Table 4 MAE in the ECD (µm) from the ensembled segmentation maps predicted for OOD image set against that measured by EBSD in the corresponding region

From: Generalizable image segmentation for microstructure characterization through integrated SEM and EBSD analysis

 

Training set modality

Model

BSE 10 keV

BSE 20 keV

SE 10 keV

SE 20 keV

ImageNet/DICE

2.04

2.49

0.87

0.85

MicroNet/DICE

19.40

11.64

0.40

2.07

MicroNet/TopoDICE

21.00

9.59

0.34

4.37

  1. The bolded and underlined entry indicates model with the lowest MAE.
Back to article page