De Beers Group

Machine Learning Classifier

Using hydrophone audio data to classify soil composition.

Main Takeaway

De Beers Group made proprietory hydrophone audio data available to a group of hackathon contenders which the Codeswop team formed part of. Our challenge was to assist in identifying soil composition based purely on drilling audio date which is available in real-time.

Features

  • Audio data vectorisation
  • Real-time reporting interface
  • Machine learning classifier

Technology and Integrations

SciKit Learn, Pandas, NumPy, Matplotlib, Seaborn, TensorFlow, Keras, MongoDB, Docker, Gitlab

Team Members

Dean Harber - Lead Architect

Tiaan Hendricks - Data Engineer

Regardt Nel - Audio Engineer

Service

Software Architecture and Development

Machine Learning Environment Development

Time to Launch

1 Week

Outcomes

Web application built with Meteor and React, MongoDB for Data, Scikit Learn for Machine Learning, Docker for containerization, and Gitlab for CI/CD.

Outsource Software Development with Confidence