Hillary Luan
Pronouns: she/her
Research Mentor(s): Carol Flannagan
Co-Presenter: Sanz Planchart, Mariana
Research Mentor School/College/Department: UMTRI / Engineering
Presentation Date: April 20
Presentation Type: Oral5
Session: Session 1 – 10am – 10:50am
Room: Breakout Room 1
Authors:
Presenter: 3
Abstract
Video recordings of drivers provide essential data for improving driver safety, but to be usable, the data must first be extracted. Compared to human coders, automated algorithms have the potential to increase efficiency and accessibility of extracted information. My project focuses on developing strategies for automating video data extraction to record vehicle occupant behavior, specifically on developing a rubric to categorize various extracted information. These techniques could be extended to other research areas that collect data in video form. However, there is a potential to introduce an unintended bias in the algorithms resulting in negative social impacts. Measuring and reducing this bias will be a key goal of our work. To address both bias and efficiency, we will evaluate the process used by human coders, eventually including recording eye gaze as they code videos. We will develop a rubric in which human coders can accurately process and categorize videos in order to have universal agreement on unanimous agreement on descriptions. We will address the questions of what roles human and computer play in the development of computer vision algorithms, finding and addressing bias, and how to decrease inefficiencies.
Engineering, Social Sciences
