|
|
MIT Electrical Engineering and Computer Science
Fall 2000 Catalogue Supplement |
TR 2:30-4, Room 36-839
Prof. Trevor Darrell, Room NE43-829, x3-8966
Prereq.: Permission of instructor
3-0-9
Qualifies as a subject in the Artificial Intelligence Engineering Concentration
The internet will soon have eyes -- computer vision systems that can detect,
track and recognize people and other objects. These systems will enable
new perceptual interfaces between man and machine, including smart videoconferencing,
expressive avatars, and rooms that recognize users and their gestures.
They will allow the widespread tracking of people in outdoor spaces, with
clear implications to notions of community, public safety, and privacy.
This class will survey the algorithms and techniques involved in vision-based
perception of people, and analyze the privacy, freedom and safety implications
of this new technology. We will focus on the questions of whether these
goals must be mutually exclusive and under what conditions this technology
empowers or constrains the individual user.
I. Identity: automatic identification
methods; face detection and recognition
Biometrics: iris, fingerprint,
skeletal
Large-scale database indexing
and search
Issues of anonymity
II. Interface: from parametric control from dialog
Automatic camera positioning
Expression analysis
Avatars
Body tracking: silhouette, background
detection, range and motion techniques
Kiosks and conversational agents
Kinematic estimation primer
III. Intelligent environments: person tracking from multiple
viewpoints
Multi-modal context integration
Distributed agent systems and
standards (Metaglue, SRI-OAA, etc)
IV. Surveillance web and CCTV security cameras
Activity classification and recognition
Statistical assessment and applications:
retail, architecture, service
Vitrual tourism
Privacy vs security