pith. sign in

arxiv: 2512.09664 · v2 · submitted 2025-12-10 · 💻 cs.DC · cs.CV· cs.LG· eess.IV

SynthPix: A lightspeed PIV image generator

Antonio Terpin , Alan Bonomi , Francesco Banelli , Raffaello D'Andrea This is my paper

Pith reviewed 2026-05-16 23:26 UTC · model grok-4.3

classification 💻 cs.DC cs.CVcs.LGeess.IV
keywords synthetic image generationparticle image velocimetryPIVJAXon-the-fly data streamingflow field imagingmachine learning datasetsimage synthesis
0
0 comments X

The pith

SynthPix generates PIV image pairs from flow fields and streams them live into learning pipelines using JAX.

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

SynthPix is a synthetic image generator for particle image velocimetry that creates image pairs directly from a user-supplied flow field. It incorporates standard experimental controls such as particle seeding density, image size, nonuniform illumination, noise, and blur. Implemented in JAX, the tool runs on accelerators and delivers images in real time without writing them to disk first. This removes the storage barrier that normally limits large-scale training and allows closed-loop procedures where acquisition parameters can be adjusted while the model learns. The authors show the generator works for both laboratory flows and field measurements such as river velocimetry, and that it supports quick sweeps over imaging settings to test robustness.

Core claim

SynthPix produces PIV image pairs from prescribed flow fields while exposing a configuration interface aligned with common PIV imaging and acquisition parameters and streams them on-the-fly into learning pipelines without prohibitive storage costs.

What carries the argument

The JAX-parallel image synthesis pipeline that converts an input velocity field into particle positions and then applies optical and noise models to produce realistic image pairs.

If this is right

  • Machine-learning models for flow estimation can train continuously without first building and storing a fixed dataset.
  • Rapid sweeps over imaging parameters become practical, enabling systematic robustness checks during development.
  • Closed-loop co-design of acquisition settings and analysis algorithms is supported because images can be generated with live feedback.
  • The same generator can serve both controlled laboratory work and field applications such as riverine image velocimetry.

Where Pith is reading between the lines

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

  • The tool could be coupled to reinforcement-learning agents that adjust experimental parameters in real time based on model performance.
  • On-demand generation might reduce the domain gap in transfer learning by allowing fresh synthetic samples that match the exact statistics of a target experiment.
  • Integration with distributed training frameworks could let multiple learners draw from the same live stream without duplication of effort.

Load-bearing premise

The assumption that the modeled imaging effects produce images close enough to real camera captures that algorithms trained on them transfer to physical experiments.

What would settle it

Train a flow estimator exclusively on SynthPix images and measure its accuracy drop when tested on a held-out collection of real PIV recordings taken under matching flow conditions.

Figures

Figures reproduced from arXiv: 2512.09664 by Alan Bonomi, Antonio Terpin, Francesco Banelli, Raffaello D'Andrea.

Figure 1
Figure 1. Figure 1: The SynthPix pipeline (top) follows established PIV image-generation techniques but is [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Using SynthPix to instantiate the image generator, get a new batch of PIV images, the [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Zoomed-in images generated with SynthPix (left) and PIVlab [ [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Flows (left), original (middle) and Synthpix-generated (right) images from two datasets of [ [PITH_FULL_IMAGE:figures/full_fig_p007_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Results of the ablation studies in Section [PITH_FULL_IMAGE:figures/full_fig_p008_5.png] view at source ↗
read the original abstract

We describe SynthPix, a synthetic image generator for Particle Image Velocimetry (PIV) with a focus on performance and parallelism on accelerators, implemented in JAX. SynthPix produces PIV image pairs from prescribed flow fields while exposing a configuration interface aligned with common PIV imaging and acquisition parameters (e.g., seeding density, particle image size, illumination nonuniformity, noise, blur, and timing). In contrast to offline dataset generation workflows, SynthPix is built to stream images on-the-fly directly into learning and benchmarking pipelines, enabling data-hungry methods and closed-loop procedures -- such as adaptive sampling and acquisition/parameter co-design -- without prohibitive storage and input-output costs. We demonstrate that SynthPix is compatible with a broad range of application scenarios, including controlled laboratory experiments and riverine image velocimetry, and supports rapid sweeps over nuisance factors for systematic robustness evaluation. SynthPix is a tool that supports the flow quantification community and in this paper we describe the main ideas behind the software package.

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

1 major / 2 minor

Summary. The manuscript describes SynthPix, a JAX-based synthetic image generator for Particle Image Velocimetry (PIV) that produces image pairs from prescribed flow fields on-the-fly. It exposes a configuration interface matching common PIV parameters including seeding density, particle image size, illumination nonuniformity, noise, blur, and timing, with the goal of streaming data directly into learning and benchmarking pipelines to avoid storage and I/O costs. The work claims compatibility with laboratory experiments and riverine image velocimetry and support for rapid nuisance-factor sweeps.

Significance. If the implementation delivers the claimed performance and image fidelity, SynthPix would provide a practical tool for the flow quantification community, lowering barriers to large-scale synthetic data use in data-hungry machine-learning methods and enabling closed-loop procedures such as adaptive sampling and parameter co-design.

major comments (1)
  1. Abstract: the central claim of 'lightspeed' performance together with broad compatibility and on-the-fly streaming advantages is presented without any quantitative benchmarks, timing results, error metrics, or validation against real PIV data, leaving the performance and utility assertions unsubstantiated.
minor comments (2)
  1. The manuscript would benefit from a concise table listing all exposed configuration parameters with their default values and physical units to improve reproducibility.
  2. Consider including a short pseudocode or API usage example illustrating how a flow field is supplied and an image pair is generated and streamed.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the positive overall assessment and the recommendation for minor revision. The single major comment is addressed point-by-point below; we have revised the manuscript to incorporate the suggested improvements.

read point-by-point responses

  1. Referee: [—] Abstract: the central claim of 'lightspeed' performance together with broad compatibility and on-the-fly streaming advantages is presented without any quantitative benchmarks, timing results, error metrics, or validation against real PIV data, leaving the performance and utility assertions unsubstantiated.

    Authors: We agree that the abstract would be strengthened by explicit quantitative support. The body of the manuscript (Sections 4 and 5) already reports concrete timing benchmarks on accelerator hardware, throughput comparisons with existing generators, RMS error metrics between synthetic and laboratory PIV images, and compatibility demonstrations on both controlled experiments and riverine datasets. In the revised version we have updated the abstract to include the key performance numbers (e.g., sustained generation rates and validation errors) so that the central claims are directly substantiated. revision: yes

Circularity Check

0 steps flagged

No significant circularity

full rationale

The manuscript is a software tool description paper. It documents the implementation of SynthPix, a JAX-based generator that produces PIV image pairs from prescribed flow fields using standard imaging parameters. No mathematical derivations, theorems, fitted parameters, or empirical predictions are advanced that could reduce to self-referential inputs. The value claim rests on the tool's interface and on-the-fly streaming capability, which are externally verifiable through use rather than internal equations or self-citations. No load-bearing steps of any enumerated circularity kind are present.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

The work is a software engineering contribution rather than a scientific derivation. No free parameters are fitted inside the paper; configuration values are user inputs. No new physical axioms or invented entities are introduced.

pith-pipeline@v0.9.0 · 5485 in / 1206 out tokens · 45546 ms · 2026-05-16T23:26:51.062388+00:00 · methodology

discussion (0)

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

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Flow Gym: A framework for the development, benchmarking, training, and deployment of flow-field quantification methods

    physics.flu-dyn 2025-12 accept novelty 7.0

    Flow Gym supplies a JAX-based framework with standardized interfaces, modular components, and utilities to develop, benchmark, train, and deploy flow-field quantification methods such as PIV on both synthetic and expe...

Reference graph

Works this paper leans on

41 extracted references · 41 canonical work pages · cited by 1 Pith paper

  1. [1]

    Banelli, A

    F. Banelli, A. Bonomi, A. Terpin, Particle Image Velocimetry Refinement via Consensus ADMM, Working paper (2025)

    work page 2025

  2. [2]

    Banelli, A

    F. Banelli, A. Terpin, A. Bonomi, R. D’Andrea, Flow Gym, Working paper (2025)

    work page 2025

  3. [3]

    Terpin, R

    A. Terpin, R. D’Andrea, Using reinforcement learning to probe the role of feedback in skill acquisition, arXiv preprint arXiv:2512.08463 (2025)

    work page arXiv 2025

  4. [4]

    Probst, synpivimage: Tool to build synthetic particle image velocimetry (PIV) images (2024)

    M. Probst, synpivimage: Tool to build synthetic particle image velocimetry (PIV) images (2024)

    work page 2024

  5. [5]

    Mendes, A

    L. Mendes, A. Bernardino, R. Ferreira, piv-image-generator: An image generating software package for planar PIV and optical flow benchmarking, SoftwareX (2020)

    work page 2020

  6. [6]

    Stamhuis, W

    E. Stamhuis, W. Thielicke, PIVlab–towards user-friendly, affordable and accurate digital particle image velocimetry in MATLAB, Journal of open research software (2014)

    work page 2014

  7. [7]

    Willert, S

    C. Willert, S. T. Wereley, J. Kompenhans, Particle image velocimetry: a practical guide (2007)

    work page 2007

  8. [8]

    Rayleigh, On the stability, or instability, of certain fluid motions, Proceedings of the London Mathematical Society (1879)

    L. Rayleigh, On the stability, or instability, of certain fluid motions, Proceedings of the London Mathematical Society (1879). 9

    work page

  9. [9]

    T. Von Karman, Über den mechanismus des widerstandes, den ein bewegter körper in einer flüssigkeit erfährt, Nachrichten von der Gesellschaft der Wissenschaften zu Göttingen, Mathematisch- Physikalische Klasse (1911)

    work page 1911

  10. [10]

    C. E. Willert, M. Gharib, Digital particle image velocimetry, Experiments in fluids (1991)

    work page 1991

  11. [11]

    Scarano, Iterative image deformation methods in PIV, Measurement science and technology (2001)

    F. Scarano, Iterative image deformation methods in PIV, Measurement science and technology (2001)

    work page 2001

  12. [12]

    Westerweel, G

    J. Westerweel, G. E. Elsinga, R. J. Adrian, Particle image velocimetry for complex and turbulent flows, Annual Review of Fluid Mechanics (2013)

    work page 2013

  13. [13]

    H. Wang, G. He, S. Wang, Globally optimized cross-correlation for particle image velocimetry, Experiments in Fluids (2020)

    work page 2020

  14. [14]

    Astarita, Analysis of weighting windows for image deformation methods in PIV, Experiments in fluids (2007)

    T. Astarita, Analysis of weighting windows for image deformation methods in PIV, Experiments in fluids (2007)

    work page 2007

  15. [15]

    F. F. J. Schrijer, F. Scarano, Effect of predictor–corrector filtering on the stability and spatial resolution of iterative PIV interrogation, Experiments in Fluids (2008)

    work page 2008

  16. [16]

    Q. Gao, H. Lin, H. Tu, H. Zhu, R. Wei, G. Zhang, X. Shao, A robust single-pixel particle image velocimetry based on fully convolutional networks with cross-correlation embedded, Physics of Fluids (2021)

    work page 2021

  17. [17]

    Y. Lee, F. Gu, Z. Gong, D. Pan, W. Zeng, Surrogate-based cross-correlation for particle image velocimetry, Physics of Fluids (2024)

    work page 2024

  18. [18]

    B. K. Horn, B. G. Schunck, Determining optical flow, Artificial intelligence (1981)

    work page 1981

  19. [19]

    Baker, I

    S. Baker, I. Matthews, Lucas-kanade 20 years on: A unifying framework, International journal of computer vision (2004)

    work page 2004

  20. [20]

    Corpetti, D

    T. Corpetti, D. Heitz, G. Arroyo, E. Mémin, A. Santa-Cruz, Fluid experimental flow estimation based on an optical-flow scheme, Experiments in fluids (2006)

    work page 2006

  21. [21]

    Zhong, H

    Q. Zhong, H. Yang, Z. Yin, An optical flow algorithm based on gradient constancy assumption for PIV image processing, Measurement Science and Technology (2017)

    work page 2017

  22. [22]

    S. Cai, J. Liang, Q. Gao, C. Xu, R. Wei, Particle image velocimetry based on a deep learning motion estimator, IEEE Transactions on Instrumentation and Measurement (2019)

    work page 2019

  23. [23]

    S. Cai, J. Liang, S. Zhou, Q. Gao, C. Xu, R. Wei, S. Wereley, J.-S. Kwon, Deep-PIV: A new framework of PIV using deep learning techniques, in: Proceedings of the 13th International Symposium on Particle Image Velocimetry—ISPIV, 2019

    work page 2019

  24. [24]

    S. Cai, S. Zhou, C. Xu, Q. Gao, Dense motion estimation of particle images via a convolutional neural network, Experiments in Fluids (2019)

    work page 2019

  25. [25]

    Manickathan, C

    L. Manickathan, C. Mucignat, I. Lunati, Kinematic training of convolutional neural networks for particle image velocimetry, Measurement Science and Technology (2022). 10

    work page 2022

  26. [26]

    Lagemann, K

    C. Lagemann, K. Lagemann, S. Mukherjee, W. Schröder, Deep recurrent optical flow learning for particle image velocimetry data, Nature Machine Intelligence (2021)

    work page 2021

  27. [27]

    Rabault, J

    J. Rabault, J. Kolaas, A. Jensen, Performing particle image velocimetry using artificial neural networks: a proof-of-concept, Measurement Science and Technology (2017)

    work page 2017

  28. [28]

    Y. Lee, H. Yang, Z. Yin, PIV-DCNN: cascaded deep convolutional neural networks for particle image velocimetry, Experiments in Fluids (2017)

    work page 2017

  29. [29]

    Q. Zhu, J. Wang, J. Hu, J. Ai, Y. Lee, PIV-FlowDiffuser: Transfer-learning-based denoising diffusion models for PIV, arXiv preprint arXiv:2504.14952 (2025)

    work page arXiv 2025

  30. [30]

    Y. A. Reddy, J. Wahl, M. Sjödahl, Twins-PIVNet: Spatial attention-based deep learning framework for particle image velocimetry using Vision Transformer, Ocean Engineering (2025)

    work page 2025

  31. [31]

    Huang, X

    Z. Huang, X. Shi, C. Zhang, Q. Wang, K. C. Cheung, H. Qin, J. Dai, H. Li, Flowformer: A transformer architecture for optical flow, in: European conference on computer vision, 2022

    work page 2022

  32. [32]

    Lecordier, J

    B. Lecordier, J. Westerweel, The EUROPIV synthetic image generator (SIG), in: Particle Image Velocimetry: Recent Improvements: Proceedings of the EUROPIV 2 Workshop held in Zaragoza, Spain, March 31–April 1, 2003, Springer, 2004, pp. 145–161

    work page 2003

  33. [33]

    Y. Li, E. Perlman, M. Wan, Y. Yang, C. Meneveau, R. Burns, S. Chen, A. S. Szalay, G. L. Eyink, A public turbulence database cluster and applications to study lagrangian evolution of velocity increments in turbulence, Journal of Turbulence (2008)

    work page 2008

  34. [34]

    Perlman, R

    E. Perlman, R. Burns, Y. Li, C. Meneveau, Data exploration of turbulence simulations using a database cluster, in: Proceedings of the 2007 ACM/IEEE Conference on Supercomputing (SC ’07), 2007

    work page 2007

  35. [35]

    Jassal, B

    G. Jassal, B. E. Schmidt, Synthetic particle image datasets for benchmarking PIV processing algorithms (2024)

    work page 2024

  36. [36]

    Smirnov, S

    A. Smirnov, S. Shi, I. Celik, Random flow generation technique for large eddy simulations and particle-dynamics modeling, Journal of Fluids Engineering (2001)

    work page 2001

  37. [37]

    Graham, K

    J. Graham, K. Kanov, X. Yang, M. Lee, N. Malaya, C. Lalescu, R. Burns, G. Eyink, A. Szalay, R. Moser, et al., A web services accessible database of turbulent channel flow and its use for testing a new integral wall model for les, Journal of Turbulence (2016)

    work page 2016

  38. [38]

    Y. Li, E. Perlman, M. Wan, Y. Yang, C. Meneveau, R. Burns, S. Chen, A. Szalay, G. Eyink, A public turbulence database cluster and applications to study lagrangian evolution of velocity increments in turbulence, Journal of Turbulence (2008)

    work page 2008

  39. [39]

    Szeliski, Computer vision: algorithms and applications, Springer Nature, 2022

    R. Szeliski, Computer vision: algorithms and applications, Springer Nature, 2022

    work page 2022

  40. [40]

    C. Yu, X. Bi, Y. Fan, Y. Han, Y. Kuai, Lightpivnet: An effective convolutional neural network for particle image velocimetry, IEEE Transactions on Instrumentation and Measurement (2021)

    work page 2021

  41. [41]

    P. D. Grontas, A. Terpin, E. C. Balta, R. D’Andrea, J. Lygeros, Pinet: Optimizing hard-constrained neural networks with orthogonal projection layers, arXiv preprint arXiv:2508.10480 (2025). 11

    work page arXiv 2025