About this Abstract |
| Meeting |
2026 TMS Annual Meeting & Exhibition
|
| Symposium
|
Artificial Intelligence Applications in Integrated Computational Materials Engineering (AI-ICME)
|
| Presentation Title |
Breaking Boundaries: Texture-Aware AI for Metallography |
| Author(s) |
Ofer Beeri, Boaz Meivar, Inbal Cohen, Julien Robitaille, Francis Quintal Lauzon, Shai Avidan, Gal Oren |
| On-Site Speaker (Planned) |
Gal Oren |
| Abstract Scope |
Accurate grain boundary detection is vital in metallography but remains challenging due to its reliance on texture, not semantics. Can deep learning models built for semantic segmentation handle such texture-driven tasks? Our work explores this through a series of studies. Initial efforts applying semantic segmentation showed promise [Rusanovsky et al., 2022], but limitations soon emerged. A second study highlighted these issues and introduced the Texture Boundary Metallography (TBM) dataset, a benchmark tailored to the unique challenges of texture-based segmentation [Rusanovsky et al., 2023]. In our latest work [Cohen et al., 2024], combining partial annotations with contextual input greatly improved accuracy. Integrating AutoSAM, a fine-tuned, no-prompt adaptation of the Segment Anything Model, further enhanced performance. Building on this, we are developing SAMaterial—a model designed to generalize across metallographic segmentation tasks using large, diverse datasets, promising broader applicability and improved material analysis. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Characterization, ICME, Machine Learning |