Contact
Info
Email: bpotetz @
cs.cmu.edu
Office: Wean Hall 3708
Tel:
(412) 268-8184
Advisor: Tai Sing Lee
Mail:
Computer Science Department
Carnegie Mellon University
5000 Forbes Avenue
Pittsburgh, PA 15206 USA
|
|
|
|
|
|
|
In 2002, Tai Sing
and I
rented an LMS-Z360 range scanner
with integrated color photosensor, and acquired about a hundred scans
of
various outdoor natural scenes. Aside from several exciting research
applications, the database also offers some nifty eyecandy
opportunities. Here’s
a movie I made to demonstrate the database. This was generated from a
single
scan. Link goes to video.
|
|
|
|
Our
range camera also had
the ability to take 360 degree
panoramic shots. Here is a quicktime VR of a garden outside Phipps
conservatory, which is adjacent to CMU campus. Note that this image has
nothing to do
with the range-finding component of the database, which is what cost us
thousands of dollars. If you've got Quicktime installed, just drag the
image around to see more.
|
|
|
|
|
|
|
|
Snake is a simple
matlab implementation of
the A* search
approach to active contour finding. I wrote this for Tai Sing to use as
a
teaching tool in his computer vision course (15-385). This is loosely
based on
a tool I wrote to help the US National Ice Center to trace icebergs
back when I
was working at Sensor Systems.
This matlab program help you to digitize (vectorize) edges in images,
while providing lots of behind-the-scene features to help computer
vision students learn how the algorithm works. [screenshot, program, docs]
|
|
|
|
At one point, while
working on a grant proposal, Tai
Sing asked me if Ramachandran’s famous contour-dependent strips
could be
parameterized into a family of strips that smoothly warp from one shape
to the
other while maintaining an invariant intensity profile. Here’s what I
came up
with. Anyway, we never used it, so I'm posting it here for you to do
whatever you want with. Decorate your webpage or something.
The idea here is that the strip is changing shape, but the light source
is moving at the same time to cause the intensity at any interiorpoint
to remain the same. The strip is generated using Lambertian shading (I
didn't just fake the result by guessing some reasonable contour shapes).
|
|
|
|
|
|
|
|
My first year at
CMU, I decided that
animated gifs would be
a good way to communicate high dimensional patterns in natural scene
statistics
with my advisor. But explaining the gifs was harder than explaining the
original
concept. These obscure animations were all there was to my homepage for
years.
But some people like them, so here they are. |
|
|
|
