Abstract

Exciting research is being conducted using Google's Street View imagery. Researchers can have access to training data that allows CNN training for topics ranging from assessing neighborhood environments to estimating the age of a building. However, due to the uncontrolled nature of imagery available via Google's Street View API, data collection can be lengthy and tedious. In an effort to help researchers gather address specific dwelling images efficiently, we developed an innovative and novel way of automatically performing this task. It was accomplished by exploiting Google's publicly available platform with a combination of 3 separate network types and post-processing techniques. Our uniquely developed non-maximum suppression (NMS) strategy helped achieve 99.4%, valid, address specific, dwelling images. We explored the efficacy of utilizing our newly developed mechanism to train a CNN on Unreinforced Masonry (URM) buildings. We made this selection because building collapse during an earthquake account for majority of the deaths during a disaster of this kind. An automated approach for identifying seismically vulnerable buildings using street level imagery has been met with limited success to this point with no promising results presented in the literature. We have been able to achieve the best accuracy reported to date, at 83.63%, in identifying URM, finished URM, and non-URM buildings using manually curated images. We performed an ablation study to establish synergistic parameters on ResNeXt-101-FixRes. We also present a visualization the first layer of the network to ascertain and demonstrate how a deep learning network can distinguish between various types of URM buildings. Lastly, we establish the value of our automatically generated data set for these building types by achieving an accuracy of 84.91%. This is higher than the accuracy achieved using our hand curated data set of 83.63%.

Library of Congress Subject Headings

Earthquake hazard analysis--Data processing; Dwellings--Earthquake effects; Computer vision; Automatic classification; Machine learning; Neural networks (Computer science); Convolutions (Mathematics)

Publication Date

9-2020

Document Type

Thesis

Student Type

Graduate

Degree Name

Imaging Science (MS)

Department, Program, or Center

Chester F. Carlson Center for Imaging Science (COS)

Advisor

Carl Salvaggio

Advisor/Committee Member

Anna Lang Ofstad

Advisor/Committee Member

Guoyu Lu

Campus

RIT – Main Campus

Plan Codes

IMGS-MS

Download

Share

COinS