Deep Learning Analysis of Defect and Phase Evolution During Electron Beam Induced Transformations in WS2
Supplementary code for the paper: https://arxiv.org/abs/1803.05381
In this project, we used a machine learning framework to identify, extract, cluster, analyze, and track defects in a scanning transmission electron microscopy (STEM) movie of 2D material (WS2).
This repository contains 3 notebooks.
-
WS2_Defects_DeepLearning_gmm.ipynb is the main workflow, starting, involving:
1.1) Generating traning set from the first frame of the image by utilizing Fast Fourier Transform subtraction (FFT-subtraction) to create labels.
1.2) Training a fully convolutional neural network (FCNN) for defect identification.
1.3) Extraction of defects from all frames using the trained FCNN.
1.4) Clustering the extracted defects using a Gausian Mixture Model (GMM).
1.5) Tracking the evolution of selected defects and calculating diffusion
-
WS2_Defects_LocalCryst_PCA.ipynb uses clustered defects to perfrom local crystallographic analysis using principal component analysis (PCA)
-
WS2_Defects_Markov_Transitions.ipynb uses extracted defect trajectories to calculate transition probabilities.