Information Technology Reference
In-Depth Information
O
1
I
11
V
111
O
2
I
12
V
112
O
i
I
1j
V
11k
NV
11
feature
vectors of
image
NI
1
of object
O
1
NO
objects
NI
1
images
of object
O
1
Fig. 12.11
Relation between objects, images and features
Table 12.3
Formulas for obtaining decision
d
, response
r
and weight
w
values for various classifiers
Decision
d
Classifier
Classifier outputs
(positive decision
Response
r
∈[
0
,
1
]
Weight
w
>
0
condition)
max
o
1
+
1
2
2
o
2
+
1
,
o
2
o
2
2
ANN
(
o
1
,
o
2
)
∈[−
1
,
1
]
o
1
>
o
2
0
.
5
·
(
−
+
1
)
o
1
+
1
o
2
+
2
+
2
2
— values of outputs
o
∈[
0
,
1
]
—output,
L
= {
l
:
l
∈{
0
,
1
}}
k
i
=
1
m
i
·
l
i
kNN
— vector of
k
responses,
o
>
0
.
5
o
k
i
=
1
l
i
M
}
— vector of
k
distances
={
m
:
m
≥
0
RTree
o
∈[
0
,
1
]
—probability
o
>
0
.
5
o
1
o
∈[
0
,
1
]
—probability
o
>
0
.
5
GBTree
o
1
Then, the aggregated result
R
i
for each object
O
i
is calculated as follows:
NI
i
j
0
r
ij
NI
i
=
R
i
=
.
(12.32)
Search WWH ::
Custom Search