The effectiveness of data augmentation in porous substrate, nanowire, fiber and tip images at the level of deep learning intelligence

TL;DR

This paper evaluates the impact of data augmentation on deep learning classification of SEM images of nanowires, fibers, tips, and porosity levels, achieving high accuracy and aiding nanoscience research.

Contribution

It demonstrates that data augmentation significantly improves CNN classification accuracy for nanostructure images and develops software for automated analysis in nanoscience.

Findings

Highest validation accuracy of 97.1% for nanowire, fiber, tip classification

Achieved 93% accuracy in porosity level classification

Software enables automated counting and design of nanostructures

Abstract

To prepare for identifying the composition of nanowire-fiber mixtures in Scanning Electron Microscope (SEM) images, we optimize the performance of image classification between nanowires, fibers and tips due to their geometric similarities. The SEM images are analyzed by deep learning techniques where the validation accuracies of 11 convolutional neural network (CNN) models are compared. By increasing the diversity of data such as reflection, translation and scale factor approaches, the highest validation accuracy of recognizing nanowires, fibers and tips is 97.1%. We proceed to classify the level of porosity in anodized aluminum oxide for the self-assisted nanowire growth where the validation accuracy is optimized at 93%. Our software allow scientists to count the percentage of fibers in any nanowire-fiber composite and design the porous substrate for embedding different sizes of nanowires automatically, which assists the software development in Nanoscience Foundries & Fine Analysis (NFFA) Europe Projects.