9.357 Current Topics in Vision Science: Things and Stuff
Fall 2012
Perceiving the properties of the materials in this image is vital to interpreting it.
Instructors: adelson@csail.mit.edu, Ruth Rosenholtz, Bei Xiao
Seminars: Thursday 2:30pm-4:30pm (46-4199)
Information:
One can conceptualize visual perception in terms of “things”
or
“stuff.” The distinction comes up everywhere, although it’s
often
neglected. Is motion perception about matching features across
frames
(things), or about sensing motion energy (stuff)? Is a
texture
composed of collection of elements (things), or distributions
of
filter responses (stuff)? The distinction shows up in discussions
of
low level physiology (are V1 cells detecting edges?) or higher
level
cognition (e.g., in the discretized flavor of classic AI versus
the
distributed flavor of neural nets). We think the vision community
is
sometimes led astray by prematurely choosing the wrong
representation,
and then ends up building elaborate models to fix the damage
(e.g.,
feature integration theory in visual search). We’ll survey a range
of
literature, re-evaluating the assumptions and looking for better
ways
to frame the problems and the solutions.
Announcements
Tomorrow (last class): Two speakers : Evan and Leo
Tomorrow is our last class for this semester. We will have two speakers:1. Evan from Pawan's lab will talk about "Early development of face perception and stuff"
He suggested the following readings:
Morton and Johnson (1991) CONSPEC and CONLERN: A Two-Process Theory
of Infant Face Recognition
Bednar and Miikkulainen (2003) Learning Innate Face
Preferences
Sugita (2008) Face perception in monkeys reared with no exposure to
faces
Butko and Movellan (2006) Learning about Humans During the First 6
Minutes of Life
2. Kleovoulos Tsourides will talk about "Dynamic Textures
and Flow Dynamics "
He will discuss:
Chen 2012
http://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=6248218
Vidal CVPR '05
http://cis.jhu.edu/~rvidal/publications/cvpr05-dyntextures.pdf
and then will attempt to contrast the above methods with methods
for action recognition with respect to flow dynamics
such as :
Ali 2010
http://ieeexplore.ieee.org/xpl/articleDetails.jsp?arnumber=4693713
Announced on 12 December 2012 8:31 p.m. by Bei Xiao
Tomorrow: Correction. Owen Lewis, Alternatives to stuff in computer vision?
Sorry. The earlier announcement was for next week's lecture. This week, Owen Lewis is going to talk about Alternatives to stuff in computer visionHere is the reading for tomorrow's class:
http://www.pnas.org/content/108/43/17621.abstract
Announced on 05 December 2012 4:00 p.m. by Bei Xiao
Thursday: Evan Ehrenberg "early development of face perception and stuff"
Evan Ehrenberg will discuss early development of face perception and stuff.Here are the readings:
Morton and Johnson (1991) CONSPEC and CONLERN: A Two-Process
Theory of Infant Face Recognition
Bednar and Miikkulainen (2003) Learning Innate Face
Preferences
Sugita (2008) Face perception in monkeys reared with no exposure to
faces
Butko and Movellan (2006) Learning about Humans During the First 6
Minutes of Life.
He will also discuss some beginning research in the Sinha lab.
Bei
Announced on 05 December 2012 3:39 p.m. by Bei Xiao
Thursday: Max Siegel, Development of knowledge of stuff
This Thursday Max Siegel will talk about Development of knowledge of stuff.Here is the reading:
http://groups.psych.northwestern.edu/infantcognitionlab/HesposvanMarle2011proofs.pdf
See you tomorrow,
Bei
Announced on 28 November 2012 1:58 p.m. by Bei Xiao
Thursday: Crowding paper
Here is the paper for thursday 's discussion for Crowding.http://www.sciencedirect.com/science/article/pii/S1364661311000325
Announced on 13 November 2012 9:30 p.m. by Bei Xiao
