I
n
d
on
e
s
ian
Jou
r
n
al
o
f
E
lec
t
r
ica
l
E
n
gin
e
e
r
in
g
a
n
d
Com
p
u
t
e
r
S
c
ience
Vo
l
.
3
8
,
N
o
.
2
,
M
a
y
20
2
5
,
pp.
1
367
~
1
375
I
S
S
N:
2
502
-
4
7
52
,
DO
I
:
10
.
11591/i
j
e
e
cs
.v
3
8
.
i
2
.
pp
1
367
-
1
375
1367
Jou
r
n
al
h
o
m
e
page
:
ht
tp:
//
ij
e
e
cs
.
iaes
c
or
e
.
c
om
D
is
c
r
e
t
e
w
a
ve
le
t
t
r
an
sf
o
r
m
a
n
d
c
on
v
ol
u
t
io
n
al
n
e
u
r
al
n
e
t
w
or
k
b
ase
d
h
a
n
d
w
r
itt
e
n
S
an
s
k
r
it
c
h
ar
ac
t
e
r
r
e
c
ogn
ition
S
h
r
ad
d
h
a
V.
S
h
e
l
k
e
,
Din
e
s
h
M
.
Chan
d
wadk
ar
,
S
u
n
it
a
P
.
Ugal
e
,
Rup
al
i
V.
Chot
h
e
D
e
pa
r
tm
e
nt
of
E
le
c
t
r
o
ni
c
s
a
nd
T
e
l
e
c
o
mm
uni
c
a
ti
o
n E
ngi
n
e
e
r
in
g, K
.
K
. W
a
gh I
ns
ti
tu
te
of
E
ngi
n
e
e
r
in
g
E
duc
a
ti
o
n a
nd R
e
s
e
a
r
c
h
,
S
a
v
it
r
ib
a
i
P
hul
e
P
un
e
U
ni
v
e
r
s
it
y
, P
un
e
, I
ndi
a
Ar
t
ic
l
e
I
n
f
o
AB
S
T
RA
CT
A
r
ti
c
le
h
is
tor
y
:
R
e
c
e
i
ve
d
J
u
l
18
,
202
4
R
e
vi
s
e
d
A
ug
28
,
202
4
A
c
c
e
pt
e
d
N
o
v
24
,
202
4
San
s
k
ri
t
i
s
o
n
e
o
f
t
h
e
an
ci
e
n
t
l
an
g
u
a
g
e
s
fro
m
w
h
i
c
h
t
h
e
m
a
j
o
r
i
t
y
o
f
p
r
e
s
en
t
I
n
d
i
a
n
l
an
g
u
a
g
e
s
are
d
e
v
el
o
p
e
d
.
A
l
t
h
o
u
g
h
t
h
e
n
at
i
o
n
al
m
i
s
s
i
o
n
fo
r
m
an
u
s
c
r
i
p
t
s
(N
MM)
i
s
d
i
g
i
t
i
z
i
n
g
h
a
n
d
w
ri
t
t
en
Sa
n
s
k
ri
t
m
an
u
s
c
r
i
p
t
s
,
t
h
e
r
e
ar
e
s
t
i
l
l
a
l
o
t
o
f
p
ap
e
rs
t
h
at
n
ee
d
t
o
b
e
d
i
g
i
t
i
z
ed
.
R
ec
o
g
n
i
t
i
o
n
o
f
h
a
n
d
w
ri
t
t
e
n
s
c
ri
p
t
i
s
a
c
h
a
l
l
en
g
i
n
g
t
as
k
d
u
e
t
o
i
n
d
i
v
i
d
u
al
d
i
ff
e
r
e
n
ce
s
i
n
w
ri
t
i
n
g
s
t
y
l
e
s
a
n
d
h
o
w
t
h
o
s
e
v
ari
at
i
o
n
s
a
l
t
e
r
o
v
e
r
t
i
me.
T
h
e
Sa
n
s
k
ri
t
l
a
n
g
u
ag
e
i
s
w
r
i
t
t
e
n
i
n
D
ev
an
a
g
ari
s
c
ri
p
t
.
A
n
o
v
e
l
ap
p
ro
ach
u
s
i
n
g
d
i
s
c
r
e
t
e
w
a
v
e
l
e
t
t
ran
s
fo
r
m
(D
W
T
)
an
d
c
o
n
v
o
l
u
t
i
o
n
al
n
at
u
ral
n
e
t
w
o
r
k
(CN
N
)
i
s
p
ro
p
o
s
e
d
i
n
t
h
i
s
p
ap
e
r.
D
ev
an
a
g
ari
h
an
d
w
r
i
t
t
e
n
ch
ara
c
t
e
r
d
at
as
e
t
w
h
i
c
h
i
n
c
l
u
d
e
s
2
0
0
0
h
an
d
w
ri
t
t
en
i
m
ag
e
s
o
f
3
6
c
l
as
s
e
s
(2
0
0
0
*
3
6
=
7
2
0
0
0
)
i
s
u
s
e
d
i
n
t
h
i
s
re
s
e
ar
ch
.
F
i
n
e
-
t
u
n
e
d
G
o
o
g
L
eN
e
t
m
o
d
el
i
m
p
l
eme
n
t
ed
h
e
re
g
a
v
e
o
p
t
i
m
u
m
v
al
u
e
s
o
f
e
p
o
ch
s
an
d
l
e
ar
n
i
n
g
rat
e
o
f
1
5
an
d
0
.
0
1
re
s
p
ec
t
i
v
el
y
.
Cl
as
s
i
fi
c
at
i
o
n
acc
u
ra
cy
o
b
t
ai
n
e
d
b
y
p
ro
p
o
s
e
d
D
W
T
–
C
N
N
mo
d
el
i
s
9
8
.
9
7
%
w
i
t
h
a
l
o
s
s
o
f
0
.
0
9
8
.
Fi
n
e
-
t
u
n
ed
G
o
o
g
L
eN
e
t
m
o
d
e
l
ach
i
ev
e
s
9
9
.
6
8
%
acc
u
ra
cy
w
i
t
h
a
0
.
0
6
3
5
l
o
s
s
.
Re
s
u
l
t
s
o
b
t
ai
n
ed
ar
e
a
l
s
o
co
m
p
ared
w
i
t
h
e
x
i
s
t
i
n
g
ap
p
ro
ach
e
s
an
d
fo
u
n
d
s
u
p
e
r
i
o
r
.
K
e
y
w
o
r
d
s
:
C
o
n
v
o
l
ut
i
o
n
a
l
n
a
t
ur
a
l
n
e
t
wo
r
k
G
oo
g
L
e
N
et
D
i
s
c
r
e
t
e
wa
v
e
l
e
t
t
r
a
n
s
f
o
r
m
Na
t
i
o
n
a
l
m
i
s
s
i
o
n
f
o
r
m
a
n
u
s
c
r
i
pt
s
S
a
n
s
kr
i
t
m
a
n
u
s
c
r
i
pt
s
Th
i
s
i
s
a
n
o
p
en
a
c
ces
s
a
r
t
i
c
l
e
u
n
d
e
r
t
h
e
CC
B
Y
-
SA
l
i
cen
s
e.
C
or
r
e
s
pon
din
g
A
u
th
or
:
S
h
r
a
dd
h
a
V.
S
h
e
l
k
e
De
pa
r
t
m
e
n
t
o
f
E
l
e
c
t
r
o
ni
c
s
a
n
d
T
e
l
e
c
o
m
m
u
ni
c
a
t
i
o
n
E
n
g
i
ne
e
r
i
ng,
K
.
K
.
W
a
g
h
I
n
s
t
i
t
ut
e
o
f
E
n
g
i
ne
e
r
i
n
g
E
duc
a
t
i
o
n
a
n
d
R
e
s
e
a
r
c
h
,
S
a
vi
t
r
i
ba
i
P
h
u
l
e
P
un
e
U
ni
ve
r
s
i
t
y
P
un
e
,
I
n
d
i
a
E
m
a
i
l
:
s
vs
h
e
l
ke
@
kkwa
g
h
.
e
du.
i
n
1.
I
NT
RODU
C
T
I
ON
An
c
i
e
n
t
S
a
ns
kr
i
t
t
e
x
t
s
i
n
c
l
ude
a
n
a
b
u
n
d
a
n
c
e
o
f
kn
o
w
l
e
dge
o
n
H
i
ndu
m
y
t
h
o
l
o
g
y
,
c
u
l
t
ur
e
,
s
c
i
e
n
c
e
,
m
a
t
h
e
m
a
t
i
c
s
a
n
d
I
n
d
i
a
n
hi
s
t
o
r
y
.
A
r
o
un
d
300
m
il
li
o
n
i
n
d
i
v
i
dua
l
s
u
s
e
S
a
n
s
kr
i
t
s
c
r
i
pt
gl
o
b
a
ll
y
[
1]
.
T
h
e
r
e
i
s
a
l
o
t
o
f
r
e
s
e
a
r
c
h
be
i
n
g
do
ne
i
n
t
h
e
f
i
e
l
d
o
f
r
e
c
o
g
ni
z
i
ng
h
a
n
dwr
i
t
t
e
n
t
e
x
t
[
2]
.
T
h
e
r
e
a
r
e
n
u
m
e
r
o
us
o
f
f
i
c
i
a
l
l
a
n
gua
ge
s
i
n
I
n
d
i
a
t
h
a
t
a
r
e
wr
i
t
t
e
n
i
n
De
va
n
a
ga
r
i
s
c
r
i
pt
,
s
uc
h
a
s
H
i
n
d
i
,
M
a
r
a
t
hi
,
S
i
n
d
hi
,
Ne
p
a
l
i
,
S
a
ns
kr
i
t
,
a
n
d
K
o
n
ka
ni
.
A
w
i
de
r
a
n
g
e
o
f
a
l
go
r
i
t
hm
s
ha
v
e
b
e
e
n
c
r
e
a
t
e
d
f
o
r
r
e
c
o
gni
t
i
o
n
o
f
s
c
r
i
pt
s
wr
i
t
t
e
n
i
n
s
e
v
e
r
a
l
I
n
d
i
a
n
l
a
n
gua
ge
s
,
i
n
c
l
ud
i
ng
H
i
n
d
i
,
B
a
n
g
l
a
,
T
e
l
ugu,
a
nd
Gur
m
uk
hi
.
B
ut
s
t
i
ll
t
h
e
r
e
h
a
s
b
e
e
n
n
o
t
e
n
o
u
gh
s
t
ud
y
c
o
n
duc
t
e
d
o
n
de
v
e
l
o
p
i
n
g
a
n
e
f
f
i
c
i
e
n
t
S
a
n
s
kr
i
t
s
c
r
i
pt
r
e
c
o
gni
t
i
o
n
s
y
s
t
e
m
,
h
e
nc
e
e
m
p
h
a
s
i
z
i
ng
t
h
e
n
e
e
d
to
c
o
n
s
t
r
uc
t
o
n
e
[
3]
.
A
h
u
m
a
n
c
a
n
e
a
s
il
y
u
n
de
r
s
t
a
n
d
c
o
n
t
e
n
t
s
wr
i
t
t
e
n
o
n
pa
pe
r
,
b
ut
a
m
a
c
hi
ne
f
i
nds
i
t
c
h
a
ll
e
n
g
i
ng
t
o
r
e
c
o
gni
z
e
h
a
n
dwr
i
t
t
e
n
t
e
x
t
due
to
v
e
r
i
t
y
o
f
wr
i
t
i
n
g
s
t
y
l
e
s
.
F
o
r
h
a
n
dwr
i
t
t
e
n
t
e
x
t
i
de
n
t
i
f
i
c
a
t
i
o
n
,
t
h
e
r
e
s
e
a
r
c
h
e
r
s
h
a
v
e
us
e
d
a
v
a
r
i
e
t
y
o
f
m
a
c
hi
ne
l
e
a
r
ni
ng,
de
e
p
l
e
a
r
ni
ng
t
e
c
hni
qu
e
s
[
4]
.
T
h
e
i
ni
t
i
a
l
a
n
d
c
r
uc
i
a
l
s
t
e
p
i
n
a
n
o
p
t
i
c
a
l
c
h
a
r
a
c
t
e
r
r
e
c
o
gn
i
t
i
o
n
s
y
s
t
e
m
i
s
f
e
a
t
ur
e
e
x
t
r
a
c
t
i
o
n
f
r
o
m
t
h
e
i
n
put
i
m
a
ge
f
o
l
l
o
w
i
ng
pr
e
pr
o
c
e
s
s
i
ng
a
n
d
n
o
i
s
e
r
e
duc
t
i
o
n
.
T
h
e
c
l
a
s
s
if
i
e
r
i
s
t
h
e
n
t
r
a
i
n
e
d
by
f
e
e
d
i
ng
t
h
e
e
x
t
r
a
c
t
e
d
f
e
a
t
ur
e
s
to
s
e
v
e
r
a
l
c
l
a
s
s
if
i
e
r
s
.
I
n
t
h
e
t
e
s
t
i
n
g
a
n
d
v
a
li
da
t
i
o
n
ph
a
s
e
,
t
h
e
o
u
t
pu
t
o
f
t
h
e
c
l
a
s
s
if
i
e
r
i
s
ve
r
i
f
i
e
d
by
t
e
s
t
i
n
g
i
t
s
pe
r
f
o
r
m
a
n
c
e
o
n
n
e
w
da
t
a
[
5]
.
T
h
e
i
ni
t
i
a
l
s
t
a
ge
s
o
f
a
c
o
n
v
o
l
ut
i
o
na
l
ne
ur
a
l
n
e
t
wo
r
k
(
C
NN
)
b
a
s
e
d
m
e
t
h
o
d
i
nv
o
l
v
e
a
pp
lyi
ng
v
a
r
i
o
us
e
dge
de
t
e
c
t
i
o
n
f
il
t
e
r
s
t
o
i
n
put
i
m
a
ge
s
i
n
o
r
de
r
to
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2
5
0
2
-
4
7
52
In
do
n
e
s
i
a
n
J
E
l
e
c
E
n
g
&
C
o
m
p
S
c
i
,
Vo
l
.
3
8
,
N
o
.
2
,
M
a
y
20
2
5
:
1
367
-
1
375
1368
e
x
t
r
a
c
t
f
e
a
t
ur
e
m
a
ps
.
T
h
e
i
nput
i
m
a
ge
i
s
c
o
nv
o
l
v
e
d
w
i
t
h
f
il
t
e
r
s
t
o
e
x
t
r
a
c
t
f
e
a
t
ur
e
s
,
whi
c
h
a
r
e
t
h
e
n
a
pp
l
i
e
d
t
o
t
h
e
f
u
ll
y
c
o
nn
e
c
t
e
d
l
a
y
e
r
a
f
t
e
r
poo
l
i
ng
o
pe
r
a
t
i
o
n
.
T
o
i
m
pr
o
v
e
a
m
o
de
l
's
pe
r
f
o
r
m
a
n
c
e
,
s
e
v
e
r
a
l
c
o
nv
o
l
ut
i
o
n
a
l
,
poo
l
i
ng,
a
n
d
f
u
ll
y
c
o
nn
e
c
t
e
d
l
a
y
e
r
s
a
r
e
us
e
d.
Af
t
e
r
a
pp
l
yi
ng
a
c
t
i
va
t
i
o
n
f
u
n
c
t
i
o
n
s
f
o
r
m
u
l
t
i
c
l
a
s
s
c
l
a
s
s
if
i
c
a
t
i
o
n
,
s
uc
h
a
s
r
e
c
t
i
f
i
e
d
l
i
ne
a
r
uni
t
s
(
R
e
L
U)
a
n
d
S
o
f
t
M
a
x
,
c
l
a
s
s
i
f
i
e
d
o
ut
pu
t
i
s
ge
n
e
r
a
t
e
d
[
6]
.
S
a
n
s
kr
i
t
s
c
r
i
pt
i
s
d
i
s
t
i
n
gu
i
s
he
d
by
t
h
e
h
e
a
de
r
l
i
ne
,
o
r
s
hi
r
o
l
e
k
h
a
,
t
h
a
t
a
ppe
a
r
s
a
b
o
v
e
e
a
c
h
c
h
a
r
a
c
t
e
r
.
C
o
m
pa
r
e
d
t
o
ot
h
e
r
s
c
r
i
pt
s
l
i
ke
E
n
g
li
s
h
,
m
o
r
e
c
o
n
s
o
n
a
n
t
s
a
n
d
v
o
we
l
s
a
n
d
d
i
f
f
e
r
e
n
t
c
o
m
bi
n
a
t
i
o
n
s
o
f
c
o
m
po
un
d
l
e
t
t
e
r
s
a
n
d
c
o
nj
u
nc
t
c
o
n
s
o
n
a
n
t
s
m
a
ke
s
S
a
n
s
kr
i
t
s
c
r
i
pt
r
e
c
o
gni
t
i
o
n
a
c
h
a
ll
e
n
g
i
ng
t
a
s
k.
T
h
e
r
e
a
r
e
16
v
o
we
l
s
a
n
d
39
c
o
n
s
o
n
a
n
t
s
i
n
S
a
n
s
kr
i
t
wr
i
t
i
n
g.
C
o
n
s
o
n
a
n
t
s
c
a
n
s
o
m
e
t
i
m
e
s
h
a
v
e
i
de
n
t
i
c
a
l
s
h
a
pe
s
w
i
t
h
j
u
s
t
mi
n
o
r
v
a
r
i
a
t
i
o
n
s
,
s
uc
h
a
s
dot
s
,
s
l
a
n
t
i
n
g
l
i
ne
s
,
a
n
d
c
ur
v
e
s
[
7]
a
s
s
h
o
wn
i
n
F
i
gur
e
1.
F
i
gur
e
1.
C
o
n
s
o
n
a
n
t
s
w
i
t
h
s
im
il
a
r
s
h
a
pe
w
i
t
h
m
i
n
o
r
v
a
r
i
a
t
i
o
n
s
On
e
o
f
t
h
e
c
h
a
ll
e
n
g
e
i
n
r
e
c
o
gni
t
i
o
n
i
s
d
i
f
f
e
r
e
n
t
h
a
n
dwr
i
t
i
n
g
s
t
y
l
e
o
f
d
i
f
f
e
r
e
n
t
pe
o
pl
e
w
hi
c
h
c
ha
n
ge
s
o
v
e
r
a
ge
.
O
t
h
e
r
i
m
po
r
t
a
n
t
pa
r
a
m
e
t
e
r
s
to
b
e
c
o
n
s
i
d
e
r
e
d
whi
l
e
de
ve
l
o
p
i
n
g
r
e
c
o
gni
t
i
o
n
s
y
s
t
e
m
a
r
e
c
ur
s
i
ve
s
t
y
le
o
f
wr
i
t
i
n
g,
s
l
o
pe
,
t
hi
c
k
ne
s
s
,
a
v
e
r
a
ge
s
i
z
e
o
f
l
e
t
ter
s
,
s
pa
c
i
n
g
b
e
t
we
e
n
l
e
t
t
e
r
s
a
n
d
i
n
put
i
m
a
ge
qu
a
l
i
t
y
[
8]
.
A
s
y
s
t
e
m
to
r
e
c
o
gni
z
e
h
a
n
dwr
i
t
t
e
n
De
v
a
na
ga
r
i
c
ha
r
a
c
t
e
r
s
i
s
pr
o
p
o
s
e
d
by
De
o
r
e
a
n
d
P
r
a
vi
n
[
9]
wh
e
r
e
a
u
t
h
o
r
s
h
a
v
e
im
p
l
e
m
e
n
t
e
d
t
w
o
s
t
a
ge
a
ppr
o
a
c
h
us
i
n
g
V
GG
16
to
r
e
c
o
gni
z
e
D
e
v
a
n
a
ga
r
i
h
a
n
dwr
i
t
t
e
n
c
ha
r
a
c
t
e
r
s
.
R
M
S
pr
o
p
a
n
d
A
da
m
o
pt
i
mi
z
e
r
s
a
r
e
us
e
d
whi
c
h
r
e
s
u
l
t
e
d
i
n
go
o
d
a
c
c
ur
a
c
y
w
i
t
h
R
M
S
pr
o
p
wi
t
h
f
a
s
t
e
r
c
o
n
v
e
r
ge
n
c
e
t
h
a
n
A
da
m
.
I
n
or
de
r
to
p
r
e
v
e
n
t
o
v
e
r
f
i
t
t
i
n
g,
a
ppr
o
a
c
h
e
s
us
e
d
h
e
r
e
a
r
e
da
t
a
a
ugm
e
n
t
a
t
i
o
n
a
n
d
i
nc
l
us
i
o
n
o
f
0.
5
dr
o
p
o
u
t.
R
e
s
u
l
t
s
a
r
e
o
b
t
a
i
n
e
d
by
c
h
a
n
g
i
ng
n
u
m
be
r
o
f
e
po
c
h
s
.
M
o
de
l
i
s
a
l
s
o
t
r
a
i
n
e
d
b
y
u
s
i
ng
VG
G
19
whi
c
h
g
i
ve
s
l
e
s
s
a
c
c
ur
a
c
y
a
n
d
n
e
e
ds
m
o
r
e
s
t
or
a
ge
s
pa
c
e
t
h
a
n
t
h
a
n
VG
G16,
whi
c
h
a
r
c
hi
ve
s
in
c
l
a
s
s
if
i
c
a
t
i
o
n
a
c
c
ur
a
c
y
o
f
96.
55%
.
An
a
ppr
o
a
c
h
u
s
i
ng
C
NN
t
o
r
e
c
o
gni
z
e
h
a
n
d
wr
i
tt
e
n
De
v
a
n
a
ga
r
i
c
h
a
r
a
c
t
e
r
s
i
s
im
p
l
e
m
e
n
t
e
d
by
B
h
a
r
dwa
j
[
10]
.
M
o
de
l
A
a
n
d
M
o
de
l
B
d
i
f
f
e
r
i
n
n
u
m
be
r
o
f
c
o
nv
o
l
ut
i
o
n
a
l
l
a
y
e
r
s
,
f
u
ll
y
c
o
nn
e
c
t
e
d
l
a
y
e
r
s
.
I
n
b
o
t
h
m
o
de
l
s
R
e
L
U
a
c
t
i
v
a
t
i
o
n
f
u
nc
t
i
o
n
a
n
d
l
o
c
a
l
r
e
s
po
n
s
e
n
o
r
m
a
li
z
a
t
i
o
n
i
s
us
e
d.
W
i
t
h
a
m
o
m
e
n
t
um
o
f
0.
9,
s
to
c
h
a
s
t
i
c
gr
a
d
i
e
n
t
de
s
c
e
n
t
(
S
GD
)
i
s
us
e
d
to
a
c
c
o
m
p
li
s
h
t
h
e
b
a
c
kpr
o
pa
ga
t
i
o
n
o
n
f
e
e
d
-
f
o
r
wa
r
d
n
e
t
s
.
M
o
de
l
A
i
s
a
bl
e
to
gi
v
e
c
l
a
s
s
if
i
c
a
t
i
o
n
r
a
t
e
o
f
0.
98471.
A
ut
h
o
r
s
f
o
un
d
i
m
pr
o
v
e
m
e
n
t
i
n
c
l
a
s
s
i
f
i
c
a
t
i
o
n
a
c
c
ur
a
c
y
up
to
0.
981326
by
i
nc
l
u
s
i
o
n
o
f
dr
o
p
o
u
t
l
a
y
e
r
a
n
d
e
x
t
e
n
d
i
ng
da
t
a
s
e
t
.
An
e
j
a
a
n
d
An
e
j
a
[
11]
h
a
v
e
pr
o
p
o
s
e
d
a
n
a
ppr
o
a
c
h
us
i
ng
tr
a
n
s
f
e
r
l
e
a
r
ni
n
g
to
c
l
a
s
s
i
f
y
De
v
a
n
a
ga
r
i
c
h
a
r
a
c
t
e
r
s
.
F
i
n
e
t
uni
n
g
a
n
d
C
o
nv
Ne
t
f
o
r
f
i
x
e
d
f
e
a
t
ur
e
e
x
t
r
a
c
to
r
a
r
e
t
e
c
hni
que
s
us
e
d.
I
n
c
a
s
e
o
f
f
i
xe
d
f
e
a
t
ur
e
e
x
t
r
a
c
to
r
a
b
r
o
b
t
a
i
n
e
d
f
r
o
m
we
i
g
h
t
s
o
f
c
o
n
v
o
l
ut
i
o
n
a
l
l
a
y
e
r
s
w
hi
c
h
a
r
e
f
r
o
z
e
n
a
n
d
we
i
g
h
t
v
e
c
t
or
s
o
f
f
u
l
ly
c
o
n
n
e
c
t
e
d
l
a
y
e
r
s
a
r
e
upda
t
e
d.
R
e
s
u
l
t
s
s
h
o
ws
t
h
a
t
i
nc
e
pt
i
o
n
V3
m
o
de
l
r
e
s
u
l
t
e
d
i
n
hi
g
he
s
t
a
c
c
ur
a
c
y
o
f
99%
.
Vgg11
h
a
v
e
hi
g
h
e
r
c
o
m
put
a
t
i
o
n
a
l
c
o
s
t
t
h
a
n
i
nc
e
p
t
i
o
n
V3
w
i
t
h
99%
a
c
c
ur
a
c
y
w
i
t
h
a
v
e
r
a
ge
t
i
m
e
o
f
5.
7
m
in
pe
r
e
po
c
h
.
C
l
a
s
s
if
i
a
c
t
i
o
n
a
c
c
ur
a
c
y
o
f
De
n
s
e
Ne
t
20
1
a
n
d
De
ns
e
Ne
t
121
i
s
89%
a
n
d
90%
r
e
s
pe
c
t
i
v
e
ly
.
L
o
m
t
e
a
n
d
Do
y
e
[
12]
h
a
v
e
pr
o
p
o
s
e
d
a
n
e
f
f
e
c
t
i
v
e
m
e
t
h
o
d
f
o
r
c
a
l
li
ng
up
De
v
a
na
ga
r
i
t
e
x
t
a
n
d
c
a
l
li
gr
a
p
hy
r
e
c
o
gni
t
i
o
n
t
h
a
t
c
o
m
bi
ne
s
w
a
v
e
l
e
t
,
c
o
n
to
u
r
,
a
n
d
s
uppo
r
t
v
e
c
to
r
m
a
c
hi
ne
u
s
i
n
g
c
o
r
r
e
l
a
t
i
o
n
f
u
n
c
t
i
o
n
s
I
C
F
a
n
d
AC
F
r
e
s
u
l
t
e
d
a
c
c
ur
a
c
y
o
f
98.
98
%
.
Vi
n
a
e
t
al
.
[
13]
i
m
p
l
e
m
e
n
t
e
d
m
o
d
i
f
i
e
d
C
NN
a
l
o
n
g
w
i
t
h
Al
e
xne
t
to
r
e
c
o
gni
z
e
h
a
n
dwr
i
t
t
e
n
Ve
d
i
c
S
a
n
s
kr
i
t
t
e
x
t
.
T
h
r
e
e
C
NN
m
o
de
l
s
w
i
t
h
3,
8
a
n
d
1
2
C
ON
V
l
a
y
e
r
s
r
e
s
pe
c
t
i
v
e
ly
a
r
e
pr
o
p
o
s
e
d.
I
t
i
s
o
b
s
e
r
ve
d
th
a
t
c
o
m
bi
n
i
ng
t
h
e
A
da
m
o
p
t
i
mi
z
e
r
w
i
t
h
R
e
L
U
a
ll
o
ws
f
o
r
f
a
s
t
e
r
pr
o
c
e
s
s
i
n
g
by
i
nc
r
e
a
s
i
ng
t
h
e
tr
a
i
ni
ng
r
a
t
e
a
n
d
pr
e
v
e
n
t
i
n
g
o
v
e
r
f
i
t
t
i
n
g
o
f
t
h
e
m
o
de
l
.
M
o
de
l
M
2
f
o
un
d
b
e
s
t
wi
t
h
a
c
c
ur
a
c
y
o
f
97.
42%
f
o
r
Ve
d
i
c
S
a
n
s
kr
i
t
.
A
ut
h
o
r
s
h
a
v
e
a
l
s
o
pr
o
p
o
s
e
d
a
n
o
v
e
l
s
y
s
t
e
m
to
r
e
c
o
gni
z
e
ge
s
t
ur
e
-
b
a
s
e
d
t
e
x
t
wr
i
t
t
e
n
i
n
De
v
a
n
a
ga
r
i
.
T
wo
C
NN
m
o
de
l
s
w
i
t
h
2
a
n
d
3
l
a
y
e
r
s
a
r
e
us
e
d
t
o
r
e
c
o
g
ni
z
e
h
a
n
d
t
i
p
ge
s
t
ur
e
c
h
a
r
a
c
t
e
r
wi
t
h
a
c
c
ur
a
c
y
o
f
97.
6
%
a
n
d
98.
4%
r
e
s
pe
c
t
i
v
e
ly
[
14]
.
M
a
n
o
c
h
a
a
n
d
T
e
wa
r
i
[
15]
pr
o
p
o
s
e
d
a
s
y
s
t
e
m
us
in
g
C
NN
f
o
r
f
e
a
t
ur
e
e
x
t
r
a
c
t
i
o
n
a
n
d
s
uppo
r
t
v
e
c
t
or
m
a
c
hi
ne
,
r
a
n
do
m
f
o
r
e
s
t
,
de
c
i
s
i
o
n
t
r
e
e
,
m
u
l
t
i
l
a
y
e
r
pe
r
c
e
ptr
o
n
s
a
n
d
K
-
Ne
a
r
e
s
t
n
e
i
g
hb
o
r
a
n
d
e
x
t
r
e
m
e
b
o
o
s
t
gr
a
d
i
e
n
t
f
o
r
c
l
a
s
s
i
f
i
c
a
t
i
o
n
o
f
DH
C
D
da
t
a
s
e
t
.
T
r
a
i
ni
ng
a
n
d
t
e
s
t
i
n
g
da
t
a
s
e
t
i
s
s
p
li
t
t
e
d
i
n
c
a
s
e
s
o
f
70:
30,
75:25
a
n
d
80:20.
H
i
g
h
e
s
t
c
l
a
s
s
i
f
i
c
a
t
i
o
n
a
c
c
ur
a
c
y
i
s
o
b
t
a
i
n
e
d
a
s
99%
by
u
s
i
ng
C
NN
a
n
d
S
VM
-
R
B
F
.
I
n
pr
e
vi
o
us
wo
r
k
s
e
v
e
r
a
l
f
e
a
t
ur
e
e
x
tr
a
c
t
i
o
n
m
e
t
h
o
ds
a
n
d
c
l
a
s
s
i
f
i
e
r
s
a
r
e
us
e
d
i
n
a
m
a
c
hi
ne
l
e
a
r
ni
ng
-
b
a
s
e
d
s
t
r
a
t
e
gy
f
o
r
r
e
c
o
gni
z
i
n
g
h
a
n
dwr
i
t
t
e
n
De
v
a
na
ga
r
i
c
ha
r
a
c
t
e
r
s
t
h
r
o
ugh
a
n
a
l
y
z
i
ng
t
h
e
i
r
pe
r
f
o
r
m
a
n
c
e
[
16]
.
I
n
t
hi
s
wo
r
k
a
n
o
v
e
l
a
ppr
o
a
c
h
us
i
ng
d
i
s
c
r
e
t
e
wa
ve
l
e
t
t
r
a
n
s
f
o
r
m
(
DW
T
)
a
n
d
C
NN
i
s
e
m
p
l
o
y
e
d
to
r
e
c
o
gni
z
e
h
a
n
dwr
i
t
t
e
n
c
h
a
r
a
c
t
e
r
s
.
T
h
e
f
i
ne
-
t
un
e
d
Go
o
gL
e
Ne
t
m
o
de
l
u
n
de
r
go
e
s
m
u
l
t
i
p
l
e
t
r
a
i
ni
ng
c
y
c
l
e
s
t
o
d
e
t
e
r
m
i
ne
t
h
e
o
p
t
i
m
u
m
e
po
c
h
a
n
d
l
e
a
r
ni
ng
r
a
t
e
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
do
n
e
s
i
a
n
J
E
l
e
c
E
n
g
&
C
o
m
p
S
c
i
I
S
S
N:
2
5
0
2
-
4
7
52
Dis
c
r
e
te
w
av
e
let
tr
ans
f
or
m
and
c
onv
olut
ional
ne
ur
al
ne
t
w
or
k
bas
e
d
…
(
Shr
addha
V
.
She
lke
)
1369
2.
M
E
T
HO
D
A
r
c
hi
t
e
c
t
ur
e
o
f
pr
o
p
o
s
e
d
s
y
s
t
e
m
i
s
s
h
o
w
n
i
n
F
i
g
ur
e
2.
S
a
n
kr
i
t
i
s
wr
i
t
t
e
n
us
i
ng
De
v
a
n
a
ga
r
i
s
c
r
i
pt
,
da
t
a
s
e
t
i
n
c
l
ud
i
ng
92,
000
h
a
n
d
wr
i
t
t
e
n
i
m
a
ge
s
pr
e
pa
r
e
d
by
A
c
ha
r
y
a
e
t
al.
[
17
]
a
r
e
us
e
d
i
n
t
hi
s
e
xpe
r
i
m
e
n
t
.
Da
t
a
s
e
t
i
n
c
l
ude
s
2
,
000
h
a
n
dwr
i
t
t
e
n
i
m
a
g
e
s
o
f
46
d
i
f
f
e
r
e
n
t
c
l
a
s
s
e
s
o
f
36
c
o
n
s
o
n
a
n
t
s
a
n
d
10
n
u
m
e
r
a
l
s
.
72,
000
im
a
ge
s
o
f
36
c
l
a
s
s
e
s
a
r
e
us
e
d
i
n
t
hi
s
e
x
p
e
r
i
m
e
n
t
.
T
h
e
i
m
a
ge
da
t
a
s
tor
e
o
f
a
l
l
72,
000
i
m
a
ge
s
i
s
c
r
e
a
t
e
d.
T
h
e
pr
e
pr
o
c
e
s
s
i
ng
s
t
e
p
i
nv
o
l
v
e
s
n
o
i
s
e
r
e
m
o
v
a
l
a
n
d
r
e
s
i
z
i
n
g
o
f
t
h
e
i
m
a
g
e
.
F
ur
t
h
e
r
di
s
c
r
e
t
e
wa
v
e
l
e
t
tr
a
n
s
f
or
m
i
s
a
p
pl
i
e
d
f
or
f
e
a
tu
r
e
e
x
t
r
a
c
ti
on
w
h
i
c
h
i
s
th
e
n
a
p
pl
i
e
d
to
C
N
N
m
od
e
l
to
g
e
n
e
r
a
te
c
l
a
s
s
i
f
i
c
a
t
i
on
r
e
s
ul
ts
.
F
i
gur
e
2.
DW
T
C
NN
b
a
s
e
d
s
y
s
t
e
m
a
r
c
hi
t
e
c
t
ur
e
2.
1.
Dis
c
r
e
t
e
wave
l
e
t
t
r
an
s
f
o
r
m
A
t
e
c
hni
que
f
o
r
e
x
tr
a
c
t
i
n
g
f
e
a
t
ur
e
s
t
h
a
t
di
vi
d
e
s
a
n
i
n
put
s
i
g
n
a
l
i
n
t
o
s
e
v
e
r
a
l
s
u
b
b
a
n
ds
by
a
pp
lyi
ng
a
d
i
s
c
r
e
t
e
wa
v
e
l
e
t
tr
a
n
s
f
o
r
m
to
t
h
e
i
n
put
i
m
a
ge
.
Ana
l
y
z
i
ng
s
i
g
n
a
l
da
t
a
i
n
b
o
t
h
t
h
e
t
i
m
e
a
n
d
f
r
e
qu
e
n
c
y
d
o
m
a
i
ns
i
s
o
n
e
o
f
t
h
e
b
e
s
t
f
e
a
t
ur
e
s
o
f
D
W
T
[
18]
.
F
i
gur
e
3
de
s
c
r
i
b
e
w
a
v
e
l
e
t
de
c
o
m
po
s
i
t
i
o
n
o
f
t
h
e
i
n
put
c
h
a
r
a
c
t
e
r
im
a
ge
i
n
f
o
ur
s
ub
b
a
n
ds
.
c
A
i
s
a
ppr
o
xi
m
a
t
i
o
n
c
oe
f
f
i
c
i
e
n
t
whi
c
h
i
s
l
o
w
f
r
e
que
n
c
y
s
u
b
b
a
n
d.
An
d
th
r
e
e
hi
g
h
f
r
e
qu
e
n
c
y
s
u
b
ba
n
ds
i
n
c
l
ude
h
o
r
i
z
o
n
t
a
l
de
t
a
i
l
c
o
e
f
f
i
c
i
e
n
t
(
c
H)
,
v
e
r
t
i
c
a
l
de
t
a
i
l
c
o
e
f
f
i
c
i
e
n
t
(
c
V)
,
a
n
d
d
i
a
go
n
a
l
de
t
a
i
l
c
o
e
f
f
i
c
i
e
n
t
(
c
D)
[
19]
.
Af
t
e
r
pr
e
pr
o
c
e
s
s
i
n
g
a
n
d
i
n
put
i
m
a
g
e
(
I
)
i
s
a
pp
l
i
e
d
to
di
s
c
r
e
t
e
wa
v
e
l
e
t
tr
a
n
s
f
o
r
m
ge
n
e
r
a
t
e
s
I
=
{c
A
,
c
V,
c
H,
c
D}.
T
h
e
de
t
a
i
l
c
o
e
f
f
i
c
i
e
n
t
s
o
b
t
a
i
ne
d
f
r
o
m
a
pp
l
yi
ng
DW
T
to
a
n
i
n
put
i
m
a
ge
o
f
d
i
m
e
n
s
i
o
ns
128x
128
a
r
e
o
f
s
i
z
e
64
×
64,
a
n
d
t
h
e
r
e
s
u
l
t
i
n
g
a
ppr
o
xi
m
a
t
i
o
n
c
o
e
f
f
i
c
i
e
n
t
c
A
un
de
r
go
e
s
C
NN
pr
o
c
e
s
s
i
n
g.
F
i
gur
e
3
.
F
e
a
tur
e
e
x
t
r
a
c
t
i
o
n
pr
oc
e
s
s
by
DW
T
2.
2.
CN
N
a
r
c
h
it
e
c
t
u
r
e
T
r
a
di
t
i
o
n
a
ll
y
,
a
n
a
r
t
i
f
i
c
i
a
l
n
e
ur
a
l
ne
t
wor
k
(
A
NN
)
i
s
m
a
de
up
o
f
m
a
ny
d
e
n
s
e
l
a
y
e
r
s
o
f
n
e
ur
o
n
s
,
o
f
t
e
n
r
e
f
e
r
r
e
d
to
a
s
f
u
ll
y
c
o
nn
e
c
t
e
d
l
a
y
e
r
s
.
Us
i
ng
t
h
e
s
e
f
u
ll
y
c
o
nn
e
c
t
e
d
l
a
y
e
r
s
r
e
s
u
l
t
s
i
n
a
n
e
x
t
r
e
m
e
ly
l
a
r
ge
tot
a
l
n
u
m
be
r
o
f
pa
r
a
m
e
t
e
r
s
[
6]
.
Us
i
n
g
m
u
l
t
i
p
l
e
c
o
nv
o
l
ut
i
o
n
a
l
a
n
d
po
o
l
i
ng
l
a
y
e
r
s
i
n
C
NN
a
ll
o
w
s
us
t
o
r
e
duc
e
t
h
e
n
u
m
be
r
o
f
pa
r
a
m
e
t
e
r
s
b
e
c
a
us
e
t
h
e
s
e
l
a
y
e
r
s
o
nl
y
r
e
qu
i
r
e
a
s
m
a
ll
n
u
m
be
r
o
f
pa
r
a
m
e
t
e
r
s
.
T
h
e
f
e
a
t
ur
e
pa
tt
e
r
n
s
o
f
t
h
e
i
n
put
da
t
a
s
e
t
a
r
e
s
to
r
e
d
i
n
f
e
a
t
ur
e
m
a
p
s
t
h
a
t
a
r
e
ge
n
e
r
a
t
e
d
a
s
a
r
e
s
u
l
t
o
f
c
o
nv
o
l
ut
i
o
n
a
l
a
n
d
poo
l
i
ng
l
a
y
e
r
s
[
5]
.
F
i
gur
e
4
s
h
o
ws
t
h
e
s
t
r
u
c
t
ur
e
o
f
t
h
e
pr
o
p
o
s
e
d
C
NN
m
o
de
l
.
C
o
n
v
o
l
ut
i
o
n
a
l
l
a
y
e
r
i
n
C
NN
c
o
n
s
i
s
t
s
o
f
a
ppl
yi
n
g
c
o
n
v
o
l
ut
i
o
n
a
l
o
pe
r
a
t
i
o
n
us
i
ng
a
f
i
l
t
e
r
a
n
d
t
h
e
n
a
dd
i
n
g
bi
a
s
t
o
o
u
t
pu
t
a
n
d
pa
s
s
i
n
g
i
t
to
n
o
nl
i
ne
a
r
a
c
t
i
va
t
i
o
n
f
u
n
c
t
i
o
n
l
i
ke
R
e
L
U.
F
o
r
m
u
l
t
i
p
l
e
f
il
t
e
r
s
bi
a
s
a
dde
d
w
i
ll
b
e
d
i
f
f
e
r
e
n
t
to
ge
n
e
r
a
t
e
m
u
l
t
i
p
l
e
o
ut
pu
ts
o
f
c
o
n
v
o
l
ut
i
o
n
a
l
l
a
y
e
r
s
.
(
)
=
3
(1
)
(
)
=
3
(2
)
(
)
=
(
−
+
1
)
(
−
+
1
)
(3
)
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2
5
0
2
-
4
7
52
In
do
n
e
s
i
a
n
J
E
l
e
c
E
n
g
&
C
o
m
p
S
c
i
,
Vo
l
.
3
8
,
N
o
.
2
,
M
a
y
20
2
5
:
1
367
-
1
375
1370
I
n
(
1
)
-
(
3
)
i
n
d
i
c
a
t
e
s
di
m
e
ns
i
o
n
s
o
f
i
n
put
i
m
a
g
e
,
f
il
t
e
r
a
n
d
f
e
a
t
ur
e
m
a
p
o
b
t
a
i
n
e
d
a
f
t
e
r
c
o
n
v
o
l
ut
i
o
n
a
l
o
pe
r
a
t
i
o
n
.
W
h
e
n
a
n
im
a
ge
o
f
s
i
z
e
n
×
n
×
3
i
s
c
o
nvo
l
v
e
d
w
i
t
h
f
il
t
e
r
o
f
s
i
z
e
f
×
f
×
3
×
u,
wh
e
r
e
3
i
nd
i
c
a
t
e
s
R
GB
im
a
ge
a
n
d
u
i
n
d
i
c
a
t
e
s
n
u
m
be
r
o
f
f
il
t
e
r
s
,
t
h
e
n
s
i
z
e
o
f
f
e
a
t
ur
e
m
a
p
o
b
t
a
i
n
e
d
a
f
t
e
r
c
o
n
v
o
l
ut
i
o
n
a
l
l
a
y
e
r
i
s
(
n
-
f
+
1)
×
(n
-
f
+
1)
×
u.
P
oo
l
i
n
g
o
pe
r
a
t
i
o
n
i
s
pe
r
f
o
r
m
e
d
a
f
t
e
r
c
o
n
v
o
l
ut
i
o
n
a
l
o
pe
r
a
t
i
o
n
.
F
un
c
t
i
o
n
o
f
poo
l
i
ng
l
a
y
e
r
i
s
to
r
e
duc
e
di
m
e
ns
i
o
n
s
o
f
i
nput
i
m
a
ge
by
pr
e
s
e
r
vi
ng
f
e
a
t
ur
e
s
i
n
i
t
.
I
n
m
a
x
poo
l
i
ng
o
pe
r
a
t
i
o
n
m
a
xim
u
m
v
a
l
u
e
f
e
a
t
ur
e
i
s
s
e
l
e
c
t
e
d
f
r
o
m
g
i
ve
n
f
il
t
e
r
w
i
n
do
w,
a
n
d
t
h
e
n
m
o
v
e
d
w
i
t
h
a
s
t
r
i
d
e
o
f
s
o
v
e
r
t
h
e
c
o
nv
o
l
v
e
d
im
a
ge
.
I
f
f
i
l
t
e
r
s
i
z
e
o
f
m
a
x
po
o
l
i
n
g
o
pe
r
a
t
i
o
n
i
s
m
x
m
a
n
d
s
t
r
i
de
o
f
2
t
h
e
n
f
e
a
t
ur
e
m
a
p
o
b
t
a
i
n
e
d
a
t
t
h
e
o
u
t
pu
t
w
i
ll
ha
v
e
s
i
z
e
o
f
(
m
/2)
×
(
m
/2)
.
T
h
e
r
e
a
r
e
m
u
l
t
i
p
l
e
s
uc
h
l
a
y
e
r
s
c
o
n
ne
c
t
e
d
o
n
e
a
f
t
e
r
t
h
e
ot
h
e
r
[
5
]
.
P
r
o
p
o
s
e
d
C
NN
m
o
de
l
c
o
n
s
i
s
t
s
o
f
8
c
o
n
v
o
l
ut
i
o
na
l
l
a
y
e
r
s
w
i
t
h
R
e
L
U
a
c
t
i
va
t
i
o
n
f
u
n
c
t
i
o
n
a
n
d
f
o
ur
m
a
x
po
o
l
i
ng
l
a
y
e
r
s
.
A
c
t
i
v
a
t
i
o
n
s
i
z
e
o
b
t
a
i
n
e
d
a
t
f
i
r
s
t
a
n
d
s
e
c
o
n
d
c
o
n
v
o
l
ut
i
o
n
a
l
l
a
y
e
r
s
i
s
(
62,
62,
32)
.
M
a
x
po
o
l
i
ng
o
pe
r
a
t
i
o
n
i
s
do
n
e
w
i
t
h
s
t
r
i
de
o
f
2
a
n
d
pa
dd
i
n
g
i
s
s
e
t
a
s
v
a
l
i
d.
T
hi
r
d
a
n
d
f
o
ur
t
h
c
o
n
v
o
l
ut
i
o
n
a
l
l
a
y
e
r
s
h
a
v
e
o
ut
pu
t
o
f
(
19,
19
,
32
)
wh
e
r
e
a
s
f
i
f
t
h
a
n
d
s
i
x
t
h
c
o
nv
o
l
ut
i
o
n
a
l
l
a
y
e
r
w
i
t
h
(
8,
8,
64
)
a
n
d
s
e
v
e
n
t
h
a
n
d
e
i
g
h
t
h
c
o
nv
o
l
ut
i
o
n
a
l
l
a
y
e
r
h
a
v
e
s
i
z
e
o
f
(
3,
3,
64)
.
B
e
f
o
r
e
a
ppl
yi
ng
f
u
ll
y
c
o
nn
e
c
t
e
d
l
a
y
e
r
o
u
t
pu
t
o
f
l
a
s
t
C
ON
V
l
a
y
e
r
i
s
f
l
a
t
t
e
n
i
n
o
n
e
d
i
m
e
n
s
i
o
n
a
l
v
e
c
t
o
r
.
M
u
l
t
i
p
l
e
f
u
ll
y
c
o
nne
c
t
e
d
l
a
y
e
r
s
c
a
n
b
e
a
dde
d
b
ut
a
dd
i
n
g
m
u
l
t
i
p
l
e
f
u
ll
y
c
o
nn
e
c
t
e
d
l
a
y
e
r
s
i
nc
r
e
a
s
e
s
a
m
o
u
n
t
o
f
pa
r
a
m
e
t
e
r
s
to
b
e
de
a
l
t
w
i
t
h
[
4]
.
A
c
t
i
v
a
t
i
o
n
f
u
n
c
t
i
o
n
s
li
ke
t
h
e
S
o
f
t
M
a
x
f
o
r
m
u
l
t
i
c
l
a
s
s
c
l
a
s
s
i
f
i
c
a
t
i
o
n
a
n
d
t
h
e
s
i
g
m
o
i
d
f
o
r
bi
n
a
r
y
c
l
a
s
s
if
i
c
a
t
i
o
n
a
r
e
i
n
c
l
ude
d
i
n
t
he
l
a
s
t
l
a
y
e
r
.
S
o
f
t
M
a
x
a
c
t
i
v
a
t
i
o
n
f
u
n
c
t
i
o
n
i
s
us
e
d
i
n
pr
o
p
o
s
e
d
m
o
de
l
.
S
o
f
t
M
a
x
ac
ti
v
a
t
i
on
f
un
c
t
i
on
i
s
u
s
e
d
f
or
m
ul
t
i
c
l
a
s
s
c
l
a
s
s
i
f
i
c
a
t
i
on
wh
e
r
e
ou
tp
u
t
i
s
p
r
ob
a
b
i
l
i
t
y
th
a
t
i
n
p
u
t
b
e
l
on
g
s
to
th
a
t
c
l
a
s
s
.
F
i
gur
e
4.
S
tr
uc
t
u
r
e
o
f
C
NN
m
o
de
l
M
a
t
h
e
m
a
t
i
c
a
l
m
o
d
e
l
i
ng
o
f
S
o
f
t
M
a
x
f
u
n
c
t
i
o
n
i
s
g
i
v
e
n
i
n
(
4
)
,
W
h
e
r
e
P
(
y
=
i
)
i
s
o
ut
pu
t
pr
o
b
a
bil
i
t
y
.
Z
i
i
s
r
a
w
c
l
a
s
s
s
c
o
r
e
.
P
(y
=
i
)
=
∑
=
1
(4
)
Nu
m
b
e
r
o
f
n
e
ur
o
n
s
i
n
l
a
s
t
f
u
ll
y
c
o
nn
e
c
t
e
d
l
a
y
e
r
s
h
o
u
l
d
b
e
e
qua
l
t
o
n
u
m
b
e
r
o
f
c
l
a
s
s
e
s
whi
c
h
g
i
v
e
s
pr
e
d
i
c
t
e
d
v
a
l
ue
s
o
f
c
a
t
e
go
r
i
e
s
.
A
s
t
h
e
r
e
a
r
e
36
d
i
s
ti
nc
t
o
u
t
pu
t
s
l
a
s
t
l
a
y
e
r
c
o
n
t
a
i
n
s
36
n
e
ur
o
n
s
.
C
o
s
t
f
un
c
t
i
o
n
is
c
a
l
c
u
l
a
t
e
d
f
r
o
m
pr
e
d
i
c
t
e
d
v
a
l
ue
s
w
hi
c
h
s
i
g
nif
y
a
m
o
un
t
o
f
e
r
r
o
r
t
h
a
t
m
o
de
l
i
s
ge
tt
i
n
g
w
hil
e
m
a
k
i
ng
pr
e
d
i
c
t
i
o
n
s
.
F
o
r
bi
n
a
r
y
c
l
a
s
s
i
f
i
c
a
t
i
o
n
bi
na
r
y
c
r
o
s
s
e
n
t
r
o
py
c
o
s
t
f
unc
t
i
o
n
c
a
n
b
e
us
e
d
a
n
d
f
o
r
m
u
l
t
i
c
l
a
s
s
c
l
a
s
s
if
i
c
a
t
i
o
n
c
a
t
e
gor
i
c
a
l
c
r
o
s
s
e
n
t
r
o
py
c
o
s
t
f
u
n
c
t
i
o
n
c
a
n
b
e
us
e
d.
T
o
e
n
h
a
n
c
e
t
h
e
m
o
de
l
's
a
c
c
ur
a
c
y
,
i
t
i
s
n
e
c
e
s
s
a
r
y
t
o
m
i
n
im
i
z
e
t
h
e
c
o
s
t
f
un
c
t
i
o
n
t
h
r
o
ugh
t
r
a
i
ni
ng
w
i
t
h
b
a
c
kpr
o
pa
ga
t
i
o
n
a
l
go
r
i
t
hm
s
,
s
uc
h
a
s
s
t
o
c
h
a
s
t
i
c
gr
a
d
i
e
n
t
de
s
c
e
n
t
(
S
GD
)
or
a
da
p
t
i
v
e
m
o
m
e
n
t
e
s
t
i
mat
i
o
n
(
A
da
m
)
[
2]
.
2.
3.
GoogL
e
Ne
t
ar
c
h
it
e
c
t
u
r
e
P
r
e
tr
a
i
n
e
d
Go
o
gL
e
Ne
t
m
o
de
l
a
l
s
o
c
a
l
l
e
d
a
s
I
n
c
e
pt
i
o
n
-
v
1
i
s
a
l
s
o
us
e
d
to
i
n
t
hi
s
pa
pe
r
.
I
n
t
h
e
t
r
a
n
s
f
e
r
l
e
a
r
ni
ng
a
ppr
o
a
c
h
,
a
pr
e
t
r
a
i
n
e
d
c
l
a
s
s
if
i
c
a
t
i
o
n
m
o
de
l
i
s
us
e
d
r
a
t
h
e
r
o
f
c
r
e
a
t
i
n
g
a
n
e
w
de
e
p
l
e
a
r
ni
n
g
m
o
de
l
w
i
t
h
ma
ny
c
o
nv
o
l
ut
i
o
na
l
a
n
d
f
u
ll
y
c
o
nn
e
c
t
e
d
l
a
y
e
r
s
[
20]
,
[
21]
.
F
or
tr
a
n
s
f
e
r
l
e
a
r
ni
ng,
t
h
e
M
A
T
L
A
B
de
e
p
n
e
t
wor
k
de
s
i
g
n
e
r
too
l
b
o
x
i
s
us
e
d
a
n
d
t
h
e
G
oo
gL
e
Ne
t
m
o
de
l
i
s
c
h
o
s
e
n
s
i
nc
e
i
t
h
a
s
pr
o
duc
e
d
b
e
tt
e
r
r
e
s
u
l
t
s
t
h
a
n
Al
e
xn
e
t
o
n
t
h
e
DH
C
D
da
t
a
s
e
t.
Goo
gL
e
Ne
t
,
t
h
e
2014
I
L
S
VR
C
c
h
a
m
p
i
o
n
,
gr
e
a
t
l
y
r
e
duc
e
d
i
t
s
e
r
r
o
r
r
a
t
e
t
h
a
n
ot
h
e
r
m
o
de
l
s
.
I
n
put
im
a
ge
o
f
s
i
z
e
224
×
224
w
i
t
h
R
GB
c
h
a
nn
e
l
s
i
s
a
pp
li
e
d
a
s
a
n
i
nput
to
t
h
e
m
o
de
l
.
O
n
e
o
f
t
h
e
ke
y
c
o
m
p
o
n
e
n
t
s
o
f
t
hi
s
m
o
de
l
i
s
i
t
s
1
×
1
c
o
n
v
o
l
ut
i
o
n
,
whi
c
h
s
i
g
nif
ica
n
t
l
y
l
o
we
r
s
t
h
e
n
u
m
be
r
we
i
g
h
t
s
a
n
d
pa
r
a
m
e
t
e
r
bi
a
s
e
s
,
c
r
e
a
t
i
n
g
a
de
e
pe
r
a
r
c
hi
t
e
c
t
ur
e
.
T
h
e
7
×
7
f
e
a
t
ur
e
m
a
p
i
s
a
v
e
r
a
ge
d
i
n
t
o
a
1
×
1
to
r
e
duc
e
t
h
e
n
u
m
be
r
o
f
tr
a
i
n
a
bl
e
pa
r
a
m
e
t
e
r
s
[
22]
.
W
h
e
n
g
l
o
b
a
l
a
ve
r
a
ge
poo
l
i
ng
i
s
us
e
d,
a
c
c
ur
a
c
y
i
nc
r
e
a
s
e
s
by
0.
6%
.
T
o
h
a
n
d
l
e
o
bj
e
c
t
s
m
o
r
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
do
n
e
s
i
a
n
J
E
l
e
c
E
n
g
&
C
o
m
p
S
c
i
I
S
S
N:
2
5
0
2
-
4
7
52
Dis
c
r
e
te
w
av
e
let
tr
ans
f
or
m
and
c
onv
olut
ional
ne
ur
al
ne
t
w
or
k
bas
e
d
…
(
Shr
addha
V
.
She
lke
)
1371
e
f
f
e
c
t
i
v
e
ly
a
t
d
i
f
f
e
r
e
n
t
s
c
a
l
e
s
1
×
1,
3
×
3,
a
n
d
5
×
5
p
a
r
a
l
l
e
l
3
×
3
m
a
x
po
o
l
i
ng
a
n
d
c
o
nv
o
l
ut
i
o
n
a
r
e
e
x
e
c
ut
e
d.
T
o
o
b
t
a
i
n
o
u
t
pu
t
o
f
i
n
c
e
pt
i
o
n
m
o
du
l
e
,
t
h
e
s
e
m
o
du
l
e
s
a
r
e
s
t
a
c
ke
d.
I
n
c
e
pt
i
o
n
m
o
du
l
e
d
e
s
i
g
n
e
d
i
n
M
A
T
L
A
B
De
e
p
Ne
t
w
o
r
k
de
s
i
g
n
e
r
i
s
s
h
o
wn
i
n
F
i
gur
e
5.
A
uxil
i
a
r
y
c
l
a
s
s
if
i
e
r
b
r
a
nc
h
e
s
,
i
n
c
l
ud
i
ng
a
v
e
r
a
ge
po
o
l
i
n
g,
c
o
n
v
o
l
ut
i
o
n
w
i
t
h
128
f
il
t
e
r
s
,
a
n
d
S
o
f
t
M
a
x
a
c
t
i
v
a
t
i
o
n
f
u
n
c
t
i
o
n,
a
r
e
a
dde
d
to
t
h
e
i
ni
t
i
a
l
m
o
de
l
s
.
T
h
e
v
a
ni
s
h
i
ng
gr
a
d
i
e
n
t
pr
o
bl
e
m
i
s
s
o
l
ve
d
w
i
t
h
l
a
y
e
r
s
[
23]
.
A
22
l
a
y
e
r
s
de
e
p
a
r
c
hi
t
e
c
t
ur
e
i
s
c
o
m
put
a
t
i
o
n
a
l
e
f
f
ic
i
e
n
t
a
s
i
t
c
a
n
r
un
o
n
de
vi
c
e
s
w
i
t
h
l
o
w
c
o
m
put
a
t
i
o
n
a
l
c
o
s
t.
Out
pu
t
o
f
i
nc
e
pt
i
o
n
(
4a
)
a
n
d
(
4d)
i
s
c
o
nn
e
c
t
e
d
to
two
a
uxi
l
i
a
r
y
c
l
a
s
s
if
i
e
r
s
.
W
i
t
h
f
o
l
l
o
w
i
ng
pr
o
pe
r
t
i
e
s
:
An
a
v
e
r
a
ge
po
o
l
i
n
g
l
a
y
e
r
,
5
×
5
f
il
t
e
r
a
n
d
s
t
r
i
de
o
f
3
.
1x
1
C
o
nv
o
l
ut
i
o
n
a
l
l
a
y
e
r
h
a
vi
ng
128
f
il
t
e
r
s
a
n
d
a
c
t
i
v
a
t
i
o
n
f
u
n
c
t
i
o
n
us
e
d
i
s
R
e
L
U
.
A
F
u
ll
y
C
o
nn
e
c
t
e
d
o
u
t
pu
t
l
a
y
e
r
w
i
t
h
1025
o
u
t
pu
t
s
a
n
d
a
c
t
i
v
a
t
i
o
n
f
u
n
c
t
i
o
n
us
e
d
i
s
R
e
L
U
.
Dr
o
p
o
u
t
r
a
t
i
o
0.
7
i
s
s
e
t
whi
l
e
dr
o
p
o
u
t
r
e
gu
l
a
r
i
z
a
t
i
o
n
.
A
S
o
f
t
M
a
x
c
l
a
s
s
if
i
e
r
w
i
t
h
1
,
000
c
l
a
s
s
e
s
o
ut
pu
t.
T
h
e
m
o
de
l
e
m
p
l
o
y
s
R
e
L
U
a
s
t
h
e
a
c
t
i
v
a
t
i
o
n
f
u
n
c
t
i
o
n
f
o
r
e
v
e
r
y
c
o
nv
o
l
ut
i
o
n
.
R
e
L
U
i
s
o
n
e
o
f
t
h
e
m
o
s
t
c
o
m
m
o
nly
e
m
p
l
o
y
e
d
a
c
t
i
v
a
t
i
o
n
f
u
n
c
t
i
o
n
i
s
C
NN
s
.
Ou
t
pu
t
o
f
R
e
L
U
i
s
0
wh
e
n
t
h
e
r
e
i
s
n
e
ga
t
i
v
e
i
nput
a
n
d
f
o
r
a
ny
po
s
i
t
i
ve
v
a
l
ue
o
f
i
n
put
o
u
t
pu
t
i
s
s
a
m
e
a
s
i
n
p
ut.
B
a
s
i
c
a
ll
y
i
t
i
s
us
e
d
to
c
a
n
c
e
l
a
ll
n
e
ga
t
i
v
e
o
u
t
p
ut
s
[
20]
.
M
a
t
h
e
m
a
t
i
c
a
ll
y
R
e
L
U
i
s
r
e
pr
e
s
e
n
t
e
d
by
(
5
)
A
d
v
a
n
t
a
ge
o
f
us
i
ng
R
e
L
U
i
s
i
t
c
a
n
h
e
l
p
t
o
o
v
e
r
c
o
m
e
va
ni
s
hi
n
g
gr
a
d
i
e
n
t
pr
o
bl
e
m
[
24]
.
(
)
=
(
0
,
)
(5
)
(
)
=
0
≤
0
(
)
=
>
0
F
i
gur
e
5.
G
oo
gL
e
Ne
t
i
nc
e
pt
i
o
n
m
o
du
l
e
de
s
i
g
n
e
d
in
M
A
T
L
AB
de
e
p
n
e
t
wo
r
k
de
s
i
g
n
e
r
too
l
2.
4.
S
t
oc
h
as
t
ic
gr
ad
ient
d
e
s
c
e
n
t
wit
h
m
om
e
n
t
u
m
(
S
GDM
)
Dur
i
n
g
b
a
c
k
pr
o
pa
ga
t
i
o
n
i
n
o
r
de
r
to
f
i
ne
t
un
e
pa
r
a
m
e
t
e
r
s
o
f
ne
t
wor
k
o
p
t
i
mi
z
e
r
s
a
r
e
i
m
p
l
e
m
e
n
t
e
d.
M
a
i
n
o
bj
e
c
t
i
v
e
o
f
o
p
t
i
mi
z
e
r
s
i
s
to
f
i
n
d
g
l
o
ba
l
mi
n
im
a
i
n
o
r
de
r
to
m
i
n
im
i
z
e
l
o
s
s
f
u
n
c
t
i
o
n
.
I
n
s
to
c
h
a
s
t
i
c
gr
a
d
i
e
n
t
de
s
c
e
n
t
w
i
t
h
m
o
m
e
n
t
u
m
e
x
po
n
e
n
t
i
a
ll
y
we
i
g
h
t
e
d
m
o
vi
ng
a
v
e
r
a
ge
i
s
c
a
l
c
u
l
a
t
e
d
by
(
6
)
,
wh
e
r
e
i
s
hy
pe
r
-
pa
r
a
m
e
t
e
r
r
a
n
g
i
n
g
f
r
o
m
0
to
1.
E
qua
t
i
o
n
s
ho
ws
t
h
a
t
v
e
l
o
c
i
t
y
V
t
de
pe
n
ds
o
n
β
.
T
h
e
e
x
pe
r
i
m
e
nt
a
tt
e
m
pt
to
o
b
t
a
i
n
a
n
a
v
e
r
a
ge
o
f
m
o
r
e
hi
s
t
o
r
i
c
a
l
d
a
t
a
t
h
e
gr
e
a
t
e
r
t
h
e
v
a
l
ue
o
f
β
,
a
n
d
vi
c
e
v
e
r
s
a
.
=
∗
−
1
+
(
1
−
)
∗
(
6)
ℎ
∈
[
0
,
1
]
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2
5
0
2
-
4
7
52
In
do
n
e
s
i
a
n
J
E
l
e
c
E
n
g
&
C
o
m
p
S
c
i
,
Vo
l
.
3
8
,
N
o
.
2
,
M
a
y
20
2
5
:
1
367
-
1
375
1372
W
hil
e
c
a
l
c
u
l
a
t
i
n
g
g
l
o
b
a
l
mi
n
im
a
c
h
a
n
ge
s
i
n
gr
a
d
ien
t
a
r
e
i
n
d
i
c
a
t
e
d
by
v
e
l
o
c
i
t
y
i
n
S
GD
M
g
i
ve
n
by
(
7)
,
wh
e
r
e
i
s
w
ei
g
h
t
.
+
1
=
–
(7
)
ℎ
=
∗
−
1
+
W
h
e
n
β
i
s
s
e
t
to
0,
i
t
c
a
us
e
s
a
de
c
a
y
,
h
o
we
v
e
r
w
he
n
i
t
i
s
e
qua
l
t
o
1,
t
h
e
r
e
i
s
n
o
de
c
a
y
.
Us
ua
ll
y
,
o
nl
y
0.
5,
0
.
9,
or
0.
99
a
r
e
us
e
d
f
o
r
β
.
S
i
n
c
e
S
GD
M
i
nv
o
l
ve
s
m
o
m
e
n
t
u
m
,
i
n
c
o
n
t
r
a
s
t
to
S
GD
,
a
n
d
c
a
n
e
s
c
a
pe
l
o
c
a
l
mi
n
im
a
a
n
d
a
t
t
a
i
n
g
l
o
ba
l
mi
n
im
a
,
i
t
i
s
a
b
e
t
t
e
r
o
p
t
i
o
n
t
h
a
n
S
GD
f
o
r
o
p
t
i
mi
z
a
t
i
o
n
[
25]
.
3.
RE
S
UL
T
S
AN
D
DI
S
CU
S
S
I
ON
3.
1.
Re
s
u
l
t
s
ob
t
ain
e
d
f
or
DWT
-
CN
N
m
od
e
l
M
A
T
L
A
B
de
e
p
n
e
t
wo
r
k
de
s
i
g
ne
r
i
s
us
e
d
f
o
r
i
m
p
le
m
e
n
t
a
t
i
o
n
.
T
h
e
im
a
ge
da
t
a
s
t
or
e
o
f
72000
i
m
a
ge
s
f
r
o
m
t
h
e
DH
C
D
da
t
a
s
e
t
i
s
c
r
e
a
t
e
d,
a
ppr
o
xi
m
a
t
i
o
n
c
o
e
f
f
i
c
i
e
n
t
(
c
A
)
o
b
t
a
i
n
e
d
a
f
t
e
r
d
i
s
c
r
e
t
e
wa
v
e
l
e
t
tr
a
ns
f
o
r
m
a
r
e
a
pp
l
i
e
d
to
C
NN
m
o
de
l
.
Da
t
a
s
to
r
e
i
m
a
ge
s
a
r
e
s
p
l
i
t
a
t
r
a
n
do
m
80:20,
w
i
t
h
80%
o
f
t
h
e
i
m
a
ge
s
us
e
d
f
o
r
t
r
a
i
ni
ng
a
n
d
20%
f
o
r
t
e
s
t
i
n
g.
T
h
e
m
o
de
l
i
s
i
m
p
le
m
e
n
t
e
d
o
n
a
n
i
5
C
P
U.
L
e
a
r
ni
n
g
r
a
t
e
i
s
s
e
t
to
0
.
001
a
n
d
m
o
de
l
i
s
t
r
a
i
n
e
d
f
o
r
10,
15
a
n
d
20
e
po
c
h
s
,
o
p
t
i
m
i
z
e
r
us
e
d
i
s
S
GD
M
.
T
a
bl
e
1
s
h
o
ws
r
e
s
u
l
t
s
o
b
t
a
i
n
e
d
f
o
r
DW
T
–
C
NN
m
o
de
l
.
T
a
bl
e
1.
T
h
e
r
e
s
u
l
t
s
o
f
c
l
a
s
s
i
f
i
c
a
t
i
o
n
a
c
c
ur
a
c
y
f
o
r
DW
T
-
C
NN
m
o
de
l
E
poc
h
%
T
r
a
in
in
g
a
c
c
u
r
a
c
y
%
T
e
s
ti
ng
a
c
c
ur
a
c
y
L
o
s
s
10
86.47
87.89
0.232
15
98.99
98.95
0.085
20
98.14
94.38
0.045
I
t
i
s
o
b
s
e
r
v
e
d
t
h
a
t
b
e
s
t
c
l
a
s
s
i
f
i
c
a
t
i
o
n
a
c
c
ur
a
c
y
i
s
o
b
t
a
i
n
e
d
f
o
r
15
e
po
c
h
s
.
T
e
s
t
i
n
g
a
c
c
ur
a
c
y
i
s
r
e
duc
e
d
f
o
r
20
e
p
o
c
h
s
due
to
o
v
e
r
f
i
t
t
i
n
g.
F
ur
t
h
e
r
m
o
de
l
i
s
t
r
a
i
n
e
d
f
o
r
l
e
a
r
ni
ng
r
a
t
e
s
o
f
0.
001,
0.
01
a
n
d
0.
015
by
ke
e
p
i
n
g
n
u
m
be
r
o
f
e
po
c
h
a
s
15.
R
e
s
u
l
t
s
a
r
e
s
h
o
wn
i
n
T
a
bl
e
2.
T
a
bl
e
2.
L
e
a
r
ni
ng
r
a
t
e
a
n
a
l
y
s
i
s
f
o
r
DW
T
-
C
NN
m
o
de
l
L
e
a
r
ni
ng
r
a
te
0.001
0.01
0.015
%
T
r
a
in
in
g
a
c
c
u
r
a
c
y
98.99
99.15
88.34
%
T
e
s
ti
ng
a
c
c
ur
a
c
y
98.95
98.97
74.38
L
o
s
s
0.085
0.098
0.045
R
e
s
u
l
t
s
s
h
o
ws
t
h
a
t
m
a
xim
u
m
t
e
s
t
i
n
g
a
c
c
ur
a
c
y
o
f
98.
97%
i
s
o
b
t
a
i
n
e
d
f
o
r
l
e
a
r
ni
ng
r
a
t
e
o
f
0.
01.
A
c
c
ur
a
c
y
de
c
r
e
a
s
e
s
i
f
l
e
a
r
ni
ng
r
a
t
e
i
nc
r
e
a
s
e
s
to
0.
015
b
e
c
a
us
e
a
hi
g
h
e
r
l
e
a
r
ni
n
g
r
a
t
e
c
a
us
e
s
t
h
e
m
o
de
l
t
o
l
e
a
r
n
r
a
p
i
d
ly
,
whi
c
h
a
f
f
e
c
t
s
n
e
t
wor
k
pe
r
f
o
r
m
a
n
c
e
.
I
t
i
s
a
l
s
o
o
b
s
e
r
v
e
d
t
h
a
t
tr
a
i
ni
ng
t
i
m
e
de
c
r
e
a
s
e
s
a
s
l
e
a
r
ni
ng
r
a
t
e
i
nc
r
e
a
s
e
s
b
e
c
a
us
e
n
e
t
wo
r
ks
l
e
a
r
n
qu
i
c
k
ly
.
3.
2.
Re
s
u
l
t
s
ob
t
ain
e
d
f
or
GoogL
e
Ne
t
m
od
e
l
T
h
e
Go
o
gL
e
Ne
t
m
o
de
l
a
c
c
e
pt
s
R
GB
I
n
put
i
m
a
ge
o
f
s
i
z
e
224
x
224,
'
C
o
l
o
r
P
r
e
pr
o
c
e
s
s
i
n
g
'
,
'
gr
a
y
2r
gb'
pr
o
pe
r
t
y
i
s
s
e
t
whil
e
c
r
e
a
t
i
n
g
im
a
ge
d
a
t
a
s
to
r
e
.
T
h
e
S
GD
M
o
pt
i
m
i
z
e
r
w
i
t
h
a
l
e
a
r
ni
ng
r
a
t
e
o
f
0.
001
i
s
us
e
d
t
o
a
n
a
ly
z
e
t
h
e
m
o
de
l
o
ve
r
10,
15,
a
n
d
20
e
po
c
h
s
.
T
h
e
tr
a
i
n
e
d
m
o
de
l
i
s
t
h
e
n
e
x
po
r
t
e
d
to
t
h
e
wo
r
ks
pa
c
e
,
a
n
d
20%
o
f
da
t
a
po
i
n
t
s
f
r
o
m
t
h
e
im
a
ge
da
t
a
s
t
or
e
a
r
e
us
e
d
f
o
r
t
e
s
t
i
n
g
pur
po
s
e
s
.
T
a
bl
e
3
d
i
s
p
l
a
y
s
t
h
e
r
e
s
u
l
t
s
o
b
t
a
i
n
e
d
a
f
t
e
r
t
r
a
i
ni
ng
a
n
d
t
e
s
t
i
n
g
t
h
e
m
o
de
l
f
o
r
v
a
r
i
a
bl
e
e
p
o
c
h
s
.
T
a
bl
e
3.
T
h
e
r
e
s
u
l
t
s
o
f
c
l
a
s
s
i
f
i
c
a
t
i
o
n
a
c
c
ur
a
c
y
f
o
r
f
i
ne
-
t
un
e
d
Goo
gL
e
Ne
t
m
o
de
l
e
p
oc
h
%
T
r
a
in
in
g
a
c
c
u
r
a
c
y
%
T
e
s
ti
ng
a
c
c
ur
a
c
y
L
o
s
s
10
96.97
98.34
0.1589
15
99.78
99.65
0.0875
20
99.89
98.66
0.0646
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
do
n
e
s
i
a
n
J
E
l
e
c
E
n
g
&
C
o
m
p
S
c
i
I
S
S
N:
2
5
0
2
-
4
7
52
Dis
c
r
e
te
w
av
e
let
tr
ans
f
or
m
and
c
onv
olut
ional
ne
ur
al
ne
t
w
or
k
bas
e
d
…
(
Shr
addha
V
.
She
lke
)
1373
R
e
s
u
l
t
s
s
h
o
w
t
h
a
t
a
s
t
h
e
n
u
m
be
r
o
f
e
po
c
h
s
i
nc
r
e
a
s
e
s
,
t
r
a
i
ni
ng
a
c
c
ur
a
c
y
a
l
s
o
i
n
c
r
e
a
s
e
s
a
s
t
h
e
m
o
de
l
l
e
a
r
n
s
m
o
r
e
f
r
o
m
t
r
a
i
ni
ng
da
t
a
.
I
t
a
l
s
o
r
e
s
u
l
t
s
i
n
a
n
i
nc
r
e
a
s
e
i
n
a
v
e
r
a
ge
t
i
m
e
pe
r
e
po
c
h
.
I
t
i
s
o
b
s
e
r
v
e
d
t
h
a
t
t
r
a
i
ni
ng
m
o
de
l
f
o
r
15
e
po
c
h
s
r
e
s
u
l
t
e
d
i
n
a
m
a
xi
m
u
m
t
e
s
t
i
n
g
a
c
c
ur
a
c
y
o
f
99.
65%
wi
t
h
a
l
o
s
s
o
f
0.
0875
.
L
e
a
r
ni
n
g
r
a
t
e
a
n
a
ly
s
i
s
o
f
m
o
de
l
i
s
s
h
o
wn
i
n
T
a
bl
e
4.
T
a
bl
e
4.
L
e
a
r
ni
ng
r
a
t
e
a
n
a
l
y
s
i
s
f
o
r
Goo
gL
e
Ne
t
m
o
de
l
L
e
a
r
ni
ng
r
a
te
0.001
0.01
0.015
%
T
r
a
in
in
g
a
c
c
u
r
a
c
y
99.78 %
99.81 %
99.52 %
%
T
e
s
ti
ng
a
c
c
ur
a
c
y
99.65 %
99.68 %
98.87 %
L
o
s
s
0.0875
0.0635
0.0884
T
h
e
b
e
s
t
c
l
a
s
s
if
i
c
a
t
i
o
n
a
c
c
ur
a
c
y
o
f
99.
68%
i
s
o
b
t
a
i
ne
d
f
o
r
l
e
a
r
ni
ng
r
a
t
e
o
f
0.
01
wi
t
h
l
o
s
s
o
f
0.
0875.
C
o
nf
us
i
o
n
m
a
t
r
i
x
i
n
c
l
ude
s
400
i
m
a
ge
s
o
f
36
c
l
a
s
s
e
s
e
a
c
h
a
r
e
us
e
d
f
o
r
t
e
s
t
i
n
g
pur
po
s
e
.
I
t
de
m
o
n
s
t
r
a
t
e
s
t
h
e
c
o
nf
us
i
o
n
o
f
s
i
mi
l
a
r
c
ha
r
a
c
t
e
r
s
wi
t
h
l
i
t
t
l
e
v
a
r
i
a
t
i
o
n
s
.
I
t
c
a
n
b
e
o
b
s
e
r
v
e
d
f
o
r
c
o
nf
us
i
o
n
m
a
t
r
i
x
t
h
a
t
out
o
f
400
s
a
m
p
l
e
s
o
f
c
h
a
r
a
c
t
e
r
“
भ
”
,
15
s
a
m
p
l
e
s
a
r
e
pr
e
d
i
c
t
e
d
a
s
“
म
”
a
n
d
r
e
m
a
i
n
i
ng
385
s
a
m
p
l
e
s
a
r
e
c
o
r
r
e
c
t
l
y
pr
e
d
i
c
t
e
d
a
s
“
भ
”
.
S
i
mi
l
a
r
ly
o
ut
o
f
400
s
a
m
p
l
e
o
f
c
h
a
r
a
c
t
e
r
“
व
”
,
3
s
a
m
p
l
e
s
a
r
e
i
n
c
o
r
r
e
c
t
l
y
pr
e
d
i
c
t
e
d
a
s
“
ब
”
,
1
i
s
c
l
a
s
s
if
i
e
d
a
s
“
प
”
,
1
s
a
m
p
l
e
a
s
“
”
,
1
s
a
m
p
l
e
a
s
“
न
”
a
n
d
1
s
a
m
p
l
e
a
s
“
त
”
w
i
t
h
r
e
m
a
i
n
i
ng
393
c
ha
r
a
c
t
e
r
s
a
r
e
c
o
r
r
e
c
t
l
y
c
l
a
s
s
if
i
e
d
a
s
“
व
”.
3.
3.
Val
id
at
ion
of
r
e
s
u
l
t
s
T
a
bl
e
5
g
i
ve
s
c
o
m
pa
r
i
s
o
n
o
f
pr
o
p
o
s
e
d
m
o
de
l
w
i
t
h
ot
h
e
r
m
e
t
h
o
do
l
o
g
i
e
s
.
T
h
e
f
i
nd
i
ng
s
s
h
o
w
t
h
a
t
t
h
e
pr
o
p
o
s
e
d
DW
T
-
C
NN
a
l
go
r
i
t
hm
o
ut
pe
r
f
o
r
m
e
d
ot
h
e
r
t
e
c
h
ni
que
s
de
s
c
r
i
be
d
i
n
t
h
e
l
i
t
e
r
a
t
ur
e
wi
t
h
a
n
a
c
c
ur
a
c
y
o
f
98.
97%
.
Us
i
n
g
DW
T
a
l
l
o
w
s
t
h
e
e
x
t
r
a
c
t
i
o
n
o
f
tex
t
ur
a
l
e
l
e
m
e
n
t
s
f
r
o
m
a
n
i
m
a
ge
,
s
uc
h
a
s
i
t
s
s
tr
u
c
t
ur
e
o
r
pa
tt
e
r
n
s
.
A
dd
i
t
i
o
na
l
ly
,
t
h
e
a
c
c
ur
a
c
y
r
e
a
c
h
e
d
t
o
99
.
68%
wh
e
n
c
o
m
bi
ne
d
w
i
t
h
a
f
i
ne
-
t
un
e
d
Go
o
gL
e
Ne
t
m
o
de
l
.
T
h
e
22
-
l
a
y
e
r
de
e
p
de
s
i
g
n
o
f
Go
o
gL
e
Ne
t
c
o
n
t
r
i
b
u
t
e
s
f
o
r
i
t
s
e
nha
n
c
e
d
pe
r
f
o
r
m
a
n
c
e
.
B
y
a
ll
o
w
i
ng
t
h
e
us
e
o
f
d
i
f
f
e
r
e
n
t
f
il
t
e
r
s
i
z
e
s
,
t
h
e
i
n
c
e
pt
i
o
n
m
o
du
l
e
i
m
pr
o
v
e
s
Goo
gL
e
Ne
t
'
s
e
f
f
i
c
i
e
nc
y
.
T
a
bl
e
5.
C
o
m
pa
r
i
s
o
n
o
f
r
e
s
u
l
t
s
o
b
t
a
i
ne
d
f
o
r
DH
C
D
da
t
a
s
e
t
A
ut
ho
r
Y
e
a
r
A
lg
o
r
i
th
m
P
e
r
c
e
nt
a
ge
a
c
c
ur
a
c
y
A
c
ha
r
y
a
e
t
al
.
[
17]
2015
D
e
e
p
c
o
n
vo
lu
ti
o
na
l
n
e
ur
a
l
n
e
tw
o
r
k
98.47
A
ne
ja
a
nd A
ne
ja
[
11]
2019
T
r
a
ns
f
e
r
le
a
r
n
in
g
of
D
C
N
N
A
le
x
N
e
t,
D
e
ns
e
N
e
, V
gg, i
nc
e
pt
i
o
nV
3
H
ig
h
e
s
t
a
c
c
ur
a
c
y
A
le
x
n
e
t:
98
M
a
no
c
ha
a
nd
T
e
w
a
r
i
[
15]
2021
D
e
e
p
l
e
a
r
ni
ng C
N
N
+
S
V
M
, K
N
N
, M
L
P
92
V
in
a
e
t
al
.
[
13]
2022
D
e
e
p C
N
N
m
o
d
e
ls
w
it
h 3,8,12 C
O
N
V
l
a
y
e
r
s
98.94
S
he
lk
e
e
t
al
.
[
16]
2023
H
y
br
id
D
W
T
-
D
C
T
, S
V
M
H
O
G
, S
V
M
83.09
97.10
P
r
o
p
o
s
e
d A
lg
o
r
it
h
m
2024
P
r
o
p
o
s
e
d D
W
T
-
C
N
N
m
o
d
e
l
98.97
F
in
e
-
tu
ne
d
G
oo
g
L
e
N
e
t
m
o
d
e
l
99.68%
4.
CONC
L
USI
ON
Ha
n
dwr
i
t
t
e
n
S
a
ns
kr
i
t
m
a
n
u
s
c
r
i
pt
s
m
us
t
b
e
di
g
i
t
i
z
e
d
i
n
o
r
de
r
to
b
e
pr
e
s
e
r
v
e
d
f
o
r
f
ut
ur
e
ge
n
e
r
a
t
i
o
n
s
.
I
n
t
hi
s
pa
pe
r
a
n
a
ppr
o
a
c
h
us
i
ng
D
W
T
f
o
r
f
e
a
t
ur
e
e
x
tr
a
c
t
i
o
n
a
n
d
c
o
nv
o
l
ut
i
o
n
a
l
ne
ur
a
l
n
e
t
wo
r
k
f
o
r
c
l
a
s
s
if
i
c
a
t
i
o
n
o
f
h
a
n
dwr
i
t
t
e
n
c
ha
r
a
c
t
e
r
s
i
s
i
m
p
l
e
men
t
e
d.
F
i
n
e
-
t
un
e
d
Goo
gL
e
Ne
t
m
o
de
l
i
s
a
l
s
o
i
m
p
le
m
e
n
t
e
d
h
e
r
e
.
DH
C
D
da
t
a
s
e
t
i
nc
l
ude
s
72000
h
a
n
d
wr
i
t
t
e
n
i
m
a
ge
s
i
s
a
v
a
il
a
bl
e
f
r
e
e
ly
f
o
r
r
e
s
e
a
r
c
h
i
s
us
e
d
f
o
r
thi
s
s
t
ud
y
.
A
ppr
o
xim
a
t
i
o
n
c
o
e
f
f
i
c
i
e
n
t
o
b
t
a
i
ne
d
a
f
t
e
r
a
ppl
yin
g
DW
T
o
n
i
n
put
i
m
a
ge
s
i
s
f
e
d
to
C
NN
f
o
r
i
m
pr
o
vi
ng
f
e
a
t
ur
e
l
e
a
r
ni
ng
c
a
p
a
c
i
t
y
.
Af
t
e
r
tr
a
i
ni
n
g
b
o
t
h
t
h
e
m
o
de
l
s
by
v
a
r
yi
ng
hy
pe
r
pa
r
a
m
e
t
e
r
s
,
o
p
t
i
m
u
m
va
l
ue
s
o
f
l
e
a
r
ni
ng
r
a
t
e
a
n
d
e
po
c
h
s
i
s
o
b
t
a
i
n
e
d
a
s
0.
01
a
n
d
1
5
r
e
s
pe
c
t
i
ve
l
y
.
R
e
s
u
l
t
s
s
h
o
ws
t
h
a
t
DW
T
-
C
NN
m
o
de
l
g
i
ve
s
t
r
a
i
ni
n
g
a
c
c
ur
a
c
y
o
f
99.
15%
a
n
d
t
e
s
t
i
n
g
a
c
c
ur
a
c
y
o
f
98.
97%
wi
t
h
t
h
e
l
o
s
s
o
f
0.
098.
F
i
n
e
-
t
un
e
d
Goo
gL
e
Ne
t
m
o
de
l
h
a
s
s
i
g
nif
i
c
a
n
t
l
y
i
nc
r
e
a
s
e
d
a
c
c
ur
a
c
y
t
o
99
.
81%
whi
l
e
t
r
a
i
ni
ng
a
n
d
99.
68%
whi
l
e
t
e
s
t
i
n
g
w
i
t
h
l
o
s
s
o
f
0.
0635.
A
dd
i
t
i
o
n
a
ll
y
,
t
h
e
i
n
c
e
pt
i
o
n
m
o
du
l
e
a
ll
o
ws
s
e
v
e
r
a
l
f
il
t
e
r
s
i
z
e
s
t
o
b
e
us
e
d
s
i
m
u
l
t
a
ne
o
us
l
y
.
Ho
we
v
e
r
,
a
c
h
a
ll
e
n
g
e
w
i
t
h
Goo
gL
e
Ne
t
i
s
t
h
a
t
t
h
e
m
o
de
l
r
e
qu
i
r
e
s
m
o
r
e
t
i
m
e
to
t
r
a
i
n
due
to
i
t
s
de
e
p
l
a
y
e
r
e
d
de
s
i
gn
.
T
h
e
i
nc
r
e
a
s
i
n
g
n
u
m
be
r
o
f
e
po
c
h
s
a
n
d
l
e
a
r
ni
ng
r
a
t
e
r
e
s
u
l
t
e
d
i
n
o
v
e
r
f
i
t
t
i
n
g
o
f
a
m
o
de
l
w
hi
c
h
c
a
n
b
e
r
e
duc
e
d
b
y
i
nc
r
e
a
s
i
n
g
dr
o
p
o
u
t
r
e
gu
l
a
r
i
z
a
t
i
o
n
.
Us
i
ng
t
h
e
S
G
DM
o
p
t
i
mi
z
e
r
,
w
hi
c
h
pr
o
vi
de
s
b
e
t
t
e
r
ge
n
e
r
a
l
i
z
a
t
i
o
n
t
h
a
n
ot
h
e
r
o
p
t
i
mi
z
e
r
s
de
s
p
i
t
e
b
e
i
ng
s
l
o
w
e
r
t
h
a
n
o
t
h
e
r
s
,
i
m
pr
o
v
e
s
t
h
e
o
v
e
r
a
l
l
pe
r
f
o
r
m
a
n
c
e
o
f
a
m
o
de
l
.
T
h
e
Goo
gL
e
Ne
t
m
o
de
l
'
s
e
f
f
e
c
t
i
ve
n
e
s
s
c
o
m
e
s
f
r
o
m
t
he
us
e
o
f
t
h
e
i
n
c
e
pt
i
o
n
m
o
du
l
e
,
whi
c
h
m
a
i
n
t
a
i
ns
pr
e
c
i
s
i
o
n
whil
e
l
o
we
r
i
n
g
pr
o
c
e
s
s
i
ng
c
o
m
p
l
e
xi
t
y
.
T
h
e
c
o
nf
us
i
o
n
m
a
t
r
i
x
s
h
o
ws
t
h
a
t
t
h
e
r
e
i
s
c
o
nf
us
i
o
n
b
e
t
we
e
n
c
h
a
r
a
c
t
e
r
s
“
भ
”
a
n
d
“
म
”,
“
ढ
”
a
n
d
“
द
”,
“
घ
”
a
n
d
“
ध
”
.
C
o
m
pa
r
i
s
o
n
o
f
r
e
s
u
l
t
s
s
h
o
ws
t
h
a
t
pr
o
p
o
s
e
d
D
W
T
-
C
NN
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2
5
0
2
-
4
7
52
In
do
n
e
s
i
a
n
J
E
l
e
c
E
n
g
&
C
o
m
p
S
c
i
,
Vo
l
.
3
8
,
N
o
.
2
,
M
a
y
20
2
5
:
1
367
-
1
375
1374
m
o
de
l
o
u
t
pe
r
f
o
r
m
e
d
t
h
a
n
o
t
h
e
r
m
o
de
l
s
.
H
y
b
r
i
d
D
W
T
a
n
d
hi
s
t
o
gr
a
m
o
f
o
r
i
e
n
t
e
d
gr
a
d
i
e
n
t
s
(
HO
G)
c
a
n
b
e
us
e
d
f
o
r
r
e
c
o
gni
t
i
o
n
o
f
h
a
n
d
wr
i
t
t
e
n
S
a
n
s
kr
i
t
wo
r
ds
.
Al
t
h
o
ugh
Go
o
gL
e
Ne
t
pe
r
f
o
r
m
e
d
b
e
t
t
e
r
,
t
h
e
DW
T
-
C
NN
m
o
de
l
i
s
a
go
o
d
a
l
t
e
r
n
a
t
i
v
e
t
h
a
t
r
e
qui
r
e
s
l
o
we
r
c
o
m
pu
t
a
t
i
o
n
a
l
c
o
s
t
.
RE
F
E
R
E
NC
E
S
[
1]
R
.
J
a
y
a
d
e
v
a
n,
S
.
R
.
K
o
lh
e
,
P
.
M
.
P
a
ti
l,
a
nd
U
.
P
a
l,
“
O
f
f
li
ne
r
e
c
o
gni
ti
o
n
of
D
e
v
a
na
ga
r
i
s
c
r
ip
t:
A
s
ur
ve
y
,”
I
E
E
E
T
r
ans
ac
ti
ons
on
Sy
s
te
m
s
,
M
an
and
C
y
be
r
n
e
ti
c
s
P
ar
t
C
:
A
ppl
ic
at
io
ns
and
R
e
v
ie
w
s
,
vo
l.
41,
n
o
.
6,
pp.
782
–
796,
N
ov
.
2011,
do
i:
10.1109/
T
S
M
C
C
.2010.2095841.
[
2]
D
.
T
.
M
a
ne
a
nd
U
.
V
.
K
ul
ka
r
ni
,
“
V
is
ua
li
z
in
g
a
nd
U
nde
r
s
ta
ndi
ng
C
us
to
mi
z
e
d
C
o
n
vo
lu
ti
o
na
l
N
e
ur
a
l
N
e
tw
o
r
k
f
or
R
e
c
o
gni
ti
on
of
H
a
ndw
r
it
t
e
n M
a
r
a
th
i
N
ume
r
a
ls
,”
P
r
oc
e
di
a C
om
put
e
r
Sc
ie
nc
e
,
vo
l.
132, pp. 1123
–
1137, 2018, d
o
i
:
10.1016/
j.
pr
oc
s
.2018.05.027.
[
3]
A
.
M
o
udgi
l,
S
.
S
in
gh,
V
.
G
a
ut
a
m,
S
.
R
a
ni
,
a
nd
S
.
H
.
S
ha
h,
“
H
a
ndw
r
it
t
e
n
de
v
a
na
ga
r
i
ma
nus
c
r
ip
t
c
ha
r
a
c
t
e
r
s
r
e
c
o
gni
ti
o
n
u
s
in
g
c
a
ps
ne
t,
”
I
nt
e
r
nat
io
nal
J
our
nal
of
C
ogni
ti
v
e
C
om
put
in
g
in
E
ngi
ne
e
r
in
g
,
vo
l.
4,
pp.
47
–
54,
J
un.
2023
,
do
i:
10.1016/j
.i
jc
c
e
.2023.02.001.
[
4]
P
.
P
.
N
a
ir
,
A
. J
a
me
s
,
P
.
S
im
o
n,
a
nd
P
. V
.
B
ha
gy
a
s
r
e
e
,
“
M
a
la
y
a
la
m
H
a
ndw
r
it
te
n
C
ha
r
a
c
t
e
r
R
e
c
o
gni
ti
o
n
us
in
g
C
N
N
A
r
c
hi
te
c
tu
r
e
,”
I
ndone
s
ia
n
J
our
nal
o
f
E
le
c
tr
ic
al
E
ngi
ne
e
r
in
g
and
I
nf
or
m
at
ic
s
,
v
o
l.
11,
no
.
3,
pp.
764
–
777,
S
e
p.
2023,
do
i:
10.52549/i
je
e
i.
v
11i
3.4829.
[
5]
S
.
S
in
gh,
A
.
S
ha
r
ma
,
a
nd
V
.
K
.
C
ha
uh
a
n,
“
O
nl
in
e
ha
ndw
r
it
te
n
G
ur
mukhi
w
or
d
r
e
c
o
gni
t
i
o
n
us
in
g
f
in
e
-
tu
n
e
d
D
e
e
p
C
o
n
vo
lu
ti
o
na
l
N
e
ur
a
l
N
e
tw
o
r
k
o
n
of
f
l
in
e
f
e
a
tu
r
e
s
,”
M
ac
hi
ne
L
e
ar
ni
ng
w
it
h
A
ppl
ic
at
io
ns
,
vo
l.
5,
p.
100
037,
S
e
p.
2021,
do
i:
10.1016/j
.ml
w
a
.2021.100037.
[
6]
S
. D
. P
a
nde
e
t
al
.
, “
D
ig
it
iz
a
ti
o
n
of
ha
ndw
r
it
t
e
n D
e
v
a
na
ga
r
i
te
x
t
us
in
g C
N
N
t
r
a
ns
f
e
r
l
e
a
r
ni
ng
–
A
b
e
tt
e
r
c
us
t
o
m
e
r
s
e
r
v
ic
e
s
upp
or
t,
”
N
e
ur
os
c
ie
nc
e
I
n
f
or
m
at
ic
s
, v
o
l.
2, n
o
. 3, p. 100016, S
e
p. 2022, d
o
i:
10.1
016/
j.
n
e
u
r
i.
2021.100016.
[
7]
R
.
G
h
o
s
h,
C
.
V
a
ms
hi
,
a
nd
P
.
K
u
ma
r
,
“
R
N
N
ba
s
e
d
o
n
li
ne
ha
ndw
r
it
te
n
w
or
d
r
e
c
o
gni
ti
o
n
in
D
e
v
a
na
ga
r
i
a
nd
B
e
nga
li
s
c
r
ip
ts
us
in
g
ho
r
i
z
o
nt
a
l
z
o
ni
ng,”
P
at
te
r
n R
e
c
ogni
ti
on
,
vo
l.
92, pp. 203
–
218,
A
ug. 2019, do
i:
10.1016/j
.pa
t
c
o
g.2019.03.0
30.
[
8]
R
.
S
a
r
khe
l,
N
.
D
a
s
,
A
.
D
a
s
,
M
.
K
undu,
a
nd
M
.
N
a
s
ip
ur
i,
“
A
mul
ti
-
s
c
a
l
e
de
e
p
qua
d
tr
e
e
ba
s
e
d
f
e
a
tu
r
e
e
x
tr
a
c
ti
o
n
me
th
o
d
f
o
r
th
e
r
e
c
o
gni
ti
o
n
of
is
o
la
t
e
d
ha
ndw
r
it
t
e
n
c
ha
r
a
c
t
e
r
s
of
p
o
pu
la
r
in
di
c
s
c
r
ip
ts
,”
P
at
te
r
n
R
e
c
ogni
ti
on
,
vo
l.
71,
pp.
78
–
93,
N
ov
.
201
7,
do
i:
10.1016/j
.pa
tc
o
g.2017.05.022.
[
9]
S
. P
. D
e
or
e
a
nd A
. P
r
a
v
in
, “
D
e
v
a
na
ga
r
i
H
a
ndw
r
it
t
e
n C
ha
r
a
c
t
e
r
R
e
c
o
gni
t
i
o
n us
in
g
f
in
e
-
tu
n
e
d D
e
e
p C
o
n
vo
lu
ti
o
na
l
N
e
ur
a
l
N
e
tw
o
r
k
o
n
tr
i
v
ia
l
da
ta
s
e
t,
”
Sadhana
-
A
c
ade
m
y
P
r
oc
e
e
di
ngs
in
E
ngi
ne
e
r
in
g
Sc
ie
nc
e
s
,
vo
l.
45,
n
o
.
1
,
p.
243,
D
e
c
.
2020,
do
i:
10.1007/s
12046
-
020
-
01484
-
1.
[
10]
A
.
B
ha
r
dw
a
j,
“
A
n
A
c
c
ur
a
te
a
nd
F
in
e
-
tu
n
e
d
D
e
e
p
-
L
e
a
r
ni
ng
M
o
d
e
l
f
o
r
H
a
ndw
r
it
t
e
n
D
e
v
a
na
ga
r
i
C
ha
r
a
c
te
r
R
e
c
o
gni
ti
o
n,”
vo
l.
20,
no
. 6, pp. 6717
–
6736, 2022.
[
11]
N
.
A
ne
ja
a
nd
S
.
A
n
e
ja
,
“
T
r
a
ns
f
e
r
L
e
a
r
ni
ng
u
s
in
g
C
N
N
f
o
r
H
a
ndw
r
it
t
e
n
D
e
v
a
na
ga
r
i
C
ha
r
a
c
t
e
r
R
e
c
o
gni
ti
o
n,”
1s
t
I
E
E
E
I
nt
e
r
nat
io
nal
C
onf
e
r
e
nc
e
on
A
dv
anc
e
s
in
I
nf
or
m
at
io
n
T
e
c
hnol
ogy
,
I
C
A
I
T
2019
-
P
r
oc
e
e
di
ngs
,
pp.
293
–
296,
2019,
do
i:
10.1109/I
C
A
I
T
47043.2019.8987286.
[
12]
V
.
M
.
L
o
mt
e
a
nd
D
.
D
.
D
oy
e
,
“
D
e
v
a
na
ga
r
i
T
e
x
t
a
nd
C
a
ll
ig
r
a
ph
y
R
e
c
o
gni
ti
o
n
U
s
in
g
I
C
F
&
A
C
F
,”
C
om
put
e
r
I
nt
e
gr
at
e
d
M
anuf
ac
tu
r
in
g Sy
s
te
m
s
, v
o
l.
29, n
o
. 1, pp. 88
–
114, 2023, d
o
i:
1
0.24297/j
.c
im
s
.2023.1.7.
[
13]
M
. V
in
a
, M
. L
omt
e
,
a
nd D
. D
. D
oy
e
, “
H
a
ndw
r
it
t
e
n V
e
di
c
S
a
ns
kr
it
T
e
x
t
R
e
c
o
gni
ti
o
n U
s
in
g D
e
e
p
L
e
a
r
n
in
g,”
J
our
nal
of
A
lg
e
br
ai
c
St
at
is
ti
c
s
, v
ol
. 13, n
o
. 3, pp. 2190
–
2198, 2022, [
O
nl
in
e
]
. A
v
a
il
a
bl
e
:
ht
tp
s
:/
/p
ubl
is
h
o
a
.c
om
[
14]
V
.
M
.
L
o
mt
e
a
nd
D
.
D
oy
e
,
“
G
e
s
tu
r
e
ba
s
e
d
D
e
v
a
na
ga
r
i
T
e
x
t
R
e
c
o
gni
ti
o
n
us
in
g
D
e
e
p
L
e
a
r
ni
ng,”
H
ar
bi
n
G
ongy
e
D
ax
ue
X
ue
bao/
J
our
na
l
of
H
a
r
bi
n I
ns
ti
tu
te
of
T
e
c
hnol
ogy
, v
o
l.
54, n
o
.
6, pp. 239
–
247, 2022.
[
15]
S
.
K
.
M
a
noc
ha
a
nd
P
.
T
e
w
a
r
i,
“
D
e
v
a
na
ga
r
i
H
a
ndw
r
it
t
e
n
C
ha
r
a
c
te
r
R
e
c
o
gni
t
i
o
n
us
in
g
C
N
N
a
s
F
e
a
tu
r
e
E
x
tr
a
c
t
o
r
,”
in
2
021
I
nt
e
r
nat
io
nal
C
onf
e
r
e
nc
e
on
Sm
ar
t
G
e
n
e
r
at
io
n
C
om
put
in
g,
C
o
m
m
uni
c
at
io
n
and
N
e
tw
or
k
in
g,
SM
A
R
T
G
E
N
C
O
N
2021
,
O
c
t.
2
021,
pp. 1
–
5. do
i:
10.1109/
S
M
A
R
T
G
E
N
C
O
N
51891.2021.9645786.
[
16]
S
.
V
.
S
he
lk
e
,
D
.
M
.
C
ha
ndw
a
dka
r
,
S
.
P
.
U
ga
le
,
a
nd
R
.
V
.
C
ho
th
e
,
“
C
ombi
ni
ng
M
ul
ti
pl
e
F
e
a
tu
r
e
E
x
tr
a
c
ti
o
n
a
nd
C
la
s
s
if
i
c
a
ti
o
n
M
e
th
o
ds
t
o
S
tu
d
y
P
e
r
f
or
ma
nc
e
of
H
a
ndw
r
i
tt
e
n
S
a
ns
kr
it
C
ha
r
a
c
te
r
R
e
c
o
gni
ti
o
n,”
in
2023
7
th
I
nt
e
r
nat
io
nal
C
onf
e
r
e
n
c
e
on
C
om
put
in
g,
C
om
m
uni
c
at
io
n,
C
ont
r
ol
and
A
ut
om
at
i
on,
I
C
C
U
B
E
A
2023
,
A
ug.
2023,
pp.
1
–
6.
do
i:
10.1109/I
C
C
U
B
E
A
58933.2023.10391986.
[
17]
S
.
A
c
ha
r
y
a
,
A
.
K
.
P
a
nt
,
a
nd
P
.
K
.
G
y
a
w
a
li
,
“
D
e
e
p
l
e
a
r
ni
ng
b
a
s
e
d
la
r
ge
s
c
a
l
e
ha
ndw
r
it
t
e
n
D
e
v
a
na
ga
r
i
c
ha
r
a
c
t
e
r
r
e
c
o
gni
t
i
o
n,
”
in
2015
9t
h
I
nt
e
r
nat
io
nal
C
on
f
e
r
e
nc
e
on
Sof
tw
ar
e
,
K
now
le
dge
,
I
n
f
or
m
at
io
n
M
anage
m
e
nt
and
A
ppl
ic
at
io
ns
(
SK
I
M
A
)
,
D
e
c
.
2015,
pp.
1
–
6. do
i:
10.1109/
S
K
I
M
A
.2015.7400041.
[
18]
M
. W
ul
a
nda
r
i,
R
. C
ha
i,
B
.
B
a
s
a
r
i,
a
nd D
.
G
una
w
a
n, “
H
y
br
id
F
e
a
tu
r
e
E
x
tr
a
c
t
o
r
U
s
in
g D
is
c
r
e
t
e
W
a
v
e
le
t
T
r
a
ns
f
or
m a
nd
H
is
t
o
g
r
a
m
of
O
r
i
e
nt
e
d
G
r
a
di
e
n
t
o
n
C
o
n
vol
ut
i
o
na
l
-
N
e
ur
a
l
-
N
e
tw
o
r
k
-
B
a
s
e
d
P
a
lm
V
e
in
R
e
c
o
gni
ti
o
n,”
Se
ns
or
s
,
v
o
l.
24,
no
.
2,
p.
341,
J
a
n.
2024, do
i:
10.3390/s
24020341.
[
19]
S
.
S
he
lk
e
,
“
H
a
ndw
r
it
t
e
n
C
ha
r
a
c
t
e
r
R
e
c
o
gni
ti
o
n
us
in
g
W
a
v
e
le
t
T
r
a
ns
f
o
r
m
f
or
F
e
a
tu
r
e
E
x
t
r
a
c
ti
o
n,”
I
nt
e
r
nat
io
nal
J
our
nal
o
f
M
ul
ti
di
s
c
ip
li
nar
y
E
duc
at
io
nal
R
e
s
e
ar
c
h (
I
J
M
E
R
)
,
vo
l.
3, n
o
. M
a
r
c
h 2014, pp. 3
–
7,
2016.
[
20]
N
.
A
z
a
w
i,
“
H
a
ndw
r
it
te
n
di
gi
ts
r
e
c
o
gni
ti
o
n
us
in
g
tr
a
ns
f
e
r
l
e
a
r
n
in
g,”
C
om
put
e
r
s
and
E
le
c
tr
ic
al
E
ngi
ne
e
r
in
g
,
v
o
l.
106,
p.
108604,
M
a
r
. 2023, do
i:
10.1016/j
.
c
o
mp
e
l
e
c
e
ng.2023.108604.
[
21]
P
.
P
.
N
a
ir
,
A
. J
a
me
s
,
P
.
S
im
o
n,
a
nd
P
. V
.
B
ha
gy
a
s
r
e
e
,
“
M
a
la
y
a
la
m
H
a
ndw
r
it
te
n
C
ha
r
a
c
t
e
r
R
e
c
o
gni
ti
o
n
us
in
g
C
N
N
A
r
c
hi
te
c
tu
r
e
,”
I
ndone
s
ia
n
J
our
nal
o
f
E
le
c
tr
ic
al
E
ngi
ne
e
r
in
g
and
I
nf
or
m
at
ic
s
,
v
o
l.
11,
no
.
3,
pp.
764
–
777,
S
e
p.
2023,
do
i:
10.52549/i
je
e
i.
v
11i
3.4829.
[
22]
C
.
S
z
e
ge
d
y
e
t
al
.
,
“
G
o
in
g
D
e
e
p
e
r
w
it
h
C
o
n
vo
lu
ti
o
ns
,”
i
n
P
r
o
c
e
e
di
ngs
o
f
th
e
I
E
E
E
c
onf
e
r
e
nc
e
on
c
om
put
e
r
v
is
io
n
and
pat
te
r
n
r
e
c
ogni
ti
on
, 2015, pp. 1
–
9. do
i:
10.48550/ar
X
i
v
.1409.4842.
[
23]
Z
.
Z
h
o
ng,
L
.
J
in
,
a
nd
Z
.
X
ie
,
“
H
ig
h
p
e
r
f
or
ma
nc
e
of
f
li
ne
h
a
ndw
r
it
te
n
C
hi
ne
s
e
c
ha
r
a
c
t
e
r
r
e
c
o
gni
ti
o
n
us
in
g
G
oo
g
L
e
N
e
t
a
nd
di
r
e
c
t
i
o
na
l
f
e
a
tu
r
e
ma
ps
,”
in
P
r
oc
e
e
di
ngs
o
f
th
e
I
nt
e
r
nat
io
nal
C
onf
e
r
e
nc
e
on
D
oc
um
e
nt
A
nal
y
s
i
s
and
R
e
c
ogni
ti
on,
I
C
D
A
R
,
A
ug.
2015, vo
l.
2015
-
N
ove
mb
e
r
, pp. 846
–
850. d
o
i:
10.1109
/I
C
D
A
R
.2015.7333881.
[
24]
M
.
M
is
hr
a
,
T
.
C
ho
udhur
y
,
a
nd
T
.
S
a
r
ka
r
,
“
D
e
v
a
na
ga
r
i
H
a
ndw
r
it
t
e
n
C
ha
r
a
c
t
e
r
R
e
c
o
gni
ti
o
n,”
2021
I
E
E
E
I
ndi
a
C
ou
nc
il
I
nt
e
r
nat
io
nal
Subs
e
c
ti
ons
C
on
f
e
r
e
nc
e
, I
N
D
I
SC
O
N
2021
, 2021,
do
i:
10.1109/
I
N
D
I
S
C
O
N
53343.2021.9582192.
[
25]
L
.
L
iu
a
nd
X
.
L
u
o
,
“
A
N
e
w
A
c
c
e
l
e
r
a
te
d
S
t
o
c
ha
s
ti
c
G
r
a
di
e
n
t
M
e
th
o
d
w
it
h
M
o
m
e
nt
um,”
in
P
r
oc
e
e
di
ngs
o
f
M
ac
hi
ne
L
e
ar
ni
ng
R
e
s
e
ar
c
h
, 2020, pp. 1
–
10. [
O
nl
in
e
]
. A
v
a
il
a
bl
e
:
ht
tp
:/
/a
r
x
i
v
.
o
r
g/
a
bs
/2
006.00423
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
do
n
e
s
i
a
n
J
E
l
e
c
E
n
g
&
C
o
m
p
S
c
i
I
S
S
N:
2
5
0
2
-
4
7
52
Dis
c
r
e
te
w
av
e
let
tr
ans
f
or
m
and
c
onv
olut
ional
ne
ur
al
ne
t
w
or
k
bas
e
d
…
(
Shr
addha
V
.
She
lke
)
1375
B
I
OG
RA
P
HI
E
S
OF
AU
T
HO
RS
Sh
ra
dd
h
a
V
.
Sh
e
l
k
e
i
s
w
o
r
k
i
n
g
as
an
A
s
s
i
s
t
an
t
Pro
fe
s
s
o
r
i
n
E
l
ec
t
ro
n
i
c
s
an
d
T
e
l
ec
o
mmu
n
i
c
at
i
o
n
E
n
g
i
n
ee
ri
n
g
D
e
p
art
me
n
t
o
f
K
.
K
.
W
ag
h
In
s
t
i
t
u
t
e
o
f
E
n
g
i
n
ee
r
i
n
g
E
d
u
c
at
i
o
n
an
d
r
e
s
e
ar
c
h
,
N
as
h
i
k
,
Mah
aras
h
t
ra
s
i
n
ce
l
as
t
1
3
y
e
ars
.
Sh
e
i
s
p
e
ru
s
i
n
g
h
e
r
P
h
.
D
.
i
n
E
l
ec
t
ro
n
i
c
s
an
d
T
e
l
eco
mm
u
n
i
c
at
i
o
n
fro
m
Sav
i
t
ri
b
ai
Ph
u
l
e
P
u
n
e
U
n
i
v
e
rs
i
t
y
.
S
h
e
c
a
n
b
e
c
o
n
t
ac
t
e
d
at
em
a
i
l
:
s
v
s
h
el
k
e
@
k
k
w
ag
h
.
ed
u
.
i
n
.
D
i
n
es
h
M
.
C
h
a
n
dw
a
dk
a
r
i
s
a
p
ro
fe
s
s
o
r
an
d
h
e
a
d
o
f
E
&
T
C
D
e
p
art
me
n
t
at
K
.
K
.
W
ag
h
I
n
s
t
i
t
u
t
e
o
f
E
n
g
i
n
ee
ri
n
g
E
d
u
c
at
i
o
n
&
R
e
s
e
ar
c
h
,
N
as
h
i
k
,
In
d
i
a.
H
i
s
ar
e
a
o
f
i
n
t
e
r
e
s
t
i
n
c
l
u
d
e
s
S
i
g
n
al
Pr
o
ce
s
s
i
n
g
,
Po
w
er
E
l
ec
t
ro
n
i
c
s
,
Mech
at
ro
n
i
c
s
,
a
n
d
A
u
t
o
mo
t
i
v
e
E
l
ec
t
ro
n
i
c
s
.
H
e
h
o
l
d
s
Ph
.
D
.
d
eg
r
ee
an
d
h
e
h
as
p
u
b
l
i
s
h
ed
Mo
r
e
t
h
an
5
0
re
s
e
ar
ch
p
ap
e
rs
i
n
r
e
p
u
t
ed
J
o
u
r
n
al
s
.
H
e
i
s
w
o
rk
i
n
g
as
B
o
ard
o
f
St
u
d
i
e
s
mem
b
e
r
o
f
E
l
ec
t
ro
n
i
c
s
an
d
T
el
ec
o
mmu
n
i
c
at
i
o
n
E
n
g
i
n
ee
ri
n
g
fo
r
Pu
n
e
U
n
i
v
e
rs
i
t
y
.
H
e
c
a
n
b
e
c
o
n
t
ac
t
ed
at
em
ai
l
:
d
mch
an
d
w
a
d
k
ar
@
k
k
w
a
g
h
.
e
d
u
.
i
n
.
Suni
ta
P
.
U
g
a
l
e
i
s
w
o
r
k
i
n
g
as
a
Pr
o
f
e
s
s
o
r
i
n
E
l
ec
t
ro
n
i
c
s
an
d
T
e
l
eco
mm
u
n
i
c
at
i
o
n
E
n
g
i
n
ee
r
i
n
g
D
e
p
art
men
t
o
f
K
.
K
.
W
ag
h
I
n
s
t
i
t
u
t
e
o
f
E
n
g
i
n
ee
r
i
n
g
E
d
u
c
at
i
o
n
an
d
R
e
s
e
ar
ch
,
N
as
h
i
k
,
Mah
aras
h
t
ra
s
i
n
ce
l
as
t
2
8
y
e
ars
.
S
h
e
h
o
l
d
s
P
h
.
D
.
d
eg
r
ee
an
d
h
e
r
s
p
ec
i
a
l
fi
el
d
s
o
f
i
n
t
e
r
e
s
t
i
n
cl
u
d
e
Fi
b
e
r
O
p
t
i
c
s
Co
mmu
n
i
c
at
i
o
n
,
O
p
t
i
c
al
S
en
s
o
rs
,
A
u
t
o
m
at
i
o
n
a
n
d
V
L
S
I
t
e
ch
n
o
l
o
g
y
.
S
h
e
c
a
n
b
e
c
o
n
t
ac
t
e
d
at
em
ai
l
:
s
p
u
g
al
e@
k
k
w
a
g
h
.
e
d
u
.
i
n
.
R
u
pa
l
i
V
.
C
h
o
th
e
i
s
w
o
rk
i
n
g
as
a
n
A
s
s
i
s
t
an
t
Pro
fe
s
s
o
r
i
n
E
l
ec
t
ro
n
i
c
s
a
n
d
T
e
l
ec
o
mmu
n
i
c
at
i
o
n
E
n
g
i
n
ee
ri
n
g
D
e
p
art
me
n
t
o
f
K
.
K
.
W
ag
h
In
s
t
i
t
u
t
e
o
f
E
n
g
i
n
ee
r
i
n
g
E
d
u
c
at
i
o
n
an
d
R
e
s
e
arch
,
N
as
h
i
k
,
Mah
aras
h
t
ra
s
i
n
ce
l
a
s
t
1
6
y
e
ars
.
S
h
e
i
s
p
u
rs
u
i
n
g
P
h
.
D
.
i
n
E
l
ec
t
ro
n
i
c
s
an
d
T
e
l
eco
mm
u
n
i
c
at
i
o
n
fro
m
Sav
i
t
ri
b
ai
Ph
u
l
e
P
u
n
e
U
n
i
v
e
rs
i
t
y
.
S
h
e
c
a
n
b
e
c
o
n
t
ac
t
e
d
at
em
a
i
l
:
r
v
c
h
o
t
h
e
@
k
k
w
a
g
h
.
e
d
u
.
i
n
.
Evaluation Warning : The document was created with Spire.PDF for Python.