pith. sign in

arxiv: 2605.01273 · v3 · pith:OZAXFIOKnew · submitted 2026-05-02 · ⚛️ physics.geo-ph

Learning Stratigraphically Consistent Relative Geologic Time from 3D Seismic Data via Sinusoidal Mapping

Yimin Dou , Xinming Wu , Hui Gao , Zhengfa Bi This is my paper

Pith reviewed 2026-05-21 09:01 UTC · model grok-4.3

classification ⚛️ physics.geo-ph
keywords relative geologic time3D seismic datadeep learningsinusoidal mappingstratigraphic consistencyhorizon correlationadversarial lossesgeological modeling
0
0 comments X

The pith

Mapping relative geologic time estimation into a sinusoidal space lets deep networks capture fine horizons and global stratigraphic order from seismic volumes.

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

The paper introduces RGT-Est, a deep-learning framework that reframes the task of estimating relative geologic time from 3D seismic data. Standard regression networks using pixel-wise losses often blur thin layers and produce inconsistent layer ordering across complex structures. By shifting the optimization target to a differentiable sinusoidal representation, the method directly encodes the periodic, layered character of stratigraphy. Joint pointwise, perceptual, and adversarial losses then enforce local accuracy, inter-layer continuity, and overall topological plausibility. An optional guidance module can incorporate sparse horizon priors, and tests on field surveys with faults, folds, and clinoforms show improved performance over earlier AI approaches.

Core claim

The central claim is that converting the relative geologic time field into a sinusoidal space transfers the learning problem into a representation that naturally respects stratigraphic periodicity. This change, combined with multi-term losses in the new space, yields networks that discriminate thin horizons while maintaining global ordering, outperforming prior regression-based methods on both synthetic and real data with and without sparse horizon constraints.

What carries the argument

The sinusoidal mapping of the relative geologic time field, which embeds the continuous stratigraphic sequence into a periodic differentiable space so that network outputs can be decoded back to consistent RGT values.

If this is right

  • Subsurface models built from the resulting RGT fields will exhibit fewer crossing horizons and more reliable fault offsets.
  • Depositional system reconstruction can proceed with less manual editing because local layer thicknesses and terminations are better preserved.
  • Sparse 2-D or 3-D horizon picks can be used as soft constraints to raise correlation accuracy without requiring dense manual interpretation.
  • The same network generalizes across steeply dipping strata, unconformities, and clinoform geometries without retraining for each structural style.

Where Pith is reading between the lines

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

  • The sinusoidal embedding idea may transfer to other layered inverse problems such as velocity model building or electromagnetic sounding where periodicity is intrinsic.
  • Adversarial training in the transformed space could be combined with physics-informed constraints to reduce the need for synthetic pre-training.
  • Because the method already handles large unconformities, it may serve as a starting point for automatic sequence-stratigraphic boundary detection.

Load-bearing premise

The assumption that the sinusoidal transformation will preserve stratigraphic semantics without distortion and that the combined losses will simultaneously enforce local fidelity, layer consistency, and global plausibility.

What would settle it

Run the trained model on a field volume whose horizons have been independently picked and correlated by multiple interpreters; measure whether horizon-to-horizon ordering errors decrease measurably when the sinusoidal mapping and adversarial term are ablated.

Figures

Figures reproduced from arXiv: 2605.01273 by Hui Gao, Xinming Wu, Yimin Dou, Zhengfa Bi.

Figure 1
Figure 1. Figure 1: Overview of the proposed RGT-Est framework. A 3D HRNet backbone takes a 3D seismic volume with optional 2D / 3D horizon guidance as input ˆˆ ˆ [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 1
Figure 1. Figure 1: Overview of the proposed RGT-Est framework. A 3D HRNet backbone takes a 3D seismic volume with optional 2D / 3D horizon guidance as input and predicts a continuous RGT scalar field Rˆ. The Sinusoidal Mapping module maps Rˆ into a three-channel phase space via sin(fiRˆ) with f1=2.0, f2=1.0, f3=0.5, yielding the Sinusoidal RGT. Training jointly minimizes the MAE, perceptual, and adversarial losses over the P… view at source ↗
Figure 2
Figure 2. Figure 2: Gradient back-propagation comparison between the raw RGT space and the proposed sinusoidal space. A pretrained 3D HRNet is kept frozen, and ˆ ˆˆ [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 2
Figure 2. Figure 2: Gradient back-propagation comparison between the raw RGT space and the proposed sinusoidal space. A pretrained 3D HRNet is kept frozen, and the L1 and 3D LPIPS losses are computed either directly on Rˆ or on its sinusoidal encoding T (Rˆ); the gradient at Rˆ is recorded for comparison. (a) Inline slice of the seismic input, the GT RGT, the prediction, and the resulting L1/LPIPS gradient maps in both domain… view at source ↗
Figure 3
Figure 3. Figure 3: Qualitative comparison of RGT estimation on field seismic volumes. From left to right, each group presents the input seismic volume, the result of [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figure 3
Figure 3. Figure 3: Qualitative comparison of RGT estimation on field seismic volumes. From left to right, each group presents the input seismic volume, the result of DeepRGT† (re-implementation) (Bi et al., 2021), and the result of the proposed RGT-Est. For both methods, the displayed horizons are extracted as iso-surfaces from the estimated RGT fields and overlaid on the seismic volumes for visual comparison. The colored da… view at source ↗
Figure 4
Figure 4. Figure 4: Effect of stratigraphic constraints on RGT estimation. (a) Incorporating 2D horizon constraints into RGT-Est. The purple dashed curves denote [PITH_FULL_IMAGE:figures/full_fig_p010_4.png] view at source ↗
Figure 4
Figure 4. Figure 4: Effect of stratigraphic constraints on RGT estimation. The purple lines represent the input 2D horizon constraints, which also serve as reference lines for the horizons. (a) Incorporating 2D horizon constraints into RGT-Est. The purple dashed curves denote ground-truth horizons used for visual comparison, and the yellow boxes highlight regions where the constrained result better honors the target stratigra… view at source ↗
Figure 5
Figure 5. Figure 5: Representative RGT estimation results on challenging field surveys. From left to right, each row shows the input seismic volume, the estimated RGT [PITH_FULL_IMAGE:figures/full_fig_p011_5.png] view at source ↗
Figure 5
Figure 5. Figure 5: Representative RGT estimation results on challenging field surveys. From left to right, each row shows the input seismic volume, the estimated RGT field, and the horizons extracted from the estimated RGT field and overlaid on the seismic volume. The six rows corre￾spond to the Costa Rica survey, the Poseidon survey in Australia, two Netherlands surveys, and two field surveys from a region in China. These e… view at source ↗
Figure 6
Figure 6. Figure 6: Visualization of intermediate features learned by RGT-Est. For each example, the seismic volume is shown together with the RGB visualization [PITH_FULL_IMAGE:figures/full_fig_p012_6.png] view at source ↗
Figure 6
Figure 6. Figure 6: Visualization of intermediate features learned by RGT-Est. For each example, the seismic volume is shown together with the RGB visualization of the Stage 3 HRNet feature map after PCA along the channel dimension. The feature responses highlight geologically important structures, including faults, unconformities, horizons, slope bodies, and deformed stratigraphic intervals. This indicates that RGT￾Est learn… view at source ↗
read the original abstract

Relative Geologic Time (RGT) estimation from seismic data is a cornerstone of subsurface structural modeling, depositional evolution analysis, and reservoir characterization, supporting horizon correlation and depositional system reconstruction. Yet accurate RGT estimation remains challenging: RGT is intrinsically a topologically constrained continuous field, in which local errors readily propagate globally and distort the overall result. Conventional methods rely heavily on priors, attribute extraction, and manual interaction, leading to cumbersome workflows. Existing deep-learning approaches mostly use a regression formulation with pixel-wise MSE/MAE losses, which struggle to capture thin horizons and fail to model the stratigraphic semantics of the RGT field, yielding limited generalization and unstable ordering across diverse structural and depositional settings. We propose RGT-Est, a deep-learning framework that transfers the optimization target from the topologically constrained continuous field into a differentiable sinusoidal space, which explicitly encodes the periodic stratigraphic semantics of RGT and alleviates over-smoothing of fine horizons. Pointwise, perceptual, and adversarial losses are jointly imposed in this space to enforce local fidelity, inter-layer consistency, and global structural plausibility, providing both fine-horizon discrimination and global stratigraphic awareness. An optional horizon-guidance module further accepts sparse 2D or 3D horizons as priors. Trained on synthetic data and evaluated on field surveys with densely faulted zones, large unconformities, steeply dipping strata, folded deformations, and clinoforms, RGT-Est achieves state-of-the-art performance among AI-based methods without horizon constraints, and attains substantially higher horizon-correlation accuracy and global topological consistency once sparse priors are incorporated.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 1 minor

Summary. The paper introduces RGT-Est, a deep-learning framework for estimating Relative Geologic Time (RGT) from 3D seismic data. It transfers the optimization target into a differentiable sinusoidal space (encoding periodic stratigraphic semantics) and applies a combination of pointwise, perceptual, and adversarial losses to enforce local fidelity, inter-layer consistency, and global structural plausibility. An optional horizon-guidance module incorporates sparse priors. The method is trained on synthetic data and evaluated on field surveys with faults, unconformities, and complex structures, claiming state-of-the-art performance among AI-based methods without horizon constraints and substantially improved horizon-correlation accuracy and topological consistency with priors.

Significance. If the central claims hold, the sinusoidal mapping approach could meaningfully advance automated RGT estimation by better capturing stratigraphic periodicity and reducing over-smoothing compared to standard regression losses. This would support more reliable horizon correlation and structural modeling in subsurface interpretation, with potential to decrease reliance on manual priors in complex geological settings. The optional guidance module and synthetic-to-field transfer also offer practical flexibility.

major comments (2)
  1. [Framework description (abstract and §3)] The sinusoidal mapping (described in the abstract and framework as transferring RGT into differentiable sinusoidal space, likely via sin(2π·RGT) and cos(2π·RGT)) yields only the fractional part upon phase recovery with atan2. The manuscript does not detail an explicit mechanism for recovering or enforcing the integer component of RGT, nor does it show how the pointwise + perceptual + adversarial loss combination structurally penalizes inconsistencies across faults or unconformities. This leaves the claimed global topological consistency dependent on empirical behavior rather than a property of the mapping, directly affecting the central claim of stratigraphically consistent RGT.
  2. [Abstract and Evaluation section] The abstract states that RGT-Est achieves state-of-the-art performance among AI-based methods without horizon constraints and substantially higher accuracy with priors, yet provides no quantitative metrics, error bars, ablation studies on the loss weights or mapping, or implementation details for the sinusoidal transformation and losses. This weakens the evidential support for the performance claims and makes it difficult to assess whether the losses truly enforce the intended consistencies.
minor comments (1)
  1. [Abstract] The abstract would benefit from including at least one key quantitative result (e.g., correlation coefficient or topological error metric) to substantiate the performance claims.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive and detailed review. The comments highlight important points about clarity in the framework description and the presentation of results. We address each major comment below and indicate the revisions we will make.

read point-by-point responses

  1. Referee: [Framework description (abstract and §3)] The sinusoidal mapping (described in the abstract and framework as transferring RGT into differentiable sinusoidal space, likely via sin(2π·RGT) and cos(2π·RGT)) yields only the fractional part upon phase recovery with atan2. The manuscript does not detail an explicit mechanism for recovering or enforcing the integer component of RGT, nor does it show how the pointwise + perceptual + adversarial loss combination structurally penalizes inconsistencies across faults or unconformities. This leaves the claimed global topological consistency dependent on empirical behavior rather than a property of the mapping, directly affecting the central claim of stratigraphically consistent RGT.

    Authors: We appreciate the referee's precise observation on the properties of the sinusoidal mapping. The transformation to sin(2π·RGT) and cos(2π·RGT) is chosen specifically to embed the periodic stratigraphic semantics, with each 2π cycle corresponding to one stratigraphic interval. The network outputs a continuous field; the integer component is recovered via cumulative phase integration (unwrapping) along depth, which is stabilized by the perceptual and adversarial losses that penalize violations of layer ordering and continuity. These losses do structurally discourage inconsistencies at faults and unconformities by enforcing global structural plausibility in the mapped space. We acknowledge that the original manuscript did not sufficiently detail the unwrapping procedure or loss behavior at discontinuities. We will add an explicit subsection in §3 describing the phase recovery and include additional figures illustrating loss gradients across faults. revision: yes

  2. Referee: [Abstract and Evaluation section] The abstract states that RGT-Est achieves state-of-the-art performance among AI-based methods without horizon constraints and substantially higher accuracy with priors, yet provides no quantitative metrics, error bars, ablation studies on the loss weights or mapping, or implementation details for the sinusoidal transformation and losses. This weakens the evidential support for the performance claims and makes it difficult to assess whether the losses truly enforce the intended consistencies.

    Authors: We agree that the abstract would benefit from explicit quantitative support. The evaluation section and supplementary material already contain the requested elements: mean absolute error and horizon-correlation accuracy with standard deviations across multiple runs, ablation tables on loss weights and the sinusoidal mapping, and implementation specifics (including the exact sin/cos formulation and loss definitions) in §3. To directly address the concern, we will revise the abstract to summarize the key metrics and error bars while adding a sentence referencing the ablation results and implementation details. revision: yes

Circularity Check

0 steps flagged

No significant circularity: sinusoidal mapping is an explicit design choice, not a reduction to inputs

full rationale

The paper introduces RGT-Est as a methodological framework that transfers the RGT optimization target into a differentiable sinusoidal space to encode periodic stratigraphic semantics, then applies a combination of pointwise, perceptual, and adversarial losses. This mapping and loss design are presented as deliberate choices to address over-smoothing and capture inter-layer consistency, rather than any claimed derivation that reduces by construction to the target RGT field or fitted parameters. Training occurs on synthetic data with evaluation on independent field surveys, and the optional horizon-guidance module accepts external priors. No self-citations, uniqueness theorems, or ansatzes are invoked in a load-bearing way that would make the central performance claims tautological. The approach remains self-contained with external benchmarks and does not exhibit any of the enumerated circularity patterns.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The framework rests on the domain assumption that RGT possesses periodic stratigraphic semantics that a sinusoidal representation can capture. Typical deep-learning free parameters such as loss-term weights and network architecture choices are not specified in the abstract and must be assumed to exist.

free parameters (1)
  • weights for pointwise, perceptual, and adversarial losses
    The abstract states that these losses are jointly imposed but does not report how the relative weights are chosen or tuned.
axioms (1)
  • domain assumption RGT is intrinsically a topologically constrained continuous field in which local errors readily propagate globally
    Invoked in the opening paragraph of the abstract to motivate the problem.

pith-pipeline@v0.9.0 · 5822 in / 1352 out tokens · 62359 ms · 2026-05-21T09:01:13.122086+00:00 · methodology

Review history (3 revisions) →

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

79 extracted references · 79 canonical work pages · 3 internal anchors

  1. [1]

    Geophysics , volume =

    Geng, Zhicheng and Wu, Xinming and Shi, Yunzhi and Fomel, Sergey , title =. Geophysics , volume =. 2020 , doi =

    work page 2020

  2. [2]

    IEEE Transactions on Geoscience and Remote Sensing , volume =

    Yang, Jiarun and Wu, Xinming and Bi, Zhengfa and Geng, Zhicheng , title =. IEEE Transactions on Geoscience and Remote Sensing , volume =. 2023 , doi =

    work page 2023

  3. [3]

    Patel, M. D. and McMechan, G. A. , title =. Computers & Geosciences , volume =

    work page

  4. [4]

    Stark, T. J. , title =. 73rd Annual International Meeting, SEG, Expanded Abstracts , pages =

    work page

  5. [5]

    Stark, T. J. , title =. The Leading Edge , volume =

    work page

  6. [6]

    Petroleum Exploration and Development , volume =

    Zeng, Hongliu and Zhu, Xiaomin and Zhu, Rukai and Zhang, Qisong , title =. Petroleum Exploration and Development , volume =

    work page

  7. [7]

    Structural data constraints for implicit modeling of folds , journal =

    Grose, Laurent and Laurent, Gautier and Aill. Structural data constraints for implicit modeling of folds , journal =

    work page

  8. [8]

    and Spikes, K

    Bader, S. and Spikes, K. and Fomel, S. , title =. SEG Technical Program Expanded Abstracts , pages =

    work page

  9. [9]

    Geophysical Prospecting , volume =

    Karimi, Parvaneh and Fomel, Sergey , title =. Geophysical Prospecting , volume =

    work page

  10. [10]

    Zeng, Hongliu and Backus, M. M. and Barrow, K. T. and Tyler, N. , title =. Geophysics , volume =

    work page

  11. [11]

    and Hemstra, N

    de Bruin, G. and Hemstra, N. and Pouwel, A. , title =. The Leading Edge , volume =

    work page

  12. [12]

    Geophysics , volume =

    Wu, Xinming and Zhong, Guangzhi , title =. Geophysics , volume =

    work page

  13. [13]

    Geophysics , volume =

    Fomel, Sergey , title =. Geophysics , volume =

    work page

  14. [14]

    , title =

    Lomask, Jesse and Guitton, Antoine and Fomel, Sergey and Claerbout, Jon and Valenciano, Alejandro A. , title =. Geophysics , volume =

    work page

  15. [15]

    Geophysical Journal International , volume =

    Wu, Xinming and Janson, Xavier , title =. Geophysical Journal International , volume =

    work page

  16. [16]

    Geophysics , volume =

    Wu, Xinming and Fomel, Sergey , title =. Geophysics , volume =

    work page

  17. [17]

    Figueiredo, A. M. and Gattass, M. and Szenberg, F. , title =. 10th International Congress of the Brazilian Geophysical Society , pages =

    work page

  18. [18]

    Geophysics , volume =

    Lou, Yihuai and Zhang, Bo and Fang, Huilin and Cao, Danping and Wang, Kai and Huo, Zhizhong , title =. Geophysics , volume =

    work page

  19. [19]

    Geophysics , volume =

    Luo, Simon and Hale, Dave , title =. Geophysics , volume =

    work page

  20. [20]

    Geophysics , volume =

    Wu, Xinming and Luo, Simon and Hale, Dave , title =. Geophysics , volume =

    work page

  21. [21]

    Geophysics , volume =

    Wu, Xinming and Hale, Dave , title =. Geophysics , volume =

    work page

  22. [23]

    Geophysics , volume =

    Wu, Xinming and Liang, Luming and Shi, Yunzhi and Fomel, Sergey , title =. Geophysics , volume =

    work page

  23. [24]

    Geophysics , volume =

    Gao, Hang and Wu, Xinming and Liu, Guofeng , title =. Geophysics , volume =

    work page

  24. [25]

    Interpretation , volume =

    Shi, Yunzhi and Wu, Xinming and Fomel, Sergey , title =. Interpretation , volume =

    work page

  25. [26]

    Geophysics , volume =

    Yang, Fangshu and Ma, Jianwei , title =. Geophysics , volume =

    work page

  26. [27]

    Geophysical Journal International , volume =

    Geng, Zhicheng and Zhao, Zeyu and Shi, Yunzhi and Wu, Xinming and Fomel, Sergey and Sen, Mrinal , title =. Geophysical Journal International , volume =

    work page

  27. [28]

    Journal of Geophysical Research: Solid Earth , volume =

    Bi, Zhengfa and Wu, Xinming and Geng, Zhicheng and Li, Haishan , title =. Journal of Geophysical Research: Solid Earth , volume =

    work page

  28. [29]

    Geophysics , volume =

    Wu, Xinming and Geng, Zhicheng and Shi, Yunzhi and Pham, Nam and Fomel, Sergey and Caumon, Guillaume , title =. Geophysics , volume =

    work page

  29. [30]

    and Shechtman, Eli and Wang, Oliver , title =

    Zhang, Richard and Isola, Phillip and Efros, Alexei A. and Shechtman, Eli and Wang, Oliver , title =. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , pages =

    work page

  30. [32]

    Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , pages=

    Image-to-image translation with conditional adversarial networks , author=. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , pages=

    work page

  31. [33]

    Geometric

    Lim, Jae Hyun and Ye, Jong Chul , booktitle=. Geometric

    work page

  32. [34]

    International Conference on Learning Representations (ICLR) , year=

    Adam: A Method for Stochastic Optimization , author=. International Conference on Learning Representations (ICLR) , year=

    work page

  33. [35]

    Geophysics , volume=

    Seismic horizon tracking based on the TransUnet model , author=. Geophysics , volume=. 2025 , publisher=

    work page 2025

  34. [36]

    Journal of Petroleum Science and Engineering , volume=

    Variable seismic waveforms representation: Weak-supervised learning based seismic horizon picking , author=. Journal of Petroleum Science and Engineering , volume=. 2022 , publisher=

    work page 2022

  35. [37]

    IEEE Transactions on Geoscience and Remote Sensing , volume=

    A deep learning-based seismic horizon tracking method with uncertainty encoding and vertical constraint , author=. IEEE Transactions on Geoscience and Remote Sensing , volume=. 2024 , publisher=

    work page 2024

  36. [38]

    IEEE Transactions on Geoscience and Remote Sensing , volume=

    Seismic horizon identification using semi-supervised learning with virtual adversarial training , author=. IEEE Transactions on Geoscience and Remote Sensing , volume=. 2022 , publisher=

    work page 2022

  37. [39]

    IEEE Transactions on Geoscience and Remote Sensing , volume=

    Seismic attributes aided horizon interpretation using an ensemble dense inception transformer network , author=. IEEE Transactions on Geoscience and Remote Sensing , volume=. 2024 , publisher=

    work page 2024

  38. [40]

    IEEE Transactions on Geoscience and Remote Sensing , volume=

    Automatic extraction of seismic data horizon across faults , author=. IEEE Transactions on Geoscience and Remote Sensing , volume=. 2022 , publisher=

    work page 2022

  39. [42]

    804--808, 2018

    Bader, S., Spikes, K., and Fomel, S.: Missing well-log data prediction using B ayesian approach in the relative-geologic time domain, in: SEG Technical Program Expanded Abstracts, pp. 804--808, 2018

    work page 2018

  40. [43]

    Bi, Z., Wu, X., Geng, Z., and Li, H.: Deep relative geologic time: A deep learning method for simultaneously interpreting 3-D seismic horizons and faults, Journal of Geophysical Research: Solid Earth, 126, e2021JB021\,882, 2021

    work page 2021

  41. [44]

    de Bruin, G., Hemstra, N., and Pouwel, A.: Stratigraphic surfaces in the depositional and chronostratigraphic ( W heeler-transformed) domain, The Leading Edge, 26, 883--886, 2007

    work page 2007

  42. [45]

    Dou, Y.: RGT-Est: Datasets and Trained Model Weights for Learning Stratigraphically Consistent Relative Geologic Time from 3D Seismic Data via Sinusoidal Mapping [Dataset/Software], Zenodo, doi:10.5281/zenodo.20118902, [Dataset/Software], 2026

    work page doi:10.5281/zenodo.20118902 2026

  43. [46]

    M., Gattass, M., and Szenberg, F.: Seismic horizon mapping across faults with growing neural gas, in: 10th International Congress of the Brazilian Geophysical Society, pp

    Figueiredo, A. M., Gattass, M., and Szenberg, F.: Seismic horizon mapping across faults with growing neural gas, in: 10th International Congress of the Brazilian Geophysical Society, pp. 1476--1481, 2007

    work page 2007

  44. [47]

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

    work page 1946

  45. [48]

    Fomel, S.: Predictive painting of 3D seismic volumes, Geophysics, 75, A25--A30, 2010

    work page 2010

  46. [49]

    Gao, H., Wu, X., and Liu, G.: ChannelSeg3D : Channel simulation and deep learning for channel interpretation in 3D seismic images, Geophysics, 86, IM73--IM83, 2021

    work page 2021

  47. [50]

    Geng, Z., Wu, X., Shi, Y., and Fomel, S.: Deep learning for relative geologic time and seismic horizons, Geophysics, 85, WA87--WA100, doi:10.1190/geo2019-0252.1, 2020

    work page doi:10.1190/geo2019-0252.1 2020

  48. [51]

    Geng, Z., Zhao, Z., Shi, Y., Wu, X., Fomel, S., and Sen, M.: Deep learning for velocity model building with common-image gather volumes, Geophysical Journal International, 228, 1054--1070, 2021

    work page 2021

  49. [52]

    Grose, L., Laurent, G., Aill \`e res, L., Armit, R., Jessell, M., and Caumon, G.: Structural data constraints for implicit modeling of folds, Journal of Structural Geology, 104, 80--92, 2017

    work page 2017

  50. [53]

    A.: Image-to-image translation with conditional adversarial networks, in: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp

    Isola, P., Zhu, J.-Y., Zhou, T., and Efros, A. A.: Image-to-image translation with conditional adversarial networks, in: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp. 1125--1134, 2017

    work page 2017

  51. [54]

    Jolicoeur-Martineau, A.: The relativistic discriminator: a key element missing from standard GAN , arXiv preprint arXiv:1807.00734, 2018

    work page internal anchor Pith review Pith/arXiv arXiv 2018

  52. [55]

    and Fomel, S.: Stratigraphic coordinates: A coordinate system tailored to seismic interpretation, Geophysical Prospecting, 63, 1246--1255, 2015

    Karimi, P. and Fomel, S.: Stratigraphic coordinates: A coordinate system tailored to seismic interpretation, Geophysical Prospecting, 63, 1246--1255, 2015

    work page 2015

  53. [56]

    Kingma, D. P. and Ba, J.: Adam: A Method for Stochastic Optimization, in: International Conference on Learning Representations (ICLR), 2015

    work page 2015

  54. [57]

    and Huang, W.: Automatic extraction of seismic data horizon across faults, IEEE Transactions on Geoscience and Remote Sensing, 60, 1--11, 2022

    Li, J. and Huang, W.: Automatic extraction of seismic data horizon across faults, IEEE Transactions on Geoscience and Remote Sensing, 60, 1--11, 2022

    work page 2022

  55. [58]

    Liao, Z., Zhu, P., Zhang, H., Li, Z., Li, Z., and Ali, M.: A deep learning-based seismic horizon tracking method with uncertainty encoding and vertical constraint, IEEE Transactions on Geoscience and Remote Sensing, 62, 1--13, 2024

    work page 2024

  56. [59]

    Lim, J. H. and Ye, J. C.: Geometric GAN , in: arXiv preprint arXiv:1705.02894, 2017

    work page internal anchor Pith review Pith/arXiv arXiv 2017

  57. [60]

    Liu, N., Huo, J., Li, Z., Wu, H., Lou, Y., and Gao, J.: Seismic attributes aided horizon interpretation using an ensemble dense inception transformer network, IEEE Transactions on Geoscience and Remote Sensing, 62, 1--10, 2024

    work page 2024

  58. [61]

    A.: Flattening without picking, Geophysics, 71, P13--P20, 2006

    Lomask, J., Guitton, A., Fomel, S., Claerbout, J., and Valenciano, A. A.: Flattening without picking, Geophysics, 71, P13--P20, 2006

    work page 2006

  59. [62]

    Lou, Y., Zhang, B., Fang, H., Cao, D., Wang, K., and Huo, Z.: Simulating the procedure of manual seismic horizon picking, Geophysics, 86, O1--O12, 2021

    work page 2021

  60. [63]

    and Hale, D.: Unfaulting and unfolding 3D seismic images, Geophysics, 78, O45--O56, 2013

    Luo, S. and Hale, D.: Unfaulting and unfolding 3D seismic images, Geophysics, 78, O45--O56, 2013

    work page 2013

  61. [64]

    Shi, Y., Wu, X., and Fomel, S.: SaltSeg : Automatic 3D salt segmentation using a deep convolutional neural network, Interpretation, 7, SE113--SE122, 2019

    work page 2019

  62. [65]

    J.: Unwrapping instantaneous phase to generate a relative geologic time volume, in: 73rd Annual International Meeting, SEG, Expanded Abstracts, pp

    Stark, T. J.: Unwrapping instantaneous phase to generate a relative geologic time volume, in: 73rd Annual International Meeting, SEG, Expanded Abstracts, pp. 1707--1710, 2003

    work page 2003

  63. [66]

    J.: Relative geologic time (age) volumes---Relating every seismic sample to a geologically reasonable horizon, The Leading Edge, 23, 928--932, 2004

    Stark, T. J.: Relative geologic time (age) volumes---Relating every seismic sample to a geologically reasonable horizon, The Leading Edge, 23, 928--932, 2004

    work page 2004

  64. [67]

    Wang, F., Wu, X., and Wang, H.: Seismic horizon identification using semi-supervised learning with virtual adversarial training, IEEE Transactions on Geoscience and Remote Sensing, 60, 1--11, 2022

    work page 2022

  65. [68]

    Wu, H., Li, Z., and Liu, N.: Variable seismic waveforms representation: Weak-supervised learning based seismic horizon picking, Journal of Petroleum Science and Engineering, 214, 110\,412, 2022

    work page 2022

  66. [69]

    and Fomel, S.: Least-squares horizons with local slopes and multigrid correlations, Geophysics, 83, IM29--IM40, 2018

    Wu, X. and Fomel, S.: Least-squares horizons with local slopes and multigrid correlations, Geophysics, 83, IM29--IM40, 2018

    work page 2018

  67. [70]

    and Hale, D.: Horizon volumes with interpreted constraints, Geophysics, 80, IM21--IM33, 2015

    Wu, X. and Hale, D.: Horizon volumes with interpreted constraints, Geophysics, 80, IM21--IM33, 2015

    work page 2015

  68. [71]

    and Janson, X.: Directional structure tensors in estimating seismic structural and stratigraphic orientations, Geophysical Journal International, 210, 534--548, 2017

    Wu, X. and Janson, X.: Directional structure tensors in estimating seismic structural and stratigraphic orientations, Geophysical Journal International, 210, 534--548, 2017

    work page 2017

  69. [72]

    and Zhong, G.: Generating a relative geologic time volume by 3D graph-cut phase unwrapping method with horizon and unconformity constraints, Geophysics, 77, O21--O34, 2012

    Wu, X. and Zhong, G.: Generating a relative geologic time volume by 3D graph-cut phase unwrapping method with horizon and unconformity constraints, Geophysics, 77, O21--O34, 2012

    work page 2012

  70. [73]

    Wu, X., Luo, S., and Hale, D.: Moving faults while unfaulting 3D seismic images, Geophysics, 81, IM25--IM33, 2016

    work page 2016

  71. [74]

    Wu, X., Liang, L., Shi, Y., and Fomel, S.: FaultSeg3D : Using synthetic datasets to train an end-to-end convolutional neural network for 3D seismic fault segmentation, Geophysics, 84, IM35--IM45, 2019

    work page 2019

  72. [75]

    Wu, X., Geng, Z., Shi, Y., Pham, N., Fomel, S., and Caumon, G.: Building realistic structure models to train convolutional neural networks for seismic structural interpretation, Geophysics, 85, WA27--WA39, 2020

    work page 2020

  73. [76]

    and Ma, J.: Deep-learning inversion: A next-generation seismic velocity model building method, Geophysics, 84, R583--R599, 2019

    Yang, F. and Ma, J.: Deep-learning inversion: A next-generation seismic velocity model building method, Geophysics, 84, R583--R599, 2019

    work page 2019

  74. [77]

    Yang, J., Wu, X., Bi, Z., and Geng, Z.: A multi-task learning method for relative geologic time, horizons, and faults with prior information and transformer, IEEE Transactions on Geoscience and Remote Sensing, 61, 5907\,720, doi:10.1109/TGRS.2023.3264593, 2023

    work page doi:10.1109/tgrs.2023.3264593 2023

  75. [78]

    Yu, S., Ma, J., and Wang, W.: Deep learning for denoising, arXiv preprint arXiv:1810.11614, 2018

    work page internal anchor Pith review Pith/arXiv arXiv 2018

  76. [79]

    M., Barrow, K

    Zeng, H., Backus, M. M., Barrow, K. T., and Tyler, N.: Stratal slicing--- P art 1: Realistic 3- D seismic model, Geophysics, 63, 502--513, 1998

    work page 1998

  77. [80]

    Zeng, H., Zhu, X., Zhu, R., and Zhang, Q.: Guidelines for seismic sedimentologic study in non-marine postrift basins, Petroleum Exploration and Development, 39, 295--304, 2012

    work page 2012

  78. [81]

    Zhang, R., Isola, P., Efros, A. A., Shechtman, E., and Wang, O.: The Unreasonable Effectiveness of Deep Features as a Perceptual Metric, in: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp. 586--595, 2018

    work page 2018

  79. [82]

    and Zhao, J.-g.: Seismic horizon tracking based on the TransUnet model, Geophysics, 90, IM1--IM13, 2025

    Zhao, Z. and Zhao, J.-g.: Seismic horizon tracking based on the TransUnet model, Geophysics, 90, IM1--IM13, 2025

    work page 2025