Skip to content
Close
Contact
Book Demo
Contact
Book Demo

EarthNET AI faults

2D and 3D models

EarthNET fault-interpretation models provide unparalleled user control and accuracy. Leverage pre-trained single-click models and user-interactive workflows for local optimisation.

Fault_Probability_and_Plane_02

Single-click results using 3D pretrained model with fault-probability volumes and planes

Our fault interpretation journey

Our journey began in 2020, when we started by training 2D models based on user interpretations.

By 2021, we introduced Pegasii 3, a 2D model pretrained on over 30 surveys, and launched our Fault 3D pretrained model, developed using synthetic data.

In 2022, we integrated transformers and released Pegasii 5, a 2D model pretrained on over 50 surveys.

2023 saw a significant enhancement of our 3D model, overcoming edge effect challenges linked to GPU memory limitations.

By 2024, we achieved a major milestone: optimizing ESA's pretrained 3D model for your surveys using 2D labeling, setting a new standard of excellence.

This advancement is pivotal to all our pretrained models, including those for injectites and salt, marking a key milestone in our journey.

 

  • Single-click models
  • Interactive workflows
  • Fault-plane extraction

Single-click models

Apply multiple pre-trained 2D and 3D models with different neural-network architectures. Multiple outputs enables you to explore scenarios and interpretation uncertainty. Stack your fault-probability volumes using EarthNET’s calculator to get the best from your multiple initial outputs.

 

single click

 

Interactive workflows

Optimise models for your target survey using interactive workflows. Interact with the data and models using active learning and model-assisted labelling. Extract fault-sticks from models, edit them using labelling tools and re-train models on your user-optimised training data.

 

interactive

 

Fault-plane extraction

Extract fault sticks, grouped sticks, fault planes and/or grouped pointsets for use in your geomodelling workflows. EarthNET features automatic fault-stick, fault-points and fault-plane extraction, as well as manual tools to optimise fault grouping and plane extraction

 

plane

 

Get in touch
Designed by geoscientists for geoscientists

Fault interpretation deliverables

  • Data conditioning and denoising
  • Fault-probability volumes from pre-trained 2D and 3D models
  • Fine-tuning of 2D and 3D models on target dataset
  • Locally optimised fault-probability volume
  • Fault-stick extraction
  • Fault-stick grouping and fault plane extraction
  • Faults burned into seismic volume
  • Frequency-decomposition RGB blend with faults embedded
  • Outputs can be delivered for full-stack, angle stacks and azimuth stacks
HOW IT WORKS

Streamlined interpretation process

Whether you're seeking faults as a service or prefer to handle the work yourself with a software license, EarthNET AI Faults delivers a unique, time- and cost-effective solution. Our streamlined process and comprehensive set of deliverables ensure you achieve next-generation results with maximum efficiency.

1
Data ingestion

Flexible ingestion of seismic data using formats such as .segy, .zgy and .vds
2
Cleaning

Condition, denoise and enhance your seismic data
3
One Click interpretation

Apply multiple pre-trained 2D and 3D models with different neural-network architectures
4
Iterative model training

Optimise models to your target survey using interactive workflows
5
Active learning

Use model-assisted labelling to increase speed and consistency in data annotation. Interact with the model so that it learns from you and you learn from it.
6
Deliverables and Fault extraction

Extract fault sticks, grouped sticks, fault planes and/or grouped pointsets for use in your geomodelling workflows
Book demo
RECOMMENDED READING

Explore our resources

Get in touch

We are dedicated to helping our clients unleash the full potential of their data.
Let's talk about the opportunities, and explore how we can assist you. 

Get in touch
Book demo