Addressing materials’ microstructure diversity using transfer learning - Archive ouverte HAL Accéder directement au contenu
Article Dans Une Revue npj Computational Materials Année : 2022

Addressing materials’ microstructure diversity using transfer learning

Résumé

Abstract Materials’ microstructures are signatures of their alloying composition and processing history. Automated, quantitative analyses of microstructural constituents were lately accomplished through deep learning approaches. However, their shortcomings are poor data efficiency and domain generalizability across data sets, inherently conflicting the expenses associated with annotating data through experts, and extensive materials diversity. To tackle both, we propose to apply a sub-class of transfer learning methods called unsupervised domain adaptation (UDA). UDA addresses the task of finding domain-invariant features when supplied with annotated source data and unannotated target data, such that performance on the latter is optimized. Exemplarily, this study is conducted on a lath-shaped bainite segmentation task in complex phase steel micrographs. Domains to bridge are selected to be different metallographic specimen preparations and distinct imaging modalities. We show that a state-of-the-art UDA approach substantially fosters the transfer between the investigated domains, underlining this technique’s potential to cope with materials variance.

Dates et versions

hal-03642883 , version 1 (15-04-2022)

Identifiants

Citer

Aurèle Goetz, Ali Riza Durmaz, Martin Müller, Akhil Thomas, Dominik Britz, et al.. Addressing materials’ microstructure diversity using transfer learning. npj Computational Materials, 2022, 8 (1), pp.27. ⟨10.1038/s41524-022-00703-z⟩. ⟨hal-03642883⟩
159 Consultations
0 Téléchargements

Altmetric

Partager

Gmail Facebook X LinkedIn More