Deterioration level estimation via neural network maximizing category-based ordinally supervised multi-view canonical co
- PDF / 6,148,786 Bytes
- 22 Pages / 439.642 x 666.49 pts Page_size
- 37 Downloads / 152 Views
Deterioration level estimation via neural network maximizing category-based ordinally supervised multi-view canonical correlation Keisuke Maeda1
· Sho Takahashi2 · Takahiro Ogawa3 · Miki Haseyama1
Received: 30 July 2019 / Revised: 9 September 2020 / Accepted: 7 October 2020 / © Springer Science+Business Media, LLC, part of Springer Nature 2020
Abstract A deterioration level estimation method via neural network maximizing category-based ordinally supervised multi-view canonical correlation is presented in this paper. This paper focuses on real world data such as industrial applications and has two contributions. First, a novel neural network handling multi-modal features transforms original features into features effectively representing deterioration levels in transmission towers, which are one of the infrastructures, with consideration of only correlation maximization. It can be realized by setting projection matrices maximizing correlations between multiple features into weights of hidden layers. That is, since the proposed network has only a few hidden layers, it can be trained from a small amount of training data. Second, since there exist diverse characteristics and an ordinal scale in deterioration levels, the proposed method newly derives category-based ordinally supervised multi-view canonical correlation analysis (Co-sMVCCA). Co-sMVCCA enables estimation of effective projection considering both within-class divergence and the ordinal scale between classes. Experimental results showed that the proposed method realizes accurate deterioration level estimation. Keywords Neural network · Within-class divergence · Ordinal scale · Canonical correlation · Deterioration level estimation Keisuke Maeda
[email protected] Sho Takahashi [email protected] Takahiro Ogawa [email protected] Miki Haseyama [email protected] 1
Office of Institutional Research, Hokkaido University, Sapporo, Japan
2
Faculty of Engineering, Hokkaido University, Sapporo, Japan
3
Faculty of Information Science and Technology, Hokkaido University, Sapporo, Japan
Multimedia Tools and Applications
1 Introduction With the development of hardware devices and advent of the big data era, convolutional neural networks (CNNs) have been effectively trained by using a large-scale dataset [10, 45] such as ImageNet and have achieved accurate image classification [12]. In the field of information science, although many researchers have focused on large-scale datasets, many recent studies have been conducted by using not images for generic object recognition but real data such as agricultural images [20], medical images [23] and images for infrastructure management [25] in order to efficiently support experts in several fields. In studies using images for infrastructure management, automatic detection of specific distresses [4] such as potholes and automatic deterioration level estimation [39] have attracted much attention. Since human errors often occur in manual deterioration level estimation due to ambiguity in
Data Loading...
