[Theory] 8/17 Thesis Defense: Ruotian Luo, TTIC

[Mary Marre]

*Thesis Defense: Ruotian Luo, TTIC*

*When:  *     Tuesday*,* August 17th at *8:00 - 10:00 am CT*



*Virtually: *  *Register in advance here
<https://uchicago.zoom.us/meeting/register/tJUrfuuuqD4pGtFW2b2FRbZI7dz7Y2ePyH6B>*



*Who: *        Ruotian Luo, TTIC


*Thesis title: *Goal-Driven Text Descriptions for Images

*Abstract: *While visual understanding has achieved significant progress in
recent years, only perception is not enough for building AI; an AI agent
also needs to know how to talk, especially how to communicate with a human.
In the thesis, I study how we design goals for generating more meaningful
texts given visual inputs. First, I will introduce a method to generate
informative image tags, based on the idea of information utility: how much
information does a tag convey about the image. Then, I will discuss how we
incorporate “discriminability” in generating referring expressions and
image captions. Specifically, I will introduce a speaker-listener framework
where the speaker generates the expressions/captions, and the listener
tells if they are discriminative or not.

*Thesis Advisor:* *Greg Shakhnarovich* <greg at ttic.edu>




Mary C. Marre
Faculty Administrative Support
*Toyota Technological Institute*
*6045 S. Kenwood Avenue*
*Chicago, IL  60637*
*mmarre at ttic.edu <mmarre at ttic.edu>*
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.cs.uchicago.edu/pipermail/theory/attachments/20210813/7961b97e/attachment.html>