| About this Abstract |
| Meeting |
2021 TMS Annual Meeting & Exhibition
|
| Symposium
|
Practical Tools for Integration and Analysis in Materials Engineering
|
| Presentation Title |
AMPIS: Automated Materials Particle Instance Segmentation |
| Author(s) |
Ryan Cohn, Timothy Prost, Iver Anderson, Emma White, Jordan Tiarks, Elizabeth Holm |
| On-Site Speaker (Planned) |
Ryan Cohn |
| Abstract Scope |
Instance segmentation has been shown to be a powerful tool for image analysis but has not been adopted by the materials science community. Thus, we present AMPIS, an open-source framework for applying instance segmentation to materials image data. We provide a case study applying this tool to segment individual powder particles and satellites in images of additive manufacturing (AM) feedstock powders. Detecting, quantifying, and minimizing the presence of satellites is critical to opening up low-cost feedstock options to the AM community. Despite labeling a very small number of images to train the model, segmentation is successfully performed on a wide variety of images, including images captured with different magnifications and imaging modes. The results demonstrate the first ever direct measurements of satellite contents in powders. To demonstrate the flexibility of the technique we provide a second case study characterizing the spheroidite content in microscope images of steel. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Machine Learning, Additive Manufacturing, Characterization |