home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!concert!borg!marshall!marshall
- From: marshall@marshall.cs.unc.edu (Jonathan A. Marshall)
- Newsgroups: triangle.general
- Subject: Spring Course: Neural Networks and Vision
- Message-ID: <17415@borg.cs.unc.edu>
- Date: 11 Nov 92 16:36:06 GMT
- Expires: 13 Jan 93 05:00:00 GMT
- Sender: news@cs.unc.edu
- Followup-To: triangle.general
- Distribution: nc
- Lines: 80
-
- ------------------------------------------------------------------------------
-
- Course Announcement - January 1993 - Please Post
-
- Graduate Seminar on
-
-
- NEURAL NETWORKS AND VISION:
-
- Nonlocal Cortical Interactions, Filling-in, and Object Perception
-
-
- Prof. Jonathan Marshall
- Department of Computer Science, UNC-Chapel Hill
-
- COMP 290 (section 050), 3.0 credits, Registration call number: 21617
- [Note: this course will be offered even though
- it does not appear in the directory of classes.]
-
- TuTh 9:30-10:45, Spring 1993; starts January 12
- Sitterson Hall, room 325
-
- This graduate research seminar will focus on modeling and understanding
- nonlocal interactions in visual cortex and in visual perception. Nonlocal
- interactions include such phenomena as:
-
- o "filling-in" or diffusion of color, motion, depth, and surface
- information across wide regions of an image;
- o visual transparency effects, where more than one surface can be seen
- within an image region;
- o "amodal" representations of temporarily occluded visual objects;
- o grouping operations (establishing linkages among perceptually
- related data);
- o scission operations (breaking linkages between unrelated data);
- o steering operations, whereby visible portions of a visual group or
- object can control the representations of invisible or occluded
- portions of the same group.
- o development of long-range connections across visual cortex, whereby
- data about one part of an image can become able to influence the
- processing or perception of data about some distant parts;
- o development of specific dissociation of neural connections, whereby
- some neurons representing nearby parts of an image can become unable
- to influence one another strongly, despite their proximity.
-
- Introductory information will also be provided on neural networks and on
- vision. The seminar will cover selected papers from the literature on neural
- networks and vision. Introductory topics include:
-
- o basic architecture of animal visual systems;
- o basic introduction to neural networks;
- o how to read papers in neurophysiology, psychophysics, and neural
- networks;
- o computational aspects of vision;
- o computational behavior of neural networks;
- o edge detection, orientation detection;
- o perception of visual motion, depth, transparency, and occlusion;
- o visual segmentation, boundaries, and grouping;
- o learning, adaptation, and self-organization.
-
- The field of neural network modeling is concerned with both how neurons in
- animal brains communicate and perform tasks for perception, cognition, and
- motor control, and how networks of artificial neuron-like processing elements
- can solve computational problems. Neural network research is thus highly
- interdisciplinary: computational theory helps explain and predict findings in
- neurobiology and psychophysics, while neurobiological and psychological
- results suggest new methods for solving computational problems.
-
- The course workload will consist of readings, class presentation(s), class
- participation, and a research project or paper. Prior exposure to material on
- neural networks or vision is not a prerequisite, but would likely be helpful.
- Graduate students in other departments, as well as those in computer science,
- are encouraged to take the course. This course is open to students who have
- not studied neural networks or vision, and it is also open to students who
- previously took my Neural Networks and Vision course (Spring 1991), since the
- material covered will be about 75% new.
-
- For more information, contact Prof. Jonathan Marshall at (919) 962-1887
- (marshall@cs.unc.edu).
-
- ------------------------------------------------------------------------------
-