I
AE
S
I
nte
rna
t
io
na
l J
o
urna
l o
f
Art
if
icia
l In
t
ellig
ence
(
I
J
-
AI
)
Vo
l.
9
,
No
.
2
,
J
u
n
e
2020
,
p
p
.
27
6
~
28
3
I
SS
N:
2252
-
8938
,
DOI
: 1
0
.
1
1
5
9
1
/i
j
ai.
v
9
.i
2
.
p
p
27
6
-
28
3
276
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
a
i
.
ia
esco
r
e.
co
m
Vietna
m
ese
ha
nd
w
rit
ten
cha
ra
ct
er
rec
o
g
nition us
ing
co
nv
o
lutiona
l neural netw
o
rk
T
ruo
ng
Q
ua
ng
Vinh
,
L
e
H
o
a
i D
uy
,
Ng
uy
en
T
ha
nh
Nha
n
F
a
c
u
lt
y
o
f
El
e
c
tri
c
a
l
a
n
d
El
e
c
tro
n
ics
En
g
in
e
e
rin
g
,
Ho
C
h
i
M
in
h
Cit
y
Un
iv
e
rsit
y
o
f
T
e
c
h
n
o
lo
g
y
(HC
M
UT
),
V
ietn
a
m
Na
ti
o
n
a
l
Un
iv
e
rsity
–
Ho
Ch
i
M
i
n
h
,
V
iet
n
a
m
Art
icle
I
nfo
AB
ST
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Dec
1
2
,
2
0
19
R
ev
i
s
ed
Feb
10
,
2
0
20
A
cc
ep
ted
A
pr
2
7
,
2
0
20
Ha
n
d
w
rit
in
g
re
c
o
g
n
it
io
n
is
o
n
e
o
f
th
e
c
o
re
a
p
p
li
c
a
ti
o
n
s
o
f
c
o
m
p
u
ter
v
isio
n
f
o
r
re
a
l
-
w
o
rd
p
ro
b
lem
s
a
n
d
it
h
a
s
b
e
e
n
g
a
in
in
g
m
o
re
in
tere
st
b
e
c
a
u
se
o
f
th
e
p
ro
g
re
ss
io
n
i
n
t
h
is
f
ield
.
T
h
is
p
a
p
e
r
p
re
se
n
ts
a
n
e
f
f
i
c
ien
t
m
o
d
e
l
f
o
r
V
ietn
a
m
e
se
h
a
n
d
w
rit
ten
c
h
a
ra
c
ter
re
c
o
g
n
it
io
n
b
y
Co
n
v
o
lu
ti
o
n
a
l
Ne
u
ra
l
Ne
tw
o
rk
s
(CNN
s)
–
a
k
in
d
o
f
d
e
e
p
n
e
u
ra
l
n
e
tw
o
rk
m
o
d
e
l
w
h
ich
c
a
n
a
c
h
iev
e
h
ig
h
p
e
rf
o
rm
a
n
c
e
o
n
h
a
rd
re
c
o
g
n
it
io
n
tas
k
s.
T
h
e
p
ro
p
o
se
d
a
rc
h
it
e
c
tu
re
o
f
th
e
CNN
n
e
t
w
o
rk
f
o
r
V
ietn
a
m
e
se
h
a
n
d
w
rit
ten
c
h
a
ra
c
ter
re
c
o
g
n
it
io
n
c
o
n
sists
o
f
f
iv
e
h
id
d
e
n
lay
e
r
s
in
w
h
ich
th
e
f
irst
3
la
y
e
rs
a
re
c
o
n
v
o
lu
ti
o
n
a
l
lay
e
rs
a
n
d
th
e
las
t
2
la
y
e
rs
a
re
f
u
ll
y
-
c
o
n
n
e
c
ted
lay
e
rs.
O
v
e
r
f
it
ti
n
g
p
ro
b
le
m
is
a
lso
m
in
i
m
ize
d
b
y
u
sin
g
d
ro
p
o
u
t
tec
h
n
i
q
u
e
s
w
it
h
th
e
re
a
so
n
a
b
le
d
r
o
p
ra
te.
T
h
e
e
x
p
e
ri
m
e
n
tal
re
su
lt
s
sh
o
w
th
a
t
o
u
r
m
o
d
e
l
a
c
h
iev
e
s
a
p
p
ro
x
ima
tel
y
9
7
%
a
c
c
u
ra
c
y
.
K
ey
w
o
r
d
s
:
Han
d
w
r
iti
n
g
r
ec
o
g
n
itio
n
C
o
n
v
o
lu
tio
n
al
Ne
u
r
al
Net
w
o
r
k
Ov
er
f
itti
n
g
p
r
o
b
le
m
Vietn
a
m
ese
h
a
n
d
w
r
it
in
g
T
h
is i
s
a
n
o
p
e
n
a
c
c
e
ss
a
rticle
u
n
d
e
r th
e
CC B
Y
-
SA
li
c
e
n
se
.
C
o
r
r
e
s
p
o
nd
ing
A
uth
o
r
:
T
r
u
o
n
g
Qu
a
n
g
V
in
h
,
Facu
lt
y
o
f
E
lectr
ical
a
n
d
E
lect
r
o
n
ics E
n
g
i
n
ee
r
in
g
,
Ho
C
h
i M
i
n
h
C
it
y
U
n
iv
er
s
it
y
o
f
T
ec
h
n
o
lo
g
y
(
H
C
MU
T
)
,
Vietn
a
m
Natio
n
al
U
n
i
v
er
s
it
y
–
Ho
C
h
i M
i
n
h
,
2
6
8
L
y
T
h
u
o
n
g
Kiet,
d
is
tr
ict
1
0
,
Ho
C
h
i M
in
h
cit
y
,
Viet
n
a
m
.
E
m
ail: tq
v
in
h
@
h
c
m
u
t.e
d
u
.
v
n
1.
I
NT
RO
D
UCT
I
O
N
Han
d
w
r
itte
n
ch
ar
ac
ter
r
ec
o
g
n
itio
n
h
as
r
ec
en
tl
y
b
ec
o
m
e
o
n
e
o
f
th
e
m
o
s
t
ch
al
len
g
i
n
g
r
ese
ar
ch
ar
ea
s
w
h
e
n
it
co
m
es
to
p
atter
n
r
ec
o
g
n
itio
n
an
d
i
m
a
g
e
clas
s
i
f
ica
tio
n
[1
-
3]
.
T
h
e
p
u
r
p
o
s
e
o
f
h
an
d
w
r
i
tten
c
h
ar
ac
ter
r
ec
o
g
n
itio
n
is
to
cr
ea
te
th
e
a
b
ilit
y
o
f
co
m
p
u
ter
to
id
en
ti
f
y
an
d
in
ter
p
r
et
in
tell
ig
ib
le
h
a
n
d
w
r
itte
n
.
Ge
n
er
all
y
,
th
er
e
ar
e
t
w
o
ap
p
r
o
ac
h
es
-
O
n
li
n
e
Han
d
w
r
itte
n
r
ec
o
g
n
itio
n
[4
-
5
]
w
h
ic
h
u
s
es
m
o
v
e
m
e
n
t
o
f
th
e
p
en
as
i
n
p
u
t
d
ata
an
d
Of
f
li
n
e
Ha
n
d
w
r
i
tten
R
ec
o
g
n
itio
n
[6
-
7
]
w
h
ic
h
u
s
es
o
n
l
y
i
n
f
o
r
m
atio
n
o
f
i
n
p
u
t
s
ca
n
n
ed
i
m
a
g
es.
T
h
er
e
ar
e
m
an
y
d
i
f
f
icu
l
ties
o
f
h
a
n
d
w
r
iti
n
g
r
ec
o
g
n
itio
n
d
u
e
to
t
h
e
f
ac
t
t
h
at
ea
c
h
p
er
s
o
n
h
as
a
u
n
iq
u
e
w
r
it
in
g
s
t
y
le
an
d
p
e
o
p
le
o
f
ten
w
r
ite
s
o
q
u
i
ck
l
y
t
h
at
m
ak
e
s
t
h
eir
c
h
ar
ac
t
er
’
s
s
h
ap
e
u
n
clea
r
an
d
r
ar
el
y
s
tan
d
ar
d
ized
[8
-
9]
.
Ma
n
y
r
esear
c
h
er
s
a
n
d
s
cien
t
is
ts
h
a
v
e
co
n
ce
n
tr
ated
o
n
n
e
w
t
ec
h
n
iq
u
es
a
n
d
m
et
h
o
d
s
th
a
t
wo
u
ld
ad
d
r
ess
th
ese
p
r
o
b
lem
s
w
it
h
a
h
ig
h
er
r
ate
o
f
ac
cu
r
ac
y
.
So
m
e
co
m
p
etit
i
o
n
s
w
er
e
o
p
en
ed
to
f
i
n
d
th
e
b
est
alg
o
r
it
h
m
f
o
r
h
an
d
w
r
iti
n
g
r
ec
o
g
n
i
tio
n
w
it
h
t
h
e
d
ed
icate
d
d
a
taset [
1
0
-
1
1
]
.
I
n
t
h
e
p
ast
f
e
w
d
ec
ad
es,
m
a
n
y
b
r
ea
k
th
r
o
u
g
h
s
h
a
v
e
b
ee
n
m
a
d
e
in
m
ac
h
in
e
lear
n
i
n
g
,
esp
ec
iall
y
Dee
p
L
ea
r
n
i
n
g
.
T
h
is
r
es
u
lted
i
n
t
h
e
cr
ea
tio
n
o
f
v
ar
io
u
s
al
g
o
r
ith
m
s
th
at
h
a
v
e
s
h
o
w
n
t
h
eir
i
m
p
r
o
v
ed
ef
f
icie
n
c
y
i
n
m
an
y
i
m
p
o
r
tan
t
ar
ea
s
s
u
c
h
a
s
co
m
p
u
ter
v
is
io
n
,
n
at
u
r
al
la
n
g
u
a
g
e
p
r
o
ce
s
s
in
g
,
a
n
d
b
ig
d
ata.
Ma
n
y
m
ac
h
i
n
e
lear
n
in
g
alg
o
r
it
h
m
s
in
c
lu
d
i
n
g
Su
p
p
o
r
t
Vec
to
r
Ma
ch
in
es
(
S
VM
)
[
1
2
]
,
Naiv
e
B
ay
es
C
la
s
s
if
ica
tio
n
[
1
3
]
,
an
d
d
ee
p
lear
n
in
g
al
g
o
r
ith
m
s
s
u
ch
as
C
o
n
v
o
lu
t
io
n
al
Ne
u
r
al
Net
w
o
r
k
s
(
C
NN
s
)
[
1
4
]
,
R
ec
u
r
r
en
t
Ne
u
r
al
Net
w
o
r
k
s
(
R
NNs)
[
1
5
]
,
co
m
b
i
n
ed
w
it
h
t
h
e
d
ev
elo
p
m
e
n
t
b
e
y
o
n
d
th
e
co
m
p
u
ti
n
g
s
p
ee
d
o
f
h
ar
d
w
ar
e
d
e
v
ices
h
a
v
e
cr
ea
ted
m
o
d
el
s
th
at
ca
n
o
u
tp
er
f
o
r
m
h
u
m
a
n
p
er
f
o
r
m
a
n
ce
in
m
an
y
co
m
p
u
ter
v
is
io
n
tas
k
s
,
esp
e
ciall
y
clas
s
i
f
icatio
n
task
s
s
u
ch
a
s
h
a
n
d
w
r
i
tin
g
r
ec
o
g
n
i
tio
n
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
A
r
ti
f
I
n
tell
I
SS
N:
2252
-
8938
V
ietn
a
mese
h
a
n
d
w
r
itten
ch
a
r
a
cter reco
g
n
itio
n
u
s
in
g
co
n
vo
lu
tio
n
a
l n
eu
r
a
l
n
etw
o
r
k
(
T
r
u
o
n
g
Qu
a
n
g
V
in
h
)
277
Ma
n
y
r
e
s
ea
r
ch
es
o
n
h
a
n
d
w
r
itt
en
ch
ar
ac
ter
r
ec
o
g
n
itio
n
f
o
r
E
n
g
l
is
h
,
J
ap
an
ese,
an
d
C
h
i
n
ese
lan
g
u
a
g
e
h
av
e
b
ee
n
p
u
b
lis
h
ed
[
1
6
-
1
8
]
.
Ho
w
ev
er
,
t
h
er
e
ar
e
f
e
w
o
f
s
tu
d
ie
s
o
n
Viet
n
a
m
e
s
e
h
a
n
d
w
r
iti
n
g
r
ec
o
g
n
i
tio
n
.
Vietn
a
m
ese
alp
h
ab
et
i
s
b
ased
o
n
t
h
e
L
ati
n
s
cr
ip
t
w
i
th
t
h
e
a
d
d
itio
n
o
f
n
in
e
ac
ce
n
t
m
ar
k
s
t
o
cr
ea
te
ad
d
itio
n
al
s
o
u
n
d
s
,
an
d
f
i
v
e
s
tr
o
k
es
to
ex
p
r
ess
th
e
to
n
e
o
f
w
o
r
d
s
.
T
h
er
ef
o
r
e,
it
is
a
ch
allen
g
e
f
o
r
r
esear
ch
er
s
to
d
ev
elo
p
alg
o
r
ith
m
s
to
r
ec
o
g
n
ize
Vie
tn
a
m
ese
h
an
d
w
r
itte
n
ch
ar
ac
ter
s
.
So
m
e
a
u
t
h
o
r
s
p
r
esen
ted
Vietn
a
m
e
s
e
h
a
n
d
w
r
iti
n
g
r
ec
o
g
n
itio
n
m
et
h
o
d
s
u
s
i
n
g
s
u
p
p
o
r
t v
ec
to
r
m
ac
h
i
n
e
(
SVM)
with
f
ea
tu
r
e
e
x
tr
ac
tio
n
m
et
h
o
d
s
[
1
9
-
2
0
]
.
Ho
w
ev
er
,
th
e
p
er
f
o
r
m
a
n
ce
d
ep
en
d
s
o
n
t
h
e
r
o
b
u
s
tn
e
s
s
o
f
f
ea
t
u
r
e
ex
tr
a
ctio
n
m
et
h
o
d
s
an
d
th
u
s
th
e
ac
cu
r
ac
y
is
d
ec
r
ea
s
ed
w
h
e
n
th
e
h
a
n
d
w
r
i
tten
c
h
ar
ac
ter
s
ar
e
n
atu
r
al
an
d
u
n
co
n
s
tr
ai
n
ed
.
So
m
e
m
o
d
els
o
f
d
ee
p
lea
r
n
in
g
s
u
ch
as
C
N
N,
R
NN,
a
n
d
B
id
ir
ec
tio
n
al
L
o
n
g
Sh
o
r
t
-
T
er
m
Me
m
o
r
y
h
av
e
b
ee
n
ap
p
lied
to
i
m
p
r
o
v
e
t
h
e
r
ec
o
g
n
itio
n
ac
c
u
r
ac
y
r
ate
[
2
1
-
2
2
]
.
Nev
er
th
eless
,
ac
ce
n
t
m
ar
k
s
an
d
s
tr
o
k
es
i
n
Vi
etn
a
m
ese
lan
g
u
ag
e
m
ak
e
it
d
i
f
f
icu
l
t
to
r
ec
o
g
n
ize
h
an
d
w
r
iti
n
g
co
r
r
ec
tl
y
.
I
n
th
i
s
p
ap
er
,
Vietn
a
m
ese
h
a
n
d
w
r
itte
n
ch
ar
ac
ter
d
ataset
f
o
r
tr
ain
i
n
g
a
s
u
p
er
v
i
s
ed
m
o
d
el
an
d
a
C
NN
ar
ch
itect
u
r
e
f
o
r
Vietn
a
m
ese
h
an
d
w
r
i
tten
c
h
ar
ac
ter
r
ec
o
g
n
iti
o
n
ar
e
p
r
esen
ted
.
A
tech
n
iq
u
e
ca
lled
d
r
o
p
o
u
t
is
ap
p
lied
to
p
r
ev
en
t
o
u
r
m
o
d
el
f
r
o
m
o
v
er
f
itti
n
g
,
to
g
u
ar
an
te
e
th
e
g
e
n
er
aliza
tio
n
p
r
o
p
er
ty
an
d
to
i
m
p
r
o
v
e
it
s
p
er
f
o
r
m
a
n
ce
.
T
h
is
tech
n
iq
u
e
in
cl
u
d
es
a
u
n
it
’
s
te
m
p
o
r
ar
y
r
e
m
o
v
al
f
r
o
m
t
h
e
n
et
w
o
r
k
.
T
h
is
r
e
m
o
v
ed
u
n
it
i
s
s
elec
ted
in
a
r
an
d
o
m
w
a
y
o
n
l
y
d
u
r
i
n
g
t
h
e
tr
ain
in
g
s
ta
g
e.
T
h
e
g
o
al
o
f
o
u
r
w
o
r
k
i
s
to
cr
ea
te
a
Vietn
a
m
ese
h
an
d
w
r
itte
n
c
h
ar
ac
ter
d
ataset
co
n
tain
i
n
g
2
9
Vietn
a
m
ese
c
h
ar
ac
ter
s
an
d
to
b
u
ild
a
d
ee
p
l
ea
r
n
in
g
m
o
d
el
t
h
at
ca
n
ac
h
ie
v
e
a
h
i
g
h
d
eg
r
ee
o
f
a
cc
u
r
ac
y
f
o
r
Vietn
a
m
ese
h
an
d
w
r
itte
n
c
h
ar
ac
ter
r
ec
o
g
n
itio
n
.
T
h
e
r
e
s
t
o
f
th
e
p
ap
er
is
co
m
p
o
s
ed
as
f
o
llo
w
s
.
I
n
s
ec
tio
n
I
I
,
th
e
o
v
er
v
ie
w
o
f
th
e
C
NN
m
o
d
el
an
d
th
e
p
r
o
p
o
s
ed
C
NN
ar
ch
i
tectu
r
e
f
o
r
Vietn
a
m
ese
h
a
n
d
w
r
it
ten
c
h
ar
ac
ter
r
ec
o
g
n
i
tio
n
ar
e
p
r
ese
n
ted
.
I
n
s
ec
tio
n
I
I
I
,
th
e
ex
p
er
i
m
en
tal
r
esu
l
ts
ar
e
s
h
o
w
n
in
d
etail.
Fi
n
all
y
,
in
t
h
e
la
s
t sectio
n
,
t
h
e
co
n
cl
u
s
io
n
is
g
i
v
en
.
2.
RE
S
E
ARCH
M
E
T
H
O
D
I
n
th
is
s
ec
t
io
n
,
th
e
o
v
er
v
ie
w
ar
ch
itec
tu
r
e
o
f
o
u
r
o
f
f
l
i
n
e
Vietn
a
m
e
s
e
h
an
d
w
r
it
te
n
ch
ar
ac
ter
r
ec
o
g
n
itio
n
s
y
s
te
m
b
ased
o
n
C
o
n
v
o
lu
tio
n
al
n
eu
r
al
n
et
w
o
r
k
is
p
r
esen
ted
.
B
esid
es,
a
b
r
ief
l
y
d
is
c
u
s
s
io
n
o
f
t
h
e
f
u
n
d
a
m
en
ta
l
f
ea
t
u
r
es
o
f
C
N
N
an
d
a
d
r
o
p
o
u
t
tech
n
iq
u
e
w
h
ic
h
co
m
b
in
e
s
f
u
ll
y
-
co
n
n
ec
ted
lay
er
s
to
p
r
ev
en
t
o
v
er
f
itti
n
g
p
h
e
n
o
m
e
n
o
n
f
o
r
th
e
p
r
o
p
o
s
e
m
o
d
el
ar
e
d
escr
ib
e
d
.
2
.
1
.
Co
nv
o
lutio
na
l neura
l net
w
o
rks
(
CNNs)
C
NN
i
s
t
h
e
cu
r
r
e
n
t
s
ta
te
-
of
-
th
e
-
ar
t
m
o
d
el
ar
c
h
itect
u
r
e
f
o
r
i
m
ag
e
clas
s
i
f
icatio
n
tas
k
s
[
2
3
]
.
I
t
h
as
b
ee
n
e
m
p
lo
y
ed
to
s
o
l
v
e
a
v
ar
iet
y
o
f
tas
k
s
in
cl
u
d
i
n
g
i
m
a
g
e
r
ec
o
g
n
itio
n
p
r
o
b
le
m
s
,
n
at
u
r
al
la
n
g
u
a
g
e
p
r
o
ce
s
s
i
n
g
ch
alle
n
g
e
s
b
y
ta
k
i
n
g
ad
v
an
ta
g
e
o
f
th
e
w
a
y
t
h
e
i
n
p
u
t
i
m
a
g
es
s
en
s
ib
l
y
co
n
s
tr
ai
n
th
e
ar
c
h
ite
ct
u
r
e.
C
NN
s
lear
n
s
f
ea
t
u
r
es
o
f
i
m
a
g
es
o
n
s
u
b
-
r
eg
i
o
n
s
i
n
s
tead
o
f
s
ca
lin
g
f
u
ll
i
m
a
g
es.
I
n
p
ar
tic
u
lar
,
C
NN
s
ar
r
an
g
e
la
y
er
s
an
d
t
h
eir
n
eu
r
o
n
s
i
n
3
d
i
m
e
n
s
io
n
alitie
s
:
w
id
th
,
h
ei
g
h
t a
n
d
d
ep
th
.
C
N
Ns p
as
s
es
th
e
i
m
ag
e
s
w
h
ic
h
c
o
n
tain
s
t
h
e
r
a
w
p
ix
e
l
d
ata
th
r
o
u
g
h
a
s
er
ie
s
o
f
f
ilter
s
to
ex
tr
ac
t
an
d
lear
n
h
ig
h
er
-
le
v
el
f
ea
t
u
r
es,
w
h
ich
ca
n
b
e
u
s
e
d
f
o
r
class
if
icat
io
n
p
h
ase
o
f
th
e
m
o
d
el.
C
NN
is
b
u
ilt
f
r
o
m
th
r
ee
co
m
p
o
n
e
n
ts
.
C
o
n
v
o
lu
tio
n
al
la
y
er
s
(
C
o
n
v
la
y
er
s
)
ar
e
th
e
m
aj
o
r
b
u
ild
in
g
b
lo
ck
o
f
a
C
o
n
v
o
lu
ti
o
n
al
Net
w
o
r
k
w
h
ic
h
m
ak
e
u
s
e
o
f
a
s
p
ec
i
f
ic
a
m
o
u
n
t
o
f
co
n
v
o
lu
tio
n
f
ilter
s
f
o
r
th
e
i
m
a
g
e.
Fo
r
ea
c
h
s
u
b
-
r
e
g
io
n
,
C
o
n
v
la
y
er
s
p
er
f
o
r
m
co
n
v
o
lu
tio
n
o
p
er
atio
n
o
f
th
e
s
u
b
-
r
eg
io
n
co
n
s
id
er
ed
as
a
m
atr
i
x
o
f
p
ix
el
v
alu
e
s
a
n
d
t
h
e
f
il
ter
co
n
s
id
er
ed
as
a
m
atr
i
x
o
f
r
ea
l
n
u
m
b
er
s
to
p
r
o
d
u
ce
a
s
in
g
le
v
al
u
e
in
th
e
o
u
tp
u
t
f
ea
tu
r
e
m
ap
.
C
o
n
v
o
lu
tio
n
la
y
er
s
k
ee
p
th
e
s
p
atial
r
elatio
n
s
h
ip
b
et
w
ee
n
p
ix
el
s
b
y
lear
n
i
n
g
i
m
a
g
e
f
ea
t
u
r
es
u
s
in
g
s
u
b
-
r
e
g
io
n
o
f
in
p
u
t
d
ata.
An
ex
tr
a
o
p
er
a
tio
n
ca
lled
R
e
L
U
h
a
s
b
ee
n
u
s
ed
af
ter
ev
er
y
C
o
n
v
o
lu
tio
n
la
y
er
.
R
e
L
U
ac
t
s
as
a
n
o
n
-
li
n
ea
r
a
cti
v
atio
n
f
u
n
ct
io
n
i
n
tr
o
d
u
ci
n
g
n
o
n
-
li
n
ea
r
it
y
i
n
to
t
h
e
m
o
d
el,
w
h
ic
h
is
r
ea
ll
y
co
m
m
o
n
i
n
r
ea
l
-
w
o
r
ld
d
ata.
P
o
o
lin
g
la
y
er
s
i
n
v
o
lv
e
d
o
w
n
-
s
a
m
p
li
n
g
t
h
e
i
m
ag
e
d
ata
e
x
tr
ac
ted
b
y
t
h
e
co
n
v
o
l
u
tio
n
al
la
y
er
s
to
r
ed
u
ce
th
e
d
i
m
en
s
io
n
o
f
ea
c
h
f
ea
t
u
r
e
m
ap
b
u
t
s
till
r
etai
n
s
th
e
m
o
s
t
i
m
p
o
r
tan
t
i
n
f
o
r
m
atio
n
s
o
as
to
m
i
n
i
m
iz
e
p
r
o
ce
s
s
in
g
ti
m
e.
A
co
m
m
o
n
l
y
u
s
ed
p
o
o
lin
g
a
lg
o
r
it
h
m
is
k
n
o
w
n
as
m
a
x
p
o
o
lin
g
,
w
h
ic
h
d
ef
in
es
s
u
b
-
r
eg
io
n
s
o
f
f
ea
t
u
r
e
m
ap
b
y
p
ick
i
n
g
o
u
t t
h
e
lar
g
est ele
m
en
t a
n
d
g
et
tin
g
r
i
d
o
f
o
th
er
ele
m
en
ts
.
Den
s
e
(
f
u
ll
y
co
n
n
ec
t
ed
)
la
y
e
r
s
ar
e
tr
ad
itio
n
al
Mu
lti
-
la
y
er
P
er
ce
p
tr
o
n
,
w
h
ic
h
clas
s
i
f
y
t
h
e
f
ea
t
u
r
es
ex
tr
ac
ted
b
y
t
h
e
co
n
v
o
l
u
tio
n
a
l
la
y
er
s
an
d
ar
e
d
o
w
n
-
s
a
m
p
le
d
b
y
th
e
p
o
o
lin
g
la
y
er
s
.
I
n
a
d
en
s
e
la
y
er
,
ev
er
y
n
o
d
e
in
t
h
e
la
y
er
i
s
co
n
n
ec
te
d
to
ev
er
y
n
o
d
e
in
th
e
p
r
ec
e
d
in
g
la
y
er
.
Fo
r
class
i
f
ica
tio
n
task
,
a
n
ac
ti
v
atio
n
f
u
n
ctio
n
ca
lled
s
o
f
t
m
a
x
r
eg
r
es
s
io
n
is
u
s
ed
m
o
s
t o
f
te
n
.
T
y
p
icall
y
,
a
C
NN
co
n
s
is
t
s
o
f
co
n
v
o
l
u
tio
n
al
la
y
er
s
,
ea
ch
o
f
w
h
ich
is
f
o
llo
w
ed
b
y
a
p
o
o
lin
g
la
y
er
.
T
h
ey
ar
e
ca
lled
co
n
v
o
lu
tio
n
al
m
o
d
u
les
a
n
d
r
esp
o
n
s
ib
l
e
f
o
r
p
er
f
o
r
m
in
g
f
ea
tu
r
e
e
x
tr
ac
tio
n
.
T
h
e
last
co
n
v
o
lu
tio
n
al
m
o
d
u
le
p
r
o
ce
ed
s
af
ter
o
n
e
o
r
m
o
r
e
d
en
s
e
la
y
e
r
s
th
at
p
er
f
o
r
m
s
clas
s
if
icatio
n
task
.
2
.
2
.
Dro
po
ut
t
ec
hn
iqu
e
A
lar
g
e
n
e
u
r
al
n
et
w
o
r
k
co
n
t
ain
s
m
u
ltip
le
n
o
n
-
li
n
ea
r
h
id
d
en
la
y
er
s
ca
n
h
a
v
e
a
v
er
y
e
x
p
r
ess
i
v
e
p
er
f
o
r
m
a
n
ce
o
n
tr
ain
in
g
s
et
b
u
t
ac
h
ie
v
e
a
b
ad
r
esu
lt
o
n
test
s
et
[
2
4
-
2
5
]
.
T
h
is
is
o
v
er
f
itti
n
g
p
h
en
o
m
e
n
o
n
.
T
h
i
s
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8938
I
n
t J
A
r
ti
f
I
n
tell
,
Vo
l.
9
,
No
.
2
,
J
u
n
e
20
20
:
27
6
–
28
3
278
is
t
h
e
ca
s
e
w
h
e
n
s
a
m
p
li
n
g
n
o
is
e
is
b
r
o
u
g
h
t a
b
o
u
t b
y
t
h
e
lac
k
o
f
tr
ai
n
in
g
d
ata
f
o
r
a
co
m
p
le
x
m
o
d
el.
Dr
o
p
o
u
t is
a
tech
n
ical
s
o
l
u
tio
n
to
th
i
s
p
r
o
b
lem
.
T
h
is
d
r
o
p
u
n
its
(
h
id
d
en
an
d
v
is
ib
le)
f
r
o
m
t
h
e
n
e
u
r
al
n
et
w
o
r
k
d
u
r
i
n
g
tr
ain
i
n
g
.
T
h
e
n
u
m
b
er
o
f
d
r
o
p
p
ed
u
n
i
ts
w
ill
b
e
d
eter
m
in
ed
b
y
p
r
o
b
ab
ilit
y
p
(
d
r
o
p
o
u
t
r
ate)
.
Su
c
h
d
r
o
p
p
ed
u
n
it
s
ar
e
ch
o
s
en
r
an
d
o
m
l
y
.
2
.
3
.
ReLU a
ct
iv
a
t
io
n f
un
ct
io
n
R
eL
U
(
R
ec
ti
f
ied
L
i
n
ea
r
U
n
i
t)
f
u
n
ctio
n
i
s
a
n
o
n
-
li
n
ea
r
f
u
n
ct
io
n
.
T
o
d
ay
,
m
an
y
r
esear
ch
er
s
a
n
d
s
cien
t
is
t
s
u
s
e
R
e
L
U
f
u
n
ctio
n
to
in
tr
o
d
u
ce
n
o
n
-
li
n
ea
r
p
r
o
p
er
t
y
in
to
d
ee
p
n
eu
r
al
n
et
w
o
r
k
m
o
d
el
i
n
s
tead
o
f
u
s
i
n
g
o
th
er
ac
tiv
atio
n
f
u
n
ctio
n
s
.
R
eL
U
f
u
n
ctio
n
m
a
k
es
th
e
tr
ain
i
n
g
o
f
t
h
e
m
o
d
el
f
aster
th
a
n
r
e
g
u
lar
ac
ti
v
atio
n
f
u
n
ctio
n
s
s
u
c
h
as
s
i
g
m
o
id
an
d
tan
h
f
u
n
ctio
n
s
,
b
ec
au
s
e
it
d
o
es
n
o
t
co
m
p
u
te
e
x
p
o
n
en
t
ials
an
d
d
iv
is
io
n
s
.
T
h
e
f
o
r
m
u
la
o
f
R
e
L
U
is
f
(
x
)
=
m
a
x
(
0
,
x
)
.
2
.
4
.
T
he
CNN
a
rc
hite
ct
ure
f
o
r
Viet
na
m
e
s
e
ha
nd
w
rit
t
en
cha
ra
ct
er
re
co
g
nitio
n:
T
h
e
p
r
o
p
o
s
ed
C
NN
ar
c
h
itect
u
r
e
is
s
h
o
w
n
i
n
Fi
g
u
r
e
1
.
I
t
c
o
n
tain
s
f
i
v
e
h
id
d
en
la
y
er
s
w
i
t
h
w
ei
g
h
ts
.
T
h
e
f
ir
s
t
th
r
ee
ar
e
co
n
v
o
l
u
tio
n
al
la
y
er
s
an
d
th
e
r
e
m
ai
n
in
g
t
wo
ar
e
f
u
ll
y
-
co
n
n
ec
ted
la
y
er
s
.
E
ac
h
co
n
v
o
l
u
tio
n
a
l
la
y
er
is
f
o
llo
w
ed
b
y
o
n
e
R
e
L
U
ac
tiv
atio
n
f
u
n
ctio
n
an
d
o
n
e
m
ax
p
o
o
lin
g
la
y
er
w
it
h
a
f
ilte
r
o
f
s
ize
2
x
2
an
d
a
s
tr
id
e
o
f
2
p
ix
els.
A
ls
o
,
th
e
f
ir
s
t f
u
ll
y
-
co
n
n
ec
ted
la
y
er
is
f
o
ll
o
w
ed
b
y
o
n
e
d
r
o
p
o
u
t la
y
er
.
T
h
e
f
ir
s
t
co
n
v
o
l
u
tio
n
al
la
y
er
a
p
p
lies
3
2
k
er
n
els
o
f
s
ize
1
1
x
1
1
(
ex
tr
ac
tin
g
1
1
x
1
1
s
u
b
-
r
eg
io
n
)
w
i
th
a
s
tr
id
e
o
f
2
p
i
x
els.
I
t ta
k
es a
1
0
0
x
1
0
0
im
a
g
e
a
s
a
n
i
n
p
u
t.
T
h
e
s
ec
o
n
d
co
n
v
o
lu
tio
n
al
la
y
e
r
u
s
e
s
o
u
tp
u
ts
o
f
t
h
e
f
ir
s
t
co
n
v
o
l
u
tio
n
a
l
la
y
er
as
a
n
in
p
u
t,
a
n
d
f
il
ter
s
it
w
i
th
6
4
k
er
n
els
o
f
s
ize
5
x
5
w
it
h
a
s
tr
id
e
o
f
2
p
ix
els.
T
h
e
th
ir
d
co
n
v
o
l
u
tio
n
al
la
y
er
h
as
1
2
8
k
er
n
els
o
f
s
ize
3
x
3
w
h
ich
ar
e
co
n
n
ec
ted
to
th
e
o
u
t
p
u
ts
o
f
t
h
e
s
ec
o
n
d
co
n
v
o
lu
tio
n
al
la
y
er
.
T
h
e
f
ir
s
t
f
u
ll
y
-
co
n
n
ec
ted
la
y
er
s
(
d
en
s
e
la
y
er
)
h
av
e
1
0
2
4
n
eu
r
o
n
s
.
T
h
ey
ar
e
co
m
b
i
n
ed
w
it
h
d
r
o
p
o
u
t
r
eg
u
lar
izat
io
n
r
ate
o
f
0
.
4
(
4
0
%
o
f
n
eu
r
o
n
s
w
il
l
b
e
d
r
o
p
p
e
d
d
u
r
in
g
tr
ain
i
n
g
)
.
T
h
e
la
s
t
f
u
l
l
y
-
co
n
n
ec
ted
la
y
er
s
(
d
en
s
e
la
y
er
)
h
av
e
2
9
n
e
u
r
o
n
s
eq
u
al
to
th
e
n
u
m
b
er
o
f
clas
s
.
T
h
e
o
u
tp
u
t
la
y
er
is
a
s
o
f
t
m
ax
la
y
er
w
h
ic
h
ta
k
es
th
e
o
u
tp
u
t
o
f
las
t
f
u
ll
y
-
co
n
n
ec
ted
la
y
er
s
as
an
i
n
p
u
t
an
d
w
i
ll
r
et
u
r
n
t
h
e
r
a
w
v
a
lu
e
s
f
o
r
o
u
r
p
r
ed
ictio
n
s
.
T
h
ese
v
alu
e
s
ar
e
th
e
p
r
o
b
ab
ilit
y
v
al
u
es t
h
at
r
ep
r
esen
t t
h
e
p
r
o
b
ab
ilit
y
o
f
ea
ch
c
lass
'
s
s
a
m
p
le.
Fig
u
r
e
1
.
An
ill
u
s
tr
atio
n
o
f
th
e
ar
ch
itectu
r
e
o
f
o
u
r
C
NN
T
h
e
ar
ch
itect
u
r
e
is
t
h
e
co
m
b
in
at
io
n
o
f
t
w
o
b
lo
ck
s
:
Featu
r
e
E
x
tr
ac
tio
n
b
lo
ck
a
n
d
Feat
u
r
e
C
las
s
i
f
icatio
n
b
lo
ck
.
T
h
e
Fe
atu
r
e
E
x
tr
ac
tio
n
tak
e
s
an
o
r
i
g
in
a
l
i
m
ag
e
a
s
an
i
n
p
u
t
a
n
d
ex
tr
ac
t
ab
s
tr
ac
ti
v
e
f
ea
t
u
r
e
b
y
u
s
i
n
g
C
o
n
v
o
l
u
tio
n
a
l
la
y
er
.
T
h
e
Featu
r
e
C
las
s
if
ica
tio
n
tak
e
s
th
e
o
u
tp
u
t
o
f
t
h
e
Fe
atu
r
e
E
x
tr
ac
tio
n
a
s
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
A
r
ti
f
I
n
tell
I
SS
N:
2252
-
8938
V
ietn
a
mese
h
a
n
d
w
r
itten
ch
a
r
a
cter reco
g
n
itio
n
u
s
in
g
co
n
vo
lu
tio
n
a
l n
eu
r
a
l
n
etw
o
r
k
(
T
r
u
o
n
g
Qu
a
n
g
V
in
h
)
279
an
in
p
u
t
an
d
f
ee
d
s
th
e
s
e
f
ea
t
u
r
es
in
to
a
d
ee
p
n
eu
r
al
n
et
w
o
r
k
to
p
e
r
f
o
r
m
t
h
e
class
if
ica
tio
n
t
ask
.
I
n
th
e
Feat
u
r
e
C
las
s
i
f
icatio
n
b
lo
ck
,
d
r
o
p
o
u
t
u
n
it
is
u
s
ed
to
p
r
ev
en
t
o
v
er
f
it
tin
g
p
h
e
n
o
m
e
n
o
n
,
an
d
th
u
s
i
m
p
r
o
v
e
t
h
e
p
er
f
o
r
m
a
n
ce
o
f
t
h
e
m
o
d
el.
3.
E
XP
E
R
I
M
E
NT
A
L
RE
SUL
T
S
I
n
th
i
s
s
ec
tio
n
,
th
e
p
r
o
ce
s
s
to
co
llect
th
e
i
m
a
g
es
o
f
Viet
n
a
m
ese
h
an
d
w
r
itte
n
c
h
ar
ac
ter
s
t
o
b
u
ild
th
e
d
ataset
is
d
escr
ib
ed
.
Nex
t,
t
h
e
tr
ain
i
n
g
p
r
o
ce
s
s
an
d
th
e
test
i
n
g
r
es
u
lts
ar
e
p
r
esen
ted
.
Fin
all
y
,
th
e
d
e
m
o
ap
p
licatio
n
w
h
ic
h
ca
n
r
ec
o
g
n
i
ze
Vietn
a
m
ese
h
an
d
w
r
itte
n
c
h
ar
ac
ter
s
is
s
h
o
w
n
.
3
.
1
.
Da
t
a
s
et
Un
til
r
ec
en
tl
y
,
d
atase
ts
o
f
la
b
eled
i
m
ag
e
s
o
f
Viet
n
a
m
ese
h
a
n
d
w
r
i
tten
ch
ar
ac
ter
s
h
a
v
e
n
o
t
b
ee
n
p
u
b
lis
h
ed
.
T
h
e
2
9
Vietn
a
m
e
s
e
ch
ar
ac
ter
s
w
a
s
co
n
s
tr
u
cted
f
r
o
m
2
3
L
ati
n
alp
h
ab
et
a
n
d
6
u
n
iq
u
e
c
h
ar
ac
ter
s
.
T
h
er
ef
o
r
e,
w
e
h
a
v
e
co
llected
s
ix
Vietn
a
m
e
s
e
s
p
ec
ial
c
h
ar
ac
ter
ex
a
m
p
le
s
w
r
itte
n
b
y
5
0
0
d
if
f
er
en
t
p
er
s
o
n
s
co
m
b
i
n
e
w
i
th
m
o
d
if
ied
ex
a
m
p
les
o
f
2
3
L
at
in
a
lp
h
ab
et
i
n
NI
ST
Sp
ec
ial
Data
b
ase
1
9
w
h
ich
co
n
tai
n
s
b
i
n
ar
y
i
m
a
g
es
o
f
h
an
d
w
r
itte
n
c
h
ar
a
cter
s
to
cr
ea
te
Viet
n
a
m
ese
h
an
d
w
r
i
tten
ch
ar
ac
ter
d
ataset
f
o
r
th
e
p
u
r
p
o
s
e
o
f
tr
ain
i
n
g
a
n
d
test
in
g
a
s
u
p
er
v
i
s
ed
lear
n
in
g
m
o
d
el
f
o
r
class
if
i
ca
tio
n
tas
k
.
T
h
e
s
i
x
Viet
n
a
m
e
s
e
s
p
ec
ial
c
h
ar
ac
ter
s
et
is
co
m
p
o
s
ed
o
f
1
5
,
0
0
0
p
a
tter
n
s
t
h
at
w
as
co
llected
b
y
t
w
o
m
e
th
o
d
s
.
T
h
e
f
ir
s
t
m
et
h
o
d
is
th
at
an
a
n
d
r
o
id
ap
p
licatio
n
is
d
e
v
elo
p
ed
to
co
llect
h
a
n
d
w
r
iti
n
g
i
m
a
g
es
o
n
m
o
b
ile
d
ev
ice.
T
h
e
ap
p
licatio
n
allo
w
s
u
s
er
s
w
r
ite
a
letter
o
n
s
cr
ee
n
a
n
d
th
e
n
s
e
n
d
th
is
i
m
a
g
e
o
f
letter
to
o
u
r
s
er
v
er
.
A
lt
h
o
u
g
h
th
is
m
et
h
o
d
is
v
er
y
f
a
s
t
f
o
r
co
llectin
g
a
n
d
p
r
o
ce
s
s
in
g
d
at
a
b
ec
au
s
e
i
t
is
d
o
n
e
a
u
to
m
at
i
ca
ll
y
,
t
h
e
f
ac
t
t
h
at
it
i
s
q
u
i
te
d
if
f
ic
u
lt
to
r
ea
ch
a
s
u
f
f
icie
n
t
n
u
m
b
er
o
f
u
s
er
s
en
s
u
r
i
n
g
t
h
e
d
iv
er
s
it
y
o
f
d
ata
s
o
u
r
ce
s
.
T
h
er
ef
o
r
e
,
th
e
s
ec
o
n
d
m
et
h
o
d
w
h
ich
u
s
es
h
an
d
w
r
iti
n
g
o
n
co
llected
p
ap
e
r
s
o
f
HC
M
UT
s
tu
d
en
t
s
is
p
r
o
ce
ed
ed
.
T
h
e
s
tu
d
en
ts
w
r
o
te
ch
ar
ac
ter
s
o
n
g
r
id
p
ap
er
.
T
h
en
,
w
e
s
ca
n
n
ed
th
o
s
e
p
ap
er
s
an
d
s
ep
ar
ated
th
em
in
to
s
in
g
le
ch
ar
ac
ter
i
m
ag
e
s
.
T
h
e
o
r
ig
in
al
b
lack
an
d
w
h
i
te
i
m
ag
e
s
w
er
e
s
ize
-
n
o
r
m
al
i
ze
d
an
d
ce
n
ter
ed
in
a
1
0
0
x
1
0
0
b
o
x
.
T
h
is
m
et
h
od
tak
es
a
lo
t
o
f
e
f
f
o
r
t
to
p
r
o
ce
s
s
d
ata
m
a
n
u
a
ll
y
,
b
u
t
it
en
s
u
r
es
th
e
d
iv
er
s
i
t
y
o
f
w
r
i
t
er
s
.
B
y
co
m
b
i
n
i
n
g
th
e
s
e
t
w
o
m
et
h
o
d
s
,
w
e
h
a
v
e
ac
h
iev
ed
t
w
o
cr
iter
ia:
to
co
llect
d
ata
q
u
ick
l
y
a
n
d
m
a
k
e
s
u
r
e
th
e
v
ar
iet
y
o
f
d
ata
s
o
u
r
ce
s
.
Ou
r
f
i
n
al
Viet
n
a
m
ese
h
an
d
w
r
itte
n
c
h
ar
ac
ter
d
ataset
co
n
tai
n
s
7
2
,
5
0
0
lab
elled
im
a
g
es
o
f
2
9
Vietn
a
m
ese
h
an
d
w
r
itte
n
ch
ar
ac
ter
s
th
at
w
as
co
n
s
tr
u
cted
f
r
o
m
5
7
,
5
0
0
p
atter
n
s
o
f
NI
ST
Data
s
et
an
d
1
5
,
0
0
0
p
atter
n
s
o
f
o
u
r
s
ix
Viet
n
a
m
e
s
e
u
n
iq
u
e
ch
ar
ac
ter
.
I
t
w
as
d
i
v
id
ed
in
to
th
r
ee
s
et
s
:
5
8
,
0
0
0
e
x
a
m
p
le
s
i
n
tr
ain
i
n
g
s
et,
7
,
2
5
0
ex
am
p
les i
n
v
a
lid
ati
o
n
s
et,
an
d
7
,
2
5
0
ex
am
p
le
s
in
test
i
n
g
s
e
t
.
I
n
o
r
d
er
to
co
llect
d
ata
co
n
v
en
ien
tl
y
,
w
e
d
ev
elo
p
an
An
d
r
o
id
m
o
b
ile
ap
p
licatio
n
ca
lled
Ho
d
y
P
ain
t
f
o
r
co
llectin
g
d
ata
f
r
o
m
i
n
ter
n
et
u
s
er
s
as
s
h
o
w
n
in
F
ig
u
r
e
2
.
User
s
ca
n
d
r
aw
a
ch
ar
ac
t
er
o
n
th
e
p
ain
tin
g
w
i
n
d
o
w
,
s
e
n
d
an
i
m
a
g
e
o
f
t
h
is
ch
ar
ac
ter
to
o
u
r
d
atab
ase
b
y
u
p
lo
ad
b
u
tto
n
,
p
r
ess
c
lear
b
u
tto
n
,
a
n
d
r
ep
ea
t
th
ese
s
tep
s
ab
o
v
e
f
o
r
n
e
w
let
te
r
s
.
Fig
u
r
e
2.
Mo
b
ile
ap
p
licatio
n
f
o
r
co
llectin
g
d
ata
Fig
u
r
e
3
a
s
h
o
w
s
s
o
m
e
i
m
ag
e
s
a
m
p
les
o
f
u
n
iq
u
e
Viet
n
a
m
e
s
e
ch
ar
ac
ter
s
th
at
w
er
e
co
llected
b
y
o
u
r
t
w
o
m
e
th
o
d
s
an
d
Fi
g
u
r
e
3
b
s
h
o
w
s
s
o
m
e
s
a
m
p
le
s
o
f
2
3
L
atin
alp
h
ab
ets
t
h
at
w
er
e
p
ick
ed
f
r
o
m
NI
ST
Sp
ec
ial
Data
b
ase
1
9
.
Gen
er
ally
,
d
ata
in
tr
ain
i
n
g
s
et,
v
alid
atio
n
s
e
t,
an
d
test
s
et
s
h
o
u
ld
f
o
ll
o
w
t
h
e
s
a
m
e
p
r
o
b
ab
ilit
y
d
is
tr
ib
u
tio
n
[
2
6
]
.
T
h
er
ef
o
r
e,
w
h
e
n
b
u
ild
i
n
g
o
u
r
d
ata
s
et
f
o
r
m
ac
h
in
e
lear
n
i
n
g
o
r
d
ee
p
lear
n
in
g
s
y
s
te
m
,
we
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8938
I
n
t J
A
r
ti
f
I
n
tell
,
Vo
l.
9
,
No
.
2
,
J
u
n
e
20
20
:
27
6
–
28
3
280
h
av
e
p
er
f
o
r
m
ed
a
p
r
ep
r
o
ce
s
s
i
n
g
in
c
lu
d
i
n
g
zo
o
m
i
n
g
,
r
es
izin
g
,
an
d
b
i
n
ar
izin
g
o
n
2
3
L
at
in
alp
h
ab
ets
t
h
at
co
m
e
f
r
o
m
NI
ST
Sp
ec
ial
Da
tab
ase
1
9
to
m
a
k
e
all
i
m
ag
e
s
in
o
u
r
d
ataset
h
a
v
e
t
h
e
s
a
m
e
d
is
tr
ib
u
tio
n
.
(
a)
U
n
iq
u
e
Viet
n
a
m
e
s
e
ch
ar
ac
ter
s
(
b
)
L
atin
alp
h
ab
et
c
h
ar
ac
ter
s
Fig
u
r
e
3
.
E
x
a
m
p
les o
f
Viet
n
a
m
ese
c
h
ar
ac
ter
s
in
o
u
r
d
ataset
3
.
2
.
T
ra
ini
ng
pro
ce
s
s
T
h
e
tr
ain
in
g
p
r
o
ce
s
s
w
as
d
o
n
e
b
y
th
e
Go
o
g
le
C
o
llab
o
r
ato
r
y
s
er
v
er
.
Fo
r
b
o
th
tr
ain
in
g
p
h
ase
a
n
d
test
i
n
g
p
h
a
s
e,
w
e
ch
o
s
e
c
ate
g
o
r
ical
c
r
o
s
s
-
e
n
tr
o
p
y
f
u
n
c
tio
n
t
o
b
e
a
lo
s
s
f
u
n
ctio
n
f
o
r
o
u
r
C
NN
ar
ch
itectu
r
e.
I
t
is
a
s
o
f
t
m
ax
ac
tiv
a
tio
n
p
l
u
s
a
c
r
o
s
s
-
e
n
tr
o
p
y
lo
s
s
.
T
h
is
f
u
n
c
tio
n
ta
k
es
t
h
e
m
ea
s
u
r
e
m
e
n
t
s
o
f
h
o
w
clo
s
e
l
y
t
h
e
m
o
d
el
's
p
r
ed
ictio
n
s
m
atc
h
t
h
e
tar
g
et
clas
s
es
in
p
er
f
o
r
m
i
n
g
m
u
lti
-
clas
s
cla
s
s
i
f
icat
io
n
o
p
er
atio
n
.
T
h
e
p
r
o
p
o
s
ed
m
o
d
el
w
as
tr
ai
n
ed
b
y
u
s
i
n
g
m
in
i
-
b
atch
g
r
ad
ie
n
t
d
esce
n
t
as
o
p
tim
izatio
n
al
g
o
r
ith
m
.
I
t
is
a
v
ar
ia
n
t
o
f
g
r
ad
ien
t
d
escen
t
al
g
o
r
ith
m
.
T
h
e
m
i
n
i
-
b
atch
g
r
ad
ie
n
t
d
esce
n
t
i
s
c
h
o
s
en
b
ec
a
u
s
e
it
p
r
o
v
id
e
s
a
co
m
p
u
tat
io
n
all
y
m
o
r
e
ef
f
icien
t
p
r
o
ce
s
s
th
a
n
s
to
ch
a
s
t
ic
g
r
ad
ien
t
d
esce
n
t.
I
n
t
h
is
wo
r
k
,
m
in
i
-
b
atc
h
g
r
ad
ie
n
t
d
esc
en
t
w
as
co
n
f
i
g
u
r
ed
w
it
h
a
b
atch
s
ize
o
f
1
2
8
an
d
lear
n
in
g
r
ate
o
f
0
.
0
0
1
as
th
e
o
p
ti
m
izatio
n
alg
o
r
ith
m
.
An
eq
u
al
lear
n
in
g
r
ate
w
a
s
e
m
p
lo
y
ed
f
o
r
all
la
y
er
s
.
T
h
e
n
et
w
o
r
k
w
a
s
tr
ai
n
ed
f
o
r
7
0
,
0
0
0
s
tep
s
eq
u
i
v
alen
t
to
1
5
5
ep
o
ch
s
t
h
r
o
u
g
h
5
8
,
0
0
0
ex
a
m
p
le
s
o
f
2
9
ch
ar
ac
ter
s
in
th
e
tr
ain
i
n
g
s
et.
I
n
ad
d
itio
n
to
lo
s
s
m
etr
ic,
th
e
ac
c
u
r
ac
y
m
etr
ic
w
as
u
s
ed
to
ev
alu
a
te
tr
ain
i
n
g
p
r
o
ce
s
s
.
3
.
3
.
T
esting
r
esu
lt
s
T
h
e
ac
cu
r
ac
y
m
etr
ic
is
u
til
ize
d
to
ev
alu
ate
th
e
m
o
d
el
p
er
f
o
r
m
an
ce
o
n
o
u
r
Viet
n
a
m
e
s
e
h
a
n
d
w
r
i
tten
ch
ar
ac
ter
d
ataset.
T
h
e
ac
c
u
r
a
c
y
m
etr
ic
is
d
e
f
i
n
ed
as
th
e
r
atio
b
et
w
ee
n
t
h
e
n
u
m
b
er
o
f
co
r
r
ec
t
p
r
ed
icted
s
a
m
p
les
a
n
d
t
h
e
to
tal
s
a
m
p
le
s
.
T
h
e
p
r
o
p
o
s
ed
m
o
d
el
ac
h
ie
v
ed
an
ac
c
u
r
ac
y
9
7
.
2
%
o
n
te
s
t
s
e
t
(
to
p
-
1
te
s
t
s
et
er
r
o
r
r
ate
o
f
2
.
8
%)
af
ter
tr
ain
i
n
g
6
9
,
0
1
0
s
tep
s
.
Fig
u
r
e
4
s
h
o
w
s
th
e
ac
c
u
r
ac
y
g
r
ap
h
o
f
o
u
r
m
o
d
el
o
n
t
h
e
te
s
t
s
et
an
d
Fig
u
r
e
5
p
r
esen
ts
t
h
e
tr
ain
in
g
a
n
d
v
alid
atio
n
lo
s
s
.
Fig
u
r
e
4
.
T
h
e
ac
cu
r
ac
y
o
f
o
u
r
m
o
d
el
o
n
te
s
t set a
s
a
f
u
n
c
tio
n
o
f
iter
atio
n
s
t
e
p
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
A
r
ti
f
I
n
tell
I
SS
N:
2252
-
8938
V
ietn
a
mese
h
a
n
d
w
r
itten
ch
a
r
a
cter reco
g
n
itio
n
u
s
in
g
co
n
vo
lu
tio
n
a
l n
eu
r
a
l
n
etw
o
r
k
(
T
r
u
o
n
g
Qu
a
n
g
V
in
h
)
281
Fig
u
r
e
5
.
T
r
ain
in
g
(
─
)
an
d
v
al
id
atio
n
(
─
)
lo
s
s
o
f
o
u
r
m
o
d
el
as a
f
u
n
ctio
n
o
f
iter
atio
n
I
n
o
r
d
er
t
o
co
m
p
ar
e
o
u
r
alg
o
r
ith
m
w
it
h
o
th
er
m
o
d
els,
t
w
o
m
e
th
o
d
s
o
f
[
1
9
-
2
0
]
w
er
e
u
s
ed
f
o
r
r
ef
er
en
ce
s
.
T
h
e
p
r
o
p
o
s
ed
m
et
h
o
d
in
[
1
9
]
tr
ies
to
s
e
g
m
en
t
th
e
ac
ce
n
ted
c
h
ar
ac
ter
s
i
n
to
t
w
o
p
ar
ts
:
th
e
r
o
o
t
ch
ar
ac
ter
an
d
t
h
e
ac
ce
n
t.
T
h
en
,
th
e
SVM
m
o
d
el
i
s
ap
p
lied
t
o
class
i
f
y
t
h
e
Viet
n
a
m
ese
c
h
ar
ac
ter
s
.
I
n
s
o
m
e
ca
s
es,
t
h
is
m
eth
o
d
m
ee
ts
p
r
o
b
lem
in
w
h
ich
t
h
e
c
h
ar
ac
ter
is
n
o
t
w
r
itte
n
i
n
o
n
e
co
n
n
ec
ted
co
m
p
o
n
e
n
t
.
T
h
er
ef
o
r
e,
th
e
s
e
g
m
en
ta
tio
n
p
r
o
ce
s
s
ca
n
b
e
f
ai
led
.
T
h
e
m
et
h
o
d
o
f
th
e
a
u
t
h
o
r
s
i
n
[
2
0
]
u
s
e
s
co
n
to
u
r
a
n
d
p
r
o
j
ec
t
h
is
to
g
r
a
m
f
ea
tu
r
e
s
w
ith
S
VM
class
i
f
icatio
n
to
r
ec
o
g
n
ize
h
a
n
d
w
r
itte
n
ch
ar
ac
ter
s
.
Ho
w
ev
er
,
th
o
s
e
f
ea
t
u
r
es
ar
e
n
o
t
en
o
u
g
h
to
ch
ar
ac
ter
ize
Vi
etn
a
m
ese
h
an
d
w
r
itte
n
ch
ar
ac
t
er
s
w
i
th
u
n
co
n
s
tr
ai
n
ed
ac
ce
n
t
s
an
d
s
tr
o
k
e
s
.
Ou
r
C
NN
m
o
d
el
ca
n
s
o
lv
e
t
h
e
p
r
o
b
lem
o
f
n
at
u
r
al
a
n
d
u
n
c
o
n
s
tr
ain
ed
Viet
n
a
m
ese
ch
ar
a
cter
s
.
T
h
e
f
ea
tu
r
e
ex
tr
ac
tio
n
b
lo
ck
b
y
3
co
n
v
o
lu
tio
n
al
la
y
er
s
ca
n
lear
n
h
i
g
h
er
-
lev
el
f
ea
t
u
r
es,
a
n
d
t
h
en
t
h
e
f
e
atu
r
e
cla
s
s
i
f
icatio
n
b
lo
ck
w
h
ic
h
ar
e
b
u
ilt
b
y
2
f
u
l
l
y
co
n
n
ec
ted
la
y
er
s
w
it
h
d
r
o
p
o
u
t
tech
n
iq
u
e
s
ca
n
p
r
ev
e
n
t
o
v
er
f
i
tti
n
g
is
s
u
e
a
n
d
th
u
s
p
r
o
v
id
e
t
h
e
b
est
p
er
f
o
r
m
an
ce
o
f
h
a
n
d
w
r
iti
n
g
r
ec
o
g
n
i
ti
o
n
.
A
s
a
r
esu
lt,
t
h
e
co
m
p
ar
is
o
n
w
h
ic
h
is
d
escr
ib
ed
in
T
ab
le
1
s
h
o
w
s
t
h
at
t
h
e
p
r
o
p
o
s
ed
m
e
th
o
d
o
u
tp
er
f
o
r
m
s
t
w
o
o
th
er
m
et
h
o
d
s
.
T
ab
le
1
.
T
h
e
co
m
p
ar
is
o
n
b
et
wee
n
th
e
p
r
o
p
o
s
ed
m
et
h
o
d
an
d
o
th
er
s
D
e
C
a
o
T
r
a
n
[
1
9
]
T
h
a
c
h
T
r
a
n
V
a
n
e
t
a
l
.
[
2
0
]
T
h
e
p
r
o
p
o
se
d
M
o
d
e
l
A
c
c
e
n
t
e
d
c
h
a
r
a
c
t
e
r
se
g
me
n
t
a
t
i
o
n
+
S
V
M
c
l
a
ssi
f
i
c
a
t
i
o
n
C
o
n
t
o
u
r
a
n
d
p
r
o
j
e
c
t
h
i
s
t
o
g
r
a
m
-
b
a
se
d
f
e
a
t
u
r
e
e
x
t
r
a
c
t
i
o
n
+
S
V
M
C
N
N
(
3
c
o
n
v
o
l
u
t
i
o
n
a
l
l
a
y
e
r
s +
2
f
u
l
l
y
c
o
n
n
e
c
t
e
d
l
a
y
e
r
s)
D
a
t
a
se
t
U
N
I
P
EN
+
I
R
O
N
O
F
F
+
se
l
f
-
c
o
l
l
e
c
t
e
d
V
i
e
t
n
a
me
se
c
h
a
r
a
c
t
e
r
s
(
3
9
,
8
0
0
sam
p
l
e
s)
S
e
l
f
-
c
o
l
l
e
c
t
e
d
V
i
e
t
n
a
me
se
c
h
a
r
a
c
t
e
r
s
(
2
7
,
0
0
0
sam
p
l
e
s)
N
I
S
T
d
a
t
a
se
t
+
se
l
f
-
c
o
l
l
e
c
t
e
d
V
i
e
t
n
a
me
se
c
h
a
r
a
c
t
e
r
s (7
2
,
5
0
0
samp
l
e
s)
A
c
c
u
r
a
c
y
9
1
.
3
%
9
3
.
2
3
%
9
7
.
2
%
3
.
4
.
Dem
o
a
pp
lica
t
io
n f
o
r
Viet
na
m
e
s
e
ha
nd
w
rit
t
en
cha
ra
ct
er
re
co
g
nitio
n
I
n
o
r
d
er
to
d
em
o
n
s
tr
ate
a
r
e
al
ap
p
licatio
n
f
o
r
Viet
n
a
m
es
e
h
an
d
w
r
itte
n
c
h
ar
ac
ter
r
ec
o
g
n
i
tio
n
,
w
e
d
esig
n
ed
a
d
esk
to
p
s
o
f
t
w
ar
e
b
y
Qt
to
o
l
k
it
b
ased
o
n
o
u
r
p
r
etr
ain
ed
C
NN
m
o
d
el.
T
h
e
ap
p
licatio
n
ca
n
t
a
k
e
an
in
p
u
t
as
an
i
m
ag
e
o
f
a
V
ietn
a
m
ese
ch
ar
ac
ter
th
at
w
as
d
r
aw
n
d
ir
ec
tl
y
i
n
to
p
ai
n
ti
n
g
w
i
n
d
o
w
o
f
t
h
e
ap
p
o
r
s
elec
ted
f
r
o
m
an
ex
i
s
ti
n
g
i
m
a
g
e
an
d
t
h
e
n
p
r
o
d
u
ce
a
d
ig
ita
l
ch
ar
ac
ter
o
f
t
h
i
s
ch
ar
ac
ter
i
n
in
p
u
t
i
m
ag
e.
T
h
e
ap
p
licatio
n
h
as
t
w
o
m
o
d
es:
p
r
ed
ict
d
ir
e
ctly
i
m
ag
e
f
r
o
m
p
a
in
ti
n
g
w
i
n
d
o
w
(
o
n
li
n
e
m
o
d
e)
an
d
p
r
ed
ict
f
r
o
m
ex
ter
n
al
i
m
ag
e
(
o
f
f
li
n
e
m
o
d
e)
.
Fo
r
o
n
li
n
e
m
o
d
e,
th
e
ap
p
licatio
n
ca
p
t
u
r
es
p
ai
n
ti
n
g
w
i
n
d
o
w
a
n
d
m
ak
e
s
p
r
ed
ictio
n
b
ased
o
n
th
is
ca
p
tu
r
ed
i
m
ag
e.
User
s
also
ca
n
c
h
an
g
e
b
r
u
s
h
w
ei
g
h
t a
n
d
p
ain
ti
n
g
w
i
n
d
o
w
s
ize
i
n
th
i
s
m
o
d
e.
Fo
r
o
f
f
l
in
e
m
o
d
e,
u
s
e
r
s
ca
n
s
elec
t
a
n
i
m
a
g
e
t
h
at
c
o
n
tain
s
a
c
h
ar
ac
ter
n
ee
d
to
r
ec
o
g
n
ize
b
y
Op
en
b
u
tto
n
.
T
h
e
r
esu
lt
w
in
d
o
w
s
h
o
w
s
t
w
o
in
f
o
r
m
at
io
n
o
f
p
r
ed
ictio
n
:
to
p
5
s
co
r
e
o
f
p
r
e
d
ictio
n
in
d
etail
m
o
d
e
an
d
th
e
h
ig
h
es
t sco
r
e
as t
h
e
f
i
n
al
r
esu
lt
in
R
es
u
lt o
p
tio
n
.
Th
i
s
ap
p
licatio
n
ca
n
r
u
n
o
n
b
o
th
W
i
n
d
o
w
s
O
S a
n
d
L
i
n
u
x
OS.
Fig
u
r
e
6
s
h
o
w
s
t
h
e
UI
o
f
Vietn
a
m
ese
h
an
d
w
r
itte
n
ch
ar
ac
ter
r
ec
o
g
n
itio
n
ap
p
licatio
n
.
T
h
e
d
em
o
v
id
eo
o
f
th
e
p
r
o
p
o
s
ed
m
et
h
o
d
h
as b
ee
n
u
p
lo
ad
ed
to
Yo
u
T
u
b
e
at
h
ttp
s
://
y
o
u
tu
.
b
e/NY8
GI
q
I
h
6
6
0
.
s
t
e
p
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8938
I
n
t J
A
r
ti
f
I
n
tell
,
Vo
l.
9
,
No
.
2
,
J
u
n
e
20
20
:
27
6
–
28
3
282
Fig
u
r
e
6
.
Vietn
a
m
ese
h
an
d
w
r
i
tten
ch
ar
ac
ter
r
ec
o
g
n
itio
n
ap
p
l
icatio
n
4.
CO
NCLU
SI
O
N
I
n
th
i
s
p
ap
er
,
w
e
p
r
esen
ted
an
ef
f
icie
n
t
C
NN
ar
ch
i
tectu
r
e
f
o
r
Vietn
a
m
ese
h
an
d
w
r
it
te
n
ch
ar
ac
ter
r
ec
o
g
n
itio
n
.
Ou
r
C
NN
m
o
d
el
is
b
u
ilt
w
it
h
3
co
n
v
o
l
u
tio
n
al
la
y
er
s
a
n
d
2
f
u
l
l
y
co
n
n
ec
ted
la
y
er
s
.
A
d
r
o
p
o
u
t
tech
n
iq
u
e
i
s
co
m
b
i
n
ed
w
i
th
t
h
e
f
u
ll
y
co
n
n
ec
ted
la
y
er
s
to
p
r
ev
en
t
o
v
er
f
itti
n
g
p
h
e
n
o
m
e
n
o
n
.
T
h
e
e
x
p
er
i
m
en
ts
o
n
a
h
an
d
w
r
it
in
g
d
atab
ase
s
h
o
w
t
h
at
o
u
r
m
o
d
el
ca
n
ac
h
ie
v
e
ap
p
r
o
x
i
m
a
tel
y
9
7
%
ac
c
u
r
ac
y
.
Mo
r
eo
v
er
,
a
Qt
ap
p
licatio
n
f
o
r
Vietn
a
m
e
s
e
h
an
d
w
r
i
tten
c
h
ar
ac
ter
r
ec
o
g
n
it
io
n
w
a
s
d
esig
n
ed
to
d
em
o
n
s
tr
ate
th
e
p
r
o
p
o
s
ed
C
NN
ar
ch
i
tectu
r
e.
RE
F
E
R
E
NC
E
S
[1
]
A
.
&
.
M
.
S
.
&
.
R.
S
.
&
.
M
.
S
.
&
.
D.
S
.
P
riy
a
,
"
On
li
n
e
a
n
d
o
f
f
li
n
e
c
h
a
ra
c
ter
re
c
o
g
n
it
io
n
:
A
su
rv
e
y
”,
In
ter
n
a
ti
o
n
a
l
Co
n
fer
e
n
c
e
o
n
Co
mm
u
n
ica
t
io
n
a
n
d
S
ig
n
a
l
Pro
c
e
ss
in
g
(
ICCS
P)
,
2
0
1
6
.
[2
]
M
.
S
.
a
.
N.
S
a
h
u
,
“
A
S
u
rv
e
y
o
n
Ha
n
d
w
rit
ten
Ch
a
ra
c
ter
Re
c
o
g
n
it
io
n
(HCR)
T
e
c
h
n
iq
u
e
s
f
o
r
En
g
li
sh
A
lp
h
a
b
e
ts
”,
Ad
v
a
n
c
e
s i
n
Vi
si
o
n
C
o
mp
u
ti
n
g
:
A
n
In
ter
n
a
ti
o
n
a
l
J
o
u
rn
a
l
(
AV
C),
v
o
l.
3
,
M
a
rc
h
2
0
1
6
.
[3
]
Zo
u
h
a
ira
No
u
b
ig
h
,
M
o
n
ji
Kh
e
r
a
ll
a
h
,
“
A
su
rv
e
y
o
n
h
a
n
d
w
rit
in
g
re
c
o
g
n
it
io
n
b
a
se
d
o
n
t
h
e
traje
c
to
ry
re
c
o
v
e
r
y
tec
h
n
iq
u
e
”
,
2
0
1
7
1
st
I
n
ter
n
a
ti
o
n
a
l
W
o
rk
sh
o
p
o
n
Ara
b
ic S
c
rip
t
An
a
l
y
sis a
n
d
Rec
o
g
n
it
i
o
n
(
AS
AR
)
,
2
0
1
7
.
[4
]
S
.
Esp
a
ñ
a
-
Bo
q
u
e
ra
,
M
.
J.
Ca
stro
-
Bled
a
,
J.
G
o
rb
e
-
M
o
y
a
,
F
.
Za
m
o
ra
-
M
a
rti
n
e
z
,
“
Im
p
ro
v
in
g
O
ff
li
n
e
H
a
n
d
w
rit
ten
T
e
x
t
Re
c
o
g
n
it
io
n
w
it
h
H
y
b
rid
HMM
/
A
N
N
M
o
d
e
ls
”
,
IEE
E
T
ra
n
sa
c
ti
o
n
s
o
n
Pa
tt
e
rn
A
n
a
lys
is
a
n
d
M
a
c
h
in
e
In
telli
g
e
n
c
e
,
v
o
l.
3
3
,
n
o
.
4
,
p
p
.
7
6
7
-
7
7
9
,
2
0
1
1
.
[5
]
Ya
n
w
e
i
Wan
g
;
X
iao
q
in
g
Din
g
;
Ch
a
n
g
so
n
g
L
iu
,
“
T
o
p
ic
Lan
g
u
a
g
e
M
o
d
e
l
A
d
a
p
ti
o
n
f
o
r
Re
c
o
g
n
it
io
n
o
f
Ho
m
o
lo
g
o
u
s
Off
li
n
e
Ha
n
d
w
rit
ten
Ch
i
n
e
se
T
e
x
t
I
m
a
g
e
”
,
IEE
E
S
i
g
n
a
l
Pro
c
e
ss
in
g
L
e
tt
e
rs
,
v
o
l.
2
1
,
n
o
.
5
,
p
p
.
5
5
0
-
5
5
3
,
2
0
1
4
.
[6
]
Da
n
iel
Ke
y
s
e
rs,
T
h
o
m
a
s
De
se
l
a
e
rs,
He
n
r
y
A
.
Ro
w
le
y
,
L
i
-
L
u
n
W
a
n
g
,
V
icto
r
Ca
rb
u
n
e
,
“
M
u
lt
i
-
L
a
n
g
u
a
g
e
O
n
li
n
e
Ha
n
d
w
rit
in
g
Re
c
o
g
n
it
io
n
”
,
IEE
E
T
ra
n
sa
c
ti
o
n
s
o
n
Pa
tt
e
rn
A
n
a
lys
i
s
a
n
d
M
a
c
h
i
n
e
In
telli
g
e
n
c
e
,
v
o
l.
3
9
,
n
o
.
6
,
p
p
.
1
1
8
0
-
1
1
9
4
,
2
0
1
7
.
[7
]
Zec
h
e
n
g
X
ie
,
Zen
g
h
u
i
S
u
n
,
L
ian
w
e
n
Jin
,
Ha
o
Ni
,
Terr
y
Ly
o
n
s,
“
L
e
a
rn
in
g
S
p
a
ti
a
l
-
S
e
m
a
n
ti
c
Co
n
tex
t
w
it
h
F
u
ll
y
Co
n
v
o
l
u
ti
o
n
a
l
R
e
c
u
rre
n
t
Ne
tw
o
rk
f
o
r
On
li
n
e
Ha
n
d
w
rit
ten
Ch
i
n
e
se
Tex
t
Re
c
o
g
n
it
io
n
”
,
IEE
E
T
ra
n
sa
c
ti
o
n
s
o
n
Pa
tt
e
rn
A
n
a
lys
is a
n
d
M
a
c
h
i
n
e
In
t
e
ll
ig
e
n
c
e
,
v
o
l.
4
0
,
n
o
.
8
,
p
p
.
1
9
0
3
-
1
9
1
7
,
2
0
1
8
.
[8
]
S
.
A
tt
ig
e
ri,
“
Ne
u
ra
l
Ne
t
w
o
rk
b
a
se
d
Ha
n
d
w
rit
ten
Ch
a
ra
c
terRe
c
o
g
n
it
io
n
sy
ste
m
”
,
In
ter
n
a
ti
o
n
a
l
J
o
u
rn
a
l
o
f
En
g
i
n
e
e
rin
g
An
d
Co
m
p
u
ter
S
c
ien
c
e
,
v
o
l.
7
,
n
o
.
3
,
p
p
.
2
3
7
6
1
-
2
3
7
6
8
,
M
a
rc
h
2
0
1
8
.
[9
]
N.
T
.
M
.
K.
Ra
n
ia
M
A
AL
EJ,
“
On
li
n
e
A
ra
b
ic
Ha
n
d
w
rit
in
g
Re
c
o
g
n
it
io
n
w
it
h
Dro
p
o
u
t
a
p
p
li
e
d
in
De
e
p
Re
c
u
rre
n
t
Ne
u
ra
l
Ne
tw
o
rk
s
”,
1
2
th
IAP
R
W
o
rk
sh
o
p
o
n
D
o
c
u
me
n
t
A
n
a
lys
is S
y
ste
ms
,
2
0
1
6
.
[1
0
]
Ch
e
n
g
-
L
in
L
iu
,
F
e
i
Yin
,
Qiu
-
F
e
n
g
W
a
n
g
,
Da
-
Ha
n
W
a
n
g
,
“
IC
DA
R
2
0
1
1
C
h
i
n
e
se
Ha
n
d
w
rit
in
g
Re
c
o
g
n
it
io
n
Co
m
p
e
ti
ti
o
n
”
,
2
0
1
1
I
n
ter
n
a
ti
o
n
a
l
Co
n
fer
e
n
c
e
o
n
Do
c
u
me
n
t
A
n
a
lys
is a
n
d
Rec
o
g
n
it
io
n
,
2
0
1
1
.
[1
1
]
M
ich
a
e
l
M
u
rd
o
c
k
,
Ja
c
k
Re
e
s
e
,
S
h
a
w
n
Re
id
,
"
ICF
H
R2
0
1
6
C
o
m
p
e
ti
ti
o
n
o
n
L
o
c
a
l
A
tt
rib
u
te
De
tec
ti
o
n
f
o
r
Ha
n
d
w
rit
in
g
Re
c
o
g
n
it
io
n
"
,
2
0
1
6
1
5
th
In
ter
n
a
ti
o
n
a
l
Co
n
fer
e
n
c
e
o
n
Fro
n
ti
e
rs
in
Ha
n
d
writ
in
g
Rec
o
g
n
it
io
n
(
ICFHR)
,
2
0
1
6
.
[1
2
]
Bo
u
c
h
ra
El
q
a
c
imy
,
M
o
u
n
ir
A
it
k
e
rro
u
m
,
A
h
m
e
d
Ha
m
m
o
u
c
h
,
“
Ha
n
d
w
rit
ten
d
ig
it
re
c
o
g
n
it
io
n
b
a
s
e
d
o
n
DCT
f
e
a
tu
re
s a
n
d
S
V
M
c
las
sif
ier
”
,
2
0
1
4
S
e
c
o
n
d
W
o
rl
d
C
o
n
fer
e
n
c
e
o
n
C
o
mp
lex
S
y
ste
ms
(
W
CCS
)
,
2
0
1
4
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
A
r
ti
f
I
n
tell
I
SS
N:
2252
-
8938
V
ietn
a
mese
h
a
n
d
w
r
itten
ch
a
r
a
cter reco
g
n
itio
n
u
s
in
g
co
n
vo
lu
tio
n
a
l n
eu
r
a
l
n
etw
o
r
k
(
T
r
u
o
n
g
Qu
a
n
g
V
in
h
)
283
[1
3
]
S
e
ld
a
G
ü
n
e
y
,
Ce
y
h
u
n
Ça
k
a
r,
“
C
o
m
p
a
riso
n
o
f
e
x
p
e
c
tatio
n
m
a
x
i
m
iza
ti
o
n
a
n
d
Na
ïv
e
Ba
y
e
s
a
l
g
o
rit
h
m
s
in
c
h
a
ra
c
ter
re
c
o
g
n
it
io
n
”
,
2
0
1
6
2
4
th
S
ig
n
a
l
Pr
o
c
e
ss
in
g
a
n
d
Co
mm
u
n
ica
ti
o
n
Ap
p
li
c
a
ti
o
n
Co
n
fer
e
n
c
e
(
S
IU)
,
2
0
1
6
.
[1
4
]
M
.
Ra
jala
k
sh
m
i,
P
.
S
a
ra
n
y
a
,
P
.
S
h
a
n
m
u
g
a
v
a
d
iv
u
,
“
P
a
tt
e
rn
Re
c
o
g
n
it
io
n
-
Re
c
o
g
n
it
io
n
o
f
Ha
n
d
w
rit
ten
Do
c
u
m
e
n
t
Us
in
g
Co
n
v
o
lu
ti
o
n
a
l
Ne
u
ra
l
Ne
tw
o
rk
s
”
,
2
0
1
9
IEE
E
In
ter
n
a
ti
o
n
a
l
Co
n
fer
e
n
c
e
o
n
I
n
t
e
ll
ig
e
n
t
T
e
c
h
n
i
q
u
e
s in
Co
n
tro
l,
Op
ti
miza
ti
o
n
a
n
d
S
ig
n
a
l
Pro
c
e
ss
i
n
g
(
INCOS
)
,
2
0
1
9
.
[1
5
]
K
.
Du
tt
a
,
P
.
Krish
n
a
n
,
M
.
M
a
t
h
e
w
,
C.
V
.
Ja
wa
h
a
r,
“
I
m
p
ro
v
in
g
CNN
-
RNN
H
y
b
rid
Ne
t
w
o
rk
s
fo
r
Ha
n
d
w
rit
in
g
Re
c
o
g
n
it
io
n
”
,
2
0
1
8
1
6
th
I
n
ter
n
a
ti
o
n
a
l
Co
n
fer
e
n
c
e
o
n
Fro
n
ti
e
rs
in
Ha
n
d
writ
in
g
Rec
o
g
n
it
i
o
n
(
ICFHR)
,
2
0
1
9
.
[1
6
]
P
ri
tam
Dh
a
n
d
e
,
Re
e
n
a
Kh
a
ra
t,
“
Re
c
o
g
n
it
io
n
o
f
c
u
rsiv
e
En
g
li
sh
h
a
n
d
w
rit
ten
c
h
a
ra
c
ters
”
,
2
0
1
7
In
ter
n
a
t
io
n
a
l
Co
n
fer
e
n
c
e
o
n
T
re
n
d
s in
El
e
c
tro
n
ics
a
n
d
In
f
o
rm
a
ti
c
s (
ICEI)
,
2
0
1
7
.
[1
7
]
Qiu
-
F
e
n
g
W
a
n
g
,
F
e
i
Yin
,
Ch
e
n
g
-
L
in
L
iu
,
"
Ha
n
d
w
rit
ten
Ch
in
e
se
Tex
t
Re
c
o
g
n
it
io
n
b
y
In
teg
r
a
ti
n
g
M
u
lt
ip
l
e
Co
n
tex
ts"
,
IEE
E
T
r
a
n
sa
c
ti
o
n
s
o
n
Pa
tt
e
rn
A
n
a
lys
is a
n
d
M
a
c
h
i
n
e
In
t
e
ll
ig
e
n
c
e
,
v
o
l.
3
4
,
n
o
.
8
,
p
p
.
1
4
6
9
-
1
4
8
1
,
2
0
1
2
.
[1
8
]
X
ian
g
-
Do
n
g
Zh
o
u
,
Da
-
Ha
n
W
a
n
g
,
F
e
n
g
T
ian
,
Ch
e
n
g
-
L
in
L
iu
,
M
a
sa
k
i
Na
k
a
g
a
w
a
,
"
Ha
n
d
w
rit
ten
Ch
in
e
se
/Ja
p
a
n
e
se
T
e
x
t
Re
c
o
g
n
it
io
n
Us
in
g
S
e
m
i
-
M
a
rk
o
v
Co
n
d
it
i
o
n
a
l
Ra
n
d
o
m
F
ield
s"
,
IEE
E
T
ra
n
sa
c
ti
o
n
s
o
n
P
a
tt
e
r
n
An
a
lys
is
a
n
d
M
a
c
h
in
e
In
telli
g
e
n
c
e
,
v
o
l
.
3
5
,
n
o
.
1
0
,
p
p
.
2
4
1
3
-
2
4
2
6
,
2
0
1
3
.
[1
9
]
De
Ca
o
T
ra
n
,
“
A
n
e
ff
icie
n
t
m
e
th
o
d
f
o
r
o
n
li
n
e
Vie
tn
a
m
e
se
h
a
n
d
w
rit
ten
c
h
a
ra
c
ter
re
c
o
g
n
it
io
n
”
,
T
h
e
T
h
ird
S
y
mp
o
si
u
m o
n
In
fo
rm
a
ti
o
n
a
n
d
C
o
mm
u
n
ic
a
ti
o
n
T
e
c
h
n
o
l
o
g
y
–
S
o
IC
T
,
2
0
1
2
.
[2
0
]
T
h
a
c
h
T
ra
n
V
a
n
,
P
h
i
Ng
u
y
e
n
Hu
u
,
T
ra
n
g
Ho
a
n
g
,
“
Iso
late
d
v
i
e
tn
a
m
e
se
h
a
n
d
w
rit
in
g
re
c
o
g
n
it
io
n
e
m
b
e
d
d
e
d
sy
ste
m
a
p
p
li
e
d
c
o
m
b
in
e
d
f
e
a
tu
re
e
x
tr
a
c
ti
o
n
m
e
th
o
d
”
,
In
ter
n
a
t
io
n
a
l
Co
n
fer
e
n
c
e
o
n
Ad
v
a
n
c
e
d
T
e
c
h
n
o
l
o
g
ies
fo
r
Co
mm
u
n
ica
ti
o
n
s (
AT
C)
,
2
0
1
5
.
[2
1
]
H.
T
.
Ng
u
y
e
n
,
C.
T
.
Ng
u
y
e
n
,
P
.
T
.
Ba
o
,
a
n
d
M
.
Na
k
a
g
a
w
a
,
“
P
re
p
a
ra
ti
o
n
o
f
a
n
Un
c
o
n
stra
in
e
d
V
iet
n
a
m
e
se
On
li
n
e
Ha
n
d
w
rit
in
g
Da
tab
a
se
a
n
d
Re
c
o
g
n
it
io
n
Ex
p
e
rim
e
n
ts
b
y
Re
c
u
rre
n
t
Ne
u
ra
l
Ne
tw
o
rk
s
”
,
1
5
th
In
ter
n
a
t
io
n
a
l
Co
n
fer
e
n
c
e
o
n
Fro
n
ti
e
rs
in
Ha
n
d
writi
n
g
Rec
o
g
n
it
i
o
n
(
ICFHR)
,
2
0
1
6
.
[2
2
]
A
n
h
Du
c
L
e
,
Hu
n
g
T
u
a
n
Ng
u
y
e
n
,
M
a
sa
k
i
Na
k
a
g
a
wa
,
“
Re
c
o
g
n
iz
in
g
Un
c
o
n
stra
in
e
d
Vie
t
n
a
m
e
se
H
a
n
d
w
rit
in
g
By
A
tt
e
n
ti
o
n
Ba
se
d
En
c
o
d
e
r
De
c
o
d
e
r
M
o
d
e
l
”
,
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
Co
m
p
u
t
in
g
a
n
d
A
p
p
l
ica
ti
o
n
s
(
ACOM
P)
,
2
0
1
8
.
[2
3
]
W
.
&
.
W
.
Z.
Ra
w
a
t,
"
De
e
p
Co
n
v
o
lu
ti
o
n
a
l
Ne
u
ra
l
Ne
tw
o
rk
s
f
o
r
Im
a
g
e
Clas
si
f
ica
ti
o
n
:
A
Co
m
p
re
h
e
n
siv
e
Re
v
ie
w”
,
J
o
u
rn
a
l
of
Ne
u
ra
l
Co
m
p
u
t
a
ti
o
n
,
v
o
l.
29
,
iss.9
,
p
p
.
2
3
5
2
-
2
4
4
9
,
S
e
p
.
2
0
1
7
.
[2
4
]
C.
G
.
S
.
L
a
w
re
n
c
e
,
"
Ov
e
r
f
it
ti
n
g
a
n
d
n
e
u
ra
l
n
e
tw
o
rk
s:
c
o
n
ju
g
a
te
g
r
a
d
ien
t
a
n
d
b
a
c
k
p
ro
p
a
g
a
ti
o
n
”,
IEE
E
-
INNS
-
ENN
S
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
IJ
CNN
,
2
0
0
0
.
[2
5
]
G
.
H.
A
.
K.
I.
S
.
R
.
S
.
Niti
sh
S
ri
v
a
sta
v
a
,
"
Dro
p
o
u
t:
A
S
im
p
le
W
a
y
to
P
re
v
e
n
t
Ne
u
ra
l
Ne
tw
o
rk
s
f
ro
m
o
v
e
r
f
it
ti
n
g
"
,
J
o
u
rn
a
l
o
f
M
a
c
h
i
n
e
L
e
a
rn
in
g
Res
e
a
rc
h
,
v
o
l.
1
5
,
2
0
1
4
.
[2
6
]
J.
Qu
iñ
o
n
e
ro
-
Ca
n
d
e
la,
M
.
S
u
g
iy
a
m
a
,
A
.
S
c
h
w
a
i
g
h
o
f
e
r,
N.
D.
Law
r
e
n
c
e
,
“
D
a
tas
e
t
S
h
if
t
in
M
a
c
h
in
e
L
e
a
rn
in
g
”
,
Ne
u
ra
l
In
fo
rm
a
t
io
n
Pro
c
e
ss
in
g
se
rie
s
,
T
h
e
M
IT
P
re
ss
,
IS
BN 9
7
8
-
0
-
2
6
2
-
1
7
0
0
5
-
5
,
2
0
0
8
.
B
I
O
G
RAP
H
I
E
S
O
F
AUTH
O
RS
Dr.
T
ru
o
n
g
Qu
a
n
g
V
in
h
re
c
e
iv
e
d
th
e
B.
E
.
d
e
g
re
e
f
ro
m
Ho
Ch
i
M
i
n
h
Cit
y
Un
iv
e
rsit
y
o
f
T
e
c
h
n
o
lo
g
y
,
V
ietn
a
m
,
in
1
9
9
9
,
th
e
M
.
E.
d
e
g
re
e
in
Co
m
p
u
ter
S
c
ien
c
e
f
ro
m
th
e
As
ian
In
stit
u
te
o
f
T
e
c
h
n
o
lo
g
ies
,
Ba
n
g
k
o
k
,
T
h
a
il
a
n
d
,
i
n
2
0
0
3
;
a
n
d
P
h
.
D.
d
e
g
re
e
in
Co
m
p
u
ter
E
n
g
in
e
e
rin
g
a
t
Ch
o
n
n
a
m
Na
ti
o
n
a
l
Un
iv
e
rsit
y
,
Ko
re
a
,
in
2
0
1
0
.
Cu
rre
n
lt
y
,
h
e
is
a
lec
tu
re
r
in
De
p
a
rtme
n
t
o
f
El
e
c
tro
n
ics
,
F
a
c
u
lt
y
o
f
El
e
c
tri
c
a
l
a
n
d
El
e
c
tro
n
ics
En
g
in
e
e
rin
g
,
HCM
Cit
y
Un
iv
e
rsit
y
o
f
T
e
c
h
n
o
lo
g
y
,
V
ietn
a
m
Na
ti
o
n
a
l
Un
iv
e
rsit
y
–
Ho
Ch
i
M
in
h
,
V
ietn
a
m
.
His
h
o
m
e
p
a
g
e
:
h
tt
p
:
//
ww
w
4
.
h
c
m
u
t.
e
d
u
.
v
n
/~
tq
v
in
h
/
M
r.
L
e
Ho
a
i
Du
y
re
c
e
iv
e
d
th
e
B.
E.
d
e
g
re
e
in
El
e
c
tro
n
ics
f
ro
m
Ho
Ch
i
M
i
n
h
Cit
y
Un
iv
e
rsit
y
o
f
T
e
c
h
n
o
lo
g
y
,
V
ietn
a
m
in
2
0
1
8
.
No
w
h
e
w
o
rk
s
a
s
a
n
A
rti
f
icia
l
I
n
telli
g
e
n
c
e
En
g
in
e
e
r
a
t
Em
a
g
e
De
v
e
lo
p
m
e
n
t
Co
.
L
td
,
V
iet
n
a
m
.
M
r.
Ng
u
y
e
n
T
h
a
n
h
Nh
a
n
re
c
e
iv
e
d
th
e
B
.
E.
d
e
g
re
e
in
El
e
c
tro
n
ics
f
r
o
m
Ho
Ch
i
M
i
n
h
Cit
y
Un
iv
e
rsit
y
o
f
Tec
h
n
o
lo
g
y
,
V
ietn
a
m
in
2
0
1
8
.
C
u
rre
n
tl
y
,
h
e
w
o
rk
s a
s
a
n
V
isio
n
E
n
g
in
e
e
r
a
t
Em
a
g
e
De
v
e
lo
p
m
e
n
t
Co
.
L
td
,
V
ie
tn
a
m
.
Evaluation Warning : The document was created with Spire.PDF for Python.