New plugin for both operational geo-scientists, experimental geo-scientists and research geo-scientists. Machine Learning links the OpendTect Pro environment to the research world of Python, TensorFlow, Keras, PyTorch & Scikit Learn. The plugin is the successor of our popular Neural Networks plugin, which has been fully integrated into the new plugin.
Machine Learning offers workflows for: seismic, wells, and seismic-to-wells applications.
Train on real data extracted from multiple surveys, or on synthetic data generated by SynthRock, OpendTect's stochastic simulator.
Includes trained models for off-the-shelf applications such as fault prediction by a U-Net.
"Having experience of more than 35 years in the field of Oil & gas exploration, we never came across such innovative solutions."
Dr. N.P. Singh Dhillon Chief General Manager ONGC
All commercial plugins require OpendTect Pro.
Among others Machine Learning offers the following features:
- Create input training data from multiple surveys.
- Convolutional Neural Networks, Random Forest Algorithms, Support Vector Machines, ADaBoost, ...
- Trained models for direct application.
- Option to import your own trained models.
- Training options: new, resume and transfer training.
- Image-to-Image workflows (seismic faults, facies, horizons, ...).
- Image-to-Point workflows (seismic facies, chimneys, salt, ...).
- Log-log prediction workflows.
- All supervised and unsupervised workflows supported in the original Neural Networks plugin.
In combination with SynthRock:
- Create synthetic data sets for seismic rock-property prediction workflows.