Model Card for ThinkOnward's Denoizer

This is a fine-tuned model based on ThinkOnward's Geophysical Foundation Model.

Model Details

Model Description

ThinkOnward's Denoizer is a fine-tuned version of the Geophysical Foundation Model, specifically designed to denoise synthetic seismic data. As part of the Image Impeccable challenge, this model was fine-tuned on a unique dataset consisting of 250 synthetically generated 3D seismic volumes, each paired with their corresponding noisy and denoised volumes. This extensive training dataset enables the Denoizer to effectively learn the patterns and characteristics of seismic data, allowing it to accurately remove noise and produce high-quality denoised outputs. By leveraging the strengths of the Geophysical Foundation Model and the specialized training data, ThinkOnward's Denoizer is well-suited for denoising synthetic seismic data and can be a valuable tool in various geophysical applications.

• Developed by: Ognjen Tanovic, Mike McIntire and Jesse Pisel of ThinkOnward
• Model type: MAE
• License: Apache 2.0

Model Sources

Link to the model repository listed below.

• Repository: https://github.com/thinkonward/geophysical-foundation-model

Uses

Direct Use

This model is for denoising synthetic seismic volumes and is a fine-tuned model based on the Geophysical Foundation Model architecture. The model was fine tuned on 250 synthetically generated seismic volumes. The goal of this project is to demonstrate that Vision Transformers (ViT) with Masked Autoencoders (MAE) can be used to leverage large amounts of paired seismic data for a specific downstream task (denoising).

Out-of-Scope Use

The model is for denoising synthetic seismic data. Use of this model on anything outside of seismic data, or similar technologies would be out-of-scope and likely have poor performance.

How to Get Started with the Model

After downloading the model architecture from the Geophysical Foundation Model Repository you can load the model using:

import torch
from huggingface_hub import snapshot_download

MODEL_REPO_ID = "thinkonward/denoizer"
snapshot_download(repo_id=MODEL_REPO_ID, repo_type="model", local_dir="./denoizer-weights")

# import the architecture from the GFM GitHub repository and make minor architecture changes 
from GFM import ElasticViTMAE
model = ElasticViTMAE.ElasticViTMAE()

Once the model architecture has been defined, you can use .from_pretrained() to extract weights!

model = model.from_pretrained("./denoizer-weights")

Check out the tutorial on GitHub for more help getting started

Training Details

Training Data

The data used to train the Denoizer was 250 synthetically generated seismic volumes. The data was generated using the Synthoseis package, which is a synthetic seismic data generator. The data was generated using the default rock properties model in the code repository. The data was genereated for the Image Impeccable Challenge, hosted by ThinkOnward.

Evaluation

Testing Data

Test data was generated using the same Synthoseis package as the training data. The test data was generated using the same rock properties model as the training data. The test data was generated for the Image Impeccable Challenge, hosted by ThinkOnward.

Metrics

Structural Similarity Index (SSIM) - The primary metric for comparison of denoising results is the scikit-image implementation of the Structural Similarity Index. The Structural Similarity Index is a metric used to measure the similarity between two images. When the SSI equals 1, the images are identical. When the SSI equals 0, the images are completely dissimilar. Please refer to the scikit-image docs for more information about the metric, as well as examples of implementation. Similarity is be calculated for all predictions. The minimum and maximum SSI values are dropped, and the mean SSI score across all predictions are be the final score.

Mean Squared Error (MSE): - The Mean Squared Error is a metric used as a loss metric for this model to measure the average of the squares of the errors between the true and predicted values. The lower the MSE, the better the model is at predicting the values. MSE is used for regression tasks.

Cross Entropy Loss: - The Cross Entropy Loss is a metric was used as a loss metric for this model to measure the average of the loss function for all predictions. The lower the Cross Entropy Loss, the better the model is at predicting the values. Cross Entropy Loss is used for downstream classification and segmentation tasks.

Results

We use 50 3D seismic volumes from the Image Impeccable Challenge, hosted by ThinkOnward as our benchmark hold-out dataset. Using a Structural Similarity Index Metric (SSIM) to compare results we document a benchmark score of 0.901743.

Citations

@misc {thinkonward_2025, author = { {ThinkOnward} }, title = { denoizer (Revision 1b0a784) }, year = 2025, url = { https://huggingface.co/thinkonward/denoizer }, doi = { 10.57967/hf/4715 }, publisher = { Hugging Face } }

Model Card Contact

Please contact [email protected] for questions, comments, or concerns about this model.