Scribd
Upload
127 views11 pages

Hack, Math and Stat

This document discusses the need to reconcile extreme views between those focused on hacks, statistics, and mathematics in computer vision. It provides examples of how hacks, statistics, and mathematics have contributed to developments in areas like texture modeling, optimization algorithms, and historic trends. The document also notes that the field has become polarized and imbalanced, with hacks dominating at the expense of original work on statistics and mathematics. This risks impacting student training. Quotes from reviewers demonstrate polarized views between those only interested in performance and those rejecting any work not providing certainty or solving entirely unsolved problems.

Uploaded by

davidrojasv
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
127 views11 pages

Hack, Math and Stat

This document discusses the need to reconcile extreme views between those focused on hacks, statistics, and mathematics in computer vision. It provides examples of how hacks, statistics, and mathematics have contributed to developments in areas like texture modeling, optimization algorithms, and historic trends. The document also notes that the field has become polarized and imbalanced, with hacks dominating at the expense of original work on statistics and mathematics. This risks impacting student training. Quotes from reviewers demonstrate polarized views between those only interested in performance and those rejecting any work not providing certainty or solving entirely unsolved problems.

Uploaded by

davidrojasv
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 11

Hack, Math and Stat

--- call for a reconciliation between extreme views


Song-Chun Zhu
The Frontiers of Vision Workshop, August 20-23, 2011.
Math: under certain conditions, things can be said analytically.
Hack, Math and Stat: co-existing ingredients
Stat: is, essentially, regression, fitting by weighted features.
Hack: something, somehow, works better somewhere.
Hack
Stat

Math


breadth
depth
* Berry 1898, A Short History of Astronomy, p.240. ** Stigler 1986,The History of Statistics, p.26.
Hack: Celestial navigation
For example, when people looked at the sky
Math: Newtonian Gravitational Theory. 1680s
Newton reportedly told Halley that lunar theory
(to represent mathematically the motion of moon )
made his head ache and kept him awake so often that he would think of it no more.*
2
2 1
r
m m
G F
e.g. the Chinese expedition in Ming Dynasty, 1405-1433.
Stat: Rescue the solar system by lsq (Mayer, Legendre, Laplace, Euler)
n
1
75
2
.
.
.
x
1
, x
2
, , x
k
y
2 sin " 32 2 sin " 168 sin " 23525 q q
" ) 2 sin( " 243 ) sin( " 257 m q q
) sin( " 2 sin " sin " p z q y q x
" ) cos( ) 360 ( Nn p p v u
q k q 2 cos " 600 / 1 cos " 11405 . 0
** Euler 1749
Math: --- 3D geometry, lighting models, PDEs,
Pattern Theory.
When people look at images
Stat: --- Boosting, Support vector machines
(Regression with various loss functions and algorithms.)
Hack: --- good features (corner, SIFT, HoG, LBP, ),
algorithms that are not supposed to apply
(such as Loopy Belief Propagation) BTW, true for most vision algorithms,
it is unclear what state spaces they are operating on, and whether or how it may converge.

What makes research exciting are the discoveries that connection them !
Julesz 1960-70s
Daugman1985
Bergen, Adelson
Heeger 1991, 95
Julesz ensemble, 1999
Gibbs 1902
equivalence, 1999
Lewis etc. 95
Case study I: Development of MRF/texture models
Besag 1973
Ising 1920
Potts 1957
Cross and
Jain 1983
FRAME, 1996
Case study II: Dev. MCMC/optimizing algorithms
Metropolis 1946
Hastings 1970
Waltz 1972 (labeling)
Rosenfeld, Hummel, Zucker 1976
(relaxation-labeling)
Geman brothers 1984, (Gibbs sampler)
Swendsen-Wang 1987 (cluster sampling)
Jump-diffusion,
Miller & Grenander,1994
Kirkpatrick, 1983
Reversible jump, Green 1995
Swendsen-Wang Cut 2003
DDMCMC 2001-2005
C4: Clustering w. +/- Constraints, 2009
Historic trends in vision, based on my perception


Hack Stat Math


Hack Stat Math
Before 1990
1990-2000


Hack Stat Math
2000-2005
The field is out of balance and polarized


Hack Stat Math
2006-now
Not only that hacks have become dominating, but original work
on stat and math have been diminishing. A longer term problem
is its impacts on student training.
Vision = Classification + e
= Datasets + Features + Classifiers + Curves + e
Quotes from review comments
From the far-right wing:

Why not you compare against X1, X2, X3 on datasets Y1, Y2, Y3 ???
Mathematically rigorous methods usually do not work !
I have no interest reading your theory until you show me performance
better than the state-of-the-art.
If you are smart, why arent you rich?
Quotes from review comments
From the far-left wing.

Your method is trained and tested only on a finite, artificial dataset,
therefore it does not contribute to the solution of any real vision problem
I am talking with mathematical certainty that it is impossible to reconstruct
a 3D scene from a single image. So this paper must be rejected.
With so many sub-problems unsolved, it makes no sense to solve a problem
as complex as image parsing.
Professor, your question was wrong
Finally: How to reach the moon, in 20 years?

You might also like