Recent Advances in Scripting, Automation, Artificial Intelligence, Machine Learning for advanced TEM applications
In the past few years, machine learning and deep learning have matured significantly in performance and speed, becoming integratable into closed-loop automation sequences. Thermo Fisher Scientific's AutoScript, a cross-platform Python-based API, allows to automate and optimize electron microscopy workflows. By employing instance segmentation, we can separate features in an image, count them, and determine properties like size and shape. Incorporating this into automated workflow enables the targeted experimental setting saving the microscopist’s efforts and precious time. In the case of HR TEM imaging, the positions and diameters of atoms can be detected within the automated workflow. The integration of a neural network enables rapid prediction of atomic structures, which can be performed at multiple locations. Additionally, the neural network can help searching for experimentally relevant interfaces. This automation sequence as well provides the opportunity to optimize experimental set-up during the process. The booth presentation will discuss advances at Thermo Fisher Scientific in fields of automation, scripting and deep learning TEM based data acquisition applications.