Semantic Segmentation of Porosity in 4D Spatio-Temporal X-ray μCT of Titanium Coated Ni wires using Deep Learning
Abstract: A fully convolutional neural network was used to measure the evolution of the volume fraction of two different Kirkendall pores during the homogenization of Ti coated Ni wires. Traditional methods like Otsus thresholding and the largest connected component analysis were used to obtain the masks for training the segmentation model. Once trained, the model was used to semantically segment the two types of pores at different stages in their evolution. Masks of the pores predicted by the network were then used to measure the volume fraction of porosity at 0 mins, 240 mins, and 480 mins of homogenization. The model predicted an increase in porosity for one type of pore and a decrease in porosity for another type of pore due to pore sintering, and it achieved an F1 Score of 0.95.
- Nobuyuki Otsu. A threshold selection method from gray-level histograms. IEEE Transactions on Systems, Man, and Cybernetics, 9(1):62–66, 1979.
- Porosity segmentation in x-ray computed tomography scans of metal additively manufactured specimens with machine learning. Additive Manufacturing, 36:101460, 08 2020.
- Understanding important features of deep learning models for segmentation of high-resolution transmission electron microscopy images. npj Computational Materials, 6:108, 07 2020.
- Phase segmentation of x-ray computer tomography rock images using machine learning techniques: An accuracy and performance study. Solid Earth, 7:1125–1139, 07 2016.
- Automated segmentation of computed tomography images of fiber-reinforced composites by deep learning. Journal of Materials Science, 55:16273 – 16289, 2020.
- Kirkendall pore evolution during interdiffusion and homogenization of titanium-coated nickel microwires. Intermetallics, 134:107199, 07 2021.
- Deep learning vs. traditional computer vision. 04 2019.
- Reconciling modern machine-learning practice and the classical bias–variance trade-off. Proceedings of the National Academy of Sciences, 116:15849 – 15854, 2018.
- U-net: Convolutional networks for biomedical image segmentation. volume 9351, pages 234–241, 10 2015.
- Shortcut learning in deep neural networks. Nature Machine Intelligence, 2:665–673, 11 2020.
- Invariant risk minimization, 2019.
- Eric Kauderer-Abrams. Quantifying translation-invariance in convolutional neural networks. 12 2017.
- Deep Learning. MIT Press, Cambridge, MA, USA, 2016. http://www.deeplearningbook.org.
- Chris M. Bishop. Training with noise is equivalent to tikhonov regularization. Neural Computation, 7(1):108–116, 1995.
- Towards theoretically understanding why sgd generalizes better than adam in deep learning. 10 2020.
- Sgd generalizes better than gd (and regularization doesn’t help). 02 2021.
- Improving generalization performance by switching from adam to sgd. 12 2017.
Paper Prompts
Sign up for free to create and run prompts on this paper using GPT-5.
Top Community Prompts
Collections
Sign up for free to add this paper to one or more collections.