Abstract
Machine-learning (ML) techniques hold the potential of enabling efficient quantitative micrograph analysis, but the robustness of ML models with respect to real-world micrograph quality variations has not been carefully evaluated. We collected thousands of scanning electron microscopy (SEM) micrographs for molecular solid materials, in which image pixel intensities vary due to both the microstructure content and microscope instrument conditions. We then built ML models to predict the ultimate compressive strength (UCS) of consolidated molecular solids, by encoding micrographs with different image feature descriptors and training a random forest regressor, and by training an end-to-end deep-learning (DL) model. Results show that instrument-induced pixel intensity signals can affect ML model predictions in a consistently negative way. As a remedy, we explored intensity normalization techniques. It is seen that intensity normalization helps to improve micrograph data quality and ML model robustness, but microscope-induced intensity variations can be difficult to eliminate.
Highlights
Micrographs constitute an important class of scientific data and play a key role in the interpretation of material process–structure–property (PSP) linkage by revealing material microstructures
We differentiate two classes of image signals: (1) microstructure-induced signals, which come from the microstructure content of the material sample, and (2) microscopeinduced signals, which come from instrument conditions and user-controlled settings
We discuss the quantification of micrograph pixel intensity, the variation of pixel intensities within the original dataset, the ML model hyperparameter choices, and the effect of intensity normalization
Summary
Micrographs constitute an important class of scientific data and play a key role in the interpretation of material process–structure–property (PSP) linkage by revealing material microstructures. The content of a micrograph depends on multiple factors, including the microstructure being captured, the manufacturer of the microscope[16], the user-controlled settings (such as magnification, brightness, contrast), and instantaneous microscope conditions (such as filament aging)[17]. We differentiate two classes of image signals: (1) microstructure-induced signals, which come from the microstructure content of the material sample, and (2) microscopeinduced signals, which come from instrument conditions and user-controlled settings. Even if all micrographs are collected by the same person on the same microscope using high-throughput auto collection techniques, the microscope-induced signals can still vary if the experiment takes a prolonged time and instrument drift happens during the collection[18]. Material chemistry (e.g., atomic number) contributes to micrograph pixel intensities, but holistically they are intrinsic to the material sample and are reproducible
Sign-in/Register to view highlights & summary Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
