Aprendizaje profundo para clasificación de nanopartículas en micrografías SEM: mapeo sistemático

In recent years the microscopy industry has had significant technological advances.Today these advances allow us to obtain micrographs taken by microscopes with a high resolution at very small scales that allow us to observe nanoparticles, an example of this is the scanning electron microscope (SEM). The analysis of nanoparticles contained within these micrographs is necessary in order to improve manufacturing methods, purification systems, improvements in the medical industry, among others. In this article, a systematic mapping study of the literature on the use of deep learning techniques for the detection and classificationof nanoparticles contained in SEM micrographs is presented. An analysis is made of the main neural networks used and the methods implemented for the generation of an input training database for the neural networks. Results show that convolutional neural networks (CNN) are themost common techniques to analyze micrographs, obtaining a high accuracy in the projects developed in the reviewed publications.In order to create a framework and methodology that allows us to detect, classify and count, the mapping study allows us to identify the most efficient technologies to develop systems capable of performing these tasks in an automated way with the help of deep learning.