Classification of Rugosity in Plasmonic Metallic Thin Films Using Deep Learning for Speckle Images

C.N. Magaña-Barocio; Marlen Gonzalez; M.C. Peña-Gomar; M. Torres.Cisneros; Rodríguez-Sánchez, Alejandro E.

doi:10.1364/LAOP.2024.W4A.33

Classification of Rugosity in Plasmonic Metallic Thin Films Using Deep Learning for Speckle Images

Journal

Optica Latin America Optics and Photonics Conference (LAOP) 2024

Date Issued

2024

Author(s)

C.N. Magaña-Barocio

Marlen Gonzalez

M.C. Peña-Gomar

M. Torres.Cisneros

Rodríguez-Sánchez, Alejandro E.

Facultad de Ingeniería - CampGDL

Type

text::conference output::conference proceedings::conference paper

DOI

10.1364/LAOP.2024.W4A.33

URL

https://scripta.up.edu.mx/handle/20.500.12552/11901

Abstract

In this work, we report, for the first time, to the best of our knowledge, the classification of metallic samples with different roughness values. As a reference, the Ra and Rq values were obtained using a Mitutoyo roughness meter. About 2,000 Speckle images were obtained for each sample. They were processed and used as inputting neural networks such as ResNet50 and EfficientNet. We obtained 99.63 % accuracy in classifying the samples with the ResNet50 model and 99.48 % accuracy for the EfficientNet model. These accuracies can be compared with the 99.926 % and 99.932 % values obtained for aluminum and steel surfaces in a similar work that used an optics system, image processing, and a CNN.