metadata

language:
  - en
license: cc-by-4.0
pretty_name: reflection-connection

Dataset Card for Reflection Connection

Dataset Description

This data was produced by ThinkOnward for the Reflection Connection Challenge, from real seismic images from Australia and the Netherlands.

The data for this Challenge consists of four parts:

Training set (data/train) of 561 images, divided in 8 distinct categories
Validation set (data/test) of 105 images, divided in 8 distinct categories
Image corpus (data/image_corpus) of 362 unlabeled images with unknown category.
Query set (data/query) of 50 unlabeled images with unknown category

Each image is a 1-channel, 2D patch cut out from multiple sections of 3D seismic volumes.

Training and validation set both contain labeled images. The dataset has a structure similar to the well-known MNIST dataset - images are stored in subcatalogs that represent class labels. Both training and validation data can be accessed with a SeismicDataset class in the GitHub repository that offers an API very similar to a standard MNIST torch vision dataset.

Classes in the dataset are as follows:

Boring - No visible interesting features
Bright_Planar - Contains laterally continuous bright seismic amplitudes
Bright_Chaotic - Contains chaotic bright amplitudes
Channel - Visible channels (irrespective of depositional environment)
Converging_Amplitudes - Seismic amplitudes that converge together
Fault - Visible faults offsetting seismic horizons
Salt - Visible salt body which may contain the top or edge of the salt body
Transparent_Planar - Patches where the seismic reflectors are planar and the majority of the reflectors are transparent

The split between test and train was arbitrary to demonstrate capabilities of the tools available to you in this Challenge - you are free to move patches between the test and train sets. Keep in mind that class 01_Boring contains only patches of uninteresting seismic - we will not ask you to find images that belong to this class in the image corpus.

The image corpus and query contain sets of unlabeled images that can be accessed with the ImageSet class. You must not use this data in training of the model.

Note that some of images in image corpus and query set don't belong to any category defined within training and validation sets and will be used to evaluate performance of your algorithm in one-shot scenario.

Created by: Jakub Mizera, Mike McIntire, Nate Suurmeyer and Jesse Pisel at ThinkOnward
License: CC 4.0

Dataset Creation

Source Data

This data was produced by ThinkOnward for the Reflection Connection Challenge, using open source data from Australia and the Netherlands.

Who are the source data producers?

Australia Source: National Offshore Petroleum Information Management System. Available at https://www.ga.gov.au/nopims by Geoscience Australia which is © Commonwealth of Australia and is provided under a Creative Commons Attribution 4.0 International License and is subject to the disclaimer of warranties in section 5 of that license.
Netherlands Source: NAM (2020). Petrel geological model of the Groningen gas field, the Netherlands. Open access through EPOS-NL. Yoda data publication platform Utrecht University. https://doi.org/10.24416/UU01-1QH0MW

Recommendations

These data are patches of real world seismic datasets and do not capture entire seismic lines. It is recommended that this dataset be used for research purposes only.

Citation

BibTeX:

@misc {thinkonward_2024, author = { {ThinkOnward} }, title = { reflection-connection (Revision df10dc1) }, year = 2024, url = { https://huggingface.co/datasets/thinkonward/reflection-connection }, doi = { 10.57967/hf/3715 }, publisher = { Hugging Face } }

APA: Mizera, J., McIntire, M., Suurmeyer, N., & Pisel, J. (2024). reflection-connection (Revision df10dc1). doi:10.57967/hf/3715

Dataset Card Contact

Please contact [email protected] for questions, comments, or concerns about this dataset.