2017 Instructor: Vlad Wojcik
A branch of Machine Learning, Computer Vision aims at equipping computers with animal skills of vision, i.e. the ability to interpret perceived images in term of objects and their spatial and temporal relationships. This is a tall order, as full operational details of the animalian visual cortex remain poorly understood. Vigorous research continues in both directions.
Does it all seem like we are trying to teach a computer something we do not know well ourselves? You are right. Worse still: We already know that your brain can see things your eyes can't. Unbelievable? Check this out! Now, get this: overrules what your eyes see ... Check this girl out: Does she stand on her left or right leg? How does she rotate? Left or right? Are you sure ??
Sometimes the damage to a part of visual cortex leads to total or partial blindness; at other times some cognitive disorders emerge. Go figure!
COURSE TOPICS COVERED:
- Basic optics(a quick and dirty review)
- Image formation and image models
Issues: Cameras, camera models, camera calibration, radiometry, light sources, shadows, shadings, colour.
- Early vision: Just one image
Issues: Linear filters, edge detection, texture.
- Early vision: Multiple images
Issues: Geometry of multiple views, stereopsis, affine and projective structures from motion.
- Mid-level vision
Issues: Segmentation by clustering, by fitting a model. Segmentation and fitting using probabilistic methods. Tracking with linear dynamic models.
- High-level vision: Geometric methods
Issues: Model-based vision, smooth surfaces and their outlines, aspect graphs, range data.
- High-level vision: Probabilistic and inferential methods
Issues: Finding templates using classifiers, recognition by relations between templates, geometric templates from spatial relations.
Issues: Searching image libraries, image-based rendering, etc.
- D.A. Forsyth, J. Ponce: Computer Vision: A Modern Approach, 2nd ed., Prentice-Hall 2003, ISBN 9780136085928.
- R.O. Duda, P.E. Hart, D.G. Stork: Pattern Classification, 2nd ed., J. Wiley & Sons 2001, ISBN 0-471-05669-3.
- Mammalian Sensory Systems.
- Image Alignment (i.e. Registration).
- Issues in Scene / Image Segmentation.
- The Hyperball Concept plus FAQ.
- Audio Pattern Recognition and Learning.
- Visual Pattern Recognition and Learning.
- Computer Vision On-Line: The Evolving, Distributed, Non-Proprietary Compendium.
- Computer Vision Handbook.
- D.H. Ballard, C.M. Brown: Computer Vision.
- Intel Corp.: Open Source Computer Vision Library.
- Partial list of Computer Vision Industry.
- V. Wojcik, Recursive Problem of Calibration of Expert Systems, January 1998.
- V. Wojcik, The Structure of the Brain and of the Adaptive Computer Architectures, January 1998.
- Two assignments @ 15% each
- One mid-term test @ 30%
- One two-person project @ 40%
In case a given mark is perceived unjust or unclear by a student, s/he is encouraged to see the instructor to discuss the issue. Depending on the case s/he is able to make, a mark can be modified. The deadline to contact the instructor on these matters is one week after the mark has been issued. Marks not disputed within this period will be considered final.
Possible lateness in assignment submission is counted in days, each period of a day ending at 4 PM. The penalty for late submission of assignments is 25% up to three days (or a part of a day). After that period the penalty is 100%.
While honest cooperation between students is considered appropriate, the Department considers plagiarism a serious offense. For clarification on these issues you are directed to the statement of Departmental Policies and Procedures.
Instructor: Vlad Wojcik
Revised: Friday, 10-Feb-2017 11:05 PM
Copyright © 2017 Vlad Wojcik