An AI model adapted from image inpainting can reconstruct missing atoms in crystal structures with a reported 97 percent success rate. Described in npj Computational Materials, the method applies computer vision, in which AI recognizes and interprets visual information, to incomplete crystal structures. The new method instead focuses on filling missing positions within structures that are already known experimentally. The researchers turned to AI to recover that missing information. “We can now use our method, for example, to complete structures in databases with the missing hydrogen positions,” says Pizzi.