|About this Abstract
|7th World Congress on Integrated Computational Materials Engineering (ICME 2023)
|Fluoroelastomer Crystallization Kinetics
Studied by Deep Learning Segmentation of Atomic Force Microscopy Images
|Sameera Nalin Venkat, Thomas Ciardi, Jube Augustino, Jayvic Cristian Jimenez, Peter Schlueter, Mingjian Lu, Frank Ernst, Yinghui Wu, Roger H. French, Laura S. Bruckman
|On-Site Speaker (Planned)
|Sameera Nalin Venkat
Atomic force microscopy (AFM) provides valuable insights into the crystallization of fluoroelastomers, which impacts performance in applications. Time-lapsed acquisition of AFM image sequences provides quantitative information on crystallization kinetics. However, corresponding datasets are large. Open-source software and tools are routinely used for data processing, but it can be time-consuming and challenging to process thousands of images. In this work, we integrate automated feature detection and segmentation algorithms, leveraging evaluation by convolutional neural networks. End-to-end frameworks, such as UNet image segmentation, allows for batch processing such as generating binarized masks which can be used to obtain image properties. This can help in quantifying the projected area fraction of the crystalline phase in each image. It can also track individual crystallites as a function of time when combined with an open-source software for AFM-image processing, which serves as the “ground truth” for comparison.