Fig. 3: Representative results for the C-CNN and R-CNN for simulated test sets with different number of domains and domain structure. | npj Computational Materials

Fig. 3: Representative results for the C-CNN and R-CNN for simulated test sets with different number of domains and domain structure.

From: Ultrafast Bragg coherent diffraction imaging of epitaxial thin films using deep complex-valued neural networks

Fig. 3

Predicted amplitude and phase in the real space with C-CNN and R-CNN for (a) 15 domains with grid position and (b) 50 domains with random position. Quantitative comparison on synthetic test set with histograms of SSIM calculated between prediction and the ground truth of amplitude and phase, and the χ2 errors between the input and reconstructed diffraction patterns with C-CNN and R-CNN for (c) 15 domains with grid position and (d) 50 domains with random position.

Back to article page