INTERACTIVE DEEP LEARNING
GEOSCIENTISTS ARE NOW IN CONTROL WITH A FULLY INTERACTIVE EXPERIENCE AND INSTANT FEEDBACK.
Train, infer, and interpret your subsurface data in real-time without compromising accuracy or confidence in your interpretation.
Bluware INTERACTIVE DEEP LEARNING significantly improves the effectiveness of geoscientists during their interpretation workflow. Our deep learning approach allows geoscientists to train the system and interactively correct the results. This interactive experience provides geoscientists full control over the accuracy of the results. Deep learning can be a fully trusted tool, directly applicable to asset teams instead of being relegated to research.
Bluware DEEP LEARNING is interactive and driven by geoscientists. This case study demonstrated what can be achieved in two hours compared to three months using traditional techniques.
IN DEEP LEARNING WORKFLOWS
E&P companies keep spending significant resources to develop machine learning capabilities with limited success. Geoscientists are already battling uncertainty. Deep learning results with 60 to 80% accuracy are simply unacceptable as the resulting compounded accuracy decreases from 36 to 64%. As a result, interpreters are forced to spend a significant amount of time to quality control (QC) and fix the results. This low confidence in the system means deep learning is not being adopted and geoscientists resort to business-as-usual.
INTERACTIVE DEEP LEARNING TECHNOLOGY
To address these shortcomings, we integrate deep learning directly into subsurface interpretation, with the geoscientist directly and personally driving the training of the network while interpreting in real-time. The results are computed immediately, validated, and corrected interactively by the geoscientist. This workflow delivers highly accurate results that are endorsed and supported by the experts who approve the final interpretation.
80% OF DATA SCIENTISTS’ TIME IS SPENT ON DATA PREPARATION.
Bluware INTERACTIVE DEEP LEARNING is a service which creates complex geo-bodies from seismic data. Your data stays within your network. Bluware ADAPTIVE STREAMING sends data to the deep learning process. Data is never copied. The interpreter is in full control of training the model and reinforces the learning interactively.
Current implementations of deep learning used on sub-surface data begins with a lengthy data preparation step. Seismic data is broken into images of slices and then into tiles before randomizing the images and loading into deep learning tools. This data preparation step is time consuming, static in nature, and needs to be repeated for each dimension of data (inline vs. crossline).
Interpreters can now spend time interpreting instead of data preparation. Upon completion, the resulting geo-bodies are transferred to interpretation applications.
Tim Roden, Shell GeoSigns Software Manager
“Our machine-learning can process seismic data to find geologic faults faster. Depending on the geology, some fault lines can help oil and gas migrate to the surface while others can disrupt drilling operations. Machine learning allows him to do that work in two hours when it used to take geologists two months through pore through the data. That’s transformational.”
SCHEDULE A DEMO
Interested in learning more about Bluware INTERACTIVE DEEP LEARNING?
Schedule a meeting with us.