Abstract
Automating the analysis portion of materials characterization by electron microscopy (EM) has the potential to accelerate the process of scientific discovery. To this end, we present a Bayesian deep-learning model for semantic segmentation and localization of particle instances in EM images. These segmentations can subsequently be used to compute quantitative measures such as particle-size distributions, radial- distribution functions, average sizes, and aspect ratios of the particles in an image. Moreover, by making use of the epistemic uncertainty of our model, we obtain uncertainty estimates of its outputs and use these to filter out false-positive predictions and hence produce more accurate quantitative measures. We incorporate our method into the ImageDataExtractor package, as ImageDataExtractor 2.0, which affords a full pipeline to automatically extract particle information for large-scale data-driven materials discovery. Finally, we present and make publicly available the Electron Microscopy Particle Segmentation (EMPS) data set. This is the first human-labeled particle instance segmentation data set, consisting of 465 EM images and their corresponding semantic instance segmentation maps.
Highlights
The ability to automate analysis during characterization by electron microscopy (EM) is a desirable endeavor due to its capability to speed up particle analysis, as well as having the potential to collect data from EM images on a large scale
We evaluate the performance of our Bayesian particle instance segmentation model using a number of metrics and compare it to several similar algorithms as benchmarks
To quantify the improvement in performance afforded by our Bayesian formulation, we report the performance of the discriminative version of our method, as well as the Bayesian version for direct comparison
Summary
The ability to automate analysis during characterization by electron microscopy (EM) is a desirable endeavor due to its capability to speed up particle analysis, as well as having the potential to collect data from EM images on a large scale. Learning-based segmentation methods that perform well in a variety of cases (given enough training data) have taken off in the past decade with the democratization of machine-learning methods in science The robustness of these methods is well suited to automating data extraction and particle analysis from EM images, given that the need to tune parameters for edge cases can be eliminated by training on a sufficiently large and diverse set of labeled examples. Its authors used a series of image-processing methods such as thresholding, contour detection, and ellipse fitting to detect and locate particles in electron micrographs, subsequently performing particle analysis on the identified particles They were able to automate the entire process, from the extraction of EM images from scientific literature using ChemDataExtractor,[2] to the measuring of scalebars in these images, achieving accurate particle and scalebar measurements. Kim et al measure particles in SEM images using a neural network to predict particle morphologies, followed by an application of the watershed segmentation
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.
