Fig. 2 | npj Computational Materials

Fig. 2

From: Deep learning analysis of defect and phase evolution during electron beam-induced transformations in WS2

Fig. 2

Training a deep convolutional neural network to recognize defects that break lattice periodicity. a The first frame from STEM movie on Mo-doped WS2. b Global FFT and global FFT with high-pass filter applied. c Binary masks for image differences between the original data in a and inverse of filtered FFT in b. The image in a is a training image and the data in c serves as ‘ground truth’ (pixel-wise labeling). d Schematics of convolutional neural network with an encoder–decoder type of structure

Back to article page