Figure 4
From: Efficient few-shot machine learning for classification of EBSD patterns
Visual explanation of feature contributions. Shapley values are computed for each input image to gauge the importance of features in the EBSPs. The first row is the raw input image. The second row corresponds to the Shapley values for the correct prediction. Row three corresponds to the first incorrect classification as ranked by softmax probability. From left to right, the six space groups are \(Pm\overline{3}m\) \(Pm\overline{3}n\),\(Fm\overline{3}m\),\(Fd\overline{3}m\),\(Im\overline{3}m\), and \(Ia\overline{3}d\).
