Blog icon

Important information

The National AI Centre (NAIC) is transitioning from CSIRO to the Department of Industry, Science, and Resources (DISR) on 1st July 2024. You can find the latest NAIC content at industry.gov.au. For any NAIC-related questions, please email NAIC@industry.gov.au.

Data Mosaic® is an advanced software solution that allows exploration and mine geologists to rapidly and accurately turn numerical drill hole data into useful geological information.

Rapidly turn numerical drill hole data into accurate, useful geological information with Data Mosaic®

Data Mosaic ®

  • provides rapid, reliable interpretation of drill hole data
  • is designed by geologists, for geologists
  • has a simple, intuitive interface for applying sophisticated mathematical and machine learning methods to your data
  • incorporates spatial information to reduce misclassification and allows easy upscaling of results
  • produces multiscale results to generate 3D models of geological data at any scale – from large scale geology models down to detailed orebody models.

Watch our introduction video to learn more about Data Mosaic®

Web app

Our web app is easy to use and gives you access to the Data Mosaic ® algorithms and workflow.

Choose from a range of licenses or take advantage of our free trial period to try before you buy.

We offer generous discounts to academics and students.

Visit our web app to see the current price list

Technology

Rapid and consistent interpretation  

Using automated methods, Data Mosaic® provides consistent interpretation of numerical drill hole data, overcoming inconsistent subjectivity inherent in human interpretation of large and complex data sets.

Data Mosaic® can process large data sets in a matter of seconds or minutes, replacing human interpretation, which may take many weeks. It uses an automated process controlled by the geologist so that their expert geological knowledge is incorporated into the workflow, resulting in geologically meaningful outputs targeted to solving deposit specific problems.

Accurate lithological boundary detection

Data Mosaic® uses boundary detection algorithms adapted from established and reliable image analysis techniques to provide accurate detection of lithological boundaries down hole. The boundaries are used to segment the data stream into lithological units to which lithological classification is applied. This method overcomes noise and misclassification problems that result from applying machine learning to individual samples.

A wavelet transform (left) is applied to each variable (left middle) to provide multiscale analysis of data. Machine learning is applied to classify samples (right middle) and multiscale domains (right).

Multiscale spatial analysis for continuous upscaling

Geological interpretation is scale dependant. Data Mosaic® uses the continuous wavelet transform to filter input data and provide multiscale results. Machine learning techniques provide a single result at the scale of sampling. This can be very noisy and require substantial time-consuming post-processing by the geologist to upscale the data to be suitable for application, such as generating a 3D geology model.

Data Mosaic® overcomes this problem by providing results at a continuous range of scales from sample scale to drill hole scale. The geologist can inspect results and choose a scale suitable for their task and then export results at a consistent scale across the entire drill hole data base.

No black boxes

Our methods are published in international peer-reviewed journals, including several case studies.

Applications

Data Mosaic® is for anyone who collects and interprets geoscience data from mineral exploration drilling (e.g. geochemistry, mineralogy, downhole geophysics, assays).

Intellectual property

CSIRO owns all intellectual property.

The team

Data Mosaic® is developed by a team of researchers and engineers in the Discovery program of the CSIRO Mineral Resources business unit.

Meet the team

Our methods are published in international peer-reviewed journals, including several case studies.  

See our research articles