3:50 PM - 4:10 PM
[2O5-GS-13-01] A study of applicability of deep learning to extract tacit knowledge of civil engineering and disaster prevention engineers
Keywords:Civil engineer tacit knowledge, Civil engineering, Sabo, Topographic interpretation, Training data selection
Research and diagnosis in the fields of civil engineering and Sabo are often judged on empirical knowledge and tacit knowledge of experts, and it may not be possible to create 100% correct data. Topographic interpretation is one of them. The topographic interpretation is not only a method for predicting sediment disaster sites, but also can be used to analysis information, such as potentials for topographic and geological objects. However, individual differences are likely to occur due to skill of engineers. Also, it becomes a great burden for the analysis in a large area.
In this paper, we describe researches and examples of deep learning in the fields of the civil engineering and the Sabo, which are often judged by the tacit knowledge of experts, including topographic interpretation. In order to verify the applicability, we created train data based on expert knowledge and learned and predicted using pix2pix. As a result, the accuracy rate was about 70% for the correct data created by the experts, and characteristics obtained from the train data have characteristics intended by the experts who created the correct data, such as the change of the shape and slope of the cliff contained in landslide terrains. Thus, it found that the deep learning is useful in the fields of the civil engineering and the Sabo, and is also effective in labor saving, efficiency, and rapidity.
In this paper, we describe researches and examples of deep learning in the fields of the civil engineering and the Sabo, which are often judged by the tacit knowledge of experts, including topographic interpretation. In order to verify the applicability, we created train data based on expert knowledge and learned and predicted using pix2pix. As a result, the accuracy rate was about 70% for the correct data created by the experts, and characteristics obtained from the train data have characteristics intended by the experts who created the correct data, such as the change of the shape and slope of the cliff contained in landslide terrains. Thus, it found that the deep learning is useful in the fields of the civil engineering and the Sabo, and is also effective in labor saving, efficiency, and rapidity.
Authentication for paper PDF access
A password is required to view paper PDFs. If you are a registered participant, please log on the site from Participant Log In.
You could view the PDF with entering the PDF viewing password bellow.