cs.CVOct 1, 2026

DeepStratNet: A Context-Aware Coordinate Regression Framework for Seismic Horizon Tracking under Sparse Labels

Authors: Aniq Ahmad, Musham Ahmad Malik, Ahmad Mustafa, Heather Bedle

Organizations: Attribute Assisted Seismic Processing and Interpretation (AASPI), University of Oklahoma, USA · CIVIL Lab, Information Technology University, Lahore, Pakistan · SDAIA Joint Research Center for AI (JRCAI), King Fahd University of Petroleum and Minerals, Dhahran, Saudi Arabia

Read PDF
Ask AI
Cite

Copy citation

DOINot available for this paper

Abstract

Automatic horizon tracking is a foundational task in 3D seismic interpretation. Most existing deep learning approaches formulate it as dense semantic segmentation, typically using U-Net-based architectures. The model produces a probability map over all pixels that must be post-processed to extract precise horizon coordinates, while horizon picks in time/depth must be converted into dense masks for training. Unpicked seismic traces are consequently treated as background, which can hinder convergence, and both pre- and post-processing can introduce errors into the final interpretation. Moreover, 2D segmentation models do not inherently capture inter-slice context, while 3D models are often computationally prohibitive. We instead formulate horizon tracking as a bounded coordinate regression problem, where the model directly predicts the time/depth coordinate of the target horizon at each lateral position. We propose a lightweight regression head compatible with any pretrained vision backbone, coupled with an LSTM module to model inter-slice context and produce a continuous horizon surface across the volume. A combination of L1 and L2 losses supervises predictions at valid horizon picks, while a geology-informed regularization enforces lateral continuity between successive traces. Under controlled experimental conditions, we evaluate four pretrained vision backbones under both segmentation and regression configurations on a seismic volume from New Zealand. The proposed approach consistently outperforms its segmentation counterparts quantitatively, using metrics including RMSE and PCC, and qualitatively, while also demonstrating greater robustness to increasing sparsity of training picks. Finally, we show that prediction variation across successive traces captures local variations in geological complexity, providing an automated quality control measure for downstream seismic interpretation.

Figures & tables

Fig. 1: Architectural pipeline of the proposed seismic horizon picking framework. Given a sequence of seismic slices XiX_{i} , a pretrained backbone extracts slice-level dense embeddings FiF_{i} . An LSTM module models sequential spatial context across slices, followed by an attention-based aggregation layer that combines feature representations into hatth_{\text{att}} . A projection layer outputs the target pick prediction y^i∈RW\hat{y}_{i}\in\mathbb{R}^{W} . The model is trained using a composite loss function comprising masked l1/l2l_{1}/l_{2} reconstruction losses and a smoothness regularization term LregL_{\text{reg}} .
Fig. 2: Comparison of segmentation and regression model on an inline section. (a) shows the segmented mask and the horizon extracted using the argmax of the predicted logits. Extracting the horizon from the segmentation mask introduces additional error and uncertainty, as the extracted horizon does not consistently follow the reflection character. (b) shows the horizon directly predicted by the regression model, which accurately follows the reflection character of the target horizon.
Fig. 3: Time surface of the horizon generated from the model. The vertical axis shows two-way travel time (TWT) in milliseconds. Travel times increase toward the northwest, indicating that the horizon dips to the NW
Fig. 4: Another Inline showing the comparison of segmentation and regression model.(a) and (b) shows the predicted shown from segmentation and regression model respectively. (b) also shows the irregular horizon picks predicted by the model. This character is noticed on couple of Inlines and it marks the presence of the channel.
Fig. 5: Different backbones (a–d) showing the degradation in regression and segmentation model performance as the proportion of training labels decreases. Segmentation models show the fastest degradation across all backbones, with ResNet-50 and ResNet-101 (b–c) showing the largest performance drop, followed by ViT-Large (a) and ConvNeXt-Large (d). Regression models show a more consistent trend with a smaller drop as the number of training labels decreases across all backbones.
Fig. 6: Model uncertainty shown by the inter-point variance. A thin uncertainty band indicates low variance, whereas a thick uncertainty band indicates high variance. (a) and (b) show two examples of the predicted horizon, with each example including an individual view and a zoomed-in view. The high inter point variance regions coincide with chaotic amplitude patterns that indicate the presence of channel features which is confirmed by the seismic attribute (Variance) along the predicted horizon.
 Training SetTest Set
ModelRMSEPCCR2R^{2}RMSEPCCR2R^{2}
DeepLabV3-ResNet50 (segmentation)10.78 ms0.99240.973313.36 ms0.99150.9753
DeepLabV3-ResNet50 (regression - no LSTM)16.75 ms0.98260.932914.52 ms0.99680.9848
DeepLabV3-ResNet50 (regression)10.25 ms0.99690.98698.77 ms0.99850.9943
DeepLabV3-ResNet101 (segmentation)9.24 ms0.99400.978912.16 ms0.99140.9820
DeepLabV3-ResNet101 (regression - no LSTM)19.30 ms0.98730.935318.35 ms0.99670.9761
TABLE I: Benchmark for various vision backbones under the Segmentation and Regression Configurations
Figure 8
Figure 9
Figure 10
Figure 11
More in this topic

Data-Efficient Learning

Data-efficient learning develops methods that achieve accurate, robust models with fewer labeled examples, training samples, or interactions.

See all 245 papers

Explore similar work

CardsList
Jun 15, 2026

Contrastive Learning for Seismic Horizon Tracking with Domain-Specific Priors

Unsupervised 3D seismic horizon tracking faces a key limitation: signal-based propagators provide accurate trace-level alignment but often fail near faults, whereas texture-driven deep models are more robust to discontinuities, typically at the cost of labeled data requirements and reduced trace-level precision. We propose a self-supervised fusion of both paradigms in which signal-derived local horizon correspondences act as domain-specific priors to train a texture-based deep learning model. Specifically, we estimate reliable trace-to-trace flows from reflector slopes and use them to form positive pairs in a contrastive objective, while restricting training to high-confidence neighborhoods, optionally augmented with a fault mask. The objective is not to infer ambiguous correspondences close to discontinuities, but to preserve horizon identity across them. As a result, the network learns voxel-wise embeddings that preserve local signal continuity while enabling horizon propagation beyond discontinuities through similarity search. Experiments on the public F3 dataset and a faulted synthetic dataset achieve lower mean absolute error (MAE) than unsupervised baselines and competitive performance against a semi-supervised method using a single labeled slice.

Alexandre Thouvenot, Lionel Boillot, Vincent Gripon

IMT Atlantique, LAB-STICC, UMR CNRS 6285, F-29238 Brest, France · TotalEnergies, OneTech, 64000 Pau, France

Contrastive Learning3D Representation LearningSeismology
AskPDF
Jun 5, 2026

Beyond Backscatter: InSAR coherence from detected SAR images

In this work, we propose a deep learning framework for coherence regression directly from detected SAR images, without the need for accurate coregistration. A Residual U-Net is trained using coherence maps derived from precisely coregistered Sentinel-1 SLC data to learn the relationship between backscatter magnitudes and coherence. The model is trained on 12-day SLC pairs and evaluated across different datasets, including coregistered SLC products and open access analysis-ready data, covering diverse radiometric properties, geometries, and locations. Experimental results demonstrate that the proposed method achieves high-resolution coherence regression with improved accuracy compared to existing intensity-based approaches. The network generalizes well across diverse geographical locations and even across different temporal baselines that were never seen at training time. Additionally, the ability to operate on globally available analysis-ready data, such as ground range detected data, e.g., distributed through Google Earth Engine, enables its large-scale application in mission design, change monitoring, and diverse mapping tasks.

Francescopaolo Sica, Andrea Pulella, Michael Schmitt

Department of Aerospace Engineering, University of the Bundeswehr Munich, Neubiberg, Germany · Microwaves and Radar Institute, German Aerospace Center (DLR), Weßling, Germany

Remote SensingSynthetic Aperture Radar
AskPDF
Jun 14, 2026

Domain-Guided Prompting of the Segment Anything Model for Seismic Interpretation: The Role of Attributes, Visualization, and Hybrid Prompts

The advent of large pretrained foundation models for computer vision has significantly improved the efficiency of visual data interpretation. The Segment Anything Model (SAM), in particular, offers powerful zero shot segmentation capabilities through prompt based interaction, thus making it a promising tool for seismic interpretation. However, most existing applications of SAM rely on fine tuning for specific geological targets, which requires extensive labeled data, incurs high computational cost, and often compromises the model's generalization capability. In this study, we introduce a principled framework for zero shot adaptation of foundation models to seismic data. The framework is built on two key components: (1) aligning seismic attributes and visualization choices (e.g., colormaps) with the geological target of interest, and (2) employing a hybrid prompting strategy that combines sparse user defined point prompts with dense mask prompts derived from SAM's internal feature activations. We systematically evaluate this framework across multiple geological targets, datasets, prompt configurations, and seismic attribute representations. Our results demonstrate that geologic target aware selection of seismic attributes and colormaps, combined with hybrid prompting, enhances the separability of geological features and improves boundary delineation and segmentation accuracy relative to point based prompting alone. Our findings show that, when these components are jointly applied, SAM can achieve competitive segmentation performance in a fully zero shot setting, thereby eliminating the need to retrain SAM for each geologic feature. This work establishes a practical and scalable pathway to leverage foundation models in seismic interpretation, reducing reliance on labeled data while preserving model generality.

Aniq Ahmad, Heather Bedle, Ahmad Mustafa

School of Geosciences, University of Oklahoma, Norman, Oklahoma, USA · King Fahd University of Petroleum and Minerals, Dhahran, Saudi Arabia

Promptable Image SegmentationVision Foundation Model AdaptationZero-Shot Referring Image SegmentationSemantic SegmentationPrompt Engineering
AskPDF