Photoelastic Dispersion Coefficient by Holographic Reconstruction with Neural Networks and the Fresnel Method

Abstract

Here we report the characterization of the photoelastic dispersion coefficient using digital holography with two distinct reconstruction methods: one based on the Fresnel method and the other utilizing convolutional neural networks (CNN). The CNN was trained with reconstruction from the Fresnel method and was able to provide reconstructions with an average Mean Squared Error of 0.006.