2020 SEG Annual Meeting Machine Learning Interpretation Workshop


Seismic data is publicly available and provided by New Zealand Petroleum and Minerals (NZPM). See https://www.nzpam.govt.nz/maps-geoscience/exploration-database/ for details.

The associated labels, Labeled geological model of Parihaka seismic data for machine learning by Chevron U.S.A. Inc., is licensed under CC BY-SA 4.0. See https://creativecommons.org/licenses/by-sa/4.0/ for details.


Thank you for participating in the 2020 SEG Annual Meeting Machine Learning Interpretation Workshop! On this page, you will find a 3D seismic data volume and an accompanying "label" volume consisting of a seismic facies interpretation. Each pixel in the label volume is assigned a value from 1 to 6, denoting its facies classification:
  1. Basement/Other: Basement - Low S/N; Few internal Reflections; May contain volcanics in places
  2. Slope Mudstone A: Slope to Basin Floor Mudstones; High Amplitude Upper and Lower Boundaries; Low Amplitude Continuous/Semi-Continuous Internal Reflectors
  3. Mass Transport Deposit: Mix of Chaotic Facies and Low Amplitude Parallel Reflections
  4. Slope Mudstone B: Slope to Basin Floor Mudstones and Sandstones; High Amplitude Parallel Reflectors; Low Continuity Scour Surfaces
  5. Slope Valley: High Amplitude Incised Channels/Valleys; Relatively low relief
  6. Submarine Canyon System: Erosional Base is U shaped with high local relief.  Internal fill is low amplitude mix of parallel inclined surfaces and chaotic disrupted reflectors.  Mostly deformed slope mudstone filled with isolated sinuous sand-filled channels near the basal surface.
Both the seismic data and label volume are provided in SEG-Y format. If you are unfamiliar with SEG-Y, you can use freely-available Python utilities such as segyio (https://github.com/equinor/segyio) to convert the underlying data to numpy or other formats (you will find the segyio.tools.cube command especially useful). 

Your task is to use this training dataset to build a machine-learning model that can predict the facies classifications for each pixel in a different, unlabeled 3D seismic volume with the same six facies categories. Around September 1, one or more such volumes will be posted on this page (if you provided your email address before accessing this page, you will be notified when these test volumes are available). At that time, you may apply your model to the new dataset(s) and submit your "answers" to the workshop organizers. Participants will be judged primarily using the average of the Intersection over Union (IoU) scores for the six facies categories, although the Organizers reserve the right to use additional qualitative and quantitative metrics to gauge performance.

Participant and performance information will remain anonymous, although any participant may request to present their methodology during the workshop at the Annual Meeting (we may be unable to accommodate all requests). There are no particular restrictions on the methods or strategies you may employ to build your model(s), but keep in mind that one goal of the workshop is to identify broadly applicable and generalizable strategies for ML-assisted interpretation.

Thank you for your participation, and good luck! Please contact the organizers with any questions:

Download the data


Test Datasets