A Modular U-Net for Automated Segmentation of X-Ray Tomography Images in Composite Materials

• DOI: 10.3389/fmats.2021.761229
• Publication Date: 2021-11-30T23:47:16Z
• Authors: João P. C. Bertoldo; Etienne Decencière; David Ryckelynck; Henry Proudhon
• Keywords: Market Segmentation

X-Ray Computed Tomography (XCT) techniques have evolved to a point that high-resolution data can be acquired so fast classic segmentation methods are prohibitively cumbersome, demanding automated pipelines capable of dealing with non-trivial 3D images. Meanwhile, deep learning has demonstrated success in many image processing tasks, including materials science applications, showing promising alternative for human-free pipeline. However, the rapidly increasing number available architectures serious drag wide adoption this type models by end user. In paper modular interpretation U-Net (Modular U-Net) is proposed parametrized architecture easily tuned optimize it. As an example, model trained segment tomography images three-phased glass fiber-reinforced Polyamide 66. We compare 2D and versions our model, finding former slightly better than latter. observe human-comparable results achievied even only 13 annotated slices using shallow yields deeper one. consequence, neural networks show indeed venue automate XCT needing no human, adhoc intervention.