Predicting human behavior using visual media
Name
986529121-MIT.pdf
Description
Full printable version
Size
24.64 MB
Format
Adobe PDF
Checksum (MD5)
204bd5985a0c3122e157d2c37c86390e
Author(s)
Khosla, Aditya
Advisor(s)
Antonio Torralba.
Date Issued
2017
Publisher
Massachusetts Institute of Technology
Abstract
The ability to predict human behavior has applications in many domains ranging from advertising to education to medicine. In this thesis, I focus on the use of visual media such as images and videos to predict human behavior. Can we predict what images people remember or forget? Can we predict the type of images people will like? Can we use a photograph of someone to determine their state of mind? These are some of the questions I tackle in this thesis. Through my work, I demonstrate: (1) It is possible to predict with near human-level correlation, the probability with which people will remember images, (2) it is possible to predictably modify the extent to which a face photograph is remembered, (3) it is possible to predict, with a high correlation, the number of views an image will receive even before it is uploaded, (4) it is possible to accurately identify the gaze of people in images, both from the perspective of a device, and third-person. Further, I develop techniques to visualize and understand machine learning algorithms that could help humans better understand themselves through the analysis of algorithms capable of predicting behavior. Overall, I demonstrate that visual media is a rich resource for the prediction of human behavior.
Description
Thesis: Ph. D., Massachusetts Institute of Technology, Department of Electrical Engineering and Computer Science, 2017.
Cataloged from PDF version of thesis.
Includes bibliographical references (pages 161-173).
Subjects
Electrical Engineering and Computer Science.
MIT Department
Massachusetts Institute of Technology. Department of Electrical Engineering and Computer Science
Terms of Use
MIT theses are protected by copyright. They may be viewed, downloaded, or printed from this source but further reproduction or distribution in any format is prohibited without written permission.
Persistent DSpace Link
