I
nd
o
ne
s
ia
n J
o
urna
l o
f
E
lect
rica
l En
g
ineering
a
nd
Co
m
p
u
t
er
Science
Vo
l.
21
,
No
.
2
,
Feb
r
u
ar
y
2
0
2
1
,
p
p
.
1
1
7
6
~1
1
8
4
I
SS
N:
2
5
02
-
4
7
5
2
,
DOI
: 1
0
.
1
1
5
9
1
/
i
j
ee
cs.
v
2
1
.
i
2
.
pp
1
1
7
6
-
1
1
8
4
1176
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
ee
cs.ia
esco
r
e.
co
m
Cla
ss
ificatio
n
t
e
c
hniques
’
p
erfor
ma
nce
e
v
a
lua
tion f
o
r
f
a
cia
l
e
x
press
io
n
r
ecog
nition
M
a
y
y
a
da
h R.
M
a
h
m
o
o
d
1
,
M
a
iw
a
n
B
.
Abdu
lra
z
za
q
2
,
Su
b
hi R.
M
.
Z
ee
ba
re
e
3
,
Ab
ba
s
K
h.
I
bra
hi
m
4
,
Rizg
a
r
R.
Z
eba
ri
5
,
H
iv
i Is
m
a
t
Dino
6
1,
2,
4,
6
C
o
m
p
u
ter S
c
ien
c
e
De
p
a
rtm
e
n
t,
F
a
c
u
lt
y
o
f
S
c
ien
c
e
,
Un
iv
e
rsit
y
o
f
Zak
h
o
,
Du
h
o
k
,
Ira
q
3
De
p
a
rt
m
e
n
t
Du
h
o
k
P
o
ly
tec
h
n
ic Un
iv
e
rsit
y
P
re
sid
e
n
c
y
,
Du
h
o
k
P
o
ly
te
c
h
n
ic Un
iv
e
rsity
,
Du
h
o
k
,
Ira
q
5
In
f
o
rm
a
ti
o
n
T
e
c
h
n
o
l
o
g
y
De
p
a
rt
m
e
n
t,
Du
h
o
k
P
o
ly
tec
h
n
ic Un
iv
e
rs
it
y
,
Du
h
o
k
,
Ira
q
Art
icle
I
nfo
AB
ST
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
J
u
l 2
8
,
2
0
20
R
ev
i
s
ed
Sep
7
,
2
0
20
A
cc
ep
ted
Sep
15
,
2
0
2
0
F
a
c
ial
e
x
p
re
stio
n
re
c
o
g
n
it
io
n
a
s
a
re
c
e
n
tl
y
d
e
v
e
lo
p
e
d
m
e
th
o
d
in
c
o
m
p
u
ter
v
isio
n
is
f
o
u
n
d
e
d
u
p
o
n
th
e
id
e
a
o
f
a
n
a
laz
in
g
th
e
f
a
c
ial
c
h
a
n
g
e
s
in
w
h
ich
a
r
e
w
it
n
e
ss
e
d
d
u
e
to
e
m
o
ti
o
n
a
l
im
p
a
c
ts
o
n
a
n
in
d
iv
id
u
a
l.
T
h
is
p
a
p
e
r
p
ro
v
id
e
s
a
p
e
rf
o
r
m
a
n
c
e
e
v
a
lu
a
ti
o
n
o
f
a
se
t
o
f
su
p
e
rv
ise
d
c
las
sif
iers
u
se
d
f
o
r
f
a
c
ia
l
e
x
p
re
ss
io
n
re
c
o
g
n
it
io
n
b
a
se
d
o
n
m
in
i
m
u
m
fe
a
tu
re
s
se
le
c
ted
b
y
c
h
i
-
sq
u
a
re
.
T
h
e
se
f
e
a
tu
re
s
a
re
th
e
m
o
st
ico
n
ic
a
n
d
in
f
lu
e
n
ti
a
l
o
n
e
s
th
a
t
h
a
v
e
tan
g
ib
le
v
a
lu
e
f
o
r
re
su
lt
d
e
rm
in
a
ti
o
n
.
T
h
e
h
ig
h
e
st
ra
n
k
e
d
six
f
e
a
tu
re
s
a
r
e
a
p
p
li
e
d
o
n
six
c
las
sif
i
e
rs
in
c
lu
d
in
g
m
u
lt
i
-
la
y
e
r
p
re
c
e
p
tro
n
,
su
p
p
o
rt
v
e
c
to
r
m
a
c
h
in
e
,
d
e
c
isio
n
tree
,
ra
n
d
o
m
f
o
re
st,
ra
d
ial
b
a
ise
d
f
u
n
c
ti
o
n
,
a
n
d
K
-
Ne
a
re
st
n
e
io
u
g
h
b
o
r
t
o
f
ig
u
re
o
u
t
th
e
m
o
s
t
a
c
c
u
ra
te
o
n
e
w
h
e
n
th
e
m
in
u
m
n
u
m
b
e
r
o
f
f
e
a
tu
re
s are
u
ti
li
z
e
d
.
T
h
is
is
d
o
n
e
v
ia an
a
l
y
z
in
g
a
n
d
a
p
p
ra
isin
g
t
h
e
c
las
si
f
iers
’
p
e
rf
o
r
m
a
n
c
e
.
CK
+
is
u
se
d
a
s
th
e
re
se
a
r
c
h
’s
d
a
tas
e
t
.
ra
n
d
o
m
f
o
re
s
t
w
it
h
th
e
to
tal
a
c
c
u
ra
c
y
ra
ti
o
o
f
9
4
.
2
3
%
i
s
il
lu
stra
ted
a
s
th
e
m
o
st
a
c
c
u
ra
te
c
las
sif
ier
a
m
o
n
g
st t
h
e
re
st.
K
ey
w
o
r
d
s
:
B
ase
f
u
n
ctio
n
Facial
e
x
p
r
ess
io
n
r
ec
o
g
n
i
tio
n
C
h
i
-
s
q
u
ar
e
f
ea
t
u
r
e
s
elec
t
io
n
K
-
Nea
r
est
n
ei
g
h
b
o
r
Mu
lti
-
l
a
y
er
p
er
ce
p
tr
o
n
T
h
is
is
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
:
R
izg
ar
R
.
Z
eb
ar
i
I
n
f
o
r
m
a
tio
n
T
ec
h
n
o
lo
g
y
Dep
a
r
t
m
en
t
Du
h
o
k
P
o
l
y
tech
n
ic
U
n
i
v
er
s
it
y
6
1
za
k
h
o
R
o
ad
,
1
0
0
6
Ma
zi
Qr
Du
h
o
k
,
Ku
r
d
is
ta
n
-
I
r
aq
E
m
ail: r
iz
g
ar
.
r
a
m
ad
h
a
n
@
d
p
u
.
ed
u
.
k
r
d
1.
I
NT
RO
D
UCT
I
O
N
As
a
f
o
r
m
o
f
p
atter
n
r
ec
o
g
n
i
ti
o
n
an
d
m
ac
h
i
n
e
lear
n
i
n
g
,
Faci
al
E
x
p
r
ess
io
n
R
ec
o
g
n
itio
n
i
s
b
ein
g
e
y
e
-
ca
tch
i
n
g
l
y
d
e
v
elo
p
ed
an
d
h
as
p
lay
ed
a
n
i
m
p
o
r
tan
t
r
o
le
in
th
e
f
ac
ilit
a
tio
n
o
f
f
o
r
en
s
ic
p
r
o
s
i
d
u
r
es
th
r
o
u
g
h
f
ac
e
r
ec
o
g
n
itio
n
an
d
l
ie
d
etec
tio
n
.
Si
m
u
lta
n
o
u
s
l
y
,
th
e
d
ev
elo
p
m
e
n
t o
f
th
e
f
ield
h
a
s
h
ad
p
r
o
m
i
n
en
t p
o
s
iti
v
e
i
m
p
ac
t
s
o
n
v
ar
io
u
s
m
ed
ical
f
ield
s
.
W
ith
t
h
e
e
m
er
g
e
n
ce
o
f
th
e
co
m
p
licated
v
i
s
io
n
ar
y
h
ar
d
w
ar
es
s
u
ch
as
h
i
g
h
d
ef
in
i
tio
n
ca
m
er
a
s
,
th
e
f
ie
ld
s
tar
ted
a
n
e
w
p
r
ac
tical
p
h
ase
th
r
o
u
g
h
o
u
t
its
f
lo
u
r
is
h
m
en
t
s
in
c
e
it
w
as
p
r
ev
io
u
s
l
y
d
is
cu
s
s
ed
m
er
el
y
i
n
ac
ad
e
m
ic
ar
en
a
[
1
,
2
]
.
Dea
lin
g
w
it
h
F
E
R
,
d
if
f
er
en
t
m
e
th
o
d
s
ar
e
b
ei
n
g
s
u
g
g
es
ted
f
r
o
m
b
o
th
s
eq
u
e
n
cial
a
n
d
s
tatic
i
m
ag
es
[
3
]
.
Dea
lin
g
w
i
th
FER
i
n
ter
m
s
o
f
ap
p
lied
f
ield
s
a
n
d
ac
ca
d
em
ia
,
d
if
f
er
en
t
p
r
o
j
ec
ts
ar
e
b
ein
g
i
m
p
le
m
e
n
ted
.
T
h
ese
ed
u
ca
tio
n
al
en
d
e
av
o
r
s
ar
e
b
ein
g
i
m
p
le
m
e
n
te
d
in
f
ield
s
o
f
f
ac
e
d
etec
tio
n
an
d
tr
ac
in
g
,
ex
p
r
ess
i
o
n
class
i
f
icat
io
n
,
an
d
f
ea
tu
r
e
e
x
tr
ac
tio
n
[
4
,
5
]
.
Su
d
h
a
et
al.
[
6
]
,
o
f
f
er
ed
an
e
m
o
tio
n
d
etec
tio
n
p
r
o
g
r
a
m
t
h
a
t
ev
alu
ated
a
m
o
b
ile
p
h
o
n
e
u
s
er
’
s
f
ac
ial
ex
p
r
ess
i
o
n
.
T
h
e
f
r
a
m
e
w
o
r
k
w
a
s
ap
p
lied
f
o
r
o
p
er
atin
g
r
e
g
io
n
s
elec
tio
n
a
n
d
f
ac
e
-
r
e
g
is
tr
atio
n
f
o
r
co
m
p
ac
t
f
ac
ial
A
ctio
n
Un
it
r
ep
r
esen
ta
ti
o
n
.
A
ctio
n
Un
i
t
class
i
f
icat
io
n
an
d
in
telli
g
e
n
t
m
ap
p
i
n
g
o
f
t
h
e
p
r
ed
icted
A
Us
to
d
etec
t
e
m
o
tio
n
a
l
tr
ait
s
.
T
h
e
s
y
s
te
m
u
tili
ze
d
d
atab
a
s
es
s
u
c
h
a
s
Mi
n
d
R
ea
d
i
n
g
,
M
u
lti
P
ie,
J
AFFE,
F
AC
S,
I
S
L
,
an
d
C
K+
.
Ku
m
b
h
ar
a
n
d
P
atil
[
7
]
illu
s
tr
ated
t
h
at
th
e
ap
p
licatio
n
o
f
th
e
ex
p
lici
t
f
ac
ial
p
a
r
ts
p
r
o
v
id
es
m
o
r
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
d
o
n
esi
a
n
J
E
lec
E
n
g
&
C
o
m
p
Sci
I
SS
N:
2502
-
4752
C
la
s
s
i
fica
tio
n
tech
n
iq
u
es’
p
erf
o
r
ma
n
ce
ev
a
lu
a
tio
n
fo
r
fa
cia
l
ex
p
r
ess
io
n
r
ec
o
g
n
itio
n
(
R
iz
g
a
r
R
.
Zeb
a
r
i
)
1177
ac
cu
r
ac
y
t
h
a
n
t
h
e
h
o
lis
t
ic
a
p
p
r
o
ac
h
es.
T
h
e
r
esear
ch
er
s
a
p
p
lied
HOG
to
o
b
tain
f
ea
t
u
r
es
f
r
o
m
t
h
e
f
ac
ial
p
latf
o
r
m
p
o
in
ts
i
n
s
tead
o
f
co
n
s
id
er
in
g
th
e
w
h
o
le
f
ac
e
f
o
r
co
n
s
tr
u
ctin
g
a
v
i
g
o
r
o
u
s
s
y
s
te
m
.
T
h
e
au
th
o
r
s
elab
o
r
ated
Gab
o
r
f
ilter
f
o
r
f
ea
tu
r
e
ex
tr
ac
tio
n
co
m
b
in
ed
w
it
h
f
ee
d
f
o
r
w
ar
d
n
e
u
r
al
n
e
t
w
o
r
k
s
f
o
r
to
r
ec
o
g
n
ize
7
e
m
m
o
t
io
n
al
e
x
p
r
ess
io
n
s
f
r
o
m
h
u
m
a
n
cap
t
u
r
ed
i
m
ag
e
s
.
T
h
e
r
esear
ch
il
lu
s
tr
ated
6
0
%
to
7
0
%
r
ate
o
f
f
ac
ial
r
ec
o
g
n
itio
n
ac
cu
r
ac
y
r
ate
f
o
r
th
e
w
h
o
le
J
A
FEE
d
atab
ase.
Sin
g
h
an
d
Naso
z
[
8
]
d
em
o
n
s
tr
ated
an
o
r
ig
in
al
m
et
h
o
d
to
en
h
a
n
ce
f
u
t
u
r
e
ac
cu
r
ac
y
co
n
c
er
n
in
g
th
e
p
r
o
ce
s
s
f
f
ac
e
r
ec
o
g
n
itio
n
v
ia
r
ep
r
o
ce
s
s
in
g
.
I
t
e
m
b
ed
s
t
w
o
m
aj
o
r
s
tep
s
o
f
f
ac
e
d
etec
tio
n
an
d
co
r
r
ec
tio
n
o
f
th
e
illu
m
i
n
atio
n
.
I
n
o
r
d
er
to
ex
tr
ac
t
th
e
m
o
s
t
i
m
p
o
r
tan
t
f
ac
ial
e
x
p
r
ess
io
n
f
ea
tu
r
e
s
,
f
ea
tu
r
e
e
x
t
r
ac
tio
n
is
ap
p
lied
.
T
h
e
r
ea
s
er
ch
er
s
,
in
2
0
1
3
,
m
an
ag
ed
to
ac
h
iev
e
6
1
.
7
%
o
f
ac
c
u
r
ac
y
r
atio
f
o
r
Face
E
x
p
r
ess
io
n
R
ec
o
g
n
itio
n
in
7
class
es
v
ia
C
o
n
v
u
la
tio
n
al
Neu
r
al
Net
w
o
r
k
cla
s
s
i
f
icat
io
n
co
m
p
ar
i
n
g
to
7
5
.
2
% o
f
ac
c
u
r
ac
y
g
ai
n
ed
b
y
State
-
of
-
A
r
t
c
lass
if
ier
.
Z
h
o
n
g
et
al.
[
9
]
s
u
g
g
e
s
ted
a
n
o
v
el
Face
E
x
p
r
ess
io
n
R
ec
o
g
n
i
tio
n
b
ased
o
n
s
u
b
s
p
ac
e
lear
n
in
g
o
n
lo
ca
l
s
tr
u
ct
u
r
e.
B
o
th
m
e
th
o
d
s
m
an
a
g
ed
to
ac
h
iev
e
h
i
g
h
er
ac
cu
r
ac
y
r
ate
f
r
o
m
t
h
e
o
th
er
tr
a
d
itio
n
al
ap
p
r
o
ac
h
es
w
h
e
n
ap
p
lied
o
n
C
AS
-
P
E
AR
L
-
R
1
an
d
AR
d
atab
ad
es.
B
ilk
h
u
et
al.
[
1
0
]
s
u
g
g
ested
a
n
e
w
m
et
h
o
d
o
f
Face
E
x
p
r
ess
i
o
n
R
ec
o
g
n
itio
n
i
n
o
r
d
er
to
r
ec
o
g
n
ize
s
i
x
e
m
o
tio
n
a
l
f
ea
tu
r
es.
Usi
n
g
t
h
r
ee
class
i
f
ier
s
,
C
a
s
ca
d
e
R
e
g
r
ess
io
n
T
r
ee
is
u
s
ed
f
o
r
th
e
p
u
r
p
o
s
e
o
f
f
ea
t
u
r
e
ex
tr
ac
tio
n
in
t
h
is
m
eth
o
d
.
T
h
e
s
y
s
te
m
ap
p
lies
L
o
g
is
tic
R
eg
r
ess
io
n
,
S
u
p
p
o
r
t
Vec
to
r
M
ac
h
in
e,
an
d
Neu
r
al
Net
w
o
r
k
class
i
f
ier
s
.
C
o
m
p
ar
i
s
o
n
o
f
th
e
ap
p
lied
class
i
f
ier
s
,
w
h
e
n
u
tili
ze
d
C
K+
,
ill
u
s
tr
ate
s
SVM
as
th
e
m
o
s
t
ac
cu
r
ate
clas
s
i
f
ier
w
i
th
to
tal
a
cc
u
r
ac
y
r
atio
o
f
8
9
%.
Ne
u
r
al
Net
w
o
r
k
co
m
e
s
to
th
e
s
ec
o
n
d
r
an
k
w
it
h
8
0
%
o
f
ac
cu
r
ac
y
,
an
d
7
7
.
0
6
p
er
ce
n
t o
f
ac
cu
r
ac
y
is
ac
h
ie
v
ed
f
o
r
L
o
g
i
s
tic
R
e
g
r
ess
io
n
.
T
h
e
cu
r
r
en
t
r
esear
ch
s
u
g
g
est
s
an
in
f
l
u
en
tial
m
et
h
o
d
o
f
class
i
f
icatio
n
an
d
f
ea
t
u
r
e
s
elec
tio
n
f
o
r
Facial
E
x
p
r
ess
io
n
R
ec
o
g
n
it
io
n
f
r
o
m
s
eq
u
en
ce
f
ac
ia
l
i
m
ag
e
s
.
T
o
ac
h
iev
e
th
is
g
o
al,
s
i
x
cla
s
s
i
f
ier
s
ar
e
b
ein
g
ap
p
lied
an
d
th
e
ir
p
er
f
o
r
m
a
n
ce
s
ar
e
c
o
m
p
ar
ed
to
d
is
t
in
g
u
is
h
t
h
e
m
o
s
t
ac
c
u
r
ate
o
n
e.
T
h
r
ee
f
u
n
d
a
m
e
n
tal
s
ta
g
es
o
f
class
i
f
icatio
n
,
f
ac
e
ex
tr
ac
tio
n
,
an
d
f
ea
t
u
r
e
e
x
tr
ac
tio
n
ar
e
s
u
g
g
ested
t
h
r
o
u
g
h
Face
E
x
p
r
ess
i
o
n
R
ec
o
g
n
it
io
n
[
1
1
,
1
2
]
.
A
s
a
co
r
al
p
h
a
s
e,
f
ea
t
u
r
e
s
elec
tio
n
is
d
o
n
e
to
e
lecit
t
h
e
m
o
s
t
i
m
p
o
r
tan
t
f
ea
t
u
r
es
t
h
at
w
o
u
ld
h
a
v
e
th
e
m
o
s
t
s
ig
n
i
f
ica
n
ce
i
m
p
ac
t
v
al
u
e
[
1
3
]
in
w
h
ich
f
ac
to
r
s
s
u
c
h
as
co
m
p
u
tatio
n
al
p
er
f
o
r
m
an
ce
a
n
d
th
e
class
if
icatio
n
co
m
p
ete
n
ce
ar
e
m
a
n
ip
u
lated
b
y
[
1
4
,
1
5
]
.
Natu
r
all
y
,
th
e
in
f
o
r
m
atio
n
th
at
i
s
as
s
o
ciate
d
w
it
h
t
h
e
s
elec
ted
f
ea
t
u
r
es
ar
e
f
air
e
n
o
u
g
h
f
o
r
an
ac
cu
r
ate
d
eter
m
i
n
atio
n
o
f
th
e
in
p
u
t
cla
s
s
.
I
n
o
t
h
er
w
o
r
d
s
,
s
elec
tin
g
t
h
e
h
i
g
h
es
t
r
an
k
ed
f
ea
t
u
r
es
a
n
d
eli
m
i
n
at
io
n
o
f
th
e
u
n
n
ec
es
s
ar
y
f
ea
tu
r
es
w
o
u
ld
m
ak
e
t
h
e
w
h
o
le
p
r
o
ce
s
s
less
ti
m
e
-
co
n
s
u
m
i
n
g
an
d
ea
s
ier
th
r
o
u
g
h
th
e
tr
ain
i
n
g
an
d
tes
tin
g
s
ta
g
e.
Feat
u
r
e
s
elec
tio
n
m
et
h
o
d
s
th
a
t
ap
p
l
y
m
o
d
est
d
iv
is
io
n
o
f
p
atter
n
s
,
h
av
in
g
a
p
o
s
itio
n
w
ith
d
i
f
f
er
en
t
clas
s
e
s
,
f
ail
i
n
clas
s
if
icatio
n
ta
s
k
s
w
it
h
r
ep
ea
ted
ed
g
es
an
d
co
m
p
licated
d
i
s
tr
ib
u
tio
n
t
ec
h
n
iq
u
es
[
1
6
]
.
T
h
e
s
u
g
g
e
s
te
d
ap
p
r
o
ac
h
ap
p
lies
C
h
i
-
Sq
u
ar
e
to
s
elec
t
t
h
e
le
s
t
n
u
m
b
er
o
f
f
ea
t
u
r
es
t
h
at
h
a
v
e
m
o
s
t
s
ig
n
i
f
ica
n
t
v
alu
e
s
.
Af
t
er
ca
r
r
y
i
n
g
o
u
t
t
h
e
p
r
o
ce
s
s
o
f
tr
y
a
n
d
test
,
t
h
e
s
tr
ateg
y
c
h
o
s
es
t
h
e
f
ir
s
t
6
f
ea
tu
r
es
th
at
co
u
ld
b
e
co
n
s
id
er
e
d
as
t
h
e
m
o
s
t
i
m
p
o
r
tan
t
attr
ib
u
tes
o
f
th
e
p
ictu
r
es
th
at
ar
e
ca
p
tu
r
ed
f
r
o
th
e
in
d
iv
id
u
als
f
ac
e
s
.
T
h
e
ch
o
s
en
attr
ib
u
tes
ar
e
u
tili
ze
d
to
class
if
y
t
h
e
f
ac
ial
e
m
o
t
io
n
al
s
tates
[
1
7
]
v
ia
R
an
d
o
m
Fo
r
est
,
Mu
lti
-
L
a
y
er
P
r
ec
ep
tr
o
n
,
Su
p
p
o
r
t
Vec
to
r
Ma
ch
in
e
[
1
8
]
,
Dad
ial
B
ase
Fu
n
ctio
n
[
1
9
]
, K
-
Nea
r
est Ne
ig
h
b
o
r
[
2
0
]
,
an
d
Dec
is
io
n
T
r
ee
[
2
1
]
class
if
ier
s
.
2.
D
AT
AS
E
T
C
K+
d
ataset
is
o
n
e
o
f
t
h
e
m
o
s
t
ap
p
lied
b
an
k
s
o
f
in
f
o
r
m
a
t
io
n
t
h
at
i
s
w
id
el
y
u
s
ed
u
n
iv
e
r
s
all
y
th
at
e
m
b
r
ac
es
5
9
3
i
m
ag
e
cl
u
s
ter
s
o
f
1
2
3
in
d
iv
id
u
al
s
w
h
o
ar
e
m
o
s
tl
y
ad
u
l
t
an
d
m
at
u
r
e
[
2
2
]
.
E
ig
h
t
b
a
s
ic
f
ac
ial
ex
p
r
ess
io
n
o
f
a
n
g
r
y
,
d
is
g
u
s
te
d
,
h
ap
p
y
,
s
u
r
p
r
is
ed
,
s
ad
s
ca
r
e
d
co
n
te
m
p
an
d
n
o
r
m
al
ar
e
r
ec
o
r
d
ed
in
th
e
i
m
ag
e
s
[
2
3
]
.
Fro
m
all
t
h
e
p
ict
u
r
es
t
h
at
ar
e
in
cl
u
d
ed
in
t
h
e
d
atase
t,
3
1
%
ar
e
o
f
m
ale
g
e
n
d
er
b
etw
ee
n
1
8
to
5
0
y
ea
r
s
o
ld
an
d
6
9
%
ar
e
f
e
m
ales
f
r
o
m
t
h
e
s
a
m
e
ag
e
r
a
n
g
e
[
2
4
]
.
As
f
ar
a
s
it
is
co
n
ce
r
en
ed
w
i
th
t
h
e
et
h
n
ical
d
iv
er
s
it
y
o
f
th
e
d
ataset,
8
1
%,
1
3
%,
a
n
d
6
%
ar
e
r
esp
ec
tiv
el
y
f
o
r
E
u
r
o
-
Am
er
ica
n
,
Af
r
o
-
Am
er
ica
n
an
d
an
d
o
t
h
er
r
ac
ial
b
ac
k
g
r
o
u
n
d
s
.
I
n
th
e
r
esp
ec
te
d
d
ataset,
2
0
p
eo
p
le
p
r
o
j
ec
t
2
ex
p
r
ess
io
n
s
,
2
6
r
ef
lect
4
f
ac
ial
ex
p
r
ess
io
n
s
,
2
2
h
av
e
all
t
h
e
f
ac
ial
ex
p
r
es
s
io
n
s
,
8
in
d
iv
id
u
a
ls
s
h
o
w
3
ex
p
r
ess
io
n
s
,
an
d
f
i
n
all
y
2
8
p
e
o
p
le
d
em
o
n
s
tr
ate
5
e
m
o
tio
n
al
r
e
f
lectio
n
s
.
I
n
C
K+
d
ataset,
d
if
f
er
en
t
ill
u
m
in
atio
n
is
b
ein
g
ap
p
lied
an
d
th
e
p
ictu
r
es
’
r
eso
lu
tio
n
i
s
6
4
0
x
4
9
0
.
T
h
e
cu
r
r
en
t
s
t
u
d
y
h
as
ad
o
p
ted
4
0
9
0
in
s
tan
ce
s
f
r
o
m
C
K+
.
So
m
e
in
s
ta
n
ce
s
ad
o
p
ted
b
y
t
h
e
r
esear
c
h
ar
e
b
ein
g
illu
s
tr
ated
in
Fi
g
u
r
e
1
[
2
5
]
.
T
ab
le
1
d
em
o
n
s
tr
ates
a
n
u
m
b
er
o
f
s
a
m
p
le
s
th
at
il
l
u
s
tr
ate
ea
ch
f
ac
ial
ex
p
r
ess
io
n
i
n
C
K+
d
ataset.
3.
M
E
T
H
O
DO
L
O
G
Y
T
h
e
cu
r
r
en
t
ex
p
er
i
m
en
t
atte
m
p
ts
to
p
r
o
v
id
e
it
s
r
ea
d
er
s
w
it
h
a
f
ac
i
al
e
x
p
r
ess
io
n
r
ec
o
g
n
it
io
n
m
et
h
o
d
in
o
r
d
er
to
r
ec
o
g
n
ize
th
e
m
ai
n
8
e
m
o
tio
n
al
p
r
o
j
ec
tio
n
s
.
T
o
m
ee
t
th
e
s
t
u
d
y
g
o
al
s
,
th
e
r
es
ea
r
ch
is
ca
r
r
ied
o
u
t
th
r
o
u
g
h
f
o
u
r
m
ai
n
p
h
ases
t
h
at
ar
e
Data
P
r
o
ce
s
s
in
g
,
Face
d
etec
tio
n
,
Featu
r
e
Selectio
n
,
an
d
C
lass
o
f
icat
io
n
,
as
illu
s
tr
ated
in
Fi
g
u
r
e
2
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
5
0
2
-
4752
I
n
d
o
n
esia
n
J
E
lec
E
n
g
&
C
o
m
p
Sci,
Vo
l.
21
,
No
.
2
,
Feb
r
u
ar
y
2
0
2
1
:
1
1
7
6
-
1
1
8
4
1178
Fig
u
r
e
1
.
Sa
m
p
le
o
f
ex
p
r
es
s
io
n
s
f
r
o
m
C
K+
d
ataset
[
2
6
]
T
ab
le
1
.
T
h
e
n
u
m
b
er
o
f
r
an
d
o
m
s
elec
ted
e
x
p
r
i
s
s
io
n
i
n
s
ta
n
ce
s
f
r
o
m
C
K
+
d
ataset
No
Ex
p
r
e
ssi
o
n
s
N
o
.
o
f
I
n
st
a
n
c
e
s
1
A
n
g
r
y
5
2
7
2
C
o
n
t
e
m
p
t
47
3
D
i
sg
u
st
3
8
9
4
F
e
a
r
4
5
8
5
H
a
p
p
y
6
1
4
6
N
o
r
mal
9
1
3
7
S
a
d
5
4
0
8
S
u
r
p
r
i
se
d
6
0
2
Fig
u
r
e
2
.
B
lo
ck
d
iag
r
a
m
f
o
r
m
ain
s
tep
s
t
h
at
ap
p
lied
in
clas
s
i
f
icatio
n
s
y
s
te
m
3
.
1
.
Da
t
a
pro
ce
s
s
ing
Facial
E
x
p
r
ess
io
n
R
ec
o
g
n
it
io
n
tr
ig
g
er
s
w
it
h
t
h
e
s
ta
n
d
ar
d
izin
g
o
f
th
e
i
m
a
g
es,
m
ea
n
i
n
g
th
at
th
e
w
o
u
ld
-
b
e
p
r
o
ce
s
s
ed
p
ictu
r
es
h
av
e
to
g
e
t
cle
an
ed
o
f
a
n
y
u
n
w
a
n
ted
n
o
i
s
e
th
at
w
o
u
ld
r
ed
u
ce
th
e
w
o
r
k
’
s
ac
cu
r
ac
y
.
Gen
er
all
y
,
in
d
ata
p
r
o
ce
s
s
in
g
p
h
ase
t
h
e
i
m
a
g
es
ar
e
clea
r
ed
f
r
o
m
th
e
n
o
is
es,
r
esi
ze
d
in
to
id
ea
l
s
izes,
an
d
i
f
n
esa
s
ar
y
o
n
v
er
ted
in
to
ea
s
il
y
p
r
o
ce
s
s
ab
le
d
ata.
T
h
e
c
h
o
s
en
i
m
ag
e
s
f
r
o
m
t
h
e
d
atase
t
ar
e
ch
a
n
g
ed
i
n
to
b
lack
an
d
w
h
i
te
an
d
f
ed
i
n
to
Vio
la
-
J
o
n
es to
d
etec
t th
e
8
e
m
o
tio
n
al
tr
aits
.
3
.
2
.
F
a
ce
det
ec
t
io
n
As
a
w
id
el
y
ap
p
lied
m
e
th
o
d
,
Vio
la
-
J
o
n
es
[
2
7
]
is
co
n
s
id
er
ed
as
o
n
e
o
f
t
h
e
m
o
s
t
v
i
g
o
r
o
u
s
f
ac
e
-
d
etec
tio
n
ap
p
r
o
ac
h
es
b
ec
au
s
e
o
f
it
s
e
x
ce
p
tio
n
al
ac
c
u
r
ac
y
a
n
d
r
o
b
u
s
t
n
ess
[
2
8
]
.
A
d
d
to
t
h
at,
it
is
e
s
p
ec
iall
y
ap
p
lied
f
o
r
r
ea
l
ti
m
e
d
etec
tio
n
[
2
9
]
.
T
h
is
m
et
h
o
d
g
o
es
t
h
r
o
u
g
h
f
o
u
r
s
eq
u
en
t
ial
s
tep
s
o
f
s
elec
tio
n
o
f
Haa
r
f
ea
t
u
r
es
w
h
ic
h
ar
e
ex
tr
ac
ted
th
r
o
u
g
h
co
m
p
le
x
co
m
p
u
ta
tio
n
to
in
teg
r
al
i
m
ag
e
cr
ea
tio
n
,
A
d
ab
o
o
s
t
a
r
e
u
s
ed
to
d
is
tin
g
u
is
h
th
e
f
ac
e
f
ea
t
u
r
es
f
r
o
m
t
h
e
r
est,
an
d
ca
s
ca
d
in
g
th
e
s
u
b
w
i
n
d
o
w
s
t
h
at
co
n
tai
n
f
ac
e
i
m
ag
e.
W
h
e
n
th
e
f
ac
e
is
r
ec
o
g
n
ized
,
t
h
e
i
m
ag
e
s
ar
e
cr
o
p
p
e
d
r
esized
to
2
8
x
2
8
.
T
h
en
,
th
e
7
8
4
attr
ib
u
te
s
ar
e
f
ed
to
C
h
i
-
s
q
u
ar
e
f
ea
t
u
r
e
s
elec
tio
n
m
et
h
o
d
t
o
r
an
k
t
h
e
f
ea
t
u
r
es.
T
h
en
,
th
e
m
o
s
t
d
is
t
in
ct
iv
el
y
h
ig
h
r
an
k
e
d
attr
ib
u
tes
w
ill
b
e
elicited
f
r
o
m
t
h
e
r
est i
n
o
r
d
er
t
o
b
e
u
s
ed
in
s
i
x
class
if
ier
s
to
d
is
ti
n
g
u
is
h
t
h
e
m
o
s
t a
cc
u
r
ate
o
n
e.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
d
o
n
esi
a
n
J
E
lec
E
n
g
&
C
o
m
p
Sci
I
SS
N:
2502
-
4752
C
la
s
s
i
fica
tio
n
tech
n
iq
u
es’
p
erf
o
r
ma
n
ce
ev
a
lu
a
tio
n
fo
r
fa
cia
l
ex
p
r
ess
io
n
r
ec
o
g
n
itio
n
(
R
iz
g
a
r
R
.
Zeb
a
r
i
)
1179
3
.
3
.
F
ea
t
ure
s
elec
t
io
n
C
h
i
-
Sq
u
ar
e
test
i
s
u
s
e
f
u
l
i
n
m
ac
h
in
e
lear
n
in
g
.
T
h
is
m
eth
o
d
s
co
r
in
g
th
e
f
ea
t
u
r
es,
p
r
io
r
itize
th
e
f
ea
tu
r
e
s
b
ased
o
n
th
eir
d
iv
er
s
it
y
i
n
th
e
w
h
o
le
d
ataset.
T
h
e
C
h
i
-
Sq
u
ar
e
test
is
ill
u
s
tr
ated
in
(
1
)
:
(
)
(
)
(
)
(
)
(
)
(
)
(
1
)
W
h
en
A
r
ep
r
esen
ts
a
d
o
cu
m
e
n
t
f
r
eq
u
e
n
c
y
th
at
i
n
cl
u
d
es
th
e
v
ar
iav
le
t
a
n
d
b
elo
n
g
s
to
th
e
c
ateg
o
r
y
c,
th
e
B
co
u
ld
b
e
t
h
e
d
o
cu
m
e
n
t
f
r
eq
u
en
c
y
th
a
t
co
n
s
is
t
t
h
e
v
ar
i
ab
le
t
an
d
d
o
en
’
t
b
elo
n
g
to
t
h
e
ca
teg
o
r
y
c.
T
h
en
,
C
s
h
o
w
s
th
e
r
ep
ea
titi
o
n
o
f
th
e
d
o
cu
m
e
n
t
w
h
ich
d
o
es
n
o
t
in
clu
d
e
t
an
d
b
elo
n
g
s
to
s
u
b
s
et
c.
T
h
er
f
o
r
e,
D
r
ep
r
esen
ts
th
e
n
u
m
b
er
o
f
th
e
d
o
cu
m
en
ts
th
a
t
d
o
n
o
t
b
elo
n
g
to
t
an
d
d
o
es
n
o
t
in
clu
d
e
class
c.
Me
an
w
h
ile,
N
p
r
esen
ts
t
h
e
q
u
a
n
tit
y
o
f
t
h
e
d
o
cu
m
e
n
ts
in
th
e
m
as
s
[
3
0
]
.
T
h
e
m
eth
o
d
ap
p
lies
C
h
i
-
Sq
u
ar
e
f
ea
t
u
r
e
s
elec
tio
n
ap
p
r
o
ac
h
o
n
C
K+
an
d
th
e
f
ir
s
t
s
ix
o
u
tr
a
n
k
ed
f
ea
t
u
r
es a
r
e
s
el
ec
ted
as sh
o
w
n
in
T
ab
le
2
.
T
ab
le
2
.
C
hi
-
S
q
u
ar
e
f
ea
t
u
r
e
s
e
lectio
n
F
e
a
t
u
r
e
N
u
m
b
e
r
s
A
n
g
e
r
C
o
n
t
e
m
p
t
D
i
sg
u
st
F
e
a
r
H
a
p
p
y
N
o
r
mal
S
a
d
S
u
r
p
r
i
se
1
95
3
7
8
1
5
0
5
9
5
4
9
9
5
4
5
37
6
0
5
2
96
4
3
1
1
5
1
5
9
6
5
1
0
5
5
0
38
6
2
7
3
1
0
1
4
3
6
1
6
0
6
0
9
5
1
1
5
5
1
39
6
2
8
4
10
2
5
4
6
1
7
7
6
1
0
5
1
2
5
7
0
4
3
3
6
3
3
5
1
03
5
6
5
1
7
8
6
2
3
5
2
6
5
7
1
4
3
7
6
3
4
6
1
3
0
5
9
2
1
7
9
6
3
7
5
2
7
5
7
2
6
0
1
6
5
5
3
.
4
.
Cla
s
s
if
ica
t
io
n
T
h
e
s
tu
d
y
ap
p
l
ie
s
an
d
co
m
p
ar
es
th
e
s
i
x
s
u
p
er
v
is
ed
lear
n
i
n
g
m
ac
h
i
n
es
v
ia
ap
p
licatio
n
o
f
th
e
s
i
x
f
ea
t
u
r
es
ac
h
ie
v
ed
b
y
C
h
i
-
Sq
u
ar
e
ap
p
r
o
ac
h
.
1
0
-
f
o
ld
-
cr
o
s
s
is
ip
lim
e
n
ted
th
r
o
u
g
h
th
e
p
r
o
ce
s
s
o
f
tr
ain
i
n
g
an
d
test
i
n
g
.
T
h
e
ap
p
lied
s
ix
s
u
p
er
v
is
ed
m
ac
h
in
e
s
ar
e
ex
p
lai
n
ed
as:
3
.
4
.
1
.
Deci
s
io
n t
re
e
(
J
4
8
)
J
4
8
is
a
m
o
r
e
d
ev
elo
p
ed
an
d
ad
v
an
ce
d
p
atter
n
o
f
C
4
.
5
alg
o
r
ith
m
.
B
ased
o
n
t
h
e
g
r
o
w
ed
tr
e
e
f
r
o
m
t
h
e
tr
ain
ed
d
ataset,
a
n
e
w
s
a
m
p
le
w
i
ll
b
e
clas
s
if
ied
f
r
o
m
th
e
te
s
ted
d
ataset
b
y
Dec
is
io
n
T
r
ee
.
T
h
e
ti
m
e
s
p
a
n
i
n
w
h
ic
h
i
s
s
p
e
n
t t
h
r
o
u
g
h
tr
ain
in
g
s
e
t d
eter
m
i
n
e
s
t
h
e
attr
ib
u
te
t
h
at
i
s
ac
co
u
n
tab
le
f
o
r
clas
s
i
f
y
i
n
g
v
ar
io
u
s
s
a
m
p
le
s
in
a
s
h
o
r
ter
ti
m
e.
T
h
e
co
n
c
er
n
ed
b
r
an
ch
is
a
m
ar
k
to
ig
n
o
r
e
t
h
e
p
o
s
s
ib
le
f
ea
t
u
r
e
s
i
g
n
i
f
ica
n
ce
w
it
h
n
o
a
m
b
ig
u
it
y
[
3
1
]
.
T
h
is
class
i
f
ie
r
,
d
u
e
to
its
w
id
el
y
u
s
e
a
n
d
b
ein
g
co
n
s
tr
u
ct
ed
u
p
o
n
in
f
o
r
m
atio
n
en
tr
o
p
y
,
i
s
b
eliev
ed
to
b
e
as
th
e
b
est
cla
s
s
if
ier
s
a
m
o
n
g
th
e
s
i
x
alg
o
r
it
h
m
s
[
3
2
]
.
I
n
th
i
s
m
eth
o
d
ea
ch
attr
ib
u
te
o
f
t
h
e
d
ata
co
u
ld
b
e
em
p
lo
y
ed
v
ia
d
is
m
a
n
teli
n
g
in
to
s
m
aller
ele
m
e
n
ts
t
h
at
i
n
clu
d
e
n
o
d
es
o
f
th
e
tr
ee
r
o
o
t.
B
asically
,
ea
c
h
tr
ee
is
f
o
r
m
ed
f
r
o
m
th
r
ee
n
o
d
es
th
at
co
n
s
tr
u
ct
t
h
e
w
h
o
le
tr
ee
an
ato
m
y
:
leav
e
s
,
in
ter
n
al
n
o
d
e,
an
d
th
e
r
o
o
t
n
o
d
e.
I
n
th
is
ap
p
r
o
ac
h
,
a
cla
s
s
lab
el
is
d
eter
m
i
n
ed
b
y
t
h
e
leaf
n
o
d
e
s
in
ce
t
h
e
r
o
o
t
n
o
d
e
h
as
n
o
in
co
m
i
n
g
ed
g
es.
J
4
8
g
r
an
tees sati
s
f
ac
to
r
y
o
u
tco
m
e
s
f
r
o
m
all
t
h
e
ed
g
es
o
f
th
e
co
n
s
tr
u
cted
tr
ee
.
3
.
4
.
2
.
K
-
nea
re
s
t
neig
hb
o
rs
(
K
NN)
K
-
Nea
r
est
Nei
g
h
b
o
u
r
class
i
f
ie
r
is
d
esig
n
ed
to
f
u
n
ctio
n
b
ase
d
o
n
m
ea
s
u
r
i
n
g
t
h
e
d
is
tan
ce
.
I
t
atte
m
p
ts
to
class
i
f
y
n
u
m
er
al
d
ata
r
ec
o
r
d
s
af
ter
f
i
n
d
in
g
t
h
e
K
-
Nea
r
es
t
n
ei
g
h
b
o
r
v
ia
m
ea
s
u
r
in
g
t
h
e
d
is
tan
ce
b
et
w
ee
n
th
e
tr
ain
i
n
g
s
a
m
p
les
an
d
th
e
test
s
a
m
p
les
ac
co
r
d
in
g
to
E
u
c
lid
ian
[
3
3
]
.
I
n
th
is
ap
p
r
o
ac
h
,
t
h
e
o
u
tp
u
t
i
s
cla
s
s
-
lab
led
an
d
K
is
u
s
u
all
y
a
p
o
s
iti
v
e
a
n
d
in
teg
er
d
i
g
it
o
f
its
n
ei
g
h
b
o
r
s
.
A
s
a
m
p
le
i
s
clas
s
i
f
ied
ac
co
r
d
in
g
to
t
h
e
r
ec
ei
v
ed
v
o
tes
r
e
ce
iv
ed
f
r
o
m
t
h
e
n
ei
g
h
b
o
r
s
ac
co
r
d
in
g
to
th
e
n
atu
r
e
a
n
d
lo
ca
tio
n
o
f
th
e
attr
ib
u
te.
I
n
o
th
er
w
o
r
d
s
,
if
a
test
ed
s
a
m
p
le
s
h
ar
e
s
p
ar
ticu
la
r
attr
ib
u
tes
o
f
a
cla
s
s
,
co
n
s
id
er
in
g
t
h
e
d
is
ta
n
ce
o
f
t
h
e
s
a
m
p
l
e
to
th
e
class
,
t
h
e
n
th
e
s
a
m
p
le
w
ill
b
e
class
if
ied
t
o
th
at
ca
t
e
g
o
r
y
[
3
4
]
.
A
s
s
h
o
w
n
i
n
(
2
)
d
e
m
o
n
s
tr
ates
ca
lcu
lati
o
n
o
f
th
e
E
u
clid
ia
n
d
is
tan
ce
:
(
)
√
∑
(
)
(
2
)
3
.
4
.
3
.
Ra
dia
l
ba
s
e
f
un
ct
io
n
(
RB
F
)
B
ein
g
an
o
p
ti
m
al
n
et
w
o
r
k
f
o
r
in
p
u
t
an
d
o
u
tp
u
t
m
ap
p
in
g
o
p
er
atio
n
s
in
an
y
f
ee
d
f
o
r
w
ar
d
n
et
w
o
r
k
,
R
ac
ial
B
ase
F
u
n
ct
io
n
is
h
i
g
h
l
y
ac
cu
r
ate
a
n
d
f
ast
co
n
v
er
g
an
t
[
3
5
]
.
I
t
is
a
ty
p
e
o
f
ar
ti
f
icial
n
eu
r
al
n
et
w
o
r
k
t
h
a
t
co
u
l
d
b
e
u
tili
ze
d
f
o
r
m
u
lt
i
-
cla
s
s
i
f
icatio
n
ca
s
es t
h
at
i
n
cl
u
d
es
th
r
ee
-
la
y
er
an
d
n
o
n
-
li
n
ea
r
clas
s
if
ier
s
.
R
ad
ial
B
ase
Fu
n
ctio
n
co
n
s
i
s
t
a
t
w
o
-
la
y
er
e
d
n
eu
r
al
n
et
w
o
r
k
w
h
ic
h
ea
ch
p
ar
t
f
u
n
ctio
n
s
a
cir
cu
lar
ac
tiv
ated
o
p
er
atio
n
.
T
h
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
5
0
2
-
4752
I
n
d
o
n
esia
n
J
E
lec
E
n
g
&
C
o
m
p
Sci,
Vo
l.
21
,
No
.
2
,
Feb
r
u
ar
y
2
0
2
1
:
1
1
7
6
-
1
1
8
4
1180
o
u
tp
u
t
p
ar
ts
ap
p
ly
a
w
eig
h
te
d
s
u
m
o
f
h
id
d
en
s
eg
m
e
n
t
o
u
tp
u
t.
I
t
is
i
m
p
o
r
tan
t
to
n
o
te
th
at
in
R
ad
ial
B
ase
Fu
n
ctio
n
t
h
e
o
u
tp
u
t is li
n
ea
r
a
n
d
th
e
i
n
p
u
t i
s
n
o
n
-
li
n
ea
r
[
3
6
]
.
3
.
4
.
4
.
Su
pp
o
rt
v
ec
t
o
r
m
a
chine (
S
V
M
)
T
h
is
alg
o
r
ith
m
d
u
e
to
its
s
tr
o
n
g
co
m
p
atib
ilit
y
to
b
e
u
tili
ze
d
in
m
ac
h
in
e
lear
n
in
g
ca
s
es
o
f
h
i
g
h
q
u
an
tit
y
d
ata
s
u
ch
as
co
m
p
u
t
er
v
is
io
n
a
n
d
p
atter
n
r
ec
o
g
n
i
tio
n
h
a
s
tak
e
n
a
s
p
ec
ial
atten
tio
n
f
r
o
m
s
ch
o
lar
s
[
3
4
]
.
B
ein
g
h
i
g
h
l
y
u
t
ilized
,
th
is
cla
s
s
i
f
ier
atte
m
p
ts
to
b
u
i
l
d
th
e
o
p
ti
m
al
h
y
p
er
p
lan
e
-
li
k
e
m
ar
g
i
n
s
.
Su
p
p
o
r
t
Vec
to
r
Ma
ch
i
n
e
ca
r
r
ies
o
u
t
th
e
f
u
n
c
tio
n
o
f
t
h
e
m
a
x
i
m
izatio
n
o
f
t
h
e
d
is
ta
n
ce
b
et
w
ee
n
th
e
h
y
p
er
p
lan
e
an
d
t
h
e
clo
s
est
tr
ai
n
in
g
s
ata
s
a
m
p
les.
Var
io
u
s
r
esear
c
h
es
h
a
v
e
ap
p
r
o
v
ed
th
at
in
a
li
n
ea
r
l
y
d
i
v
i
d
ab
le
in
s
ta
n
ce
,
th
e
o
p
tim
a
l h
y
p
er
p
la
n
e
w
o
u
ld
p
r
o
v
id
e
m
o
r
e
ac
cu
r
ac
y
w
i
th
al
l so
r
ts
o
f
d
atasets
[
3
7
]
.
3
.
4
.
5
.
M
ulti
la
y
er
perc
ept
ro
n
(
M
L
P
)
I
n
ML
P
ap
p
r
o
ac
h
n
eu
r
o
n
s
ar
e
ar
r
an
g
ed
in
o
n
e
d
ir
ec
tio
n
.
I
n
th
i
s
class
if
ier
,
tr
an
s
m
is
s
io
n
o
f
d
ata
o
cc
u
r
s
th
r
o
u
g
h
t
h
r
ee
la
y
er
s
.
I
n
in
p
u
t
la
y
er
th
e
n
u
m
b
er
o
f
th
e
n
o
d
es
h
a
v
e
in
s
tan
t
r
elatio
n
w
it
h
th
e
n
u
m
b
er
o
f
th
e
f
ea
t
u
r
es
t
h
at
ar
e
s
elec
ted
b
y
t
h
e
o
p
er
ato
r
.
T
h
e
h
id
en
lay
er
th
at
is
r
esp
o
n
s
ib
le
f
o
r
m
a
k
i
n
g
a
b
alan
ce
d
s
t
ate
in
t
h
e
ac
c
u
r
ac
y
o
f
t
h
e
o
u
tp
u
t.
I
n
t
h
is
m
et
h
o
d
,
n
o
d
e
n
u
m
b
er
s
an
d
t
h
e
n
u
m
b
er
o
f
th
e
clas
s
es
ar
e
ev
e
n
a
n
d
th
e
r
elatio
n
s
h
ip
b
et
w
ee
n
th
e
la
y
er
s
ar
e
lab
elled
b
y
ce
r
tain
w
ei
g
h
ts
.
No
d
es
in
M
u
lti
-
L
a
y
er
P
r
e
ce
p
tr
o
n
co
u
ld
ca
r
r
y
o
u
t t
w
o
m
aj
o
r
task
s
t
h
at
ar
e
s
u
m
m
ati
o
n
a
n
d
ac
tiv
at
io
n
[
3
8
]
.
3
.
4
.
6
.
Ra
nd
o
m
f
o
re
s
t
(
RF
)
R
an
d
o
m
Fo
r
est
class
if
ier
is
a
s
u
p
er
v
i
s
ed
m
ac
h
in
e
t
h
at
f
u
n
ct
io
n
s
t
h
r
o
u
g
h
cr
ea
tio
n
an
d
m
er
g
i
n
g
v
ar
io
u
s
d
ec
is
io
n
tr
ee
s
i
n
th
e
f
o
r
m
o
f
a
f
o
r
est.
Utilizatio
n
o
f
t
h
is
m
et
h
o
d
en
ab
les
th
e
o
p
er
ato
r
s
to
g
et
ad
v
an
ta
g
e
o
f
a
p
p
l
y
i
n
g
d
i
f
f
er
en
t
lear
n
i
n
g
m
o
d
el
s
i
n
o
r
d
er
t
o
ad
v
an
ce
ac
cu
r
ac
y
to
a
n
e
w
lev
el.
W
h
at
m
a
k
e
s
th
is
ap
p
r
o
ac
h
d
is
ticti
v
l
y
d
i
f
f
er
en
t
f
r
o
m
o
th
er
lear
n
i
n
g
m
ac
h
i
n
es
i
s
t
h
at
i
n
t
h
is
m
et
h
o
d
r
o
o
t
n
o
d
es
ar
e
lin
k
ed
i
n
a
r
ed
u
n
d
an
t
m
an
n
er
[
3
9
]
.
4.
P
E
RF
O
RM
ANCE E
VA
L
U
AT
I
O
N
AN
D
RE
SUL
T
S
T
h
e
av
er
ag
e
w
eig
h
t
s
ig
n
i
f
ica
n
ce
o
f
T
P
r
atio
,
Pre
cisi
o
n
,
R
ec
all,
FP
r
atio
,
an
d
th
e
p
r
o
ce
s
s
i
n
g
t
i
m
e
o
v
er
s
ec
o
n
d
f
o
r
ea
c
h
i
n
s
ta
n
c
e
o
f
ea
ch
c
lass
if
ier
i
n
r
elatio
n
to
t
h
e
ap
p
lied
f
ea
t
u
r
e
s
ele
ctio
n
m
eth
o
d
s
ar
e
illu
s
tr
ated
in
T
ab
les
3
to
8
,
d
e
m
o
n
s
tr
ate
th
e
C
h
i
-
Sq
u
ar
e
f
e
atu
r
e
s
elec
tio
n
.
T
h
is
p
r
o
ce
s
s
’
s
ai
m
s
to
ca
r
r
y
o
u
t
a
co
m
p
ar
ati
v
e
ac
co
u
n
t
o
f
ea
c
h
class
i
f
ier
i
n
o
r
d
er
to
p
o
in
t
o
u
t
th
e
m
o
s
t
ac
cu
r
ate
o
n
e.
T
h
e
r
esear
ch
r
es
u
lt
p
o
in
t
s
o
u
t
th
at
w
h
e
n
ad
o
p
tin
g
t
h
e
s
i
x
attr
ib
u
tes
f
r
o
m
C
h
i
-
Sq
u
ar
e
m
et
h
o
d
,
R
an
d
o
m
Fo
r
e
s
t
ac
h
ie
v
es
th
e
h
i
g
h
e
s
t
r
an
k
a
m
o
n
g
all
th
e
ex
p
er
i
m
en
ted
class
i
f
ier
s
w
it
h
th
e
to
tal
r
atio
o
f
9
4
.
2
3
%.
Me
an
w
h
ile
KN
N
ac
h
iev
e
s
th
e
to
tal
ac
cu
r
ac
y
r
atio
o
f
9
4
.
1
8
%.
I
n
t
h
e
o
th
er
h
an
d
,
R
a
n
d
o
m
Fo
r
est
g
ets
th
e
t
h
ir
d
r
an
k
i
n
ter
m
o
f
ac
cu
r
ac
y
w
it
h
t
h
e
to
tal
r
atio
o
f
9
2
.
8
7
%.
Fu
r
t
h
er
m
o
r
e,
M
L
P
,
SVM,
an
d
R
B
F
g
o
to
th
e
n
ex
t
le
v
el
w
i
th
th
e
r
es
p
ec
tiv
el
y
ac
h
iev
ed
r
atio
s
o
f
9
2
.
0
9
%,
9
0
.
4
8
%,
an
d
9
0
.
3
8
% o
f
ac
cu
r
ac
y
r
atio
,
as illu
s
tr
ated
in
T
ab
les 4
,
5,
6,
7,
8
,
an
d
9
.
T
ab
le
3
.
E
x
p
r
ess
io
n
s
ac
cu
r
ac
y
f
o
r
J
4
8
class
if
ier
u
s
i
n
g
C
h
i
-
S
q
u
ar
e
Ex
p
r
e
ssi
o
n
s
T
P
R
a
t
e
F
P
R
a
t
e
P
r
e
c
i
si
o
n
R
e
c
a
l
l
F
-
M
e
a
su
r
e
A
c
c
u
r
a
c
y
A
n
g
e
r
0
.
9
7
0
.
4
2
0
.
9
4
0
.
9
9
0
.
9
5
9
1
.
9
3
C
o
n
t
e
m
p
t
1
.
0
0
0
.
6
4
0
.
9
9
1
.
0
0
1
.
0
0
9
9
.
1
2
D
i
sg
u
st
0
.
9
9
0
.
5
1
0
.
9
5
0
.
9
9
0
.
9
7
9
3
.
9
4
F
e
a
r
0
.
9
9
0
.
6
2
0
.
9
3
0
.
9
9
0
.
9
6
9
2
.
4
7
H
a
p
p
y
0
.
9
9
0
.
1
7
0
.
9
7
0
.
9
9
0
.
9
8
9
6
.
2
6
N
o
r
mal
0
.
9
5
0
.
5
8
0
.
8
5
0
.
9
5
0
.
9
0
8
3
.
0
1
S
a
d
0
.
9
8
0
.
6
2
0
.
9
1
0
.
9
8
0
.
9
5
9
0
.
0
2
S
u
r
p
r
i
se
0
.
9
9
0
.
1
9
0
.
9
7
0
.
9
9
0
.
9
8
9
6
.
1
9
A
v
g
.
R
a
t
e
0
.
9
8
0
.
4
7
0
.
9
4
0
.
9
8
0
.
9
6
9
2
.
8
7
T
ab
le
4
.
E
x
p
r
ess
io
n
s
ac
cu
r
ac
y
f
o
r
KNN
class
i
f
ier
u
s
i
n
g
C
hi
-
S
q
u
ar
e
Ex
p
r
e
ssi
o
n
s
T
P
R
a
t
e
F
P
R
a
t
e
P
r
e
c
i
si
o
n
R
e
c
a
l
l
F
-
M
e
a
su
r
e
A
c
c
u
r
a
c
y
A
n
g
e
r
0
.
9
9
0
.
3
5
0
.
9
5
0
.
9
9
0
.
9
7
9
4
.
4
8
C
o
n
t
e
m
p
t
1
.
0
0
0
.
7
5
0
.
9
9
1
.
0
0
1
.
0
0
9
9
.
1
4
D
i
sg
u
st
1
.
0
0
0
.
3
7
0
.
9
6
1
.
0
0
0
.
9
8
9
6
.
1
4
F
e
a
r
1
.
0
0
0
.
6
6
0
.
9
2
1
.
0
0
0
.
9
6
9
2
.
5
4
H
a
p
p
y
0
.
9
9
0
.
2
6
0
.
9
6
0
.
9
9
0
.
9
7
9
5
.
5
8
N
o
r
mal
0
.
9
6
0
.
4
1
0
.
8
8
0
.
9
6
0
.
9
3
8
7
.
9
5
S
a
d
0
.
9
9
0
.
6
6
0
.
9
1
0
.
9
9
0
.
9
5
9
0
.
6
9
S
u
r
p
r
i
se
1
.
0
0
0
.
1
9
0
.
9
7
1
.
0
0
0
.
9
8
9
6
.
9
4
A
v
g
.
R
a
t
e
0
.
9
9
0
.
4
6
0
.
9
4
0
.
9
9
0
.
9
7
9
4
.
1
8
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
d
o
n
esi
a
n
J
E
lec
E
n
g
&
C
o
m
p
Sci
I
SS
N:
2502
-
4752
C
la
s
s
i
fica
tio
n
tech
n
iq
u
es’
p
erf
o
r
ma
n
ce
ev
a
lu
a
tio
n
fo
r
fa
cia
l
ex
p
r
ess
io
n
r
ec
o
g
n
itio
n
(
R
iz
g
a
r
R
.
Zeb
a
r
i
)
1181
T
ab
le
5
.
E
x
p
r
ess
io
n
s
ac
cu
r
ac
y
f
o
r
R
B
F
class
if
ier
u
s
i
n
g
C
hi
-
S
q
u
ar
e
Ex
p
r
e
ssi
o
n
s
T
P
R
a
t
e
F
P
R
a
t
e
P
r
e
c
i
si
o
n
R
e
c
a
l
l
F
-
M
e
a
su
r
e
A
c
c
u
r
a
c
y
A
n
g
e
r
0
.
9
9
0
.
8
3
0
.
8
9
0
.
9
9
0
.
9
4
8
8
.
1
7
C
o
n
t
e
m
p
t
1
.
0
0
0
.
7
7
0
.
9
9
1
.
0
0
1
.
0
0
9
9
.
1
2
D
i
sg
u
st
0
.
9
9
0
.
6
7
0
.
9
3
0
.
9
9
0
.
9
6
9
2
.
9
1
F
e
a
r
0
.
9
9
0
.
7
7
0
.
9
1
0
.
9
9
0
.
9
5
9
0
.
81
H
a
p
p
y
0
.
9
6
0
.
2
9
0
.
9
5
0
.
9
6
0
.
9
5
9
2
.
20
N
o
r
mal
0
.
9
5
0
.
8
0
0
.
8
1
0
.
9
5
0.
87
78
.
41
S
a
d
0
.
9
9
0
.
8
7
0.
88
0
.
9
9
0
.
9
3
87
.
75
S
u
r
p
r
i
se
1
.
0
0
0
.
4
1
0
.
9
3
1
.
0
0
0
.
9
6
9
3
.
67
A
v
g
.
R
a
t
e
0
.
9
8
0
.
6
8
0
.
9
1
0
.
9
8
0
.
9
5
9
0
.
38
T
ab
le
6
.
E
x
p
r
ess
io
n
s
ac
cu
r
ac
y
f
o
r
SVM
class
i
f
ier
u
s
i
n
g
C
hi
-
S
q
u
ar
e
Ex
p
r
e
ssi
o
n
s
T
P
R
a
t
e
F
P
R
a
t
e
P
r
e
c
i
si
o
n
R
e
c
a
l
l
F
-
M
e
a
su
r
e
A
c
c
u
r
a
c
y
A
n
g
e
r
1
.
0
0
0
.
9
2
0
.
8
8
1
.
0
0
0
.
9
3
8
7
.
7
8
C
o
n
t
e
m
p
t
1
.
0
0
0
.
7
5
0
.
9
9
1
.
0
0
1
.
0
0
9
9
.
1
0
D
i
sg
u
st
1
.
0
0
0
.
7
9
0
.
9
2
1
.
0
0
0
.
9
6
9
2
.
9
1
F
e
a
r
0
.
9
9
0
.
7
6
0
.
9
1
0
.
9
9
0
.
9
5
9
0
.
9
5
H
a
p
p
y
0
.
9
8
0
.
3
2
0
.
9
5
0
.
9
8
0
.
9
6
9
3
.
7
7
N
o
r
mal
1
.
0
0
1
.
0
0
0
.
7
8
1
.
0
0
0
.
8
7
7
7
.
6
8
S
a
d
1
.
0
0
1
.
0
0
0
.
8
7
1
.
0
0
0
.
9
3
8
6
.
80
S
u
r
p
r
i
se
0
.
9
9
0
.
2
5
0
.
9
6
0
.
9
9
0
.
9
7
9
5
.
4
3
A
v
g
.
R
a
t
e
1
.
0
0
0
.
7
2
0
.
9
1
1
.
0
0
0
.
9
5
9
0
.
4
8
T
ab
le
7
.
E
x
p
r
ess
io
n
s
ac
cu
r
ac
y
f
o
r
ML
P
class
i
f
ier
u
s
i
n
g
C
hi
-
S
q
u
ar
e
Ex
p
r
e
ssi
o
n
s
T
P
R
a
t
e
F
P
R
a
t
e
P
r
e
c
i
si
o
n
R
e
c
a
l
l
F
-
M
e
a
su
r
e
A
c
c
u
r
a
c
y
A
n
g
e
r
0
.
9
7
0
.
5
9
0
.
9
2
0
.
9
7
0
.
9
4
8
9
.
7
6
C
o
n
t
e
m
p
t
1
.
0
0
0
.
7
5
0
.
9
9
1
.
0
0
1
.
0
0
9
9
.
1
4
D
i
sg
u
st
0
.
9
9
0
.
6
0
0
.
9
4
0
.
9
9
0
.
9
7
9
3
.
7
4
F
e
a
r
1
.
0
0
0
.
6
2
0
.
9
3
1
.
0
0
0
.
9
6
9
2
.
6
4
H
a
p
p
y
0
.
9
7
0
.
1
8
0
.
9
7
0
.
9
7
0
.
9
7
9
5
.
1
4
N
o
r
mal
0
.
9
3
0
.
6
1
0
.
8
4
0
.
9
3
0
.
8
9
8
1
.
1
7
S
a
d
0
.
9
9
0
.
7
7
0
.
8
9
0
.
9
9
0
.
9
4
8
8
.
7
3
S
u
r
p
r
i
se
0
.
9
9
0
.
2
0
0
.
9
7
0
.
9
9
0
.
9
6
9
6
.
4
1
A
v
g
.
R
a
t
e
0
.
9
8
0
.
5
4
0
.
9
3
0
.
9
8
0
.
9
6
9
2
.
0
9
T
ab
le
8
.
E
x
p
r
ess
io
n
s
ac
cu
r
ac
y
f
o
r
RF
clas
s
i
f
ier
u
s
in
g
C
hi
-
S
q
u
ar
e
Ex
p
r
e
ssi
o
n
s
T
P
R
a
t
e
F
P
R
a
t
e
P
r
e
c
i
si
o
n
R
e
c
a
l
l
F
-
M
e
a
su
r
e
A
c
c
u
r
a
c
y
A
n
g
e
r
0
.
9
7
0
.
2
7
0
.
9
2
0
.
9
7
0
.
9
7
9
4
.
1
1
C
o
n
t
e
m
p
t
1
.
0
0
0.
64
0
.
9
9
1.
00
1
.
0
0
9
9
.
1
9
D
i
sg
u
st
0
.
9
9
0.
43
0
.
9
6
0
.
9
9
0
.
9
8
9
5
.
43
F
e
a
r
0
.
9
9
0.
58
0
.
9
3
0
.
9
9
0
.
9
6
9
2
.
79
H
a
p
p
y
0
.
9
9
0
.
1
4
0
.
9
8
0
.
9
9
0
.
9
8
9
7
.
07
N
o
r
mal
0
.
9
6
0.
43
0
.
8
9
0
.
9
6
0.
92
8
7
.
24
S
a
d
0
.
9
7
0.
52
0.
93
0
.
9
7
0
.
9
5
90
.
91
S
u
r
p
r
i
se
0
.
9
9
0.
14
0
.
9
8
0
.
9
9
0
.
9
8
9
7
.
09
A
v
g
.
R
a
t
e
0
.
9
8
0.
39
0
.
9
5
0
.
9
8
0
.
9
7
9
4
.
23
T
h
e
u
lti
m
a
te
r
esu
l
t
ac
h
ie
v
ed
f
o
r
C
h
i
-
Sq
u
ar
e
f
ea
t
u
r
e
s
elec
t
i
o
n
ap
p
r
o
ac
h
ap
p
lied
o
n
eig
h
t
t
y
p
es
o
f
e
m
o
tio
n
al
r
e
f
lectio
n
s
o
n
id
i
v
i
d
u
als
f
ac
es d
e
m
o
n
s
tr
ate
t
h
at
K
NN
is
t
h
e
m
o
s
t a
cc
u
r
ate
clas
s
i
f
ier
d
etec
tin
g
a
n
g
er
w
it
h
t
h
e
to
tal
ac
cu
r
ac
y
r
atio
o
f
9
4
.
4
8
%,
d
is
g
u
s
t
g
ai
n
s
t
h
e
to
tal
r
atio
o
f
9
6
.
1
4
%,
an
d
n
o
r
m
al
ac
h
iev
e
s
8
7
.
9
5
%
o
f
ac
cu
r
ac
y
i
n
e
x
p
r
ess
io
n
r
ec
o
g
n
itio
n
.
T
h
is
is
w
h
e
n
co
n
te
m
p
t
ac
h
ie
v
es
9
9
.
1
9
%
o
f
ac
cu
r
ac
y
r
atio
,
f
ea
r
g
et
s
9
2
.
7
9
%,
h
ap
p
y
g
et
s
9
7
.
0
7
%,
s
ad
ac
h
ie
v
es
9
0
.
9
1
%,
an
d
s
u
r
p
r
is
ed
g
ain
s
9
7
.
0
9
%
o
f
ac
cu
r
ac
y
in
e
x
p
r
ess
io
n
d
etec
tio
n
u
tili
z
in
g
R
a
n
d
o
m
Fo
r
est
class
i
f
ier
.
Hen
ce
,
as
f
ar
as
it
is
co
n
ce
r
n
ed
w
it
h
f
ac
ial
e
x
p
r
ess
io
n
r
ec
o
g
n
itio
n
,
R
an
d
o
m
Fo
r
est is
th
e
m
o
s
t a
cc
u
r
ate
clas
s
i
f
ier
f
o
r
C
h
i
-
Sq
u
ar
e.
B
ased
o
n
th
e
d
e
m
o
n
s
tr
ated
o
u
tco
m
es
,
w
h
e
n
u
tili
zi
n
g
th
e
a
lg
o
r
ith
m
s
o
n
th
e
a
ttrib
u
te
s
d
em
o
n
s
tr
ated
f
r
o
m
t
h
e
C
h
i
-
s
q
u
ar
e
f
ea
t
u
r
e
s
elec
tio
n
ap
p
r
o
ac
h
,
t
h
e
e
x
p
r
ess
io
n
o
u
tco
m
e
s
eq
u
en
tia
ll
y
i
n
itiate
s
w
it
h
An
g
er
w
it
h
th
e
h
i
g
h
est
ac
c
u
r
ac
y
r
ate
9
4
.
4
8
%
b
y
KN
N,
an
d
t
h
e
lo
w
e
s
t
r
ate
b
y
SVM
o
f
8
7
.
7
8
%.
C
o
n
te
m
p
t
w
i
th
th
e
h
ig
h
e
s
t
ac
cu
r
ac
y
r
ate
9
9
.
1
9
%
b
y
R
F,
a
n
d
th
e
lo
w
est
r
ate
b
y
SVM
9
9
.
1
0
%.
Dis
g
u
s
t
w
it
h
t
h
e
h
ig
h
es
t
ac
cu
r
ac
y
r
ate
9
6
.
1
4
%
b
y
KNN,
a
n
d
t
h
e
lo
w
e
s
t
r
ate
b
y
SVM
9
2
.
3
5
%.
Fear
w
it
h
t
h
e
h
ig
h
e
s
t
ac
c
u
r
ac
y
r
ate
9
2
.
7
9
%
b
y
R
F,
a
n
d
th
e
lo
w
e
s
t
r
ate
b
y
R
B
F
9
0
.
8
1
%.
Hap
p
y
w
it
h
t
h
e
h
i
g
h
est
ac
c
u
r
ac
y
r
ate
9
7
.
0
7
%
b
y
R
F,
an
d
t
h
e
lo
w
e
s
t
r
ate
b
y
R
B
F
9
2
.
2
0
%.
No
r
m
al
w
it
h
t
h
e
h
i
g
h
e
s
t
ac
cu
r
ac
y
r
at
e
8
7
.
9
5
%
b
y
KNN,
an
d
t
h
e
l
o
w
est
r
ate
b
y
S
VM
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
5
0
2
-
4752
I
n
d
o
n
esia
n
J
E
lec
E
n
g
&
C
o
m
p
Sci,
Vo
l.
21
,
No
.
2
,
Feb
r
u
ar
y
2
0
2
1
:
1
1
7
6
-
1
1
8
4
1182
7
7
.
6
8
%.
Sad
w
i
th
t
h
e
h
i
g
h
est
ac
cu
r
ac
y
r
ate
9
0
.
9
1
%
b
y
R
F
an
d
th
e
lo
w
e
s
t
r
ate
b
y
SV
M
8
6
.
8
0
%.
Su
r
p
r
is
e
w
it
h
t
h
e
h
i
g
h
est ac
c
u
r
ac
y
r
ate
9
7
.
0
9
%
b
y
R
F,
an
d
t
h
e
lo
w
e
s
t
r
ate
b
y
R
B
F 9
3
.
6
7
% a
s
s
h
o
wn
in
Fig
u
r
e
3
.
Fig
u
r
e
3
.
E
x
p
r
ess
io
n
s
p
er
f
o
r
m
an
ce
ac
cu
r
ac
y
r
ate
5.
CO
M
P
ARIS
O
N
WI
T
H
P
RE
VIOU
S WO
RK
S
T
ab
le
9
,
s
h
o
w
s
t
h
e
co
m
p
ar
is
o
n
s
u
m
m
ar
y
b
et
w
ee
n
r
esu
lts
o
b
tain
b
y
o
u
r
p
r
o
p
o
s
ed
s
y
s
te
m
an
d
t
h
o
s
e
p
r
ev
io
u
s
r
elate
d
w
o
r
k
s
.
Fro
m
th
is
tab
le,
it
is
clea
r
t
h
at
t
h
e
s
cien
t
is
t
s
o
f
p
r
ev
io
u
s
clo
s
e
s
t
-
w
o
r
k
s
d
ep
en
d
ed
m
an
y
ap
p
r
o
ac
h
es
o
f
f
ea
tu
r
e
s
elec
tio
n
b
es
id
es
cla
s
s
i
f
icatio
n
an
d
v
ar
io
u
s
d
atase
ts
w
it
h
d
i
f
f
er
en
t
d
ir
ec
tio
n
s
o
f
f
ac
ial
e
x
p
r
ess
io
n
s
.
C
o
m
p
ar
ed
to
th
e
r
elate
d
w
o
r
k
s
,
th
e
p
r
o
v
id
ed
ap
p
r
o
ac
h
o
b
tain
s
a
g
o
o
d
r
ec
o
g
n
itio
n
r
ate
w
it
h
f
e
w
er
f
ea
tu
r
e
s
an
d
m
o
r
e
r
ec
o
g
n
ized
f
ac
ial
ex
p
r
ess
io
n
s
.
Ho
w
e
v
er
,
r
esear
ch
er
s
in
[4
-
6
]
o
b
tain
ed
a
g
o
o
d
r
ec
o
g
n
itio
n
r
ate
r
a
n
g
ed
(
6
0
%
-
7
0
%)
b
u
t
u
s
i
n
g
d
i
f
f
er
e
n
t
n
u
m
b
er
o
f
f
ea
tu
r
e
s
(
1
7
,
2
0
,
8
)
r
esp
ec
tiv
el
y
w
it
h
d
if
f
er
e
n
t
cla
s
s
i
f
ier
s
.
A
l
s
o
,
r
ese
ar
ch
er
s
in
[7
,
8
]
co
u
ld
g
ain
a
h
ig
h
ac
c
u
r
ac
y
u
s
in
g
lar
g
e
n
u
m
b
er
o
f
f
ea
tu
r
e
s
b
u
t
w
it
h
th
e
ab
ilit
y
to
r
ec
o
g
n
ize
f
e
w
er
e
x
p
r
ess
io
n
s
f
r
o
m
d
if
f
er
en
t d
ataset.
T
h
is
w
o
r
k
u
s
e
s
f
e
w
e
s
t
f
ea
tu
r
e
s
w
it
h
s
i
x
class
i
f
ier
s
to
r
ea
ch
e
x
ce
lle
n
t a
cc
u
r
ac
y
.
T
ab
le
9
.
C
o
m
p
ar
is
o
n
b
et
w
ee
n
th
is
w
o
r
k
w
it
h
th
e
o
t
h
er
s
R
e
f
e
r
e
n
c
e
D
a
t
a
se
t
Emo
t
i
o
n
N
o
.
F
e
a
t
u
r
e
N
o
.
F
e
a
t
u
r
e
S
e
l
e
c
t
i
o
n
C
l
a
ssi
f
i
e
r
R
e
su
l
t
[
4
]
C
K
+
,
I
S
L
F
A
C
S
,
JA
F
F
E,
M
i
n
d
R
e
a
d
i
n
g
6
17
S
p
i
n
S
u
p
p
o
r
t
L
G
B
P
,
G
e
o
m
e
t
r
i
c
F
e
a
t
u
r
e
s,
T
e
x
t
u
r
e
F
e
a
t
u
r
e
s
A
d
a
b
o
o
st
,
S
V
M
V
a
l
st
a
r
6
7
.
5
M
o
o
n
7
2
.
7
M
E
R
8
7
.
6
[
4
]
JA
F
EE
6
HOG
F
F
N
N
6
0
%
-
7
0
%
[
4
]
F
ER
2
0
1
3
7
C
N
N
S
t
a
t
e
-
of
-
A
r
t
6
1
.
7
%
7
5
.
2
%
[
4
]
A
r
a
n
d
C
A
S
P
EA
L
-
RI
4
A
r
<
1
2
0
C
A
S
<
6
5
L
TP
NN
A
r
9
9
.
0
%
C
A
S
9
1
.
4
%
[
4
]
C
K
+
5
68
F
ER
S
V
M
,
L
o
g
i
st
i
c
R
e
g
r
e
si
o
n
,
NN
S
V
M
8
9
%
L
o
g
i
st
i
c
R
e
g
r
e
si
o
n
8
0
%
N
N
7
7
.
0
6
%
T
h
i
s w
o
r
k
C
K
+
8
6
C
h
i
-
S
q
u
a
r
e
J4
8
,
K
N
N
,
R
B
F
,
S
V
M
,
M
L
P
,
F
M
J4
8
9
2
.
8
7
%
K
N
N
9
4
.
1
8
%
R
B
F
9
0
.
3
8
%
S
V
M
9
4
.
4
8
%
M
L
P
9
2
.
0
9
%
R
F
9
4
.
2
3
%
6.
CO
NCLU
SI
O
N
E
y
e
-
ca
tc
h
i
n
g
u
tili
za
tio
n
o
f
C
h
i
-
s
q
u
ar
e
b
y
n
u
m
u
r
o
u
s
r
esear
ch
er
s
d
u
r
in
g
th
e
las
t
d
ec
ad
e
h
as
ass
u
r
ed
s
ch
o
lar
s
’
in
te
n
s
e
attr
ac
tio
n
to
w
ar
d
s
t
h
e
ap
p
r
o
ac
h
.
T
h
e
b
est
h
ig
h
r
an
k
ed
s
ix
attr
ib
u
te
s
f
r
o
m
C
h
i
-
Sq
u
ar
e
ar
e
b
ein
g
ap
p
lied
i
n
s
ix
clas
s
i
f
ie
r
s
an
d
t
h
e
e
x
p
er
i
m
e
n
tal
o
u
tc
o
m
e
s
d
e
m
o
n
s
tr
ate
t
h
at
R
F
with
t
h
e
9
4
.
2
3
%
o
f
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
d
o
n
esi
a
n
J
E
lec
E
n
g
&
C
o
m
p
Sci
I
SS
N:
2502
-
4752
C
la
s
s
i
fica
tio
n
tech
n
iq
u
es’
p
erf
o
r
ma
n
ce
ev
a
lu
a
tio
n
fo
r
fa
cia
l
ex
p
r
ess
io
n
r
ec
o
g
n
itio
n
(
R
iz
g
a
r
R
.
Zeb
a
r
i
)
1183
ac
cu
r
ac
y
i
llu
s
tr
ates
th
e
m
o
s
t a
cc
u
r
ate
an
d
e
f
f
icie
n
t c
la
s
s
i
f
ier
in
f
ac
ial
e
x
p
r
ess
io
n
ac
co
r
d
in
g
to
C
K+
d
ataset.
I
n
th
e
o
th
er
h
a
n
d
,
KNN
co
u
ld
b
e
co
n
s
id
er
ed
as
a
s
tr
o
n
g
co
n
te
s
t
an
t
ag
ai
n
s
t
R
F
w
i
th
t
h
e
ac
cu
r
a
c
y
r
ate
o
f
9
4
.
1
8
%.
J
4
8
an
d
ML
P
w
it
h
th
e
r
esp
ec
t
iv
e
ac
cu
r
ac
y
r
ates
o
f
9
2
.
8
7
%
an
d
9
2
.
0
9
%
ar
e
av
er
ag
e
class
i
f
ier
s
.
T
h
e
last
t
wo
class
i
f
ier
s
r
esp
ec
ti
v
el
y
ar
e
S
VM
w
it
h
th
e
ac
cu
r
ac
y
r
atio
o
f
9
0
.
4
8
%
an
d
R
B
F
w
ith
t
h
e
ac
cu
r
ac
y
r
ate
o
f
9
0
.
3
8
%
o
f
ac
cu
r
ac
y
.
RE
F
E
R
E
NC
E
S
[1
]
N.
S
a
m
a
d
ian
i
e
t
a
l.
,
“
A
re
v
i
e
w
o
n
a
u
to
m
a
ti
c
f
a
c
ial
e
x
p
re
s
sio
n
re
c
o
g
n
it
io
n
sy
ste
m
s
a
ss
isted
b
y
m
u
lt
im
o
d
a
l
se
n
so
r
d
a
ta,”
S
e
n
so
rs
,
v
o
l.
1
9
,
n
o
.
8
,
p
.
1
8
6
3
,
2
0
1
9
.
[2
]
S
.
S
h
in
d
o
,
T
.
G
o
to
,
T
.
Kiris
h
im
a
,
a
n
d
K.
T
su
c
h
i
d
a
,
“
A
n
o
p
t
im
iza
t
io
n
o
f
f
a
c
ial
f
e
a
tu
re
p
o
in
t
d
e
tec
ti
o
n
p
ro
g
ra
m
b
y
u
sin
g
se
v
e
ra
l
t
y
p
e
s
o
f
c
o
n
v
o
lu
ti
o
n
a
l
n
e
u
ra
l
n
e
tw
o
rk
,
”
I
n
d
o
n
e
sia
n
J
o
u
rn
a
l
o
f
E
lec
trica
l
En
g
in
e
e
r
in
g
a
n
d
Co
m
p
u
ter
S
c
ien
c
e
,
v
o
l.
1
6
,
n
o
.
2
,
A
rt.
n
o
.
2
,
2
0
1
9
,
d
o
i:
1
0
.
1
1
5
9
1
/i
jee
c
s.v
1
6
.
i2
.
p
p
8
2
7
-
8
3
4
.
[3
]
Din
o
,
Hiv
iIsm
a
t,
M
.
B
.
A
b
d
u
lraz
z
a
q
,
S
.
RM
.
Zee
b
a
re
e
,
A
.
B.
S
a
ll
o
w
,
R
.
R.
Zeb
a
ri,
H
.
M
.
S
h
u
k
u
r
,
a
n
d
L
.
M
.
Ha
ji
.
"
F
a
c
ial
e
x
p
re
ss
io
n
re
c
o
g
n
it
io
n
b
a
se
d
o
n
h
y
b
rid
f
e
a
tu
re
e
x
trac
t
io
n
tec
h
n
iq
u
e
s
w
it
h
d
iff
e
re
n
t
c
l
a
ss
if
i
e
rs
,
"
T
ES
T
En
g
i
n
e
e
rin
g
&
M
a
n
a
g
e
me
n
t
.
v
o
l
.
8
3
,
p
p
.
2
2
3
1
9
-
2
2
3
2
9
,
2
0
2
0
.
[4
]
H.
I.
Din
o
a
n
d
M
.
B
.
A
b
d
u
lraz
z
a
q
,
“
F
a
c
ial
e
x
p
re
ss
io
n
c
las
sif
ic
a
ti
o
n
b
a
se
d
o
n
S
V
M
,
KN
N
a
n
d
M
L
P
c
las
si
f
iers
,
”
in
2
0
1
9
I
n
ter
n
a
ti
o
n
a
l
Co
n
fer
e
n
c
e
o
n
Ad
v
a
n
c
e
d
S
c
ien
c
e
a
n
d
En
g
i
n
e
e
rin
g
(
ICOAS
E)
,
p
p
.
7
0
-
75
,
2
0
1
9
.
[5
]
D.
A
.
Zeb
a
ri,
H.
Ha
ro
n
,
S
.
R.
Zee
b
a
re
e
,
a
n
d
D.
Q.
Zee
b
a
re
e
,
“
En
h
a
n
c
e
th
e
m
a
m
m
o
g
r
a
m
i
m
a
g
e
s
f
o
r
b
o
t
h
se
g
m
e
n
tatio
n
a
n
d
f
e
a
tu
re
e
x
tra
c
ti
o
n
u
sin
g
w
a
v
e
let
tran
sf
o
r
m
,
”
i
n
2
0
1
9
I
n
ter
n
a
ti
o
n
a
l
Co
n
fer
e
n
c
e
o
n
Ad
v
a
n
c
e
d
S
c
ien
c
e
a
n
d
E
n
g
i
n
e
e
rin
g
(
ICOAS
E)
,
p
p
.
1
0
0
-
1
0
5
,
2
0
1
9
.
[6
]
V
.
S
u
d
h
a
,
G
.
V
isw
a
n
a
th
,
A
.
Ba
las
u
b
ra
m
a
n
ian
,
P
.
C
h
iran
jee
v
i,
K.
P
.
Ba
sa
n
t,
a
n
d
M
.
P
ra
t
ib
h
a
,
“
A
f
a
st
a
n
d
ro
b
u
st
e
m
o
ti
o
n
re
c
o
g
n
it
io
n
sy
ste
m
f
o
r
re
a
l
-
w
o
rld
m
o
b
il
e
p
h
o
n
e
d
a
ta,”
i
n
2
0
1
5
I
EE
E
In
ter
n
a
ti
o
n
a
l
Co
n
fer
e
n
c
e
o
n
M
u
lt
ime
d
ia
&
Exp
o
W
o
rk
sh
o
p
s (
ICM
EW
)
,
p
p
.
1
-
6
,
2
0
1
5
.
[7
]
M
.
Ku
m
b
h
a
r,
A
.
Ja
d
h
a
v
,
a
n
d
M
.
P
a
ti
l,
“
F
a
c
ial
e
x
p
re
ss
io
n
re
c
o
g
n
it
io
n
b
a
se
d
o
n
im
a
g
e
f
e
a
tu
re
,
”
In
ter
n
a
ti
o
n
a
l
J
o
u
rn
a
l
o
f
Co
m
p
u
ter
a
n
d
Co
mm
u
n
ica
ti
o
n
En
g
in
e
e
rin
g
,
v
o
l
.
1
,
n
o
.
2
,
p
.
1
1
7
,
2
0
1
2
.
[8
]
S
.
S
in
g
h
a
n
d
F
.
Na
so
z
,
“
F
a
c
ial
e
x
p
re
ss
io
n
re
c
o
g
n
it
io
n
w
it
h
c
o
n
v
o
lu
ti
o
n
a
l
n
e
u
ra
l
n
e
tw
o
rk
s
,
”
i
n
2
0
2
0
1
0
th
A
n
n
u
a
l
Co
mp
u
t
in
g
a
n
d
Co
mm
u
n
ica
ti
o
n
W
o
rk
sh
o
p
a
n
d
Co
n
fer
e
n
c
e
(
CCW
C)
,
p
p
.
0
3
2
4
-
0
3
2
8
,
2
0
2
0
.
[9
]
F
.
Zh
o
n
g
,
S
.
Ya
n
,
L
.
L
iu
,
a
n
d
K.
L
iu
,
“
A
n
e
ff
e
c
ti
v
e
f
a
c
e
re
c
o
g
n
it
i
o
n
f
ra
m
e
w
o
rk
w
it
h
su
b
sp
a
c
e
lea
r
n
in
g
b
a
se
d
o
n
lo
c
a
l
tex
tu
re
p
a
tt
e
rn
s,”
i
n
2
0
1
8
5
th
In
ter
n
a
ti
o
n
a
l
Co
n
fer
e
n
c
e
o
n
S
y
ste
ms
a
n
d
In
fo
rm
a
ti
c
s
(
ICS
AI)
,
p
p
.
2
6
6
-
2
7
1
,
2
0
1
8
.
[1
0
]
M
.
S
.
Bil
k
h
u
,
S
.
G
u
p
ta,
a
n
d
V
.
K.
S
riv
a
sta
v
a
,
“
E
m
o
ti
o
n
c
las
si
fica
ti
o
n
f
ro
m
f
a
c
ial
e
x
p
re
ss
io
n
s
u
sin
g
c
a
sc
a
d
e
d
re
g
re
ss
io
n
tree
s
a
n
d
S
V
M
,
”
in
C
o
mp
u
t
a
ti
o
n
a
l
In
telli
g
e
n
c
e
:
T
h
e
o
r
ies
,
Ap
p
li
c
a
ti
o
n
s
a
n
d
F
u
tu
re
Dire
c
ti
o
n
s
-
Vo
l
u
me
II,
S
p
rin
g
e
r
,
p
p
.
5
8
5
-
5
9
4
,
2
0
1
9
.
[1
1
]
M
.
M
.
Hu
ss
e
in
,
A
.
H.
M
u
tl
a
g
,
a
n
d
H.
S
h
a
re
e
f
,
"
De
v
e
lo
p
e
d
a
rti
f
ici
a
l
n
e
u
ra
l
n
e
tw
o
rk
b
a
s
e
d
h
u
m
a
n
f
a
c
e
re
c
o
g
n
it
io
n
,
"
In
d
o
n
e
sia
n
J
o
u
rn
a
l
o
f
El
e
c
trica
l
En
g
i
n
e
e
rin
g
a
n
d
C
o
mp
u
ter
S
c
ien
c
e
,
v
o
l.
1
6
,
n
o
.
3
,
p
p
.
1
2
7
9
-
1
2
8
5
,
2
0
1
9
.
[1
2
]
M
.
Z.
Os
m
a
n
,
M
.
A
.
M
a
a
ro
f
,
M
.
F
.
Ro
h
a
n
i,
N.
N.
A
.
S
jarif
,
a
n
d
N.
S
.
A
.
Zu
lk
if
li
,
“
A
m
u
lt
i
-
c
o
lo
r
b
a
se
d
f
e
a
tu
re
s
f
ro
m
f
a
c
ial
i
m
a
g
e
s
f
o
r
a
u
to
m
a
ti
c
e
th
n
icity
id
e
n
ti
f
ica
ti
o
n
m
o
d
e
l,
”
I
n
d
o
n
e
sia
n
J
o
u
rn
a
l
o
f
El
e
c
trica
l
En
g
i
n
e
e
rin
g
a
n
d
Co
mp
u
ter
S
c
ien
c
e
,
v
o
l.
1
8
,
n
o
.
3
,
A
rt.
pp
.
1
3
8
3
-
1
3
9
0
,
2
0
2
0
,
d
o
i:
1
0
.
1
1
5
9
1
/i
jee
c
s.v
1
8
.
i
3
.
p
p
1
3
8
3
-
1
3
9
0
.
[1
3
]
Z.
Zh
o
u
,
Y.
L
in
,
Z.
Z
h
a
n
g
,
Y.
W
u
,
a
n
d
P
.
Jo
h
n
so
n
,
“
Earth
q
u
a
k
e
d
e
tec
ti
o
n
in
1
D
ti
m
e
-
se
ries
d
a
ta
w
it
h
f
e
a
tu
re
se
lec
ti
o
n
a
n
d
d
icti
o
n
a
ry
lea
rn
in
g
,
”
S
e
ismo
lo
g
ica
l
Res
e
a
rc
h
L
e
tt
e
rs
,
v
o
l.
9
0
,
n
o
.
2
A
,
p
p
.
5
6
3
-
5
7
2
,
2
0
1
9
.
[1
4
]
M
.
R.
M
a
h
m
o
o
d
a
n
d
A
.
M
.
A
b
d
u
laz
e
e
z
,
“
Di
ff
e
re
n
t
m
o
d
e
l
f
o
r
h
a
n
d
g
e
stu
re
re
c
o
g
n
it
io
n
w
it
h
a
n
o
v
e
l
li
n
e
f
e
a
tu
re
e
x
trac
ti
o
n
,
”
in
2
0
1
9
I
n
ter
n
a
ti
o
n
a
l
Co
n
fer
e
n
c
e
o
n
A
d
v
a
n
c
e
d
S
c
ien
c
e
a
n
d
E
n
g
i
n
e
e
rin
g
(
ICOAS
E)
,
p
p
.
5
2
-
57
,
2
0
1
9
.
[1
5
]
R.
Zeb
a
ri,
A
.
A
b
d
u
laz
e
e
z
,
D.
Zee
b
a
re
e
,
D.
Zeb
a
ri,
a
n
d
J.
S
a
e
e
d
,
“
A
c
o
m
p
re
h
e
n
siv
e
re
v
ie
w
o
f
d
im
e
n
sio
n
a
li
ty
re
d
u
c
ti
o
n
tec
h
n
i
q
u
e
s
f
o
r
fe
a
tu
re
se
lec
ti
o
n
a
n
d
f
e
a
tu
re
e
x
tra
c
ti
o
n
,
”
J
o
u
rn
a
l
o
f
Ap
p
li
e
d
S
c
ien
c
e
a
n
d
T
e
c
h
n
o
l
o
g
y
T
re
n
d
s
,
v
o
l
.
1
,
n
o
.
2
,
A
rt.
n
o
.
2
,
2
0
2
0
,
d
o
i:
1
0
.
3
8
0
9
4
/
jas
tt
1
2
2
4
.
[1
6
]
P
.
L
i,
S
.
L
.
P
h
u
n
g
,
A
.
Bo
u
z
e
rd
o
m
,
a
n
d
F
.
H.
C.
T
iv
iv
e
,
“
F
e
a
tu
re
se
lec
ti
o
n
f
o
r
f
a
c
i
a
l
e
x
p
re
ss
io
n
re
c
o
g
n
it
io
n
,
”
i
n
2
0
1
0
2
n
d
E
u
ro
p
e
a
n
W
o
rk
sh
o
p
o
n
Vi
su
a
l
In
f
o
rm
a
ti
o
n
Pro
c
e
ss
in
g
(
EUVI
P)
,
p
p
.
3
5
-
40
,
2
0
1
0
.
[1
7
]
S
a
m
a
r
S
.
M
o
h
a
m
e
d
e
t
a
l
.
,
“
De
e
p
lea
rn
in
g
f
a
c
e
d
e
tec
ti
o
n
a
n
d
re
c
o
g
n
it
io
n
,
”
In
ter
n
a
ti
o
n
a
l
J
o
u
rn
a
l
o
f
A
d
v
a
n
c
e
d
S
c
ien
c
e
a
n
d
T
e
c
h
n
o
l
o
g
y
,
v
o
l.
2
9
,
n
o
.
2
,
p
p
.
1
-
6,
2
0
2
0
.
[1
8
]
M
.
R.
M
a
h
m
o
o
d
a
n
d
A
.
M
.
A
b
d
u
laz
e
e
z
,
“
A
Co
m
p
a
ra
ti
v
e
stu
d
y
o
f
a
n
e
w
h
a
n
d
re
c
o
g
n
it
i
o
n
m
o
d
e
l
b
a
se
d
o
n
li
n
e
o
f
f
e
a
tu
re
s
a
n
d
o
th
e
r
tec
h
n
iq
u
e
s,
”
in
In
ter
n
a
ti
o
n
a
l
Co
n
fer
e
n
c
e
o
f
Relia
b
le
In
f
o
rm
a
ti
o
n
a
n
d
Co
mm
u
n
ica
ti
o
n
T
e
c
h
n
o
l
o
g
y
,
p
p
.
4
2
0
-
4
3
2
,
2
0
1
7
.
[1
9
]
R.
P
.
S
h
e
tt
y
,
A
.
S
a
th
y
a
b
h
a
m
a
,
a
n
d
A
.
A
.
Ra
i,
“
Op
ti
m
ize
d
ra
d
ial
b
a
sis
f
u
n
c
ti
o
n
n
e
u
ra
l
n
e
tw
o
rk
m
o
d
e
l
f
o
r
w
in
d
p
o
w
e
r
p
re
d
ictio
n
,
”
in
2
0
1
6
S
e
c
o
n
d
I
n
ter
n
a
t
io
n
a
l
Co
n
fer
e
n
c
e
o
n
C
o
g
n
it
ive
Co
m
p
u
ti
n
g
a
n
d
In
fo
rm
a
t
io
n
Pro
c
e
ss
in
g
(
CCIP)
,
p
p
.
1
-
6
,
2
0
1
6
.
[2
0
]
M
.
B.
A
b
d
u
lraz
z
a
q
a
n
d
K.
I.
Kh
a
laf
,
“
Ha
n
d
w
rit
ten
n
u
m
e
r
a
ls’
re
c
o
g
n
it
io
n
i
n
k
u
rd
ish
lan
g
u
a
g
e
u
sin
g
d
o
u
b
le
f
e
a
tu
re
se
lec
ti
o
n
,
”
in
2
0
1
9
2
n
d
I
n
ter
n
a
t
i
o
n
a
l
Co
n
fer
e
n
c
e
o
n
E
n
g
in
e
e
rin
g
T
e
c
h
n
o
l
o
g
y
a
n
d
it
s
A
p
p
li
c
a
ti
o
n
s
(
IICET
A),
p
p
.
167
-
1
7
2
,
2
0
1
9
.
[2
1
]
N.
Ja
h
a
n
a
n
d
R.
S
h
a
h
a
riar,
"
P
re
d
ictin
g
f
e
rti
li
z
e
r
trea
t
m
e
n
t
o
f
m
a
ize
u
sin
g
d
e
c
isio
n
tree
a
lg
o
rit
h
m
,
"
I
n
d
o
n
e
sia
n
J
o
u
rn
a
l
o
f
El
e
c
trica
l
En
g
in
e
e
rin
g
a
n
d
Co
m
p
u
ter
S
c
ien
c
e
,
v
o
l.
2
0
,
n
o
.
3
,
p
p
.
1
4
2
7
-
1
4
3
4
,
2
0
2
0
.
[2
2
]
C.
Ji,
S
.
S
o
n
g
,
a
n
d
H.
Xia
,
“
R
e
se
a
rc
h
a
n
d
a
n
a
ly
sis
o
f
lea
rn
in
g
ra
te
o
n
f
a
c
e
e
x
p
re
ss
io
n
re
c
o
g
n
it
io
n
,
”
i
n
2
0
1
9
In
ter
n
a
t
io
n
a
l
C
o
n
fer
e
n
c
e
o
n
E
lec
tro
n
ic E
n
g
i
n
e
e
rin
g
a
n
d
I
n
fo
rm
a
ti
c
s (
EE
I)
,
p
p
.
2
1
7
-
2
2
0
,
2
0
1
9
.
[2
3
]
F
.
W
.
S
m
it
h
a
n
d
S
.
R
o
ss
it
,
“
Id
e
n
t
ify
in
g
a
n
d
d
e
tec
ti
n
g
f
a
c
ial
e
x
p
re
s
sio
n
s
o
f
e
m
o
ti
o
n
in
p
e
rip
h
e
ra
l
v
isio
n
,
”
P
lo
S
o
n
e
,
v
o
l.
1
3
,
n
o
.
5
,
p
p
.
1
-
1
5
,
2
0
1
8
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
5
0
2
-
4752
I
n
d
o
n
esia
n
J
E
lec
E
n
g
&
C
o
m
p
Sci,
Vo
l.
21
,
No
.
2
,
Feb
r
u
ar
y
2
0
2
1
:
1
1
7
6
-
1
1
8
4
1184
[2
4
]
T
.
Ka
n
a
d
e
,
J.
F
.
Co
h
n
,
a
n
d
Y.
T
ian
,
“
Co
m
p
re
h
e
n
siv
e
d
a
tab
a
s
e
f
o
r
fa
c
ial
e
x
p
re
ss
io
n
a
n
a
ly
sis,”
i
n
Pro
c
e
e
d
in
g
s
Fo
u
rt
h
IEE
E
I
n
ter
n
a
t
io
n
a
l
C
o
n
fe
re
n
c
e
o
n
Au
t
o
ma
ti
c
Fa
c
e
a
n
d
G
e
stu
re
Rec
o
g
n
it
i
o
n
(
Ca
t.
No
.
PR
0
0
5
8
0
)
,
p
p
.
4
6
-
53
,
2
0
0
0
.
[2
5
]
P
.
L
u
c
e
y
,
J.
F
.
Co
h
n
,
T
.
K
a
n
a
d
e
,
J.
S
a
ra
g
ih
,
Z.
Am
b
a
d
a
r,
a
n
d
I.
M
a
tt
h
e
w
s,
“
T
h
e
e
x
ten
d
e
d
c
o
h
n
-
k
a
n
a
d
e
d
a
tas
e
t
(c
k
+
):
A
c
o
m
p
lete
d
a
tas
e
t
f
o
r
a
c
ti
o
n
u
n
it
a
n
d
e
m
o
ti
o
n
-
sp
e
c
if
ied
e
x
p
re
ss
io
n
,
”
in
2
0
1
0
iee
e
c
o
mp
u
ter
so
c
iety
c
o
n
fer
e
n
c
e
o
n
c
o
mp
u
ter
v
isio
n
a
n
d
p
a
tt
e
rn
re
c
o
g
n
it
io
n
-
wo
rk
sh
o
p
s
,
p
p
.
9
4
-
1
0
1
,
20
10
.
[2
6
]
D.
Y.
L
il
ian
a
,
“
E
m
o
ti
o
n
re
c
o
g
n
it
i
o
n
f
ro
m
fa
c
ial
e
x
p
re
ss
io
n
u
sin
g
d
e
e
p
c
o
n
v
o
lu
ti
o
n
a
l
n
e
u
ra
l
n
e
tw
o
rk
,
”
in
J
o
u
rn
a
l
o
f
Ph
y
sic
s: Co
n
fer
e
n
c
e
S
e
rie
s
,
v
o
l.
1
1
9
3
,
n
o
.
1
,
p
.
0
1
2
0
0
4
,
2
0
1
9
.
[2
7
]
K.
D.
Is
m
a
e
l
a
n
d
S
.
Irin
a
,
"
F
a
c
e
re
c
o
g
n
it
io
n
u
sin
g
V
i
o
la
-
Jo
n
e
s
d
e
p
e
n
d
i
n
g
o
n
P
y
th
o
n
,
"
I
n
d
o
n
e
s
ia
n
J
o
u
r
n
a
l
o
f
El
e
c
trica
l
En
g
in
e
e
rin
g
a
n
d
Co
m
p
u
ter
S
c
ien
c
e
,
v
o
l.
2
0
,
n
o
.
3
,
p
p
.
1
5
1
3
-
1
5
2
1
,
2
0
2
0
.
[2
8
]
T
.
P
a
u
l
,
U.
A
.
S
h
a
m
m
i
,
M
.
U.
A
h
m
e
d
,
R.
Ra
h
m
a
n
,
S
.
K
o
b
a
sh
i
,
a
n
d
M
.
A
.
R.
A
h
a
d
,
“
A
stu
d
y
o
n
f
a
c
e
d
e
te
c
ti
o
n
u
sin
g
Vio
la
-
Jo
n
e
s
a
lg
o
ri
th
m
in
v
a
rio
u
s
b
a
c
k
g
ro
u
n
d
s,
a
n
g
les
a
n
d
d
istan
c
e
s
,
”
I
n
ter
n
a
t
io
n
a
l
J
o
u
rn
a
l
o
f
Bi
o
me
d
ica
l
S
o
ft
C
o
mp
u
ti
n
g
a
n
d
Hu
ma
n
S
c
ien
c
e
s:
th
e
o
ff
ici
a
l
j
o
u
rn
a
l
o
f
t
h
e
Bi
o
me
d
ica
l
F
u
zz
y
S
y
ste
ms
Asso
c
ia
ti
o
n
,
v
o
l.
2
3
,
n
o
.
1
,
p
p
.
2
7
-
3
6
,
2
0
1
8
.
[2
9
]
M
.
R.
M
a
h
m
o
o
d
,
A
.
M
.
A
b
d
u
laz
e
e
z
,
a
n
d
Z.
Orm
a
n
,
“
D
y
n
a
m
ic
h
a
n
d
g
e
stu
re
re
c
o
g
n
it
io
n
sy
ste
m
fo
r
k
u
rd
is
h
sig
n
lan
g
u
a
g
e
u
sin
g
t
w
o
li
n
e
s
o
f
f
e
a
t
u
re
s
,
”
in
2
0
1
8
In
ter
n
a
ti
o
n
a
l
C
o
n
fer
e
n
c
e
o
n
Ad
v
a
n
c
e
d
S
c
ien
c
e
a
n
d
En
g
in
e
e
rin
g
(
ICOAS
E)
,
p
p
.
4
2
-
47
,
2
0
1
8
.
[3
0
]
J.
S
u
n
,
X
.
Z
h
a
n
g
,
D.
L
iao
,
a
n
d
V
.
C
h
a
n
g
,
“
Ef
f
icie
n
t
m
e
th
o
d
f
o
r
f
e
a
tu
re
se
le
c
ti
o
n
in
tex
t
c
las
sif
i
c
a
ti
o
n
,
”
i
n
2
0
1
7
In
ter
n
a
t
io
n
a
l
C
o
n
fer
e
n
c
e
o
n
En
g
i
n
e
e
rin
g
a
n
d
T
e
c
h
n
o
l
o
g
y
(
ICET
)
,
A
u
g
.
2
0
1
7
,
p
p
.
1
-
6
,
d
o
i:
1
0
.
1
1
0
9
/IC
En
g
T
e
c
h
n
o
l.
2
0
1
7
.
8
3
0
8
2
0
1
.
[3
1
]
H.
Is
m
a
n
to
,
A
.
A
z
h
a
ri,
S
.
S
u
h
a
rt
o
,
a
n
d
L
.
A
rs
y
a
d
,
"
Clas
si
f
ic
a
ti
o
n
o
f
t
h
e
m
a
in
sta
y
e
c
o
n
o
m
ic
re
g
io
n
u
sin
g
d
e
c
isio
n
tree
m
e
th
o
d
,
"
In
d
o
n
e
sia
n
J
o
u
rn
a
l
o
f
El
e
c
trica
l
En
g
i
n
e
e
rin
g
a
n
d
C
o
mp
u
ter
S
c
ien
c
e
,
v
o
l.
1
2
,
n
o
.
3
,
p
p
.
1
0
3
7
-
1
0
4
4
,
2
0
1
8
.
[3
2
]
N.
Bh
a
rg
a
v
a
,
G
.
S
h
a
r
m
a
,
R.
Bh
a
r
g
a
v
a
,
a
n
d
M
.
M
a
th
u
ria,
“
De
c
isio
n
tree
a
n
a
l
y
sis
o
n
j4
8
a
lg
o
rit
h
m
fo
r
d
a
ta m
in
in
g
,
”
Pro
c
e
e
d
in
g
s
o
f
In
ter
n
a
ti
o
n
a
l
J
o
u
rn
a
l
o
f
Ad
v
a
n
c
e
d
Res
e
a
rc
h
in
Co
mp
u
ter
S
c
ien
c
e
a
n
d
S
o
ft
w
a
re
En
g
i
n
e
e
rin
g
,
v
o
l
.
3
,
n
o
.
6
,
p
p
.
1
1
1
4
-
1
1
1
9
,
2
0
1
3
.
[3
3
]
N.
A
li
,
D.
Ne
a
g
u
,
a
n
d
P
.
T
ru
n
d
le,
“
Ev
a
lu
a
ti
o
n
o
f
k
-
n
e
a
re
st n
e
ig
h
b
o
u
r
c
las
sif
i
e
r
p
e
r
f
o
rm
a
n
c
e
f
o
r
h
e
te
ro
g
e
n
e
o
u
s d
a
ta
se
ts,
”
S
N
Ap
p
li
e
d
S
c
ien
c
e
s,
v
o
l.
1
,
n
o
.
1
2
,
p
.
1
5
5
9
,
2
0
1
9
.
[3
4
]
A
.
A
.
S
a
li
h
a
n
d
M
.
B
.
A
b
d
u
lraz
a
q
,
“
Co
m
b
in
i
n
g
b
e
st
f
e
a
tu
re
s
se
lec
ti
o
n
u
si
n
g
th
re
e
c
las
sif
iers
in
in
tr
u
sio
n
d
e
tec
ti
o
n
s
y
ste
m
,
”
in
2
0
1
9
I
n
ter
n
a
t
io
n
a
l
C
o
n
fer
e
n
c
e
o
n
Ad
v
a
n
c
e
d
S
c
ien
c
e
a
n
d
E
n
g
in
e
e
rin
g
(
ICOAS
E)
,
p
p
.
9
4
-
99
,
2
0
1
9
.
[3
5
]
J.
Ch
e
n
,
Q.
L
i,
H.
Wan
g
,
a
n
d
M
.
De
n
g
,
“
A
m
a
c
h
in
e
lea
rn
in
g
e
n
se
m
b
le
a
p
p
ro
a
c
h
b
a
se
d
o
n
ra
n
d
o
m
f
o
re
st
a
n
d
ra
d
ial
b
a
sis f
u
n
c
ti
o
n
n
e
u
ra
l
n
e
tw
o
rk
f
o
r
risk
e
v
a
lu
a
ti
o
n
o
f
re
g
io
n
a
l
f
lo
o
d
d
isa
ste
r:
a
c
a
s
e
stu
d
y
o
f
th
e
Y
a
n
g
tze
Riv
e
r
De
lt
a
,
Ch
in
a
,
”
I
n
ter
n
a
t
io
n
a
l
jo
u
rn
a
l
o
f
e
n
v
iro
n
me
n
t
a
l
re
se
a
rc
h
a
n
d
p
u
b
l
ic h
e
a
lt
h
,
v
o
l
.
1
7
,
n
o
.
1
,
p
.
4
9
,
2
0
2
0
.
[3
6
]
H.
R.
Ra
n
i
a
n
d
A
.
T
.
V
icto
ire,
“
T
ra
in
in
g
ra
d
ial
b
a
sis
f
u
n
c
ti
o
n
n
e
tw
o
rk
s
f
o
r
w
in
d
sp
e
e
d
p
re
d
ict
io
n
u
sin
g
P
S
O
e
n
h
a
n
c
e
d
d
if
fe
re
n
ti
a
l
se
a
rc
h
o
p
ti
m
iz
e
r.
,
”
Pl
o
S
o
n
e
,
v
o
l.
1
3
,
n
o
.
5
,
p
p
.
e
0
1
9
6
8
7
1
–
e
0
1
9
6
8
7
1
,
2
0
1
8
.
[3
7
]
T
.
Le,
D.
T
ra
n
,
W
.
M
a
,
T
.
P
h
a
m
,
P
.
Du
o
n
g
,
a
n
d
M
.
Ng
u
y
e
n
,
“
Ro
b
u
st
su
p
p
o
r
t
v
e
c
to
r
m
a
c
h
in
e
,
”
i
n
2
0
1
4
In
ter
n
a
t
io
n
a
l
J
o
in
t
Co
n
fer
e
n
c
e
o
n
Ne
u
ra
l
Ne
two
rk
s
,
2
0
1
4
,
p
p
.
4
1
3
7
-
4
1
4
4
,
d
o
i:
1
0
.
1
1
0
9
/I
JCN
N.
2
0
1
4
.
6
8
8
9
5
8
7
.
[3
8
]
A
.
A
.
He
id
a
ri,
H.
F
a
ris,
I.
A
lj
a
ra
h
,
a
n
d
S
.
M
irj
a
li
li
,
“
A
n
e
ff
icie
n
t
h
y
b
rid
m
u
lt
il
a
y
e
r
p
e
rc
e
p
tro
n
n
e
u
r
a
l
n
e
tw
o
rk
w
it
h
g
ra
ss
h
o
p
p
e
r
o
p
ti
m
iza
ti
o
n
,
”
S
o
ft
C
o
mp
u
t
in
g
,
v
o
l
.
2
3
,
n
o
.
1
7
,
p
p
.
7
9
4
1
-
7
9
5
8
,
2
0
1
9
.
[3
9
]
Y.
El
y
u
su
f
i
a
n
d
Z.
El
y
u
su
f
i,
“
S
o
c
ial
n
e
tw
o
rk
s
fa
k
e
p
ro
f
il
e
s
d
e
tec
t
io
n
u
sin
g
m
a
c
h
in
e
lea
rn
in
g
a
lg
o
rit
h
m
s
,”
in
T
h
e
Pro
c
e
e
d
in
g
s
o
f
t
h
e
T
h
ird
I
n
ter
n
a
t
io
n
a
l
C
o
n
fer
e
n
c
e
o
n
S
m
a
rt Ci
ty A
p
p
li
c
a
ti
o
n
s,
p
p
.
3
0
-
40
,
2
0
1
9
.
Evaluation Warning : The document was created with Spire.PDF for Python.