ECE 590: Graduate Seminar Course in Computer Vision

Fridays, 2-3pm, 3169 Beckman Institute

Web site: http://vision.ai.uiuc.edu/~sintod/ECE590.html

Course Director: Professor Narendra Ahuja (n-ahuja at uiuc edu)
Course Coordinators: Dr. Silvio Savarese (silvio at uiuc edu) and Dr. Sinisa Todorovic (sintod at uiuc edu)

Course Description: In recent years, important breakthroughs have been made in computer vision. New vision problems have been addressed, and novel solutions to challenging classical problems have been proposed. These accomplishments have been made feasible by recent advances in statistical image modeling, linear-algebra techniques, graph/information-theoretic methods, and other methodologies. We will be reading an eclectic mix of classic and recent papers with the goal to become more familiar with some of these new techniques and methodologies. In particular, the course will survey recent developments in, but not limited to, the following topics: 1) Dimensionality reduction (PCA, ICA, LDA); 2) Statistical graphical models (MRFs, CRFs, generative models); 3) Discriminative learning and classifiers (AdaBoost, SVMs, Neural nets); 4) Graph-theoretic approaches in vision (Graph matching, Spectral clustering); 5) Feature extraction and selection (Filtering, SIFT, FOA, segmentation); 6) 3D structure extraction and representation.

See the guidelines how to prepare a good presentation!


Date
 Topics: Papers
 Presenter(s)
 Talks
02/23
 Dimensionality reduction:
- H. Wang and N. Ahuja, "Rank-R Approximation of Tensors Using Image-as-Matrix Representation", CVPR 2005
 Bernard Ghanem
 Bernard's talk
03/02
 Graph-theoretic approaches in computer vision:
- A. Shokoufandeh, D. Macrini, S. Dicikinson, K. Siddiqi, and S. Zucker, "Indexing hierarchical structures using graph spectra", PAMI 2005
 Varsha Hedau
 Varsha's talk
03/09
 3D structure extraction and representation:
- F. Rothganger, S. Lazebnik, C. Schmid, and J. Ponce, "3D object modeling and recognition from photographs and image sequences", LNCS 4170, 2006
- M. Brown and D. G. Lowe, "Unsupervised 3D Object Recognition and Reconstruction in Unordered Datasets", 3DIM 2005
 Zachariah Gossman
 Zack's talk
03/16
 AdaBoost:
-  P. Viola and M. J. Jones, "Robust Real-Time Face Detection", IJCV 2004
- A. Torralba, K. P. Murphy and W. T. Freeman, "Sharing visual features for multiclass and multiview object detection", PAMI 2007
 Emre Akbas
 Emre's talk
03/23
 Spring Break

03/30
 Dimensionality reduction:
- V. de Silva and J. B. Tenenbaum, "Global versus local methods in nonlinear dimensionality reduction", NIPS 2003
- L. K. Saul and S. T. Roweis, "Think globally, fit locally: unsupervised learning of low dimensional manifolds", JMLR 2003
 Sanketh Shetty
 Sanketh's talk
04/06
 Feature extraction and selection:
- K. Mikolajczyk, T. Tuytelaars, C. Schmid, A. Zisserman, J. Matas, F. Schaffalitzky, T. Kadir and L. Van Gool, "A comparison of affine region detectors", IJCV 2005
- E. N. Mortensen, D. Hongli, and L. Shapiro, "A SIFT descriptor with global context", CVPR 2005
 Esther Resendiz,

Myra Nam

 Myra's talk
04/13
 No class

04/20 CRFs and DRFs:
- S. Kumar and M. Hebert, "Discriminative random fields", IJCV 2006
- A. Quattoni, M. Collins, T. Darrell, "Conditional random fields for object recognition", NIPS 2004

Spatiotemporal feature extraction:
- A. Oikonomopoulos, I. Patras, and M. Pantic, "Spatiotemporal salient points for visual recognition of human actions", IEEE Trans. Systems, Man, and Cybernetics - Part B: Cybernetics 2006
- P. Dollar, V. Rabaud, G. Cottrell, and S. Belongie, "Behavior recognition via sparse spatio-temporal features", S-PETS 2005
- I. Laptev and T. Lindeberg, "Space-time interest points", ICCV 2003
 Mert Dikmen,

Ousman Azy,

Juan Carlos
 Juan's talk

Ousman's talk
04/27
 Statistical generative models:
- L. Fei-Fei and P. Perona, "A Bayesian hierarchical model for learning natural scene categories", CVPR 2005
 Gang Wang Gang's talk
05/04
 Graph cuts:
- Y. Boykov and M.P. Jolly," Interactive graph cuts for optimal boundary and region segmentation of objects in N-D images", ICCV 2001
- H. Lombaert, Y. Sun, L. Grady and C. Xu," A multilevel banded graph cuts method for fast image segmentation", ICCV 2005 		Theodore Ha


Guidelines for Presenters:
- The talk should not be longer than 40min. This will allow for the discussion during and after the presentation, which may take additional 20min.
- Given that on average one spends 2min per slide, the presentation should not have more than 25 slides.
- The presentation should include the following segments:
  1. Motivation: What vision problems have been addressed by this methodology? (5min)
  2. Review of the theoretical background: For example, the presentation about CRFs should cover a brief review on graphical models, and point out the advantages/disadvantages of CRFs as compared to other modeling paradigms. (5min)
  3. High-level tutorial on the specific topic: Give insights and main ideas, and try to avoid slides with a lot of equations. (15min)
  4. Specific vision problem that has been addressed by this methodology: That is, what the paper you selected to present is all about. (10min)
  5. Discussion: Is this a promising framework for addressing vision problems? List strengths and deficiencies. What are the latest trends regarding this methodology? (5min)