:1. 9
  37. [Sta14] STANTON M.: Data-Driven Methods for Interactive Simulation of Complex Phenomena . PhD thesis, Carnegie Mellon Univ., 2014. 2
  38. [Thu17] THUEREY N.: Interpolations of smoke and liquid simulations. ACM Transactions on Graphics (TOG) 36 , 1 (2017), 3. 10
  39. [TLP06] TREUILLE A., LEWIS A., POPOVI ´ C Z.: Model reduction for real-time fluids. ACM Transactions on Graphics 25 , 3 (2006), 826. 1, 2
  40. [TSSP17] TOMPSON J., SCHLACHTER K., SPRECHMANN P., PERLIN K.: Accelerating Eulerian fluid simulation with convolutional networks. In Proceedings of Machine Learning Research (Aug 2017), vol. 70, pp. 3424-3433. 1, 2
  41. [UB18] UMETANI N., BICKEL B.: Learning three-dimensional flow for interactive aerodynamic design. ACM Transactions on Graphics (SIGGRAPH Proceedings) (2018). 2
  42. [UHT18] UM K., HU X., THUEREY N.: Liquid splash modeling with neural networks. Computer Graphics Forum 37 , 8 (2018), 171-182. 2
  43. [VLL ∗ 10] VINCENT P., LAROCHELLE H., LAJOIE I., BENGIO Y., MANZAGOL P.-A.: Stacked denoising autoencoders: Learning useful representations in a deep network with a local denoising criterion. J. Mach. Learn. Res. 11 (2010), 3371-3408. 5
  44. [WBT18] WIEWEL S., BECHER M., THUEREY N.: Latent-space Physics: Towards Learning the Temporal Evolution of Fluid Flow. arXiv:1802.10123 . 2 · arXiv
  45. [WST09] WICKE M., STANTON M., TREUILLE A.: Modular bases for fluid dynamics. In ACM SIGGRAPH (2009). 2, 9
  46. [XFCT18] XIE Y., FRANZ E., CHU M., THUEREY N.: tempogan: A temporally coherent, volumetric gan for super-resolution fluid flow. ACM Transactions on Graphics (TOG) 37 , 4 (2018), 95. 1, 2
  47. [YYX16] YANG C., YANG X., XIAO X.: Data-driven projection method in fluid simulation. CAVW 27 , 3-4 (2016), 415-424. 2
  48. [ZGFK17] ZHAO H., GALLO O., FROSIO I., KAUTZ J.: Loss functions for image restoration with neural networks. IEEE Transactions on Computational Imaging 3 , 1 (2017), 47-57. 3

Cited by

Sorted by citations

Authors
Kim, B., Azevedo, V. C., Thuerey, N., Kim, T., Gross, M., & Solenthaler, B.
Affiliations
  • ETH Zurich (CH)
  • ETH Zürich
  • Technical University of Munich (DE)
  • Pixar Animation Studios
ORCID
Venue
Computer Graphics Forum
Publisher
Wiley
Published
2019-05
Volume
38
Issue
2
Pages
59-70
ISSN
  • 0167-7055 (ISSN-L)
  • 1467-8659
DOI
10.1111/cgf.13619
OpenAlex
W2805869290
Cited by
372
References
3
Citations by year
2026: 16 · 2025: 34 · 2024: 47 · 2023: 57 · 2022: 59 · 2021: 78 · 2020: 52 · 2019: 28
FWCI
117.97
Type
journal article
Language
en
Open access
Licence
onlinelibrary.wiley.com/termsAndConditions#vor (vor, 37 d)
Locations
  • Computer Graphics Forum (publishedVersion)
  • arXiv (Cornell University) (submittedVersion)
  • arXiv (Cornell University)
Topics
  • Physical Sciences › Computer Science › Computer Graphics and Computer-Aided Design › Computer Graphics and Visualization Techniques
  • Physical Sciences › Engineering › Computational Mechanics › 3D Shape Modeling and Analysis
  • Physical Sciences › Physics and Astronomy › Statistical and Nonlinear Physics › Model Reduction and Neural Networks
Keywords
fluid simulation, generative models, parameterized simulation, divergence-free velocity, loss function design, latent space interpolation, convolutional neural networks

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review