Fig. 2: Bragg disk detection using cross-correlation and deep-learning methods. | npj Computational Materials

Fig. 2: Bragg disk detection using cross-correlation and deep-learning methods.

From: Disentangling multiple scattering with deep learning: application to strain mapping from electron diffraction patterns

Fig. 2: Bragg disk detection using cross-correlation and deep-learning methods.

a Examples of simulated diffraction patterns for crystals of different thicknesses and orientations. Scale bar is 0.5 Å−1. b, c The positions of the ground truth structure factor coefficients of the crystal lattice are plotted below as blue circles, with a size proportional to the structure factor amplitudes Vg. The structure factor positions were computed using (b) template matching by cross-correlation with the vacuum probe signal, and (c) the FCU-Net network. Both measurements are overlaid as black crosses, with a size proportion to the estimated disk amplitude (square root of the disk intensity) and Vg amplitudes, for the correlation and FCU-Net predictions, respectively. The total intensity-weighted accuracy is listed above for all measurements.

Back to article page