Computational methods that automatically extract knowledge from materials science data are critical for enabling the discovery of novel nanomaterials for technological applications.

A reliable identification of nanomaterials' crystal type is a crucial first step for materials characterization and analytics. Current methods are unable to identify the correct crystal type for defective structures, i.e. structures with displaced or missing atoms.

Here, we present a new machine-learning-based approach to automatically classify nanomaterials by crystal type. First, we represent nanomaterials by their diffraction pattern, expand it on spherical harmonics, and then use a convolutional neural network for classification.

We show that this approach can correctly classify a data set comprising more than 80,000 structures, including heavily defective ones, and allows to build easily interpretable structural-similarity maps.

The internal operations of the convolutional neural network are unraveled through attentive response maps, demonstrating that it uses the same landmarks a materials scientist would use, although never explicitly instructed to do so.