Talks: 18 December 2014

IEEE Signal Processing Society, Bangalore Chapter, IEEE Bangalore Section,
and Supercomputer Education Research Centre, Indian Institute of Science

Invite you to the following talks:

1) Title: Beyond Mindless Labeling: *Really* Leveraging Humans to Build
Intelligent Machines

by Dr. Devi Parikh, Assistant Prof., Virginia Tech.


Human ability to understand natural images far exceeds machines today. One
reason for this gap is an artificially restrictive learning set up –
humans today teach machines via Morse code (e.g. providing binary labels
on images, such as “this is a horse” or “this is not”), and machines are
typically silent. These systems have the potential to be significantly
more accurate if they tap into the vast common-sense knowledge humans have
about the visual world. I will talk about our work on enriching the
communication between humans and machines by exploiting mid-level visual
properties or attributes. I will also talk about the more difficult
problem of directly learning common-sense knowledge simply by observing
the structure of the visual world around us. Unfortunately, this requires
automatic and accurate detection of objects, their attributes, poses, etc.
in images, leading to a chicken-and-egg problem. I will argue that the
solution is to give up on photorealism. Specifically, I will talk about
our work on exploiting human-generated abstract visual scenes to learn
common-sense knowledge and study high-level vision problems.


2) Title: Hedging Against Uncertainty in Machine Perception via Multiple
Diverse Predictions

by Dr. Duruv Bhatra, Assistant Prof., Virginia Tech.


What does a young child or a high-school student with no knowledge of
probability do when faced with a problem whose answer they are uncertain
of? They make guesses.

Modern machine perception algorithms (for object detection, pose
estimation, or semantic scene understanding), despite dealing with
tremendous amounts of ambiguity, do not.

In this talk, I will describe a line of work in my lab where we have been
developing machine perception models that output not just a single-best
solution, rather a /diverse/ set of plausible guesses. I will discuss
inference in graphical models, connections to submodular maximization over
a “doubly-exponential” space, and how/why this achieves state-of-art
performance on challenging Pascal VOC 2012 segmentation dataset.


Leave a Reply