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. Meanwhile, deep learning has demonstrated success in many image processing tasks, including materials science applications, showing a promising alternative for a human-free segmentation pipeline. However, the rapidly increasing number of available architectures can be a serious drag to the wide adoption of this type of models by the end user. In this paper a modular interpretation of U-Net (Modular U-Net) is proposed with a parametrized architecture that can be easily tuned to optimize it. As an example, the model is 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 13 annotated slices and using a shallow U-Net yields better results than a deeper one. As a consequence, neural networks show indeed a promising venue to automate XCT data processing pipelines needing no human, adhoc intervention.