Seminar Home
Fall 2012 Schedule
Winter 2013 Schedule
Summer 2013 Schedule
Archives

|
Date (
Summer 2008
) |
Category |
Seminar Info |
| 2007/08/31 |
Vision, Graphics, and Robotics |
Place: MC437
Time: 11:00 - 12:00
Speaker:
Marc
Pollefeys
Affiliation: ETH Zurich and UNC Chapel Hill
Area:
Computer Vision
Title: Towards capturing dynamic 3D events from video
Abstract: Images and videos form a rich source of information about the visual world. The extraction of 3D information from images is an important research problem in computer vision and graphics. The ubiquitous presence of cameras and the tremendous advances of processing and communication technologies yields important opportunities and challenges in those areas.
This talk focuses on approaches to obtain 3D reconstruction of dynamic scenes. We present two approaches. The first approach uses Bayesian inference to recover dynamic shapes from silhouettes seen from multiple viewpoints, even in the presence of partial occlusion. Our approach not only recovers the dynamic shape, but over time is also able to infer the shape of the occluders present in the scene.
Next we present our approach to analyze and recover articulated motion with non-rigid parts, e.g. the human body motion with non-rigid facial motion, from a single viewpoint under affine projection from feature trajectories. We model the motion using a set of intersecting linear subspaces. Based on this model, we can analyze and recover the articulated motion using subspace methods. Our framework consists of motion segmentation, kinematic chain building, and shape recovery. We test our approach through experiments and demonstrate its potential to recover articulated structure with non-rigid parts via a single-view camera without prior knowledge of its kinematic structure.
Biography of Speaker:
-
Marc Pollefeys is a full professor in the Dept. of Computer Science at ETH Zürich since 2007. He currently also remains associated with the Dept. of Computer Science of the University of North Carolina at Chapel Hill where he started as an assistant professor in 2002 and became an associate professor in 2005. Before he was a postdoctoral researcher at the Katholieke Universiteit Leuven in Belgium, where he also received his M.S. and Ph.D. degrees in 1994 and 1999, respectively. His main area of research is computer vision. One of his main research goals is to develop flexible approaches to capture visual representations of real world objects, scenes and events. Dr. Pollefeys has received several prizes for his research, including a Marr prize, an NSF CAREER award and a Packard Fellowship. He is the author or co-author of more than 90 peer-reviewed or invited papers. He was co-chair of the Third Symposium on 3D Data Processing, Visualization and Transmission and has organized workshops and courses at major vision and graphics conferences and has served on the program committees of many conferences. He is a regular reviewer for most of the major vision, graphics and photogrammetry journals. Prof. Pollefeys is on the Editorial Board of the IEEE Transactions on Pattern Analysis and Machine Intelligence, the International Journal of Computer Vision and Foundations and Trends in Computer Graphics and Computer Vision.
|
| 2007/06/08 |
Software Engineering |
Place: McConnell 103
Time: 11:00 - 12:00
Speaker:
Barthelemy
Dagenais
Affiliation: McGill University
Title: ICSE 2007 Conference Report
Abstract: During this seminar, we will present an overview of the ICSE 2007
conference and review in more details the following research projects/tracks
: the future of software engineering, the SIGSOFT project assessing the
impact of soft. eng. research on soft. eng. practice, an API usability study
and the use of program analysis to trace back exception chains in java web
applications.
|
| 2007/06/01 |
Software Engineering |
Place: McConnell 103
Time: 11:00 - 12:00
Speaker:
Eric
Bodden
Affiliation: McGill University, Sable group
Title: A staged static program analysis to improve the performance of runtime monitoring
Abstract: In runtime monitoring, a programmer specifies a piece of code to
execute when a trace of events occurs during program execution. Our
work is based on tracematches, an extension to AspectJ, which allows
programmers to specify traces via regular expressions with free
variables. In this paper we present a staged static analysis which
speeds up trace matching by reducing the required runtime
instrumentation.
The first stage is a simple analysis that rules out entire
tracematches, just based on the names of symbols. In the second stage,
a points-to analysis is used, along with a flow-insensitive analysis
that eliminates instrumentation points with inconsistent variable
bindings. In the third stage the points-to analysis is combined with a
flow-sensitive analysis that also takes into consideration the order
in which the symbols may execute.
To examine the effectiveness of each stage, we experimented with a set
of nine tracematches applied to the DaCapo benchmark suite. We found
that about 25% of the tracematch/benchmark combinations had
instrumentation overheads greater than 10%. In these cases the first
two stages work well for certain classes of tracematches, often
leading to significant performance improvements. Somewhat
surprisingly, we found the third, flow-sensitive, stage did not add
any improvements.
|
| 2007/05/25 |
Software Engineering |
Place: McConnell 103
Time: 11:00 - 12:00
Speaker:
Chris
Pickett
Affiliation: McGill University
Title: libspmt: A Library for Speculative Multithreading
Abstract: Speculative multithreading (SpMT) is a dynamic parallelisation technique
that uses out-of-order execution and memory buffering to achieve speedup
on multiprocessors. We present the design of a Java-specific, software
SpMT system that operates at the bytecode level, and fully addresses the
problems and requirements imposed by the Java language and VM
environment. Our experimental results show that while we can extract
significant parallelism, this comes at the cost of considerable overhead.
Our implementation is built in SableVM, a Java interpreter, and to
experiment with just-in-time (JIT) compilation, we are working to
enable similar functionality in IBM's J9 VM and Testarossa JIT. In
order to promote code reuse between these two disparate environments,
we have created libspmt, a new library for speculative multithreading.
libspmt provides a virtualization of speculative execution components
that are unavailable in commercial multiprocessors, and is designed to
support method level speculation in its clients or hosts when fully
integrated. It can accommodate hosts that implement garbage
collection, exceptions, and non-speculative multithreading. We have
isolated a reasonably minimal library interface, and the corresponding
implementation is modular. The library was developed by refactoring
our SpMT implementation in SableVM, and we found that by focusing on
modularity we have made some complex optimizations more feasible.
|
| 2007/05/04 |
Faculty Candidate Talk |
Place: MC437
Time: 9:30 - 10:30
Speaker:
Vida
Dujmovic
Affiliation: Department of Mathematics and Statistics, McGill University
Area:
Graph partitions
Title: Geometric Thickness Parameters of Graphs
Abstract: Partitions of the edge set of a graph into a small number of `nice'
subgraphs are in the mainstream of graph theory. For example, in a proper
edge colouring, the subgraphs of the partition are matching. If each
subgraph of a partition is required to be planar (respectively, a forest),
then the minimum number of subgraphs in a partition of a graph G is the
_thickness_ (_arboricity_) of G. These are classical graph parameters that
have been studied since the early 1960s.
We study graph partitions with additional geometric properties. Namely,
there is a drawing of the graph, and each subgraph in the partition is drawn
without crossings. This type of drawing has applications in graph
visualization (where each noncrossing subgraph is coloured by a distinct
colour), and in multilayer VLSI (where each noncrossing subgraph corresponds
to a set of wires that can be routed without crossings in a single layer).
With no restriction on how the edges are drawn, the minimum number of
noncrossing subgraphs, taken over all drawings of G, is again the thickness
of G. By restricting the edges to be drawn straight, we obtain the
_geometric thickness_ of G. By further restricting the vertices to be in
convex position, we obtain _book thickness_ of G.
We study these (and other) geometric graph parameters for several graph
classes, including graphs of treewidth k. Treewidth, is a modern graph
parameter that is particularly important in structural and algorithmic graph
theory. Our results settle couple of open problems.
|
|