Deep Learning, Feature Learning, and Clustering Analysis for SEM Image Classification

Abstract

In this paper, we report upon our recent work aimed at improving and adapting machine learning algorithms to automatically classify nanoscience images acquired by the Scanning Electron Microscope (SEM). This is done by coupling supervised and unsupervised learning approaches. We first investigate supervised learning on a ten-category data set of images and compare the performance of the different models in terms of training accuracy. Then, we reduce the dimensionality of the features through autoencoders to perform unsupervised learning on a subset of images in a selected range of scales (from 1 μm to 2 μm). Finally, we compare different clustering methods to uncover intrinsic structures in the images.

Hydrothermal polymerization of porous aromatic polyimide networks and machine learning-assisted computational morphology evolution interpretation

“…59 Computational SEM image analysis, notably automated image recognition and categorization, has been recently reported. [60][61][62][63] Yet, the images analyzed in these prior reports are signicantly less complex than our images. In fact, most computational SEM image analyses show low accuracy when confronted with images containing elements of multiple categories.…”

“…65 ANN approaches are well suited for image analysis in general, and to our knowledge, all reports on automated SEM image analysis employ ANN approaches. [60][61][62][63] To implement ANNs for analyzing our SEM images, we decided to use pre-trained ANNs. Furthermore, as conventionally done for image clustering, we cut away the output layer of the used ANNs, since output layers are conventionally used for prediction (e.g., in a chemical context, to predict a physicochemical property).…”

Towards the FAIRification of Scanning Tunneling Microscopy Images

“…different set of images. A plethora of theoretical results [28], as well as applications to datasets of microscopy images [29,30,31], show that models trained on ImageNet [32] capture features that are relevant in an extremely heterogeneous set of image classification tasks.…”

Deep Learning for Materials Recognition from FTIR Images

“…Machine learning has greatly accelerated the development of materials and chemical science in the big-data era. For example, it has been applied to design multidimensional materials virtual space and select suitable functional materials upon suitable filtering criteria [1][2][3][4][5][6][7]. In addition, the machine learning method is applied to understand the features and descriptors to accurately describe the chemicals and materials, which is critical for the establishment of structure-property relationships [8][9][10][11].…”