@knc6, can you please implement whatever you think is a reasonable atomistic mask generation scheme? For atom localization (#3) let's start with a U-Net strategy, so we'll need simulated ST(E)M images paired with binary label images
We have trained our model to perform atom segmentation, atomic-column Gaussian mapping, intensity-preserving super-resolution (deblur) processing, denoising and background removal. Their respective ground truth labels are shown in Fig. 3 and Table 2. The width of the circular mask is defined by the full width at half maximum of the point spread function and the width of the Gaussian mask is 0.2 angstrom.
to get an initial baseline system going, we may want to just skip the superresolution bit, or maybe do something simpler?
@knc6, can you please implement whatever you think is a reasonable atomistic mask generation scheme? For atom localization (#3) let's start with a U-Net strategy, so we'll need simulated ST(E)M images paired with binary label images
We have trained our model to perform atom segmentation, atomic-column Gaussian mapping, intensity-preserving super-resolution (deblur) processing, denoising and background removal. Their respective ground truth labels are shown in Fig. 3 and Table 2. The width of the circular mask is defined by the full width at half maximum of the point spread function and the width of the Gaussian mask is 0.2 angstrom.
to get an initial baseline system going, we may want to just skip the superresolution bit, or maybe do something simpler?
from the jarvis-materials-design repo: