A modular U-Net for automated segmentation of X-ray tomography images in composite materials

TL;DR

This paper introduces a modular U-Net architecture for automated segmentation of 3D X-ray tomography images in composite materials, demonstrating effective results with minimal annotations and shallow networks, advancing automated material analysis.

Contribution

A novel modular U-Net design tailored for 3D XCT image segmentation, showing that shallow networks and limited annotations can achieve human-level accuracy.

Findings

2D model slightly outperforms 3D model

Shallow U-Net yields better results than deeper variants

Effective segmentation with only 10 annotated layers

Abstract

X-ray Computed Tomography (XCT) techniques have evolved to a point that high-resolution data can be acquired so fast that classic segmentation methods are prohibitively cumbersome, demanding automated data pipelines capable of dealing with non-trivial 3D images. Deep learning has demonstrated success in many image processing tasks, including material science applications, showing a promising alternative for a humanfree segmentation pipeline. In this paper a modular interpretation of UNet (Modular U-Net) is proposed and trained to segment 3D tomography images of a three-phased glass fiber-reinforced Polyamide 66. We compare 2D and 3D versions of our model, finding that the former is slightly better than the latter. We observe that human-comparable results can be achievied even with only 10 annotated layers and using a shallow U-Net yields better results than a deeper one. As a consequence, Neural Network (NN) show indeed a promising venue to automate XCT data processing pipelines needing no human, adhoc intervention.