I
nte
rna
t
io
na
l
J
o
urna
l o
f
E
lect
rica
l a
nd
Co
m
p
ute
r
E
ng
in
ee
ring
(
I
J
E
CE
)
Vo
l.
9
,
No
.
3
,
J
u
n
e
201
9
,
p
p
.
2
1
9
6
~
2
2
0
4
I
SS
N:
2
0
8
8
-
8708
,
DOI
: 1
0
.
1
1
5
9
1
/
i
j
ec
e
.
v9
i
3
.
p
p
2
1
9
6
-
2204
2196
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ia
e
s
co
r
e
.
co
m/
jo
u
r
n
a
ls
/in
d
ex
.
p
h
p
/
I
JE
C
E
A
d
esig
n of
l
icens
e
p
la
te
r
ecog
nitio
n
s
y
ste
m
u
sing
c
o
nv
o
lutiona
l
n
eura
l
n
etw
o
rk
P
.
M
a
rz
uk
i,
A.
R.
Sy
a
f
ee
za
,
Y.
C.
Wo
ng
,
N.
A.
H
a
m
i
d
,
A.
Nur
Alis
a
,
M
.
M
.
I
bra
hi
m
De
p
a
rt
m
e
n
t
o
f
Co
m
p
u
ter E
n
g
in
e
e
rin
g
,
T
e
c
h
n
ica
l
Un
iv
e
rsit
y
o
f
M
a
la
y
sia
M
a
lac
c
a
,
M
a
la
y
sia
Art
icle
I
nfo
AB
ST
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Sep
9
,
2
0
1
8
R
ev
i
s
ed
Dec
2
1
,
2
0
1
8
A
cc
ep
ted
J
an
1
8
,
2
0
1
9
T
h
is
p
a
p
e
r
p
ro
p
o
se
s
a
n
im
p
ro
v
e
d
Co
n
v
o
l
u
ti
o
n
a
l
Ne
u
ra
l
Ne
tw
o
rk
(CNN
)
a
lg
o
rit
h
m
a
p
p
ro
a
c
h
f
o
r
li
c
e
n
se
p
late
re
c
o
g
n
it
io
n
sy
ste
m
.
T
h
e
m
a
in
c
o
n
tri
b
u
ti
o
n
o
f
th
is
w
o
rk
i
s
o
n
th
e
m
e
th
o
d
o
lo
g
y
to
d
e
ter
m
in
e
th
e
b
e
st
m
o
d
e
l
f
o
r
f
o
u
r
-
lay
e
re
d
CNN
a
rc
h
it
e
c
t
u
re
th
a
t
h
a
s
b
e
e
n
u
se
d
a
s
th
e
re
c
o
g
n
it
io
n
m
e
th
o
d
.
T
h
is
is
a
c
h
iev
e
d
b
y
v
a
li
d
a
ti
n
g
th
e
b
e
st
p
a
ra
m
e
ter
s
o
f
th
e
e
n
h
a
n
c
e
d
S
to
c
h
a
stic
Dia
g
o
n
a
l
L
e
v
e
n
b
e
rg
M
a
rq
u
a
rd
t
(S
DL
M
)
lea
rn
i
n
g
a
lg
o
rit
h
m
a
n
d
n
e
tw
o
rk
siz
e
o
f
CNN
.
S
e
v
e
ra
l
p
re
p
ro
c
e
ss
in
g
a
lg
o
rit
h
m
s
su
c
h
a
s
S
o
b
e
l
o
p
e
ra
to
r
e
d
g
e
d
e
te
c
ti
o
n
,
m
o
rp
h
o
l
o
g
ica
l
o
p
e
ra
ti
o
n
a
n
d
c
o
n
n
e
c
ted
c
o
m
p
o
n
e
n
t
a
n
a
ly
sis
h
a
v
e
b
e
e
n
u
se
d
to
lo
c
a
li
z
e
th
e
li
c
e
n
se
p
late
,
iso
late
a
n
d
se
g
m
e
n
t
th
e
c
h
a
ra
c
ters
r
e
sp
e
c
ti
v
e
l
y
b
e
f
o
re
f
e
e
d
in
g
th
e
in
p
u
t
to
CNN
.
It
is
f
o
u
n
d
th
a
t
th
e
p
ro
p
o
se
d
m
o
d
e
l
is
s
u
p
e
ri
o
r
w
h
e
n
su
b
jec
ted
t
o
m
u
lt
i
-
sc
a
li
n
g
a
n
d
v
a
riatio
n
s
o
f
in
p
u
t
p
a
tt
e
rn
s.
A
s
a
re
su
lt
,
t
h
e
li
c
e
n
se
p
late
p
re
p
r
o
c
e
ss
in
g
sta
g
e
a
c
h
iev
e
d
7
4
.
7
%
a
c
c
u
ra
c
y
a
n
d
CNN
re
c
o
g
n
it
io
n
sta
g
e
a
c
h
iev
e
d
9
4
.
6
%
a
c
c
u
ra
c
y
.
K
ey
w
o
r
d
s
:
C
o
n
v
o
lu
tio
n
al
n
e
u
r
al
n
e
t
w
o
r
k
Dete
ctio
n
L
ice
n
s
e
p
late
P
r
ep
r
o
ce
s
s
in
g
R
ec
o
g
n
itio
n
Co
p
y
rig
h
t
©
2
0
1
9
In
stit
u
te o
f
A
d
v
a
n
c
e
d
E
n
g
i
n
e
e
rin
g
a
n
d
S
c
ien
c
e
.
Al
l
rig
h
ts re
se
rv
e
d
.
C
o
r
r
e
s
p
o
nd
ing
A
uth
o
r
:
A.
R
.
S
y
a
f
ee
za
,
Dep
ar
t
m
en
t o
f
C
o
m
p
u
ter
E
n
g
i
n
ee
r
in
g
,
T
ec
h
n
ical
Un
i
v
er
s
i
t
y
o
f
Ma
la
y
s
ia
Ma
lacc
a,
7
6
1
0
0
Du
r
ian
T
u
n
g
g
al,
Ma
lac
ca
,
Ma
la
y
s
ia.
E
m
ail:
s
y
a
f
ee
za
@
u
te
m
.
ed
u
.
m
y
1.
I
NT
RO
D
UCT
I
O
N
L
ice
n
s
e
p
l
ate
r
ec
o
g
n
itio
n
(
L
P
R
)
s
y
s
te
m
[
1
]
-
[
3
]
h
as
b
ee
n
e
x
ten
s
i
v
el
y
u
ti
lized
as
a
p
ar
t
o
f
r
ea
l
lif
e
ap
p
licatio
n
s
s
u
ch
as
cr
i
m
i
n
al
p
u
r
s
u
i
t,
au
to
m
a
tic
to
ll
co
llecti
o
n
[
4
]
an
d
en
h
an
ci
n
g
o
f
A
u
to
m
ated
E
n
f
o
r
ce
m
e
n
t
S
y
s
te
m
(
A
E
S)
p
er
f
o
r
m
a
n
ce
wh
ich
ai
m
s
to
co
n
tr
o
l
tr
af
f
ic
e
f
f
icien
c
y
.
I
n
ter
m
s
o
f
s
ec
u
r
it
y
,
L
P
R
h
a
s
b
ee
n
u
s
ed
in
tr
af
f
ic
m
a
n
a
g
e
m
e
n
t
to
d
ete
ct
th
e
o
w
n
er
o
f
th
e
ca
r
w
h
o
h
as
b
r
ea
ch
ed
t
h
e
tr
af
f
ic
la
w
s
an
d
to
f
i
n
d
s
to
len
v
eh
ic
les.
L
P
R
s
y
s
te
m
i
s
al
s
o
u
s
ed
f
o
r
ac
ce
s
s
co
n
tr
o
l
to
e
n
ter
a
b
u
ild
in
g
.
T
h
e
Au
to
m
at
ic
L
P
R
s
y
s
te
m
w
as
in
tr
o
d
u
ce
d
in
1
9
7
9
at
th
e
P
o
li
ce
Scien
ti
f
ic
De
v
elo
p
m
e
n
t B
r
an
ch
i
n
U
n
ited
Kin
g
d
o
m
f
o
r
s
ec
u
r
it
y
p
u
r
p
o
s
es.
I
m
ag
e
p
r
o
ce
s
s
i
n
g
i
s
t
h
e
m
ain
tech
n
iq
u
e
to
b
e
u
s
ed
i
n
L
P
R
r
ec
o
g
n
itio
n
s
y
s
te
m
.
Dev
elo
p
i
n
g
t
h
e
L
P
R
s
y
s
te
m
u
s
i
n
g
i
m
a
g
e
p
r
o
ce
s
s
in
g
is
ch
alle
n
g
i
n
g
d
u
e
to
li
m
ite
d
ab
ilit
y
to
d
ea
l
w
it
h
m
u
lti
-
s
c
alin
g
s
i
n
ce
th
e
L
P
R
i
m
a
g
e
ca
n
ap
p
ea
r
to
b
e
d
ir
ty
,
m
o
tio
n
b
lu
r
,
p
o
o
r
r
eso
lu
tio
n
,
p
o
o
r
lig
h
ti
n
g
,
lo
w
co
n
tr
as
t
a
n
d
etc.
T
h
e
licen
s
e
p
late
ca
n
also
ap
p
ea
r
to
b
e
d
i
r
t
y
an
d
m
o
tio
n
b
lu
r
.
T
h
er
e
ar
e
f
iv
e
p
r
i
m
ar
y
s
ta
g
es
to
id
en
t
if
y
a
lice
n
s
e
p
late.
I
n
itiall
y
,
lo
ca
lizatio
n
tech
n
iq
u
e
is
u
s
ed
to
f
in
d
an
d
is
o
lat
e
th
e
licen
s
e
p
late
o
n
th
e
i
n
p
u
t
i
m
a
g
e.
T
h
is
is
f
o
llo
w
ed
b
y
p
late
o
r
ien
tatio
n
to
co
m
p
en
s
ate
th
e
s
k
e
w
co
n
d
iti
o
n
o
f
t
h
e
p
late
a
n
d
r
esi
zin
g
to
ad
j
u
s
t
th
e
d
i
m
en
s
io
n
s
to
t
h
e
r
eq
u
ir
ed
s
ize
r
esp
ec
tiv
el
y
.
T
h
en
,
i
m
a
g
e
n
o
r
m
aliza
t
io
n
w
i
ll
b
e
p
er
f
o
r
m
ed
to
ad
j
u
s
t
t
h
e
b
r
ig
h
t
n
es
s
an
d
co
n
tr
ast
o
f
th
e
i
m
ag
e.
T
h
e
ch
ar
ac
ter
s
eg
m
e
n
tatio
n
i
n
te
n
d
s
to
s
eg
r
eg
ate
i
n
d
iv
id
u
al
ch
ar
ac
ter
fr
o
m
t
h
e
licen
s
e
p
late.
T
h
e
r
ec
o
g
n
itio
n
p
ar
t
o
f
th
e
L
P
R
s
y
s
te
m
h
a
s
al
m
o
s
t
a
r
o
u
tin
e
al
g
o
r
ith
m
.
I
t
i
n
v
o
lv
e
s
ad
ap
tiv
e
th
r
es
h
o
ld
in
g
,
co
m
p
o
n
e
n
t
lab
e
lin
g
,
f
ea
tu
r
e
e
x
tr
ac
tio
n
an
d
class
i
f
icatio
n
.
Am
o
n
g
t
h
e
f
i
v
e
m
aj
o
r
p
ar
ts
,
th
e
ch
ar
ac
ter
r
ec
o
g
n
itio
n
p
r
o
ce
s
s
i
s
th
e
m
o
s
t c
h
alle
n
g
i
n
g
p
ar
t.
T
h
is
i
s
b
ec
au
s
e,
t
h
e
r
ec
o
g
n
itio
n
o
f
th
e
c
h
ar
ac
ter
s
is
h
ig
h
l
y
d
ep
en
d
e
n
t
o
n
th
e
t
y
p
e
o
f
al
g
o
r
ith
m
s
ap
p
lied
in
t
h
e
f
ir
s
t
f
o
u
r
m
aj
o
r
p
ar
ts
.
I
n
f
ac
t,
th
e
s
eg
m
e
n
ted
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2
0
8
8
-
8708
A
d
esig
n
o
f lice
n
s
e
p
la
te
r
ec
o
g
n
itio
n
s
ystem
u
s
in
g
co
n
v
o
lu
tio
n
a
l n
e
u
r
a
l n
etw
o
r
k
(
P
.
Ma
r
z
u
ki
)
2197
ch
ar
ac
ter
s
ca
n
ap
p
ea
r
in
v
ar
io
u
s
lo
o
k
s
.
T
h
er
ef
o
r
e,
a
r
o
b
u
s
t c
h
ar
ac
ter
r
ec
o
g
n
it
io
n
m
et
h
o
d
is
r
eq
u
ir
ed
an
d
C
NN
h
as p
o
s
s
ib
ili
t
y
s
o
lv
e
to
th
e
m
e
n
tio
n
ed
ch
a
llen
g
e
s
.
C
NN
i
s
w
ell
-
k
n
o
w
n
as
a
s
ca
l
e
an
d
r
o
tatio
n
i
n
v
ar
ia
n
t
i
n
p
at
ter
n
r
ec
o
g
n
itio
n
ta
s
k
s
.
C
NN
a
cc
ep
ts
r
a
w
i
m
a
g
es
t
h
at
h
a
v
e
b
ee
n
p
r
ep
r
o
ce
s
s
ed
w
ith
t
h
e
m
i
n
i
m
al
p
r
ep
r
o
ce
s
s
in
g
al
g
o
r
ith
m
an
d
tr
ai
n
th
e
in
p
u
t
s
a
m
p
les
i
n
s
u
p
er
v
i
s
ed
m
o
d
e.
I
t
co
m
b
in
e
s
co
m
p
r
ess
io
n
(
d
i
m
e
n
s
io
n
alit
y
r
ed
u
ctio
n
)
,
f
ea
tu
r
e
ex
tr
ac
tio
n
an
d
class
i
f
icatio
n
p
r
o
ce
s
s
es
in
a
s
i
n
g
le
ar
c
h
itec
tu
r
e.
Un
t
il
n
o
w
,
C
NN
h
as
b
e
en
ap
p
lied
to
v
ar
io
u
s
ap
p
lica
tio
n
s
s
u
c
h
as
f
ac
e
d
etec
ti
on
[
5
]
-
[
1
0
]
,
f
ac
e
r
ec
o
g
n
itio
n
[
1
1
]
-
[
1
5
]
,
g
en
d
er
r
ec
o
g
n
itio
n
[
1
6
]
-
[
1
9
]
,
o
b
j
ec
t
class
i
f
icatio
n
a
n
d
r
ec
o
g
n
itio
n
[
2
0
]
-
[
2
2
]
,
ch
ar
ac
ter
r
ec
o
g
n
itio
n
[
2
3
]
-
[
2
5
]
,
tex
tu
r
e
r
ec
o
g
n
itio
n
[
2
6
]
,
f
in
g
er
-
v
ei
n
[
2
7
]
,
etc.
Desp
ite
th
e
lis
ted
ad
v
a
n
ta
g
es,
C
NN
h
as
li
m
i
tatio
n
s
in
ter
m
s
o
f
co
s
t
an
d
s
p
ee
d
.
T
h
is
is
d
u
e
to
th
e
co
m
p
u
te
in
te
n
s
iv
e
i
m
a
g
e
p
r
o
ce
s
s
in
g
alg
o
r
ith
m
b
ein
g
i
n
co
r
p
o
r
ated
in
th
e
d
esig
n
s
u
ch
as
co
n
v
o
lu
t
io
n
an
d
s
u
b
s
a
m
p
li
n
g
.
T
h
e
co
n
v
o
lu
tio
n
p
r
o
ce
s
s
ta
k
es
al
m
o
s
t
9
0
%
o
f
th
e
p
r
o
ce
s
s
i
n
g
ti
m
e
[
2
8
]
.
T
h
er
ef
o
r
e,
in
o
r
d
er
to
o
v
er
co
m
e
th
e
li
m
ita
tio
n
,
d
esi
g
n
i
n
g
a
s
m
all
C
NN
s
ize
co
u
ld
aid
in
r
ed
u
cin
g
th
e
p
r
o
ce
s
s
i
n
g
ti
m
e.
T
h
e
L
P
R
u
s
in
g
C
NN
h
as
b
ee
n
r
ep
o
r
ted
in
[
2
9
]
.
Ho
w
e
v
er
,
th
e
c
h
ar
ac
ter
s
ar
e
m
a
n
u
all
y
s
eg
m
e
n
te
d
w
h
ile
t
h
e
r
ea
l
p
r
o
b
lem
o
f
L
P
R
s
tar
ted
f
r
o
m
t
h
e
p
r
ep
r
o
ce
s
s
i
n
g
s
ta
g
e.
I
n
[3
0
]
,
th
e
y
i
m
p
le
m
en
ted
L
eNe
t
-
5
ar
ch
itect
u
r
e
w
i
th
7
la
y
er
s
b
y
in
s
er
ti
n
g
th
e
w
h
o
le
lice
n
s
e
p
l
ate
as
in
p
u
t
an
d
r
ep
o
r
ted
9
8
.
2
5
%
ac
cu
r
ac
y
.
T
h
is
w
o
r
k
cla
s
s
i
f
ie
s
b
et
w
ee
n
t
h
e
li
ce
n
s
e
p
late
an
d
n
o
n
-
licen
s
e
p
late
an
d
n
o
t
r
ec
o
g
n
izi
n
g
th
e
c
h
ar
ac
ter
s
.
T
h
ey
u
s
ed
2
4
0
0
lice
n
s
e
p
lates
a
n
d
4
0
0
0
n
o
n
-
lice
n
s
e
p
late
d
ataset
a
n
d
d
iv
id
ed
in
to
tr
ai
n
a
n
d
test
d
ataset.
B
esid
es,
t
h
e
ac
cu
r
ac
y
r
ate
o
n
lice
n
s
e
p
late
d
etec
tio
n
is
in
co
m
p
ar
ab
le
w
i
t
h
t
h
is
w
o
r
k
h
av
e
s
h
o
w
n
t
h
at
s
y
s
te
m
p
er
f
o
r
m
ed
at
th
e
o
th
er
r
esear
ch
e
s
i
m
p
r
o
v
ed
o
n
th
e
p
r
ep
r
o
ce
s
s
i
n
g
p
ar
t to
im
p
r
o
v
e
th
e
r
es
u
lt o
b
tain
ed
o
n
r
ec
o
g
n
itio
n
.
In
[
3
1
]
,
th
e
y
p
r
o
p
o
s
ed
t
w
o
l
o
ca
l
b
in
ar
y
m
et
h
o
d
s
,
w
h
ic
h
a
r
e
lo
ca
l
Ots
u
an
d
a
n
i
m
p
r
o
v
ed
B
er
n
s
e
n
alg
o
r
ith
m
.
C
o
n
n
ec
ted
C
o
m
p
o
n
en
t
A
n
al
y
s
is
(
C
C
A
)
is
u
s
ed
f
o
r
b
in
ar
y
i
m
a
g
e
s
s
ea
r
ch
in
g
in
a
n
eig
h
t
-
co
n
n
ec
tiv
it
y
s
itu
at
io
n
.
B
esid
es,
ac
co
r
d
in
g
to
[
3
2
]
,
th
e
lab
ellin
g
alg
o
r
it
h
m
u
s
e
s
a
‘
4
-
co
n
n
ec
tiv
it
y
’
m
et
h
o
d
to
m
ar
k
t
h
e
g
r
o
u
p
o
f
co
n
n
ec
ted
p
i
x
els
a
n
d
lab
els
th
e
m
u
s
i
n
g
d
if
f
er
e
n
t
n
u
m
b
er
s
.
Fo
r
th
e
r
ec
o
g
n
itio
n
p
ar
t
[
3
3
]
,
u
s
ed
te
m
p
late
m
atc
h
in
g
a
n
d
ac
h
ie
v
ed
t
h
e
ac
cu
r
a
c
y
o
f
8
7
%.
B
ased
o
n
[
3
4
]
,
th
e
ch
ar
ac
ter
r
eg
io
n
is
ca
lcu
lated
b
y
u
s
i
n
g
v
ar
ia
n
ce
p
r
o
j
ec
tio
n
alg
o
r
ith
m
.
T
h
is
is
u
s
ed
to
e
n
h
a
n
ce
its
n
o
is
e
i
m
m
u
n
it
y
a
n
d
i
m
p
r
o
v
e
th
e
s
e
g
m
e
n
tatio
n
ac
c
u
r
ac
y
.
A
n
iter
ati
v
e
m
ea
n
f
i
lter
is
u
s
ed
t
o
s
m
o
o
th
th
e
o
r
i
g
in
a
l
v
er
tica
l
v
ar
ian
ce
p
r
o
j
ec
tio
n
g
r
ap
h
i
n
o
r
d
er
to
f
in
d
th
e
co
r
r
esp
o
n
d
in
g
p
ea
k
to
d
eter
m
i
n
e
t
h
e
n
u
m
b
er
o
f
ch
ar
ac
ter
in
t
h
e
licen
s
e
p
late.
T
h
e
ac
cu
r
ac
y
ac
h
iev
ed
u
n
s
atis
f
ied
.
2.
M
E
T
H
O
DO
L
O
G
Y
2
.
1
.
Da
t
a
ba
s
e
c
o
llect
io
n
I
n
f
ac
t,
th
e
lice
n
s
e
p
late
d
ataset
ar
e
d
if
f
ic
u
lt
to
o
b
tain
s
in
ce
th
eir
p
r
iv
ac
y
co
n
ce
r
n
.
T
h
er
ef
o
r
e
th
e
i
m
a
g
es
o
f
v
e
h
icle
licen
s
e
p
lat
e
ar
e
r
an
d
o
m
l
y
ca
p
t
u
r
ed
ar
o
u
n
d
Ma
lacc
a
ar
ea
as
d
ataset
s
.
T
h
e
i
m
a
g
e
s
ar
e
s
et
u
p
as
R
GB
,
2
5
6
b
it
w
it
h
1
2
8
0
x
8
0
0
r
eso
lu
tio
n
.
T
h
e
d
ataset
ta
k
en
ex
ce
ed
1
0
0
0
o
f
R
GB
i
m
a
g
es.
7
0
0
o
f
1
0
0
0
a
r
e
u
s
ed
as tr
ai
n
in
g
a
n
d
th
e
r
e
m
ai
n
in
g
as te
s
ti
n
g
d
ata
s
et.
2
.
2
.
L
P
R
s
y
s
t
e
m
f
lo
w
cha
rt
A
cc
o
r
d
in
g
to
Fi
g
u
r
e
1
,
it
is
illu
s
tr
ated
th
e
f
lo
w
ch
ar
t
o
f
o
v
er
all
L
P
R
s
y
s
te
m
p
h
ase
s
.
T
h
e
d
etail
alg
o
r
ith
m
o
f
e
n
h
a
n
ce
d
SD
L
M
ca
n
b
e
r
ef
er
r
ed
in
[
3
5
]
.
MA
T
L
A
B
an
d
C
la
n
g
u
a
g
e
h
as
b
ee
n
u
s
ed
as
t
h
e
p
latf
o
r
m
.
T
h
e
o
v
er
all
s
y
s
te
m
co
n
s
i
s
ts
o
f
t
h
r
ee
m
ai
n
p
h
a
s
es:
P
r
ep
r
o
ce
s
s
in
g
,
Seg
m
e
n
tat
io
n
an
d
C
h
ar
ac
t
er
R
ec
o
g
n
itio
n
.
T
h
e
u
n
iq
u
e
n
es
s
o
f
th
i
s
ap
p
r
o
ac
h
co
m
p
ar
ed
to
o
th
er
ex
is
t
in
g
w
o
r
k
s
o
n
Ma
la
y
s
ia
n
’
s
lice
n
s
e
p
lat
e
is
th
e
i
m
p
le
m
e
n
tatio
n
o
f
C
N
N
at
th
e
r
ec
o
g
n
itio
n
p
ar
t.
T
h
e
w
h
o
le
m
et
h
o
d
o
lo
g
y
w
ill
b
e
ex
p
lain
ed
in
n
e
x
t
s
ec
tio
n
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
n
t J
E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
9
,
No
.
3
,
J
u
n
e
2
0
1
9
:
2
1
9
6
-
2
2
0
4
2198
Fig
u
r
e
1
.
Flo
w
c
h
ar
t o
f
L
P
R
s
y
s
te
m
2
.
3
.
P
re
pro
ce
s
s
ing
P
r
ep
r
o
ce
s
s
in
g
i
s
t
h
e
i
n
it
ial
s
ta
g
e
o
f
i
m
a
g
e
p
r
o
ce
s
s
i
n
g
tas
k
s
to
en
h
an
ce
th
e
q
u
al
it
y
o
f
th
e
i
m
a
g
e.
I
n
th
is
s
ta
g
e,
n
o
is
es
ar
e
r
ed
u
ce
d
an
d
u
n
w
a
n
ted
f
ea
t
u
r
es
ar
e
e
li
m
i
n
ated
to
ea
s
e
t
h
e
b
u
r
d
en
o
f
t
h
e
C
NN
at
th
e
r
ec
o
g
n
itio
n
s
ta
g
e.
T
h
e
p
r
ep
r
o
c
ess
i
n
g
s
tep
s
i
n
v
o
lv
e
f
o
r
L
P
R
in
clu
d
es
th
e
f
o
llo
w
in
g
s
eq
u
e
n
ce
.
2
.
3
.
1
.
L
icense pla
t
e
lo
ca
liza
t
io
n
T
h
e
ca
p
tu
r
ed
im
a
g
es
ar
e
in
R
GB
f
o
r
m
at.
T
h
e
i
m
a
g
es
ar
e
co
n
v
er
ted
to
g
r
a
y
s
ca
le
t
o
ea
s
e
th
e
co
m
p
u
tatio
n
al
p
r
o
ce
s
s
(
Fig
u
r
e
2
)
.
A
f
ter
th
at,
t
h
e
g
r
a
y
s
ca
l
e
i
m
ag
e
s
ar
e
p
r
o
ce
s
s
ed
b
y
S
o
b
el
o
p
e
r
ato
r
e
d
g
e
d
etec
tio
n
.
A
t
h
r
es
h
o
ld
o
f
th
e
ed
g
e
d
etec
tio
n
is
s
et
in
o
r
d
er
to
d
ec
r
ea
s
e
th
e
s
e
n
s
it
iv
it
y
.
B
y
d
o
in
g
t
h
is
,
th
e
ed
g
e
s
th
at
ar
e
n
o
t
s
i
g
n
if
ican
t
w
ill
b
e
ig
n
o
r
ed
.
T
w
o
h
is
to
g
r
a
m
g
r
ap
h
s
ar
e
p
r
o
d
u
ce
d
b
ased
o
n
th
e
ed
g
e
d
et
ec
t
io
n
p
r
o
ce
s
s
.
T
h
e
h
is
to
g
r
a
m
g
r
ap
h
r
ep
r
esen
ts
th
e
co
lu
m
n
-
w
i
s
e
a
n
d
r
o
w
-
w
i
s
e
h
is
to
g
r
a
m
o
f
th
e
i
m
a
g
e
as
s
h
o
w
n
in
Fig
u
r
e
3
.
Fig
u
r
e
2
.
C
o
n
v
er
s
io
n
f
r
o
m
R
GB
in
to
g
r
a
y
s
ca
le
i
m
a
g
e
Fig
u
r
e
3
.
L
icen
s
e
p
late
lo
ca
liz
atio
n
u
s
in
g
So
b
el
o
p
er
ato
r
ed
g
e
d
etec
tio
n
(
a)
His
to
g
r
a
m
o
f
co
lu
m
n
-
w
i
s
e
(
b
)
His
to
g
r
a
m
o
f
r
o
w
-
w
i
s
e
L
i
c
e
n
s
e
p
l
a
t
e
l
o
c
a
l
i
z
a
t
i
o
n
-
G
r
a
y
s
c
a
l
e
-
S
o
b
e
l
o
p
e
r
a
t
o
r
e
d
g
e
d
e
t
e
c
t
i
o
n
S
t
a
r
t
L
o
a
d
I
m
a
g
e
C
h
a
r
a
c
t
e
r
I
s
o
l
a
t
i
o
n
-
M
o
r
p
h
o
l
o
g
y
O
p
e
r
a
t
o
r
C
h
a
r
a
c
t
e
r
S
e
g
m
e
n
t
a
t
i
o
n
-
C
o
n
n
e
c
t
e
d
C
o
m
p
o
n
e
n
t
A
n
a
l
y
s
i
s
C
h
a
r
a
c
t
e
r
R
e
c
o
g
n
i
t
i
o
n
-
C
N
N
R
e
c
o
g
n
i
t
i
o
n
S
y
s
t
e
m
E
n
d
P
r
e
p
r
o
c
e
s
s
i
n
g
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2
0
8
8
-
8708
A
d
esig
n
o
f lice
n
s
e
p
la
te
r
ec
o
g
n
itio
n
s
ystem
u
s
in
g
co
n
v
o
lu
tio
n
a
l n
e
u
r
a
l n
etw
o
r
k
(
P
.
Ma
r
z
u
ki
)
2199
I
n
th
e
co
l
u
m
n
-
w
i
s
e
g
r
ap
h
,
t
h
e
h
is
to
g
r
a
m
ca
lcu
la
tes
t
h
e
n
u
m
b
er
o
f
p
ix
els
at
th
e
co
l
u
m
n
p
a
r
t
o
f
th
e
d
iag
r
a
m
.
I
n
th
i
s
h
is
to
g
r
a
m
,
t
h
e
lice
n
s
e
p
late
a
n
d
ch
ar
ac
te
r
ed
g
es
ar
e
tr
ac
ed
.
A
t
h
r
es
h
o
ld
is
s
et
to
i
g
n
o
r
e
th
o
s
e
t
h
at
ar
e
n
o
t
l
icen
s
e
p
l
ate
ca
n
d
id
ates.
I
n
t
h
e
r
o
w
-
wis
e
g
r
ap
h
,
t
h
e
h
ig
h
er
t
h
e
n
u
m
b
er
o
f
ed
g
es
i
n
h
o
r
izo
n
tal
o
f
t
h
e
i
m
ag
e
w
ill
c
r
ea
te
h
ig
h
p
ea
k
s
.
W
it
h
t
h
e
m
e
an
t
h
r
es
h
o
ld
of
ed
g
es,
t
h
e
lo
ca
tio
n
o
f
th
e
licen
s
e
p
late
ca
n
b
e
d
eter
m
i
n
ed
.
T
h
e
f
in
al
o
u
tco
m
e
of
th
e
lice
n
s
e
p
lat
e
lo
ca
lizatio
n
as
s
h
o
w
n
in
Fi
g
u
r
e
4
.
2
.
3
.
2
.
Cha
ra
ct
er
is
o
la
t
i
o
n
W
h
en
t
h
e
lo
ca
tio
n
o
f
t
h
e
lice
n
s
e
p
late
is
d
eter
m
in
ed
,
th
e
h
ei
g
h
t
an
d
w
id
t
h
ar
e
ca
lcu
lated
a
n
d
cr
o
p
p
ed
o
u
t.
T
h
e
licen
s
e
p
late
i
m
a
g
e
is
cr
o
p
p
ed
d
ir
ec
tly
f
r
o
m
t
h
e
o
r
i
g
in
a
l i
n
p
u
t
i
m
a
g
e
w
h
ic
h
i
s
in
R
GB
f
o
r
m
at.
Af
ter
th
at,
t
h
e
licen
s
e
p
late
is
co
n
v
er
ted
in
to
th
e
g
r
a
y
s
ca
le
f
o
r
m
at
an
d
f
i
n
all
y
to
b
i
n
ar
y
f
o
r
m
a
t
f
o
r
m
o
r
p
h
o
lo
g
ical
o
p
er
atio
n
.
Mo
r
p
h
o
lo
g
y
o
p
er
atio
n
ca
n
b
e
u
s
ed
to
r
e
m
o
v
e
t
h
e
u
n
w
an
ted
n
o
is
e
a
n
d
is
o
late
th
e
c
h
ar
ac
ter
s
f
r
o
m
th
e
lice
n
s
e
p
late.
I
n
t
h
e
n
ex
t
s
tep
,
a
r
ec
tan
g
le
s
tr
u
ct
u
r
i
n
g
ele
m
en
t
is
cr
ea
ted
.
T
h
e
d
ilatio
n
a
n
d
er
o
s
io
n
p
r
o
ce
s
s
es
ar
e
u
s
ed
to
s
ep
ar
ate
th
e
f
o
r
eg
r
o
u
n
d
an
d
b
ac
k
g
r
o
u
n
d
p
ix
el
o
f
th
e
licen
s
e
p
late
im
ag
e.
T
h
e
d
ilatio
n
p
r
o
ce
s
s
en
lar
g
es
t
h
e
f
o
r
eg
r
o
u
n
d
p
ix
el
w
h
ile
th
e
er
o
s
io
n
s
h
r
i
n
k
s
t
h
e
f
o
r
eg
r
o
u
n
d
p
ix
el.
B
y
ap
p
ly
in
g
s
u
b
tr
ac
tio
n
o
f
t
h
ese
t
w
o
o
u
tp
u
t
p
r
o
ce
s
s
e
s
,
th
e
ed
g
e
o
f
t
h
e
f
o
r
eg
r
o
u
n
d
p
ix
el
ca
n
be
o
b
tain
ed
.
F
u
r
th
er
m
o
r
e,
th
r
o
u
g
h
co
n
v
o
lu
tio
n
a
n
d
co
n
tr
ast
co
l
o
r
ad
j
u
s
tin
g
,
th
e
f
o
r
eg
r
o
u
n
d
o
b
ject
w
ill
b
e
m
o
r
e
s
ig
n
i
f
i
ca
n
t.
Fin
a
ll
y
,
th
e
ch
ar
ac
ter
of
th
e
lice
n
s
e
p
late
ch
ar
ac
ter
an
d
b
ac
k
g
r
o
u
n
d
ar
ea
can
be
d
if
f
er
en
tiated
an
d
f
ilt
er
ed
o
u
t.
T
h
e
f
i
n
al
o
u
tco
m
e
in
p
r
o
ce
s
s
c
h
ar
ac
ter
is
o
latio
n
is
s
h
o
w
n
i
n
Fi
g
u
r
e
5
.
Fig
u
r
e
4
.
Ou
tp
u
t o
f
licen
s
e
p
late
lo
ca
lizatio
n
Fig
u
r
e
5
.
L
icen
s
e
p
late
is
o
lati
o
n
u
s
i
n
g
Mo
r
p
h
o
lo
g
ical
o
p
er
atio
n
2
.
3
.
3
.
Cha
ra
ct
er
s
eg
m
ent
a
t
i
o
n
Af
ter
i
m
p
le
m
e
n
ti
n
g
th
e
p
r
e
v
io
u
s
s
tep
,
th
e
r
e
m
a
in
i
n
g
ite
m
s
i
n
th
e
i
m
a
g
e
ar
e
o
n
l
y
ch
ar
ac
ter
s
.
B
y
th
i
s
,
C
o
n
n
ec
ted
C
o
m
p
o
n
en
t
An
al
y
s
is
i
s
u
s
ed
to
s
eg
m
e
n
t
th
e
c
h
a
r
ac
ter
s
u
s
in
g
co
n
n
ec
ted
p
i
x
el
in
t
h
e
i
m
a
g
e.
E
ac
h
o
f
t
h
e
ch
ar
ac
ter
s
s
e
g
m
en
ted
in
to
a
n
i
n
d
iv
id
u
al
i
m
a
g
e
f
o
r
r
ec
o
g
n
it
io
n
.
T
h
e
o
u
tp
u
t
o
f
t
h
e
c
h
ar
ac
ter
s
eg
m
e
n
tatio
n
p
r
o
ce
s
s
i
s
s
h
o
w
n
in
Fig
u
r
e
6.
2
.
3
.
4
.
Cha
ra
ct
er
re
s
ize
a
nd
pa
dd
ing
T
h
e
s
eg
m
en
ted
c
h
ar
ac
ter
s
ar
e
b
ein
g
r
esized
in
to
1
8
×8
p
ix
el
s
ize
to
r
ed
u
ce
t
h
e
co
m
p
le
x
it
y
o
f
th
e
i
m
a
g
e.
B
esid
es
th
a
t,
t
h
e
i
m
a
g
e
i
s
f
u
r
t
h
er
p
ad
d
in
g
b
y
2
p
i
x
els
o
n
th
e
s
u
r
r
o
u
n
d
i
n
g
o
f
th
e
i
m
a
g
e
to
b
ec
o
m
e
2
2
×1
2
p
ix
el
i
m
a
g
e.
P
ad
d
in
g
i
s
ca
r
r
ied
o
u
t
to
e
n
s
u
r
e
t
h
at
a
ll
th
e
f
ea
tu
r
es
ar
e
a
v
ailab
le
d
u
r
in
g
t
h
e
r
ec
o
g
n
itio
n
p
r
o
ce
s
s
.
T
h
e
im
a
g
e
af
ter
p
ad
d
in
g
i
s
s
h
o
w
n
i
n
Fi
g
u
r
e
7.
Fig
u
r
e
6
.
E
ac
h
o
f
th
e
c
h
ar
ac
te
r
is
s
eg
m
e
n
ted
in
to
o
n
e
i
m
ag
e
Fig
u
r
e
7
.
C
h
ar
ac
ter
s
i
m
a
g
e
is
r
esized
an
d
p
ad
d
ed
in
to
2
2
×1
2
p
ix
el
i
m
a
g
e
2
.
3
.
5
.
No
rm
a
liza
t
io
n
In
C
NN
r
ec
o
g
n
it
io
n
s
y
s
te
m
,
t
h
e
i
m
a
g
e
is
tr
ai
n
ed
u
s
i
n
g
n
u
m
er
ic
d
ata
o
f
t
h
e
i
m
a
g
e.
T
h
e
ac
ce
p
ted
r
an
g
e
o
f
th
e
n
u
m
er
ic
d
ata
is
f
r
o
m
-
1
to
1
.
T
h
e
m
i
n
-
m
a
x
n
o
r
m
al
izatio
n
is
ap
p
lied
to
th
e
in
p
u
t
i
m
a
g
e
s
to
r
escale
w
i
th
i
n
t
h
e
m
e
n
tio
n
ed
r
an
g
e.
T
h
e
eq
u
at
io
n
o
f
th
e
m
i
n
-
m
ax
n
o
r
m
aliza
t
io
n
is
s
h
o
w
n
b
elo
w
:
2
.
4
.
Cha
ra
c
t
er
re
co
g
nitio
n
C
h
ar
ac
ter
r
ec
o
g
n
itio
n
is
th
e
s
e
co
n
d
s
tag
e
af
ter
p
r
ep
r
o
ce
s
s
in
g
.
I
n
th
is
s
tag
e,
al
l
t
h
e
ch
ar
ac
te
r
s
ar
e
in
n
u
m
er
ic
d
ata
f
o
r
m
a
n
d
ca
n
b
e
th
e
in
p
u
t
f
o
r
C
NN
s
y
s
te
m
.
T
h
e
f
in
al
o
u
tp
u
t
o
f
t
h
e
p
r
ep
r
o
ce
s
s
i
n
g
s
tag
e
is
a
s
et
o
f
2
2
×1
2
p
ix
els
o
f
th
e
i
m
a
g
e
to
f
ee
d
th
e
C
NN
ar
ch
itect
u
r
e.
T
h
e
C
NN
m
o
d
el
u
s
ed
in
t
h
is
w
o
r
k
is
a
f
o
u
r
-
l
a
y
er
ed
ar
ch
itectu
r
e.
T
h
e
f
ir
s
t
an
d
s
ec
o
n
d
la
y
e
r
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
n
t J
E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
9
,
No
.
3
,
J
u
n
e
2
0
1
9
:
2
1
9
6
-
2
2
0
4
2200
ap
p
lied
to
th
e
f
u
s
io
n
m
eth
o
d
p
r
o
p
o
s
ed
b
y
Ma
m
alet
a
n
d
Ga
r
cia.
I
n
o
r
d
er
to
f
i
n
d
t
h
e
b
est
ar
ch
itect
u
r
e
m
o
d
el
an
d
th
e
b
est p
ar
a
m
eter
s
,
1
0
-
f
o
ld
cr
o
s
s
-
v
a
lid
atio
n
tec
h
n
iq
u
e
h
as b
ee
n
u
s
ed
.
A
to
tal
o
f
8
0
%
f
r
o
m
th
e
o
v
er
a
ll
s
a
m
p
le
h
a
v
e
b
ee
n
u
s
ed
f
o
r
1
0
-
f
o
ld
cr
o
s
s
-
v
alid
atio
n
tec
h
n
i
q
u
e.
T
h
e
in
itial
w
ei
g
h
t
u
s
ed
f
o
r
tr
ain
i
n
g
a
n
d
v
al
id
atio
n
ar
e
t
h
e
s
a
m
e
f
o
r
all
e
x
p
er
i
m
e
n
t
to
en
s
u
r
e
a
f
air
e
v
al
u
atio
n
.
T
h
er
e
ar
e
f
e
w
p
ar
a
m
eter
s
te
s
ted
u
s
i
n
g
1
0
-
f
o
ld
cr
o
s
s
v
ali
d
atio
n
.
T
h
e
p
ar
am
eter
s
i
n
cl
u
d
e
th
e
n
u
m
b
er
o
f
f
ea
t
u
r
e
m
ap
at
ea
c
h
la
y
er
,
t
h
e
p
atter
n
o
f
co
n
n
ec
tio
n
o
f
t
h
e
f
i
r
s
t
la
y
er
an
d
s
ec
o
n
d
la
y
er
,
t
y
p
e
of
w
eig
h
t,
v
a
lu
e
o
f
th
e
r
eg
u
lated
p
ar
a
m
eter
an
d
th
e
-
co
n
s
ta
n
t
(
Fi
g
u
r
e
8
)
.
Fig
u
r
e
8
.
Me
an
s
q
u
ar
e
er
r
o
r
v
er
s
u
s
f
ea
t
u
r
e
m
ap
g
r
ap
h
2
.
5
.
L
P
R
s
y
s
t
e
m
dev
elo
ped i
n M
AT
L
AB
G
U
I
I
n
o
r
d
er
to
cr
ea
te
a
u
s
er
f
r
ien
d
ly
L
P
R
s
y
s
te
m
,
th
e
p
r
o
g
r
a
m
is
d
ev
elo
p
ed
in
M
A
T
L
A
B
s
o
f
t
w
ar
e
a
n
d
u
s
i
n
g
a
MA
T
L
A
B
Gr
ap
h
ical
User
I
n
ter
f
ac
e
(
GUI
)
f
o
r
th
e
s
y
s
te
m
r
ep
r
esen
tatio
n
.
As
s
h
o
w
n
in
F
ig
u
r
e
9
,
th
e
GUI
is
a
p
latf
o
r
m
to
p
r
o
v
id
e
a
u
s
er
f
r
ie
n
d
l
y
i
n
ter
f
ac
e
an
d
i
m
p
r
o
v
e
th
e
s
u
s
tai
n
ab
ilit
y
o
f
t
h
is
s
y
s
te
m
.
Fig
u
r
e
9
.
L
P
R
s
y
s
te
m
GUI
3.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
N
3
.
1
.
CNN
a
rc
hite
ct
ure
a
na
ly
s
is
3
.
1
.
1
.
F
ea
t
ure
m
a
p
Fro
m
t
h
e
Fi
g
u
r
e
10,
th
er
e
ar
e
12
f
ea
tu
r
e
m
ap
co
n
f
ig
u
r
a
tio
n
s
ar
e
b
ein
g
te
s
ted
th
r
o
u
g
h
1
0
-
f
o
ld
cr
o
s
s
-
v
alid
atio
n
.
T
h
e
1
2
co
n
f
ig
u
r
atio
n
s
i
n
cl
u
d
e
3
-
14
-
6
0
,
4
-
14
-
6
0
,
4
-
15
-
6
0
,
5
-
11
-
6
0
,
5
-
12
-
3
0
,
5
-
12
-
4
0
,
5
-
12
-
5
0
,
5
-
12
-
6
0
,
5
-
13
-
6
0
,
5
-
14
-
6
0
,
5
-
15
-
6
0
an
d
6
-
14
-
6
0
.
E
q
u
atio
n
(
1
)
a
n
d
E
q
u
atio
n
(
2
)
i
n
d
icate
s
th
e
s
ize
s
o
f
th
e
o
u
tp
u
t
m
ap
(
M
x
, M
y
)
,
in
x
an
d
y
d
ir
ec
tio
n
an
d
t
h
e
co
n
v
o
lu
tio
n
/
s
u
b
s
a
m
p
li
n
g
p
r
o
ce
s
s
r
es
p
ec
tiv
el
y
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2
0
8
8
-
8708
A
d
esig
n
o
f lice
n
s
e
p
la
te
r
ec
o
g
n
itio
n
s
ystem
u
s
in
g
co
n
v
o
lu
tio
n
a
l n
e
u
r
a
l n
etw
o
r
k
(
P
.
Ma
r
z
u
ki
)
2201
(
)
=
(
−
1
)
−
(
(
)
−
(
)
)
(
)
;
(
)
=
(
−
1
)
−
(
(
)
−
(
)
)
(
)
(
1
)
(
)
=
(
∑
∑
(
,
)
(
)
(
,
)
∈
(
)
∈
(
−
1
)
0
(
−
1
)
(
+
,
+
)
+
(
)
)
(
2
)
w
h
er
e
y
x
k
v
k
u
N
v
u
K
0
;
0
|
,
2
,
,
k
x
an
d
k
y
ar
e
th
e
w
i
d
th
an
d
th
e
h
e
ig
h
t
o
f
t
h
e
c
o
n
v
o
l
u
tio
n
k
er
n
el
s
w
ji
(l)
o
f
la
y
er
(
l)
a
n
d
b
j
(l)
is
t
h
e
b
ias o
f
f
ea
tu
r
e
m
ap
n
i
n
la
y
er
(
l
)
,
c
a
n
d
r
r
e
f
er
s
to
t
h
e
cu
r
r
en
t p
i
x
el
a
n
d
p
r
ef
er
s
to
th
e
p
ar
ticu
lar
tr
ai
n
i
n
g
s
a
m
p
le.
T
h
e
s
et
M
j
(l
-
1)
co
n
t
ain
s
t
h
e
f
ea
tu
r
e
m
ap
s
i
n
th
e
p
r
ec
ed
in
g
la
y
er
(
l
-
1
)
th
at
ar
e
co
n
n
ec
ted
to
f
ea
t
u
r
e
m
ap
n
i
n
la
y
er
(
l
)
.
T
h
e
n
o
ta
tio
n
f
i
s
t
h
e
ac
ti
v
atio
n
f
u
n
ct
i
o
n
o
f
la
y
er
(
l
)
.
T
h
e
v
ar
iab
le
u
an
d
v
d
escr
ib
es th
e
h
o
r
izo
n
tal
an
d
v
er
tical
s
tep
s
iz
e
o
f
th
e
co
n
v
o
lu
t
io
n
k
er
n
el
i
n
la
y
er
(
l
)
.
Fig
u
r
e
1
0
.
Fin
al
ar
ch
itect
u
r
e
o
f
C
N
N
m
o
d
el
T
h
e
in
itial
f
ea
t
u
r
e
m
ap
i
n
C
1
,
C
2
an
d
C
3
la
y
er
s
i
s
5
,
1
4
an
d
6
0
f
ea
tu
r
e
s
m
ap
s
.
Me
an
w
h
ile
t
h
e
o
u
tp
u
t
la
y
er
co
n
tain
s
3
3
n
eu
r
o
n
s
s
i
n
ce
th
er
e
ar
e
a
to
tal
o
f
3
3
ch
ar
ac
ter
class
es
f
o
r
Ma
la
y
s
ia
n
lice
n
s
e
p
late.
B
u
t
th
e
n
u
m
b
er
o
f
f
ea
tu
r
e
m
ap
s
at
C
1
,
C
2
an
d
C
3
la
y
er
ca
n
v
ar
y
to
s
u
it
to
th
e
co
m
p
l
ex
it
y
o
f
t
h
e
in
p
u
t
i
m
a
g
e.
Fig
u
r
e
1
0
is
th
e
f
i
n
al
r
esu
lt a
f
ter
test
in
g
ea
ch
f
ea
t
u
r
e
m
ap
co
n
f
ig
u
r
atio
n
.
3
.
1
.
2
.
Co
n
nect
io
n p
a
t
t
er
n
T
h
er
e
ar
e
6
ty
p
es
o
f
p
atter
n
co
n
n
ec
tio
n
b
et
w
ee
n
C
1
a
n
d
C
2
la
y
er
a
s
s
h
o
w
n
i
n
T
ab
le
1
.
T
h
e
s
elec
t
io
n
o
f
co
n
n
ec
tio
n
in
e
v
er
y
la
y
er
is
ev
al
u
ated
b
y
t
h
e
lo
w
e
s
t
v
alid
atio
n
er
r
o
r
in
th
ese
6
t
y
p
es
o
f
co
n
n
ec
tio
n
.
E
ac
h
t
y
p
e
o
f
co
n
n
ec
tio
n
i
s
b
e
test
ed
f
r
o
m
t
h
e
f
ir
s
t
co
l
u
m
n
to
th
e
la
s
t
co
lu
m
n
i
n
a
s
eq
u
e
n
tia
l
m
an
n
er
.
T
h
e
lo
w
est
v
alid
atio
n
er
r
o
r
am
o
n
g
s
i
x
t
y
p
es
o
f
co
n
n
ec
t
io
n
w
ill
b
e
ch
o
s
e
n
.
T
h
r
o
u
g
h
o
u
t
th
e
1
0
-
f
o
ld
cr
o
s
s
-
v
alid
atio
n
,
t
h
e
b
est co
n
n
ec
tio
n
p
atter
n
is
a
s
s
h
o
w
n
i
n
T
ab
le
2
.
T
ab
le
1
.
C
o
n
n
ec
tio
n
b
et
w
ee
n
C
1
an
d
C
2
la
y
er
s
L
a
y
e
r
C
o
n
n
e
c
t
i
o
n
1
st
l
a
y
e
r
0
0
×
×
×
2
n
d
l
a
y
e
r
×
0
×
0
×
3
r
d
l
a
y
e
r
0
×
×
×
×
4
t
h
l
a
y
e
r
×
0
×
0
×
5
t
h
l
a
y
e
r
×
×
×
×
0
6
t
h
l
a
y
e
r
0
×
×
×
×
7
t
h
l
a
y
e
r
0
0
×
×
×
8
t
h
l
a
y
e
r
0
0
×
×
×
9
t
h
l
a
y
e
r
0
×
×
×
×
10
t
h
l
a
y
e
r
0
0
×
×
×
11
t
h
l
a
y
e
r
0
×
×
×
×
12
t
h
l
a
y
e
r
×
×
×
×
×
T
ab
le
2
.
Six
T
y
p
e
o
f
C
o
n
n
ec
ti
o
n
Ty
p
e
o
f
c
o
n
n
e
c
t
i
o
n
P
a
t
t
e
r
n
C
o
n
n
e
c
t
i
o
n
1
×
×
×
0
0
C
o
n
n
e
c
t
i
o
n
2
0
×
×
×
0
C
o
n
n
e
c
t
i
o
n
3
0
0
×
×
×
C
o
n
n
e
c
t
i
o
n
4
×
0
×
0
×
C
o
n
n
e
c
t
i
o
n
5
×
×
×
×
0
C
o
n
n
e
c
t
i
o
n
6
0
×
×
×
×
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
n
t J
E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
9
,
No
.
3
,
J
u
n
e
2
0
1
9
:
2
1
9
6
-
2
2
0
4
2202
3
.
1
.
3
.
T
y
pe
o
f
w
eig
ht
s
elec
t
i
o
n
T
h
e
w
ei
g
h
t i
n
itializa
tio
n
i
s
ca
r
r
ied
o
u
t b
y
u
s
i
n
g
1
0
-
f
o
ld
cr
o
s
s
-
v
a
lid
atio
n
.
T
h
er
e
ar
e
4
t
y
p
es
o
f
w
ei
g
h
t
b
ein
g
te
s
ted
n
a
m
el
y
Ga
u
s
s
ian
,
Ng
u
y
e
n
,
Fa
n
-
i
n
a
n
d
u
n
i
f
o
r
m
.
Fro
m
T
ab
le
3
,
th
e
Gau
s
s
ia
n
weig
h
t
p
r
o
d
u
ce
s
th
e
b
est
ac
cu
r
ac
y
w
h
ic
h
is
8
1
.
4
8
%
f
o
r
r
ec
o
g
n
i
tio
n
.
T
h
er
ef
o
r
e,
Gau
s
s
ia
n
w
eig
h
t
h
as
b
ee
n
u
s
e
d
f
o
r
th
e
r
est
o
f
t
h
e
test
i
n
g
.
T
ab
l
e
3
.
T
y
p
e
o
f
w
e
ig
h
t
v
er
s
u
s
ac
cu
r
ac
y
Ty
p
e
s o
f
w
e
i
g
h
t
A
c
c
u
r
a
c
y
(
%)
G
a
u
ssi
a
n
8
1
.
4
8
N
g
u
y
e
n
4
8
.
0
0
F
a
n
-
in
0
.
0
0
U
n
i
f
o
r
m
5
9
.
0
0
T
h
e
lear
n
in
g
alg
o
r
it
h
m
u
s
ed
in
tr
ain
i
n
g
th
e
C
NN
is
a
n
en
h
an
ce
d
v
er
s
io
n
o
f
Sto
ch
as
tic
Diag
o
n
al
L
e
v
en
b
er
g
Ma
r
q
u
ad
t
(
SD
L
M)
[
3
5
]
alg
o
r
ith
m
.
T
h
is
lear
n
in
g
al
g
o
r
ith
m
is
b
ett
er
th
an
s
ta
n
d
ar
d
b
ac
k
p
r
o
p
ag
atio
n
in
w
h
ic
h
i
t
av
o
id
s
t
h
e
g
r
ad
ie
n
t
f
r
o
m
g
etti
n
g
tr
ap
in
to
t
h
e
lo
ca
l
m
i
n
i
m
a
.
As
a
r
es
u
lt
w
h
e
n
SDLM
alg
o
r
it
h
m
i
s
u
s
ed
,
a
s
m
o
o
t
h
er
er
r
o
r
g
r
ad
ien
t
is
ac
h
i
ev
ed
.
I
n
th
e
m
en
tio
n
ed
alg
o
r
i
th
m
,
t
h
er
e
ar
e
t
w
o
p
ar
am
eter
s
n
ee
d
to
b
e
tu
n
ed
n
a
m
el
y
r
e
g
u
lar
izatio
n
,
p
ar
a
m
et
er
μ
an
d
th
e
p
ar
am
eter
.
3
.
1
.
4
.
Reg
ula
riza
t
io
n pa
r
a
m
et
er
T
h
e
r
eg
u
lar
izatio
n
p
ar
a
m
eter
is
th
e
p
ar
a
m
ete
r
o
f
t
h
e
lea
r
n
in
g
al
g
o
r
ith
m
w
h
ich
i
s
S
DL
M.
T
h
is
p
ar
am
eter
ca
n
v
ar
y
f
r
o
m
0
.
0
4
u
n
t
il
0
.
0
9
.
T
h
e
o
u
tco
m
e
o
f
ea
ch
v
alu
e
o
f
t
h
e
r
e
g
u
lar
izatio
n
p
ar
am
eter
is
s
h
o
w
n
in
Fi
g
u
r
e
1
1
.
Fro
m
F
ig
u
r
e
1
1
,
t
h
e
r
eg
u
lar
i
za
tio
n
p
ar
a
m
eter
=
0
.
0
4
h
as
p
r
o
d
u
ce
th
e
lo
w
e
s
t
m
ea
n
s
q
u
ar
e
er
r
o
r
(
MSE
)
in
co
m
p
ar
is
o
n
to
o
th
er
s
.
T
h
er
ef
o
r
e,
0
.
0
4
is
th
e
b
est
v
alu
e
to
b
e
u
s
ed
f
o
r
licen
s
e
p
late
r
ec
o
g
n
itio
n
.
A
cc
o
r
d
in
g
to
th
e
g
r
ap
h
,
t
h
e
f
ea
tu
r
e
m
ap
5
-
12
-
6
0
p
er
f
o
r
m
a
n
ce
is
h
ig
h
er
th
a
n
o
th
er
co
n
f
i
g
u
r
atio
n
w
h
ic
h
also
in
cl
u
d
in
g
t
h
e
i
n
itial
f
ea
t
u
r
e
5
-
14
-
6
0
.
T
h
r
o
u
g
h
th
i
s
e
x
p
er
i
m
en
t,
a
b
est
f
ea
t
u
r
e
m
ap
co
n
f
ig
u
r
atio
n
i
s
o
b
tain
ed
w
h
ic
h
is
5
-
12
-
60.
Fig
u
r
e
1
1
.
Me
an
s
q
u
ar
e
er
r
o
r
v
er
s
u
s
r
eg
u
latio
n
p
ar
a
m
eter
g
r
ap
h
3
.
1
.
5
.
y
–
c
o
ns
t
a
nt
-
co
n
s
ta
n
t
i
s
al
s
o
th
e
p
ar
a
m
e
ter
f
r
o
m
S
DL
M
lear
n
i
n
g
al
g
o
r
ith
m
.
T
h
e
v
al
u
e
o
f
t
h
e
-
co
n
s
ta
n
t
ca
n
eith
er
b
e
0
.
1
,
0
.
0
1
o
r
0
.
0
0
1
.
T
h
e
r
esu
lt
o
f
t
h
e
p
er
f
o
r
m
a
n
ce
i
s
o
b
tain
ed
b
y
t
h
e
ac
c
u
r
ac
y
o
f
r
ec
o
g
n
itio
n
d
u
r
in
g
v
alid
atio
n
b
y
u
s
in
g
1
0
-
f
o
ld
cr
o
s
s
-
va
lid
atio
n
e
x
p
er
i
m
e
n
t.
T
h
e
T
ab
le
4
s
h
o
w
ed
th
at
t
h
e
-
c
o
n
s
ta
n
t
0
.
0
1
h
a
v
in
g
th
e
h
i
g
h
est
ac
c
u
r
ac
y
o
f
r
ec
o
g
n
i
tio
n
w
h
ic
h
is
8
1
.
4
8
%.
So
th
at,
-
co
n
s
tan
t
=
0
.
0
1
is
u
s
ed
in
t
h
e
C
N
N
r
ec
o
g
n
itio
n
s
y
s
te
m
.
T
h
is
r
esu
lt
ca
n
b
e
f
u
r
th
er
i
m
p
r
o
v
ed
if
th
e
p
r
o
b
le
m
at
t
h
e
p
r
ep
r
o
ce
s
s
in
g
p
ar
t
h
as
s
o
lv
ed
.
T
a
b
le
5
s
h
o
w
s
t
h
e
ac
c
u
r
ac
y
o
f
p
r
ep
r
o
ce
s
s
i
n
g
a
n
d
C
NN
r
ec
o
g
n
itio
n
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2
0
8
8
-
8708
A
d
esig
n
o
f lice
n
s
e
p
la
te
r
ec
o
g
n
itio
n
s
ystem
u
s
in
g
co
n
v
o
lu
tio
n
a
l n
e
u
r
a
l n
etw
o
r
k
(
P
.
Ma
r
z
u
ki
)
2203
T
ab
le
4
.
-
co
n
s
tan
t
v
er
s
u
s
A
c
cu
r
ac
y
V
a
l
u
e
A
c
c
u
r
a
c
y
(
%)
0
.
1
7
0
.
3
7
0
.
0
1
8
1
.
4
8
0
.
0
0
1
7
7
.
7
8
T
ab
le
5
.
A
cc
u
r
ac
y
o
f
L
P
R
S
y
s
te
m
P
r
o
c
e
ss
A
c
c
u
r
a
c
y
(
%)
P
r
e
p
r
o
c
e
ssi
n
g
7
4
.
7
%
o
u
t
o
f
3
0
0
s
a
m
p
l
e
s
C
N
N
r
e
c
o
g
n
i
t
i
o
n
9
4
.
6
%
o
u
t
o
f
5
2
8
s
a
m
p
l
e
s
3
.
1
.
6
.
P
er
f
o
r
m
a
nce
o
f
t
he
des
ig
ned L
P
R
s
y
s
t
e
m
I
n
t
h
is
p
ap
er
,
t
w
o
t
y
p
es
o
f
ac
c
u
r
ac
ies
ar
e
tak
e
n
at
d
if
f
er
en
t
s
tag
es.
T
h
e
f
ir
s
t
ac
cu
r
ac
y
i
s
ta
k
en
at
t
h
e
p
r
ep
r
o
ce
s
s
in
g
p
ar
t
th
e
s
y
s
te
m
an
d
th
e
s
ec
o
n
d
o
n
e
is
ta
k
en
at
th
e
class
i
f
icatio
n
s
ta
g
e.
T
h
e
c
lass
i
f
icatio
n
r
esu
l
t
is
tak
e
n
ac
co
r
d
in
g
to
t
h
e
n
u
m
b
er
o
f
ch
ar
ac
ter
s
s
u
cc
e
s
s
f
u
ll
y
r
ec
o
g
n
ized
.
T
ab
le
5
d
escr
ib
es
th
at
th
e
p
r
ep
r
o
ce
s
s
in
g
p
ar
t
ac
h
ie
v
es
7
4
.
7
%
o
u
t
o
f
3
0
0
s
a
m
p
les
te
s
ted
w
h
ic
h
d
o
es
n
o
t
ac
h
iev
e
t
h
e
e
x
p
ec
tatio
n
l
ev
el.
T
h
e
p
r
ep
r
o
ce
s
s
in
g
al
g
o
r
ith
m
n
ee
d
s
to
b
e
f
u
r
th
er
i
m
p
r
o
v
ed
in
o
r
d
er
to
ef
f
ec
tiv
e
l
y
f
i
lter
o
u
t
n
o
i
s
es
s
u
c
h
as
en
v
ir
o
n
m
e
n
t
f
ac
to
r
(
illu
m
i
n
atio
n
)
to
ac
h
ie
v
e
h
i
g
h
er
ac
cu
r
ac
y
.
T
h
e
p
r
ep
r
o
ce
s
s
in
g
p
r
o
ce
s
s
is
p
r
o
m
i
n
en
t
b
ec
a
u
s
e
w
h
en
e
v
er
t
h
e
L
P
R
s
y
s
t
e
m
f
ailed
to
p
r
ep
r
o
ce
s
s
t
h
e
in
p
u
t
i
m
ag
e
co
r
r
ec
tly
,
t
h
e
ac
cu
r
ac
y
w
ill b
e
af
f
ec
ted
at
t
h
e
clas
s
if
icatio
n
s
tag
e.
C
NN
ab
le
to
ac
h
ie
v
e
9
4
.
6
%
o
u
t
o
f
5
2
8
s
a
m
p
les
test
ed
.
T
h
is
C
NN
r
ec
o
g
n
itio
n
d
esi
g
n
ed
m
o
d
el
h
a
s
ac
h
iev
ed
t
h
e
ex
p
ec
tatio
n
le
v
el
th
at
ca
n
p
r
ac
ticall
y
u
s
e
in
f
u
t
u
r
e.
T
h
e
C
NN
r
ec
o
g
n
itio
n
ca
n
b
e
c
o
n
s
id
er
ed
as
a
r
o
b
u
s
t r
ec
o
g
n
itio
n
tech
n
iq
u
e
t
o
b
e
u
s
ed
in
licen
s
e
p
late
ch
ar
ac
ter
r
ec
o
g
n
itio
n
.
4.
CO
NCLU
SI
O
N
A
L
P
R
s
y
s
te
m
u
s
i
n
g
t
h
e
C
N
N
r
ec
o
g
n
itio
n
m
e
th
o
d
i
s
s
u
c
ce
s
s
f
u
l
l
y
d
ev
elo
p
ed
.
T
h
e
p
r
ep
r
o
ce
s
s
in
g
s
tag
e
o
f
L
P
R
s
y
s
te
m
t
h
at
i
s
d
ev
elo
p
ed
in
M
A
T
L
A
B
s
o
f
t
w
ar
e
h
as
s
u
cc
e
s
s
f
u
ll
y
m
er
g
ed
w
it
h
th
e
C
NN
r
ec
o
g
n
itio
n
s
y
s
te
m
i
n
C
lan
g
u
ag
e.
T
h
is
s
y
s
te
m
n
ee
d
s
i
m
p
r
o
v
e
m
e
n
t
at
th
e
p
r
ep
r
o
ce
s
s
in
g
s
tag
e
to
ac
h
iev
e
a
b
etter
ac
cu
r
ac
y
le
v
el.
Fo
r
f
u
t
u
r
e
r
ec
o
m
m
e
n
d
atio
n
,
t
h
is
p
r
o
p
o
s
ed
w
o
r
k
ca
n
b
e
u
s
ed
to
en
h
a
n
ce
t
h
e
A
E
S
in
Ma
la
y
s
ia.
C
u
r
r
en
tl
y
th
e
A
E
S
i
s
u
s
ed
to
ca
p
t
u
r
e
t
h
e
v
e
h
icle
t
h
at
e
x
ce
ed
s
th
e
s
p
ee
d
li
m
it
o
n
l
y
.
T
h
e
a
n
al
y
s
i
s
o
f
th
e
ca
p
tu
r
ed
i
m
a
g
es i
s
d
o
n
e
b
y
h
u
m
an
.
ACK
NO
WL
E
D
G
E
M
E
NT
Au
t
h
o
r
s
w
o
u
ld
lik
e
to
th
a
n
k
Un
iv
er
s
iti
T
ek
n
i
k
al
Ma
la
y
s
i
a
Me
lak
a
(
UT
eM
)
f
o
r
s
u
p
p
o
r
tin
g
t
h
i
s
r
esear
ch
u
n
d
er
P
J
P
/2
0
1
8
/FKEKK
(
9
D)
/S0
1
6
2
2
.
RE
F
E
R
E
NC
E
S
[1
]
N.
A
.
Ba
k
a
r,
e
t
a
l
.
,
“
M
a
la
y
sia
n
V
e
h
icle
L
ice
n
se
P
late
Re
c
o
g
n
it
i
o
n
Us
in
g
Do
u
b
le
Ed
g
e
De
tec
ti
o
n
,
”
2
0
1
2
IE
EE
In
ter
n
a
t
io
n
a
l
C
o
n
fer
e
n
c
e
o
n
C
o
n
t
ro
l
S
y
ste
m,
Co
m
p
u
t
in
g
a
n
d
En
g
in
e
e
rin
g
,
Pen
a
n
g
,
p
p
.
4
2
2
-
4
2
6
,
2
0
1
2
.
[2
]
M
.
F
.
Zak
a
ria
a
n
d
S
.
A
.
S
u
a
n
d
i,
“
M
a
la
y
sia
n
Ca
r
Nu
m
b
e
r
P
late
De
tec
ti
o
n
S
y
ste
m
B
a
se
d
o
n
T
e
m
p
lat
e
M
a
tch
in
g
a
n
d
Co
lo
u
r
In
f
o
rm
a
ti
o
n
,
”
In
ter
n
a
ti
o
n
a
l
J
o
u
rn
a
l
o
n
Co
m
p
u
ter
S
c
ien
c
e
a
n
d
En
g
i
n
e
e
rin
g
,
v
o
l
/i
ss
u
e
:
2
(
4
)
,
p
p
.
1
1
5
9
–
11
6
4
,
2
0
1
0
.
[3
]
W
.
W
.
Ke
o
n
g
a
n
d
V
.
Ira
n
m
a
n
e
sh
,
“
M
a
la
y
sia
n
A
u
to
m
a
ti
c
Nu
m
b
e
r
P
late
Re
c
o
g
n
it
io
n
S
y
ste
m
u
sin
g
P
e
a
rso
n
Co
rre
latio
n
,
”
2
0
1
6
IE
EE
S
y
mp
o
si
u
m o
n
C
o
mp
u
ter
Ap
p
li
c
a
ti
o
n
s &
In
d
u
stria
l
El
e
c
tro
n
ics
(
IS
CAIE
),
Ba
tu
Fer
i
n
g
g
h
i
,
p
p
.
4
0
–
4
5
,
2
0
1
6
.
[4
]
S
.
R.
S
o
o
m
ro
,
e
t
a
l
.,
“
V
e
h
icle
Nu
m
b
e
r
Re
c
o
g
n
it
io
n
S
y
ste
m
f
o
r
A
u
to
m
a
ti
c
T
o
ll
T
a
x
Co
ll
e
c
ti
o
n
,
”
2
0
1
2
In
t.
Co
n
f
.
Ro
b
o
t.
Arti
f.
I
n
tell.
(
ICRA
I
2
0
1
2
)
,
p
p
.
1
2
5
–
1
2
9
,
2
0
1
2
.
[5
]
F
.
H.
C.
T
iv
iv
e
a
n
d
A
.
Bo
u
z
e
rd
o
u
m
,
“
A
F
a
c
e
De
tec
ti
o
n
S
y
st
e
m
Us
in
g
S
h
u
n
ti
n
g
I
n
h
i
b
it
o
ry
Co
n
v
o
lu
ti
o
n
a
l
Ne
u
ra
l
Ne
tw
o
r
k
s,”
Ne
u
ra
l
Ne
two
rk
s,
2
0
0
4
.
Pro
c
e
e
d
i
n
g
s.
2
0
0
4
IEE
E
In
t
.
J
t
.
Co
n
f
.
,
v
o
l.
4
,
p
p
.
2
5
7
1
–
2
5
7
5
,
2
0
0
4
.
[6
]
C.
G
a
rc
i
a
a
n
d
M
.
De
lak
is,
“
Co
n
v
o
lu
ti
o
n
a
l
F
a
c
e
F
in
d
e
r:
A
Ne
u
ra
l
A
rc
h
it
e
c
tu
re
f
o
r
F
a
st an
d
Ro
b
u
st
F
a
c
e
De
tec
ti
o
n
,
”
IEE
E
T
ra
n
s.
P
a
tt
e
rn
An
a
l.
M
a
c
h
.
In
tell
.
,
v
o
l
/is
su
e
:
26
(
11
)
,
p
p
.
1
4
0
8
–
1
4
2
3
,
2
0
0
4
.
[7
]
M
.
De
lak
is
a
n
d
C.
G
a
r
c
ia,
“
T
ra
i
n
in
g
Co
n
v
o
lu
t
io
n
a
l
F
il
ters
f
o
r
Ro
b
u
st
F
a
c
e
De
tec
ti
o
n
,
”
Ne
u
ra
l
N
e
two
rk
s
S
ig
n
a
l
Pro
c
e
ss
.
-
Pro
c
.
IEE
E
W
o
rk
.
,
v
o
l
.
2
0
0
3
,
p
p
.
7
3
9
–
7
4
8
,
2
0
0
3
.
[8
]
N.
F
a
rru
g
ia,
e
t.
a
l
.
,
“
F
a
st
a
n
d
Ro
b
u
st F
a
c
e
De
tec
ti
o
n
o
n
a
P
a
ra
ll
e
l
Op
ti
m
ize
d
A
r
c
h
it
e
c
tu
re
Im
p
le
m
e
n
ted
o
n
F
P
G
A
,
”
IEE
E
T
ra
n
s.
Circ
u
it
s S
y
st.
Vi
d
e
o
T
e
c
h
n
o
l
.
,
v
o
l
/
issu
e
:
19
(
4
)
,
p
p
.
5
9
7
–
6
0
2
,
2
0
0
9
.
[9
]
C.
P
o
u
let,
e
t
a
l
.,
“
CN
P
:
A
n
F
GP
A
-
b
a
se
d
P
r
o
c
e
ss
o
r
f
o
r
Co
n
v
o
lu
ti
o
n
a
l
Ne
tw
o
rk
s,”
FP
L
0
9
1
9
t
h
In
t.
Co
n
f.
F.
Pro
g
ra
m.
L
o
g
.
Ap
p
l.
,
v
o
l
/i
ss
u
e
:
1
(
1
)
,
p
p
.
3
2
–
3
7
,
2
0
0
9
.
[1
0
]
N.
F
a
rru
g
ia,
e
t
a
l
.
,
“
De
sig
n
o
f
a
Re
a
l
-
T
i
m
e
F
a
c
e
De
tec
ti
o
n
P
a
ra
ll
e
l
A
r
c
h
it
e
c
tu
re
Us
in
g
Hig
h
-
L
e
v
e
l
S
y
n
th
e
sis,”
EURA
S
IP
J
.
Emb
e
d
.
S
y
st
.
,
v
o
l.
2
0
0
8
,
p
p
.
1
-
9
,
2
0
0
8
.
[1
1
]
P
.
Bu
y
ss
e
n
s
a
n
d
M
.
Re
v
e
n
u
,
“
L
e
a
rn
in
g
sp
a
rse
f
a
c
e
f
e
a
t
u
re
s:
A
p
p
li
c
a
ti
o
n
t
o
F
a
c
e
V
e
rif
ica
ti
o
n
,
”
Pr
o
c
.
-
In
t.
C
o
n
f.
Pa
tt
e
rn
Rec
o
g
n
it
.
,
p
p
.
6
7
0
–
6
7
3
,
2
0
1
0
.
[1
2
]
S
.
Ch
o
p
ra
,
e
t
a
l
.
,
“
L
e
a
rn
in
g
a
S
im
il
iart
y
M
e
tri
c
Di
sc
ri
m
in
a
ti
v
e
l
y
,
W
it
h
A
p
p
li
c
a
ti
o
n
t
o
F
a
c
e
V
e
rif
ica
ti
o
n
,
”
Pro
c
.
IEE
E
Co
n
f.
C
o
mp
u
t.
Vi
s.
P
a
tt
e
rn
Rec
o
g
n
it
.,
p
p
.
3
4
9
–
3
5
6
,
2
0
0
5
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
n
t J
E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
9
,
No
.
3
,
J
u
n
e
2
0
1
9
:
2
1
9
6
-
2
2
0
4
2204
[1
3
]
H.
G
h
ias
sira
d
a
n
d
M
.
T
e
sh
n
e
h
lab
,
“
S
im
il
a
rit
y
M
e
a
su
re
m
e
n
t
in
Co
n
v
o
lu
ti
o
n
a
l
S
p
a
c
e
,
”
IS
’2
0
1
2
-
2
0
1
2
6
th
IE
EE
In
t.
Co
n
f.
In
tell
.
S
y
st.
Pro
c
.
,
p
p
.
2
5
0
–
2
5
5
,
2
0
1
2
.
[1
4
]
Y.
N.
Ch
e
n
,
e
t
a
l
.
,
“
T
h
e
A
p
p
li
c
a
ti
o
n
o
f
a
Co
n
v
o
l
u
ti
o
n
Ne
u
ra
l
Ne
two
rk
o
n
F
a
c
e
a
n
d
L
ice
n
se
P
late
De
tec
ti
o
n
,
”
Pr
o
c
.
-
In
t.
C
o
n
f.
P
a
tt
e
rn
Rec
o
g
n
it
.
,
v
o
l
.
3
,
p
p
.
5
5
2
–
5
5
5
,
2
0
0
6
.
[1
5
]
G
.
B.
Hu
a
n
g
,
e
t
a
l
.
,
“
L
e
a
rn
in
g
H
iera
rc
h
ica
l
Re
p
re
se
n
tatio
n
s
f
o
r
F
a
c
e
V
e
rif
ic
a
ti
o
n
w
it
h
Co
n
v
o
lu
ti
o
n
a
l
De
e
p
Be
li
e
f
Ne
tw
o
rk
s,”
Pro
c
.
IEE
E
C
o
mp
u
t.
S
o
c
.
C
o
n
f
.
Co
m
p
u
t
.
Vi
s.
Pa
tt
e
rn
Rec
o
g
n
it
.
,
p
p
.
2
5
1
8
–
2
5
2
5
,
2
0
1
2
.
[1
6
]
S
.
S
.
L
ie
w
,
e
t
a
l
.
,
“
G
e
n
d
e
r
Clas
sifica
ti
o
n
:
A
Co
n
v
o
lu
ti
o
n
a
l
Ne
u
ra
l
Ne
tw
o
rk
A
p
p
ro
a
c
h
,
”
p
p
.
1
2
4
8
–
1
2
6
4
,
2
0
1
6
.
[1
7
]
F
.
Hin
g
,
e
t
a
l
.
,
“
A
G
e
n
d
e
r
Re
c
o
g
n
it
io
n
S
y
ste
m
u
sin
g
S
h
u
n
ti
n
g
I
n
h
ib
it
o
ry
Co
n
v
o
lu
ti
o
n
a
l
Ne
u
ra
l
Ne
t
w
o
rk
s,”
In
t.
J
t.
Co
n
f.
Ne
u
ra
l
Ne
two
rk
s (
IJ
CNN ’
0
6
)
,
p
p
.
5
3
3
6
–
5
3
4
1
,
2
0
0
6
.
[1
8
]
F
.
H.
C.
T
i
v
iv
e
a
n
d
A
.
Bo
u
z
e
rd
o
u
m
,
“
A
S
h
u
n
ti
n
g
In
h
i
b
it
o
r
y
Co
n
v
o
lu
ti
o
n
a
l
Ne
u
ra
l
Ne
t
w
o
rk
f
o
r
G
e
n
d
e
r
Clas
sif
ic
a
ti
o
n
,
”
Pro
c
.
-
In
t
.
Co
n
f.
Pa
tt
e
rn
Rec
o
g
n
it
.
,
v
o
l.
4
,
p
p
.
4
2
1
–
4
2
4
,
2
0
0
6
.
[1
9
]
S
.
F
.
A
b
d
u
l
la
h
,
e
t
a
l
.
,
“
M
u
l
ti
lay
e
r
P
e
rc
e
p
tro
n
Ne
u
ra
l
Ne
tw
o
rk
i
n
Clas
sify
in
g
Ge
n
d
e
r
u
sin
g
F
in
g
e
rp
r
in
t
G
lo
b
a
l
L
e
v
e
l
F
e
a
tu
re
s,”
v
o
l
/i
ss
u
e
:
9
(
9
)
,
2
0
1
6
.
[2
0
]
C.
H.
T
e
o
,
e
t
a
l
.
,
“
A
No
v
e
l
Ap
p
r
o
a
c
h
to
Im
p
ro
v
e
th
e
T
ra
in
in
g
T
i
m
e
o
f
Co
n
v
o
lu
ti
o
n
a
l
Ne
two
rk
s
f
o
r
Ob
jec
t
Re
c
o
g
n
it
io
n
.
”
Re
t
riv
e
f
ro
m
h
tt
p
:/
/
u
se
rs.ce
c
s.a
n
u
.
e
d
u
.
a
u
/~
c
h
te
o
/p
u
b
/T
e
o
Ta
y
L
a
i0
5
.
p
d
f
[2
1
]
F
.
J.
Hu
a
n
g
a
n
d
Y.
L
e
Cu
n
,
“
L
a
r
g
e
-
sc
a
le
L
e
a
rn
in
g
w
it
h
S
VM
a
n
d
Co
n
v
o
l
u
ti
o
n
a
l
Ne
ts
f
o
r
G
e
n
e
ric
Ob
jec
t
Ca
teg
o
riza
ti
o
n
,
”
Pro
c
.
I
EE
E
C
o
mp
u
t.
S
o
c
.
Co
n
f.
C
o
mp
u
t.
Vi
s.
P
a
tt
e
rn
Rec
o
g
n
it
.
,
v
o
l.
1
,
p
p
.
2
8
4
–
2
9
1
,
2
0
0
6
.
[2
2
]
M
.
M
.
P
iram
li
,
e
t
a
l
.
,
“
Rice
G
r
a
in
G
ra
d
in
g
Clas
si
f
ic
a
ti
o
n
Ba
se
d
On
P
e
rim
e
ter
Us
in
g
M
o
o
r
e
-
Ne
ig
h
b
o
r
T
ra
c
in
g
M
e
th
o
d
,
”
v
o
l
/i
ss
u
e
:
8
(
2
)
,
p
p
.
2
3
–
2
7
,
1
8
4
3
.
[2
3
]
S
.
S
.
A
h
ra
n
jan
y
,
e
t
a
l
.
,
“
A
V
e
r
y
Hig
h
Ac
c
u
ra
c
y
Ha
n
d
w
rit
ten
Ch
a
r
a
c
ter
Re
c
o
g
n
it
io
n
S
y
ste
m
f
o
r
F
a
r
si/A
ra
b
ic
Di
g
it
s
Us
in
g
Co
n
v
o
lu
ti
o
n
a
l
Ne
u
ra
l
Ne
tw
o
rk
s,”
Pro
c
.
2
0
1
0
IE
EE
5
th
I
n
t
.
Co
n
f
.
Bi
o
-
In
sp
ire
d
Co
mp
u
t.
T
h
e
o
r.
Ap
p
l.
BIC
-
T
A
2
0
1
0
,
p
p
.
1
5
8
5
–
1
5
9
2
,
2
0
1
0
.
[2
4
]
S
.
A
ro
ra
,
e
t
a
l
.
,
“
P
e
rf
o
rm
a
n
c
e
Co
m
p
a
riso
n
o
f
S
VM
a
n
d
A
NN
f
o
r
Ha
n
d
w
rit
ten
De
v
n
a
g
a
ri
Ch
a
ra
c
te
r
Re
c
o
g
n
it
io
n
,
”
In
ter
n
a
t
io
n
a
l
J
o
u
rn
a
l
o
f
C
o
mp
u
ter
S
c
ien
c
e
Iss
u
e
s
,
v
o
l
/i
ss
u
e
:
7
(
3
)
,
p
p
.
1
–
1
0
,
2
0
1
0
.
[2
5
]
H.
S
w
e
th
a
la
k
sh
m
i,
e
t
a
l
.
,
“
On
li
n
e
Ha
n
d
w
rit
ten
Ch
a
ra
c
ter
R
e
c
o
g
n
it
io
n
o
f
De
v
a
n
a
g
a
ri
a
n
d
T
e
lu
g
u
C
h
a
ra
c
ters
u
sin
g
S
u
p
p
o
rt
V
e
c
to
r
M
a
c
h
in
e
s,”
Gu
y
L
o
re
tt
e
.
T
e
n
t
h
In
ter
n
a
ti
o
n
a
l
W
o
r
k
sh
o
p
o
n
Fr
o
n
t
ier
s
in
Ha
n
d
writi
n
g
Rec
o
g
n
i
ti
o
n
,
L
a
B
a
u
le (
Fra
n
c
e
),
S
u
v
iso
ft
,
p
p
.
1
–
6
,
2
0
0
6
.
[2
6
]
F
.
H.
C.
T
iv
iv
e
a
n
d
A
.
Bo
u
z
e
rd
o
u
m
,
“
T
e
x
tu
re
Clas
si
f
ica
ti
o
n
u
sin
g
Co
n
v
o
lu
ti
o
n
a
l
Ne
u
ra
l
Ne
tw
o
rk
s,”
T
ENCON
2
0
0
6
-
2
0
0
6
IEE
E
Reg
.
1
0
Co
n
f.
,
Ho
n
g
Ko
n
g
,
p
p
.
1
4
–
1
7
,
2
0
0
6
.
[2
7
]
A
.
R.
S
y
a
f
e
e
z
a
,
e
t
a
l
.
,
“
A
Re
v
i
e
w
o
f
F
in
g
e
r
-
V
e
in
Bi
o
m
e
tri
c
s
Id
e
n
ti
f
ica
ti
o
n
A
p
p
ro
a
c
h
e
s,”
In
d
i
a
n
J
o
u
rn
a
l
o
f
S
c
ien
c
e
a
n
d
T
e
c
h
n
o
l
o
g
y
,
v
o
l
/i
ss
u
e
:
9
(
32
)
,
2
0
1
6
.
[2
8
]
D.
R.
T
o
b
e
rg
te
a
n
d
S
.
Cu
rti
s,
“
A
Un
if
i
e
d
A
rc
h
it
e
c
tu
re
f
o
r
th
e
De
tec
ti
o
n
a
n
d
Clas
sif
ica
ti
o
n
o
f
L
i
c
e
n
se
P
late
s,”
J
.
Ch
e
m.
In
f.
M
o
d
e
l
.
,
v
o
l
/i
ss
u
e
:
53
(
9
)
,
p
p
.
1
6
8
9
–
1
6
9
9
,
2
0
1
3
.
[2
9
]
S
.
A
.
R
a
d
z
i
a
n
d
M
.
Kh
a
li
l
-
h
a
n
i,
“
Ch
a
ra
c
ter
Re
c
o
g
n
it
io
n
o
f
L
ice
n
se
P
late
Nu
m
b
e
r
U
sin
g
Co
n
v
o
lu
ti
o
n
a
l
Ne
u
ra
l
Ne
tw
o
rk
,
”
VIC'
1
1
Pr
o
c
e
e
d
in
g
s
o
f
th
e
S
e
c
o
n
d
i
n
ter
n
a
ti
o
n
a
l
c
o
n
fer
e
n
c
e
o
n
Vi
su
a
l
i
n
fo
rm
a
ti
c
s:
su
sta
in
in
g
re
se
a
rc
h
a
n
d
in
n
o
v
a
ti
o
n
s,
S
e
la
n
g
o
r,
M
a
l
a
y
sia
,
p
p
.
4
5
–
5
5
,
2
0
1
1
.
[3
0
]
Z.
Zh
a
o
,
e
t
a
l
.
,
“
Ch
i
n
e
se
L
ice
n
se
P
late
Re
c
o
g
n
it
i
o
n
Us
in
g
a
Co
n
v
o
lu
ti
o
n
a
l
Ne
u
ra
l
Ne
tw
o
rk
,
”
2
0
0
8
IEE
E
Pa
c
if
ic
-
Asia
W
o
rk
.
Co
m
p
u
t
.
In
tell.
I
n
d
.
A
p
p
l.
,
p
p
.
2
7
–
3
0
,
2
0
0
8
.
[3
1
]
Y.
W
e
n
,
e
t
a
l
.
,
“
A
n
A
lg
o
rit
h
m
f
o
r
L
ice
n
se
P
late
Re
c
o
g
n
it
io
n
A
p
p
l
ied
to
In
telli
g
e
n
t
T
ra
n
sp
o
rtatio
n
S
y
st
e
m
,
”
IEE
E
T
ra
n
s.
I
n
tell.
T
ra
n
sp
.
S
y
st
.
,
v
o
l
/i
s
su
e
:
12
(
3
)
,
p
p
.
8
3
0
–
8
4
5
,
2
0
1
1
.
[3
2
]
X
.
Z
h
a
i,
e
t
a
l
.
,
“
L
ice
n
se
P
late
Lo
c
a
li
sa
ti
o
n
b
a
se
d
o
n
M
o
r
p
h
o
l
o
g
ica
l
Op
e
ra
ti
o
n
s,”
1
1
t
h
In
t.
C
o
n
f.
Co
n
tro
l.
Au
t
o
m
.
Ro
b
o
t.
Vi
sio
n
,
ICA
RCV
2
0
1
0
,
p
p
.
1
1
2
8
–
1
1
3
2
,
2
0
1
0
.
[3
3
]
S
.
Ch
a
k
ra
b
o
rty
,
“
A
n
I
m
p
ro
v
e
d
T
e
m
p
late
M
a
tch
in
g
A
lg
o
rit
h
m
fo
r
Ca
r
L
ice
n
se
P
late
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
Co
m
p
u
ter
A
p
p
l
ica
ti
o
n
s
,
v
o
l
/i
ss
u
e
:
1
1
8
(
25
)
,
p
p
.
1
6
–
2
2
,
2
0
1
5
.
[3
4
]
Y.
M
a
,
e
t
a
l
.
,
“
A
n
e
w
a
lg
o
rit
h
m
fo
r
c
h
a
ra
c
ters
se
g
m
e
n
tatio
n
o
f
li
c
e
n
se
p
late
b
a
se
d
o
n
v
a
rian
c
e
p
r
o
jec
ti
o
n
a
n
d
m
e
a
n
f
il
ter,”
Pro
c
.
2
0
1
1
I
EE
E
5
th
In
t
.
Co
n
f.
Cy
b
e
rn
.
I
n
te
ll
.
S
y
st.
CI
S
2
0
1
1
,
p
p
.
1
3
2
–
1
3
5
,
2
0
1
1
.
[3
5
]
A
.
R.
S
y
a
fe
e
z
a
,
e
t
a
l
.
,
“
Co
n
v
o
lu
ti
o
n
a
l
Ne
u
ra
l
Ne
tw
o
rk
s
w
it
h
F
u
se
d
L
a
y
e
rs
A
p
p
li
e
d
t
o
F
a
c
e
Re
c
o
g
n
it
io
n
,
”
In
ter
n
a
t
io
n
a
l
J
o
u
rn
a
l
o
f
C
o
mp
u
ta
ti
o
n
a
l
I
n
telli
g
e
n
c
e
a
n
d
A
p
p
li
c
a
ti
o
n
s
,
v
o
l
/i
ss
u
e
:
14
(
3
)
,
2
0
1
5
.
B
I
O
G
RAP
H
I
E
S O
F
AUTH
O
RS
M
u
h
a
m
a
d
M
a
r
z
u
k
i
Pi
r
a
m
li
wa
s
b
o
rn
in
M
a
lay
sia
in
y
e
a
r
1
9
9
2
.
He
w
a
s
c
o
m
p
lete
d
b
a
c
h
e
lo
r
d
e
g
r
e
e
in
Co
m
p
u
ter
S
c
ien
c
e
(A
rti
f
icia
l
In
telli
g
e
n
t)
i
n
y
e
a
r
2
0
1
5
f
ro
m
T
e
c
h
n
ica
l
Un
iv
e
rsit
y
o
f
M
a
la
y
sia
M
a
lac
c
a
(U
T
e
M
).
P
re
se
n
t
h
e
is
p
u
rsu
i
n
g
M
S
c
.
in
El
e
c
tro
n
ic
a
n
d
Co
m
p
u
ter
En
g
in
e
e
rin
g
a
t
T
e
c
h
n
i
c
a
l
Un
iv
e
rsit
y
o
f
M
a
la
y
sia
M
a
lac
c
a
(UT
e
M
).
His
Re
se
a
rc
h
in
c
lu
d
e
s
D
e
e
p
L
e
a
rn
in
g
,
P
a
tt
e
rn
Re
c
o
g
n
it
io
n
,
M
a
c
h
in
e
L
e
a
rn
in
g
a
n
d
A
rti
f
icia
l
In
telli
g
e
n
t.
Dr
.
S
y
a
fe
e
z
a
A
h
m
a
d
R
a
d
z
i
w
a
s
b
o
rn
in
M
a
lay
sia
in
y
e
a
r
1
9
8
1
.
S
h
e
wa
s a
wa
rd
e
d
d
e
g
re
e
a
n
d
M
.
En
g
.
in
El
e
c
tri
c
a
l
a
n
d
El
e
c
tro
n
ics
En
g
in
e
e
rin
g
in
T
e
c
h
n
o
lo
g
y
Un
iv
e
rsit
y
o
f
M
a
la
y
si
a
(UT
M
)
Jo
h
o
r
Ba
h
r
u
,
M
a
la
y
sia
.
S
h
e
c
o
m
p
lete
d
P
h
.
D
d
e
g
re
e
in
El
e
c
tri
c
a
l
En
g
in
e
e
rin
g
f
ro
m
Tec
h
n
o
lo
g
y
Un
iv
e
rsit
y
o
f
M
a
la
y
sia
(U
T
M
)
in
y
e
a
r
2
0
1
4
.
Cu
rr
e
n
tl
y
,
sh
e
is
w
o
rk
in
g
a
s
S
e
n
io
r
L
e
c
tu
re
r
in
th
e
De
p
a
rtm
e
n
t
o
f
Co
m
p
u
ter
En
g
in
e
e
rin
g
a
t
T
e
c
h
n
ica
l
Un
iv
e
rsit
y
o
f
M
a
la
y
sia
M
a
la
c
c
a
(U
T
e
M
).
He
r
r
e
se
a
r
c
h
a
re
a
in
Em
b
e
d
d
e
d
S
y
ste
m
,
P
a
tt
e
rn
Re
c
o
g
n
it
io
n
,
M
a
c
h
in
e
L
e
a
rn
in
g
,
Im
a
g
e
P
r
o
c
e
ss
in
g
.
Evaluation Warning : The document was created with Spire.PDF for Python.