Mixed-Scale Dense Convolutional Neural Network based Improvement of Glass Fiber-reinforced Composite CT Images

TitleMixed-Scale Dense Convolutional Neural Network based Improvement of Glass Fiber-reinforced Composite CT Images
Publication TypeConference Abstract
Year of Publication2019
AuthorsElberfeld, T., S. Bazrafkan, J. De Beenhouwer, and J. Sijbers
Conference Name4th International Conference on Tomography of Materials & Structures
Date Published07/2019
Abstract

For the study of glass fiber-reinforced polymers (GFRP), µCT is the method of choice. Obtaining
GFRP parameters from a µCT scan is difficult, due to the presence of noise and artifacts. We propose a method
to improve GFRP image quality using a recently introduced deep neural network. We describe the network’s
setup and the data generation and show how the trained network improves the reconstruction.

Research area: