Ryan's work is featured on the cover of the July 2021 JOM (the member journal of the Minerals, Metals, and Materials Society). images/pdf-documents/cohn-jom-mag-j7uly21-cover-min.pdf

It can be challenging for conventional image analysis software to identify individual objects when many objects overlap. That is the case in this SEM micrograph of gas atomized Ni superalloy powder produced for additive manufacturing.To address this challenge, advanced computer vision techniques were applied to locate and segment each individual powder particle, termed ‘instance segmentation.’ Overlaid on the original image are colored bounding boxes and segmentation masks for each particle, as determined by the computer vision system. (Note that the color of each mask is randomly assigned for visual clarity and does not correspond to particle properties.) After powder particles have been segmented, they can be used as data for subsequent machine learning investigations to correlate particle characteristics, such as surface morphology, with powder flow properties to improve the quality and decrease the cost of feedstock powders for additive manufacturing. See “Instance Segmentation for Direct Measurements of Satellites in Metal Powders and Automated Microstructural Characterization from Image Data” by Ryan Cohn et al. for details.