Computer Vision Seminar
Instructors: Rama Chellappa, Tapas Kanungo and Azriel Rosenfeld
Presentation Schedule and Slides
The general theme for this semester will be representations for
computer vision. This includes representations of
camera and imaging geometry, lighting, motion and action, texture,
curves and objects, uncertainty and noise, etc. Emphasis will be on
understanding the mathematical details of the representations
and methods for deriving those representations from noisy image data.
Students are responsible for doing the following things.
- Select one or a few articles from
conference proceedings or chapter(s) from books. Discuss
your choices with one of the course coordinators. Ideally
you should make your articles available for copying at least one day
prior to the previous class period, so that copies can be distributed
- Give a one hour presentation discussing the details of algorithms.
The first 15 minutes should be spent on the background and
introduction, and the rest of the time should be spent on technical
details and methods.
- Lead a discussion on the paper presented: What was the key problem?
What representation was used? What is the mathematical theory underlying
the representation? How is the representation computationally derived
from data? How do you represent the uncertainty in the derived
representation? How stable is the representation?
Are there other ways of addressing this problem? Are there faster
methods? What are the strengths and weaknesses of the representation?
What other improvements or tradeoffs are there?
- Submit a summary report of roughly four pages with a bibliography
citing related research. The report should summarize the paper and
the discussion. The document should be in Latex and the bibliography
should be in Bibtex. We plan on combining these summaries at the
end of the semester to form a survey of the area and a bibliography.
Samples will be handed out to help you in preparing these.
The following list is intended to serve as a starting point for ideas
for topics; feel free to suggest others that you think may be
interesting and appropriate:
- Light: Color spaces, surface reflectance models
- Camera: Spherical and parabolic cameras; etc.
- Action: Hidden Markov models (HMMs) and
HMM-based representations of action;
grammatical models; declarative models; Kalman filtering approaches
- Geometry: Quaternions, projective geometry, invariants
- 3D Shape: Generalized cylinders, superquadrics, surface meshes,
- 2D Shape: Moments, Fourier coefficients, skeletal representations,
splines, non-uniform rational b-splines (NURBS), deformable models
- Texture: Markov random field models for texture;
Participants will be provided with pointers to relevant conference proceedings
and books. Possible sources are:
- Proceedings of the IEEE Conference on Computer Vision
and Pattern Recognition
- Proceedings of the IEEE International Conference on Computer Vision
- Proceedings of the DARPA Image Understanding Workshop
- Proceedings of the IEEE International Conference on Image Processing
- Proceedings of the IAPR International Conference on Pattern Recognition
Grades will be based on the presentation and class participation.
This is a 2-credit seminar. However, if a student is interested in
doing a project he/she can earn 3 credits. Please contact one of
the coordinators to discuss possible projects.
Any computer vision, image processing, computer graphics, or
statistical pattern recognition class or permission of the instructor.
Meeting Time and Place
|Day ||Time ||Room |
|Tuesday || 3:30- 5:30 || AVW 4424 |