AI visualization in Nanoscale Microscopy

Rajagopal A,Nirmala V,Andrew J,Arun Muthuraj Vedamanickam.

DOI: https://doi.org/10.48550/arXiv.2201.00966

2022-01-04

Abstract:Artificial Intelligence & Nanotechnology are promising areas for the future of humanity. While Deep Learning based Computer Vision has found applications in many fields from medicine to automotive, its application in nanotechnology can open doors for new scientific discoveries. Can we apply AI to explore objects that our eyes can't see such as nano scale sized objects? An AI platform to visualize nanoscale patterns learnt by a Deep Learning neural network can open new frontiers for nanotechnology. The objective of this paper is to develop a Deep Learning based visualization system on images of nanomaterials obtained by scanning electron microscope. This paper contributes an AI platform to enable any nanoscience researcher to use AI in visual exploration of nanoscale morphologies of nanomaterials. This AI is developed by a technique of visualizing intermediate activations of a Convolutional AutoEncoder. In this method, a nano scale specimen image is transformed into its feature representations by a Convolution Neural Network. The Convolutional AutoEncoder is trained on 100% SEM dataset, and then CNN visualization is applied. This AI generates various conceptual feature representations of the nanomaterial. While Deep Learning based image classification of SEM images are widely published in literature, there are not much publications that have visualized Deep neural networks of nanomaterials. There is a significant opportunity to gain insights from the learnings extracted by machine learning. This paper unlocks the potential of applying Deep Learning based Visualization on electron microscopy to offer AI extracted features and architectural patterns of various nanomaterials. This is a contribution in Explainable AI in nano scale objects. This paper contributes an open source AI with reproducible results at URL (<a class="link-external link-https" href="https://sites.google.com/view/aifornanotechnology" rel="external noopener nofollow">this https URL</a>)

Computer Vision and Pattern Recognition,Artificial Intelligence,Human-Computer Interaction

What problem does this paper attempt to address?

The problem that this paper attempts to solve is to apply artificial intelligence (AI) technology to the analysis of microscopic images at the nanoscale, especially to explore and visualize the feature representations of nanomaterials through deep - learning methods. Specifically, the paper focuses on how to use the Convolutional AutoEncoder (CAE) to generate the feature representations of nanomaterials and provide a new understanding of the structural features of nanomaterials by visualizing these features. In addition, the paper also explores how to extract common structural patterns from a large number of scanning electron microscope (SEM) images through machine - learning techniques, thereby providing a new perspective for researchers in nanotechnology to understand and classify nanomaterials.

The main contribution of the paper lies in the development of an open - source AI platform - NanoAID, which enables nanoscience researchers to use AI technology for visual exploration of materials at the nanoscale, including intuitively presenting the feature representations of nanomaterials through the concept of "AI lens", and visualizing the common patterns of specific types of nanomaterials by training convolutional neural networks (CNN). These contributions fill the gaps in the existing literature in the field of nanomaterial image visualization and promote the application of explainable AI in the field of nanotechnology.