Data from the subsurface of the earth such as borehole imagery and seismic data is especially important in energy exploration. Automated interpretation of this complex data has great potential to be useful for tasks like monitoring and detecting natural resources.
The innovations in energy innovation area will achieve robust and reliable methods for automatic analysis of complex imagery from the digital subsurface for more efficient and detailed energy exploration.
The amount and quality of labelled training data is a challenge in this field. Existing interpretations are not made for machine learning purposes. Hence, the annotation quality for the task is unknown and interpretations are generally incomplete. Generating realistic simulated data is difficult as simulated data tends to be too simple. For many of these problems context and dependencies, through exploitation of prior knowledge of the geology, dependencies in space and time or results derived from existing solutions could improve predictions.
Visual Intelligence is advancing deep learning to overcome these challenges.