| About this Abstract |
| Meeting |
MS&T22: Materials Science & Technology
|
| Symposium
|
Advanced Characterization of Materials for Nuclear, Radiation, and Extreme Environments III
|
| Presentation Title |
Recent Innovations in Machine Learning-based Techniques for In-situ Microscopy Data Analysis
|
| Author(s) |
Kevin Field, Priyam Patki, Matthew Lynch, Ryan Jacobs, T.M. Kelsy Green, Robert Renfrow, Wei-Ying Chen, Dane D. Morgan, Christopher R. Field |
| On-Site Speaker (Planned) |
Kevin Field |
| Abstract Scope |
Machine learning (ML) techniques are emerging as an attractive means for in-situ microscopy data analysis. The driving factors are inference at exceptional speeds with minimal tuning of hyperparameters during in-situ experiments. Given this, a range of challenges exist including the need for significant amounts of training data, the applicability to ML techniques to variations in imaging or materials domains (e.g., transmission versus scanning transmission electron microscopy imaging – S/TEM), and the limited computational and software infrastructure for the adoption of ML techniques on-the-microscope. Here, we will present recent studies and advances to overcome these challenges including the adoption of synthetic data generation for training workflows in cavity-based imaging. Additional discussion will be centered on imaging domain applicability for dislocation loops and evaluation of performance of in-situ microscopy data analysis compared to conventional, human-based ex-situ analysis during in-situ TEM dual ion irradiations. |