(→All Colloquia at the Department)
Topic: |
|
Date: |
06.10.08 |
Time: |
10:15 |
Place: |
Ü2, LMS2 (Übungsraum Ü2 des Instituts für Informatik (Vorbau Ludewig-Meyn-Str. 2)) |
Guest: |
Assistent Prof. Jan-Michael Frahm |
|
University of North Carolina, Chapel Hill |
Invited by: |
Prof. Dr.-Ing. Reinhard Koch
|
Abstract: |
Abstract:
With the recent availability of large collections of images and videos
through websites like Flickr and You Tube researchers have started to
use the data for 3D reconstruction. The talk will discuss our recent
approaches to tackle the challenges in this research area like our
real-time scene reconstruction system, efficient 3D reconstruction form
large-scale image collections and our recently proposed location
recognition from images using local geometry. Our real-time scene
reconstruction system uses the video of a moving (un-)calibrated camera.
To enable image based path intersection detection we developed an image
based location recognition, which will be introduced during the talk. It
is based on the viewpoint invariant patches (VIP). The location
recognition combines image based scene recognition with local geometry
to enable a drift free construction from several hundred thousands of
images. |
Topic: |
How and Why to Prove W-hardness |
Date: |
03.11.08 |
Time: |
17:15 |
Place: |
Ü2, LMS2 (Übungsraum Ü2 des Instituts für Informatik (Vorbau Ludewig-Meyn-Str. 2)) |
Guest: |
Prof. Dr. Mike Fellows |
|
University of Newcastle |
Invited by: |
Prof. Dr. Klaus Jansen
|
Abstract: |
Abstract: Parameterized complexity takes into account
secondary measurements about problem inputs or objectives,
codified as a parameter k, in addition to the overall input size n.
It is thus a kind of fine-grained, multivariate sequel to P vs NP.
Central to studying complexity issues in this two-dimensional framework
are methods for showing W-hardness (the parameterized analog of
NP-hardness). The talk will describe some new tricks for constructing
such proofs, and why such results are interesting.
|
