Published in 2018 ASEE Annual Conference & Exposition Proceedings, June 23, 2018.
Original paper found at: https://peer.asee.org/30830
To address existing challenges with filtering and classification of post-earthquake structural damage images, the authors are engaged in a multidisciplinary project to develop and train a machine-learning algorithm that identifies relevant photographs and assigns damage tags to those images. The research team is predominantly comprised of undergraduate students and is led by a structural engineering and a computer science faculty. While machine-learning algorithms have been successfully used for image tagging in a variety of fields (health care, manufacturing, etc.), the extension of this approach for earthquake reconnaissance is only just beginning. As such, the creation and development of this tool is a new and dynamic project-based learning experience for both the students and faculty involved.
This collaborative project emphasizes student initiative and innovation where they are active in all development stages of the tool ranging from collection and tagging of earthquake damage images, coding and testing the machine-learning algorithm, to writing papers for and presenting at conferences. In addition, the unique nature of this project exposes students to a field and possible career path they may not have encountered in their typical course of study. The authors provide a comprehensive discussion of the results of faculty and student surveys/ interviews and conclude by highlighting some of the greatest benefits of the multidisciplinary project. They also point out lessons learned engaging in a project with a large scope, diverse experts (who have limited knowledge of the partnering disciplines), and a number of undergraduate students who began as novices in their respective research area.
Number of Pages