Electron microscopy measurements are usually performed by collecting all points in a 2D grid. Here, using deep learning in real time, only sites of interest are measured (colored circles), allowing experiments to be conducted on a much larger variety of materials, even those that change under the beam.