|
Hokkaido University Collection of Scholarly and Academic Papers >
Research Institute for Electronic Science >
Peer-reviewed Journal Articles, etc >
Visible fingerprint of X-ray images of epoxy resins using singular value decomposition of deep learning features
This item is licensed under:Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International
| Title: | Visible fingerprint of X-ray images of epoxy resins using singular value decomposition of deep learning features |
| Authors: | Avalos, Edgar Browse this author | | Akagi, Kazuto Browse this author | | Nishiura, Yasumasa Browse this author →KAKEN DB |
| Keywords: | Epoxy resin | | X-ray CT scan | | Deep learning | | Convolutional neural network | | Computer vision | | Structure-property mapping |
| Issue Date: | Jan-2021 |
| Publisher: | Elsevier |
| Journal Title: | Computational materials science |
| Volume: | 186 |
| Start Page: | 109996 |
| Publisher DOI: | 10.1016/j.commatsci.2020.109996 |
| Abstract: | Although the process variables of epoxy resins alter their mechanical properties, recently it was found that the total variation of the X-ray images of these resins is one of the key features that affect the toughness of these materials. However it is still not clear how to visualize such a difference in a clear way. To facilitate the visualization, we use a robust approximation of the gradient of the intensity field of the X-ray images of different kinds of epoxy resins and then we use deep learning to discover the most representative features of the transformed images. In this solution of the inverse problem to find characteristic features to discriminate samples of heterogeneous materials, we use the eigenvectors obtained from the singular value decomposition of all the channels of the response maps of the early layers in a convolutional neural network. While the strongest activated channel gives a visual representation of the characteristic features, often these are not robust enough in some practical settings. On the other hand, the left singular vectors of the matrix decomposition of the response maps barely change when variables such as the capacity of the network or the network architecture change. High classification accuracy and robustness of characteristic features are presented in this work. |
| Rights: | © 2021. This manuscript version is made available under the CC-BY-NC-ND 4.0 license http://creativecommons.org/licenses/by-nc-nd/4.0/ | | http://creativecommons.org/licenses/by-nc-nd/4.0/ |
| Type: | article (author version) |
| URI: | http://hdl.handle.net/2115/87840 |
| Appears in Collections: | 電子科学研究所 (Research Institute for Electronic Science) > 雑誌発表論文等 (Peer-reviewed Journal Articles, etc)
|
Submitter: 西浦 廉政
|
