MinDet1: A deep learning-enabled approach for plagioclase textural studies

Main Article Content

Norbert Toth
John Maclennan


Quantitative textural attributes, such as crystal size distributions or aspect ratios, provide important constraints on the thermal history of rocks. They facilitate the investigation of crystal nucleation, growth, and mixing as well as cooling rate. However, they require large volumes of crystal segmentations and measurements often obtained with manual methods. Here, a deep learning-based technique—instance segmentation—is proposed to automate the pixel-by-pixel detection of plagioclase crystals in thin-section images. Using predictions from a re-trained model, the physical properties of the detected crystals (size and aspect ratio) are rapidly generated to provide textural insights. These are validated against published results from manual approaches to demonstrate the accuracy of the method. The power and efficiency of this approach is showcased by analysing an entire sample suite, segmenting over 48,000 crystals in a matter of days. The approach is available as MinDet1 software for users with moderate expertise in Python. Widespread use of MinDet may facilitate significant developments in igneous petrography and related fields.


Download data is not yet available.

Article Details

How to Cite
Toth, N. and Maclennan, J. (2024) “MinDet1: A deep learning-enabled approach for plagioclase textural studies”, Volcanica, 7(1), pp. 135–151. doi: 10.30909/vol.07.01.135151.
Citation information
This article has citations in Scopus
Link to article page on Google Scholar
Connected papers
Received 2023-05-16
Accepted 2024-02-08
Published 2024-03-06
Alternative metrics via Dimensions and altmetric


Smart Citations via scite_