Algorithms and

Applications in Computer
Vision
Lihi Zelnik-Manor
lihi@ee.technion.ac.il
ROC

Precision-Recall
circles Not circles
Truth:
Not circles
Result:
circles

Precision-Recall
circles Not circles
Truth:
Not circles

FN

Result:

circles

TP FP TN

Confusion matrix
P = Positive
N = Negative

Ground-truth
E1
E2
E1

tp
fp
(true
(false
positive) positive)

E2

fn
tn
(false
(true
negative negative
)
)

obtained
result

ROC curve

TP
TPR
P

AUC = Area
Under Curve

FP
FPR
N

Precision-Recall
TP
Re call
TP FN

TP
Pr ecision
TP FP

F-measure

The harmonic mean (mean of rates) of the precision and reca

precision recall
F
precision recall

True/false positives
50
75

true match

200

false match

feature distance

The distance threshold affects performance

True positives = # of detected matches that are correct
False positives = # of detected matches that are incorrect

Evaluating the results

How can we measure the performance of a feature matcher?
1
0.7
true
# true positives
# matching features (positives) positive
rate

0.1

false positive rate

# false positives
# unmatched features (negatives)

Evaluating the results

How can we measure the performance of a feature matcher?

ROC curve

(Receiver Operator Characteristic)

0.7
true
# true positives
# matching features (positives) positive
rate

0.1

false positive rate

# false positives
# unmatched features (negatives)

ROC Curves

Generated by counting # correct/incorrect matches, for different threholds

Want to maximize area under the curve (AUC)
Useful for comparing different feature matching methods
For more info: http://en.wikipedia.org/wiki/Receiver_operating_characteristic