I will talk about work I did last semester on model-based vision, the
application being car-tracking for the Navlab autonomous vehicles. If
I'm lucky, and there is time left, I will also sneak-preview a new
texture mapping-based tracking technique I am currently developing.
Both subjects do not involve learning per se, but combine 3D imaging
with optimal estimation (Bayes law, right !?), and so I think will be
of interest to the usual RL bunch.

(BTW. This will be more or less the same talk that I gave at the VASC
seminar, so if you saw that, you probably don't want to hear it all
again..)

If you're interested in knowing more, here is the abstract of one
paper I wrote with Dean and Chuck on this stuff. This paper and an
earlier, bigger paper can be downloaded from:
http://www.cs.cmu.edu/~dellaert/publications.html