CS8113g: Computational Perception
Description
Fall Quarter 1997
4:35pm - 5:55pm, Monday,& Wednesday.
CoC Room # 102
Instructor: Irfan Essa (irfan@cc.gatech.edu,
www.cc.gatech.edu/~irfan, CoC 320)
Contents of the Course:
This course will introduce various concepts related to machine perception.
We will undertake an in depth study of existing methods and will also work
on projects in teams of 3 to develop a system with perceptual abilities.
The main aim is to introduce you to computational perception both on a
theoretical and practical levels. To achieve this we will study
-
what sensory signals can be captured from a scene or an environment?
-
this year we will mostly concentrate on video and audio signals, bu we
will look into tactile sensors too.
-
how can these signals be classified and categorized (ie. signal
understanding)?
-
we will study a few pattern recognition techniques, leading upto HMMs.
-
what learning methods would be useful?
-
we will undertake only a brief review.
-
what are different types of analysis and interpretations?
-
issues of domains and context,
-
issues of static vs. dynamic analysis.
-
what systems can we build?
-
system integration
-
multi-modality (combine sesnory information).
-
can we do this without language, discourse ?
-
build a prototype system.
Typical Projects:
Here is a list of typical projects that would be ideal for students to
undertake in groups of 3-4. I will help assign group members based
on your interests and expertise. Preferences from students as to
what projects they would like to work on will be considered. For
all of these projects I will provide additional information and also examples.
-
Analyze a video of a Professor teaching a class. Correlate the various
visual cues and audio cues. Can you teach your system to detect when
the professor is about to make a very important point?
-
Analyze a video of a stand-up comic. Build a joke detector.
-
Recognize America Sign Language (ASL) from video.
-
Develop a "smart" kiosk, that recognizes you as you walk up to it
and interacts with you.
-
Build a home messaging system that is hooked up to a fridge. It recognizes
you and gives you video messages.
-
Develop a system to track you in the room and provide you with cues pertaining
to where you are in the room (using both audio and video)
-
Develop a system that recognize and track Humphery Bogart in the movie
Casablanca (using both audio and video)
-
etc.
Projects are not limited to this list at all. If you would like to
suggest another project, please let me know. During this term we
will study the concepts that would let us achieve a good prototype system
to achive the above.
Prerequisites:
-
Working knowldge of Linear Algebra, Calculas and Probabilty.
-
Programming ability.
-
Interested in working on projects in teams.
-
Persmission of Instructor.
-
(Undergraduates, Non-CS, Non-Engineering majors encouraged, please come
talk to me).
Evlaluation
The grading in this class will be based on
-
Class participation/discussion (24%)
-
Commentary/Critique of assigned readings (16%)
-
Final Project (60%), [Proposal: 10%, 2 Preliminary Reports: 10% each, Final
Report & Presentation: 30%]
Readings/Text
No Textbook for this class. I will hand out readings in class.
CS8113g
CoC
GT
