Computing That Serves

Colloquium: Computer Vision for Everyone: Fine-grained Object Recognition


Thursday, February 20, 2014 - 11:00am


Ryan Farrell


Parris Egbert

As we contemplate "Computing that Serves", computer vision holds great promise as a technology which can have widespread impact.    With a camera on nearly every personal device (cellphone/tablet/laptop) and soon on many cars and street corners, the ability to capture images is becoming ubiquitous.  As an example, Facebook users collectively upload more than 350 million new photos every day.  While there are clearly photos of interest present in such collections, perhaps there are even more compelling, and ultimately valuable, things that we can do with this increasing abundance of cameras and sensors?

 In this talk, I will discuss how fine-grained recognition is fundamental to many important and compelling applications.  Fine-grained object recognition is an emerging area of computer vision research which seeks to recognize objects with great precision (e.g. that's a "Subaru Outback" instead of a "car", or that's an "Ivory-billed Woodpecker" instead of a "bird").  This is a task which generally requires a very high level of expertise.  I will describe some of the techniques that my collaborators and I have previously developed for this challenging task and additionally discuss several exciting problems that we are working on now and will be working on in the future.


Ryan Farrell joined the Computer Science Department at Brigham Young University in August 2013 as an Assistant Professor. Prior to joining BYU, he worked as a research scientist at the International Computer Science Institute (ICSI), a non-profit research institute affiliated with UC Berkeley. Dr. Farrell leads the ViREO Lab (Visual Recognition and Expertise with Objects), working on research problems in computer vision and related fields such as machine learning, artificial intelligence and robotics. He is currently looking for motivated students to work in his lab.