Physics-Based X-ray CT Simulation for Synthetic Data Generation and Deep Learning Segmentation of Energy Materials

Nadia Mariana Leal Reyes, Salvatore De Angelis, Estrid Buhl Naver, Luise Theil Kuhn, Peter Stanley Jørgensen

Keywords: Simulation, Deep Learning, Materials

Abstract: This work presents the use of simulated physics-based synthetic X-ray data to train deep learning models for segmenting energy-material microstructures. Preliminary results with a U-Net trained on simulated data show that the U-Net can effectively learn to accurately segment microstructure in the presence of artifacts.

Serve As Reviewer: ~Anders_Dahl1

Submission Number: 25