I
AE
S In
t
er
na
t
io
na
l J
o
urna
l o
f
Art
if
icia
l In
t
ellig
ence
(
I
J
-
AI
)
Vo
l.
15
,
No
.
1
,
Feb
r
u
ar
y
20
26
,
p
p
.
798
~
809
I
SS
N:
2
2
5
2
-
8
9
3
8
,
DOI
: 1
0
.
1
1
5
9
1
/ijai.v
15
.i
1
.
p
p
7
9
8
-
8
0
9
798
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
a
i
.
ia
esco
r
e.
co
m
Cla
ss
ificatio
n of
Ciha
teup
duck
eg
g
f
ertili
ty using
c
o
nv
o
lutiona
l
neura
l net
wo
rk
E
ff
icien
tNe
t
-
B3
E
v
i D
ewi
Sri
M
uly
a
ni
1
,
T
e
u
k
u M
ufiz
a
r
1
,
Da
ni Ro
hp
a
nd
i
2
,
Ay
u Dj
uli
a
ni
1
,
E
g
i R
a
hm
a
t
ull
o
h
1
,
Rina
ldi
Sa
t
ia
Aulia
Ra
hm
a
t
1
1
D
e
p
a
r
t
me
n
t
o
f
I
n
f
o
r
mat
i
c
s
E
n
g
i
n
e
e
r
i
n
g
,
F
a
c
u
l
t
y
o
f
E
n
g
i
n
e
e
r
i
n
g
,
U
n
i
v
e
r
si
t
a
s Perj
u
a
n
g
a
n
T
a
si
k
m
a
l
a
y
a
,
Ta
si
k
m
a
l
a
y
a
,
I
n
d
o
n
e
s
i
a
2
D
e
p
a
r
t
me
n
t
o
f
I
n
f
o
r
mat
i
c
s
E
n
g
i
n
e
e
r
i
n
g
,
S
T
M
I
K
M
a
r
d
i
r
a
I
n
d
o
n
e
s
i
a
,
B
a
n
d
u
n
g
,
I
n
d
o
n
e
si
a
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Sep
23
,
2
0
2
5
R
ev
is
ed
Dec
22
,
2
0
2
5
Acc
ep
ted
J
an
10
,
2
0
2
6
Ac
c
u
ra
te
d
e
tec
ti
o
n
o
f
e
g
g
fe
rti
l
it
y
is
c
ru
c
ial
to
imp
ro
v
e
h
a
tch
in
g
su
c
c
e
ss
in
d
u
c
k
fa
rm
in
g
.
Co
n
v
e
n
ti
o
n
a
l
c
a
n
d
li
n
g
m
e
th
o
d
s
re
ly
h
e
a
v
il
y
o
n
h
u
m
a
n
e
x
p
e
rti
se
,
m
a
k
i
n
g
t
h
e
m
su
b
jec
ti
v
e
a
n
d
e
rro
r
-
p
ro
n
e
.
T
h
is
stu
d
y
p
r
o
p
o
se
s
a
n
a
u
to
m
a
ted
c
las
sifica
ti
o
n
sy
ste
m
fo
r
Cih
a
teu
p
d
u
c
k
e
g
g
fe
rti
l
it
y
u
si
n
g
c
a
n
d
li
n
g
ima
g
e
s
a
n
d
a
c
o
n
v
o
lu
ti
o
n
a
l
n
e
u
ra
l
n
e
tw
o
rk
(CNN
)
b
a
se
d
o
n
th
e
Eff
icie
n
tNe
t
-
B3
a
rc
h
it
e
c
tu
re
.
I
m
a
g
e
e
n
h
a
n
c
e
m
e
n
t
tec
h
n
iq
u
e
s,
in
c
lu
d
in
g
c
o
n
tras
t
li
m
it
e
d
a
d
a
p
ti
v
e
h
ist
o
g
ra
m
e
q
u
a
li
z
a
ti
o
n
(CLAHE),
u
n
sh
a
r
p
m
a
sk
in
g
,
a
n
d
a
d
a
p
ti
v
e
t
h
re
sh
o
ld
i
n
g
,
we
re
a
p
p
l
ied
t
o
imp
ro
v
e
ima
g
e
q
u
a
li
ty
a
n
d
fe
a
tu
re
v
isi
b
il
it
y
.
Th
e
d
a
ta
se
t
c
o
n
siste
d
o
f
fe
rti
le
a
n
d
in
f
e
rti
le
e
g
g
ima
g
e
s
c
a
p
tu
re
d
a
t
two
in
c
u
b
a
ti
o
n
sta
g
e
s:
th
e
f
irst
2
4
h
o
u
rs
a
n
d
th
e
8
th
–
1
5
th
d
a
y
s.
Da
ta
we
re
sp
li
t
in
t
o
trai
n
i
n
g
,
v
a
li
d
a
ti
o
n
,
a
n
d
tes
ti
n
g
se
ts
wit
h
a
ra
ti
o
o
f
7
0
:
1
5
:
1
5
.
E
x
p
e
rime
n
tal
re
su
lt
s
s
h
o
w
th
a
t
ima
g
e
e
n
h
a
n
c
e
m
e
n
t
si
g
n
ifi
c
a
n
t
ly
imp
ro
v
e
s
c
las
sifica
ti
o
n
p
e
rf
o
rm
a
n
c
e
.
Wi
th
o
u
t
e
n
h
a
n
c
e
m
e
n
t,
t
h
e
m
o
d
e
l
a
c
h
iev
e
d
a
n
a
c
c
u
ra
c
y
o
f
4
9
%
with
a
n
a
re
a
u
n
d
e
r
c
u
rv
e
(
AUC
)
o
f
0
.
4
2
2
6
,
in
d
ica
ti
n
g
p
o
o
r
d
isc
rimin
a
ti
o
n
c
a
p
a
b
il
it
y
.
Wi
t
h
ima
g
e
e
n
h
a
n
c
e
m
e
n
t,
th
e
p
ro
p
o
se
d
m
e
th
o
d
a
c
h
iev
e
d
a
c
c
u
r
a
c
ies
o
f
7
7
%
fo
r
t
h
e
first
2
4
h
o
u
rs
d
a
tas
e
t
a
n
d
8
0
%
fo
r
t
h
e
8
th
–
1
5
th
d
a
y
s
d
a
tas
e
t,
wi
th
AU
C
v
a
lu
e
s
o
f
0
.
9
9
6
2
a
n
d
0
.
9
3
1
7
,
re
sp
e
c
ti
v
e
l
y
.
Th
e
se
re
su
lt
s
d
e
m
o
n
stra
te
th
a
t
Eff
ici
e
n
tNe
t
-
B3
c
o
m
b
in
e
d
wit
h
ima
g
e
e
n
h
a
n
c
e
m
e
n
t
p
ro
v
id
e
s
a
n
e
ffe
c
ti
v
e
a
n
d
c
o
m
p
u
tati
o
n
a
ll
y
e
fficie
n
t
so
l
u
ti
o
n
f
o
r
a
u
to
m
a
ted
fe
rti
li
t
y
d
e
t
e
c
ti
o
n
o
f
Cih
a
teu
p
d
u
c
k
e
g
g
s.
K
ey
w
o
r
d
s
:
C
ih
ateu
p
d
u
ck
e
g
g
s
C
o
n
v
o
lu
tio
n
al
n
eu
r
al
n
etwo
r
k
E
f
f
icien
tNet
-
B3
E
g
g
f
er
tili
ty
class
if
icatio
n
I
m
ag
e
en
h
an
ce
m
en
t
T
h
is i
s
a
n
o
p
e
n
a
c
c
e
ss
a
rticle
u
n
d
e
r th
e
CC B
Y
-
SA
li
c
e
n
se
.
C
o
r
r
e
s
p
o
nd
ing
A
uth
o
r
:
E
v
i D
ewi
Sri
Mu
ly
an
i
Dep
ar
tm
en
t o
f
I
n
f
o
r
m
atics E
n
g
in
ee
r
in
g
,
Facu
lty
o
f
E
n
g
in
ee
r
in
g
,
Un
iv
er
s
itas
Per
ju
an
g
a
n
T
asik
m
alay
a
P
eta
St.
No
1
7
7
,
T
asik
m
alay
a
,
W
est J
av
a
,
I
n
d
o
n
esia
E
m
ail: e
v
iajad
ec
h
@
g
m
ail.
co
m
1.
I
NT
RO
D
UCT
I
O
N
I
n
d
o
n
esian
lo
ca
l
d
u
ck
s
in
cl
u
d
in
g
C
ih
ateu
p
b
r
ee
d
s
in
W
est
J
av
a
,
I
n
d
o
n
esia
h
av
e
c
o
n
tr
i
b
u
ted
m
ea
t
an
d
eg
g
to
th
e
h
o
u
s
eh
o
ld
d
u
r
i
n
g
th
e
r
ec
e
n
t
p
ast
[
1
]
.
T
h
e
s
u
c
ce
s
s
o
f
d
u
ck
e
g
g
h
atc
h
in
g
h
ea
v
ily
d
ep
e
n
d
s
o
n
t
h
e
ab
ilit
y
to
d
is
tin
g
u
is
h
b
etwe
en
f
er
tile
an
d
in
f
er
tile
eg
g
s
.
T
r
ad
itio
n
al
m
eth
o
d
s
,
s
u
ch
as
ca
n
d
lin
g
,
ar
e
o
f
ten
in
ac
cu
r
ate
an
d
r
eq
u
ir
e
s
p
ec
ial
ized
s
k
ills
b
ased
o
n
th
e
f
ar
m
er
’
s
ex
p
er
ien
ce
,
wh
ich
c
an
le
ad
to
er
r
o
r
s
in
e
g
g
s
elec
tio
n
an
d
af
f
ec
t
p
r
o
d
u
ctio
n
ef
f
icien
cy
.
I
n
p
r
ev
io
u
s
r
esear
ch
,
ch
ick
en
eg
g
class
if
icatio
n
b
ased
o
n
s
e
g
m
en
tatio
n
an
d
f
e
r
tili
ty
s
tatu
s
h
as
b
ee
n
ca
r
r
ied
o
u
t
u
s
in
g
a
m
ask
r
e
g
io
n
-
b
ased
co
n
v
o
lu
tio
n
al
n
e
u
r
al
n
etwo
r
k
(
Ma
s
k
R
-
C
NN
)
b
ased
ap
p
r
o
ac
h
to
s
u
cc
ess
f
u
lly
d
etec
t,
class
if
y
,
an
d
s
eg
m
en
t
f
er
tile
a
n
d
in
f
er
t
ile
eg
g
s
[
2
]
.
I
n
an
o
th
er
s
tu
d
y
,
it
h
ig
h
li
g
h
ted
t
h
e
ef
f
icac
y
o
f
h
y
p
e
r
s
p
ec
tr
al
im
a
g
in
g
c
o
m
b
in
e
d
with
m
ac
h
in
e
lear
n
in
g
as
a
p
o
ten
tial
g
r
ee
n
t
ec
h
n
o
lo
g
y
f
o
r
eg
g
f
er
tili
ty
d
etec
tio
n
to
war
d
s
a
s
u
s
tain
ab
le
eg
g
in
d
u
s
tr
y
[
3
]
.
Fu
r
th
er
m
o
r
e,
r
esear
ch
o
n
d
etec
tin
g
th
e
f
er
tili
ty
o
f
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
C
la
s
s
i
fica
tio
n
o
f Ci
h
a
teu
p
d
u
c
k
eg
g
fert
ilit
y
u
s
in
g
co
n
vo
l
u
tio
n
a
l n
eu
r
a
l
n
etw
o
r
k
…
(
E
vi
Dewi S
r
i Mu
lya
n
i)
799
ch
ick
en
e
g
g
s
u
s
in
g
th
e
e
x
tr
a
ctio
n
tech
n
i
q
u
e
u
s
in
g
f
ir
s
t
-
o
r
d
er
s
tatis
tics
(
FOS)
f
r
o
m
i
m
ag
e
s
eg
m
en
tatio
n
r
esu
lts
class
if
ied
with
a
b
ac
k
p
r
o
p
a
g
atio
n
ar
tific
ial
n
e
u
r
al
n
etwo
r
k
(
ANN)
p
r
o
d
u
ce
d
r
elat
iv
ely
lo
w
ac
cu
r
ac
y
,
th
u
s
n
o
t a
ch
iev
i
n
g
o
p
tim
al
r
esu
lts
[
4
]
.
Ho
wev
er
,
s
p
ec
if
ic
a
p
p
licatio
n
s
f
o
r
C
ih
ateu
p
d
u
ck
eg
g
f
er
tili
ty
h
av
e
n
o
t
y
et
b
ee
n
co
n
d
u
cte
d
,
d
esp
ite
th
is
s
p
ec
ies
p
r
o
d
u
cin
g
eg
g
s
with
ch
ar
ac
ter
is
tics
d
is
tin
ct
f
r
o
m
o
th
er
p
o
u
ltry
.
T
h
is
s
tu
d
y
o
p
en
s
o
p
p
o
r
tu
n
ities
to
au
to
m
ate
th
e
p
r
o
ce
s
s
o
f
class
if
y
in
g
e
g
g
f
e
r
tili
ty
b
ased
o
n
c
an
d
lin
g
im
a
g
es
[
5
]
,
[
6
]
.
Fo
r
t
h
e
class
if
icatio
n
o
f
C
ih
ateu
p
d
u
ck
eg
g
f
er
tili
ty
,
th
e
E
f
f
icien
tNet
ar
ch
itectu
r
e
will
b
e
s
elec
ted
to
o
p
tim
ize
m
o
d
el
p
er
f
o
r
m
an
c
e,
aim
in
g
f
o
r
h
ig
h
er
ac
cu
r
ac
y
with
m
o
r
e
co
m
p
u
tatio
n
al
ef
f
icien
cy
,
th
er
eb
y
p
r
o
d
u
cin
g
a
h
ig
h
-
ac
c
u
r
ac
y
au
to
m
ated
p
r
o
to
ty
p
e.
2.
M
E
T
H
O
D
T
h
e
ca
n
d
lin
g
m
eth
o
d
h
as
lo
n
g
b
ee
n
u
s
ed
i
n
th
e
p
o
u
ltry
i
n
d
u
s
tr
y
to
ass
ess
eg
g
q
u
ality
an
d
f
er
tili
ty
b
y
o
b
s
er
v
in
g
t
h
eir
in
ter
n
al
s
tr
u
c
tu
r
e
th
r
o
u
g
h
lig
h
tin
g
[
7
]
.
T
h
is
ass
e
s
s
m
en
t
is
co
n
d
u
cted
m
an
u
ally
,
m
ak
i
n
g
it
p
r
o
n
e
to
s
u
b
jectiv
ity
an
d
in
c
o
n
s
is
ten
cy
,
wh
ich
h
as
led
to
th
e
d
ev
elo
p
m
e
n
t
o
f
au
to
m
ate
d
ap
p
r
o
ac
h
es
u
s
in
g
im
ag
e
p
r
o
ce
s
s
in
g
tech
n
o
lo
g
y
as
a
s
o
lu
tio
n
[
8
]
,
[
9
]
.
I
n
th
e
e
x
p
er
im
en
ts
th
at
h
av
e
b
ee
n
ca
r
r
ied
o
u
t,
it
ca
n
b
e
s
ee
n
th
at
co
n
tr
ast
lim
ited
ad
a
p
tiv
e
h
is
to
g
r
am
e
q
u
aliza
tio
n
(
C
L
AHE
)
ca
n
in
cr
ea
s
e
th
e
lo
ca
l
co
n
tr
ast
o
f
th
e
im
ag
e
with
o
u
t
i
n
cr
ea
s
in
g
th
e
n
o
is
e
s
o
th
at
it
ca
n
im
p
r
o
v
e
t
h
e
q
u
ality
o
f
t
h
e
r
esu
ltin
g
im
a
g
e
[
1
0
]
–
[
1
2
]
,
wh
ile
u
n
s
h
ar
p
m
ask
in
g
s
h
ar
p
en
s
t
h
e
ed
g
es
s
o
th
at
d
etails
ar
e
clea
r
er
[
1
3
]
,
[
1
4
]
,
wh
ile
a
d
ap
tiv
e
th
r
esh
o
ld
i
n
g
co
n
v
er
ts
g
r
a
y
s
ca
le
im
ag
es
in
t
o
b
in
ar
y
ad
ap
tiv
el
y
s
o
th
at
th
ey
ar
e
ef
f
ec
tiv
e
in
u
n
ev
en
lig
h
tin
g
co
n
d
itio
n
s
s
o
th
at
th
ey
will im
p
r
o
v
e
th
e
q
u
a
lity
o
f
th
e
r
esu
ltin
g
im
ag
e
f
r
o
m
th
e
f
ea
tu
r
e
ex
tr
ac
tio
n
p
r
o
ce
s
s
[
1
5
]
,
[
1
6
]
.
C
o
n
v
o
l
u
t
i
o
n
a
l
n
e
u
r
a
l
n
e
tw
o
r
k
(
C
N
N
)
h
a
v
e
b
e
c
o
m
e
t
h
e
s
t
a
n
d
a
r
d
f
o
r
i
m
a
g
e
c
l
a
s
s
i
f
i
c
at
i
o
n
t
as
k
s
d
u
e
t
o
t
h
e
i
r
a
b
i
l
it
y
t
o
a
u
t
o
m
a
ti
c
a
ll
y
a
n
d
h
i
e
r
a
r
c
h
i
c
a
l
l
y
e
x
t
r
ac
t
f
e
a
t
u
r
es
[
1
7
]
–
[
1
9
]
.
V
a
r
i
o
u
s
C
NN
a
r
c
h
i
t
e
c
t
u
r
es
h
a
v
e
b
e
e
n
d
e
v
e
l
o
p
e
d
a
n
d
a
p
p
l
i
e
d
a
c
r
o
s
s
d
i
f
f
e
r
e
n
t
d
o
m
a
i
n
s
.
I
n
2
0
1
9
,
r
es
ea
r
c
h
e
r
s
f
r
o
m
G
o
o
g
l
e
A
I
i
n
t
r
o
d
u
c
e
d
E
f
f
i
c
i
e
n
tN
e
t
,
a
f
a
m
i
l
y
o
f
C
N
N
m
o
d
e
ls
t
h
at
o
p
t
i
m
i
z
es
p
e
r
f
o
r
m
a
n
c
e
u
s
i
n
g
a
c
o
m
p
o
u
n
d
s
c
a
l
i
n
g
te
c
h
n
i
q
u
e
[
2
0
]
,
[
2
1
]
.
T
h
i
s
a
p
p
r
o
a
c
h
u
n
i
f
o
r
m
l
y
s
c
a
l
e
s
t
h
e
n
e
t
w
o
r
k
’
s
d
e
p
t
h
,
w
i
d
t
h
,
a
n
d
r
e
s
o
l
u
t
i
o
n
u
s
i
n
g
a
s
i
n
g
l
e
s
ca
l
a
r
p
a
r
a
m
e
t
e
r
,
r
e
s
u
l
ti
n
g
i
n
m
o
d
e
l
s
t
h
a
t
a
r
e
m
o
r
e
e
f
f
i
c
i
e
n
t
a
n
d
h
i
g
h
e
r
-
p
e
r
f
o
r
m
i
n
g
c
o
m
p
a
r
e
d
t
o
p
r
e
v
i
o
u
s
a
r
c
h
i
t
e
c
t
u
r
e
s
.
O
n
e
v
a
r
i
a
n
t
o
f
t
h
i
s
f
a
m
i
l
y
i
s
E
f
f
i
c
i
e
n
tN
e
t
-
B
3
,
w
h
ic
h
o
f
f
e
r
s
a
b
a
l
a
n
c
e
b
e
tw
e
e
n
a
cc
u
r
a
c
y
a
n
d
c
o
m
p
u
t
a
t
i
o
n
a
l
e
f
f
i
ci
e
n
c
y
[
2
2
]
.
T
h
e
s
elec
tio
n
o
f
E
f
f
icien
tNet
-
B
3
as
th
e
C
NN
ar
ch
itectu
r
e
f
o
r
eg
g
f
er
tili
ty
class
if
icatio
n
i
s
b
ased
o
n
its
s
u
p
er
io
r
ity
in
ac
h
iev
in
g
an
o
p
tim
al
b
ala
n
ce
b
etwe
en
ac
c
u
r
ac
y
a
n
d
c
o
m
p
u
tatio
n
al
ef
f
ic
ien
cy
.
W
h
ile
o
th
e
r
ar
ch
itectu
r
es
lik
e
R
esNet
an
d
Den
s
eNe
t
o
f
f
er
h
ig
h
ac
cu
r
a
cy
,
E
f
f
icie
n
tNet
-
B
3
s
u
r
p
ass
es
th
em
b
y
y
ield
in
g
s
im
ilar
o
r
e
v
en
b
etter
p
r
ed
icti
o
n
p
er
f
o
r
m
an
ce
,
y
et
ac
h
iev
e
d
with
a
s
ig
n
if
ican
tly
lo
wer
n
u
m
b
er
o
f
p
ar
am
eter
s
an
d
co
m
p
u
tatio
n
al
r
eq
u
ir
em
e
n
ts
(
g
ig
a
f
l
o
atin
g
-
p
o
in
t
o
p
er
atio
n
s
p
er
s
ec
o
n
d
s
(
GFLO
Ps
)
)
.
T
h
is
a
d
v
an
tag
e
s
tem
s
f
r
o
m
th
e
u
n
i
q
u
e
c
o
m
p
o
u
n
d
s
ca
lin
g
m
et
h
o
d
,
wh
ich
in
tellig
en
tly
o
p
tim
izes
an
d
b
a
lan
ce
s
th
e
n
etwo
r
k
d
ep
th
,
wid
th
,
an
d
in
p
u
t
r
eso
lu
tio
n
s
im
u
ltan
eo
u
s
ly
[
2
3
]
.
C
o
n
s
eq
u
en
tly
,
E
f
f
icien
tNet
-
B
3
p
r
o
v
id
es
a
co
m
p
ac
t
y
et
p
o
wer
f
u
l
m
o
d
el,
m
ak
in
g
i
t
id
ea
l
f
o
r
th
e
s
p
ec
if
ic
an
d
s
en
s
itiv
e
im
ag
e
class
if
icatio
n
task
o
f
d
etec
tin
g
eg
g
f
er
tili
ty
wh
ile
also
e
n
s
u
r
in
g
f
a
s
t
in
f
er
en
ce
an
d
p
r
ac
tical
im
p
l
em
en
tatio
n
o
n
r
eso
u
r
ce
-
co
n
s
tr
ain
ed
h
a
r
d
war
e
in
r
ea
l
-
wo
r
ld
o
p
er
atio
n
al
e
n
v
ir
o
n
m
en
ts
.
E
f
f
icien
tNet
h
as
b
ee
n
ap
p
lie
d
ac
r
o
s
s
v
ar
io
u
s
d
o
m
ain
s
,
an
d
s
tu
d
ies
s
h
o
w
th
at
it
ca
n
ac
h
iev
e
h
ig
h
ac
cu
r
ac
y
with
b
etter
co
m
p
u
t
atio
n
al
ef
f
icien
c
y
co
m
p
ar
e
d
to
p
r
e
v
io
u
s
m
o
d
els
[
2
4
]
–
[
2
6
]
.
I
n
teg
r
atin
g
th
e
ca
n
d
lin
g
m
eth
o
d
with
C
NNs
u
s
in
g
E
f
f
icien
tNet
-
B
3
o
f
f
e
r
s
s
ig
n
if
ican
t
p
o
ten
tial
to
im
p
r
o
v
e
b
o
th
ac
cu
r
ac
y
an
d
ef
f
icien
cy
[
2
7
]
,
[
2
8
]
,
p
ar
ticu
l
ar
ly
in
class
if
y
in
g
th
e
f
er
tili
t
y
o
f
C
ih
ateu
p
d
u
c
k
e
g
g
s
.
Fig
u
r
e
1
illu
s
tr
ates
th
e
r
esear
ch
wo
r
k
f
lo
w
co
n
d
u
cte
d
to
class
if
y
eg
g
f
er
tili
ty
.
T
h
e
p
r
o
ce
s
s
b
eg
in
s
with
im
ag
e
ac
q
u
is
itio
n
o
f
two
eg
g
class
es
(
f
er
tile
an
d
in
f
er
tile)
,
f
o
llo
wed
b
y
s
p
litt
in
g
th
e
d
ata
s
et
in
to
tr
ain
in
g
,
v
alid
atio
n
,
a
n
d
test
in
g
s
ets.
T
h
e
p
r
ep
r
o
ce
s
s
in
g
s
tag
e
in
clu
d
es
i
m
ag
e
en
h
an
ce
m
en
t
(
u
s
in
g
C
L
AHE
,
u
n
s
h
ar
p
m
ask
in
g
,
an
d
a
d
ap
tiv
e
th
r
esh
o
ld
)
,
b
u
ild
in
g
a
T
e
n
s
o
r
Flo
w
d
ata
p
ip
elin
e
(
i
n
v
o
lv
i
n
g
r
esizin
g
,
n
o
r
m
aliza
tio
n
,
b
atch
i
n
g
,
an
d
p
r
ef
etc
h
in
g
)
,
an
d
ap
p
ly
in
g
d
ata
au
g
m
e
n
tatio
n
te
ch
n
iq
u
es
s
u
ch
as
f
lip
p
in
g
,
r
o
t
atio
n
,
an
d
zo
o
m
in
g
t
o
in
cr
ea
s
e
v
ar
iab
ilit
y
.
Nex
t,
th
e
im
ag
es
ar
e
p
r
o
ce
s
s
ed
u
s
in
g
a
C
NN
m
o
d
el
b
ased
o
n
E
f
f
i
cien
tNet
-
B
3
,
wh
ich
is
tr
ain
e
d
an
d
f
i
n
e
-
tu
n
e
d
with
ca
llb
ac
k
s
s
u
ch
as
Mo
d
elC
h
ec
k
p
o
in
t,
E
ar
ly
Sto
p
p
in
g
,
an
d
R
ed
u
ce
L
R
On
Plateau
.
T
h
e
m
o
d
el’
s
p
er
f
o
r
m
an
ce
is
ev
alu
ated
u
s
in
g
a
c
o
n
f
u
s
io
n
m
atr
ix
,
r
ec
eiv
er
o
p
er
ati
n
g
ch
a
r
ac
ter
is
tic
(
R
O
C
)
cu
r
v
e,
an
d
a
r
ea
u
n
d
e
r
th
e
cu
r
v
e
(
AUC)
to
ass
e
s
s
cla
s
s
if
icat
io
n
ac
cu
r
ac
y
.
Fin
ally
,
th
e
co
n
clu
s
io
n
s
u
m
m
ar
izes th
e
ef
f
ec
tiv
en
ess
o
f
th
e
p
r
o
p
o
s
e
d
m
eth
o
d
in
au
to
m
atica
lly
d
etec
tin
g
f
er
tile e
g
g
s
.
I
n
th
is
s
tu
d
y
,
two
ex
p
er
im
e
n
ts
will
b
e
co
n
d
u
cted
:
th
e
f
ir
s
t
f
o
r
class
if
y
in
g
eg
g
f
er
tili
ty
b
ased
o
n
ca
n
d
lin
g
r
esu
lts
with
im
ag
e
en
h
an
ce
m
en
t
a
n
d
with
o
u
t
i
m
ag
e
en
h
an
c
em
en
t
f
o
r
f
er
til
e
eg
g
s
at
th
e
f
ir
s
t
2
4
h
o
u
r
s
,
f
e
r
tile
eg
g
s
at
th
e
8
th
an
d
1
5
th
d
ay
s
an
d
in
f
e
r
tile
eg
g
s
u
s
in
g
t
h
e
E
f
f
icien
tNet
-
B
3
C
NN
alg
o
r
ith
m
.
Fig
u
r
e
2
is
an
ex
am
p
le
o
f
a
d
ataset
:
Fig
u
r
e
2
(
a)
f
er
tile
eg
g
s
in
th
e
f
i
r
s
t
2
4
h
o
u
r
s
,
Fig
u
r
e
2
(
b
)
f
e
r
tile
eg
g
s
at
8
an
d
1
5
d
ay
s
o
f
ag
e
,
an
d
Fig
u
r
e
2
(
c)
in
f
er
tile
eg
g
s
.
T
h
is
clas
s
if
icatio
n
will
u
s
e
two
cla
s
s
e
s
,
n
am
ely
f
er
tile
an
d
in
f
er
tile,
with
th
e
d
etailed
q
u
an
titi
es
p
r
esen
ted
in
T
ab
le
1
.
Nex
t,
th
e
d
ata
is
s
p
lit
with
a
r
atio
o
f
7
0
%
f
o
r
tr
ain
in
g
,
1
5
% f
o
r
test
in
g
,
an
d
1
5
% f
o
r
v
alid
atio
n
,
p
er
f
o
r
m
ed
r
an
d
o
m
ly
.
T
h
e
d
etails ar
e
p
r
e
s
en
ted
in
T
ab
le
2
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2252
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
15
,
No
.
1
,
Feb
r
u
ar
y
20
26
:
798
-
8
0
9
800
Fig
u
r
e
1
.
E
x
p
er
im
e
n
tal
m
eth
o
d
s
(
a)
(
b
)
(
c)
Fig
u
r
e
2
.
Data
s
et
of
(
a)
f
er
tile
th
e
f
ir
s
t 2
4
h
o
u
r
s
,
(
b
)
th
e
8
th
a
n
d
1
5
t
h
d
ay
s
,
a
n
d
(
c
)
in
f
e
r
tile
T
ab
le
1
.
Data
s
et
N
o
.
Ex
p
e
r
i
m
e
n
t
F
e
r
t
i
l
e
I
n
f
e
r
t
i
l
e
A
mo
u
n
t
1
Th
e
f
i
r
st
2
4
h
o
u
r
s
8
6
7
8
4
9
1
7
1
6
2
Th
e
8
t
h
a
n
d
1
5
t
h
d
a
y
s
7
8
4
7
4
5
1
5
2
9
T
ab
le
2
.
Data
s
p
lit
D
a
t
a
Th
e
f
i
r
st
2
4
h
o
u
r
s
Th
e
8
t
h
a
n
d
1
5
t
h
d
a
y
s
Tr
a
i
n
i
n
g
1
2
0
0
1
0
6
9
V
a
l
i
d
a
t
i
o
n
2
5
7
2
2
8
Te
st
i
n
g
2
5
9
2
3
2
3.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
N
3
.
1
.
P
re
pro
ce
s
s
ing
Fig
u
r
e
3
is
th
e
r
esu
lts
o
f
im
ag
e
en
h
an
ce
m
e
n
t
ap
p
lied
t
o
th
r
ee
eg
g
s
am
p
les
:
f
er
tile
at
th
e
f
ir
s
t
2
4
h
o
u
r
s
as
s
h
o
w
n
in
Fig
u
r
e
3
(
a)
,
f
er
tile
at
th
e
8
th
an
d
1
5
t
h
d
ay
s
in
Fig
u
r
e
3
(
b
)
,
an
d
in
f
er
tile
in
Fig
u
r
e
3
(
c)
th
r
o
u
g
h
th
r
ee
k
ey
p
r
o
ce
s
s
in
g
m
eth
o
d
s
:
C
L
AHE
,
u
n
s
h
ar
p
m
ask
in
g
,
an
d
ad
ap
tiv
e
t
h
r
esh
o
ld
in
g
.
C
L
AHE
en
h
an
ce
s
lo
ca
l
co
n
tr
ast,
m
ak
i
n
g
f
in
e
s
tr
u
ctu
r
es
o
n
t
h
e
s
u
r
f
ac
e
o
f
th
e
eg
g
an
d
th
e
em
b
r
y
o
m
o
r
e
v
is
ib
le.
I
n
Fig
u
r
e
3
(
b
)
,
t
h
e
em
b
r
y
o
an
d
b
lo
o
d
v
ess
els
ar
e
s
ig
n
if
ican
tl
y
m
o
r
e
p
r
o
m
in
en
t
co
m
p
ar
ed
to
th
e
u
n
p
r
o
ce
s
s
ed
im
ag
e.
I
n
s
am
p
les
Fig
u
r
es
3
(
a)
an
d
3
(
c)
,
s
h
ell
s
u
r
f
ac
e
p
atter
n
s
an
d
th
e
eg
g
’
s
co
n
to
u
r
s
ar
e
r
en
d
er
e
d
with
g
r
ea
ter
clar
ity
.
Un
s
h
ar
p
m
ask
in
g
tech
n
iq
u
e
s
h
ar
p
e
n
s
ed
g
es
b
y
r
ed
u
cin
g
b
lu
r
ef
f
ec
ts
.
As
a
r
esu
lt,
th
e
eg
g
's
b
o
u
n
d
ar
ies
an
d
tex
t
u
r
e
d
etail
s
b
ec
o
m
e
cr
is
p
er
,
f
ac
ilit
atin
g
d
is
tin
ctio
n
b
etwe
en
in
ter
n
al
an
d
ex
ter
n
al
ar
ea
s
.
Acr
o
s
s
all
s
am
p
les,
ed
g
e
co
n
tr
ast is
im
p
r
o
v
ed
,
s
im
p
lify
in
g
th
e
d
etec
tio
n
o
f
em
b
r
y
o
n
ic
f
ea
tu
r
es.
Ad
ap
tiv
e
th
r
esh
o
ld
m
eth
o
d
c
o
n
v
er
ts
g
r
ay
s
ca
le
im
ag
es
in
to
b
in
ar
y
o
n
es
b
y
ad
ap
tin
g
to
lo
ca
l
lig
h
tin
g
co
n
d
itio
n
s
.
I
n
Fig
u
r
e
3
(
a)
,
th
e
eg
g
'
s
s
h
ap
e
is
clea
r
ly
d
ef
in
e
d
,
th
o
u
g
h
em
b
r
y
o
n
ic
d
etails
ar
e
less
v
is
ib
le
d
u
e
t
o
th
e
d
o
m
in
an
ce
o
f
n
o
is
e.
I
n
Fi
g
u
r
e
3
(
b
)
,
th
e
em
b
r
y
o
’
s
s
tr
u
ctu
r
e
an
d
in
ter
io
r
r
eg
io
n
s
o
f
th
e
eg
g
ar
e
v
is
u
alize
d
,
alb
eit
with
g
r
ain
y
n
o
is
e.
I
n
Fig
u
r
e
3
(
c)
,
th
e
eg
g
'
s
co
n
to
u
r
s
tan
d
s
o
u
t
m
o
s
t
p
r
o
m
in
en
tly
,
w
h
ile
in
ter
n
al
d
etails
r
em
ain
r
elativ
ely
in
s
ig
n
if
ican
t
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
C
la
s
s
i
fica
tio
n
o
f Ci
h
a
teu
p
d
u
c
k
eg
g
fert
ilit
y
u
s
in
g
co
n
vo
l
u
tio
n
a
l n
eu
r
a
l
n
etw
o
r
k
…
(
E
vi
Dewi S
r
i Mu
lya
n
i)
801
C
L
AHE
was
ap
p
lied
to
im
p
r
o
v
e
lo
c
al
co
n
tr
ast
an
d
en
h
an
ce
em
b
r
y
o
v
is
ib
ilit
y
with
o
u
t
am
p
lify
in
g
n
o
is
e.
Un
s
h
ar
p
m
ask
in
g
was
u
s
ed
to
s
h
ar
p
en
ed
g
es
an
d
clar
if
y
s
tr
u
ctu
r
al
b
o
u
n
d
ar
i
es,
wh
ile
ad
ap
tiv
e
th
r
esh
o
ld
in
g
was
em
p
lo
y
ed
t
o
h
an
d
le
u
n
ev
en
illu
m
in
atio
n
b
y
a
d
ap
tiv
ely
b
in
ar
izi
n
g
th
e
im
ag
es.
T
o
g
eth
er
,
th
ese
tech
n
iq
u
es im
p
r
o
v
e
f
ea
t
u
r
e
v
is
ib
ilit
y
an
d
s
u
p
p
o
r
t m
o
r
e
ef
f
ec
tiv
e
C
NN
f
ea
tu
r
e
ex
tr
ac
tio
n
.
(
a)
(
b
)
(
c)
Fig
u
r
e
3
.
I
m
ag
e
e
n
h
an
ce
m
e
n
t
o
f
(
a)
f
er
tile a
t th
e
f
ir
s
t 2
4
h
o
u
r
s
,
(
b
)
f
e
r
tile a
t th
e
8
th
a
n
d
1
5
t
h
d
ay
s
,
an
d
(
c)
in
f
er
tile
3
.
2
.
Aug
m
ent
a
t
io
n
Fig
u
r
e
4
s
h
o
ws
th
e
im
ag
e
af
ter
au
g
m
en
tatio
n
.
Fig
u
r
e
4
(
a)
d
is
p
lay
s
th
e
r
esu
lts
o
f
im
ag
e
au
g
m
en
tatio
n
with
im
ag
e
e
n
h
an
ce
m
e
n
t
ap
p
lied
to
eg
g
im
ag
es
u
s
in
g
s
ev
er
al
tr
an
s
f
o
r
m
atio
n
s
u
s
in
g
au
g
m
en
tatio
n
tech
n
iq
u
es
o
n
T
ab
le
3
to
en
h
an
ce
d
eg
g
im
ag
es
an
d
th
eir
ef
f
ec
ts
o
n
m
o
d
el
g
en
er
aliza
tio
n
.
Fig
u
r
e
4
(
b
)
d
is
p
lay
s
th
e
r
esu
lt
s
o
f
im
ag
e
au
g
m
en
tatio
n
with
o
u
t im
ag
e
e
n
h
an
ce
m
e
n
t.
(
a)
Fig
u
r
e
4
.
I
m
ag
e
af
te
r
au
g
m
en
t
atio
n
o
f
(
a)
with
im
ag
e
e
n
h
an
c
em
en
t a
n
d
(
b
)
with
o
u
t im
ag
e
e
n
h
an
ce
m
e
n
t
A
f
t
e
r
a
ug
m
e
nt
at
i
o
n
A
f
t
e
r
a
ug
m
e
nt
at
i
o
n
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2252
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
15
,
No
.
1
,
Feb
r
u
ar
y
20
26
:
798
-
8
0
9
802
T
ab
le
3
.
I
m
ag
e
au
g
m
en
tatio
n
tech
n
iq
u
es
Te
c
h
n
i
q
u
e
P
a
r
a
me
t
e
r
P
u
r
p
o
se
/
i
m
p
a
c
t
o
n
t
h
e
m
o
d
e
l
V
e
r
t
i
c
a
l
f
l
i
p
—
H
e
l
p
s
t
h
e
mo
d
e
l
l
e
a
r
n
o
r
i
e
n
t
a
t
i
o
n
-
i
n
v
a
r
i
a
n
t
f
e
a
t
u
r
e
s
H
o
r
i
z
o
n
t
a
l
f
l
i
p
—
I
mp
r
o
v
e
s
r
o
b
u
s
t
n
e
ss
t
o
h
o
r
i
z
o
n
t
a
l
o
r
i
e
n
t
a
t
i
o
n
v
a
r
i
a
t
i
o
n
s
R
o
t
a
t
i
o
n
0
.
0
5
r
a
d
i
a
n
s (
≈
2
.
9
°
)
En
h
a
n
c
e
s m
o
d
e
l
g
e
n
e
r
a
l
i
z
a
t
i
o
n
t
o
mi
n
o
r
a
n
g
u
l
a
r
v
a
r
i
a
t
i
o
n
s
Zo
o
m
0
.
2
(
+
2
0
%)
En
a
b
l
e
s
t
h
e
m
o
d
e
l
t
o
r
e
c
o
g
n
i
z
e
r
e
l
e
v
a
n
t
f
e
a
t
u
r
e
s
a
c
r
o
ss
d
i
f
f
e
r
e
n
t
s
c
a
l
e
s
a
n
d
d
i
s
t
a
n
c
e
s
3
.
3
.
M
o
delin
g
I
n
th
e
m
o
d
elin
g
p
ip
elin
e
f
o
r
b
in
ar
y
class
if
icatio
n
with
E
f
f
icien
tNet
-
B
3
as
th
e
b
ac
k
b
o
n
e
th
r
o
u
g
h
tr
an
s
f
er
lear
n
in
g
u
s
in
g
p
ar
am
eter
an
d
f
i
n
e
-
tu
n
in
g
co
n
f
ig
u
r
atio
n
o
n
T
a
b
les
4
an
d
5
.
T
h
e
in
p
u
t
r
eso
lu
tio
n
o
f
3
0
0
×3
0
0
p
ix
els
wer
e
s
elec
ted
to
m
atch
th
e
o
p
tim
al
co
n
f
ig
u
r
atio
n
o
f
E
f
f
icien
tNet
-
B
3
,
p
r
o
v
id
in
g
a
b
alan
ce
b
etwe
en
f
ea
tu
r
e
d
etail
p
r
eser
v
atio
n
an
d
co
m
p
u
tatio
n
al
ef
f
icien
cy
.
T
h
is
r
eso
lu
tio
n
is
s
u
f
f
icien
t
to
ca
p
tu
r
e
cr
itical
ca
n
d
lin
g
f
ea
tu
r
es,
s
u
ch
as e
m
b
r
y
o
s
tr
u
ct
u
r
es,
wh
ile
a
v
o
id
in
g
u
n
n
ec
ess
ar
y
co
m
p
u
tatio
n
al
o
v
er
h
ea
d
.
T
ab
le
4
.
E
f
f
icien
tNet
-
B
3
m
o
d
el
an
d
tr
ain
in
g
p
ar
a
m
eter
s
P
a
r
a
me
t
e
r
S
e
t
t
i
n
g
I
n
p
u
t
i
m
a
g
e
s
i
z
e
3
0
0
×
3
0
0
×
3
C
N
N
a
r
c
h
i
t
e
c
t
u
r
e
Ef
f
i
c
i
e
n
t
N
e
t
-
B3
P
r
e
t
r
a
i
n
e
d
w
e
i
g
h
t
s
I
mag
e
N
e
t
I
n
c
l
u
d
e
t
o
p
l
a
y
e
r
F
a
l
se
B
a
se
m
o
d
e
l
t
r
a
i
n
a
b
l
e
(
i
n
i
t
i
a
l
)
F
a
l
se
P
o
o
l
i
n
g
l
a
y
e
r
G
l
o
b
a
l
a
v
e
r
a
g
e
p
o
o
l
i
n
g
D
r
o
p
o
u
t
r
a
t
e
0
.
3
O
u
t
p
u
t
l
a
y
e
r
D
e
n
se
(
1
n
e
u
r
o
n
,
s
i
g
m
o
i
d
)
Lo
ss f
u
n
c
t
i
o
n
B
i
n
a
r
y
c
r
o
ss
-
e
n
t
r
o
p
y
O
p
t
i
mi
z
e
r
A
d
a
m
Le
a
r
n
i
n
g
r
a
t
e
(
i
n
i
t
i
a
l
t
r
a
i
n
i
n
g
)
3
×
1
0
-
4
B
a
t
c
h
si
z
e
16
Ep
o
c
h
s (i
n
i
t
i
a
l
t
r
a
i
n
i
n
g
)
15
Ev
a
l
u
a
t
i
o
n
me
t
r
i
c
A
c
c
u
r
a
c
y
D
a
t
a
a
u
g
m
e
n
t
a
t
i
o
n
F
l
i
p
,
r
o
t
a
t
i
o
n
,
z
o
o
m
C
a
l
l
b
a
c
k
s
M
o
d
e
l
C
h
e
c
k
p
o
i
n
t
,
E
a
r
l
y
S
t
o
p
p
i
n
g
,
R
e
d
u
c
e
L
R
O
n
P
l
a
t
e
a
u
T
ab
le
5
.
Fin
e
-
tu
n
in
g
c
o
n
f
ig
u
r
a
tio
n
P
a
r
a
me
t
e
r
S
e
t
t
i
n
g
B
a
se
m
o
d
e
l
t
r
a
i
n
a
b
l
e
Tr
u
e
Tr
a
i
n
a
b
l
e
l
a
y
e
r
s
La
st
2
0
l
a
y
e
r
s
F
r
o
z
e
n
l
a
y
e
r
s
A
l
l
l
a
y
e
r
s
e
x
c
e
p
t
l
a
st
2
0
O
p
t
i
mi
z
e
r
A
d
a
m
Le
a
r
n
i
n
g
r
a
t
e
(
f
i
n
e
-
t
u
n
i
n
g
)
1
×
1
0
-
5
Lo
ss f
u
n
c
t
i
o
n
B
i
n
a
r
y
c
r
o
ss
-
e
n
t
r
o
p
y
Ev
a
l
u
a
t
i
o
n
me
t
r
i
c
A
c
c
u
r
a
c
y
Ep
o
c
h
s (f
i
n
e
-
t
u
n
i
n
g
)
5
Tr
a
i
n
i
n
g
d
a
t
a
t
r
a
i
n
_
d
s
V
a
l
i
d
a
t
i
o
n
d
a
t
a
v
a
l
_
d
s
C
a
l
l
b
a
c
k
s
S
a
me
a
s i
n
i
t
i
a
l
t
r
a
i
n
i
n
g
Fiftee
n
ep
o
ch
s
wer
e
u
s
ed
f
o
r
in
itial
tr
ain
in
g
to
allo
w
ef
f
e
ctiv
e
ad
ap
tatio
n
to
th
e
d
atase
t
with
o
u
t
o
v
er
f
itti
n
g
.
Fin
e
-
tu
n
in
g
was
lim
ited
to
f
iv
e
ep
o
ch
s
b
ec
au
s
e
th
e
p
r
etr
ain
ed
weig
h
ts
ar
e
alr
ea
d
y
n
ea
r
o
p
tim
al,
an
d
ex
ce
s
s
iv
e
f
in
e
-
tu
n
in
g
m
ay
d
eg
r
a
d
e
p
r
e
v
io
u
s
ly
lear
n
e
d
r
ep
r
esen
tatio
n
s
.
On
ly
t
h
e
last
2
0
lay
er
s
wer
e
u
n
f
r
o
z
en
d
u
r
in
g
f
in
e
-
tu
n
in
g
t
o
ad
ap
t
h
ig
h
-
lev
el,
task
-
s
p
ec
i
f
ic
f
ea
tu
r
es
wh
ile
p
r
eser
v
in
g
lo
w
-
lev
el
g
e
n
er
ic
f
ea
tu
r
es lea
r
n
ed
f
r
o
m
I
m
ag
eN
et.
T
h
is
s
tr
ateg
y
r
ed
u
ce
s
o
v
er
f
itti
n
g
an
d
p
r
ev
en
ts
ca
tast
r
o
p
h
i
c
f
o
r
g
ettin
g
.
Fig
u
r
e
5
s
h
o
ws
th
e
tr
ain
in
g
an
d
v
alid
atio
n
g
r
ap
h
.
I
n
Fig
u
r
e
5
(
a)
,
th
e
g
r
ap
h
s
h
o
ws
th
at
tr
ain
in
g
ac
cu
r
ac
y
f
lu
ct
u
ates
s
h
ar
p
ly
an
d
d
ec
r
ea
s
es
to
war
d
th
e
f
i
n
al
ep
o
ch
s
,
wh
ile
v
alid
atio
n
ac
cu
r
ac
y
r
em
ain
s
s
tag
n
an
t
at
ar
o
u
n
d
0
.
5
,
in
d
icat
in
g
th
at
th
e
m
o
d
el
f
ails
to
lear
n
ef
f
ec
tiv
ely
.
On
th
e
o
th
er
h
a
n
d
,
tr
ain
i
n
g
lo
s
s
is
u
n
s
tab
le
with
s
h
ar
p
r
is
es
an
d
d
r
o
p
s
,
wh
er
ea
s
v
alid
atio
n
l
o
s
s
i
s
m
o
r
e
s
tab
le
an
d
d
ec
r
ea
s
es
g
r
ad
u
ally
.
T
h
is
p
atter
n
s
u
g
g
ests
is
s
u
es
with
co
n
v
er
g
en
ce
an
d
m
o
d
el
o
p
tim
izatio
n
,
p
o
s
s
ib
ly
ca
u
s
ed
b
y
an
u
n
s
u
itab
le
ar
ch
itectu
r
e,
in
a
p
p
r
o
p
r
iate
lea
r
n
in
g
r
ate,
o
r
d
ata
im
b
alan
ce
.
I
n
Fig
u
r
e
5
(
b
)
,
th
e
g
r
ap
h
s
s
h
o
w
th
at
b
o
th
tr
ain
in
g
a
n
d
v
a
lid
atio
n
ac
cu
r
ac
y
in
c
r
ea
s
e
s
ig
n
if
ican
tly
u
n
til
th
e
f
in
al
ep
o
ch
,
with
v
alid
atio
n
r
ea
ch
in
g
ar
o
u
n
d
0
.
9
.
T
r
ain
in
g
an
d
v
alid
ati
o
n
lo
s
s
al
s
o
d
ec
r
ea
s
e
co
n
s
is
ten
tly
,
alth
o
u
g
h
th
e
r
e
is
a
b
r
ief
s
p
ik
e
at
ep
o
ch
1
4
.
Ov
er
all,
th
e
m
o
d
el
lear
n
s
well,
g
en
er
alize
s
ef
f
ec
tiv
ely
o
n
v
alid
atio
n
d
ata,
an
d
s
h
o
ws o
n
ly
m
in
o
r
s
ig
n
s
o
f
in
s
tab
ilit
y
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
C
la
s
s
i
fica
tio
n
o
f Ci
h
a
teu
p
d
u
c
k
eg
g
fert
ilit
y
u
s
in
g
co
n
vo
l
u
tio
n
a
l n
eu
r
a
l
n
etw
o
r
k
…
(
E
vi
Dewi S
r
i Mu
lya
n
i)
803
On
th
e
r
ig
h
t
Fig
u
r
e
5
(
c)
,
th
e
g
r
ap
h
s
s
h
o
w
th
at
tr
ain
in
g
ac
cu
r
ac
y
f
lu
ctu
ates
s
h
ar
p
ly
an
d
ev
e
n
d
r
o
p
s
at
ce
r
tain
p
o
in
ts
,
wh
ile
v
alid
atio
n
ac
cu
r
ac
y
in
c
r
ea
s
es
s
tead
il
y
,
r
ea
ch
i
n
g
clo
s
e
to
0
.
9
.
T
r
ain
in
g
lo
s
s
is
also
u
n
s
tab
le
with
s
ig
n
if
ican
t
s
p
ik
es,
wh
er
ea
s
v
alid
atio
n
lo
s
s
d
ec
r
ea
s
es
co
n
s
is
ten
tly
.
T
h
is
in
d
ic
ates
th
at
th
e
m
o
d
el
g
en
er
alize
s
well
o
n
v
alid
atio
n
d
ata,
alth
o
u
g
h
th
e
tr
ai
n
in
g
p
r
o
ce
s
s
r
em
ain
s
s
o
m
ewh
at
u
n
s
tab
le.
(
a)
(
b
)
(c
)
Fig
u
r
e
5
.
T
r
ain
in
g
a
n
d
v
alid
atio
n
g
r
a
p
h
(
a
)
with
o
u
t im
a
g
e
en
h
an
ce
m
en
t
,
(
b
)
f
ir
s
t 2
4
h
o
u
r
s
with
im
ag
e
en
h
an
ce
m
e
n
t,
an
d
(
c)
th
e
8
th
a
n
d
1
5
t
h
d
ay
s
with
im
a
g
e
en
h
a
n
ce
m
en
t
3
.
4
.
E
v
a
lua
t
i
o
n
Fig
u
r
e
6
s
h
o
ws
th
e
c
o
n
f
u
s
io
n
m
atr
ix
.
I
n
Fig
u
r
e
6
(
a
)
,
th
is
co
n
f
u
s
io
n
m
atr
i
x
s
h
o
ws
th
at
th
e
m
o
d
el
p
r
ed
icts
all
s
am
p
les
as
in
f
er
tile,
ca
u
s
in
g
all
1
1
9
f
er
tile
s
am
p
les
to
b
e
m
is
class
if
ied
(
0
%
r
ec
all)
,
wh
ile
all
1
1
3
in
f
e
r
tile
s
am
p
les
ar
e
co
r
r
ec
tly
id
en
tifie
d
(
1
0
0
%
r
ec
all)
,
in
d
icatin
g
a
s
tr
o
n
g
b
ias
to
war
d
th
e
in
f
er
tile
class
an
d
a
c
o
m
p
lete
f
ailu
r
e
to
r
ec
o
g
n
ize
f
er
tile.
T
h
e
m
o
d
el'
s
to
ta
l
f
ailu
r
e
(
f
e
r
tile
r
ec
all
0
%)
is
ca
u
s
ed
b
y
e
x
tr
em
e
class
im
b
alan
ce
,
m
ea
n
in
g
th
e
p
r
im
ar
y
s
o
lu
tio
n
s
lie
in
d
ata
b
alan
cin
g
an
d
lo
s
s
f
u
n
ctio
n
ad
j
u
s
tm
en
t.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2252
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
15
,
No
.
1
,
Feb
r
u
ar
y
20
26
:
798
-
8
0
9
804
T
o
en
s
u
r
e
th
e
m
o
d
el
h
as
s
u
f
f
icien
t
s
am
p
les
f
r
o
m
t
h
e
m
in
o
r
ity
class
(
f
er
tile)
,
Ov
e
r
s
am
p
lin
g
tech
n
iq
u
es
s
u
ch
as
s
y
n
th
etic
m
in
o
r
ity
o
v
er
-
s
am
p
lin
g
tech
n
iq
u
e
(
SMOT
E
)
m
u
s
t
b
e
a
p
p
l
ied
to
th
e
tr
ain
in
g
d
ata.
Su
b
s
eq
u
en
tly
,
to
p
r
ev
e
n
t
th
e
m
o
d
el
f
r
o
m
i
g
n
o
r
in
g
th
e
m
in
o
r
ity
class
,
a
weig
h
te
d
lo
s
s
f
u
n
ctio
n
o
r
f
o
ca
l
lo
s
s
s
h
o
u
ld
b
e
u
tili
ze
d
.
T
h
is
ap
p
r
o
ac
h
in
h
e
r
en
tly
ass
ig
n
s
a
s
ig
n
if
ican
tly
h
i
g
h
er
lo
s
s
weig
h
t
wh
en
th
e
m
o
d
el
m
is
class
if
ies
a
f
er
tile
s
am
p
le,
ef
f
ec
tiv
ely
f
o
r
cin
g
th
e
m
o
d
el
t
o
p
ay
g
r
e
ater
atten
tio
n
to
th
e
m
in
o
r
ity
class
[
2
3
]
.
T
h
e
s
u
cc
ess
o
f
th
ese
in
ter
v
en
tio
n
s
m
u
s
t
b
e
m
ea
s
u
r
ed
u
s
in
g
m
in
o
r
ity
-
class
s
en
s
itiv
e
m
etr
ics,
s
u
ch
as
f
er
tile
class
r
ec
all
an
d
F1
-
s
co
r
e
,
r
ath
er
th
an
ju
s
t
ac
cu
r
ac
y
.
I
n
Fig
u
r
e
6
(
b
)
,
th
is
co
n
f
u
s
io
n
m
atr
ix
s
h
o
ws
th
at
o
u
t
o
f
1
3
1
f
er
tile
s
am
p
les,
o
n
ly
7
2
wer
e
co
r
r
ec
tly
id
en
tifie
d
wh
ile
5
9
we
r
e
m
is
class
if
ied
as
in
f
er
tile,
wh
er
ea
s
a
ll
1
2
8
in
f
er
tile
s
am
p
les
wer
e
c
lass
if
ied
co
r
r
ec
tly
.
T
h
e
m
o
d
el
p
e
r
f
o
r
m
s
v
er
y
well
in
d
etec
tin
g
in
f
er
tile
(
1
0
0
%
r
ec
all)
b
u
t
is
less
ef
f
ec
tiv
e
f
o
r
f
er
tile
(
5
5
%
r
ec
all)
,
r
esu
ltin
g
in
u
n
b
ala
n
ce
d
p
er
f
o
r
m
an
ce
b
etwe
en
th
e
two
class
es.
Fig
u
r
e
6
(
c)
s
h
o
ws
th
e
r
esu
lt
s
o
f
th
e
co
n
f
u
s
io
n
m
atr
ix
an
d
class
if
icatio
n
r
ep
o
r
t
f
o
r
th
e
two
-
class
class
if
icat
io
n
m
o
d
el,
n
am
ely
f
er
tile
an
d
in
f
er
tile
.
Fr
o
m
th
e
co
n
f
u
s
io
n
m
atr
ix
,
it
ca
n
b
e
o
b
s
er
v
ed
th
at
o
u
t
o
f
1
1
9
f
er
tile
s
am
p
les,
1
1
1
wer
e
co
r
r
ec
tly
p
r
ed
icted
(
tr
u
e
p
o
s
itiv
es),
wh
ile
8
wer
e
m
is
clas
s
if
ied
as
in
f
er
tile
(
f
alse
n
eg
ativ
es).
Me
an
wh
ile,
o
u
t
o
f
1
1
3
in
f
er
tile
s
am
p
les,
7
5
wer
e
co
r
r
ec
tly
p
r
e
d
icted
(
tr
u
e
n
eg
ativ
es),
a
n
d
3
8
wer
e
m
is
class
if
ied
as f
er
tile (
f
alse p
o
s
itiv
es).
(
a)
(
b
)
(
c)
F
i
g
u
r
e
6
.
C
o
n
f
u
s
i
o
n
m
a
t
r
i
x
o
f
(
a
)
w
i
t
h
o
u
t
i
m
a
g
e
e
n
h
a
n
c
e
m
e
n
t
,
(
b
)
t
h
e
f
i
r
s
t
2
4
h
o
u
r
s
w
it
h
i
m
ag
e
e
n
h
a
n
c
e
m
e
n
t
,
a
n
d
(
c
)
t
h
e
8
t
h
a
n
d
1
5
t
h
d
a
y
s
wi
t
h
i
m
a
g
e
e
n
h
a
n
c
e
m
e
n
t
T
h
e
class
if
icatio
n
r
ep
o
r
t
s
h
o
ws
o
n
T
ab
le
6
th
at
in
d
icate
s
th
at
th
e
m
o
d
el
ac
h
iev
e
d
4
9
%
ac
cu
r
ac
y
,
co
r
r
ec
tly
id
en
tify
i
n
g
all
in
f
er
ti
le
s
am
p
les b
u
t c
o
m
p
letely
f
ailin
g
to
d
etec
t th
e
f
er
tile c
lass
,
s
h
o
win
g
s
tr
o
n
g
b
ias
an
d
u
n
b
alan
ce
d
p
er
f
o
r
m
an
ce
.
T
h
e
class
if
icatio
n
r
ep
o
r
t
s
h
o
ws
o
n
T
ab
le
7
t
h
at
in
d
icat
es
th
at
th
e
m
o
d
el
ac
h
iev
es
p
er
f
ec
t
p
r
ec
is
io
n
b
u
t
lo
w
r
ec
all
f
o
r
th
e
f
er
tile
class
,
wh
ile
f
o
r
th
e
in
f
er
tile
class
it
s
h
o
ws
h
ig
h
r
ec
all
b
u
t
lo
wer
p
r
ec
is
io
n
,
r
esu
ltin
g
in
a
n
o
v
er
all
ac
cu
r
ac
y
o
f
7
7
%
with
b
etter
s
en
s
itiv
ity
to
wa
r
d
in
f
er
tile
s
am
p
les
th
an
f
er
tile
o
n
es.
B
ased
o
n
th
e
class
if
icatio
n
r
ep
o
r
t
o
n
T
ab
l
e
8
,
t
h
e
m
o
d
el
ac
h
iev
ed
g
o
o
d
p
er
f
o
r
m
an
ce
with
8
0
%
ac
cu
r
ac
y
,
s
h
o
win
g
h
ig
h
r
ec
all
f
o
r
th
e
f
er
tile
class
an
d
h
ig
h
p
r
ec
is
io
n
b
u
t
lo
wer
r
ec
all
f
o
r
th
e
in
f
er
tile
class
,
in
d
icatin
g
th
e
n
ee
d
t
o
i
m
p
r
o
v
e
d
etec
tio
n
o
f
m
is
s
ed
in
f
er
tile c
ases
.
T
ab
le
6
.
C
lass
if
icatio
n
r
ep
o
r
t
with
o
u
t im
ag
e
en
h
an
ce
m
e
n
t
P
r
e
c
i
s
i
o
n
R
e
c
a
l
l
F1
-
sc
o
r
e
S
u
p
p
o
r
t
F
e
r
t
i
l
0
.
0
0
0
.
0
0
0
.
0
0
1
1
9
I
n
f
e
r
t
i
l
e
0
.
4
9
1
.
0
0
0
.
6
6
1
1
3
A
c
c
u
r
a
c
y
0
.
4
9
2
3
2
M
a
c
r
o
a
v
g
0
.
2
4
0
.
5
0
0
.
3
3
2
3
2
W
e
i
g
h
t
a
v
g
0
.
2
4
0
.
4
9
0
.
3
2
2
3
2
T
ab
le
7
.
C
lass
if
icatio
n
r
ep
o
r
ts
th
e
f
ir
s
t 2
4
h
o
u
r
s
with
im
a
g
e
en
h
an
ce
m
e
n
t
P
r
e
c
i
s
i
o
n
R
e
c
a
l
l
F1
-
sc
o
r
e
S
u
p
p
o
r
t
F
e
r
t
i
l
1
.
0
0
0
.
5
5
0
.
7
1
1
3
1
I
n
f
e
r
t
i
l
e
0
.
6
8
1
.
0
0
0
.
8
1
1
2
8
A
c
c
u
r
a
c
y
0
.
7
7
2
5
9
M
a
c
r
o
a
v
g
0
.
8
4
0
.
7
7
0
.
7
6
2
5
9
W
e
i
g
h
t
a
v
g
0
.
8
4
0
.
7
7
0
.
7
6
2
5
9
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
C
la
s
s
i
fica
tio
n
o
f Ci
h
a
teu
p
d
u
c
k
eg
g
fert
ilit
y
u
s
in
g
co
n
vo
l
u
tio
n
a
l n
eu
r
a
l
n
etw
o
r
k
…
(
E
vi
Dewi S
r
i Mu
lya
n
i)
805
T
ab
le
8
.
C
lass
if
icatio
n
r
ep
o
r
ts
th
e
8
th
a
n
d
1
5
th
d
a
y
s
with
im
ag
e
en
h
a
n
ce
m
en
t
P
r
e
c
i
s
i
o
n
R
e
c
a
l
l
F1
-
sc
o
r
e
S
u
p
p
o
r
t
F
e
r
t
i
l
0
.
7
4
0
.
9
3
0
.
8
3
1
1
9
I
n
f
e
r
t
i
l
e
0
.
9
0
0
.
6
6
0
.
7
7
1
1
3
A
c
c
u
r
a
c
y
0
.
8
0
2
3
2
M
a
c
r
o
a
v
g
0
.
8
2
0
.
8
0
0
.
8
0
2
3
2
W
e
i
g
h
t
a
v
g
0
.
8
2
0
.
8
0
0
.
8
0
2
3
2
Fig
u
r
e
7
s
h
o
ws
th
e
AUC
v
alu
e
in
th
e
R
OC
cu
r
v
e,
in
Fig
u
r
e
7
(
a)
s
h
o
ws
th
at
R
OC
cu
r
v
e
s
h
o
ws
an
AUC
v
alu
e
o
f
0
.
4
2
2
6
,
m
ea
n
in
g
th
e
m
o
d
el
p
er
f
o
r
m
s
wo
r
s
e
t
h
an
r
an
d
o
m
g
u
ess
in
g
(
AUC=
0
.
5
)
.
T
h
is
in
d
icate
s
th
at
th
e
m
o
d
el
f
ails
to
p
r
o
p
er
l
y
d
is
tin
g
u
is
h
b
etwe
en
p
o
s
itiv
e
an
d
n
eg
ativ
e
class
es,
s
u
g
g
esti
n
g
im
p
r
o
v
em
e
n
ts
ar
e
n
ee
d
ed
i
n
th
e
d
ata,
f
ea
tu
r
es,
o
r
m
o
d
el
u
s
ed
.
R
OC
cu
r
v
e
in
Fig
u
r
e
7
(
b
)
s
h
o
w
s
ex
ce
llen
t
m
o
d
el
p
er
f
o
r
m
an
ce
with
an
AUC=0
.
9
9
6
2
,
wh
ic
h
is
v
er
y
clo
s
e
to
1
.
0
.
T
h
e
c
u
r
v
e
lies
f
ar
ab
o
v
e
th
e
d
iag
o
n
al
lin
e
,
in
d
icatin
g
t
h
at
th
e
m
o
d
el
h
as
an
alm
o
s
t
p
er
f
ec
t
ab
ilit
y
t
o
d
is
tin
g
u
is
h
b
etwe
en
p
o
s
itiv
e
a
n
d
n
e
g
ativ
e
class
es.
Ho
wev
er
,
s
u
c
h
a
h
ig
h
v
alu
e
s
h
o
u
ld
b
e
in
ter
p
r
eted
with
ca
u
t
io
n
,
as
it
m
ay
s
u
g
g
est
o
v
e
r
f
itti
n
g
o
r
ev
alu
atio
n
o
n
non
-
r
e
p
r
esen
tativ
e
d
ata.
T
h
e
AUC
v
alu
e
in
th
e
R
OC
cu
r
v
e
in
Fig
u
r
e
7
(
c
)
is
0
.
9
3
1
7
,
wh
i
ch
in
d
icate
s
th
at
th
e
m
o
d
el
h
as
a
9
3
.
1
7
%
ab
ilit
y
to
d
is
tin
g
u
is
h
b
etwe
en
p
o
s
itiv
e
an
d
n
eg
ativ
e
class
es.
T
h
is
m
ea
n
s
th
e
m
o
d
el
p
er
f
o
r
m
s
v
er
y
well.
T
h
e
s
h
ap
e
o
f
th
e
cu
r
v
e
s
h
o
ws a
h
ig
h
tr
u
e
p
o
s
itiv
e
r
ate
(
T
PR
)
an
d
a
lo
w
f
alse p
o
s
itiv
e
r
ate
(
FP
R
)
,
in
d
icatin
g
th
at
th
e
m
o
d
el
h
as st
r
o
n
g
d
etec
tio
n
ca
p
ab
il
ity
with
m
in
im
al
f
alse p
o
s
itiv
e
s
.
Fo
r
th
e
v
is
u
aliza
tio
n
o
f
th
e
p
r
ed
ictio
n
r
esu
lts
,
s
ee
Fig
u
r
e
8
.
Fig
u
r
e
8
(
a)
s
h
o
ws
p
r
ed
ictio
n
v
is
u
aliza
tio
n
with
o
u
t
im
ag
e
e
n
h
an
ce
m
e
n
t
.
Fig
u
r
e
8
(
b
)
s
h
o
ws
p
r
ed
ictio
n
v
is
u
aliza
tio
n
th
e
f
ir
s
t
2
4
h
o
u
r
s
with
im
ag
e
en
h
a
n
ce
m
en
t
.
Fig
u
r
e
8
(
c)
s
h
o
ws th
e
8
th
a
n
d
1
5
th
d
a
y
s
with
im
ag
e
en
h
an
ce
m
e
n
t.
(
a)
(
b
)
(
c)
Fig
u
r
e
7
.
R
OC
cu
r
v
e
o
f
(
a)
with
o
u
t im
ag
e
e
n
h
an
ce
m
en
t
,
(
b
)
th
e
f
ir
s
t 2
4
h
o
u
r
s
with
im
ag
e
e
n
h
an
ce
m
e
n
t,
an
d
(
c)
th
e
8
th
an
d
1
5
th
d
ay
s
with
im
ag
e
en
h
a
n
ce
m
en
t
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2252
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
15
,
No
.
1
,
Feb
r
u
ar
y
20
26
:
798
-
8
0
9
806
(
a)
(
b
)
(
c)
Fig
u
r
e
8
.
Pre
d
ictio
n
v
is
u
aliza
tio
n
of
(
a)
with
o
u
t im
ag
e
en
h
a
n
ce
m
en
t
,
(
b
)
th
e
f
ir
s
t 2
4
h
o
u
r
s
with
im
ag
e
en
h
an
ce
m
e
n
t,
an
d
(
c)
th
e
8
th
a
n
d
1
5
t
h
d
ay
s
with
im
a
g
e
en
h
a
n
ce
m
en
t
4.
CO
NCLU
SI
O
N
T
h
e
r
esear
ch
r
esu
lts
s
h
o
w
th
at
th
e
ap
p
licatio
n
o
f
im
a
g
e
e
n
h
an
ce
m
e
n
t
h
as
a
s
ig
n
if
ican
t
im
p
ac
t
o
n
m
o
d
el
p
e
r
f
o
r
m
an
ce
.
W
ith
o
u
t
i
m
ag
e
en
h
a
n
ce
m
en
t,
th
e
m
o
d
el
f
ailed
to
lear
n
p
r
o
p
er
l
y
,
as
in
d
icate
d
b
y
s
tag
n
an
t
v
alid
atio
n
ac
cu
r
ac
y
ar
o
u
n
d
0
.
5
,
a
lo
w
AUC
v
alu
e
(
0
.
4
2
2
6
)
,
an
d
a
co
m
p
lete
f
ailu
r
e
to
r
e
co
g
n
ize
th
e
f
er
tile
class
.
I
n
co
n
tr
ast,
with
im
ag
e
en
h
an
ce
m
en
t
a
p
p
lied
t
o
th
e
f
ir
s
t
2
4
h
o
u
r
s
o
f
d
ata
as
well
as
th
e
8
th
an
d
1
5
th
d
ay
s
,
th
e
m
o
d
el
d
em
o
n
s
tr
ated
m
u
ch
b
etter
g
e
n
er
aliza
tio
n
,
a
ch
iev
in
g
v
alid
atio
n
ac
cu
r
ac
y
o
f
u
p
to
9
5
%,
h
ig
h
AUC
v
alu
es
(
0
.
9
9
6
2
an
d
0
.
9
3
1
7
)
,
an
d
m
o
r
e
b
alan
ce
d
class
if
icatio
n
p
er
f
o
r
m
a
n
ce
ac
r
o
s
s
class
e
s
,
alth
o
u
g
h
s
o
m
e
m
is
class
if
icatio
n
s
s
til
l
o
cc
u
r
r
ed
.
Ov
er
all,
im
ag
e
e
n
h
an
ce
m
en
t
im
p
r
o
v
ed
tr
ain
in
g
s
tab
ilit
y
,
ac
cu
r
ac
y
,
an
d
th
e
m
o
d
el’
s
d
is
cr
im
in
ativ
e
ca
p
ab
ilit
y
,
alth
o
u
g
h
f
u
r
th
er
o
p
tim
izatio
n
is
r
e
q
u
ir
ed
to
m
in
im
ize
o
v
er
f
itti
n
g
a
n
d
im
p
r
o
v
e
class
p
er
f
o
r
m
a
n
ce
b
alan
ce
.
Fu
r
t
h
er
m
o
r
e,
th
e
s
e
r
esear
ch
f
in
d
in
g
s
ca
n
b
e
ex
ten
d
e
d
with
h
y
p
er
p
ar
am
eter
o
p
tim
izatio
n
tech
n
iq
u
es
s
u
ch
as
lear
n
in
g
r
ate,
b
atch
s
ize,
an
d
n
u
m
b
e
r
o
f
ep
o
ch
s
to
ac
h
iev
e
m
o
r
e
s
tab
le
tr
ain
in
g
.
T
h
e
ap
p
licatio
n
o
f
r
eg
u
la
r
izatio
n
tech
n
iq
u
es
s
u
ch
as
d
r
o
p
o
u
t
o
r
we
ig
h
t
d
ec
ay
is
also
im
p
o
r
tan
t
to
r
ed
u
ce
th
e
r
is
k
o
f
o
v
er
f
itti
n
g
,
wh
ile
m
eth
o
d
s
f
o
r
h
an
d
lin
g
d
ata
im
b
alan
ce
s
u
c
h
as
o
v
er
s
am
p
lin
g
,
u
n
d
er
s
am
p
lin
g
,
o
r
c
o
s
t
-
s
en
s
it
iv
e
lear
n
in
g
m
ay
h
elp
ad
d
r
ess
class
b
ias.
I
n
ad
d
itio
n
,
th
e
u
s
e
o
f
alter
n
ativ
e
m
o
d
el
ar
ch
itectu
r
es
s
u
ch
as
R
esNet,
Den
s
eNe
t,
o
r
v
is
io
n
tr
an
s
f
o
r
m
er
ca
n
b
e
co
m
p
a
r
e
d
to
o
b
tain
o
p
tim
al
p
er
f
o
r
m
an
ce
.
E
x
p
e
r
im
en
ts
o
n
lar
g
er
an
d
m
o
r
e
d
iv
er
s
e
d
atasets
ar
e
also
r
eq
u
ir
ed
to
en
s
u
r
e
g
en
er
aliza
tio
n
,
an
d
th
e
im
p
lem
en
tatio
n
o
f
ex
p
lain
ab
le
ar
tific
ial
in
tellig
en
ce
(
XAI
)
ap
p
r
o
ac
h
es
s
u
ch
as
g
r
a
d
ien
t
-
weig
h
ted
class
ac
tiv
atio
n
m
ap
p
i
n
g
(
Gr
a
d
-
C
AM
)
ca
n
ass
is
t in
u
n
d
er
s
tan
d
in
g
th
e
k
ey
f
ea
tu
r
es lea
r
n
e
d
b
y
th
e
m
o
d
el.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
C
la
s
s
i
fica
tio
n
o
f Ci
h
a
teu
p
d
u
c
k
eg
g
fert
ilit
y
u
s
in
g
co
n
vo
l
u
tio
n
a
l n
eu
r
a
l
n
etw
o
r
k
…
(
E
vi
Dewi S
r
i Mu
lya
n
i)
807
ACK
NO
WL
E
DG
E
M
E
NT
S
T
h
e
au
t
h
o
r
s
w
o
u
ld
lik
e
t
o
th
eir
ap
p
r
ec
iatio
n
to
Un
iv
er
s
itas
Per
ju
an
g
an
T
asik
m
alay
a
f
o
r
th
e
co
n
tin
u
o
u
s
s
u
p
p
o
r
t a
n
d
f
ac
ilit
atio
n
th
r
o
u
g
h
o
u
t th
e
r
esear
ch
p
r
o
ce
s
s
.
F
UNDING
I
NF
O
R
M
A
T
I
O
N
T
h
is
r
esear
ch
was
f
u
n
d
ed
b
y
th
e
Dir
ec
to
r
ate
o
f
R
esear
ch
an
d
C
o
m
m
u
n
ity
Ser
v
ice,
Dir
ec
to
r
ate
Gen
er
al
o
f
R
esear
ch
an
d
De
v
elo
p
m
en
t,
Min
is
tr
y
o
f
Hig
h
er
E
d
u
ca
tio
n
,
Scien
ce
,
a
n
d
T
ec
h
n
o
lo
g
y
o
f
th
e
R
ep
u
b
lic
o
f
I
n
d
o
n
esia,
th
r
o
u
g
h
th
e
2
0
2
5
Fu
n
d
am
en
tal
R
esear
ch
Gr
an
t
Sch
em
e
(
R
eg
u
lar
Sch
em
e)
u
n
d
er
co
n
tr
ac
t n
u
m
b
er
s
1
2
5
/C
3
/DT
.
0
5
.
0
0
/PL/2
0
2
5
,
7
9
3
0
/LL
4
/PG
/2
0
2
5
,
an
d
0
2
9
/KP/LP2
M
-
UP/0
7
/2
0
2
5
.
AUTHO
R
CO
NT
RI
B
UT
I
O
NS ST
A
T
E
M
E
N
T
T
h
is
jo
u
r
n
al
u
s
es
th
e
C
o
n
tr
ib
u
to
r
R
o
les
T
ax
o
n
o
m
y
(
C
R
ed
iT)
to
r
ec
o
g
n
ize
in
d
iv
id
u
al
au
th
o
r
co
n
tr
ib
u
tio
n
s
,
r
ed
u
ce
au
th
o
r
s
h
ip
d
is
p
u
tes,
an
d
f
ac
ilit
ate
co
llab
o
r
atio
n
.
Na
m
e
o
f
Aut
ho
r
C
M
So
Va
Fo
I
R
D
O
E
Vi
Su
P
Fu
E
v
i D
ewi
Sri
Mu
ly
an
i
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
T
eu
k
u
Mu
f
izar
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
Dan
i Ro
h
p
an
d
i
✓
✓
✓
✓
✓
✓
✓
Ay
u
Dju
lian
i
✓
✓
✓
✓
✓
✓
✓
E
g
i Rah
m
atu
llo
h
✓
✓
✓
✓
✓
✓
✓
R
in
ald
i Satia
Au
lia
R
ah
m
at
✓
✓
✓
✓
✓
✓
✓
C
:
C
o
n
c
e
p
t
u
a
l
i
z
a
t
i
o
n
M
:
M
e
t
h
o
d
o
l
o
g
y
So
:
So
f
t
w
a
r
e
Va
:
Va
l
i
d
a
t
i
o
n
Fo
:
Fo
r
mal
a
n
a
l
y
s
i
s
I
:
I
n
v
e
s
t
i
g
a
t
i
o
n
R
:
R
e
so
u
r
c
e
s
D
:
D
a
t
a
C
u
r
a
t
i
o
n
O
:
W
r
i
t
i
n
g
-
O
r
i
g
i
n
a
l
D
r
a
f
t
E
:
W
r
i
t
i
n
g
-
R
e
v
i
e
w
&
E
d
i
t
i
n
g
Vi
:
Vi
su
a
l
i
z
a
t
i
o
n
Su
:
Su
p
e
r
v
i
s
i
o
n
P
:
P
r
o
j
e
c
t
a
d
mi
n
i
st
r
a
t
i
o
n
Fu
:
Fu
n
d
i
n
g
a
c
q
u
i
si
t
i
o
n
CO
NF
L
I
C
T
O
F
I
N
T
E
R
E
S
T
ST
A
T
E
M
E
NT
Au
th
o
r
s
s
tate
n
o
co
n
f
lict o
f
in
t
er
est.
I
NF
O
RM
E
D
CO
NS
E
N
T
W
e
h
av
e
o
b
tain
ed
in
f
o
r
m
ed
c
o
n
s
en
t f
r
o
m
all
in
d
iv
id
u
als in
c
lu
d
ed
in
t
h
is
s
tu
d
y
.
E
T
H
I
CAL AP
P
RO
V
AL
T
h
e
r
esear
ch
r
elate
d
to
an
i
m
a
l
u
s
e
h
as
b
ee
n
co
m
p
lied
with
all
th
e
r
elev
an
t
n
atio
n
al
r
eg
u
l
atio
n
s
an
d
in
s
titu
tio
n
al
p
o
licies f
o
r
th
e
ca
r
e
an
d
u
s
e
o
f
an
im
als.
DATA AV
AI
L
AB
I
L
I
T
Y
T
h
e
d
ataset
th
at
s
u
p
p
o
r
ts
th
e
f
in
d
in
g
s
o
f
t
h
is
s
tu
d
y
ar
e
a
v
ailab
le
f
r
o
m
th
e
c
o
r
r
esp
o
n
d
i
n
g
au
th
o
r
,
[
E
DSM
]
,
u
p
o
n
r
ea
s
o
n
a
b
le
r
eq
u
est.
RE
F
E
R
E
NC
E
S
[
1
]
L.
L
a
n
,
A
.
A
n
a
n
g
,
I
.
S
e
t
i
a
w
a
n
,
a
n
d
H
.
I
n
d
r
i
j
a
n
i
,
“
C
o
m
p
a
r
i
si
o
n
o
f
f
o
u
r
s
i
g
mo
i
d
a
l
m
o
d
e
l
s
t
o
d
e
s
c
r
i
b
e
g
r
o
w
t
h
c
u
r
v
e
s
o
f
C
i
h
a
t
e
u
p
l
o
c
a
l
d
u
c
k
s
i
n
W
e
s
t
Ja
v
a
,
”
I
n
t
e
r
n
a
t
i
o
n
a
l
J
o
u
rn
a
l
o
f
I
n
n
o
v
a
t
i
v
e
S
c
i
e
n
c
e
a
n
d
R
e
se
a
rc
h
T
e
c
h
n
o
l
o
g
y
,
v
o
l
.
5
,
n
o
.
1
0
,
p
p
.
1
1
1
6
–
1
1
2
1
,
2
0
2
0
.
[
2
]
K
.
K
.
Ç
e
v
i
k
,
H
.
E.
K
o
ç
e
r
,
a
n
d
M
.
B
o
ğ
a
,
“
D
e
e
p
l
e
a
r
n
i
n
g
b
a
se
d
e
g
g
f
e
r
t
i
l
i
t
y
d
e
t
e
c
t
i
o
n
,
”
Ve
t
e
ri
n
a
ry
S
c
i
e
n
c
e
s
,
v
o
l
.
9
,
n
o
.
1
0
,
2
0
2
2
,
d
o
i
:
1
0
.
3
3
9
0
/
v
e
t
sc
i
9
1
0
0
5
7
4
.
[
3
]
M
.
W
.
A
h
m
e
d
,
A
.
S
p
r
i
g
l
e
r
,
J.
L.
Em
mert,
R
.
N
.
D
i
l
g
e
r
,
G
.
C
h
o
w
d
h
a
r
y
,
a
n
d
M
.
K
a
mr
u
z
z
a
m
a
n
,
“
N
o
n
-
d
e
s
t
r
u
c
t
i
v
e
d
e
t
e
c
t
i
o
n
o
f
p
r
e
-
i
n
c
u
b
a
t
e
d
c
h
i
c
k
e
n
e
g
g
f
e
r
t
i
l
i
t
y
u
s
i
n
g
h
y
p
e
r
s
p
e
c
t
r
a
l
i
ma
g
i
n
g
a
n
d
ma
c
h
i
n
e
l
e
a
r
n
i
n
g
,
”
S
m
a
r
t
Ag
r
i
c
u
l
t
u
r
a
l
T
e
c
h
n
o
l
o
g
y
,
v
o
l
.
1
0
,
2
0
2
5
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
a
t
e
c
h
.
2
0
2
5
.
1
0
0
8
5
7
.
[
4
]
S
.
S
a
i
f
u
l
l
a
h
a
n
d
A
.
P
.
S
u
r
y
o
t
o
m
o
,
“
C
h
i
c
k
e
n
e
g
g
f
e
r
t
i
l
i
t
y
i
d
e
n
t
i
f
i
c
a
t
i
o
n
u
si
n
g
F
O
S
a
n
d
B
P
-
n
e
u
r
a
l
n
e
t
w
o
r
k
s
o
n
i
m
a
g
e
p
r
o
c
e
ssi
n
g
,
”
J
u
r
n
a
l
RES
T
I
(
Re
k
a
y
a
s
a
S
i
s
t
e
m
d
a
n
T
e
k
n
o
l
o
g
i
I
n
f
o
rm
a
si
)
,
v
o
l
.
5
,
n
o
.
5
,
p
p
.
9
1
9
–
9
2
6
,
O
c
t
.
2
0
2
1
,
d
o
i
:
1
0
.
2
9
2
0
7
/
r
e
st
i
.
v
5
i
5
.
3
4
3
1
.
[
5
]
L.
G
e
n
g
,
Y
.
H
u
,
Z
.
X
i
a
o
,
a
n
d
J.
X
i
,
“
F
e
r
t
i
l
i
t
y
d
e
t
e
c
t
i
o
n
o
f
h
a
t
c
h
i
n
g
e
g
g
s
b
a
se
d
o
n
a
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
u
r
a
l
n
e
t
w
o
r
k
,
”
A
p
p
l
i
e
d
S
c
i
e
n
c
e
s
,
v
o
l
.
9
,
n
o
.
7
,
A
p
r
.
2
0
1
9
,
d
o
i
:
1
0
.
3
3
9
0
/
a
p
p
9
0
7
1
4
0
8
.
Evaluation Warning : The document was created with Spire.PDF for Python.