Event Location: NSH 1507
Bio: Please see http://cs.nyu.edu/~gwtaylor/

Abstract: Recent developments in hierarchical feature learning and unsupervised
learning have fueled a resurgence of interest in convolutional networks.
In this talk I will focus on our progress in developing convolutional
architectures and learning algorithms for the analysis of human activity
from images and video. I will describe a method of unsupervised
convolutional low-level feature extraction from video, as well as a form
of convolutional nonlinear embedding for the retrieval of images of
people in similar pose but with different clothes, background, and other
appearance changes. Additionally, I I will briefly survey some of the
work on unsupervised and supervised convolutional learning taking place
at NYU and elsewhere.