Safe, efficient management of underground reservoirs is critical to the sustainability of Earth's natural resources. This task arises in many areas of applied geoscience: from efficient production of petroleum reservoirs during the transition to renewable energy, to safe geologic sequestration of carbon dioxide in efforts to mitigate climate change, to sustainable production of groundwater for human and agricultural consumption. Geologic interpretation of survey data—especially raw seismic images of Earth's interior—remains a key, time-consuming step in this process. Like the interpretation of medical images, the process is still done largely by human specialists—that is, by expert geologists assisted by specialists in geophysical data collection, processing, imaging, and display. The task can take from months to years. 

The goal of this global hackathon is to explore the application of modern tools of machine learning and artificial intelligence to speed up and improve both qualitative and quantitative interpretation of geophysical images of Earth's interior using GPUs. Geophysical datasets pose special problems to machine learning because of their huge size (terabytes to petabytes), complexity (geologic structures have a near-infinite range of variations), and multi-parametric physical description (dozens of independent variables are needed to specify the properties of fluid-filled rocks). 

The SEAM AI Applied Geoscience Hackathon will be hosted across multiple time zones is a global event that aims to gather multidisciplinary teams of geologists, geophysicists, petrophysicists, reservoir engineers, and data scientists to brain-storm ideas in this space using examples of real and synthetic geophysical datasets.

Hackathon Flow

  1. Call for applications opens on January 18, 2021. Teams can apply to participate by completing and submitting the questionnaire in the official application. GPU knowledge or use is not required at this stage.
  2. Call closes on March 1, 2021. The Review Committee starts evaluating applications.
  3. Acceptance letters are sent to selected teams on March 15, 2021. Accepted teams are identified by a review committee and notifications are sent.
  4. Accepted teams participate in a Bootcamp on March 22, 2021. Teams will be trained on GPU tools, technologies, and compute resources in order to be prepared for the main part of the hackathon.  
  5. Teams work with mentors March 23 through April 2, 2021. Communication will be facilitated through Slack and Zoom. Teams will be presenting to their mentors through Zoom during SCRUM or presentations sessions that will be scheduled for every other day except weekends. Detailed instructions will be provided to the accepted teams.
  6. Final Teams Presentations – April 5-9, 2021. Team will be presenting their final solutions the week of April 5th.  
  7. Winner announcement – April 12, 2021. Winners identified by a Jury will be announced on April 12th.  

Problem Statement

During the event, participants will be working on problems of classification and interpretation of 3D seismic (sound-wave) images of geologic structures, using AI and machine-learning algorithms running on GPUs. For example, a typical problem consists of labeling each pixel (voxel) in a 3D seismic image according to different geologic facies (rock types), as identified by an expert human interpreter. The hackathon will begin with a specific example based on the public-domain “Parihaka” seismic dataset from New Zealand. Several other examples of seismic images and their classification, along with examples of seismic data processing, will be studied during the event.

One of the goals of the hackathon will be to identify new metrics for scoring the geologic fidelity and similarity of different classifications of seismic images (see, for example, the article, “Ground-truth uncertainty-aware metrics for machine learning applications on seismic image interpretation: Application to faults and horizon extraction,” by Sébastien Guillon, Frédéric Joncour, Pierre-Emmanuel Barrallon, and Laurent Castanié, in the October 2020 issue of The Leading Edge,

Prospective participants can learn more about seismic facies classification and download the Parihaka data set by visiting the Open Data page of the Wiki maintained by the Society of Exploration Geophysicists (SEG): Seismic facies identification was the subject of a workshop at the 2020 SEG Annual Meeting and is topic #1 on the Open Data page: Machine Learning Blind-test Challenge.

How to Apply

Fill out and submit the official application form by March 1, 2021. The application calls for teams to provide a proposed solution to the problem statement. Please ensure your solution is presented in PowerPoint template provided.


Each team will have access to experienced mentors for research, industry and government institutions. Mentors will be helping with AI and GPU technologies as well as with basic aspects for geoscience to make sure that teams succeed at the event.

Compute Infrastructure

All accepted teams will be using a Hackathon cluster with Tesla V100 GPUs for the duration of the event.

Winning Criteria

Different metrics for scoring seismic image interpretations will be developed and used at the event. One simple example is a weighted average of the number of pixels in the image that have been classified correctly, when compared with an expert human interpretation. 


The top 3 teams on the leaderboard at the end of the event will be able to choose from a selection of awards, including DJI Mavic 2 drones, travel vouchers for SEG meetings and workshops, and coupons redeemable at the SEG bookstore. Two additional prizes will be awarded to recognize outstanding presentations at the event.  

Terms of Award

  • Participants claiming awards will be asked to publish their code(s). SEAM AI project and hosting partners will not request any special IP rights to code generated as part of the competition. 

  • Awards will be given to selected teams irrespective of the number of members in the team (team members may choose to share awards among themselves).

  • The decision of the Jury on the awards will be final and binding.

  • An award will be given to a team as submitted in the application for the Hackathon and cannot be transferred later.

About SEAM

SEAM is a research arm of SEG. It organizes collaborations among industry, government, and academia to address major industry subsurface challenges. By doing so, we provide a forum for industry leaders to resolve geophysical problems of common interest and advance subsurface management research and development through the art and science of numerical modeling and computation.

Event Focus

Important Event Dates

  • Application Deadline
  • Acceptance Letters Sent
  • Bootcamp
  • Teams work with Mentors
  • Final Team Presentations
  • Winners Announced