Professional Documents
Culture Documents
02-11-11 Hws Slides
02-11-11 Hws Slides
Henry Schneiderman
Robotics Institute
Carnegie Mellon University
Object Detection
“Object is present”
(at fixed size and alignment)
Classifier
P ( S1 | 1 ) P( S 2 | 1 ) P( S n | 1 )
H ( x1 ,... x r ) log log ... log
P ( S1 | 2 ) P( S 2 | 2 ) P( S n | 2 )
P ( S1 | 1 ) P( S 2 | 1 ) P( S n | 1 )
H ( x1 ,... x r ) log log ... log
P ( S1 | 2 ) P( S 2 | 2 ) P( S n | 2 )
Approach: Selection by Competition
x 1 x2 x3 . . . x m
q functions
Select combination of n candidates
log [pj1(Sj1|1) / pj1(Sj1|2)] log [pj2(Sj2|1) / pj2(Sj2|2)] . . . log [pjn(Sjn|1) / pjn(Sjn|2)]
n functions, n << q
H(x1,…,xr) =
log [pj1(Sj1|1) / pj1(Sj1|2)]+log [pj2(Sj2|1) / pj2(Sj2|2)] +. . .+ log [pjn(Sjn|1) / pjn(Sjn|2)]
Approach: Selection by Competition
x 1 x2 x3 . . . x m
H(x1,…,xr) =
log [pj1(Sj1|1) / pj1(Sj1|2)]+log [pj2(Sj2|1) / pj2(Sj2|2)] +. . .+ log [pjn(Sjn|1) / pjn(Sjn|2)]
Generation of Subsets
• “modeling error for assuming independence”
P ( x ... x | ) P ( x | )
C1 ... P ( x1...xq )abs[log P ( x11... xqq | 12 ) log PP(( xx11||12 )) P ( x2 |1 )
P ( x2 | 2 ) ... P ( xqq | 12 ) ]dx1...dxq
P ( x ... x | )
C2 ... P ( x1...xq )abs[log P ( x11... xqq | 12 ) ]dx1...dxq
C ( x j , xk ) C1 ( x j , xk ) C2 ( x j , xk ) Pair-affinity
Visualization of C(x,*)
(frontal faces)
x
x x
Measure over a Subset
D( Si ) 1
N ( N 1) C(x , x )
x j , xk S i
j k N Si Subset-affinity
x j xk
Generation of Candidate Subsets
x1 x2 x3 . . . . . . . . . . . . . . . . . . . . . . . . x m
S1 S2 .................. Sp
subset size vs. modeling power
H(x1,…,xr) =
log [pj1(Sj1|1) / pj1(Sj1|2)]+log [pj2(Sj2|1) / pj2(Sj2|2)] +. . .+ log [pjn(Sjn|1) / pjn(Sjn|2)]
Log-likelihood function = Table
Si = (xi1, xi2, . . ., xiq)
vector quantization
f i (1...N i )
table look-up
fi
Pi (fi |1)
fi
Pi (fi |2)
Example of VQ
xi1 xi2 xi3 . . . xiq
projection on to 3 principal components
c1 c2 c 3
quantization to m levels
z1 z2 z3
H(x1,…,xr) =
log [pj1(Sj1|1) / pj1(Sj1|2)]+log [pj2(Sj2|1) / pj2(Sj2|2)] +. . .+ log [pjn(Sjn|1) / pjn(Sjn|2)]
Candidate
log-likelihood h1(S1) h2(S2) ... hP(SP)
functions
Evaluate on
E E1,2 E2,1 E2,2 . . . Ep,1 Ep,2
training data 1,1
Evaluate
ROCs
ROC1 ROC2 ... ROCP
Order top Q
log-likelihood
hj1(Sj1) hj2(Sj2) . . . hjQ(SjQ)
functions
Form pQ pairs hj1(Sj1) + h1(S1) ... hjQ(SjQ) + hp(Sp)
of log-likelihood
functions
Sum
Evaluations Ej1,1+ E1,1 Ej1,2+ E1,2 . . . EjQ,1+ Ep,1 EjQ,2+ Ep,2
Evaluate
ROCs
ROC1 ... ROCQP
Order top Q
pairs of
hk1,1(Sk1,1) + hk1,2(Sk1,2) . . . hkQ,1(SkQ,1) + hkQ,2(SkQ,2)
log-likelihood
functions
. . . Repeat for n iterations
Cross-Validation Selects Classifier
Q Candidates:
H1(x1, x2, . . ., xr) = hk1,1(Sk1,1) + hk1,2(Sk1,2) +. . .+ hk1,n(Sk1,n)
...
HQ(x1, x2, . . ., xr) = hkQ,1(SkQ,1) + hkQ,2(SkQ,2) + . . . + hQ,n(SkQ,n)
Cross-validation
“Object is present”
(at fixed size and alignment)
Classifier
f2 = #3214
fn = #723
2) Look-Up Log-Likelihoods
P1( #5710 | 1)
f1 = #5710 log = 0.53
P1( #5710 | 2)
“Object is present”
(at fixed size and alignment)
Classifier
Face
Classifier #2
Face
Classifier #3
Detection: Apply Classifier
Exhaustively
Search in position
Search in scale
Decision can be made by partial evaluation
Apply log [p1(S1|1) / p1(S1|2)] Apply log [p2(S2|1) / p2(S2|2)] Apply log [p3(S3|1) / p3(S3|2)]
exhaustively to scaled input image reduced search space further reduced search space
Candidate-Based Evaluation
Compute N2 + M2 +2MN feature values Look-up M2 log-likelihood values
Feature-Based Evaluation
Cascade Implementation
Create candidate subsets
Increment stage
Face, eye, ear detection
Frontal Face Detection
• MIT-CMU Frontal Face Test Set [Sung and Poggio, 1995; Rowley, Baluja and Kanade, 1997]
False detections 6 13 44 64 79
(this method)
False Detections -- 31 50 167 --
[Viola and Jones,
CVPR, 2001]
AMD Athalon
1.2GHz
Face & Eye Detection for Red-Eye
Removal from Consumer Photos