Dataset of 18,577 SEM images produced at CNR-IOM (Trieste, Italy). Images are classified into 10 categories in a folder structure, which have been used for convolutional neural network training. Results obtained from this dataset have been published in Modarres et al., Scientific Reports volume 7, Article number: 13282 (2017), doi:10.1038/s41598-017-13565-z The dataset is appropriate for the purposes of this study and in general for visual object recognition software research. Any scientific metadata associated to the measure is not present in the images. The dataset is therefore relevant as a whole, being the single images entirely detached from any specific information or scientific detail related to the displayed subject. This work has been done within the NFFA-EUROPE project (www.nffa.eu) and has received funding from the European Union’s Horizon 2020 Research and Innovation Programme under grant agreement No. 654360 NFFA-Europe.