Figure 1
Sketch of the data processing and DL model training. (a) Large Bragg coherent diffraction pattern where small 32 × 32 × 32 pixel-size portions are randomly selected. (b) Small portions in renormalized logarithmic scale, artificially masked with zeros to simulate detector gaps and used as input to the DL model. (c) DL model predictions for the corresponding masked inputs with the inpainted gap. |