Teaching Machines to Classify and Organize Reconnaissance Image Data
Post-event building reconnaissance teams have a clear mission. These teams of trained engineers are charged with collecting perishable data to be used for learning from disasters before that data is destroyed. An enormous amount of visual data (images and videos) can be generated in just a few days. However, the ability to utilize these large volumes of data are limited due to tedious and time-consuming processes needed to sift through and analyze them.
To distil such information in an efficient and expeditious manner, this project aims to implement images classification methods and computer vision methods to make time-critical decisions in the field regarding the subsequent collection of perishable data. The methods developed and validated in this research will enable rapid information extraction from these images to make timely decisions regarding where to collect the highest value perishable visual data.
Key enablers of this research are the current powerful computer vision and machine learning algorithms and modern parallel computing platforms (GPUs and multi-core CPUs).
Project website: https://youtu.be/WO3XmXKu4uI
Faculty Investigators: Shirley Dyke, Bedrich Benes, Thomas Hacker
Postdoctoral Researcher: Chul Min Yeum
Undergraduate Researchers: Corey Beck, Seth Lindsey, Haoyue Wang
Sponsor: National Science Foundation