Machine learning based Software Toolkit for Automated identification in atomic-REsolution operando nanoscopy

ended

In recent years, the analysis of large data sets is becoming increasingly important in the fields of material science and engineering. There is a strong demand for real-time automated identification algorithms in electron microscopy (EM) for the analysis of atomic-structure, phases, and defects. Unfortunately, it is non-trivial to obtain or extract meaningful scientific information from raw EM output digital data. It requires a tedious process of filtering/fitting and the expertise of a seasoned microscopist. With the rapid development of information technology and computer science, automated computer-assisted analysis of electron microscopy images/data is becoming a reality. In the past decade, different techniques have been developed and applied to digital data analysis. Meanwhile, the rapid development of novel microscopy techniques and instrumentation, e.g. in situ/operando and pixelated detector-based techniques, require high-speed data execution and analysis. Currently, several groups worldwide are concentrating their efforts into implementing machine learning and deep learning algorithms for image/data analysis. However, this is still a very undeveloped direction in the field of electron microscopy for materials science, especially in Europe. According to the Digital Transformation Monitor, artificial intelligence-based technologies will play a major role in future economy. The ability to analyse levels of data that are beyond human comprehension will allow business to personalize experiences, customize products and services and identify growth opportunities with a speed and accuracy that has never been possible before. The objective of this PoC is to generate an innovative software package that enables the analysis of large sets of EM data (i) at high throughput with (ii) low costs, in (iii) a standardized approach and (iv) under operando conditions, based on advanced machine learning algorithms.