I
AE
S In
t
er
na
t
io
na
l J
o
urna
l o
f
Art
if
icia
l In
t
ellig
ence
(
I
J
-
AI
)
Vo
l.
1
5
,
No
.
3
,
J
u
n
e
2
0
2
6
,
p
p
.
2655
~
2
6
6
3
I
SS
N:
2
2
5
2
-
8
9
3
8
,
DOI
: 1
0
.
1
1
5
9
1
/ijai.v
15
.i
3
.
p
p
2
6
5
5
-
2
6
6
3
2655
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
a
i
.
ia
esco
r
e.
co
m
Visio
n t
ra
nsfo
rm
er and hy
brid
mo
dels for Ma
la
y
a
la
m
ha
ndwritt
en wor
d recog
nition
Anj
u Ar
a
ng
il T
ha
zha
t
h
1,
2
,
B
inu
P
o
o
t
ha
k
uzhiy
il
Cha
ck
o
3
,
M
o
ha
m
ed
B
a
s
heer
K
izha
k
ke
P
a
ra
m
ba
t
h
4
1
D
e
p
a
r
t
me
n
t
o
f
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
,
S
u
l
l
a
mu
ss
a
l
a
m Sc
i
e
n
c
e
C
o
l
l
e
g
e
,
U
n
i
v
e
r
si
t
y
o
f
C
a
l
i
c
u
t
,
K
e
r
a
l
a
,
I
n
d
i
a
2
D
e
p
a
r
t
me
n
t
o
f
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
,
F
a
c
u
l
t
y
o
f
D
a
t
a
S
c
i
e
n
c
e
a
n
d
A
n
a
l
y
t
i
c
s,
U
n
i
v
e
r
si
t
y
o
f
C
a
l
i
c
u
t
,
K
e
r
a
l
a
,
I
n
d
i
a
3
D
e
p
a
r
t
me
n
t
o
f
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
,
P
r
a
j
y
o
t
i
N
i
k
e
t
a
n
C
o
l
l
e
g
e
,
K
e
r
a
l
a
,
I
n
d
i
a
4
D
e
p
a
r
t
me
n
t
o
f
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
,
A
mal
C
o
l
l
e
g
e
o
f
A
d
v
a
n
c
e
d
S
t
u
d
i
e
s,
K
e
r
a
l
a
,
I
n
d
i
a
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Oct
1
2
,
2
0
2
4
R
ev
is
ed
Ap
r
1
,
2
0
2
6
Acc
ep
ted
Ap
r
2
1
,
2
0
2
6
Tran
sfo
rm
e
r
-
b
a
se
d
a
rc
h
it
e
c
tu
re
s
a
n
d
a
tt
e
n
ti
o
n
m
e
c
h
a
n
is
m
s
h
a
v
e
re
v
o
lu
ti
o
n
ize
d
th
e
field
o
f
ima
g
e
re
c
o
g
n
it
i
o
n
.
T
h
is
stu
d
y
fo
c
u
se
s
o
n
o
ffl
in
e
h
a
n
d
writ
ten
M
a
la
y
a
lam
wo
rd
re
c
o
g
n
i
ti
o
n
,
a
d
d
re
ss
in
g
t
h
e
lac
k
o
f
p
u
b
li
c
l
y
a
v
a
il
a
b
le
d
a
tas
e
ts
fo
r
th
is
l
o
w
-
re
so
u
rc
e
lan
g
u
a
g
e
.
A
n
e
w
M
a
lay
a
lam
wo
rd
d
a
tas
e
t
(M
WD)
c
o
m
p
risin
g
2
0
,
8
5
0
sa
m
p
les
a
c
ro
ss
1
3
9
c
las
se
s
w
a
s
d
e
v
e
lo
p
e
d
to
su
p
p
o
r
t
re
se
a
rc
h
i
n
th
is
d
o
m
a
in
.
T
h
e
v
isio
n
t
ra
n
sfo
r
m
e
r
(ViT)
wa
s
e
m
p
lo
y
e
d
fo
r
a
d
v
a
n
c
e
d
fe
a
tu
re
e
x
trac
ti
o
n
,
a
n
d
m
u
l
ti
p
le
r
e
c
o
g
n
it
i
o
n
m
o
d
e
ls
—
fe
e
d
-
fo
rwa
rd
n
e
u
ra
l
n
e
two
rk
(
F
F
NN
),
g
l
o
b
a
l
a
v
e
ra
g
e
p
o
o
li
n
g
(G
AP),
b
id
irec
ti
o
n
a
l
l
o
n
g
sh
o
r
t
-
term
m
e
m
o
ry
(B
i
LS
TM
)
,
a
n
d
a
tt
e
n
ti
o
n
-
b
a
se
d
fe
e
d
-
fo
rwa
rd
n
e
u
ra
l
n
e
t
wo
rk
(AF
F
NN
)
—
we
re
e
v
a
lu
a
ted
.
Am
o
n
g
th
e
se
,
AFF
NN
a
c
h
iev
e
d
th
e
h
i
g
h
e
st
a
c
c
u
ra
c
y
o
f
9
8
.
5
6
%
,
e
sta
b
l
ish
in
g
t
h
e
p
ro
p
o
se
d
v
isio
n
tra
n
sfo
rm
e
r
-
b
a
se
d
a
tt
e
n
ti
o
n
h
a
n
d
writt
e
n
w
o
rd
r
e
c
o
g
n
it
i
o
n
(
ViTA
-
HWR
)
m
o
d
e
l
a
s
a
ro
b
u
st
fra
m
e
wo
rk
fo
r
h
a
n
d
writt
e
n
M
a
lay
a
lam
wo
rd
re
c
o
g
n
it
io
n
a
nd
v
a
l
u
a
b
le co
n
tri
b
u
ti
o
n
to
re
g
io
n
a
l
lan
g
u
a
g
e
p
r
o
c
e
ss
in
g
.
K
ey
w
o
r
d
s
:
Atten
tio
n
m
ec
h
an
is
m
Feed
-
f
o
r
war
d
n
eu
r
al
n
etwo
r
k
Han
d
wr
itten
wo
r
d
r
ec
o
g
n
itio
n
Ma
lay
alam
wo
r
d
d
ataset
Vis
io
n
tr
an
s
f
o
r
m
er
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
:
An
ju
Ar
an
g
il T
h
az
h
ath
Dep
ar
tm
en
t o
f
C
o
m
p
u
ter
Scie
n
ce
,
Su
llam
u
s
s
alam
Scien
ce
C
o
lleg
e,
Un
iv
er
s
ity
o
f
C
alicu
t
Ar
ee
k
o
d
e,
Ma
la
p
p
u
r
a
m
,
Ker
al
a,
I
n
d
ia
E
m
ail: a
n
ju
at@
s
s
co
lleg
e.
ac
.
in
1.
I
NT
RO
D
UCT
I
O
N
Op
tical
ch
ar
ac
ter
r
ec
o
g
n
itio
n
(
OC
R
)
h
as
lo
n
g
s
er
v
ed
as
a
v
ital
tech
n
o
lo
g
y
f
o
r
c
o
n
v
e
r
tin
g
p
r
in
ted
o
r
m
ac
h
in
e
-
g
e
n
er
ated
te
x
t
in
to
d
i
g
ital
f
o
r
m
ats.
Ho
wev
e
r
,
as
th
e
n
ee
d
f
o
r
d
ig
itizatio
n
e
x
p
an
d
e
d
,
atten
tio
n
s
h
if
ted
to
war
d
o
f
f
lin
e
h
an
d
wr
itten
wo
r
d
r
ec
o
g
n
itio
n
(
HW
R
)
,
a
ch
allen
g
in
g
s
u
b
s
et
o
f
OC
R
th
at
d
ea
ls
with
h
an
d
wr
itten
tex
t.
Un
lik
e
p
r
i
n
ted
ch
ar
ac
te
r
s
,
h
an
d
wr
itten
wo
r
d
s
ex
h
ib
it
s
u
b
s
tan
tial
v
ar
iab
ilit
y
d
u
e
to
d
if
f
er
en
ce
s
in
in
d
i
v
id
u
al
wr
iti
n
g
s
ty
les,
p
r
ess
u
r
e,
s
tr
o
k
e
wi
d
th
,
an
d
ch
a
r
ac
ter
c
o
n
n
ec
tiv
it
y
.
T
h
is
co
m
p
lex
ity
in
ten
s
if
ies
f
o
r
s
cr
ip
ts
s
u
ch
as
Ma
lay
alam
,
a
Dr
av
id
ian
lan
g
u
ag
e
s
p
o
k
en
p
r
ed
o
m
in
an
tly
in
th
e
I
n
d
ian
s
tate
o
f
Ker
ala,
k
n
o
w
n
f
o
r
its
co
m
p
o
u
n
d
ch
a
r
ac
ter
s
,
r
o
u
n
d
e
d
s
h
ap
es,
an
d
i
n
tr
icate
lig
atu
r
es.
HW
R
p
lay
s
a
cr
u
cial
r
o
le
in
ap
p
licatio
n
s
s
u
ch
as
d
o
cu
m
en
t
d
ig
itizatio
n
,
in
f
o
r
m
a
tio
n
r
etr
iev
al
f
r
o
m
h
a
n
d
wr
i
tten
r
ec
o
r
d
s
,
a
n
d
p
r
eser
v
atio
n
o
f
h
is
to
r
ical
m
an
u
s
cr
ip
ts
,
y
et
its
ef
f
ec
tiv
en
ess
i
s
o
f
ten
h
in
d
er
e
d
b
y
s
cr
ip
t
-
s
p
e
cif
ic
ch
allen
g
es
an
d
th
e
ab
s
en
ce
o
f
r
o
b
u
s
t a
n
n
o
tate
d
d
atasets
.
De
s
p
ite
r
e
m
ar
k
ab
l
e
p
r
o
g
r
e
s
s
in
O
C
R
an
d
h
an
d
wr
it
ten
tex
t
r
ec
o
g
n
i
tio
n
ac
r
o
s
s
l
an
g
u
ag
e
s
s
u
ch
a
s
E
n
g
l
is
h
,
A
r
ab
ic,
an
d
C
h
in
e
s
e
,
Ma
lay
a
la
m
r
e
m
ai
n
s
a
l
o
w
-
r
e
s
o
u
r
ce
lan
g
u
ag
e,
w
ith
li
m
i
ted
p
u
b
li
cly
av
a
i
lab
le
d
ata
s
et
s
f
o
r
h
an
d
wr
i
tt
en
wo
r
d
s
.
T
h
e
ab
s
en
ce
o
f
s
u
c
h
d
a
ta
s
et
s
im
p
e
d
e
s
th
e
d
ev
elo
p
m
en
t
an
d
b
en
ch
m
ar
k
i
n
g
o
f
r
el
ia
b
le
r
e
co
g
n
i
t
io
n
m
o
d
e
ls
.
F
u
r
th
er
m
o
r
e,
Ma
lay
al
am
h
an
d
wr
i
tt
en
wo
r
d
s
o
f
t
en
v
a
r
y
in
s
p
a
cin
g
an
d
cu
r
v
a
tu
r
e
,
l
ea
d
in
g
to
s
eg
m
en
ta
tio
n
d
if
f
icu
lt
ie
s
an
d
r
e
co
g
n
i
tio
n
er
r
o
r
s
.
T
h
er
ef
o
r
e
,
d
ev
elo
p
in
g
a
s
p
e
ci
al
iz
e
d
d
ata
s
et
an
d
an
ef
f
i
ci
en
t
r
e
co
g
n
it
io
n
f
r
am
ewo
r
k
i
s
cr
it
ic
al
f
o
r
ad
v
a
n
c
in
g
r
es
ea
r
ch
in
Ma
l
ay
al
am
h
an
d
wr
i
t
ten
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
1
5
,
No
.
3
,
J
u
n
e
2
0
2
6
:
2
6
5
5
-
2
6
6
3
2656
tex
t
p
r
o
c
e
s
s
in
g
[
1
]
.
E
ar
li
er
wo
r
k
s
i
n
H
W
R
p
r
im
ar
ily
r
el
i
ed
o
n
h
an
d
cr
a
f
t
ed
f
e
atu
r
e
ex
tr
a
ct
io
n
t
ec
h
n
iq
u
es
,
wh
er
e
g
e
o
m
e
tr
ic,
s
ta
ti
s
ti
ca
l,
an
d
t
ex
tu
r
e
-
b
a
s
ed
f
ea
tu
r
e
s
w
e
r
e
u
s
ed
in
co
m
b
in
a
t
io
n
wi
th
t
r
ad
i
tio
n
a
l
m
a
ch
in
e
lea
r
n
in
g
c
la
s
s
if
i
er
s
[
2
]
–
[
7
]
.
T
h
e
ad
v
en
t
o
f
d
e
ep
le
ar
n
in
g
in
t
r
o
d
u
c
ed
p
o
we
r
f
u
l
a
r
ch
i
tec
tu
r
e
s
s
u
ch
a
s
co
n
v
o
lu
t
io
n
a
l
n
e
u
r
a
l
n
e
t
wo
r
k
s
(
C
NN
s
)
,
r
e
cu
r
r
en
t
n
eu
r
al
n
et
wo
r
k
s
(
R
NN
s
)
,
an
d
lo
n
g
s
h
o
r
t
-
ter
m
m
e
m
o
r
y
(
L
S
T
M)
n
e
two
r
k
s
,
wh
i
ch
s
ig
n
if
i
ca
n
tl
y
i
m
p
r
o
v
e
d
r
e
co
g
n
it
i
o
n
ac
cu
r
a
cy
b
y
au
to
m
a
ti
ca
ll
y
lea
r
n
in
g
s
p
a
ti
al
an
d
s
eq
u
en
t
ia
l
d
ep
en
d
en
ci
e
s
[
8
]
–
[
1
8
]
.
Mo
r
e
r
ec
en
tly
,
tr
an
s
f
o
r
m
er
-
b
a
s
ed
ar
ch
i
te
ctu
r
e
s
h
av
e
r
ev
o
lu
tio
n
iz
ed
th
e
f
ie
ld
b
y
c
ap
t
u
r
in
g
lo
n
g
-
r
an
g
e
d
ep
en
d
en
ci
es
an
d
e
n
ab
lin
g
p
ar
a
ll
el
ize
d
f
ea
tu
r
e
co
m
p
u
t
at
io
n
.
Stu
d
ie
s
s
u
ch
a
s
tr
an
s
f
o
r
m
er
-
b
a
s
ed
o
p
t
ica
l
c
h
ar
ac
te
r
r
e
co
g
n
i
t
io
n
(
T
r
O
C
R
)
f
o
r
E
n
g
li
s
h
h
an
d
wr
i
tt
en
t
ex
t
[
1
9
]
an
d
o
p
t
ic
al
ch
ar
a
ct
er
f
o
r
m
er
(
O
C
Fo
r
m
er
)
f
o
r
A
r
ab
ic
h
an
d
wr
it
in
g
h
a
v
e
d
em
o
n
s
t
r
a
ted
th
e
s
u
p
er
i
o
r
p
er
f
o
r
m
an
ce
o
f
tr
an
s
f
o
r
m
er
-
b
a
s
ed
m
o
d
e
l
s
in
r
ec
o
g
n
iz
in
g
d
iv
e
r
s
e
h
an
d
wr
it
in
g
s
t
y
le
s
[
1
8
]
,
[
2
0
]
,
[
2
1
]
.
I
n
th
e
co
n
tex
t
o
f
Ma
l
ay
a
la
m
,
ex
i
s
tin
g
s
t
u
d
i
es
r
em
a
in
li
m
i
ted
to
C
NN
-
b
a
s
e
d
f
e
atu
r
e
ex
tr
ac
ti
o
n
c
o
u
p
l
ed
w
ith
co
n
v
en
t
io
n
al
cla
s
s
if
i
er
s
s
u
ch
a
s
s
u
p
p
o
r
t
v
ec
to
r
m
ac
h
in
e
(
S
VM
)
o
r
f
ee
d
-
f
o
r
w
ar
d
n
eu
r
a
l
n
et
wo
r
k
s
(
F
F
NN)
[
1
5
]
,
h
ig
h
l
ig
h
tin
g
a
s
ig
n
if
i
ca
n
t g
ap
f
o
r
ad
v
an
c
ed
t
r
an
s
f
o
r
m
er
-
b
a
s
ed
ap
p
r
o
ac
h
e
s
.
I
n
r
esp
o
n
s
e
to
th
ese
lim
itati
o
n
s
,
th
e
p
r
esen
t
s
tu
d
y
in
tr
o
d
u
ce
s
a
co
m
p
r
eh
en
s
iv
e
Ma
lay
alam
wo
r
d
d
ataset
(
MWD)
co
m
p
r
is
in
g
2
0
,
8
5
0
h
an
d
wr
itten
wo
r
d
s
am
p
les
ac
r
o
s
s
1
3
9
class
es,
ea
ch
co
n
tain
in
g
1
5
0
b
alan
ce
d
s
am
p
les.
Fu
r
t
h
er
m
o
r
e
,
a
n
o
v
el
r
ec
o
g
n
itio
n
f
r
am
ew
o
r
k
ter
m
ed
v
is
io
n
tr
an
s
f
o
r
m
e
r
-
b
ased
atten
tio
n
h
an
d
wr
itten
wo
r
d
r
e
co
g
n
itio
n
(
ViT
A
-
HW
R
)
is
p
r
o
p
o
s
ed
.
T
h
e
m
o
d
el
in
teg
r
ates
a
v
is
io
n
tr
an
s
f
o
r
m
er
(
ViT
)
f
o
r
f
ea
tu
r
e
ex
tr
ac
tio
n
an
d
an
atten
tio
n
-
b
ased
f
ee
d
f
o
r
war
d
n
eu
r
al
n
etwo
r
k
(
AFFNN)
f
o
r
r
ec
o
g
n
itio
n
.
T
h
e
ViT
m
o
d
u
le
ca
p
t
u
r
es
h
o
lis
tic
g
lo
b
al
r
ep
r
esen
tatio
n
s
o
f
h
an
d
wr
itten
wo
r
d
im
ag
es,
wh
ile
th
e
atten
tio
n
m
ec
h
an
is
m
in
AFFNN
d
y
n
am
ically
em
p
h
asizes
th
e
m
o
s
t
d
is
cr
im
in
ativ
e
f
ea
tu
r
es,
lead
in
g
to
r
o
b
u
s
t
r
ec
o
g
n
itio
n
ev
en
i
n
co
m
p
lex
h
an
d
wr
itten
s
cr
ip
ts
.
T
h
e
k
ey
co
n
tr
i
b
u
tio
n
s
o
f
th
i
s
r
esear
ch
ar
e
f
o
u
r
f
o
ld
:
i)
d
ev
elo
p
m
en
t
o
f
a
d
ed
icate
d
Ma
lay
alam
h
an
d
wr
itten
wo
r
d
d
ataset,
wh
i
ch
s
er
v
es
as
a
n
ew
b
en
c
h
m
ar
k
f
o
r
th
e
r
esear
c
h
co
m
m
u
n
ity
;
ii)
in
tr
o
d
u
ctio
n
o
f
a
tr
an
s
f
o
r
m
er
-
b
ased
h
y
b
r
id
r
e
co
g
n
itio
n
ar
ch
itectu
r
e
co
m
b
in
in
g
ViT
an
d
atten
tio
n
-
en
h
an
ce
d
FF
NN
f
o
r
im
p
r
o
v
e
d
p
er
f
o
r
m
a
n
ce
;
iii)
c
o
m
p
r
eh
e
n
s
iv
e
ev
alu
atio
n
o
f
m
u
ltip
le
r
ec
o
g
n
itio
n
s
tr
ateg
ie
s
,
in
clu
d
in
g
FF
NN,
g
lo
b
al
av
er
a
g
e
p
o
o
lin
g
(
GAP)
,
b
id
ir
ec
tio
n
al
lo
n
g
s
h
o
r
t
-
ter
m
m
em
o
r
y
(
B
iLST
M
)
,
an
d
AFFNN,
r
ev
ea
lin
g
th
e
s
u
p
er
io
r
ity
o
f
atten
tio
n
m
ec
h
a
n
is
m
;
an
d
iv
)
ac
h
iev
em
en
t
o
f
9
8
.
5
6
%
ac
cu
r
ac
y
o
n
p
r
o
p
o
s
ed
d
ataset,
m
ar
k
in
g
a
s
ig
n
if
ican
t
ad
v
an
ce
m
e
n
t
in
Ma
lay
alam
HW
R
.
T
h
is
s
tu
d
y
th
u
s
b
r
id
g
es
th
e
r
esear
ch
g
ap
f
o
r
lo
w
-
r
eso
u
r
ce
s
cr
ip
ts
an
d
co
n
tr
i
b
u
tes
b
o
t
h
a
n
ew
d
ataset
an
d
an
i
n
n
o
v
a
tiv
e
tr
an
s
f
o
r
m
e
r
-
d
r
i
v
en
r
ec
o
g
n
itio
n
f
r
a
m
ewo
r
k
.
Ov
er
all,
it sets
a
f
o
u
n
d
atio
n
f
o
r
f
u
tu
r
e
ad
v
a
n
ce
m
en
ts
in
r
e
g
io
n
al
h
an
d
wr
itin
g
r
ec
o
g
n
itio
n
s
y
s
tem
s
.
2.
M
E
T
H
O
D
T
h
is
s
ec
tio
n
d
escr
ib
es
th
e
s
y
s
tem
atic
wo
r
k
f
lo
w
f
o
llo
wed
in
th
is
s
tu
d
y
.
T
h
e
wo
r
k
f
lo
w
in
clu
d
es
d
ataset
cr
ea
tio
n
,
f
ea
t
u
r
e
e
x
tr
ac
tio
n
u
s
in
g
th
e
ViT
,
a
n
d
HW
R
th
r
o
u
g
h
v
ar
io
u
s
d
ee
p
le
ar
n
in
g
m
o
d
els.
T
h
e
o
v
er
all
p
r
o
ce
s
s
is
illu
s
tr
ated
in
Fig
u
r
e
1.
Fig
u
r
e
1
.
W
o
r
k
f
lo
w
d
iag
r
am
o
f
th
e
p
r
o
p
o
s
ed
HW
R
s
y
s
tem
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
V
is
io
n
tr
a
n
s
fo
r
mer a
n
d
h
yb
r
id
mo
d
els fo
r
Ma
la
ya
la
m
h
a
n
d
w
r
itten
…
(
A
n
ju
A
r
a
n
g
il Th
a
z
h
a
th
)
2657
2
.
1
.
Da
t
a
s
et
cr
ea
t
io
n
Sin
ce
n
o
p
u
b
licly
a
v
ailab
le
h
an
d
wr
itten
MWD
ex
is
ts
,
a
n
ew
d
ataset
titl
ed
M
W
D
was
cr
ea
ted
to
f
ac
ilit
ate
tr
ain
in
g
an
d
ev
alu
at
io
n
o
f
th
e
p
r
o
p
o
s
ed
m
o
d
els.
T
h
e
d
ataset
was
co
n
s
tr
u
cted
u
s
in
g
h
an
d
wr
itten
p
o
lice
f
ir
s
t
in
f
o
r
m
atio
n
s
tatem
en
ts
(
FIS)
,
wh
ich
wer
e
ch
o
s
en
d
u
e
to
th
eir
p
r
ac
tical
im
p
o
r
tan
ce
an
d
th
e
co
n
s
is
ten
t h
an
d
wr
itten
f
o
r
m
at
u
s
ed
ac
r
o
s
s
Ker
ala
Po
lice
Dep
ar
tm
en
ts
.
T
h
e
FIS
d
o
cu
m
en
ts
wer
e
s
o
u
r
ce
d
f
r
o
m
th
e
Dis
tr
ict
C
r
im
e
R
ec
o
r
d
B
u
r
ea
u
(
DC
R
B
)
in
Ko
zh
ik
o
d
e
a
n
d
s
ca
n
n
ed
in
to
d
ig
ital
f
o
r
m
at
f
o
r
p
r
o
ce
s
s
in
g
.
T
o
ex
tr
ac
t
in
d
iv
id
u
al
wo
r
d
s
am
p
l
es,
an
im
ag
e
p
r
o
ce
s
s
in
g
p
ip
elin
e
was
ap
p
lied
,
in
v
o
l
v
in
g
c
o
n
to
u
r
d
etec
tio
n
an
d
b
o
u
n
d
in
g
b
o
x
r
ec
tan
g
le
m
eth
o
d
s
.
I
n
th
is
wo
r
k
,
t
h
e
d
ilatio
n
p
r
o
ce
s
s
was
m
o
d
if
ied
to
co
n
n
e
ct
in
d
iv
id
u
al
wo
r
d
s
r
ath
er
th
a
n
en
tire
tex
t
lin
es,
en
h
an
cin
g
s
eg
m
en
tatio
n
p
r
ec
is
io
n
.
Su
b
s
eq
u
en
tly
,
wo
r
d
cl
u
s
ter
in
g
tech
n
i
q
u
es
wer
e
u
s
ed
to
g
r
o
u
p
s
eg
m
en
ted
wo
r
d
s
in
to
d
is
tin
ct
clu
s
ter
s
r
e
p
r
esen
tin
g
u
n
iq
u
e
wo
r
d
class
es.
T
h
e
f
in
al
MWD
co
m
p
r
is
es
2
0
,
8
5
0
h
an
d
wr
itten
wo
r
d
s
am
p
les,
d
is
tr
ib
u
ted
e
v
en
ly
ac
r
o
s
s
1
3
9
class
es,
with
ea
ch
class
co
n
tain
in
g
1
5
0
s
am
p
les.
W
o
r
d
s
elec
tio
n
was
g
u
id
ed
b
y
th
e
lin
g
u
i
s
tic
f
r
eq
u
en
cy
a
n
d
co
n
tex
tu
al
s
ig
n
if
ican
ce
o
f
w
o
r
d
s
in
Ma
lay
alam
d
o
cu
m
e
n
t
s
,
en
s
u
r
in
g
b
r
o
ad
r
ep
r
esen
tat
io
n
al
co
v
er
a
g
e.
T
o
en
s
u
r
e
s
ty
le
d
iv
er
s
ity
,
h
an
d
wr
itin
g
s
am
p
les
wer
e
co
lle
cted
f
r
o
m
7
5
0
in
d
i
v
id
u
als
ag
ed
1
0
–
6
5
y
ea
r
s
,
r
ep
r
esen
tin
g
d
if
f
er
en
t
e
d
u
ca
ti
o
n
al
an
d
r
eg
io
n
al
b
ac
k
g
r
o
u
n
d
s
.
T
h
is
d
iv
er
s
ity
im
p
r
o
v
es
m
o
d
el
g
en
e
r
aliza
tio
n
b
y
ex
p
o
s
in
g
it
to
a
wid
e
r
a
n
g
e
o
f
h
a
n
d
wr
itin
g
v
ar
iatio
n
s
.
Qu
ality
co
n
tr
o
l
c
h
ec
k
s
wer
e
p
e
r
f
o
r
m
ed
t
o
elim
in
ate
in
co
m
p
lete
o
r
n
o
is
y
s
am
p
les.
Fig
u
r
e
2
s
h
o
ws r
ep
r
esen
tativ
e
s
am
p
les f
r
o
m
th
e
d
ev
elo
p
e
d
MWD d
ataset.
Fig
u
r
e
2
.
Sam
p
le
w
o
r
d
s
f
r
o
m
MWD d
ataset
2
.
2
.
Resea
rc
h
wo
r
k
f
lo
w
R
ec
o
g
n
izin
g
h
an
d
wr
itten
Ma
lay
alam
wo
r
d
s
is
a
c
o
m
p
lex
p
r
o
b
lem
d
u
e
to
th
e
c
o
m
p
o
u
n
d
s
cr
ip
t
s
tr
u
ctu
r
e,
wr
itin
g
v
ar
iatio
n
s
,
a
n
d
cu
r
s
iv
e
n
atu
r
e
o
f
th
e
lan
g
u
ag
e.
T
o
ad
d
r
ess
th
ese
ch
allen
g
es,
a
tr
an
s
f
o
r
m
er
-
b
ased
f
r
am
ewo
r
k
was
d
ev
el
o
p
ed
th
at
c
o
m
b
in
es
th
e
ViT
f
o
r
f
ea
tu
r
e
ex
tr
ac
tio
n
with
d
if
f
er
en
t
r
ec
o
g
n
itio
n
m
o
d
els.
T
h
e
ex
p
er
im
en
tal
wo
r
k
f
lo
w
is
s
u
m
m
ar
ized
in
Alg
o
r
ith
m
1
.
E
m
p
ir
ical
ev
alu
atio
n
r
ev
ea
led
th
at
th
e
AFFNN o
u
tp
er
f
o
r
m
ed
th
e
o
th
er
m
o
d
els,
lead
in
g
to
th
e
f
in
al
h
y
b
r
id
ar
ch
itectu
r
e
te
r
m
ed
Vi
T
A
-
HW
R
.
Alg
o
r
ith
m
1
.
Ma
lay
alam
HW
R
wo
r
k
f
lo
w
I
n
p
u
t: Ha
n
d
wr
itten
wo
r
d
d
ataset
with
1
3
9
class
es,
ea
ch
co
n
ta
in
in
g
1
5
0
s
am
p
les.
Ou
tp
u
t: R
ec
o
g
n
ized
Ma
lay
ala
m
h
an
d
wr
itten
w
o
r
d
s
.
1
:
Data
s
et
p
r
ep
ar
atio
n
:
p
er
f
o
r
m
p
r
ep
r
o
ce
s
s
in
g
an
d
s
eg
m
en
tati
o
n
to
o
b
tain
in
d
iv
id
u
al
wo
r
d
s
am
p
les
f
r
o
m
s
ca
n
n
ed
h
an
d
wr
itten
FIS
d
o
c
u
m
en
ts
.
2
:
Featu
r
e
ex
tr
ac
tio
n
u
s
in
g
ViT
:
tr
an
s
f
o
r
m
im
a
g
es in
to
p
atch
es
an
d
ex
tr
ac
t
f
ea
tu
r
es u
s
in
g
ViT
.
Fo
r
m
a
f
ea
tu
r
e
m
atr
ix
×
,
wh
er
e
is
th
e
n
u
m
b
e
r
o
f
s
am
p
les an
d
th
e
f
ea
tu
r
e
d
im
en
s
io
n
.
3
:
R
ec
o
g
n
itio
n
m
o
d
els
:
̶
FF
NN
=
(
+
)
̶
GAP
=
(
)
̶
B
i
L
STM
=
(
+
)
̶
AFFNN
=
(
(
⊙
)
+
)
W
h
er
e
d
en
o
tes th
e
atten
tio
n
m
atr
ix
,
⊙
elem
en
t
-
wis
e
m
u
ltip
licatio
n
,
an
d
th
e
ac
tiv
atio
n
f
u
n
ctio
n
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
1
5
,
No
.
3
,
J
u
n
e
2
0
2
6
:
2
6
5
5
-
2
6
6
3
2658
4
:
C
o
m
p
a
r
a
t
i
v
e
e
v
al
u
a
t
i
o
n
:
a
s
s
e
s
s
a
l
l
m
o
d
e
ls
u
s
i
n
g
m
e
t
r
i
c
s
s
u
c
h
a
s
a
c
c
u
r
a
c
y
,
p
r
e
ci
s
i
o
n
,
r
e
c
al
l
,
an
d
F
1
-
s
c
o
r
e
.
5
:
Mo
d
el
s
elec
tio
n
: id
en
tify
t
h
e
m
o
d
el
with
th
e
h
ig
h
est r
ec
o
g
n
itio
n
p
er
f
o
r
m
an
ce
.
2
.
3
.
F
e
a
t
ure
ex
t
r
a
ct
io
n us
ing
v
is
io
n t
ra
ns
f
o
rm
er
T
h
e
ViT
was
eq
u
ip
p
ed
to
e
x
tr
ac
t
d
is
cr
im
in
ativ
e
v
is
u
al
f
ea
tu
r
es
f
r
o
m
h
an
d
wr
itten
wo
r
d
im
a
g
es.
Un
lik
e
C
NNs,
wh
ich
f
o
cu
s
o
n
lo
ca
l
p
atter
n
s
v
ia
co
n
v
o
l
u
tio
n
al
k
er
n
els,
ViT
ca
p
tu
r
e
s
g
lo
b
al
co
n
tex
tu
al
d
ep
en
d
e
n
cies
th
r
o
u
g
h
th
e
s
e
lf
-
atten
tio
n
m
ec
h
a
n
is
m
.
T
h
is
th
er
eb
y
im
p
r
o
v
es
f
ea
tu
r
e
r
ep
r
esen
tatio
n
an
d
r
ec
o
g
n
itio
n
r
o
b
u
s
tn
ess
[
2
2
]
.
2
.
3
.
1
.
Vis
io
n t
ra
ns
f
o
rm
er
a
r
chit
ec
t
ure
T
h
e
ViT
p
r
o
ce
s
s
es
an
im
ag
e
as
a
s
eq
u
en
ce
o
f
f
ix
ed
-
s
ize
p
atch
es
[
2
3
]
.
E
ac
h
in
p
u
t
im
a
g
e
is
d
iv
id
e
d
in
to
p
atch
es
as
,
=
(
)
,
wh
er
e
(
∙
)
d
en
o
t
es
th
e
p
atch
ex
tr
ac
tio
n
o
p
er
at
io
n
.
Po
s
itio
n
al
em
b
ed
d
i
n
g
s
ar
e
th
en
ad
d
e
d
to
p
r
eser
v
e
s
p
atia
l
in
f
o
r
m
atio
n
:
,
=
,
+
(
,
)
,
wh
er
e
(
,
)
is
th
e
p
o
s
itio
n
al
en
co
d
in
g
f
o
r
th
e
p
atch
lo
ca
ted
at
(
,
)
.
T
h
e
r
esu
ltin
g
to
k
en
s
eq
u
en
ce
is
lin
ea
r
ly
p
r
o
jecte
d
an
d
p
ass
ed
th
r
o
u
g
h
m
u
ltip
le
tr
an
s
f
o
r
m
er
en
co
d
er
lay
er
s
,
e
ac
h
co
m
p
r
is
in
g
m
u
lti
-
h
ea
d
s
e
lf
-
atten
tio
n
a
n
d
f
ee
d
f
o
r
war
d
s
u
b
-
lay
er
s
.
Fo
r
ea
c
h
atten
tio
n
h
ea
d
,
th
e
s
ca
led
d
o
t
-
p
r
o
d
u
ct
atten
tio
n
is
co
m
p
u
ted
as in
(
1
)
.
(
,
,
)
=
(
|
√
)
(
1
)
W
h
er
e
,
,
an
d
r
ep
r
esen
t
th
e
q
u
er
y
,
k
ey
,
a
n
d
v
alu
e
m
a
tr
ices,
an
d
is
th
e
k
ey
d
im
e
n
s
io
n
.
T
h
e
m
u
lti
-
h
ea
d
m
ec
h
an
is
m
allo
ws
th
e
m
o
d
el
to
atten
d
to
d
if
f
er
en
t
s
u
b
s
p
ac
es
o
f
t
h
e
f
ea
t
u
r
e
s
p
ac
e,
en
h
an
cin
g
its
ab
ilit
y
to
m
o
d
el
r
elatio
n
s
h
ip
s
b
etwe
en
d
is
tan
t
im
ag
e
r
eg
io
n
s
.
T
h
e
o
u
tp
u
t
f
ea
tu
r
e
em
b
ed
d
i
n
g
s
ar
e
n
o
r
m
alize
d
an
d
u
s
ed
as in
p
u
t to
th
e
r
ec
o
g
n
itio
n
m
o
d
u
le.
Fig
u
r
e
3
d
e
p
icts
th
e
ViT
ar
ch
itectu
r
e.
Fig
u
r
e
3
.
ViT
ar
c
h
itectu
r
e
2
.
4
.
Rec
o
g
nitio
n
m
o
dels
Fo
llo
win
g
ViT
-
b
ased
f
ea
tu
r
e
ex
tr
ac
tio
n
,
f
o
u
r
d
is
tin
ct
r
ec
o
g
n
itio
n
s
tr
ateg
ies
wer
e
im
p
lem
en
ted
to
ev
alu
ate
p
er
f
o
r
m
an
c
e
d
if
f
e
r
en
ce
s
.
i)
FF
NN:
th
e
F
FNN
co
m
p
r
is
es
two
f
u
lly
co
n
n
ec
ted
lay
er
s
.
T
h
e
f
ir
s
t
lay
er
r
e
d
u
ce
s
f
ea
tu
r
e
d
im
en
s
io
n
ality
:
ℎ
1
=
(
1
×
+
1
)
,
wh
er
e
1
is
th
e
weig
h
t
m
atr
ix
o
f
th
e
f
i
r
s
t
lay
er
,
1
is
th
e
b
ias
v
ec
to
r
f
o
r
th
e
f
ir
s
t
la
y
er
,
a
n
d
is
th
e
ac
tiv
ati
o
n
f
u
n
ctio
n
.
T
h
e
s
ec
o
n
d
lay
er
p
r
o
jects
to
class
s
co
r
es:
=
2
×
ℎ1
+
2
,
wh
er
e
2
is
th
e
weig
h
t
m
atr
ix
o
f
th
e
s
ec
o
n
d
lay
er
,
2
is
th
e
b
ias
v
ec
to
r
f
o
r
th
e
s
ec
o
n
d
lay
er
,
is
th
e
o
u
tp
u
t v
e
cto
r
,
wh
er
e
ea
c
h
elem
en
t c
o
r
r
esp
o
n
d
s
to
a
class
s
co
r
e
[
2
4
]
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
V
is
io
n
tr
a
n
s
fo
r
mer a
n
d
h
yb
r
id
mo
d
els fo
r
Ma
la
ya
la
m
h
a
n
d
w
r
itten
…
(
A
n
ju
A
r
a
n
g
il Th
a
z
h
a
th
)
2659
ii)
GAP:
th
e
GAP
ag
g
r
eg
ates
ea
ch
f
ea
tu
r
e
ch
an
n
el
b
y
av
e
r
ag
in
g
:
=
1
∑
,
=
1
.
T
h
e
n
,
a
d
en
s
e
lay
e
r
with
s
o
f
tm
ax
ac
tiv
atio
n
class
if
ies th
e
p
o
o
led
r
ep
r
esen
tatio
n
[
2
5
]
.
iii)
B
iLST
M:
th
e
B
i
L
STM
to
ex
p
lo
it
s
eq
u
en
tial
d
ep
e
n
d
en
cies
with
in
f
ea
tu
r
es,
B
i
L
STM
p
r
o
ce
s
s
es
in
p
u
t
in
b
o
th
d
ir
ec
tio
n
s
:
ℎ
=
(
,
ℎ
−
1
)
.
T
h
e
f
i
n
al
h
id
d
e
n
s
tate
is
p
ass
ed
th
r
o
u
g
h
d
e
n
s
e
lay
er
s
f
o
r
class
if
icatio
n
[
2
5
]
.
iv
)
A
F
F
N
N
:
t
h
i
s
m
o
d
e
l
e
m
b
e
d
s
m
u
l
t
i
-
h
e
a
d
s
e
l
f
-
a
t
t
e
n
t
i
o
n
m
e
c
h
a
n
i
s
m
w
i
t
h
i
n
F
F
N
N
t
o
e
m
p
h
a
s
i
z
e
c
r
u
c
i
a
l
s
p
a
t
i
a
l
f
e
a
t
u
r
e
s
.
F
o
r
e
a
c
h
a
t
t
e
n
t
i
o
n
h
e
a
d
:
=
(
,
,
)
.
T
h
e
o
u
t
p
u
t
s
f
r
o
m
a
l
l
h
e
a
d
s
a
r
e
c
o
n
c
a
t
e
n
a
t
e
d
a
n
d
l
i
n
e
a
r
l
y
t
r
a
n
s
f
o
r
m
e
d
:
=
(
(
1
,
2
,
3
,
4
)
)
.
A
r
e
s
i
d
u
a
l
c
o
n
n
e
c
t
i
o
n
e
n
s
u
r
e
s
i
n
f
o
r
m
a
t
i
o
n
p
r
e
s
e
r
v
a
t
i
o
n
:
=
+
(
)
.
T
wo
d
en
s
e
lay
er
s
th
en
p
e
r
f
o
r
m
class
if
icatio
n
.
T
h
e
atten
tio
n
m
ec
h
a
n
is
m
en
ab
l
es
th
e
n
etwo
r
k
to
d
y
n
am
icall
y
f
o
c
u
s
o
n
th
e
m
o
s
t in
f
o
r
m
ativ
e
f
ea
tu
r
es,
im
p
r
o
v
in
g
r
o
b
u
s
tn
ess
ag
ain
s
t h
a
n
d
wr
itin
g
v
a
r
iatio
n
s
an
d
n
o
is
e
.
2
.
5
.
P
r
o
po
s
ed
v
is
io
n t
ra
ns
f
o
rm
er
-
ba
s
ed
a
t
t
ent
io
n ha
nd
writ
t
en
wo
rd
re
co
g
nitio
n
m
o
d
el
T
h
e
f
i
n
al
a
r
c
h
i
tec
tu
r
e,
n
am
ed
ViT
A
-
HW
R
,
i
n
te
g
r
at
es
ViT
-
b
ase
d
f
ea
t
u
r
e
e
x
t
r
ac
ti
o
n
wit
h
th
e
AFFN
N
r
e
co
g
n
it
io
n
m
o
d
u
l
e
as
s
h
o
w
n
in
F
ig
u
r
e
4
.
T
h
is
h
y
b
r
i
d
a
p
p
r
o
a
ch
ca
p
t
u
r
es
b
o
t
h
g
l
o
b
al
c
o
n
tex
tu
al
p
atte
r
n
s
an
d
lo
c
ali
ze
d
d
is
c
r
i
m
i
n
ati
v
e
d
e
tai
ls
.
I
t
e
n
s
u
r
es
ef
f
i
cie
n
t
a
n
d
ac
cu
r
ate
r
ec
o
g
n
iti
o
n
o
f
Ma
l
ay
al
am
h
a
n
d
w
r
itt
e
n
w
o
r
d
s
.
Fig
u
r
e
4
.
ViT
A
-
HW
R
ar
ch
itectu
r
e
3.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
N
T
h
is
s
t
u
d
y
c
o
m
p
r
e
h
e
n
s
i
v
e
ly
c
o
m
p
ar
es
f
o
u
r
r
ec
o
g
n
iti
o
n
s
t
r
a
t
eg
i
es:
FF
NN,
a
v
e
r
a
g
e
p
o
o
l
in
g
,
B
i
L
ST
M,
an
d
AFFN
.
N
o
t
ab
ly
,
th
e
att
e
n
ti
o
n
-
b
ase
d
ap
p
r
o
ac
h
s
ta
n
d
s
o
u
t
w
it
h
r
e
m
a
r
k
a
b
l
e
p
e
r
f
o
r
m
a
n
c
e.
T
h
is
h
ig
h
l
ig
h
ts
th
a
t
in
te
g
r
ati
n
g
a
n
att
e
n
ti
o
n
m
ec
h
an
is
m
i
m
p
r
o
v
es
t
h
e
m
o
d
el'
s
ca
p
ac
it
y
t
o
i
d
e
n
ti
f
y
in
tr
i
ca
t
e
p
a
tte
r
n
s
wit
h
i
n
Ma
la
y
ala
m
h
a
n
d
w
r
it
te
n
w
o
r
d
s
.
T
h
r
o
u
g
h
co
m
p
a
r
a
ti
v
e
e
x
p
er
i
m
e
n
ts
,
th
e
V
iTA
-
HW
R
ac
h
ie
v
ed
t
h
e
h
ig
h
est
r
e
co
g
n
it
io
n
a
cc
u
r
a
cy
o
f
9
8
.
5
6
%,
d
e
m
o
n
s
tr
ati
n
g
t
h
e
ef
f
e
ct
iv
e
n
ess
o
f
c
o
m
b
in
in
g
t
r
a
n
s
f
o
r
m
er
-
b
as
ed
f
e
at
u
r
e
ex
t
r
ac
ti
o
n
wit
h
at
te
n
ti
o
n
-
d
r
i
v
e
n
r
e
c
o
g
n
i
ti
o
n
.
T
a
b
l
e
1
s
u
m
m
a
r
izes
t
h
e
e
x
p
e
r
i
m
e
n
ta
l
r
es
u
l
ts
o
f
al
l t
este
d
m
o
d
e
ls
.
T
ab
le
1
.
E
x
p
er
im
en
tal
r
esu
lts
M
o
d
e
l
V
a
l
i
d
a
t
i
o
n
l
o
ss
V
a
l
i
d
a
t
i
o
n
a
c
c
u
r
a
c
y
(
%)
Te
st
l
o
ss
Te
st
a
c
c
u
r
a
cy
(
%)
A
F
F
N
N
0
.
2
4
3
2
9
8
.
6
6
0
.
2
3
6
4
9
8
.
5
6
F
F
N
N
0
.
1
0
6
8
9
7
.
5
4
0
.
0
9
5
7
9
7
.
8
9
B
LST
M
0
.
1
4
4
4
9
7
.
0
3
0
.
1
4
8
3
9
7
.
0
9
GAP
0
.
2
0
3
1
9
5
.
1
4
0
.
1
9
5
0
9
5
.
5
3
3
.
1
.
Co
m
pa
riso
n
wit
h sta
t
e
-
of
-
t
he
-
a
r
t
t
ec
hn
iqu
es
T
h
e
co
m
p
ar
ativ
e
a
n
aly
s
is
p
r
esen
ted
in
th
e
ab
o
v
e
tab
le
h
ig
h
l
ig
h
ts
th
e
ev
o
l
u
tio
n
o
f
HW
R
ap
p
r
o
ac
h
es
ac
r
o
s
s
B
en
g
ali
an
d
Ma
lay
ala
m
s
cr
ip
ts
.
T
h
ese
s
cr
ip
ts
ar
e
k
n
o
wn
f
o
r
th
eir
co
m
p
lex
c
h
ar
ac
ter
s
tr
u
ctu
r
es
an
d
s
im
ilar
s
tr
o
k
e
p
atter
n
s
,
m
ak
i
n
g
th
em
s
u
itab
le
f
o
r
cr
o
s
s
-
s
cr
ip
t
co
m
p
ar
is
o
n
.
E
ar
lier
m
et
h
o
d
s
p
r
im
ar
ily
r
elied
o
n
tr
ad
itio
n
al
class
if
ier
s
lik
e
SV
M
with
h
an
d
-
cr
a
f
ted
f
ea
tu
r
es.
At
th
e
s
am
e
tim
e
r
ec
en
t
s
tu
d
ies
h
av
e
s
h
if
ted
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
1
5
,
No
.
3
,
J
u
n
e
2
0
2
6
:
2
6
5
5
-
2
6
6
3
2660
to
war
d
s
d
ee
p
lear
n
in
g
-
b
ased
a
p
p
r
o
ac
h
es
s
u
ch
as
C
NN
an
d
FF
NN,
co
u
p
led
with
d
ata
au
g
m
en
tatio
n
s
tr
ateg
ies
to
im
p
r
o
v
e
g
en
er
aliza
tio
n
.
Desp
ite
th
ese
ad
v
an
ce
m
en
ts
,
m
o
s
t
ex
is
tin
g
wo
r
k
s
eith
er
f
o
cu
s
o
n
lim
ited
d
atasets
o
r
d
o
n
o
t in
c
o
r
p
o
r
ate
atten
tio
n
m
ec
h
an
is
m
s
f
o
r
e
n
h
an
ce
d
f
ea
tu
r
e
d
is
cr
im
in
atio
n
.
I
n
co
n
tr
ast,
th
e
p
r
o
p
o
s
ed
Vi
T
A
-
HW
R
m
o
d
el
in
tr
o
d
u
ce
s
a
h
y
b
r
id
a
r
ch
itectu
r
e
t
h
at
co
m
b
in
es
th
e
g
lo
b
al
f
ea
tu
r
e
ex
tr
ac
tio
n
ca
p
a
b
ilit
y
o
f
ViT
with
th
e
ad
a
p
tiv
e
lear
n
in
g
s
tr
en
g
th
o
f
atten
tio
n
-
b
ased
f
ee
d
f
o
r
war
d
n
etwo
r
k
s
.
T
h
is
in
teg
r
ated
d
esi
g
n
elim
in
ates
th
e
n
ee
d
f
o
r
d
at
a
au
g
m
en
tatio
n
an
d
ac
h
iev
es
s
u
p
er
io
r
ac
cu
r
ac
y
,
th
er
eb
y
d
em
o
n
s
tr
atin
g
a
s
ig
n
if
ican
t
ad
v
an
ce
m
en
t
o
v
e
r
ex
is
tin
g
m
o
d
els.
T
ab
le
2
co
m
p
a
r
es
th
e
p
r
esen
ted
wo
r
k
with
ex
is
tin
g
o
f
f
lin
e
HW
R
wo
r
k
s
in
b
o
th
Ma
lay
alam
an
d
B
en
g
ali
lan
g
u
a
g
es.
T
ab
le
2
.
C
o
m
p
a
r
is
o
n
o
f
s
tate
-
of
-
th
e
-
a
r
t o
f
f
lin
e
HW
R
s
tu
d
ie
s
an
d
p
r
o
p
o
s
ed
wo
r
k
S
o
u
r
c
e
D
a
t
a
s
e
t
(
samp
l
e
s)
D
a
t
a
a
u
g
m
e
n
t
a
t
i
o
n
F
e
a
t
u
r
e
e
x
t
r
a
c
t
i
o
n
R
e
c
o
g
n
i
t
i
o
n
A
c
c
u
r
a
c
y
(
%)
B
h
o
w
mi
k
e
t
a
l
.
[
6
]
1
8
,
0
0
0
N
i
l
S
h
a
p
e
-
b
a
se
d
f
e
a
t
u
r
e
d
e
s
c
r
i
p
t
o
r
S
V
M
8
3
.
6
4
M
a
l
a
k
a
r
e
t
a
l
.
[
1
7
]
1
8
,
0
0
0
N
i
l
C
N
N
C
N
N
mo
d
e
l
,
l
o
t
t
e
r
y
Ti
c
k
e
t
h
y
p
o
t
h
e
s
i
s
9
1
.
3
0
D
a
s
e
t
a
l
.
[
1
6
]
1
8
,
0
0
0
N
i
l
C
N
N
F
F
N
N
(
H
W
o
r
d
N
e
t
)
9
6
.
1
7
Ji
n
o
e
t
a
l
.
[
1
5
]
2
9
,
5
1
6
R
o
t
a
t
i
o
n
,
G
a
u
ss
i
a
n
n
o
i
se
,
s
h
i
f
t
i
n
g
C
N
N
S
V
M
9
6
.
9
0
P
r
o
p
o
se
d
w
o
r
k
2
0
,
8
5
0
N
i
l
V
i
T
A
F
F
N
N
9
8
.
5
6
T
o
estab
lis
h
th
e
ef
f
ec
tiv
en
ess
o
f
th
e
p
r
o
p
o
s
ed
ViT
A
-
H
W
R
m
o
d
el,
a
co
m
p
a
r
ativ
e
a
n
aly
s
is
was
co
n
d
u
cte
d
ag
ain
s
t e
x
is
tin
g
b
e
n
ch
m
ar
k
m
o
d
els f
o
r
HW
R
.
T
wo
s
tate
-
of
-
th
e
-
ar
t a
p
p
r
o
ac
h
es
wer
e
s
elec
ted
b
ased
o
n
th
eir
r
elev
an
ce
to
th
e
task
:
i)
Pre
-
tr
ain
ed
C
NN
+
SVM
m
o
d
el:
th
is
m
o
d
el
was
ch
o
s
en
as
it
r
ep
r
esen
ts
th
e
s
tate
-
of
-
t
h
e
-
ar
t
ap
p
r
o
ac
h
f
o
r
Ma
lay
alam
HW
R
,
u
tili
zin
g
a
p
r
e
-
tr
ain
ed
C
NN
f
o
r
f
ea
tu
r
e
e
x
tr
ac
tio
n
f
o
llo
wed
b
y
an
SVM
class
if
ier
f
o
r
r
ec
o
g
n
itio
n
[
1
5
]
.
I
t
p
r
o
v
id
es
a
d
ir
ec
t
co
m
p
ar
is
o
n
with
ex
i
s
tin
g
m
eth
o
d
s
d
esig
n
ed
s
p
ec
if
ically
f
o
r
th
e
Ma
lay
alam
s
cr
ip
t.
T
h
is
m
o
d
el
ac
h
iev
ed
a
test
ac
cu
r
ac
y
o
f
8
7
.
8
5
% o
n
th
e
MWD
.
ii)
HW
o
r
d
Net
C
NN
mod
el:
a
w
el
l
-
estab
lis
h
ed
b
en
ch
m
ar
k
m
o
d
el
f
o
r
lo
w
-
r
eso
u
r
ce
lan
g
u
ag
e
s
,
p
ar
ticu
lar
l
y
B
an
g
la,
wh
ich
s
h
ar
es
s
tr
u
ctu
r
al
s
im
ilar
ities
with
Ma
lay
al
am
r
eg
ar
d
in
g
c
o
m
p
lex
s
cr
ip
t
-
b
ased
wr
itin
g
s
ty
les
[
1
6
]
.
Sin
ce
B
an
g
la
an
d
Ma
lay
alam
b
o
th
p
r
esen
t
c
h
allen
g
es
in
h
a
n
d
wr
itten
tex
t
r
ec
o
g
n
itio
n
,
ev
alu
atin
g
HW
o
r
d
Net
o
n
th
e
MWD
al
lo
ws
f
o
r
a
b
r
o
ad
er
ass
es
s
m
en
t
o
f
its
ap
p
licab
ilit
y
.
T
h
is
m
o
d
el
ac
h
iev
ed
a
test
ac
cu
r
ac
y
o
f
9
6
.
6
9
%.
T
h
e
r
esu
lts
in
d
icate
th
at
th
e
p
r
o
p
o
s
ed
ViT
A
-
HW
R
m
o
d
el
s
i
g
n
if
ican
tly
o
u
t
p
er
f
o
r
m
s
b
o
th
b
en
ch
m
ar
k
m
o
d
els,
d
em
o
n
s
tr
atin
g
s
u
p
er
i
o
r
r
ec
o
g
n
itio
n
ac
c
u
r
ac
y
.
T
h
e
ViT
A
-
HW
R
m
o
d
el'
s
atten
tio
n
-
en
h
a
n
ce
d
f
ea
t
u
r
e
ex
tr
ac
tio
n
an
d
class
if
icatio
n
f
r
am
ewo
r
k
e
f
f
ec
tiv
ely
ca
p
tu
r
e
th
e
co
m
p
lex
ities
o
f
h
an
d
wr
itten
Ma
lay
alam
wo
r
d
s
,
f
u
r
th
er
r
ei
n
f
o
r
ci
n
g
its
s
u
itab
il
ity
f
o
r
lo
w
-
r
eso
u
r
ce
h
an
d
wr
itten
tex
t
r
ec
o
g
n
itio
n
.
T
o
en
s
u
r
e
r
ep
r
o
d
u
cib
ilit
y
,
p
r
o
p
o
s
ed
m
o
d
el
was tr
ain
ed
with
lear
n
in
g
r
a
te
o
f
0
.
0
0
0
1
,
b
atch
s
ize
o
f
3
2
,
an
d
f
o
r
1
0
ep
o
c
h
s
.
3
.
2
.
B
enchm
a
r
k
ing
t
he
mo
del
o
n CMAT
E
R
db
2
.
1
.
2
pu
bl
ic
da
t
a
s
et
T
o
ev
alu
ate
th
e
g
en
er
aliza
tio
n
ca
p
ab
ilit
y
o
f
th
e
p
r
o
p
o
s
ed
m
o
d
el,
an
ad
d
itio
n
al
ex
p
e
r
im
en
t
was
co
n
d
u
cte
d
u
s
in
g
t
h
e
p
u
b
licly
av
ailab
le
C
MA
T
E
R
d
b
2
.
1
.
2
d
ataset
[
5
]
.
T
h
is
d
ataset
co
n
s
is
ts
o
f
h
an
d
wr
itten
B
an
g
la
wo
r
d
s
am
p
les
r
ep
r
esen
tin
g
1
2
0
city
n
am
es
f
r
o
m
W
est
B
en
g
al,
I
n
d
ia.
T
h
e
d
ata
was
co
llected
i
n
an
o
f
f
lin
e
s
ettin
g
an
d
in
clu
d
es
d
iv
er
s
e
h
an
d
wr
itin
g
s
ty
les
in
f
l
u
en
ce
d
b
y
f
ac
to
r
s
s
u
ch
as
wr
iter
d
em
o
g
r
a
p
h
ics,
in
clu
d
in
g
ag
e,
g
en
d
er
,
ed
u
ca
tio
n
al
b
ac
k
g
r
o
u
n
d
,
a
n
d
p
r
o
f
ess
io
n
.
T
h
ese
i
n
h
er
e
n
t
v
ar
iatio
n
s
m
ak
e
C
MA
T
E
R
d
b
2
.
1
.
2
a
ch
allen
g
in
g
b
en
c
h
m
ar
k
f
o
r
HW
R
.
T
h
e
m
o
d
el
was
tr
ain
e
d
an
d
te
s
ted
o
n
C
MA
T
E
R
d
b
2
.
1
.
2
an
d
th
e
r
esu
lts
d
em
o
n
s
tr
ate
th
at
t
h
e
m
o
d
el
ef
f
ec
tiv
ely
ad
ap
ts
to
th
is
n
e
w
d
ataset,
ac
h
iev
in
g
a
test
ac
cu
r
ac
y
o
f
9
7
.
7
1
%.
Ad
d
itio
n
ally
,
p
er
f
o
r
m
an
ce
m
etr
ics̶
f
u
r
th
er
̶
v
alid
ate̶
th
e̶
m
o
d
el’
s
̶
r
eliab
ilit
y
,
̶
with
̶
a̶
m
icr
o
p
r
ec
is
io
n
o
f
0
.
9
8
1
1
,
m
icr
o
r
ec
all
o
f
0
.
9
8
1
1
,
an
d
m
icr
o
F1
-
s
co
r
e
o
f
0
.
9
8
1
1
.
T
h
e
m
ac
r
o
p
r
ec
is
io
n
,
m
ac
r
o
r
ec
a
ll,
an
d
m
ac
r
o
F1
-
s
co
r
e
wer
e
r
ec
o
r
d
ed
as
0
.
9
8
1
8
,
0
.
9
8
1
5
,
an
d
0
.
9
8
0
8
,
r
esp
ec
tiv
e
ly
.
3
.
3
.
Cro
s
s
-
v
a
lid
a
t
io
n
a
na
ly
s
is
T
o
ev
alu
ate
th
e
r
o
b
u
s
tn
ess
an
d
g
e
n
er
aliza
b
ilit
y
o
f
t
h
e
p
r
o
p
o
s
ed
r
ec
o
g
n
itio
n
m
o
d
el,
a
f
iv
e
-
f
o
l
d
cr
o
s
s
-
v
alid
atio
n
ex
p
er
im
en
t
was
co
n
d
u
cted
o
n
b
o
th
th
e
p
r
iv
ate
Ma
lay
alam
h
an
d
wr
itten
w
o
r
d
d
ataset
an
d
th
e
p
u
b
lic
C
MA
T
E
R
d
b
2
.
1
.
2
B
an
g
la
d
ataset.
T
h
is
a
p
p
r
o
ac
h
h
el
p
ed
m
i
n
im
ize
d
ata
v
ar
iab
ilit
y
ef
f
ec
ts
an
d
p
r
o
v
id
ed
a
r
eliab
le
ass
ess
m
en
t
o
f
m
o
d
el
p
er
f
o
r
m
an
ce
ac
r
o
s
s
d
if
f
e
r
en
t
tr
ain
in
g
-
v
alid
atio
n
s
p
lits
.
T
h
e
d
ataset
was
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
V
is
io
n
tr
a
n
s
fo
r
mer a
n
d
h
yb
r
id
mo
d
els fo
r
Ma
la
ya
la
m
h
a
n
d
w
r
itten
…
(
A
n
ju
A
r
a
n
g
il Th
a
z
h
a
th
)
2661
r
an
d
o
m
l
y
p
ar
titi
o
n
ed
in
to
f
iv
e
f
o
ld
s
,
wh
er
e
ea
ch
f
o
ld
s
er
v
ed
as
a
v
alid
atio
n
s
et
o
n
ce
,
wh
ile
th
e
r
em
ain
in
g
f
o
u
r
wer
e
u
s
ed
f
o
r
tr
ain
in
g
.
T
h
is
p
r
o
ce
s
s
was
r
ep
ea
ted
f
o
r
all
f
iv
e
f
o
ld
s
,
an
d
th
e
p
er
f
o
r
m
an
ce
m
etr
ics
wer
e
av
er
ag
ed
al
o
n
g
with
t
h
eir
s
tan
d
ar
d
d
ev
iatio
n
t
o
o
b
tai
n
a
m
o
r
e
r
eliab
le
esti
m
ate
o
f
m
o
d
el
ac
cu
r
ac
y
a
n
d
lo
s
s
.
T
h
e
tr
ain
in
g
p
r
o
ce
d
u
r
e
in
c
o
r
p
o
r
ated
an
ea
r
l
y
s
to
p
p
in
g
m
ec
h
an
is
m
to
p
r
ev
e
n
t
o
v
er
f
itti
n
g
.
T
ab
le
3
s
h
o
ws
th
e
f
iv
e
-
f
o
ld
cr
o
s
s
-
v
alid
atio
n
r
esu
lts
f
o
r
b
o
th
d
atasets
.
T
ab
le
3
.
Fiv
e
-
f
o
ld
cr
o
s
s
-
v
alid
atio
n
r
esu
lts
o
n
MWD
an
d
C
MA
T
E
R
DB
2
.
1
.
2
F
o
l
d
M
W
D
C
M
A
T
ER
d
b
2
.
1
.
2
B
a
n
g
l
a
d
a
t
a
se
t
V
a
l
i
d
a
t
i
o
n
a
c
c
u
r
a
c
y
(
%)
V
a
l
i
d
a
t
i
o
n
l
o
s
s
Ep
o
c
h
s
V
a
l
i
d
a
t
i
o
n
a
c
c
u
r
a
c
y
(
%)
V
a
l
i
d
a
t
i
o
n
l
o
s
s
Ep
o
c
h
s
F
o
l
d
1
9
7
.
8
2
0
.
0
8
5
8
9
9
7
.
7
6
0
.
0
9
3
4
9
F
o
l
d
2
9
9
.
1
4
0
.
0
3
5
4
6
9
9
.
4
5
0
.
0
1
6
3
8
F
o
l
d
3
9
9
.
1
4
0
.
0
3
6
0
6
9
9
.
7
0
0
.
0
1
2
7
6
F
o
l
d
4
9
8
.
8
3
0
.
0
9
2
5
6
9
9
.
4
2
0
.
0
2
3
2
8
F
o
l
d
5
9
9
.
5
7
0
.
0
1
5
8
5
9
9
.
8
6
0
.
0
0
8
7
4
A
v
e
r
a
g
e
9
8
.
9
0
0
.
0
5
3
1
9
9
.
2
4
0
.
0
3
0
9
S
t
a
n
d
a
r
d
d
e
v
i
a
t
i
o
n
0
.
5
9
0
.
0
3
0
4
0
.
7
6
0
.
0
3
1
6
T
h
e
f
i
v
e
-
f
o
l
d
cr
o
s
s
-
v
alid
atio
n
r
esu
lts
d
em
o
n
s
tr
ate
th
e
p
r
o
p
o
s
ed
r
ec
o
g
n
itio
n
m
o
d
el'
s
h
ig
h
r
eliab
ilit
y
an
d
g
e
n
er
aliza
tio
n
ca
p
ab
ilit
y
ac
r
o
s
s
b
o
th
d
atasets
.
T
h
e
m
in
im
al
s
tan
d
ar
d
d
ev
iatio
n
i
n
ac
cu
r
ac
y
an
d
lo
s
s
f
u
r
th
er
̶
co
n
f
i
r
m
s
̶
th
e̶
m
o
d
el’
s
̶
s
tab
ilit
y
̶
ac
r
o
s
s
̶
d
if
f
er
en
t̶
f
o
ld
s
.
̶
T
h
ese̶
f
in
d
in
g
s
̶
in
d
icat
e̶
th
at̶
th
e̶
m
o
d
el
̶
ef
f
ec
tiv
ely
lear
n
s
d
is
cr
im
in
ati
v
e
f
ea
tu
r
es,
en
s
u
r
in
g
r
o
b
u
s
t r
e
co
g
n
itio
n
ac
r
o
s
s
d
iv
er
s
e
h
an
d
wr
itten
s
cr
ip
ts
.
4.
CO
NCLU
SI
O
N
T
h
is
s
tu
d
y
p
r
esen
ts
a
s
ig
n
if
ic
an
t
ad
v
an
ce
m
en
t
in
o
f
f
lin
e
HW
R
b
y
in
teg
r
atin
g
ViT
ca
p
a
b
ilit
ies
f
o
r
f
ea
tu
r
e
ex
tr
ac
tio
n
.
T
h
e
p
r
o
p
o
s
ed
ViT
A
-
HW
R
m
o
d
el
ef
f
ec
tiv
ely
id
en
tifie
s
an
d
ex
tr
ac
ts
ess
en
tial
f
ea
tu
r
es,
en
ab
lin
g
a
m
o
r
e
co
m
p
r
eh
e
n
s
iv
e
ev
alu
atio
n
th
r
o
u
g
h
f
o
u
r
r
ec
o
g
n
itio
n
s
tr
ateg
ies.
T
h
e
c
o
m
p
ar
ativ
e
a
n
aly
s
is
d
em
o
n
s
tr
ated
th
at
th
e
AFFNN
m
o
d
el
ac
h
iev
ed
s
u
p
er
io
r
p
er
f
o
r
m
a
n
ce
,
attr
ib
u
ted
to
its
atten
tio
n
m
ec
h
an
is
m
th
at
d
y
n
am
ically
ass
ig
n
s
im
p
o
r
tan
ce
t
o
d
if
f
er
en
t
r
e
g
io
n
s
w
ith
in
th
e
f
ea
tu
r
e
m
atr
ix
.
T
h
is
ap
p
r
o
ac
h
e
n
h
an
ce
s
th
e
m
o
d
el'
s
ab
ilit
y
to
ca
p
tu
r
e
cr
u
cial
p
atter
n
s
i
n
h
a
n
d
wr
itte
n
tex
t.
Fu
t
u
r
e
wo
r
k
will
f
o
cu
s
o
n
f
u
r
t
h
er
r
e
f
in
in
g
th
e
p
r
o
p
o
s
ed
atten
tio
n
m
ec
h
a
n
is
m
an
d
ex
ten
d
in
g
its
ev
alu
a
tio
n
ac
r
o
s
s
m
o
r
e
d
iv
er
s
e
an
d
ch
allen
g
in
g
d
atasets
to
en
h
a
n
ce
r
o
b
u
s
tn
ess
an
d
a
d
ap
tab
ilit
y
.
I
n
ad
d
itio
n
,
th
e
c
o
m
p
ar
ativ
e
a
n
aly
s
is
will
b
e
e
x
p
an
d
ed
to
in
clu
d
e
r
ec
en
t
tr
an
s
f
o
r
m
er
-
b
ased
an
d
h
y
b
r
i
d
ar
ch
itectu
r
es,
s
u
ch
as
T
r
OC
R
an
d
s
win
tr
an
s
f
o
r
m
e
r
,
en
ab
lin
g
a
m
o
r
e
co
m
p
r
eh
e
n
s
iv
e
ass
ess
m
en
t
o
f
p
er
f
o
r
m
an
ce
tr
a
d
e
-
o
f
f
s
.
A
d
e
ep
er
in
v
esti
g
atio
n
in
t
o
co
m
p
u
tatio
n
al
ef
f
icien
cy
,
er
r
o
r
ch
a
r
ac
ter
is
tics
,
an
d
r
o
b
u
s
tn
ess
u
n
d
er
ch
allen
g
in
g
c
o
n
d
itio
n
s
will
al
s
o
b
e
co
n
d
u
cte
d
to
b
etter
u
n
d
er
s
tan
d
th
e̶
m
o
d
el’
s
̶
p
r
ac
tical̶
ap
p
licab
ilit
y
̶
in
̶
r
ea
l
-
wo
r
ld
s
ce
n
ar
i
o
s
.
F
UNDING
I
NF
O
R
M
A
T
I
O
N
Au
th
o
r
s
s
tate
n
o
f
u
n
d
in
g
in
v
o
lv
ed
.
AUTHO
R
CO
NT
RI
B
UT
I
O
NS ST
A
T
E
M
E
N
T
T
h
is
jo
u
r
n
al
u
s
es
th
e
C
o
n
tr
ib
u
to
r
R
o
les
T
ax
o
n
o
m
y
(
C
R
ed
iT)
to
r
ec
o
g
n
ize
in
d
iv
id
u
al
au
th
o
r
co
n
tr
ib
u
tio
n
s
,
r
ed
u
ce
au
th
o
r
s
h
ip
d
is
p
u
tes,
an
d
f
ac
ilit
ate
co
llab
o
r
atio
n
.
Na
m
e
o
f
Aut
ho
r
C
M
So
Va
Fo
I
R
D
O
E
Vi
Su
P
Fu
An
ju
Ar
an
g
il T
h
az
h
ath
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
B
in
u
Po
o
th
ak
u
z
h
iy
il
C
h
ac
k
o
✓
✓
✓
✓
✓
✓
Mo
h
am
ed
B
ash
ee
r
Kizh
ak
k
e
Par
am
b
ath
✓
✓
✓
✓
✓
✓
C
:
C
o
n
c
e
p
t
u
a
l
i
z
a
t
i
o
n
M
:
M
e
t
h
o
d
o
l
o
g
y
So
:
So
f
t
w
a
r
e
Va
:
Va
l
i
d
a
t
i
o
n
Fo
:
Fo
r
mal
a
n
a
l
y
s
i
s
I
:
I
n
v
e
s
t
i
g
a
t
i
o
n
R
:
R
e
so
u
r
c
e
s
D
:
D
a
t
a
C
u
r
a
t
i
o
n
O
:
W
r
i
t
i
n
g
-
O
r
i
g
i
n
a
l
D
r
a
f
t
E
:
W
r
i
t
i
n
g
-
R
e
v
i
e
w
&
E
d
i
t
i
n
g
Vi
:
Vi
su
a
l
i
z
a
t
i
o
n
Su
:
Su
p
e
r
v
i
s
i
o
n
P
:
P
r
o
j
e
c
t
a
d
mi
n
i
st
r
a
t
i
o
n
Fu
:
Fu
n
d
i
n
g
a
c
q
u
i
si
t
i
o
n
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
1
5
,
No
.
3
,
J
u
n
e
2
0
2
6
:
2
6
5
5
-
2
6
6
3
2662
CO
NF
L
I
C
T
O
F
I
N
T
E
R
E
S
T
ST
A
T
E
M
E
NT
Au
th
o
r
s
s
tate
n
o
co
n
f
lict o
f
in
t
er
est.
DATA AV
AI
L
AB
I
L
I
T
Y
T
h
e
d
ata
t
h
at
s
u
p
p
o
r
t
th
e
f
i
n
d
in
g
s
o
f
th
is
s
tu
d
y
ar
e
av
a
ilab
le
f
r
o
m
th
e
c
o
r
r
esp
o
n
d
in
g
au
th
o
r
,
[
AAT
]
,
u
p
o
n
r
ea
s
o
n
ab
le
r
eq
u
e
s
t.
RE
F
E
R
E
NC
E
S
[
1
]
K
.
̶
M
a
n
j
u
sh
a
,
̶
M
.
̶
A
.
̶
K
u
m
a
r
,
̶
a
n
d
̶
K
.
̶
P
.
̶
S
o
m
a
n
,
̶
“
O
n
̶
d
e
v
e
l
o
p
i
n
g
̶
h
a
n
d
w
r
i
t
t
e
n
̶
c
h
a
r
a
c
t
e
r
̶
i
m
a
g
e
̶
d
a
t
a
b
a
s
e
̶
f
o
r
̶
M
a
l
a
y
a
l
a
m̶
l
a
n
g
u
a
g
e
̶
scri
p
t
,
”
̶
En
g
i
n
e
e
r
i
n
g
S
c
i
e
n
c
e
a
n
d
T
e
c
h
n
o
l
o
g
y
,
a
n
I
n
t
e
r
n
a
t
i
o
n
a
l
J
o
u
rn
a
l
,
v
o
l
.
2
2
,
n
o
.
2
,
p
p
.
6
3
7
–
6
4
5
,
2
0
1
9
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
j
e
s
t
c
h
.
2
0
1
8
.
1
0
.
0
1
1
.
[
2
]
M
.
C
h
e
r
i
e
t
,
N
.
K
h
a
r
m
a
,
C
.
-
L.
L
i
u
,
a
n
d
C
.
S
u
e
n
,
C
h
a
r
a
c
t
e
r
r
e
c
o
g
n
i
t
i
o
n
s
y
st
e
m
s:
A
g
u
i
d
e
f
o
r
st
u
d
e
n
t
s
a
n
d
p
r
a
c
t
i
t
i
o
n
e
rs
.
Jo
h
n
W
i
l
e
y
& So
n
s
,
2
0
0
7
.
[
3
]
S
.
̶
M
a
d
h
v
a
n
a
t
h
̶
a
n
d
̶
V
.
̶
G
o
v
i
n
d
a
r
a
j
u
,
̶
“
T
h
e
̶
r
o
l
e
̶
o
f
̶
h
o
l
i
s
t
i
c
̶
p
a
r
a
d
i
g
ms
̶
i
n
̶
h
a
n
d
w
r
i
t
t
e
n
̶
w
o
r
d
̶
r
e
c
o
g
n
i
t
i
o
n
,
”
̶
I
EE
E
T
r
a
n
s
a
c
t
i
o
n
s
o
n
Pa
t
t
e
r
n
An
a
l
y
si
s
a
n
d
Ma
c
h
i
n
e
I
n
t
e
l
l
i
g
e
n
c
e
,
v
o
l
.
2
3
,
n
o
.
2
,
p
p
.
1
4
9
–
1
6
4
,
2
0
0
1
,
d
o
i
:
1
0
.
1
1
0
9
/
3
4
.
9
0
8
9
6
6
.
[
4
]
S
.
̶
M
a
l
a
k
a
r
,
̶
S
.
̶
S
a
h
o
o
,
̶
A
.
̶
C
h
a
k
r
a
b
o
r
t
y
,
̶
R
.
̶
S
a
r
k
a
r
,
̶
a
n
d
̶
M
.
̶
N
a
si
p
u
r
i
,
̶
“
H
a
n
d
w
r
i
t
t
e
n
̶
A
r
a
b
i
c
̶
a
n
d
̶
R
o
m
a
n
̶
w
o
r
d
̶
r
e
c
o
g
n
i
t
i
o
n
̶
u
si
n
g
̶
h
o
l
i
s
t
i
c
̶
a
p
p
r
o
a
c
h
,
”
̶
V
i
su
a
l
C
o
m
p
u
t
e
r
,
v
o
l
.
3
9
,
n
o
.
7
,
p
p
.
2
9
0
9
–
2
9
3
2
,
J
u
l
.
2
0
2
3
,
d
o
i
:
1
0
.
1
0
0
7
/
s
0
0
3
7
1
-
022
-
0
2
5
0
0
-
7.
[
5
]
S
.
̶
M
a
l
a
k
a
r
,
̶
M
.
̶
G
h
o
s
h
,
̶
S
.
̶
B
h
o
w
mi
k
,
̶
R
.
̶
S
a
r
k
a
r
,
̶
a
n
d
̶
M
.
̶
N
a
s
i
p
u
r
i
,
̶
“
A
̶
G
A
b
a
se
d
h
i
e
r
a
r
c
h
i
c
a
l
f
e
a
t
u
r
e
se
l
e
c
t
i
o
n
a
p
p
r
o
a
c
h
f
o
r
h
a
n
d
w
r
i
t
t
e
n
̶
w
o
r
d
̶
r
e
c
o
g
n
i
t
i
o
n
,
”
̶
N
e
u
r
a
l
C
o
m
p
u
t
i
n
g
a
n
d
A
p
p
l
i
c
a
t
i
o
n
s
,
v
o
l
.
3
2
,
n
o
.
7
,
p
p
.
2
5
3
3
–
2
5
5
2
,
A
p
r
.
2
0
2
0
,
d
o
i
:
1
0
.
1
0
0
7
/
s
0
0
5
2
1
-
018
-
3
9
3
7
-
8.
[
6
]
S
.
̶
B
h
o
w
m
i
k
,
̶
S
.
̶
M
a
l
a
k
a
r
,
̶
R
.
̶
S
a
r
k
a
r
,
̶
S
.
̶
B
a
s
u
,
̶
M
.
̶
K
u
n
d
u
,
̶
a
n
d
̶
M
.
̶
N
a
s
i
p
u
r
i
,
̶
“
O
f
f
-
l
i
n
e
B
a
n
g
l
a
h
a
n
d
w
r
i
t
t
e
n
w
o
r
d
r
e
c
o
g
n
i
t
i
o
n
:
a
h
o
l
i
s
t
i
c
a
p
p
r
o
a
c
h
,
”
̶
N
e
u
r
a
l
C
o
m
p
u
t
i
n
g
a
n
d
A
p
p
l
i
c
a
t
i
o
n
s
,
v
o
l
.
3
1
,
n
o
.
1
0
,
p
p
.
5
7
8
3
–
5
7
9
8
,
O
c
t
.
2
0
1
9
,
d
o
i
:
1
0
.
1
0
0
7
/
s
0
0
5
2
1
-
0
1
8
-
3
3
8
9
-
1.
[
7
]
S
.
S
a
h
o
o
e
t
a
l
.
,
̶
“
H
a
n
d
w
r
i
t
t
e
n
̶
B
a
n
g
l
a
̶
w
o
r
d
̶
r
e
c
o
g
n
i
t
i
o
n
̶
u
s
i
n
g
̶
n
e
g
a
t
i
v
e
̶
r
e
f
r
a
c
t
i
o
n
̶
b
a
se
d
̶
s
h
a
p
e
̶
t
r
a
n
sf
o
r
m
a
t
i
o
n
,
”
̶
J
o
u
r
n
a
l
o
f
I
n
t
e
l
l
i
g
e
n
t
a
n
d
F
u
zzy
S
y
st
e
m
s
,
v
o
l
.
3
5
,
n
o
.
2
,
p
p
.
1
7
6
5
–
1
7
7
7
,
2
0
1
8
,
d
o
i
:
1
0
.
3
2
3
3
/
JI
F
S
-
1
6
9
7
1
2
.
[
8
]
K
.
̶
S
h
a
r
m
a
,
̶
P
.
̶
K
.
̶
S
a
r
a
n
g
i
,
̶
L
.
̶
R
a
n
i
,
̶
G
.
̶
S
i
n
g
h
,
̶
A
.
̶
K
.
̶
S
a
h
o
o
,
̶
a
n
d
̶
B
.
̶
P
.
̶
R
a
t
h
,
̶
“
H
a
n
d
w
r
i
t
t
e
n
̶
d
i
g
i
t
c
l
a
ss
i
f
i
c
a
t
i
o
n
u
si
n
g
H
O
G
f
e
a
t
u
r
e
s
a
n
d
S
V
M
c
l
a
ssi
f
i
e
r
,
”
̶
i
n
̶
2
0
2
2
2
n
d
I
n
t
e
rn
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
Ad
v
a
n
c
e
C
o
m
p
u
t
i
n
g
a
n
d
I
n
n
o
v
a
t
i
v
e
T
e
c
h
n
o
l
o
g
i
e
s
i
n
E
n
g
i
n
e
e
ri
n
g
,
I
C
AC
I
T
E
2
0
2
2
,
2
0
2
2
,
p
p
.
2
0
7
1
–
2
0
7
4
.
d
o
i
:
1
0
.
1
1
0
9
/
I
C
A
C
I
TE5
3
7
2
2
.
2
0
2
2
.
9
8
2
3
7
8
2
.
[
9
]
N
.
A
l
z
r
r
o
g
,
J
.
F
.
B
o
u
s
q
u
e
t
,
a
n
d
I
.
E
l
-
F
e
g
h
i
,
̶
“
D
e
e
p
̶
l
e
a
r
n
i
n
g
a
p
p
l
i
c
a
t
i
o
n
f
o
r
h
a
n
d
w
r
i
t
t
e
n
A
r
a
b
i
c
w
o
r
d
r
e
c
o
g
n
i
t
i
o
n
,
”
̶
i
n
̶
C
a
n
a
d
i
a
n
C
o
n
f
e
re
n
c
e
o
n
El
e
c
t
ri
c
a
l
a
n
d
C
o
m
p
u
t
e
r E
n
g
i
n
e
e
r
i
n
g
,
2
0
2
2
,
p
p
.
9
5
–
1
0
0
,
d
o
i
:
1
0
.
1
1
0
9
/
C
C
EC
E4
9
3
5
1
.
2
0
2
2
.
9
9
1
8
3
7
5
.
[
1
0
]
S
.
̶
K
h
o
sr
a
v
i
̶
a
n
d
̶
A
.
̶
C
h
a
l
e
c
h
a
l
e
,
̶
“
C
h
i
m
p
̶
o
p
t
i
mi
z
a
t
i
o
n
a
l
g
o
r
i
t
h
m
t
o
o
p
t
i
mi
z
e
a
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
u
r
a
l
n
e
t
w
o
r
k
f
o
r
r
e
c
o
g
n
i
z
i
n
g
P
e
r
si
a
n
/
A
r
a
b
i
c
h
a
n
d
w
r
i
t
t
e
n
w
o
r
d
s
,
”
̶
M
a
t
h
e
m
a
t
i
c
a
l
Pro
b
l
e
m
s
i
n
En
g
i
n
e
e
ri
n
g
,
v
o
l
.
2
0
2
2
,
p
p
.
1
–
1
2
,
A
p
r
.
2
0
2
2
,
d
o
i
:
1
0
.
1
1
5
5
/
2
0
2
2
/
4
8
9
4
9
2
2
.
[
1
1
]
A
.
̶
Z
o
h
r
e
v
a
n
d
̶
a
n
d
̶
Z.
̶
I
man
i
,
̶
“
H
o
l
i
s
t
i
c
̶
P
e
r
s
i
a
n
̶
h
a
n
d
w
r
i
t
t
e
n
̶
w
o
r
d
̶
r
e
c
o
g
n
i
t
i
o
n
̶
u
si
n
g
̶
c
o
n
v
o
l
u
t
i
o
n
a
l
̶
n
e
u
r
a
l
̶
n
e
t
w
o
r
k
,
”
̶
I
n
t
e
r
n
a
t
i
o
n
a
l
J
o
u
rn
a
l
o
f
En
g
i
n
e
e
ri
n
g
,
T
r
a
n
s
a
c
t
i
o
n
s
B:
Ap
p
l
i
c
a
t
i
o
n
s
,
v
o
l
.
3
4
,
n
o
.
8
,
p
p
.
2
0
2
8
–
2
0
3
7
,
2
0
2
1
,
d
o
i
:
1
0
.
5
8
2
9
/
i
j
e
.
2
0
2
1
.
3
4
.
0
8
b
.
2
4
.
[
1
2
]
F
.
̶
A
b
d
u
r
a
h
ma
n
,
̶
E.
̶
S
i
sa
y
,
̶
a
n
d
̶
K
.
̶
A
.
̶
F
a
n
t
e
,
̶
“
A
H
W
R
-
N
e
t
:
o
f
f
l
i
n
e
h
a
n
d
w
r
i
t
t
e
n
A
m
h
a
r
i
c
w
o
r
d
r
e
c
o
g
n
i
t
i
o
n
u
s
i
n
g
c
o
n
v
o
l
u
t
i
o
n
a
l
r
e
c
u
r
r
e
n
t
̶
n
e
u
r
a
l
̶
n
e
t
w
o
r
k
,
”
̶
S
N
A
p
p
l
i
e
d
S
c
i
e
n
c
e
s
,
v
o
l
.
3
,
n
o
.
8
,
p
p
.
1
–
1
1
,
A
u
g
.
2
0
2
1
,
d
o
i
:
1
0
.
1
0
0
7
/
s
4
2
4
5
2
-
0
2
1
-
0
4
7
4
2
-
x.
[
1
3
]
R
.
̶
P
r
a
m
a
n
i
k
̶
a
n
d
̶
S
.
̶
B
a
g
,
̶
“
H
a
n
d
w
r
i
t
t
e
n
̶
B
a
n
g
l
a
̶
c
i
t
y
̶
n
a
me
̶
w
o
r
d
̶
r
e
c
o
g
n
i
t
i
o
n
̶
u
si
n
g
̶
C
N
N
-
b
a
se
d
̶
t
r
a
n
sf
e
r
̶
l
e
a
r
n
i
n
g
̶
a
n
d
̶
F
C
N
,
”
̶
N
e
u
r
a
l
C
o
m
p
u
t
i
n
g
a
n
d
Ap
p
l
i
c
a
t
i
o
n
s
,
v
o
l
.
3
3
,
n
o
.
1
5
,
p
p
.
9
3
2
9
–
9
3
4
1
,
A
u
g
.
2
0
2
1
,
d
o
i
:
1
0
.
1
0
0
7
/
s
0
0
5
2
1
-
0
2
1
-
0
5
6
9
3
-
5.
[
1
4
]
M
.
̶
B
o
n
y
a
n
i
,
̶
S
.
̶
J
a
h
a
n
g
a
r
d
,
̶
a
n
d
̶
M
.
̶
D
a
n
e
s
h
ma
n
d
,
̶
“
P
e
r
si
a
n
̶
h
a
n
d
w
r
i
t
t
e
n
̶
d
i
g
i
t
,
̶
c
h
a
r
a
c
t
e
r
̶
a
n
d
̶
w
o
r
d
̶
r
e
c
o
g
n
i
t
i
o
n
̶
u
si
n
g
̶
d
e
e
p
̶
l
e
a
r
n
i
n
g
,
”
̶
I
n
t
e
rn
a
t
i
o
n
a
l
J
o
u
r
n
a
l
o
n
D
o
c
u
m
e
n
t
A
n
a
l
y
si
s
a
n
d
R
e
c
o
g
n
i
t
i
o
n
,
v
o
l
.
2
4
,
n
o
.
1
–
2
,
p
p
.
1
3
3
–
1
4
3
,
2
0
2
1
,
d
o
i
:
1
0
.
1
0
0
7
/
s
1
0
0
3
2
-
021
-
0
0
3
6
8
-
2.
[
1
5
]
P
.
̶
J.̶
Ji
n
o
,
̶
K
.
̶
B
a
l
a
k
r
i
s
h
n
a
n
,
̶
a
n
d
̶
U
.
̶
B
h
a
t
t
a
c
h
a
r
y
a
,
̶
“
O
f
f
l
i
n
e
̶
h
a
n
d
w
r
i
t
t
e
n
̶
M
a
l
a
y
a
l
a
m
w
o
r
d
̶
r
e
c
o
g
n
i
t
i
o
n
̶
u
si
n
g
̶
a
̶
d
e
e
p
̶
a
r
c
h
i
t
e
c
t
u
r
e
,
”
̶
i
n
̶
Ad
v
a
n
c
e
s i
n
I
n
t
e
l
l
i
g
e
n
t
S
y
st
e
m
s
a
n
d
C
o
m
p
u
t
i
n
g
,
S
p
r
i
n
g
e
r
S
i
n
g
a
p
o
r
e
,
2
0
1
9
,
p
p
.
9
1
3
–
9
2
5
.
d
o
i
:
1
0
.
1
0
0
7
/
9
7
8
-
9
8
1
-
13
-
1
5
9
2
-
3
_
7
3
.
[
1
6
]
D
.
̶
D
a
s,
̶
D
.
̶
R
.
̶
N
a
y
a
k
,
̶
R
.
̶
D
a
s
h
,
̶
B
.
̶
M
a
j
h
i
,
̶
a
n
d
̶
Y
.
̶
D
.
̶
Z
h
a
n
g
,
̶
“
H
-
W
o
r
d
N
e
t
:
a
h
o
l
i
st
i
c
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
u
r
a
l
n
e
t
w
o
r
k
a
p
p
r
o
a
c
h
f
o
r
h
a
n
d
w
r
i
t
t
e
n
w
o
r
d
r
e
c
o
g
n
i
t
i
o
n
,
”
̶
I
ET I
m
a
g
e
Pr
o
c
e
ssi
n
g
,
v
o
l
.
1
4
,
n
o
.
9
,
p
p
.
1
7
9
4
–
1
8
0
5
,
J
u
l
.
2
0
2
0
,
d
o
i
:
1
0
.
1
0
4
9
/
i
e
t
-
i
p
r
.
2
0
1
9
.
1
3
9
8
.
[
1
7
]
S
.
̶
M
a
l
a
k
a
r
,
̶
S
.
̶
P
a
u
l
,
̶
S
.
̶
K
u
n
d
u
,
̶
S
.
̶
B
h
o
w
m
i
k
,
̶
R
.
̶
S
a
r
k
a
r
,
̶
a
n
d
̶
M
.
̶
N
a
si
p
u
r
i
,
̶
“
H
a
n
d
w
r
i
t
t
e
n
̶
w
o
r
d
̶
r
e
c
o
g
n
i
t
i
o
n
̶
u
s
i
n
g
̶
l
o
t
t
e
r
y
̶
t
i
c
k
e
t
̶
h
y
p
o
t
h
e
s
i
s̶
b
a
se
d
̶
p
r
u
n
e
d
̶
C
N
N
̶
mo
d
e
l
:
̶
a
̶
n
e
w
̶
b
e
n
c
h
mark
̶
o
n
̶
C
M
A
T
ER
d
b
2
.
1
.
2
,
”
̶
N
e
u
r
a
l
C
o
m
p
u
t
i
n
g
a
n
d
Ap
p
l
i
c
a
t
i
o
n
s
,
v
o
l
.
3
2
,
n
o
.
1
8
,
p
p
.
1
5
2
0
9
–
1
5
2
2
0
,
S
e
p
.
2
0
2
0
,
d
o
i
:
1
0
.
1
0
0
7
/
s
0
0
5
2
1
-
020
-
0
4
8
7
2
-
0.
[
1
8
]
M
.
̶
A
w
n
i
,
̶
M
.
̶
I
.
̶
K
h
a
l
i
l
,
̶
a
n
d
̶
H
.
̶
M
.
̶
A
b
b
a
s,
̶
“
D
e
e
p
-
l
e
a
r
n
i
n
g
e
n
s
e
m
b
l
e
f
o
r
o
f
f
l
i
n
e
A
r
a
b
i
c
h
a
n
d
w
r
i
t
t
e
n
̶
w
o
r
d
s
̶
r
e
c
o
g
n
i
t
i
o
n
,
”
̶
i
n
̶
2
0
1
9
1
4
t
h
I
n
t
e
rn
a
t
i
o
n
a
l
C
o
n
f
e
r
e
n
c
e
o
n
C
o
m
p
u
t
e
r
E
n
g
i
n
e
e
ri
n
g
a
n
d
S
y
s
t
e
m
s
(
I
C
C
ES
)
,
C
a
i
r
o
,
Eg
y
p
t
,
2
0
1
9
,
p
p
.
4
0
–
45
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
C
ES4
8
9
6
0
.
2
0
1
9
.
9
0
6
8
1
8
4
.
[
1
9
]
M
.
Li
e
t
a
l
.
,
̶
“
Tr
O
C
R
:
t
r
a
n
sf
o
r
m
e
r
-
b
a
se
d
o
p
t
i
c
a
l
c
h
a
r
a
c
t
e
r
r
e
c
o
g
n
i
t
i
o
n
w
i
t
h
p
r
e
-
t
r
a
i
n
e
d
m
o
d
e
l
s
,
”
̶
i
n
̶
Pro
c
e
e
d
i
n
g
s
o
f
t
h
e
AAAI
C
o
n
f
e
re
n
c
e
o
n
Art
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
,
2
0
2
3
,
p
p
.
1
3
0
9
4
–
1
3
1
0
2
,
d
o
i
:
1
0
.
1
6
0
9
/
a
a
a
i
.
v
3
7
i
1
1
.
2
6
5
3
8
.
[
2
0
]
A
.
M
o
s
t
a
f
a
e
t
a
l
.
,
̶
“
O
C
F
o
r
mer:
̶
a
t
r
a
n
s
f
o
r
mer
-
b
a
se
d
m
o
d
e
l
f
o
r
A
r
a
b
i
c
h
a
n
d
w
r
i
t
t
e
n
t
e
x
t
r
e
c
o
g
n
i
t
i
o
n
,
”
̶
i
n
̶
2
0
2
1
I
n
t
e
rn
a
t
i
o
n
a
l
Mo
b
i
l
e
,
I
n
t
e
l
l
i
g
e
n
t
,
a
n
d
U
b
i
q
u
i
t
o
u
s C
o
m
p
u
t
i
n
g
C
o
n
f
e
re
n
c
e
,
MIU
C
C
2
0
2
1
,
2
0
2
1
,
p
p
.
1
8
2
–
1
8
6
,
d
o
i
:
1
0
.
1
1
0
9
/
M
I
U
C
C
5
2
5
3
8
.
2
0
2
1
.
9
4
4
7
6
0
8
.
[
2
1
]
K
.
̶
B
a
r
r
e
r
e
,
̶
Y
.
̶
S
o
u
l
l
a
r
d
,
̶
A
.
̶
Le
mai
t
r
e
,
̶
a
n
d
̶
B
.
̶
C
o
ü
a
s
n
o
n
,
̶
“
A
̶
l
i
g
h
t
t
r
a
n
sf
o
r
m
e
r
-
b
a
se
d
a
r
c
h
i
t
e
c
t
u
r
e
f
o
r
h
a
n
d
w
r
i
t
t
e
n
t
e
x
t
r
e
c
o
g
n
i
t
i
o
n
,”̶
D
o
c
u
m
e
n
t
An
a
l
y
si
s
S
y
s
t
e
m
s:
1
5
t
h
I
APR
I
n
t
e
r
n
a
t
i
o
n
a
l
Wo
r
k
s
h
o
p
,
D
AS
2
0
2
2
,
C
h
a
m
:
S
p
r
i
n
g
e
r
,
2
0
2
2
,
p
p
.
2
7
5
–
2
9
0
,
d
o
i
:
1
0
.
1
0
0
7
/
9
7
8
-
3
-
031
-
0
6
5
5
5
-
2
_
1
9
.
[
2
2
]
A
.
V
a
sw
a
n
i
e
t
a
l
.
,
̶
“
A
t
t
e
n
t
i
o
n
̶
i
s̶
a
l
l
̶
y
o
u
̶
n
e
e
d
,
”
̶
A
d
v
a
n
c
e
s
i
n
N
e
u
r
a
l
I
n
f
o
rm
a
t
i
o
n
Pr
o
c
e
ssi
n
g
S
y
s
t
e
m
s
,
p
p
.
5
9
9
9
–
6
0
0
9
,
2
0
1
7
,
d
o
i
:
1
0
.
1
2
0
1
/
9
7
8
1
0
0
3
5
6
1
4
6
0
-
19.
[
2
3
]
A
.
D
o
so
v
i
t
s
k
i
y
e
t
a
l
.
,
̶
“
A
n
i
ma
g
e
i
s
w
o
r
t
h
16
x
1
6
w
o
r
d
s
:
t
r
a
n
sf
o
r
m
e
r
s
f
o
r
i
m
a
g
e
r
e
c
o
g
n
i
t
i
o
n
a
t
sca
l
e
,
”
̶
2
0
2
1
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
L
e
a
rn
i
n
g
Re
p
res
e
n
t
a
t
i
o
n
s
,
2
0
2
1
,
p
p
.
1
.
2
1
.
[
2
4
]
D
.
̶
S
v
o
z
i
l
,
̶
V
.
̶
K
v
a
s
n
i
c
k
a
,
̶
a
n
d
̶
J.̶
P
o
sp
i
c
h
a
l
,
̶
“
I
n
t
r
o
d
u
c
t
i
o
n
̶
t
o
̶
mu
l
t
i
-
l
a
y
e
r
f
e
e
d
-
f
o
r
w
a
r
d
̶
n
e
u
r
a
l
̶
n
e
t
w
o
r
k
s,
”
̶
C
h
e
m
o
m
e
t
ri
c
s
a
n
d
I
n
t
e
l
l
i
g
e
n
t
L
a
b
o
r
a
t
o
r
y
S
y
st
e
m
s
,
v
o
l
.
3
9
,
n
o
.
1
,
p
p
.
4
3
–
6
2
,
1
9
9
7
,
d
o
i
:
1
0
.
1
0
1
6
/
S
0
1
6
9
-
7
4
3
9
(
9
7
)
0
0
0
6
1
-
0
.
[
2
5
]
A
.
A
l
-
S
a
b
a
a
w
i
,
H
.
M
.
I
b
r
a
h
i
m,
Z
.
M
.
A
r
k
a
h
,
M
.
A
l
-
A
m
i
d
i
e
,
̶
a
n
d
̶
L.
̶
A
l
z
u
b
a
i
d
i
,
̶
“
A
m
e
n
d
e
d
̶
c
o
n
v
o
l
u
t
i
o
n
a
l
̶
n
e
u
r
a
l
̶
n
e
t
w
o
r
k
̶
w
i
t
h
̶
g
l
o
b
a
l
̶
a
v
e
r
a
g
e
̶
p
o
o
l
i
n
g
̶
f
o
r
̶
i
ma
g
e
̶
c
l
a
ssi
f
i
c
a
t
i
o
n
,
”
̶
i
n
̶
I
n
t
e
l
l
i
g
e
n
t
S
y
st
e
m
s
D
e
si
g
n
a
n
d
A
p
p
l
i
c
a
t
i
o
n
s
,
C
h
a
m
:
S
p
r
i
n
g
e
r
I
n
t
e
r
n
a
t
i
o
n
a
l
P
u
b
l
i
s
h
i
n
g
,
2
0
2
1
,
p
p
.
1
7
1
–
1
8
0
,
d
o
i
:
1
0
.
1
0
0
7
/
9
7
8
-
3
-
0
3
0
-
7
1
1
8
7
-
0
_
1
6
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
V
is
io
n
tr
a
n
s
fo
r
mer a
n
d
h
yb
r
id
mo
d
els fo
r
Ma
la
ya
la
m
h
a
n
d
w
r
itten
…
(
A
n
ju
A
r
a
n
g
il Th
a
z
h
a
th
)
2663
B
I
O
G
RAP
H
I
E
S O
F
AUTH
O
RS
Anju
Ar
a
n
g
il
Th
a
z
h
a
t
h
is
a
re
se
a
rc
h
sc
h
o
lar
in
th
e
De
p
a
rtm
e
n
t
o
f
Co
m
p
u
ter
S
c
ien
c
e
a
t
S
u
ll
a
m
u
ss
a
lam
S
c
ien
c
e
Co
ll
e
g
e
,
Are
e
k
o
d
e
,
M
a
la
p
p
u
ra
m
,
Ke
ra
la,
In
d
ia.
S
h
e
is
a
ls
o
wo
rk
i
n
g
a
s
a
n
a
ss
istan
t
p
ro
fe
ss
o
r
in
d
a
ta
sc
ien
c
e
a
n
d
a
n
a
ly
ti
c
s
a
t
th
e
Un
iv
e
rsit
y
o
f
Ca
li
c
u
t
,
Ke
ra
la.
He
r
re
se
a
rc
h
in
tere
sts
in
c
lu
d
e
a
rti
ficia
l
in
telli
g
e
n
c
e
,
m
a
c
h
i
n
e
lea
rn
in
g
,
d
e
e
p
lea
rn
i
n
g
,
ima
g
e
p
ro
c
e
ss
in
g
,
a
n
d
h
a
n
d
writt
e
n
d
o
c
u
m
e
n
t
ima
g
e
a
n
a
ly
sis,
wi
th
a
f
o
c
u
s
o
n
M
a
lay
a
lam
lan
g
u
a
g
e
wo
rd
re
c
o
g
n
it
io
n
.
S
h
e
h
a
s b
e
e
n
a
c
ti
v
e
ly
e
n
g
a
g
e
d
i
n
re
se
a
rc
h
a
n
d
a
c
a
d
e
m
ic i
n
it
iati
v
e
s
in
th
e
field
o
f
in
telli
g
e
n
t
sy
ste
m
s
a
n
d
p
a
tt
e
rn
re
c
o
g
n
i
ti
o
n
.
S
h
e
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
a
n
ju
a
t@ss
c
o
ll
e
g
e
.
a
c
.
in
.
Dr
.
Bin
u
Po
o
th
a
k
u
z
h
i
y
il
Cha
c
k
o
is
a
n
a
ss
o
c
iate
p
r
o
fe
ss
o
r
i
n
t
h
e
De
p
a
rtme
n
t
o
f
Co
m
p
u
ter
S
c
ien
c
e
a
t
P
ra
jy
o
ti
Nik
e
tan
Co
ll
e
g
e
,
a
ffil
iate
d
wit
h
th
e
Un
iv
e
rsity
o
f
Ca
li
c
u
t,
Ke
ra
la,
In
d
ia.
He
h
o
ld
s
a
d
o
c
to
r
o
f
p
h
il
o
so
p
h
y
(P
h
.
D.)
i
n
Co
m
p
u
ter
S
c
ien
c
e
fro
m
Ka
n
n
u
r
Un
iv
e
rsity
,
Ke
ra
la.
He
h
a
s
e
x
ten
s
iv
e
tea
c
h
in
g
a
n
d
re
se
a
rc
h
e
x
p
e
rien
c
e
a
n
d
c
u
rre
n
tl
y
se
rv
e
s
a
s
a
n
a
p
p
r
o
v
e
d
re
se
a
rc
h
g
u
i
d
e
in
c
o
m
p
u
ter
sc
ien
c
e
.
His
a
c
a
d
e
m
ic
a
n
d
re
se
a
rc
h
in
tere
sts
in
c
lu
d
e
a
rti
ficia
l
in
tell
ig
e
n
c
e
,
m
a
c
h
i
n
e
le
a
rn
in
g
,
d
a
ta
a
n
a
ly
ti
c
s,
a
n
d
c
o
m
p
u
tatio
n
a
l
in
tell
ig
e
n
c
e
.
He
h
a
s
g
u
i
d
e
d
se
v
e
ra
l
p
o
stg
ra
d
u
a
te
re
se
a
rc
h
p
r
o
jec
ts
a
n
d
p
u
b
li
sh
e
d
p
a
p
e
rs
in
re
p
u
te
d
n
a
ti
o
n
a
l
a
n
d
in
tern
a
ti
o
n
a
l
j
o
u
r
n
a
ls.
He
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
b
i
n
u
p
c
h
a
c
k
o
@g
m
a
il
.
c
o
m
.
Pro
f.
(Dr
.
)
Mo
h
a
m
e
d
B
a
shee
r
K
izha
k
k
e
P
a
r
a
m
b
a
t
h
is
t
h
e
p
rin
c
ip
a
l
o
f
AMAL
Co
ll
e
g
e
o
f
A
d
v
a
n
c
e
d
S
tu
d
ies
,
Nilam
b
u
r,
Ke
ra
la,
In
d
ia.
He
h
o
l
d
s
a
n
M
.
P
h
il
.
a
n
d
P
h
.
D.
i
n
Co
m
p
u
ter
S
c
ie
n
c
e
fro
m
Bh
a
ra
th
id
a
sa
n
U
n
iv
e
rsit
y
,
Tam
il
Na
d
u
,
I
n
d
ia.
Wi
t
h
e
x
ten
si
v
e
a
c
a
d
e
m
ic
a
n
d
a
d
m
in
istrativ
e
e
x
p
e
rien
c
e
,
h
e
h
a
s
d
e
m
o
n
stra
ted
e
x
c
e
ll
e
n
c
e
in
tea
c
h
in
g
a
n
d
re
se
a
rc
h
su
p
e
rv
isio
n
.
His
re
se
a
rc
h
in
tere
sts
in
c
lu
d
e
m
a
c
h
i
n
e
lea
rn
i
n
g
,
in
tern
e
t
o
f
t
h
i
n
g
s
(I
o
T),
a
rti
ficia
l
in
telli
g
e
n
c
e
,
sp
e
e
c
h
tec
h
n
o
lo
g
y
,
a
n
d
b
i
o
m
e
tri
c
se
c
u
rit
y
sy
ste
m
s.
He
h
a
s
a
c
ti
v
e
ly
g
u
i
d
e
d
se
v
e
ra
l
re
se
a
rc
h
p
ro
jec
ts
a
n
d
c
o
n
tr
ib
u
ted
to
th
e
a
d
v
a
n
c
e
m
e
n
t
o
f
e
m
e
rg
in
g
tec
h
n
o
l
o
g
ies
th
ro
u
g
h
h
is
p
u
b
li
c
a
ti
o
n
s
a
n
d
a
c
a
d
e
m
ic
in
it
iati
v
e
s.
He
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
m
b
a
sh
e
e
rk
p
@g
m
a
il
.
c
o
m
.
Evaluation Warning : The document was created with Spire.PDF for Python.