This paper introduces Construction Zone, a Python package for generating complex nanoscale atomic structures, enabling the creation of large, diverse synthetic datasets for training machine learning models to analyze HRTEM images. The authors develop an end-to-end workflow, achieving state-of-the-art nanoparticle image segmentation performance on experimental benchmarks using only simulated data. The study investigates how simulation fidelity, atomic structure distribution, and imaging conditions affect model performance.
Publisher
npj Computational Materials
Published On
Jul 29, 2024
Authors
Luis Rangel DaCosta, Katherine Sytwu, C. K. Groschner, M. C. Scott
Tags
Construction Zone
nanoscale atomic structures
machine learning
HRTEM images
image segmentation
simulation fidelity
synthetic datasets
Related Publications
Explore these studies to deepen your understanding of the subject.
