dGB Earth Sciences is thrilled to unveil the latest addition to our Machine Learning library - a state-of-the-art pre-trained model that is designed to predict faults and fractures in 2D/3D seismic data.

The pre-trained Fault Net is a deep learning model for fault prediction, based on a Convolutional Neural Network (CNN). The model was trained on real and synthetic data.

Current test results show that the Fault Net pre-trained model outputs a superior fault mask with very minimum artifacts compared to other fault prediction models (Unet Fault Predictor). While we show VGG19 pretrained model results here, this model is not meant to be applied “AS IS”, but to be tuned by the transfer training.

Fault Net is the brainchild of Yimin Dou et.all

Our team at dGB has seamlessly converted the model to the ONNX format. Additionally, we've integrated pre and post transpose operators to ensure compatibility with the input and output data order expected by the OpendTect Machine Learning plugin.

Let us know what you think. If you would like to know more or request a demo license you can contact us by sending an email to This email address is being protected from spambots. You need JavaScript enabled to view it.