Figure 3 | Scientific Reports

Figure 3

From: Resolution enhancement in scanning electron microscopy using deep learning

Figure 3

Histograms of the gap sizes inferred from the network input and the output images compared to the ground truth image. Total count changes among the histograms due to some of the gaps only being visible in specific images. In the input SEM images, 13.9% of the gaps were not detectible; the percentage of undetected gaps dropped to 3.7% for the output images. A Gaussian distribution, fitted to the gap histograms, with the corresponding mean gap size is also shown for each plot. The number of unresolved gaps in both the input and output images is also shown using a different color; unresolved gaps were not used for mean gap estimation. Pixel size per image is 7.1 nm; the input image is upsampled by a factor of 2.

Back to article page