“…In recent years, convolutional neural networks (CNN) have dominated the image segmentation space by reducing the number of parameters and thus computing time and power when compared to that of a fully connected neural network [ 1 , 2 ]. Similar to multiple growing fields, materials science and engineering have benefitted from the application of CNNs in property prediction [ 3 , 4 , 5 ], the homogenization of heterogeneous materials [ 6 ], and various uses in transmission electron microscopy [ 7 , 8 , 9 ]. Surprisingly, although there have been several machine learning and semi-automated approaches applied in microstructural characterization (i.e., metallographic or ceramographic analysis) [ 10 , 11 , 12 , 13 , 14 , 15 , 16 , 17 , 18 , 19 , 20 , 21 , 22 , 23 , 24 ], the use of CNNs in grain-size determination, while growing, has been comparatively minimal.…”