Geodata from the ground up

Share your data cleaning notebooks, howtos and demos on GeoHacker.



Poseidon: Extracting and cleaning core images

The Poseidon dataset includes a large number of high-resolution core photographs, primarily stored in PDF format. As is, this data is not ready for application to machine learning. Geohacker makes the core-photos accessible to Python. We then demonstrate a deep-learning workflow for feature extraction using semantic segmentation. The example identifies the location of core-plugs in the data. This then enables the location of thin-section images in the data, for further analysis.


Poseidon: Seismic data, wells and model building

Datasets include well-logs, multi-stack seismic, VSP, processing reports,and pdf.

Seismic, VSP, and well-log data are better together! GeoHacker Jupyter notebooks interrogate the various meta-data to demonstrate how to co-locate Poseidon Pharos-1 well-data with seismic lines. Some basic deterministic sparse spike and AVO inversions are demonstrated. The data is then prepared for unsupervised learning using features extracted by a deep learning CNN.


UK-OGA: Open Well Dataset

The Oil and Gas Authority of the UK has made a collection of well-log data and interpreted tops available for download. GeoHacker provides a Jupyter notebook for profiling and visualizing this data. The GeoHacker well-analysis tool automatically profiles this dataset and determines if there is a minimally viable selection of curves, and which data has been cleaned for application to machine learning. Our Geohacker exploration scientists also provide a brief tutorial on standard well log analysis for those data scientist that are new to geology and geophysics.