Pith. sign in

REVIEW 3 major objections 5 minor 1 cited by

PIV-FlowDiffuser:Transfer-learning-based denoising diffusion models for PIV

T0 review · 3 major / 5 minor · reviewed 2026-08-16 · deepseek-v4-flash

Pith's one-line read A diffusion model fine-tuned on synthetic PIV data cuts average velocity-estimation error by 59.4 percent over the leading recurrent baseline.

desk verdict Plausible first diffusion model for PIV, but the headline 59.4% reduction rests on an incompletely documented comparison and one secondary claim contradicts the paper's own tables. read the letter →

arxiv 2504.14952 v1 pith:4GFR7FLM submitted 2025-04-21 cs.CV eess.IV

classification cs.CVeess.IV
keywords particleimagevelocimetrydenoisingdiffusionmodeltransferlearningopticalflowestimationend-pointerrordomaingeneralizationturbulentDiffuser
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper sets out to show that a denoising diffusion model, adapted from optical-flow estimation and fine-tuned on synthetic particle image velocimetry (PIV) data, can estimate velocity fields more accurately and more robustly than the leading recurrent deep-learning baseline. The central reported result is a 59.4 percent reduction in average endpoint error over RAFT256-PIV on the classic synthetic PIV dataset, with the diffusion model reaching an AEE of 0.0352 versus 0.0866; it also degrades far less when tested on out-of-domain particle images (0.5537 versus 4.7564). The claim matters because PIV is a standard experimental tool in fluid mechanics, and deep-learning estimators often carry systematic residual noise patterns that the diffusion process is designed to remove step by step. A secondary practical point is that transfer learning makes the data-hungry diffusion model trainable in about two hours rather than many hours from scratch.

What carries the argument

The load-bearing component is the conditional recurrent denoising decoder (Conditional-RDD), a Conv-GRU-based network that takes the current noisy flow field, a diffusion timestep embedding, and image-conditioning features, and outputs a less noisy estimate $p_\theta(\mathbf{v}_{t-1} \mid \mathbf{v}_t, \mathbf{x}_c, \mathbf{x}_{cv}, \mathbf{x}_h)$. Iterating this step over the diffusion chain converts the estimator's systematic residual patterns into removable noise. Transfer learning is the supporting mechanism: the model starts from FlowDiffuser weights pre-trained on optical flow data and is fine-tuned on synthetic PIV data; a 2x bilinear upsampling of the inputs, with a corresponding downsampling of the outputs, lets the model resolve small-scale turbulent structures.

What would settle it

Re-running RAFT256-PIV and PIV-FlowDiffuser on a shared Problem Class 1 test split with identical preprocessing and computing AEE per test case would settle the claim; a reduction much smaller than 59.4 percent would indicate the comparison is not controlled. The discrepancy between the reported 0.0352 average and the 0.0292 mean of the five per-category values in Table 1 is a concrete place to look.

Watch

Extended reading notes

Core claim

The paper claims that conditional denoising diffusion models provide an explicit correction mechanism for PIV velocity estimation. Instead of predicting the flow in one pass or a few deterministic updates, PIV-FlowDiffuser starts from a noisy flow field and repeatedly applies a conditional recurrent denoising decoder that is conditioned on image features (a context feature and a 4D correlation volume) and on the hidden state of a Conv-GRU; each step removes part of the residual error. Pre-trained on natural optical flow datasets and fine-tuned on synthetic PIV data, the model reaches an average endpoint error of 0.0352 on Problem Class 1, a 59.4 percent reduction from the RAFT256-PIV baseline's 0.0866, and on out-of-domain Problem Class 2 images an AEE of 0.5537 versus 4.7564.

Load-bearing premise

The headline 59.4 percent reduction assumes the published baseline errors were measured on the same test split, with the same preprocessing and the same endpoint-error definition as the PIV-FlowDiffuser runs, and that the 0.0352 class-1 average is computed over the same test subset as the baseline average.

Editorial extensions

If this is right

  • PIV-FlowDiffuser-class1 reaches an AEE of 0.0352 on Problem Class 1, a 59.4 percent drop from RAFT256-PIV-class1's 0.0866, and it beats all baselines on four of the five test subsets.
  • On Problem Class 2, out-of-domain AEE is 0.5537 for PIV-FlowDiffuser-class1 versus 4.7564 for RAFT256-PIV-class1, meaning transfer-learned diffusion models degrade less under domain shift.
  • The 2x upsampling adaptation matters: removing it (PIV-FlowDiffuser-class1(*)) raises errors on Problem Class 1 well above the full model, showing the resolution trick contributes to the gain.
  • Fine-tuning is substantially cheaper than training from scratch: about 2 hours for class 1 and 5 hours for class 2 on a single GPU, versus 18 to 32 hours for RAFT256-PIV and Twins-PIV; inference time is about 0.27 seconds, roughly three times RAFT's 0.08 seconds.
  • On experimental turbulent wavy channel flow images, PIV-FlowDiffuser-class1 produces velocity components comparable to the PascalPIV reference, while RAFT32-PIV-class1 fails outright, suggesting the method handles unseen real recordings.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • If the 59.4 percent figure survives a controlled same-split comparison, it implies that a large share of the residual error in deterministic deep PIV estimators is correctable structure rather than irreducible noise, and that an explicit denoising prior can act as a general error-correction layer for flow estimators.
  • A direct ablation the paper does not fully run is fine-tuning the same FlowDiffuser architecture on PIV data without the optical-flow pre-training; such an experiment would separate the contribution of transfer learning from the contribution of the diffusion architecture itself.
  • The upsampling lever suggests a testable scaling prediction: feeding higher-resolution input pairs should continue to improve accuracy on small-scale turbulence up to the limits of the training distribution, which could be checked on the turbulent wavy channel data or on higher-resolution synthetic PIV.
  • The unreported subset weighting behind the 0.0352 average is worth pinning down; until the average is reproducible, the 59.4 percent headline should be read as an estimate of the method's gain rather than a precisely controlled benchmark number.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. The paper proposes PIV-FlowDiffuser, a denoising-diffusion model for particle image velocimetry that is initialized from a FlowDiffuser model pre-trained on natural optical flow datasets (Sintel, KITTI-2015, HD1K) and then fine-tuned on synthetic PIV data, with a two-fold input upsampling adaptation. The central claims are that the method reduces average end-point error (AEE) by 59.4% over the RAFT256-PIV baseline on Cai's Problem Class 1 dataset, that it generalizes better to out-of-domain particle images, and that transfer learning reduces training time relative to training from scratch. Evaluations are presented on Problem Classes 1 and 2 and on the experimental TWCF case, together with runtime comparisons. The paper also states that a code repository is available.

Significance. If the headline results are correct, the paper would demonstrate a substantial accuracy improvement over a leading recurrent baseline in synthetic PIV estimation, as well as improved cross-domain generalization, at the cost of roughly 3.4× the inference time and a fraction of the training time. The transfer-learning approach for diffusion-based PIV is a sensible direction, and the inclusion of in-domain and out-of-domain benchmarks, the TWCF practical case, and a released repository are strengths. However, the central quantitative claim is not currently reproducible from the paper's own reported numbers, and the baseline comparison is not demonstrably controlled. These issues must be resolved before the significance of the claimed improvement can be assessed.

major comments (3)
  1. [Section 4.1, Tables 1 and 3] The headline 59.4% AEE reduction is not reproducible from the paper's own per-class data. The five per-class AEE values for PIV-FlowDiffuser-class1 in Table 1 (0.007, 0.029, 0.039, 0.019, 0.052) have an unweighted mean of 0.0292, not the 0.0352 reported in Table 3; the corresponding RAFT256-PIV-class1 values (0.016, 0.137, 0.093, 0.014, 0.117) have an unweighted mean of 0.0754, not 0.0866. Unless the Table 3 averages use a different weighting or a different test subset, the abstract's central claim cannot be verified from the reported data. Please report per-class test sample sizes and the exact averaging formula used, or present the per-class results for the actual compared test set.
  2. [Section 2.3 and Table 1 caption] The baseline comparison is not controlled. Table 1 states that the WIDIM, PIV-DCNN, PIV-LiteFlowNet-en, RAFT256-PIV, and Twins-PIV values are 'sourced from [24, 29]', while the present work defines an 8:1:1 split of Problem Class 1 in Section 2.3. If the cited baselines were evaluated on a different train/test split or with different preprocessing, then the 0.0866 value in Table 3 is not necessarily the AEE of RAFT256-PIV on the same test set used for PIV-FlowDiffuser-class1, and the 59.4% reduction as well as the out-of-domain comparison (0.5537 vs 4.7564 in Table 3) would not be controlled. Please re-run the baseline methods on the exact test splits used for PIV-FlowDiffuser, or explicitly document any shared evaluation protocol.
  3. [Section 4.1, text after Table 2] The claim that 'PIV-FlowDiffuser-class2 reduces the RAFT error by half' is contradicted by the paper's own numbers. In Table 3, the in-domain AEE for PIV-FlowDiffuser-class2 is 0.3124 versus 0.3540 for RAFT256-PIV-class2, a reduction of about 11.7%, not half. In Table 2, PIV-FlowDiffuser-class2 is worse than RAFT256-PIV-class2 on Backstep (0.155 vs 0.131), Cylinder (0.138 vs 0.124), Uniform (0.328 vs 0.174), and Other (0.587 vs 0.380). Please clarify which comparison supports the 'by half' statement, or correct it.
minor comments (5)
  1. [Abstract and Section 1] The phrase 'special noise' and 'special residual patterns' appears several times; the intended term is likely 'spatial noise' or 'specific residual patterns'. Please revise for clarity.
  2. [Equation 2] Equation (2) writes the loss as an expectation over v0 ∼ q(v0|c) and t ∼ [1,T], but in standard diffusion training the noisy variable vt is drawn from q(vt|v0). Please clarify the exact training objective, including how v0 is obtained from the denoising process.
  3. [Section 2.3] The text says Problem Class 2 consists of 19,000 training image pairs and 1,000 validation pairs, and then lists the class composition of 'these 1,000 cases'. It is unclear whether these 1,000 cases are the validation set or the test set used in Table 2. Please disambiguate the terms 'validation' and 'test'.
  4. [Table 3 caption] The caption says the table reports results 'evaluated on out-of-domain dataset', but the table also contains in-domain results. Please revise the caption to state that both in-domain and out-of-domain results are shown.
  5. [Section 2.1] The number of denoising steps used at inference is not stated in the text. Since the abstract emphasizes the denoising process, please specify the inference step budget and the number of fine-tuning epochs/learning-rate schedule hyperparameters used.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the AEE reductions are empirical held-out comparisons; the unverified baseline split is a benchmark-protocol concern, not an equation-level circularity.

full rationale

The paper's derivation chain is: take a pretrained FlowDiffuser model trained on external optical flow datasets (Sintel, KITTI-2015, HD1K), fine-tune all parameters on synthetic PIV data using the L1 loss in Eq. (2), and evaluate on held-out Problem Class 1, Problem Class 2, and TWCF test cases. The reported AEE, RMSE, and AAE values are empirical metrics computed from model outputs on those test sets; no equation defines these metrics in terms of a fitted constant or the training objective, and no parameter is fitted to the test metrics. Self-citations (refs. 7, 15, 21) appear in background statements about iterative correction and cross-correlation methods and are not load-bearing for the central accuracy or generalization claims. The only substantive concern is that Table 1's caption says the WIDIM, PIV-DCNN, PIV-LiteFlowNet-en, RAFT256-PIV, and Twins-PIV numbers are 'sourced from [24, 29]', while the paper's own runs use an 8:1:1 split described in Section 2.3, and the Table 3 averages are not reproducible from the Table 1 per-class means without undocumented weighting or subsetting. That is a benchmark-protocol and reproducibility risk, not a circularity, because the baseline numbers are external empirical results rather than quantities derived from this paper's own model or equations. No self-definitional step, fitted-input-called-prediction, author-imported uniqueness claim, or renaming of a known result was found.

Assumptions & free parameters 5 free parameters · 6 assumptions · 0 invented entities

The paper adds no invented physical entities; its output is a standard dense velocity field, so the invented-entities list is empty. The ledger captures the hand-chosen adaptations (upsampling factor, denoising steps, fine-tuning budget, one-cycle schedule, split ratio) and the domain assumptions that the fine-tuned model's value rests on. The key assumptions are that pretrained natural-image optical-flow features transfer to particle images, that pseudo-color replication preserves the relevant signal, that 2x bilinear upsampling exposes sub-pixel structure, and that synthetic PIV distributions stand in for real measurements.

free parameters (5)
  • Input upsampling factor = 2 (bilinear)
    Chosen by hand in Sec. 2.1 and Fig. 2 to 'resolve small-scale turbulent flow structures'; the Table 1 ablation (class1 vs class1(*)) shows it is the dominant contributor to the reported accuracy.
  • Number of denoising steps at inference = 4 (conditional-RDD steps, Fig. 2)
    Number of iterative denoising updates used at inference; not justified by a convergence study.
  • Fine-tuning budget = about 2 h (class 1) and 5 h (class 2) on one RTX 3090
    Proxy for epoch count; the paper reports time instead of epochs, batch size, or exact schedule.
  • One-cycle learning-rate schedule hyperparameters = not stated
    Ref. 43 is invoked in Sec. 2.2 without giving peak learning rate, warm-up length, or decay shape; these affect the fine-tuned weights.
  • Train/validation/test split ratio = 8:1:1 for Problem Class 1
    Chosen in Sec. 2.3; the undocumented handling of the Uniform subset in Table 1 depends on this split.
assumptions (6)
  • standard math Conditional denoising formulation p(v_{t-1}|v_t, x_c, x_cv, x_h) from FlowDiffuser estimates flow as a generative process
    The entire method inherits the DDPM framework (Eq. 1) and its pretrained weights; the paper does not re-derive it (Sec. 2.1).
  • domain assumption Particle image intensity is transported by the flow (brightness constancy)
    Standard PIV/optical-flow premise underlying the RAFT correlation volumes used as conditioning features (Sec. 2.1); violated by noise, which the paper acknowledges as motivation but does not model.
  • domain assumption Features learned from natural video optical flow (Sintel, KITTI, HD1K) transfer to particle images via fine-tuning
    The entire transfer-learning strategy (Sec. 2.2) rests on this; if the gap is architectural rather than statistical, fine-tuning would not close it.
  • domain assumption Replicating the grayscale PIV channel three times preserves the information needed for pre-trained color features
    Pseudo-color conversion (Sec. 2.2) is asserted to make transfer feasible without a study of its effect.
  • domain assumption 2x bilinear upsampling resolves small-scale turbulence rather than inserting interpolation artifacts
    Sec. 2.1 claims upsampling lets the network 'better capture fine details'; the claim is empirical via ablation but the mechanism is asserted.
  • domain assumption Synthetic Problem Class 1/2 distributions are representative of real PIV measurements
    Generalization conclusions (Secs. 4.2 and 4.3) extend from synthetic training data to the TWCF experiment; validation there uses a cross-correlation proxy as gold standard.

how reviews work

0 comments
Cite this review

Pith. "Pith review of PIV-FlowDiffuser:Transfer-learning-based denoising diffusion models for PIV." pith.science (2026). https://pith.science/paper/4GFR7FLM

@misc{pith2026250414952,
  author       = {Pith},
  title        = {Pith review of: PIV-FlowDiffuser:Transfer-learning-based denoising diffusion models for PIV},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/4GFR7FLM}},
  note         = {Machine review of arXiv:2504.14952}
}
abstract

Deep learning algorithms have significantly reduced the computational time and improved the spatial resolution of particle image velocimetry~(PIV). However, the models trained on synthetic datasets might have a degraded performance on practical particle images due to domain gaps. As a result, special residual patterns are often observed for the vector fields of deep learning-based estimators. To reduce the special noise step-by-step, we employ a denoising diffusion model~(FlowDiffuser) for PIV analysis. And the data-hungry iterative denoising diffusion model is trained via a transfer learning strategy, resulting in our PIV-FlowDiffuser method. Specifically, (1) pre-training a FlowDiffuser model with multiple optical flow datasets of the computer vision community, such as Sintel, KITTI, etc; (2) fine-tuning the pre-trained model on synthetic PIV datasets. Note that the PIV images are upsampled by a factor of two to resolve the small-scale turbulent flow structures. The visualized results indicate that our PIV-FlowDiffuser effectively suppresses the noise patterns. Therefore, the denoising diffusion model reduces the average end-point error~($AEE$) by 59.4% over RAFT256-PIV baseline on the classic Cai's dataset. Besides, PIV-FlowDiffuser exhibits enhanced generalization performance on unseen particle images due to transfer learning. Overall, this study highlights the transfer-learning-based denoising diffusion models for PIV. And a detailed implementation is recommended for interested readers in the repository https://github.com/Zhu-Qianyu/PIV-FlowDiffuser.

Figures

Figures reproduced from arXiv: 2504.14952 by the authors.

Figure 1
Figure 1. Results of the RAFT256-PIV method [24] on two test cases. The left gives the vector fields, while the right part presents corresponding error maps. Note that some special error patterns are observed in the residuals, which could be further reduced with noise removal. Despite the advancements of neural networks, their per￾formance in practical measurements is not always satisfac￾tory, as illustrated in [PITH_FULL_IM… view at source ↗
Figure 2
Figure 2. (a) The FlowDiffuser [36] includes two encoders (basic encoder and context encoder) and a series of conditional recurrent denoising decoder (RDD). (b) The training method for PIV-FlowDiffuser. The initial weights are from the pre-trained model, and a simple adaptation module (scale up & scale down) is adopted to better predict small-scale turbulence. The entire model was subsequently fine-tuned using a PIV-specified… view at source ↗
Figure 3
Figure 3. Velocity fields and corresponding absolute residuals of Problem Class 1 computed by different methods. Two cases (left: JHTDB, right: SQG) are considered. The color backgrounds denote the corresponding velocity/residual magnitude. Best viewed in color. (unit: pixels per frame) Zhu et al.: Preprint submitted to Elsevier Page 6 of 12 [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Velocity fields and corresponding absolute residuals computed by different methods. Two cases (left: a JHTDB from Problem Class 1, right: a JHTDB from Problem Class 2) are considered. The color backgrounds denote the corresponding velocity/residual magnitude. Best view…
Figure 5
Figure 5. Figure 5: Two cases of TWCF data are visualized with separate velocity components. The color backgrounds denote the corresponding component value. Best viewed in color. (unit: pixels per frame) Zhu et al.: Preprint submitted to Elsevier Page 9 of 12 [PITH_FULL_IMAGE:figures/ful…

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

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

  1. Particle Image Velocimetry Refinement via Consensus ADMM for Active Fluid Control

    physics.flu-dyn 2025-12 conditional novelty 6.0 of 10

    Running several PIV estimators in parallel and reconciling them with consensus ADMM consistently lowers endpoint error versus the best single estimator.

Reference graph

Works this paper leans on

51 extracted references · 49 canonical work pages · cited by 1 Pith paper

  1. [1]

    R. J. Adrian, Scattering particle characteristics and their effect on pulsed laser measurements of fluid flow: speckle velocimetry vs particle image velocimetry, Applied optics 23 (1984) 1690–1691

  2. [2]

    Raffel, C

    M. Raffel, C. E. Willert, F. Scarano, C. J. Kähler, S. T. Were- ley, J. Kompenhans, Particle image velocimetry: a practical guide, springer, 2018

  3. [3]

    R.Panciroli,A.Shams,M.Porfiri, Experimentsonthewaterentryof curved wedges: high speed imaging and particle image velocimetry, Ocean Engineering 94 (2015) 213–222

  4. [4]

    Bardera, M

    R. Bardera, M. Barcala-Montejano, A. Rodríguez-Sevillano, M. León-Calero, Wind flow investigation over an aircraft carrier deck by piv, Ocean Engineering 178 (2019) 476–483

  5. [5]

    Capone, F

    A. Capone, F. Di Felice, F. A. Pereira, On the flow field induced by two counter-rotating propellers at varying load conditions, Ocean Engineering 221 (2021) 108322

  6. [6]

    Y. Lee, S. Zhang, M. Li, X. He, Blind inverse gamma correction with maximized differential entropy, Signal Processing 193 (2022) 108427

  7. [7]

    Y. Lee, F. Gu, Z. Gong, D. Pan, W. Zeng, Surrogate-based cross- correlation for particle image velocimetry, Physics of Fluids 36 (2024)

  8. [8]

    C. J. Kähler, T. Astarita, P. P. Vlachos, J. Sakakibara, R. Hain, S.Discetti,R.LaFoy,C.Cierpka,Mainresultsofthe4thinternational piv challenge, Experiments in Fluids 57 (2016) 1–71

Show all 51 references
  1. [9]

    Sciacchitano, Uncertainty quantification in particle image ve- locimetry, Measurement Science and Technology 30 (2019) 092001

    A. Sciacchitano, Uncertainty quantification in particle image ve- locimetry, Measurement Science and Technology 30 (2019) 092001

  2. [10]

    Westerweel, Fundamentals of digital particle image velocimetry, Measurement science and technology 8 (1997) 1379

    J. Westerweel, Fundamentals of digital particle image velocimetry, Measurement science and technology 8 (1997) 1379

  3. [11]

    Corpetti, D

    T. Corpetti, D. Heitz, G. Arroyo, E. Mémin, A. Santa-Cruz, Fluid experimental flow estimation based on an optical-flow scheme, Ex- periments in fluids 40 (2006) 80–97

  4. [12]

    C. Pan, D. Xue, Y. Xu, J. Wang, R. Wei, Evaluating the accuracy performance of lucas-kanade algorithm in the circumstance of piv application, Science China Physics, Mechanics & Astronomy 58 (2015) 1–16

  5. [13]

    Zhong, H

    Q. Zhong, H. Yang, Z. Yin, An optical flow algorithm based on gra- dient constancy assumption for piv image processing, Measurement Science and Technology 28 (2017) 055208

  6. [14]

    H. Wang, G. He, S. Wang, Globally optimized cross-correlation for particle image velocimetry, Experiments in Fluids 61 (2020) 1–17

  7. [15]

    Y. Lee, S. Mei, Diffeomorphic particle image velocimetry, IEEE Transactions on Instrumentation and Measurement 71 (2021) 1–10

  8. [16]

    J. Ai, Z. Chen, J. Li, Y. Lee, Rethinking asymmetric image deforma- tion with post-correction for particle image velocimetry, Physics of Fluids 37 (2025)

  9. [17]

    S. T. Wereley, C. D. Meinhart, Second-order accurate particle image velocimetry, Experiments in fluids 31 (2001) 258–268

  10. [18]

    F.Scarano, Iterativeimagedeformationmethodsinpiv, Measurement science and technology 13 (2001) R1

  11. [19]

    Scharnowski, C

    S. Scharnowski, C. J. Kähler, Particle image velocimetry-classical operating rules from today’s perspective, Optics and Lasers in Engineering 135 (2020) 106185

  12. [20]

    T.Liu,A.Merat,M.Makhmalbaf,C.Fajardo,P.Merati, Comparison between optical flow and cross-correlation methods for extraction of velocityfieldsfromparticleimages, ExperimentsinFluids56(2015) 1–23

  13. [21]

    Y. Lee, H. Yang, Z. Yin, Piv-dcnn: cascaded deep convolutional neuralnetworksforparticleimagevelocimetry,ExperimentsinFluids 58 (2017) 171

  14. [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 69 (2019) 3538–3554

  15. [23]

    Zhang, M

    M. Zhang, M. D. Piggott, Unsupervised learning of particle image velocimetry, in: High Performance Computing: ISC High Perfor- mance 2020 International Workshops, Frankfurt, Germany, June 21– 25, 2020, Revised Selected Papers 35, Springer, pp. 102–115

  16. [24]

    Lagemann, K

    C. Lagemann, K. Lagemann, S. Mukherjee, W. Schröder, Deep recurrent optical flow learning for particle image velocimetry data, Nature Machine Intelligence 3 (2021) 641–651

  17. [25]

    C.Yu,H.Luo,X.Bi,Y.Fan,M.He,Aneffectiveconvolutionalneural network for liquid phase extraction in two-phase flow piv experiment of an object entering water, Ocean Engineering 237 (2021) 109502

  18. [26]

    H. Wang, Y. Liu, S. Wang, Dense velocity reconstruction from parti- cle image velocimetry/particle tracking velocimetry using a physics- informed neural network, Physics of fluids 34 (2022)

  19. [27]

    S.Cai,C.Gray,G.E.Karniadakis, Physics-informedneuralnetworks enhanced particle tracking velocimetry: An example for turbulent jet flow, IEEE Transactions on Instrumentation and Measurement (2024)

  20. [28]

    C.Yu,Y.Chang,X.Liang,C.Liang,Z.Xie,Deeplearningforparticle image velocimetry with attentional transformer and cross-correlation embedded, Ocean Engineering 292 (2024) 116522

  21. [29]

    Y. A. Reddy, J. Wahl, M. Sjödahl, Twins-pivnet: Spatial attention- based deep learning framework for particle image velocimetry using vision transformer, Ocean Engineering 318 (2025) 120205

  22. [30]

    S. Cai, S. Zhou, C. Xu, Q. Gao, Dense motion estimation of particle imagesviaaconvolutionalneuralnetwork, ExperimentsinFluids60 (2019) 1–16

  23. [31]

    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 70 (2021) 1–15

  24. [32]

    W.Zhang,X.Nie,X.Dong,Z.Sun, Pyramidaldeep-learningnetwork for dense velocity field reconstruction in particle image velocimetry, Experiments in Fluids 64 (2023) 12

  25. [33]

    Tobin, R

    J. Tobin, R. Fong, A. Ray, J. Schneider, W. Zaremba, P. Abbeel, Domain randomization for transferring deep neural networks from simulation to the real world, in: 2017 IEEE/RSJ international con- ference on intelligent robots and systems (IROS), IEEE, pp. 23–30

  26. [34]

    Dhariwal, A

    P. Dhariwal, A. Nichol, Diffusion models beat gans on image synthesis, Advances in neural information processing systems 34 (2021) 8780–8794

  27. [35]

    J. Ho, A. Jain, P. Abbeel, Denoising diffusion probabilistic models, Advances in neural information processing systems 33 (2020) 6840– 6851

  28. [36]

    19167–19176

    A.Luo,X.Li,F.Yang,J.Liu,H.Fan,S.Liu,Flowdiffuser:Advancing optical flow estimation with diffusion models, in: Proceedings of the IEEE/CVFConferenceonComputerVisionandPatternRecognition, pp. 19167–19176

  29. [37]

    Z.Zhao,L.Alzubaidi,J.Zhang,Y.Duan,Y.Gu,Acomparisonreview oftransferlearningandself-supervisedlearning:Definitions,applica- tions, advantages and limitations, Expert Systems with Applications 242 (2024) 122807

  30. [38]

    Y. Hu, Q. Jia, Y. Yao, Y. Lee, M. Lee, C. Wang, X. Zhou, R. Xie, F. R. Yu, Industrial internet of things intelligence empowering smart manufacturing: A literature review, IEEE Internet of Things Journal 11 (2024) 19143–19167

  31. [39]

    X. Han, Z. Zhang, N. Ding, Y. Gu, X. Liu, Y. Huo, J. Qiu, Y. Yao, A. Zhang, L. Zhang, et al., Pre-trained models: Past, present and future, AI Open 2 (2021) 225–250

  32. [40]

    Z.Teed,J.Deng, Raft:Recurrentall-pairsfieldtransformsforoptical flow, in: Computer Vision–ECCV 2020: 16th European Conference, Glasgow,UK,August23–28,2020,Proceedings,PartII16,Springer, pp. 402–419

  33. [41]

    Oquab, L

    M. Oquab, L. Bottou, I. Laptev, J. Sivic, Learning and transferring mid-levelimagerepresentationsusingconvolutionalneuralnetworks, in: Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 1717–1724

  34. [42]

    Chang, J

    H. Chang, J. Han, C. Zhong, A. M. Snijders, J.-H. Mao, Unsuper- visedtransferlearningviamulti-scaleconvolutionalsparsecodingfor biomedical applications, IEEE transactions on pattern analysis and Zhu et al.: Preprint submitted to Elsevier Page 11 of 12 PIV-FlowDiffuser:Transfer...

  35. [43]

    L.N.Smith,N.Topin,Super-convergence:Veryfasttrainingofneural networks using large learning rates, in: Artificial intelligence and machine learning for multi-domain operations applications, volume 11006, SPIE, pp. 369–386

  36. [44]

    3061–3070

    M.Menze,A.Geiger, Objectsceneflowforautonomousvehicles, in: Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 3061–3070

  37. [45]

    Rubbert, M

    A. Rubbert, M. Albers, W. Schröder, Streamline segment statistics propagation in inhomogeneous turbulence, Physical Review Fluids 4 (2019) 034605

  38. [46]

    Sharmin, R

    N. Sharmin, R. Brad, Optimal filter estimation for lucas-kanade optical flow, Sensors 12 (2012) 12694–12709

  39. [47]

    J.Lu,H.Yang,Q.Zhang,Z.Yin, Anaccurateopticalflowestimation of piv using fluid velocity decomposition, Experiments in Fluids 62 (2021) 1–16

  40. [48]

    T.-W. Hui, X. Tang, C. C. Loy, Liteflownet: A lightweight convolu- tional neural network for optical flow estimation, in: Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 8981–8989

  41. [49]

    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, Springer, pp. 668–685

  42. [50]

    Scharnowski, C

    S. Scharnowski, C. J. Kähler, On the effect of curved streamlines on the accuracy of piv vector fields, Experiments in fluids 54 (2013) 1–11

  43. [51]

    Y. Wald, A. Feder, D. Greenfeld, U. Shalit, On calibration and out- of-domaingeneralization, Advancesinneuralinformationprocessing systems 34 (2021) 2215–2227. Zhu et al.: Preprint submitted to Elsevier Page 12 of 12

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.