Observing the interphase and surface morphologies at the microscale level with a multi-component machine learning protocol
Microscale imaging analysis is an indispensable technique across various application domains, including the investigation of micro-interfaces in electrode materials, the exploration of catalytic surface morphologies, and the examination of alloy microstructures. However, most available image processing software lacks specialization for addressing the unique challenges at the microscale level, such as noise interference and intricate structural compositions. Additionally, current deep learning frameworks tend to be highly complex and require significant expertise and resources to implement and train. Here, we introduce "Visualization through Classification, Class Activation Mapping and Clustering at a microscale level" (Visual-CCC), a novel AI toolkit approach that provides precise visualization of latent features in imagery samples, at a microscale level. Visual-CCC leverages digital image processing algorithms supplemented by a supervised neural network and a class activation mapping technique. Additionally, an unsupervised learning approach was implemented to complement the toolkit’s visualizations. Using imagery data of solid electrolyte interphase (SEI) as a case study, Visual-CCC achieves a weighted accuracy of 93.8\% in morphology determination. To validate the generalization of the toolkit, microscale imagery of alloy substances was tested. We have released a user-friendly graphical user interface (GUI) to facilitate the use of the toolkit. The results are promising, positioning Visual-CCC to make significant contributions to microscale imaging research in material characterization and electron microscopy, with the potential for expansion into other fields such as medicine and biology.
Karl Luigi Loza Vidaurre1,2, Zhilong Wang1,2, and Jinjin Li1,2
1National Key Laboratory of Advanced Micro and Nano Manufacture Technology, Shanghai Jiao Tong University, Shanghai, 200240, China
2Department of Micro/Nano Electronics, School of Electronic Information and Electrical Engineering, Shanghai Jiao Tong University, Shanghai, 200240, China