Pith. sign in

REVIEW 4 major objections 8 minor 46 references

Joint Velocity Slope Diffusion Prior for Structurally Constrained Velocity Model Building

T0 review · 4 major / 8 minor · reviewed 2026-07-11 · grok-4.5

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read A joint velocity–slope diffusion prior, guided by sparse wells and plane-wave structure, reconstructs more continuous high-resolution subsurface velocity models than classical structural preconditioning alone.

desk verdict Solid methods stitch of known pieces into a joint velocity–slope diffusion prior; Volve train–test overlap and purely visual metrics leave the claimed gain unquantified. read the letter →

arxiv 2607.04982 v1 pith:NQZZCEY5 submitted 2026-07-06 physics.geo-ph cs.AIcs.LG

Francesco Brandolin , Tariq Alkhalifah This is my paper · ORCID
classification physics.geo-phcs.AIcs.LG
keywords velocitymodelbuildingdiffusionmodelsplane-wavedestructionstructuralpreconditioningwell-loginterpolationDDIMjointvelocity-slopepriorgeophysicalinversion
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

High-resolution subsurface velocity models are needed for reservoir work and monitoring, yet surface seismic data are band-limited and wells are sparse. This paper presents a unified reconstruction that paints well values along local geological dips, enforces a plane-wave PDE so velocity stays constant along those dips, and steers a diffusion sampler with a generative prior trained on paired velocity and slope fields. Guidance from the well-fitting inverse problem corrects only the velocity channel during DDIM sampling; the joint prior keeps slope and structure consistent. On a Volve synthetic model and Viking Graben field data, the combination improves lateral continuity, structural realism, and blind-well agreement relative to conventional structurally preconditioned inversion, while inference stays practical once the prior is trained.

What carries the argument

Joint velocity–slope diffusion prior with velocity-only measurement guidance: the network learns (velocity, plane-wave-destruction slope) pairs; each DDIM reverse step forms a clean estimate, solves a structurally preconditioned Tikhonov system (spray operator S plus plane-wave PDE operator D(γ)) to correct only the velocity channel, then recomposes the next noisy state so structure remains coupled.

What would settle it

On a hold-out field line with blind wells, if diffusion-guided models with plane-wave PDE fail to improve blind-log match and reflection continuity over classical structural preconditioning—especially when training geology mismatches the target—the claimed gains in continuity and realism would be falsified.

Watch

Extended reading notes

Core claim

Coupling plane-wave PDE regularization and structurally preconditioned least-squares well fitting with measurement-guided diffusion posterior sampling under a joint velocity–slope generative prior yields velocity models with better structural continuity, lateral consistency, and geological realism from sparse well logs than conventional structural preconditioning, at practical DDIM inference cost.

Load-bearing premise

The method assumes the trained velocity–slope prior and the slopes used to build the structural operators correctly represent the target geology, so well information is painted along the right dips rather than the wrong ones.

Editorial extensions

If this is right

  • Sparse well logs can be propagated farther along geological structure without isotropic oversmoothing.
  • Slopes from a migrated image supply stronger structural guidance than slopes from a smooth interpolated starting model when the background lacks dip detail.
  • After one-time prior training, DDIM inference cost becomes comparable to classical preconditioned inversion.
  • Explicit PDE structural constraints plus a learned prior can reduce early reliance on migration images that themselves depend on an accurate background velocity.
  • Joint generative modeling of velocity and dip offers a template for other ill-posed seismic inversions that need structural consistency.

Reading between the lines

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

  • The same joint prior could condition multiparameter elastic or time-lapse velocity updates if training pairs include those attributes.
  • Steep dips and faulted zones are natural stress tests of whether the PDE and spray operators still dominate the generative prior.
  • Training-set diversity is the practical bottleneck; domains far from the training mix may need adaptation rather than pure transfer.
  • Explicitly updating slopes from the evolving velocity during reverse sampling, as the discussion suggests, is a direct next experiment for fuller structural coupling.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 8 minor

Summary. The paper proposes a unified framework for reconstructing high-resolution velocity models from sparse well logs by combining (i) plane-wave PDE regularization, (ii) structurally preconditioned Tikhonov inversion with a slope-guided spray operator S, and (iii) measurement-guided DDIM posterior sampling under a joint two-channel velocity–slope diffusion prior trained with PWD slopes. Local slopes enter both the structural operators (S and D(γ)) and the generative state; guidance is applied only to the velocity clean estimate (Eqs. 9–15), after which the reverse step continues jointly. Experiments on the Volve synthetic model and the Viking Graben field line (with Well 4 conditioning and Well 5 held out) are used to argue improved structural continuity, lateral consistency, and geological realism relative to conventional structurally preconditioned inversion, with practical inference via few-step DDIM.

Significance. If the claimed gains hold under proper generalization tests, the work is a useful contribution to well-constrained velocity model building: it couples a joint generative prior with explicit plane-wave structural physics rather than relying solely on RTM-derived structure (as in the authors’ prior work), and it demonstrates a field-data path with a blind well and RTM-initialized slopes. Strengths include a clear stacked least-squares formulation (Eqs. 9–11), a concrete velocity-only guidance recipe (Eqs. 13–15), and an honest Discussion of slope quality and prior representativeness. The main scientific value is methodological integration rather than a new physical principle; that value depends on showing that improvements are not artifacts of prior leakage or purely visual preference.

major comments (4)
  1. §3.1 states that the 5000-model training corpus includes Volve; §3.2 then reconstructs “the Volve synthetic model” as the only ground-truth synthetic showcase. No hold-out, leave-one-family-out, or domain-shift protocol is described. Because the central claim is relative improvement over structural inversion under a learned prior, this train–test overlap is load-bearing: the Volve figures cannot establish generalization of the joint prior. A non-overlapping synthetic (or explicit leave-Volve-out retraining) is needed before the synthetic results can support the Abstract claim.
  2. §3.2–3.3 and Figs. 3–8 support “improved structural continuity, lateral consistency, and geological realism” almost entirely by visual comparison and overlaid well profiles. No RMSE/MAE/SSIM (or equivalent) versus true Volve velocity, no quantitative blind-well misfit at Well 5, and no seismic residual norms for Figs. 4, 6, 8 are reported. Without such metrics—and preferably an ablation isolating PW-PDE, structural preconditioning alone, and the joint slope channel—the magnitude and even existence of a genuine gain over conventional preconditioned inversion remain unestablished.
  3. The joint velocity–slope prior is a stated contribution (Abstract; §2.1; contribution bullets), yet guidance never corrects the slope channel (Eq. 14: γ̃0 = γ̂0), and Discussion §4 notes that generated slopes often stay close to the initial structural estimate. The manuscript does not show that joint training improves reconstructions relative to a velocity-only diffusion prior with the same S/D(γ) operators. A controlled comparison (joint vs velocity-only prior; fixed vs recomputed slopes) is needed to justify the two-channel design as load-bearing rather than incidental.
  4. Free parameters κ, λ, μ, η, warm-start t/T, and LSQR iteration count are fixed by statement (§3.2: κ=10−6, λ=0.01, μ=0.6, η=0.3, T=20) without sensitivity or stability analysis. Given that Discussion §4 already stresses strong dependence on slope quality (initial-velocity PWD vs RTM), the relative ranking of methods in Figs. 5 and 7 could shift under modest retuning. At least a limited sensitivity study on κ and μ (and slope source) should accompany the field claims.
minor comments (8)
  1. Throughout: “V olve” appears with a spurious space (Abstract, §1, §3.2, figure captions); fix consistently to “Volve”.
  2. §3.3: “trevltime tomography” → “traveltime tomography”; “full-wavefor inversion” → “full-waveform inversion”.
  3. §6 Acknowledgment: “DeepWave sponsors fort their support” → “for their support”.
  4. Eq. (1)–(2) and surrounding text: γγγ notation is heavy; a single bold γ would improve readability without loss of meaning.
  5. Fig. 3i / 5g–h / 7g–h: well profiles would be clearer with a residual panel or tabulated misfit; currently the eye must judge “higher resolution agreement.”
  6. §2.3 Eq. (8): the plane-wave PDE is written as an approximate equality; state the discrete residual norm used in D(γ) more explicitly (finite-difference stencil, boundary treatment).
  7. References [43] is cited as arXiv:2603.01231 (future-dated relative to this manuscript’s stamp); ensure citation metadata and priority relative to the present work are accurate.
  8. §3.1: “500 training epochs” in Discussion vs “50 epochs” in the training paragraph—reconcile the training budget statement.

Circularity Check

1 steps flagged · score 2.0 of 10

Mild Volve train–test overlap can inflate synthetic “geological realism”; the inversion/guidance math is not circular by construction.

  1. fitted input called prediction [§3.1 training set; §3.2 Volve evaluation; Fig. 3i / text claiming closer match to true model]
    "The training dataset consists of 5000 high-resolution velocity models drawn from diverse geological settings, including SEAM Arid, SEAM Arid Barrett, SEAM Phase I, Otway, and Volve. ... We evaluate the proposed guided diffusion strategy on the Volve synthetic model ... the diffusion-based reconstruction better preserves the surrounding geological structure and produces a model closer to the true velocity away from the direct well constraints."

    The generative prior is fit on a corpus that explicitly includes Volve, then the same Volve synthetic is the only ground-truth showcase used to claim improved geological realism and closer match to the true model under that prior. High-frequency structure attributed to the “learned joint prior” is therefore partly statistically forced by training-distribution membership rather than independent generalization; the relative visual improvement over structural inversion alone can still be real, but the absolute “closer to truth” claim is contaminated by construction of the train/eval split.

full rationale

The load-bearing reconstruction is not tautological. Sparse well residuals r_w = w − Mv, the structural spray S(γ), and the plane-wave PDE operator D(γ) are external constraints; the Gauss–Newton update (Eqs. 9–14) and DDIM guidance inject those measurements into a pre-trained prior rather than redefining the target as the input. Blind Well 5 on Viking is held out of conditioning, so field validation is not forced by construction. Self-citations ([43], PINN-slope papers) are motivational/prior-art, not uniqueness theorems that forbid alternatives. The only mild circularity is evaluation contamination: Volve is listed in the 5000-model training set and then used as the sole ground-truth synthetic showcase, so claims that diffusion yields models “closer to the true model under the learned joint prior” partly reflect training-distribution match rather than pure out-of-distribution reconstruction. That is a fitted-distribution / evaluation issue, not a derivation that reduces Eq. X to Eq. Y by definition. Score 2 (not 0) for that single non-load-bearing contamination; core method remains self-contained against external well and PDE constraints.

Assumptions & free parameters 5 free parameters · 4 assumptions · 1 invented entities

The central claim rests on standard inverse-problem and diffusion machinery plus several hand-chosen regularization/guidance weights and the assumption that a finite training corpus of velocity–slope pairs plus PWD/RTM slopes adequately represent target geology. No new physical particles or forces are postulated; the ‘joint prior’ is a learned statistical object, not an independent physical entity.

free parameters (5)
  • plane-wave PDE weight κ = 1e-6 (Volve)
    Controls strength of D(γ)SΔt term in Eq. 9–10; set to 10⁻⁶ for Volve without a systematic selection procedure.
  • Tikhonov damping λ = 0.01 (Volve)
    Damping in preconditioned domain (Eq. 9–11); chosen as 0.01 for Volve.
  • guidance weight μ = 0.6 (Volve)
    Scales the clean-sample velocity correction Δv in Eq. 14; set to 0.6 for Volve.
  • DDIM stochasticity η and step count = η≈0.3–0.8; T≈20–200
    η and number of reverse steps (e.g., η=0.3, T=20 effective steps for Volve; η=0.8, 200 steps in unconditional tests) are hand-tuned for stability vs detail.
  • training corpus composition and augmentation = 5000 models, 256×512 patches, 50 epochs, 1000 timesteps
    5000 models from SEAM/Otway/Volve plus flips and background trends define the learned prior; choice of which models and patches is a free design choice that shapes all ‘geological realism’ claims.
assumptions (4)
  • domain assumption Plane-wave PDE ∂v/∂x + γ ∂v/∂z ≈ 0 adequately enforces structural continuity of velocity along local dip.
    Introduced in §2.3 Eq. 8 and used to build D(γ); standard in structure-oriented processing but approximate in complex geology.
  • domain assumption Structural reparameterization v = S t with plane-wave spray operator S = P^H P guided by local slopes correctly propagates sparse well information.
    §2.2 following Chen et al. [18]; load-bearing for both the linear inversion and the diffusion guidance residual.
  • ad hoc to paper A DDPM/DDIM trained on joint (v, γ) pairs yields a generative prior that remains valid under measurement-guided posterior sampling for field data.
    §2.1 and §3.1; standard DPS assumptions [41,42] applied to a new two-channel geophysical state.
  • domain assumption Local slopes from PWD (or from RTM images) are sufficiently accurate to construct S and D(γ) even when the background velocity is smooth or early-stage.
    Used throughout §3; Discussion §4 explicitly notes strong sensitivity to slope quality.
invented entities (1)
  • joint velocity–slope diffusion generative prior (two-channel DDPM state)
    purpose: Encode statistical coupling between high-resolution velocity and local dip so reverse sampling stays structurally coherent while only velocity is measurement-guided.
    Methodological construct trained on paired patches; not an independent physical object. independent_evidence is false because its only validation is the same reconstruction experiments that use it.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Joint Velocity Slope Diffusion Prior for Structurally Constrained Velocity Model Building." pith.science (2026). https://pith.science/paper/NQZZCEY5

@misc{pith2026260704982,
  author       = {Pith},
  title        = {Pith review of: Joint Velocity Slope Diffusion Prior for Structurally Constrained Velocity Model Building},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/NQZZCEY5}},
  note         = {Machine review of arXiv:2607.04982}
}
read the original abstract

High-resolution velocity models are crucial for reservoir characterization and subsurface delineation. However, the band limited nature of our surface recorded data limits resolution. Utilizing well measurements to enhance the resolution of our subsurface models is an important objective. To this end, we present a diffusion-guided framework for structurally preconditioned velocity-model reconstruction from sparse well-log information. The proposed approach combines plane-wave PDE regularization, structurally preconditioned inversion, and measurement-guided diffusion posterior sampling within a unified formulation. Local structural slopes estimated through plane-wave destruction are used both to propagate well information along geological dip directions and to guide the diffusion sampling process through a joint velocity--slope generative prior. Numerical experiments on the Volve synthetic model and the Viking Graben field dataset demonstrate that the proposed framework improves structural continuity, lateral consistency, and geological realism compared with conventional structurally preconditioned inversion approaches while maintaining computationally practical inference through DDIM sampling.

Figures

Figures reproduced from arXiv: 2607.04982 by the authors.

Figure 1
Figure 1. Unconditionals samples from the learned velocity–slope prior: generated velocity, predicted slope, and slope [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. (a) Initial smooth velocity model used as input to the diffusion process. (b) Velocity realization generated by [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Velocity reconstruction results on the Volve example. (a) Original velocity model. (b) Initial smooth [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Comparison of modeled seismic data. (a) Data computed from the initial velocity model. (b) Data [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]
Figure 5
Figure 5. Figure 5: Velocity reconstruction results on the Viking Graben field dataset. (a) Initial smooth velocity model [PITH_FULL_IMAGE:figures/full_fig_p009_5.png]
Figure 6
Figure 6. Figure 6: Comparison of modeled seismic data for the Viking Graben field example. (a) Data computed from the initial [PITH_FULL_IMAGE:figures/full_fig_p010_6.png]
Figure 7
Figure 7. Figure 7: Velocity reconstruction results on the Viking Graben field dataset using RTM-derived structural guidance. (a) [PITH_FULL_IMAGE:figures/full_fig_p011_7.png]
Figure 8
Figure 8. Figure 8: Comparison of modeled seismic data for the Viking Graben field example using RTM-derived structural [PITH_FULL_IMAGE:figures/full_fig_p012_8.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

46 extracted references · 2 linked inside Pith

  1. [1]

    C. H. Dix. Seismic velocities from surface measurements.Geophysics, 20(1):68–86, 1955

    1955

  2. [2]

    Turhan Taner and Fulton Koehler

    M. Turhan Taner and Fulton Koehler. Velocity spectra-digital computer derivation and applications of velocity functions.Geophysics, 34(6):859–881, 12 1969

    1969

  3. [3]

    T. N. Bishop, K. P. Bube, R. T. Cutler, R. T. Langan, P. L. Love, J. R. Resnick, R. T. Shuey, D. A. Spindler, and H. W. Wyld. Tomographic determination of velocity and depth in laterally varying media.Geophysics, 50(6):903–923, 1985

    1985

  4. [4]

    Zelt and R

    Colin A. Zelt and R. Bruce Smith. Seismic traveltime inversion for 2-d crustal velocity structure.Geophysical Journal International, 108(1):16–34, 1992

    1992

  5. [5]

    Kamal M. AlYahya. Velocity analysis by iterative profile migration. SEG Technical Program Expanded Abstracts 1988:901–903, 01 1988

    1988

  6. [6]

    Christopher Stork and Robert W. Clayton. Migration velocity analysis.Geophysics, 56(12):1961–1972, 1991

    1961

  7. [7]

    Angle-domain common image gathers for anisotropic migration

    Biondo Biondi. Angle-domain common image gathers for anisotropic migration. SEG Technical Program Expanded Abstracts 2005:1922–1925, 01 2005

    2005

  8. [8]

    An overview of full-waveform inversion in exploration geophysics.Geophysics, 74(6):WCC1–WCC26, 2009

    Jean Virieux and Stéphane Operto. An overview of full-waveform inversion in exploration geophysics.Geophysics, 74(6):WCC1–WCC26, 2009

    2009

Show all 46 references
  1. [9]

    Target-oriented full-waveform inversion using marchenko redatumed wavefields.Geophysical Journal International, 223(2):792–810, 07 2020

    Tianci Cui, James Rickett, Ivan Vasconcelos, and Ben Veitch. Target-oriented full-waveform inversion using marchenko redatumed wavefields.Geophysical Journal International, 223(2):792–810, 07 2020

    2020

  2. [10]

    V oronoi

    G. V oronoi. Nouvelles applications des paramètres continus à la théorie des formes quadratiques. premier mémoire. sur quelques propriétés des formes quadra- tiques positives parfaites.Journal für die reine und an- gewandte Mathematik, 133:97–102, 1908

    1908

  3. [11]

    D. G. Krige. A statistical approach to some mine val- uation and allied problems on the witwatersrand: M.sc. dissertation.M.Sc. dissertation, University of the Witwatersrand., 1951

    1951

  4. [12]

    D. Shepard. A two-dimensional interpolation func- tion for irregularly-spaced data.Proceedings of the 23rd ACM National Conference, ACM, page 517–524, 1968

    1968

  5. [13]

    Anisotropic model building with wells and horizons: Gulf of mexico case study comparing different approaches.The Leading Edge, 29(12):1450–1460, 12 2010

    Andrey Bakulin, Yangjun (Kevin) Liu, Olga Zdraveva, and Kevin Lyons. Anisotropic model building with wells and horizons: Gulf of mexico case study comparing different approaches.The Leading Edge, 29(12):1450–1460, 12 2010

    2010

  6. [14]

    3d interpolation and extrapolation of sparse well data using rock-physics principles: Applications to anisotropic velocity model building.The Leading Edge, 33(3):248–254, 03 2014

    Ran Bachrach and Konstantin Osypov. 3d interpolation and extrapolation of sparse well data using rock-physics principles: Applications to anisotropic velocity model building.The Leading Edge, 33(3):248–254, 03 2014

    2014

  7. [15]

    Building 3d subsurface models conforming to seismic structural and stratigraphic features.Geo- physics, 82(3):IM21–IM30, 03 2017

    Xinming Wu. Building 3d subsurface models conforming to seismic structural and stratigraphic features.Geo- physics, 82(3):IM21–IM30, 03 2017

    2017

  8. [16]

    Image-guided 3d interpolation of borehole data.Seg Technical Program Expanded Abstracts, 2010

    Dave Hale. Image-guided 3d interpolation of borehole data.Seg Technical Program Expanded Abstracts, 2010

    2010

  9. [17]

    Naeini and D

    E. Naeini and D. Hale. Image- and horizon-guided interpolation.Geophysics, 80:V47–V56, 2015. 14

    2015

  10. [18]

    Geological structure guided well log in- terpolation for high-fidelity full waveform inversion.Geophysical Journal International, 207(2):1313–1331, 09 2016

    Yangkang Chen, Hanming Chen, Kui Xiang, and Xiaohong Chen. Geological structure guided well log in- terpolation for high-fidelity full waveform inversion.Geophysical Journal International, 207(2):1313–1331, 09 2016

    2016

  11. [19]

    Creating detailed subsurface models using predictive image- guided well-log interpolation.Interpretation, 5(3):T279–T285, 04 2017

    Parvaneh Karimi, Sergey Fomel, and Rui Zhang. Creating detailed subsurface models using predictive image- guided well-log interpolation.Interpretation, 5(3):T279–T285, 04 2017

    2017

  12. [20]

    Joint preconditioned least-squares inversion of simultaneous source time-lapse seismic data sets

    Gboyega Ayeni, Yaxun Tang, and Bindo Biondi. Joint preconditioned least-squares inversion of simultaneous source time-lapse seismic data sets. SEG Technical Program Expanded Abstracts 2009:3914–3918, 01 2009

    2009

  13. [21]

    Constrained full-waveform inversion by model reparameteri- zation.Geophysics, 77(2):R117–R127, 03 2012

    Antoine Guitton, Gboyega Ayeni, and Esteban Díaz. Constrained full-waveform inversion by model reparameteri- zation.Geophysics, 77(2):R117–R127, 03 2012

    2012

  14. [22]

    Regularized tomographic inversion with geological constraints.Geophysical Prospecting, 65(1):305–315, 2017

    Vincenzo Lipari, Domenico Urbano, Emmanuel Spadavecchia, Jacopo Panizzardi, and Nicola Bienati. Regularized tomographic inversion with geological constraints.Geophysical Prospecting, 65(1):305–315, 2017

    2017

  15. [23]

    Pinheiro, Edwin F

    Mesay Geletu Gebre, Daniel N. Pinheiro, Edwin F. Duarte, Carlos A. N. da Costa, Ramon C. F. Araújo, Danyelle da Silva, and João M. de Araújo. Preconditioning full wave inversion using geological structure-oriented gradient filtering.IEEE Geoscience and Remote Sensing Letters, ...

    2025

  16. [24]

    Estimators for orientation and anisotropy in digitized images

    Lucas J Van Vliet and Piet W Verbeek. Estimators for orientation and anisotropy in digitized images. InASCI, volume 95, pages 16–18, 1995

    1995

  17. [25]

    Marfurt, R

    Kurt J. Marfurt, R. Lynn Kirlin, Steven L. Farmer, and Michael S. Bahorich. 3-d seismic attributes using a semblance-based coherency algorithm.Geophysics, 63(4):1150–1165, 08 1998

    1998

  18. [26]

    Applications of plane-wave destruction filters.Geophysics, 67:1946–1960, 2002

    S Fomel. Applications of plane-wave destruction filters.Geophysics, 67:1946–1960, 2002

    1946

  19. [27]

    Griffiths, André J.-M

    Matthew P. Griffiths, André J.-M. Pugin, and Dariush Motazedian. Estimating local slope in the time-frequency domain: Velocity-independent seismic imaging in the near surface.Geophysics, 85(5):U99–U107, 07 2020

    2020

  20. [28]

    Seismic volumetric local slope estimation using multiscale gradient structure tensor.IEEE Geoscience and Remote Sensing Letters, 20:1–5, 2023

    Yu He, Ping Yang, Feng Qian, Weifeng Geng, Bingwei Zheng, Xiaoqiao Ren, and Guangmin Hu. Seismic volumetric local slope estimation using multiscale gradient structure tensor.IEEE Geoscience and Remote Sensing Letters, 20:1–5, 2023

    2023

  21. [29]

    A deep learning network for estimation of seismic local slopes.Petroleum Science, 18:T279–T285, 02 2021

    Wei-Lin Huang, Fei Gao, Jian-Ping Liao, and Xiao-Yu Chuai. A deep learning network for estimation of seismic local slopes.Petroleum Science, 18:T279–T285, 02 2021

    2021

  22. [30]

    Saad, and Yangkang Chen

    Shaohuan Zu, Junxing Cao, Sergey Fomel, Liuqing Yang, Omar M. Saad, and Yangkang Chen. Robust local slope estimation by deep learning.Geophysical Prospecting, 70(5):847–864, 2022

    2022

  23. [31]

    Bahia and M

    B. Bahia and M. Sacchi. Neural estimation of seismic local slopes. 2022(1):1–5, 2022

    2022

  24. [32]

    Pinnslope: Physics-informed neural network slope prediction and interpolation with positional encoding

    Francesco Brandolin, Matteo Ravasi, and Tariq Alkhalifah. Pinnslope: Physics-informed neural network slope prediction and interpolation with positional encoding. SEG/AAPG International Meeting for Applied Geoscience & Energy:SEG–2023–3907270, 08 2023

    2023

  25. [33]

    Pinnslope: Seismic data interpolation and local slope estimation with physics informed neural networks.GEOPHYSICS, 89(4):V331–V345, 2024

    Francesco Brandolin, Matteo Ravasi, and Tariq Alkhalifah. Pinnslope: Seismic data interpolation and local slope estimation with physics informed neural networks.GEOPHYSICS, 89(4):V331–V345, 2024

    2024

  26. [34]

    Francesco Brandolin, Matteo Ravasi, and Tariq Alkhalifah. Slope assisted physics-informed neural networks for seismic signal separation with applications on ground roll removal and interpolation.Geophysical Prospecting, 73(5):1337–1363, 2025

    2025

  27. [35]

    Deep-learning tomography.The Leading Edge, 37(1):58–66, 01 2018

    Mauricio Araya-Polo, Joseph Jennings, Amir Adler, and Taylor Dahlke. Deep-learning tomography.The Leading Edge, 37(1):58–66, 01 2018

    2018

  28. [36]

    Velocity model building with a modified fully convolutional network

    Wenlong Wang, Fangshu Yang, and Jianwei Ma. Velocity model building with a modified fully convolutional network. 2018 SEG International Exposition and Annual Meeting:SEG–2018–2997566, 10 2018

    2018

  29. [37]

    Deep-learning inversion: A next-generation seismic velocity model building method.Geophysics, 84(4):R583–R599, 06 2019

    Fangshu Yang and Jianwei Ma. Deep-learning inversion: A next-generation seismic velocity model building method.Geophysics, 84(4):R583–R599, 06 2019

    2019

  30. [38]

    Lukas Mosser, Olivier Dubrule, and Martin J. Blunt. Stochastic seismic waveform inversion using generative adversarial networks as a geological prior.Mathematical Geosciences, 52:52–79, 2020

    2020

  31. [39]

    Full-waveform inversion using a learned regularization.IEEE Transactions on Geoscience and Remote Sensing, 61:1–15, 2023

    Pengpeng Sun, Fangshu Yang, Hongxian Liang, and Jianwei Ma. Full-waveform inversion using a learned regularization.IEEE Transactions on Geoscience and Remote Sensing, 61:1–15, 2023

    2023

  32. [40]

    Wavenumber-aware diffusion sampling to regularize multiparameter elastic full waveform inversion.Geophysical Journal International, 240(2):1215–1233, 12 2024

    Mohammad H Taufik and Tariq Alkhalifah. Wavenumber-aware diffusion sampling to regularize multiparameter elastic full waveform inversion.Geophysical Journal International, 240(2):1215–1233, 12 2024

    2024

  33. [41]

    McCann, Scott M

    Hyungjin Chung, Jaewook Kim, Michael T. McCann, Scott M. Klasky, and Jong Chul Ye. Diffusion posterior sampling for general noisy inverse problems. InICLR, 2023. 15

    2023

  34. [42]

    Geophysical inverse problems with measurement-guided diffusion models.arXiv:2501.04881, 2025

    Matteo Ravasi. Geophysical inverse problems with measurement-guided diffusion models.arXiv:2501.04881, 2025

    2025 arXiv

  35. [43]

    Velocity model building and editing with guided denoising diffusion implicit models.arXiv preprint arXiv:2603.01231, (2603.01231), 2026

    Francesco Brandolin and Tariq Alkhalifah. Velocity model building and editing with guided denoising diffusion implicit models.arXiv preprint arXiv:2603.01231, (2603.01231), 2026

    2026 arXiv

  36. [44]

    Denoising diffusion implicit models

    Jiaming Song, Chenlin Meng, and Stefano Ermon. Denoising diffusion implicit models. InICLR, 2021

    2021

  37. [45]

    Denoising diffusion probabilistic models

    Jonathan Ho, Ajay Jain, and Pieter Abbeel. Denoising diffusion probabilistic models. InNeurIPS, page (33), 2020

    2020

  38. [46]

    Madiba and George A

    Gislain B. Madiba and George A. McMechan. Processing, inversion, and interpretation of a 2d seismic data set from the north viking graben, north sea.Geophysics, 68(3):837–848, 01 2003. 16

    2003

Pith tools

Reviewed July 11, 2026 · model on record in the stance chip above.