CS 600.441

             VISION-BASED INTERACTION FOR MAN AND MACHINE

Gregory D. Hager
Th 4:15-5:30, F 2:30-3:45
 Room 302 Shaffer Hall

 
 

Overview:

    This seminar course surveys recent research results on algorithms for dynamic vision and their application. Specific emphasis will be placed on approaches which derive novel and efficient algorithms using generalizable mathmetical and/or computational principles. Application papers will cover problems ranging from the automated control of devices using vision (e.g. automated driving) to perception-based user interfaces. Students will be expected to participate in class by reading, presenting, and discussing research papers.  Grading will be based on participation, and class project.
 

Clientele:

    Graduate students in computer science, electrical engineering, mechanical engineering and related disciplines.  The ideal student should have a background in engineering mathematics, programming experience, and some exposure to signal processing and/or computer vision.
 

Content and Format:

    I will spend the first few lectures introducing the course and covering some basic background material on camera models and imaging. We will then move into some geometrical issues, then into topics involving motion and control using Jacobian-based algorithms.  Those taking the course for credit will be required to write a paper and/or do a project of their choice.  Depending on the size of the class, students may also be asked to present some of the material.
 

Background Reading:

Computer Vision Compendium 
Trucco and Verri: Computer Vision
Ponce and Forsyth: Computer Vision: A Modern Approach
Ballard and Brown: Computer Vision (and older standby)

Related Links

Euron Summer School on Visual Servoing

 

Course Outline

This is the current plan for the course.  Links point to notes or to relevant papers.

Week 1:

Thursday: intro/Kumar lecture in ME
Friday: Imaging geometry and pose estimation

Reading:

Week 2:

Thursday: Projective geometry and calibration cont'd
Friday: CLASS CANCELLED!

Primary Reading:

Secondary Reading:

Week 3:

Thursday: Object pose estimation as an example of point geometry (notes)
Friday: Introduction to Visual servoing; review of Jacobians for point and line motion

Primary Reading:

Secondary Reading:

Nicolas Andreff's notes on Servoing from Lines and a related paper

Week 4:

Thursday: Basic monocular and stereo systems. 
Friday: Task constructions for stereo systems.

Reading:

Week 5:

Thursday: Guest Lecture: Noah Cowan on provably convergent systems (ICRA 1999 paper)
Friday: Finish Stereo 

Primary Reading:

Secondary Reading:

Week 6:

Thursday: 2.5d visual servoing (James Kinsey)
Friday: Panoramic sensor (Darius Burschka)

Primary Reading:

Secondary Reading:

 

BREAK!

Week 7:

Thursday: Virtual fixtures (Nim/Ming)
Friday: Review of basic image processing, optical flow, edge detection, warping

Primary Reading:

Week 8: 

Thursday: Region tracking (Maneesh Dewan)
Friday: More region tracking

Primary Reading:

Week 9:

Thursday: Snakes I (Jacques/Leven)
Friday: Dani Kragic guest lecture --- 3D model-based tracking

Reading:

Week 10:

Thursday: Snakes cleanup/Particle filters (Le/Donald)
Friday: more particle filters (Le/Donald)

http://www.robots.ox.ac.uk/~vdg/
 
http://research.microsoft.com/vision/cambridge/pub.htm
 
P. Pérez   http://research.microsoft.com/users/pperez/

Articulated Body Motion Capture by Annealed Particle Filtering

Jon Deutscher, Andrew Blake and Ian Reid Proc. Conf. Computer Vision and Pattern Recognition (CVPR), 2000
 

A probabilistic exclusion principle for tracking multiple objects

John MacCormick and Andrew Blake Proc Int. Conf. Computer Vision, 572-578, (1999).
 

ICONDENSATION: Unifying low-level and high-level tracking in a stochastic framework

Michael Isard and Andrew Blake Proc 5th European Conf. Computer Vision, vol. 1 893-908, (1998).
 

CONDENSATION -- conditional density propagation for visual tracking

Michael Isard and Andrew Blake Int. J. Computer Vision, 29, 1, 5--28, (1998)

Partitioned sampling, articulated objects, and interface-quality hand tracking

John MacCormick and Michael Isard Proc European Conf. Computer Vision, vol. 2, 3-19 (2000)

Week 11:

Thursday: Mean shift/Jeremy Mullendore
Friday:  Pfinder and Active Blobs

Other Reading

Week 12:

Thursday: HCI with Vision (Corso)
Friday: Temporal modeling (Ye)

Reading:

Secondary Reading:

Week 13

Thursday: People tracking
Friday:  Wrap-up

 

Previous Year's papers

Visual Tracking:

Incremental Focus of Attention for Robust Vision-Based Tracking (with K. Toyama) Accepted to appear in IJCV. (33 pages, 2.7M compressed postscript)
 

The XVision System: A General-Purpose Substrate for Portable Real-Time Vision Applications (with K. Toyama). In Computer Vision and Image Understanding 69(1) pp. 23 - 37. (32 pages, 1.5M compressed postscript).

C.Bregler, J.Malik, Video Motion Capure, UCB//CSD-97-973
Paper in gziped PS (843KB) or PDF (2.3MB)

Vision-Based Control:

What Tasks Can Be Performed with an Uncalibrated Stereo Vision System? (with J. Hespanha, Z.Dodds, and A.S. Morse) To appear, IJCV. (21 pages, 300K compressed postscript)
Read the Abstract.

Selected articles from The Confluence of Vision and Control (D. Kriegman, G.D. Hager and A.S. Morse, Editors) LNCIS series, Springer-Verlag, 1998.

Interaction:

Selected articles from Computer Vision for Human-Machine Interaction