FORCE Hackathon 2019

Code for translating formation descriptions to transcriptions/classifications (grain size) using machine learning.

Retrieving dataset

Download the following Excel spreadsheet to a folder named data: [RealPore Por Perm Lithology data 1240 Wells Norway public.xlsx]https://drive.google.com/open?id=17kOO2637vXV1jADdo6i4PpYt7X6BYY4N

If named exactly RealPore Por Perm Lithology data 1240 Wells Norway public.xlsx, it will be picked up automatically when running the machine_translation.py script. Otherwise, specify the path to the dataset with the --dataset option.

Original link seems to be dead.

The spreadsheet is also available at https://zenodo.org/record/4419060

A version 2.0 of the data is available at https://zenodo.org/record/4723018

Python dependencies

Tested with Python 3.9.

See requirements.txt file for further dependencies

Optionally install tensorflow-gpu instead of tensorflow package if a GPU is available, to speed up training.

Training formation description to transcriptions or formation descriptions to grain size

Run the machine_translation.py script to start the machine learning

Example usages:

python machine_translation.py transcription
# or
python machine_translation.py "grain size"
# or
python machine_translation.py --help # to see other options.