Artificial intelligence (Al) and machine learning (ML) have made their way into materials science and have led to a real hype within a very short period. However, this leads to the fact that many experts without a background in materials science also use ML as a universal remedy for dealing with issues in our field - however, without precisely grasping and considering the complex material-specific questions. On closer look, though, it quickly becomes clear that only a small fraction of a ML-based segmentation or classification workflow is composed of the ML code itself. In particular, the application of machine learning to the classification and segmentation of complex microstructures poses special challenges - especially when it comes to assigning the ground truth. Therefore, despite all the progress made in artificial intelligence, our materials science domain knowledge is still indispensable in this field of research and in applications.

The goal of the talk is to focus on a holistic approach with an emphasis on the underlying data and interpretation of the same, in addition to a general introduction to the applications of ML in microstructure classification.