Figure 2
TXM synchrotron experimental result and ground-truth (manually segmented) images with 1024 × 1024 × z voxels from a ∼25 µm-diameter micropillar extracted from an additively manufactured 316L stainless steel sample (z represents the number of slices along its direction, e.g. four in the figure). Dashed yellow squares represent the 192 × 192 × 192 voxels sub-volumes used to train the neural network. |