I
n
t
e
r
n
at
io
n
al
Jou
r
n
al
of
A
d
van
c
e
s
i
n
A
p
p
li
e
d
S
c
ie
n
c
e
s
(
I
JA
A
S
)
V
ol
.
14
, N
o.
4
,
D
e
c
e
m
be
r
20
25
, pp.
1
350
~
1
358
I
S
S
N
:
2252
-
8814
,
D
O
I
:
10.11591/
ij
a
a
s
.
v
14
.
i
4
.
pp
1350
-
1358
1350
Jou
r
n
al
h
om
e
page
:
ht
tp
:
//
ij
aas
.i
ae
s
c
or
e
.c
om
C
on
vol
u
t
i
on
al
n
e
u
r
al
n
e
t
w
or
k
m
od
e
l
f
or
f
i
n
ge
r
p
r
i
n
t
-
b
ase
d
ge
n
d
e
r
c
l
ass
i
f
i
c
at
i
on
u
si
n
g or
i
gi
n
al
a
n
d
d
e
gr
ad
e
d
i
m
age
s
R
is
q
y S
iwi
P
r
ad
in
i
1
, Wah
y
u
T
e
j
a K
u
s
u
m
a
1
, A
gu
n
g S
e
t
ia
B
u
d
i
2
1
P
r
ogr
a
m
of
I
n
f
or
m
a
t
i
c
s
, F
a
c
ul
t
y of
S
c
i
e
nc
e
a
nd T
e
c
hnol
ogy, I
ns
t
i
t
ut
T
e
knol
og
i
S
a
i
ns
da
n K
e
s
e
h
a
t
a
n R
S
. D
R
. S
oe
pr
a
oe
n
K
e
s
da
m
V
/
B
R
W
, M
a
l
a
ng, I
ndone
s
i
a
2
D
e
pa
r
t
e
m
e
nt
of
I
nf
or
m
a
t
i
c
s
E
ngi
ne
e
r
i
ng, F
a
c
ul
t
y o
f
C
om
put
e
r
S
c
i
e
nc
e
, U
ni
ve
r
s
i
t
a
s
B
r
a
w
i
j
a
ya
,
M
a
l
a
ng, I
ndone
s
i
a
A
r
t
ic
le
I
n
f
o
A
B
S
T
R
A
C
T
A
r
ti
c
le
h
is
to
r
y
:
R
e
c
e
iv
e
d
S
e
p 6, 2025
R
e
vi
s
e
d
O
c
t
29, 2025
A
c
c
e
pt
e
d
N
ov 4, 2025
Fingerp
rint
-
based
gender
classification
is
a
crucial
component
o
f
soft
biometrics,
providing
valuable
additional
information
to
n
arrow
the
search
space
in
forensic
investigations
and
large
-
scale
identification
sy
stems.
Although
deep
learning
models,
particularly
convolut
ional
neural
networks
(CNNs),
have
demonstrated
significant
potential
,
performance
valida
tion
is
typically
performe
d
on
high
-
quality
fingerprint
images.
This
creates
a
gap
between
laboratory
results
and
real
-
world
applications,
where
fing
erp
rint
evidence
is
often
found
in
a
degraded
state,
such
as
smudged,
distor
ted,
or
partially
damaged.
This
study
attempts
to
bridge
this
gap
by
propo
sing
a
more
realistic
training
approach.
We
design
a
lightweight
and
computat
ionally
efficient
CNN
and
train
it
on
a
comprehens
ive
co
mbined
dataset.
The
main
contribution
of
this
study
lies
in
the
data
training
st
rategy,
which explicitl
y combines real and sy
nthetically mo
dified fingerprint
i
mages
from
the
Sokoto
coventry
fingerprin
t
(SOCOFin
g)
dataset
into
a
sing
le,
unified
training
set.
Experimental
results
show
that
the
proposed
model
achieves
very
high
classifi
cation
accuracy
(97
.
39%)
on
a
test
set
th
at
also
includes
a
combination
of
original
and
degrade
d
images.
This
findi
ng
not
only
confirms
the
effectiveness
of
diverse
data
-
based
training
to
produce
more
robust
models
but
also
establishes
a
new
benchmark
for
fingerprint
-
based
gender
classification
research
under
conditions
more
representa
tive
of
practical scenarios.
K
e
y
w
o
r
d
s
:
C
om
bi
ne
d da
ta
t
r
a
in
in
g
C
onvolut
io
na
l
ne
ur
a
l
ne
twor
ks
F
in
ge
r
pr
in
t
a
na
ly
s
is
G
e
nde
r
c
la
s
s
if
ic
a
ti
on
S
O
C
O
F
in
g
This is an
open
acce
ss artic
le unde
r the
CC BY
-
SA
license.
C
or
r
e
s
pon
di
n
g A
u
th
or
:
A
gung S
e
ti
a
B
udi
D
e
pa
r
tm
e
nt
of
I
nf
or
m
a
ti
c
s
E
ngi
ne
e
r
in
g, F
a
c
ul
ty
of
C
om
put
e
r
S
c
ie
nc
e
, U
ni
ve
r
s
it
a
s
B
r
a
w
ij
a
ya
J
l.
V
e
te
r
a
n N
o.
10
-
11, M
a
la
ng, E
a
s
t
J
a
va
65145
, I
ndone
s
ia
E
m
a
il
:
a
gungs
e
ti
a
budi
@
ub.a
c
.i
d
1.
I
N
T
R
O
D
U
C
T
I
O
N
B
io
m
e
tr
ic
s
is
a
f
ie
ld
of
s
c
ie
nc
e
th
a
t
s
tu
di
e
s
th
e
pr
oc
e
s
s
of
id
e
nt
if
yi
ng
a
nd
ve
r
if
yi
ng
in
di
vi
dua
ls
by
ut
il
iz
in
g unique
a
nd dif
f
ic
ul
t
-
to
-
im
i
ta
te
phys
io
lo
gi
c
a
l
a
nd be
ha
vi
or
a
l
c
ha
r
a
c
te
r
is
ti
c
s
[
1]
. T
hi
s
f
ie
ld
i
s
t
he
m
a
in
f
ounda
ti
on
in
m
ode
r
n
id
e
nt
if
ic
a
ti
on
a
nd
s
e
c
ur
it
y
s
ys
t
e
m
s
[
2]
,
be
c
a
u
s
e
it
ut
il
iz
e
s
uni
que
phy
s
io
lo
gi
c
a
l
or
be
ha
vi
or
a
l
c
ha
r
a
c
te
r
is
ti
c
s
to
ve
r
if
y
a
pe
r
s
on'
s
id
e
nt
it
y.
I
ni
ti
a
ll
y,
bi
om
e
tr
ic
s
ys
t
e
m
s
f
oc
u
s
e
d
s
ol
e
ly
on
pr
im
a
r
y
id
e
nt
if
ie
r
s
,
s
uc
h
a
s
f
in
ge
r
pr
in
t
or
ir
is
pa
tt
e
r
ns
,
de
s
ig
n
e
d
to
pr
ovi
de
a
s
in
gl
e
id
e
nt
if
ic
a
ti
on
of
a
n
in
di
vi
dua
l.
H
ow
e
ve
r
,
r
e
c
e
nt
de
ve
lo
pm
e
nt
s
ha
ve
e
xpa
nde
d
to
in
c
lu
de
s
of
t
bi
om
e
tr
ic
s
,
w
hi
c
h
r
e
f
e
r
to
pr
e
di
c
ta
bl
e
phys
ic
a
l
or
be
ha
vi
or
a
l
a
tt
r
ib
ut
e
s
,
s
uc
h
a
s
ge
nd
e
r
,
a
ge
,
or
e
th
ni
c
it
y
[
3]
.
U
nl
ik
e
pr
im
a
r
y
id
e
nt
if
ie
r
s
,
th
e
s
e
tr
a
it
s
a
r
e
not
uni
que
to
e
ve
r
yone
.
H
ow
e
ve
r
,
th
e
a
dde
d
va
lu
e
of
s
of
t
b
io
m
e
t
r
ic
s
li
e
s
in
th
e
ir
a
bi
li
ty
to
pr
ovi
de
c
ont
e
xt
ua
l
in
f
or
m
a
ti
on t
ha
t
s
uppor
ts
t
he
i
de
nt
if
ic
a
ti
on pr
oc
e
s
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
dv A
ppl
S
c
i
I
S
S
N
:
2252
-
8814
C
onv
ol
ut
io
nal
ne
ur
al
ne
tw
o
r
k
m
ode
l
fo
r
f
in
ge
r
pr
in
t
-
bas
e
d g
e
n
de
r
c
la
s
s
if
ic
at
io
n
…
(
R
is
qy
Si
w
i
P
r
adi
ni
)
1351
I
n
f
or
e
ns
ic
in
ve
s
ti
ga
ti
ons
,
s
of
t
bi
om
e
tr
ic
s
pl
a
ys
a
c
r
uc
ia
l
r
ol
e
[
4]
.
F
or
e
xa
m
pl
e
,
w
he
n
la
te
nt
f
in
ge
r
pr
in
ts
a
r
e
f
ound
a
t
a
c
r
im
e
s
c
e
ne
,
a
ddi
ti
ona
l
in
f
or
m
a
ti
on
s
uc
h
a
s
e
s
ti
m
a
te
d
ge
nde
r
c
a
n
he
lp
na
r
r
ow
dow
n
th
e
li
s
t
of
s
u
s
pe
c
t
s
f
r
om
a
l
a
r
ge
bi
om
e
tr
ic
da
t
a
ba
s
e
[
5]
.
T
hi
s
s
c
r
e
e
ni
ng
pr
oc
e
s
s
s
ig
ni
f
ic
a
nt
ly
im
pr
ove
s
th
e
e
f
f
ic
ie
nc
y
a
nd
s
pe
e
d
of
in
ve
s
ti
ga
ti
ons
,
a
ll
ow
in
g
la
w
e
nf
or
c
e
m
e
nt
of
f
ic
ia
ls
to
f
oc
us
th
e
ir
r
e
s
our
c
e
s
m
or
e
e
f
f
e
c
ti
ve
ly
.
T
he
r
e
f
or
e
,
de
v
e
lo
pi
ng
r
e
li
a
bl
e
m
e
th
od
s
f
or
e
xt
r
a
c
ti
ng
s
of
t
bi
om
e
tr
ic
in
f
or
m
a
ti
on
f
r
om
f
or
e
ns
ic
e
vi
de
nc
e
i
s
a
hi
ghl
y r
e
le
v
a
nt
a
nd pr
e
s
s
in
g r
e
s
e
a
r
c
h a
r
e
a
.
A
m
ong
th
e
va
r
io
us
ty
pe
s
of
bi
om
e
tr
ic
s
,
f
in
ge
r
pr
in
ts
oc
c
upy
a
s
pe
c
ia
l
pos
it
io
n
be
c
a
us
e
th
e
y
ha
ve
a
pa
tt
e
r
n t
ha
t
is
uni
que
t
o
e
ve
r
yone
a
nd r
e
m
a
in
s
c
ons
is
te
nt
t
hr
ou
ghout l
if
e
. T
hi
s
uni
que
ne
s
s
m
a
ke
s
i
t
one
of
t
he
m
os
t
r
e
li
a
bl
e
bi
om
e
tr
ic
id
e
nt
it
ie
s
in
th
e
a
ut
he
nt
ic
a
ti
on
pr
oc
e
s
s
[
6]
.
I
n
a
ddi
ti
on
to
f
unc
ti
oni
ng
a
s
a
s
in
gl
e
id
e
nt
it
y
[
7]
,
f
in
ge
r
p
r
in
ts
a
ls
o
ha
ve
f
in
e
m
or
phol
ogi
c
a
l
de
ta
il
s
th
a
t
ha
ve
be
e
n
pr
ove
n
to
h
a
ve
a
s
ta
ti
s
ti
c
a
l
r
e
la
ti
ons
hi
p
w
it
h
a
n
in
di
vi
dua
l'
s
ge
nd
e
r
[
8]
.
T
hi
s
f
in
di
ng
s
how
s
th
a
t
bi
ol
ogi
c
a
l
di
f
f
e
r
e
nc
e
s
be
twe
e
n
m
e
n
a
nd
w
om
e
n
a
r
e
a
ls
o
r
e
f
le
c
te
d
in
th
e
m
ic
r
o
c
ha
r
a
c
t
e
r
is
ti
c
s
of
f
in
ge
r
pr
in
ts
[
9]
.
T
he
r
e
f
or
e
,
f
in
ge
r
pr
in
t
pa
tt
e
r
n
a
na
ly
s
is
i
s
not
onl
y
u
s
e
f
ul
f
or
id
e
nt
if
yi
ng
in
di
vi
dua
ls
but
c
a
n
a
ls
o
be
us
e
d
a
s
a
s
c
ie
nt
if
ic
ba
s
is
in
ge
nde
r
c
la
s
s
if
ic
a
ti
on, a
f
ie
ld
t
ha
t
c
ont
in
ue
s
t
o de
ve
lo
p i
n m
ode
r
n bi
om
e
tr
ic
s
tu
di
e
s
.
O
ne
of
th
e
m
os
t
f
r
e
que
nt
ly
us
e
d
tr
a
it
s
is
r
id
ge
de
ns
it
y,
w
hi
c
h
is
th
e
num
be
r
of
r
id
ge
s
pe
r
uni
t
a
r
e
a
[
8]
,
[
10]
.
F
e
m
a
le
s
ge
ne
r
a
ll
y
ha
ve
a
hi
ghe
r
r
id
ge
de
ns
it
y
th
a
n
m
a
le
s
[
5]
.
F
or
e
xa
m
pl
e
,
va
lu
e
s
gr
e
a
te
r
th
a
n
14
r
id
ge
s
/2
5
m
m
²
a
r
e
m
or
e
li
ke
ly
to
be
f
e
m
a
le
,
w
hi
le
va
lu
e
s
le
s
s
th
a
n
12
r
id
ge
s
/2
5
m
m
²
a
r
e
m
or
e
l
ik
e
ly
to
be
m
a
le
.
T
hi
s
p
a
tt
e
r
n
ha
s
b
e
e
n
f
ound
a
c
r
os
s
popula
ti
ons
[
10]
–
[
12]
,
a
nd
is
th
e
r
e
f
or
e
th
ought
to
be
uni
ve
r
s
a
l.
I
n
a
ddi
ti
on, othe
r
t
r
a
it
s
s
uc
h a
s
t
he
r
id
ge
t
hi
c
kne
s
s
t
o va
ll
e
y t
hi
c
kn
e
s
s
r
a
ti
o
(
R
T
V
T
R
)
, t
he
numbe
r
of
w
hi
te
l
in
e
s
,
a
nd
th
e
a
s
ym
m
e
tr
y
of
th
e
num
b
e
r
of
r
id
ge
s
ha
ve
a
ls
o
be
e
n
s
ho
w
n
to
be
s
ig
ni
f
ic
a
nt
in
di
s
ti
ngui
s
hi
ng
s
e
x
[
5]
.
T
he
c
om
bi
na
ti
on of
t
he
s
e
c
ha
r
a
c
te
r
is
ti
c
s
f
or
m
s
a
s
tr
ong ba
s
i
s
f
or
a
ut
om
a
te
d c
la
s
s
if
ic
a
ti
on s
y
s
te
m
s
.
E
a
r
ly
a
tt
e
m
pt
s
to
a
ut
om
a
te
ge
nde
r
c
la
s
s
if
ic
a
ti
on
f
r
om
f
in
ge
r
pr
in
ts
ge
ne
r
a
ll
y
us
e
d
m
a
c
hi
ne
l
e
a
r
ni
ng
te
c
hni
que
s
.
T
he
pr
oc
e
s
s
ty
pi
c
a
ll
y
c
ons
i
s
te
d
of
two
s
ta
ge
s
of
m
a
nua
l
f
e
a
tu
r
e
e
xt
r
a
c
ti
on,
w
he
r
e
r
e
s
e
a
r
c
h
e
r
s
de
s
ig
ne
d
a
lg
or
it
hm
s
to
c
om
put
e
f
e
a
tu
r
e
s
s
uc
h
a
s
r
id
ge
d
e
ns
it
y
a
nd
R
T
V
T
R
,
a
nd
th
e
n
f
e
d
th
e
obt
a
in
e
d
f
e
a
tu
r
e
s
in
to
s
ta
nda
r
d
c
la
s
s
if
ie
r
s
[
13]
,
[
14]
.
W
hi
le
th
e
s
e
m
e
th
ods
w
e
r
e
qui
te
s
uc
c
e
s
s
f
ul
,
th
e
ir
pe
r
f
or
m
a
nc
e
w
a
s
hi
ghl
y
de
pe
nde
nt
on
th
e
qua
li
ty
a
nd
r
e
le
va
nc
e
of
th
e
m
a
nua
ll
y
e
ngi
ne
e
r
e
d
f
e
a
tu
r
e
s
,
w
hi
c
h
w
e
r
e
of
te
n
uns
ta
bl
e
a
nd dif
f
ic
ul
t
to
ge
ne
r
a
li
z
e
a
c
r
os
s
di
f
f
e
r
e
nt
i
m
a
ge
c
ondi
ti
ons
.
A
dva
nc
e
s
in
D
e
e
p
L
e
a
r
ni
ng,
pa
r
ti
c
ul
a
r
ly
us
in
g
c
onvolut
io
na
l
ne
ur
a
l
ne
twor
ks
(
C
N
N
)
,
ha
v
e
ha
d
a
s
ig
ni
f
ic
a
nt
im
pa
c
t
on
pa
tt
e
r
n
r
e
c
ogni
ti
on
a
nd
c
om
put
e
r
vi
s
io
n
.
C
N
N
s
h
a
ve
th
e
a
dva
nt
a
ge
of
a
ut
om
a
ti
c
a
ll
y
e
xt
r
a
c
ti
ng
im
por
ta
nt
f
e
a
tu
r
e
s
di
r
e
c
tl
y
f
r
om
pi
xe
l
da
ta
w
it
ho
ut
th
e
ne
e
d
f
or
c
om
pl
e
x
f
e
a
tu
r
e
e
ngi
ne
e
r
in
g
pr
oc
e
s
s
e
s
[
15]
.
B
y
ut
il
iz
in
g
c
onvolut
io
na
l,
pool
in
g,
a
nd
f
ul
ly
c
onne
c
te
d
la
ye
r
s
,
C
N
N
s
c
a
n
f
or
m
th
e
m
os
t
s
ui
ta
bl
e
da
ta
r
e
pr
e
s
e
nt
a
ti
on
f
or
c
la
s
s
if
ic
a
ti
on
ta
s
ks
[
16]
.
T
hi
s
a
ppr
oa
c
h
not
onl
y
s
im
pl
if
ie
s
th
e
a
na
ly
s
is
pr
oc
e
s
s
but
a
l
s
o
of
te
n
r
e
s
ul
ts
in
be
tt
e
r
pe
r
f
or
m
a
nc
e
,
a
s
it
i
s
a
bl
e
to
r
e
c
ogni
z
e
d
e
ta
il
e
d
pa
tt
e
r
ns
th
a
t
a
r
e
di
f
f
ic
ul
t
to
c
a
pt
ur
e
us
in
g m
a
nua
l
m
e
th
ods
.
T
he
a
ppl
ic
a
ti
on of
C
N
N
s
i
n f
in
ge
r
pr
in
t
-
ba
s
e
d ge
nde
r
c
la
s
s
if
ic
a
ti
on
ha
s
de
m
ons
tr
a
te
d
good
pe
r
f
or
m
a
nc
e
,
w
it
h
m
a
ny
s
tu
di
e
s
r
e
por
ti
ng
hi
gh
a
c
c
ur
a
c
y
r
a
te
s
[
13]
,
[
17]
–
[
21]
.
H
ow
e
ve
r
,
m
os
t
of
th
e
s
e
e
va
lu
a
ti
ons
w
e
r
e
c
ondu
c
te
d
u
s
in
g
hi
gh
-
qua
li
ty
f
in
ge
r
pr
in
t
im
a
ge
s
a
c
qui
r
e
d
unde
r
c
ont
r
ol
le
d
la
bor
a
to
r
y
c
ondi
ti
ons
.
T
hi
s
c
r
e
a
te
s
a
ga
p
w
it
h
r
e
a
l
-
w
or
ld
f
or
e
ns
ic
c
ondi
ti
ons
,
w
he
r
e
la
te
nt
f
in
ge
r
pr
in
ts
a
r
e
of
te
n i
nc
om
pl
e
te
, bl
ur
r
e
d, or
di
s
to
r
te
d.
T
o
a
ddr
e
s
s
th
is
c
h
a
ll
e
nge
,
th
e
S
okot
o
c
ove
nt
r
y
f
in
ge
r
pr
in
t
da
t
a
s
e
t
(
S
O
C
O
F
in
g)
w
a
s
de
ve
lo
pe
d
by
pr
ovi
di
ng
two
ty
pe
s
of
im
a
ge
s
:
r
e
a
l
a
nd
a
lt
e
r
e
d
,
w
hi
c
h
m
im
ic
di
f
f
e
r
e
nt
f
or
m
s
of
de
gr
a
da
ti
on
[
22]
.
U
nf
or
tu
na
te
ly
,
m
a
ny
pr
e
vi
ous
s
tu
di
e
s
do
not
e
xpl
ic
it
ly
s
pe
c
if
y
th
e
us
e
of
th
e
A
lt
e
r
e
d
s
ubs
e
t,
r
e
s
ul
ti
ng
in
w
id
e
ly
r
e
por
te
d
r
e
s
ul
ts
r
a
ngi
ng
f
r
om
72%
[
21]
to
99%
[
23]
a
nd
m
a
ki
ng
th
e
m
di
f
f
ic
ul
t
to
c
om
pa
r
e
f
a
ir
ly
.
T
hi
s
s
tu
dy
c
ont
r
ib
ut
e
s
by
of
f
e
r
in
g
a
tr
a
ns
pa
r
e
nt
a
nd
pr
a
c
ti
c
a
l
tr
a
in
in
g
m
e
th
odol
ogy.
T
he
e
nt
ir
e
S
O
C
O
F
in
g
da
ta
s
e
t,
bot
h
r
e
a
l
a
nd
a
lt
e
r
e
d
,
is
c
om
bi
ne
d
in
to
a
s
in
gl
e
uni
f
i
e
d
da
ta
s
e
t
f
or
tr
a
in
in
g
a
nd
te
s
ti
ng.
T
he
m
a
in
nove
lt
y
of
th
is
a
ppr
oa
c
h
i
s
th
e
d
e
m
ons
tr
a
ti
on
th
a
t
by
e
xpl
ic
it
ly
tr
a
in
in
g
th
e
m
ode
l
on
di
ve
r
s
e
da
ta
,
in
c
lu
di
ng
bot
h
or
ig
in
a
l
a
nd
de
gr
a
de
d
im
a
ge
s
,
C
N
N
s
c
a
n
a
c
hi
e
ve
hi
gh
a
c
c
ur
a
c
y.
T
hi
s
s
tr
a
te
gy
e
s
ta
bl
is
h
e
s
a
s
tr
ong
ba
s
e
li
ne
f
or
pr
a
c
ti
c
a
l
f
or
e
ns
ic
a
ppl
ic
a
ti
ons
,
a
s
th
e
m
ode
l
is
not
onl
y
te
s
te
d
f
or
r
obus
tn
e
s
s
a
f
te
r
be
in
g
tr
a
in
e
d
on c
le
a
n da
ta
, but
i
s
a
l
s
o t
r
a
in
e
d t
o r
e
c
ogni
z
e
va
r
io
us
d
e
gr
a
da
ti
on c
ondi
ti
ons
f
r
om
t
he
out
s
e
t.
2.
R
E
L
A
T
E
D
WORK
A
dva
nc
e
s
in
c
om
put
in
g
po
w
e
r
ha
ve
pus
h
e
d
de
e
p
le
a
r
ni
ng,
p
a
r
ti
c
ul
a
r
ly
C
N
N
s
,
to
be
c
om
e
a
le
a
di
ng
a
ppr
oa
c
h
in
va
r
io
us
im
a
ge
-
ba
s
e
d c
la
s
s
if
ic
a
ti
on
ta
s
ks
.
C
N
N
s
ou
tp
e
r
f
or
m
tr
a
di
t
io
na
l
m
e
th
ods
by
a
ut
om
a
ti
c
a
ll
y
e
xt
r
a
c
ti
ng
im
por
ta
nt
f
e
a
tu
r
e
s
di
r
e
c
tl
y
f
r
om
r
a
w
im
a
ge
da
ta
[
15
]
.
V
a
r
io
us
C
N
N
a
r
c
hi
te
c
tu
r
e
s
ha
ve
be
e
n
us
e
d
f
or
ge
nde
r
c
la
s
s
if
ic
a
ti
on
ba
s
e
d
on
bi
om
e
tr
ic
da
ta
.
F
or
e
xa
m
pl
e
,
H
a
be
e
b
e
t
al
.
[
24]
a
ppl
ie
d
e
f
f
ic
ie
n
t
N
e
t
-
B
2,
R
e
s
N
e
t5
0,
R
e
s
N
e
t1
8,
a
nd
L
ig
ht
ni
ng
a
r
c
hi
te
c
tu
r
e
s
f
or
ge
nd
e
r
pr
e
di
c
ti
on
ba
s
e
d
on
f
a
c
ia
l
im
a
g
e
s
,
w
it
h
R
e
s
N
e
t1
8
a
c
hi
e
vi
ng
th
e
hi
gh
e
s
t
a
c
c
ur
a
c
y
of
98%
.
K
um
a
r
e
t
al
.
[
25]
us
in
g
A
le
xN
e
t
r
e
por
te
d
a
n
a
c
c
ur
a
c
y
of
a
ppr
oxi
m
a
te
ly
95.31%
f
or
ge
nde
r
pr
e
di
c
ti
on.
M
e
a
nw
hi
le
,
A
r
or
a
e
t
al
.
[
26]
s
ho
w
s
th
a
t
C
N
N
s
c
a
n
a
c
hi
e
ve
hi
gh
a
c
c
ur
a
c
y
e
ve
n
w
it
h
lo
w
-
qua
li
ty
in
put
da
ta
.
T
hi
s
f
in
di
ng
s
ugge
s
ts
th
a
t
a
r
c
hi
te
c
tu
r
e
s
pr
e
-
tr
a
in
e
d
on
im
a
ge
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8814
I
nt
J
A
dv A
ppl
S
c
i
,
V
ol
.
14
, N
o.
4
,
D
e
c
e
m
be
r
20
25
:
1350
-
1358
1352
da
ta
s
e
ts
c
a
n
be
e
f
f
e
c
ti
ve
ly
a
da
pt
e
d
f
or
s
pe
c
if
ic
ta
s
ks
,
s
uc
h
a
s
ge
nde
r
c
la
s
s
if
ic
a
ti
on
f
r
om
f
in
ge
r
pr
in
ts
,
e
ve
n
w
it
h l
ow
-
qua
li
ty
i
nput
da
ta
A
ddi
ti
ona
ll
y,
hybr
id
a
ppr
oa
c
he
s
ha
ve
a
ls
o
be
e
n
e
xpl
or
e
d
to
c
om
bi
ne
th
e
a
dva
nt
a
ge
s
of
di
f
f
e
r
e
nt
m
e
th
ods
.
O
ne
e
xa
m
pl
e
is
th
e
C
N
N
-
S
V
M
f
r
a
m
e
w
or
k,
w
he
r
e
th
e
C
N
N
s
e
r
ve
s
a
s
a
f
e
a
tu
r
e
e
xt
r
a
c
to
r
to
c
a
pt
ur
e
s
pa
ti
a
l
in
f
or
m
a
ti
on
f
r
om
f
in
ge
r
pr
in
ts
,
a
nd
th
e
obt
a
in
e
d
f
e
a
tu
r
e
s
a
r
e
th
e
n c
la
s
s
if
ie
d
us
in
g
S
V
M
.
T
hi
s
a
ppr
oa
c
h
ha
s
be
e
n
r
e
por
te
d
to
a
c
hi
e
ve
a
ve
r
y
hi
gh
a
c
c
ur
a
c
y
of
99.25
%
on
a
gi
ve
n
da
ta
s
e
t,
th
us
d
e
m
ons
tr
a
ti
ng
th
e
pot
e
nt
ia
l
s
yne
r
gy be
twe
e
n de
e
p l
e
a
r
ni
ng
-
ba
s
e
d f
e
a
tu
r
e
e
xt
r
a
c
ti
o
n a
nd t
r
a
di
ti
ona
l
c
la
s
s
if
ic
a
ti
on me
th
ods
[
27]
.
T
he
S
O
C
O
F
in
g
da
ta
s
e
t,
c
on
s
is
ti
ng
of
bot
h
r
e
a
l
a
nd
a
lt
e
r
e
d
im
a
ge
s
.
H
ow
e
ve
r
,
th
e
r
e
s
ul
t
s
of
s
tu
di
e
s
us
in
g
th
is
d
a
ta
s
e
t
s
how
c
on
s
id
e
r
a
bl
e
va
r
ia
ti
on,
in
di
c
a
ti
ng
t
he
la
c
k
of
a
c
ons
is
t
e
nt
e
va
lu
a
ti
on
pr
ot
oc
ol
.
F
or
e
xa
m
pl
e
,
O
la
de
le
e
t
al
.
[
21]
r
e
por
te
d
72
%
a
c
c
ur
a
c
y
us
in
g
a
s
im
pl
e
s
e
ve
n
-
la
ye
r
C
N
N
.
I
n
c
ont
r
a
s
t,
I
lo
a
nus
i
a
nd
E
ji
ogu
[
27]
us
e
d
a
d
e
e
pe
r
C
N
N
w
it
h
20
la
ye
r
s
a
nd
a
c
hi
e
v
e
d
91.3%
a
c
c
ur
a
c
y,
in
di
c
a
ti
ng
th
a
t
a
r
c
hi
te
c
tu
r
e
de
pt
h
c
a
n
im
pa
c
t
m
ode
l
pe
r
f
or
m
a
nc
e
.
F
ur
th
e
r
m
or
e
,
T
hongli
m
e
t
al
.
[
23]
r
e
por
te
d
up
to
99%
a
c
c
ur
a
c
y
w
it
h
C
N
N
c
on
s
is
ti
ng
of
onl
y
two
c
onvolut
io
na
l
la
ye
r
s
,
two
pool
in
g
la
ye
r
s
,
a
nd
two
d
e
ns
e
la
ye
r
s
.
T
hi
s
c
ont
r
a
s
ti
ng
r
a
nge
of
r
e
s
ul
ts
(
72%
to
99%
)
hi
ghl
ig
ht
s
t
he
di
f
f
ic
ul
ty
of
di
r
e
c
t
c
om
pa
r
is
ons
be
twe
e
n
s
tu
di
e
s
.
O
ne
of
th
e
m
a
in
f
a
c
to
r
s
c
a
u
s
in
g
th
e
di
f
f
e
r
e
nc
e
in
r
e
s
ul
ts
is
th
e
la
c
k
of
c
la
r
it
y
r
e
ga
r
di
ng
th
e
us
e
of
m
odi
f
ie
d
S
O
C
O
F
in
g
s
ubs
e
ts
in
th
e
tr
a
in
in
g
a
nd
te
s
ti
ng.
T
hi
s
l
a
c
k
of
m
e
th
odol
ogi
c
a
l
tr
a
ns
pa
r
e
nc
y
r
e
in
f
or
c
e
s
th
e
ne
e
d
f
or
m
or
e
s
ys
t
e
m
a
ti
c
a
nd
c
om
pr
e
he
ns
iv
e
s
tu
di
e
s
.
T
h
is
s
tu
dy
a
ddr
e
s
s
e
s
th
is
g
a
p
by
c
om
bi
ni
ng
a
ll
S
O
C
O
F
in
g
da
ta
,
bot
h
r
e
a
l
a
nd
a
lt
e
r
e
d,
to
bui
ld
a
n
in
he
r
e
nt
ly
m
or
e
r
obus
t
m
ode
l.
W
it
h
th
is
a
ppr
oa
c
h,
w
e
a
im
to
ge
ne
r
a
te
a
r
obus
t
ba
s
e
li
ne
w
hi
le
e
ns
ur
in
g f
a
ir
c
om
pa
r
is
ons
of
t
he
r
e
s
ul
ts
i
n t
he
f
ut
ur
e
.
3.
M
E
T
H
O
D
3.1.
S
O
C
O
F
in
g
d
a
t
as
e
t
T
he
da
ta
s
e
t
us
e
d
in
th
is
s
tu
dy
is
S
O
C
O
F
in
g
w
hi
c
h
is
a
publ
ic
bi
om
e
tr
ic
da
ta
ba
s
e
[
22]
.
T
hi
s
da
ta
s
e
t
is
w
e
ll
-
s
ui
te
d
f
or
our
pur
pos
e
s
be
c
a
us
e
it
pr
ovi
de
s
bot
h
r
e
a
l
a
nd
s
ynt
he
ti
c
a
ll
y
a
lt
e
r
e
d
f
in
ge
r
pr
in
t
im
a
ge
s
,
a
ll
ow
in
g f
or
r
obus
t
m
ode
l
tr
a
in
in
g. T
he
da
ta
s
e
t
c
ons
is
ts
of
6,00
0 “
r
e
a
l”
f
in
ge
r
pr
in
t
im
a
ge
s
c
ol
le
c
te
d f
r
om
600
s
ubj
e
c
ts
of
A
f
r
ic
a
n
de
s
c
e
nt
.
E
a
c
h
s
ubj
e
c
t
c
ont
r
ib
ut
e
d
f
in
ge
r
pr
in
ts
f
r
om
a
ll
te
n
f
in
ge
r
s
.
O
ne
im
por
ta
nt
c
om
pone
nt
of
th
e
S
O
C
O
F
in
g
da
ta
s
e
t
is
th
e
A
lt
e
r
e
d
s
ubs
e
t,
w
hi
c
h
c
ons
is
ts
of
te
ns
of
th
ous
a
nds
of
f
in
ge
r
pr
in
t
im
a
ge
s
m
odi
f
ie
d
f
r
om
th
e
o
r
ig
in
a
l
(
r
e
a
l)
im
a
ge
s
.
T
he
s
e
m
odi
f
ic
a
ti
ons
a
r
e
pe
r
f
or
m
e
d
us
in
g
th
e
S
T
R
A
N
G
E
to
ol
box,
w
hi
c
h
is
de
s
ig
ne
d
to
s
im
ul
a
te
va
r
io
us
f
or
m
s
of
f
o
r
e
ns
ic
de
gr
a
da
ti
on.
T
hi
s
s
ubs
e
t
is
to
s
im
ul
a
te
r
e
a
l
da
m
a
ge
c
ondi
ti
ons
s
o t
ha
t
th
e
r
obus
tn
e
s
s
of
t
he
f
in
ge
r
pr
in
t
r
e
c
o
gni
ti
on mode
l
c
a
n be
t
e
s
te
d m
or
e
r
e
a
li
s
ti
c
a
ll
y
.
T
he
r
e
a
r
e
th
r
e
e
m
a
in
ty
pe
s
of
m
odi
f
ic
a
ti
ons
in
th
e
A
lt
e
r
e
d
s
u
bs
e
t:
c
e
nt
r
a
l
r
ot
a
ti
on
(
C
R
)
,
obl
it
e
r
a
ti
on
(
O
bl
)
,
a
nd
z
-
c
ut
(
Z
c
ut
)
.
O
bl
it
e
r
a
ti
on
r
e
pr
e
s
e
nt
s
th
e
e
f
f
e
c
t
of
s
m
e
a
r
in
g
or
bl
ur
r
in
g
on
th
e
f
in
ge
r
pr
in
t
im
a
ge
,
c
e
nt
r
a
l
r
ot
a
ti
on
de
s
c
r
ib
e
s
di
s
to
r
ti
on
due
to
to
r
s
io
n,
w
hi
le
z
-
c
ut
s
im
ul
a
te
s
phys
ic
a
l
da
m
a
ge
s
uc
h
a
s
s
c
r
a
tc
he
s
on
th
e
f
in
ge
r
pr
in
t
s
ur
f
a
c
e
.
F
or
e
xa
m
pl
e
,
th
e
c
om
pa
r
is
on
of
th
e
th
u
m
b
f
in
ge
r
pr
in
t
be
twe
e
n
th
e
r
e
a
l
a
nd
a
lt
e
r
e
d
im
a
ge
s
is
s
how
n
in
F
ig
ur
e
1.
T
he
or
ig
in
a
l
th
um
b
f
in
ge
r
pr
in
t
im
a
ge
w
a
s
m
odi
f
ie
d
u
s
in
g
th
e
S
T
R
A
N
G
E
to
ol
box to pr
oduc
e
a
de
gr
a
de
d i
m
a
ge
w
it
h t
hr
e
e
l
e
ve
ls
of
di
f
f
ic
ul
ty
:
e
a
s
y, m
e
di
um
, a
nd ha
r
d.
F
ig
ur
e
1. I
ll
us
tr
a
ti
on c
om
pa
r
in
g t
he
r
e
a
l
th
um
bpr
in
t
a
nd de
gr
a
de
d ve
r
s
io
n of
t
he
s
a
m
e
i
ndi
vi
dua
l
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
dv A
ppl
S
c
i
I
S
S
N
:
2252
-
8814
C
onv
ol
ut
io
nal
ne
ur
al
ne
tw
o
r
k
m
ode
l
fo
r
f
in
ge
r
pr
in
t
-
bas
e
d g
e
n
de
r
c
la
s
s
if
ic
at
io
n
…
(
R
is
qy
Si
w
i
P
r
adi
ni
)
1353
3.
2
.
D
at
a
p
r
e
p
ar
at
io
n
an
d
p
r
e
p
r
oc
e
s
s
in
g
T
o
bui
ld
a
r
obus
t
m
ode
l,
th
is
s
tu
dy
im
pl
e
m
e
nt
e
d
a
n
in
te
gr
a
te
d
a
nd
s
ys
t
e
m
a
ti
c
da
ta
pr
e
pa
r
a
ti
on
w
or
kf
lo
w
.
T
he
f
ir
s
t
c
r
uc
ia
l
s
te
p
is
da
ta
m
e
r
gi
ng.
A
ll
im
a
ge
s
f
r
om
th
e
r
e
a
l
di
r
e
c
to
r
y
a
nd
a
ll
a
lt
e
r
e
d
s
ub
-
di
r
e
c
to
r
ie
s
(
a
lt
e
r
e
d
-
e
a
s
y,
a
lt
e
r
e
d
-
m
e
di
um
,
a
nd
a
lt
e
r
e
d
-
ha
r
d
)
w
e
r
e
m
e
r
ge
d
in
to
a
s
in
gl
e
c
om
pr
e
he
ns
iv
e
da
ta
s
e
t.
A
ll
a
lt
e
r
e
d
im
a
ge
s
r
e
pr
e
s
e
nt
de
gr
a
de
d
im
a
ge
s
.
T
hi
s
da
ta
pr
e
pa
r
a
ti
on
s
tr
a
te
gy
e
ns
ur
e
s
th
a
t
th
e
m
ode
l
is
t
r
a
in
e
d on a
w
id
e
va
r
ie
ty
of
da
ta
, i
nc
lu
di
ng both c
le
a
n a
nd d
e
gr
a
de
d i
m
a
ge
s
.
A
f
te
r
m
e
r
gi
ng,
th
e
ne
xt
s
te
p
is
la
be
l
e
xt
r
a
c
ti
on.
A
c
us
to
m
s
c
r
ip
t
w
a
s
de
ve
lo
pe
d
to
r
e
a
d
th
e
f
il
e
na
m
e
of
e
a
c
h
im
a
ge
in
th
e
m
e
r
ge
d
da
ta
s
e
t.
T
he
g
e
nde
r
la
be
l
w
a
s
de
r
iv
e
d
f
r
om
s
pe
c
if
ic
c
ha
r
a
c
te
r
s
in
th
e
f
il
e
na
m
e
a
nd
th
e
n
c
onve
r
te
d
in
to
a
num
e
r
ic
f
or
m
by
m
a
ppi
ng
m
a
le
(
'
M
'
)
to
0
a
nd
f
e
m
a
le
(
'F'
)
to
1.
T
he
s
e
num
e
r
ic
la
be
ls
w
e
r
e
t
he
n c
onv
e
r
te
d i
nt
o a
c
a
te
gor
ic
a
l
f
or
m
a
t
to
m
a
tc
h t
he
l
os
s
f
unc
ti
on us
e
d i
n m
ode
l
tr
a
in
in
g.
T
he
ne
xt
s
ta
ge
is
im
a
ge
pr
e
-
pr
oc
e
s
s
in
g.
A
ll
im
a
ge
s
w
e
r
e
f
ir
s
t
c
onve
r
te
d
in
to
gr
a
ys
c
a
le
f
or
m
a
t
to
e
ns
ur
e
c
ons
i
s
te
nc
y
of
vi
s
u
a
l
r
e
pr
e
s
e
nt
a
ti
on
.
A
f
te
r
th
a
t,
th
e
im
a
ge
s
w
e
r
e
r
e
s
iz
e
d
to
a
uni
f
or
m
r
e
s
ol
ut
io
n
of
96×
96
pi
xe
ls
,
w
hi
c
h
is
c
ho
s
e
n
to
b
a
la
nc
e
c
om
put
a
ti
ona
l
e
f
f
ic
ie
nc
y
a
nd
r
e
le
va
nt
m
or
phol
ogi
c
a
l
de
ta
il
s
.
A
f
te
r
th
a
t,
pi
xe
l
va
lu
e
s
a
r
e
nor
m
a
li
z
e
d
by
di
vi
di
ng
e
a
c
h
pi
xe
l
va
lu
e
by
255.0,
s
o
th
a
t
a
ll
va
lu
e
s
a
r
e
in
th
e
r
a
nge
of
0
to
1
.
T
hi
s
nor
m
a
li
z
a
ti
on
pr
oc
e
s
s
h
e
lp
s
s
ta
bi
li
z
e
tr
a
in
in
g
w
hi
le
a
c
c
e
le
r
a
ti
ng
m
ode
l
c
onv
e
r
ge
nc
e
.
N
e
xt
, t
he
nume
r
ic
l
a
be
ls
a
r
e
c
onv
e
r
te
d i
nt
o a
c
a
te
gor
ic
a
l
f
or
m
a
t
to
m
a
tc
h t
he
l
os
s
f
unc
ti
on us
e
d.
A
f
te
r
m
e
r
gi
ng
a
nd
pr
e
pr
oc
e
s
s
in
g,
a
ll
da
ta
is
r
a
ndomi
z
e
d
a
nd
th
e
n
di
vi
de
d
in
to
two
pa
r
ts
:
tr
a
in
in
g
(
80%
)
a
nd
te
s
ti
ng
(
20
%
)
.
D
ur
in
g
th
e
tr
a
in
in
g
p
r
oc
e
s
s
,
a
por
ti
on
of
th
e
tr
a
in
in
g
da
ta
is
a
ls
o
a
ll
oc
a
te
d
a
s
a
va
li
da
ti
on
s
e
t
to
m
oni
to
r
m
ode
l
pe
r
f
or
m
a
nc
e
a
nd
pr
e
ve
nt
ove
r
f
it
ti
ng.
T
hi
s
s
e
pa
r
a
ti
on
s
tr
a
te
gy
e
n
s
ur
e
s
th
a
t
bot
h
th
e
tr
a
in
in
g
a
nd
te
s
ti
ng
da
ta
ha
ve
a
r
e
pr
e
s
e
nt
a
ti
ve
di
s
t
r
ib
ut
io
n
of
bot
h
'
R
e
a
l'
a
nd
'
A
lt
e
r
e
d'
im
a
ge
s
,
a
ll
ow
in
g f
or
f
a
ir
a
nd c
ons
is
te
nt
m
ode
l
e
va
lu
a
ti
on.
3.
3
.
P
r
op
os
e
d
C
N
N
a
r
c
h
it
e
c
t
u
r
e
T
he
p
r
im
a
r
y
g
oa
l
o
f
t
hi
s
r
e
s
e
a
r
c
h
is
t
o
de
ve
lo
p
a
m
ode
l
th
a
t
not
o
nl
y
a
c
hi
e
ve
s
h
ig
h
a
c
c
ur
a
c
y
b
ut
is
a
ls
o
c
o
m
p
ut
a
ti
ona
ll
y
e
f
f
ic
i
e
nt
.
T
o
a
c
hi
e
ve
th
is
go
a
l
,
t
hi
s
s
t
udy
pr
op
os
e
s
a
m
ul
ti
-
in
p
ut
C
N
N
a
r
c
hi
te
c
t
u
r
e
.
T
hi
s
m
o
de
l
n
ot
o
nl
y
pr
oc
e
s
s
e
s
f
in
ge
r
p
r
i
nt
i
m
a
g
e
s
bu
t
a
ls
o
in
c
or
po
r
a
t
e
s
a
dd
it
io
n
a
l
f
e
a
t
ur
e
s
in
to
t
he
de
c
is
io
n
-
m
a
k
in
g
p
r
oc
e
s
s
.
T
he
m
a
i
n
b
r
a
n
c
h
o
f
th
e
m
o
de
l
pr
oc
e
s
s
e
s
f
i
nge
r
p
r
i
nt
im
a
ge
s
th
r
o
ugh
th
r
e
e
c
onv
ol
ut
io
n
a
l
b
lo
c
ks
.
E
a
c
h
bl
o
c
k
c
o
ns
is
ts
o
f
a
C
onv
2D
l
a
ye
r
w
it
h
R
e
L
U
a
c
t
iv
a
ti
on
a
n
d
a
M
a
xP
oo
li
ng
2D
la
ye
r
.
T
he
n
um
be
r
o
f
f
i
lt
e
r
s
is
g
r
a
d
ua
l
ly
i
nc
r
e
a
s
e
d
f
r
om
3
2,
64
,
to
12
8,
a
ll
o
w
i
ng
t
he
m
o
de
l
t
o
le
a
r
n
f
e
a
tu
r
e
s
f
r
om
s
im
p
le
to
c
om
p
le
x.
I
n
a
ddi
ti
on
to
th
e
im
a
ge
,
th
e
m
ode
l
a
ls
o
r
e
c
e
iv
e
s
two
a
dd
it
io
na
l
in
put
s
in
th
e
f
o
r
m
o
f
c
a
te
g
or
ic
a
l
f
e
a
t
ur
e
s
(
ha
n
d
t
ype
a
nd
f
in
ge
r
ty
pe
)
w
hi
c
h
a
r
e
p
r
oc
e
s
s
e
d
us
i
ng
a
s
e
pa
r
a
te
D
e
ns
e
la
ye
r
.
N
e
xt
,
th
e
r
e
s
ul
ts
o
f
th
e
t
hr
e
e
b
r
a
nc
he
s
t
he
f
e
a
tu
r
e
m
a
p
f
r
om
t
he
f
la
t
te
n
e
d
im
a
ge
a
n
d
th
e
o
ut
p
ut
f
r
om
th
e
D
e
ns
e
la
ye
r
f
o
r
a
d
di
ti
ona
l
f
e
a
t
ur
e
s
a
r
e
c
om
bi
n
e
d
t
hr
ou
gh
a
c
o
nc
a
te
n
a
te
l
a
ye
r
.
T
h
is
c
o
m
bi
ne
d
ve
c
t
o
r
is
f
u
r
t
he
r
p
r
o
c
e
s
s
e
d
w
i
th
a
D
e
ns
e
la
ye
r
c
on
ta
i
ni
ng
12
8
ne
u
r
o
ns
,
f
o
ll
ow
e
d
by
a
D
r
opo
ut
la
ye
r
t
o
p
r
e
ve
n
t
ove
r
f
it
ti
ng
, a
nd
f
in
a
ll
y w
it
h a
n o
ut
pu
t
la
ye
r
c
on
ta
in
in
g
t
w
o
ne
ur
ons
f
o
r
bi
na
r
y
c
la
s
s
i
f
ic
a
ti
on
.
3.
4
.
T
r
ai
n
in
g an
d
e
val
u
at
io
n
p
r
ot
oc
ol
T
o
e
ns
ur
e
f
a
ir
a
nd
c
ons
i
s
te
nt
e
v
a
lu
a
ti
on,
th
is
s
tu
dy
f
ol
lo
w
e
d
a
c
le
a
r
ly
de
f
in
e
d
tr
a
in
in
g
a
nd
t
e
s
ti
ng
pr
ot
oc
ol
.
I
n
th
is
s
tu
dy,
th
e
m
ode
l
w
a
s
bui
lt
by
ut
il
iz
in
g
th
e
K
e
r
a
s
f
r
a
m
e
w
or
k
r
unni
ng
on
T
e
ns
or
F
lo
w
a
s
a
ba
c
ke
nd,
th
u
s
a
ll
ow
in
g
th
e
tr
a
in
in
g
pr
oc
e
s
s
to
be
c
a
r
r
ie
d
out
e
f
f
ic
ie
nt
ly
a
nd
in
a
s
tr
uc
tu
r
e
d
m
a
nne
r
.
H
ype
r
pa
r
a
m
e
te
r
s
e
le
c
ti
on
w
a
s
ba
s
e
d
on
be
s
t
pr
a
c
ti
c
e
s
w
id
e
l
y
us
e
d
in
th
e
li
te
r
a
tu
r
e
.
T
he
m
ode
l
tr
a
in
in
g
pr
oc
e
s
s
w
a
s
pe
r
f
or
m
e
d us
in
g t
he
A
da
m
opt
im
iz
e
r
, c
om
bi
ne
d
w
it
h t
he
c
a
te
gor
ic
a
l
c
r
os
s
-
e
nt
r
opy los
s
f
unc
ti
on,
a
s
it
is
s
ui
ta
bl
e
f
or
m
ul
t
ic
la
s
s
c
la
s
s
if
ic
a
ti
on
ta
s
ks
w
it
h
s
of
tm
a
x
-
ba
s
e
d
out
put
.
T
he
le
a
r
ni
ng
r
a
te
w
a
s
s
e
t
a
t
0.001,
w
hi
le
th
e
m
a
xi
m
um
num
be
r
of
e
poc
hs
w
a
s
li
m
it
e
d
to
50,
w
it
h
a
n
e
a
r
ly
s
to
ppi
ng
m
e
c
ha
ni
s
m
im
pl
e
m
e
nt
e
d
to
pr
e
ve
nt
ove
r
f
it
ti
ng.
F
ur
th
e
r
m
or
e
,
a
ba
tc
h
s
iz
e
of
32
w
a
s
us
e
d
to
m
a
in
ta
in
a
ba
la
nc
e
be
twe
e
n
gr
a
di
e
nt
s
ta
bi
li
ty
a
nd c
om
put
a
ti
ona
l
e
f
f
ic
ie
nc
y dur
in
g t
r
a
in
in
g.
M
ode
l
pe
r
f
or
m
a
nc
e
e
va
lu
a
ti
on w
a
s
pe
r
f
or
m
e
d
us
in
g s
ta
nda
r
d m
e
tr
ic
s
c
om
m
onl
y us
e
d i
n c
la
s
s
if
ic
a
ti
on t
a
s
ks
t
o
pr
ovi
de
a
c
om
pr
e
he
ns
iv
e
ove
r
vi
e
w
of
s
ys
te
m
pe
r
f
or
m
a
nc
e
.
T
he
m
e
tr
ic
s
us
e
d i
nc
lu
de
a
c
c
ur
a
c
y, pr
e
c
is
io
n, r
e
c
a
ll
, a
nd F
1 s
c
or
e
.
4.
R
E
S
U
L
T
S
A
N
D
D
I
S
C
U
S
S
I
O
N
4
.1.
M
od
e
l
p
e
r
f
or
m
an
c
e
on
c
o
m
b
in
e
d
d
at
as
e
t
T
he
m
ode
l
w
a
s
tr
a
in
e
d
us
in
g
a
tr
a
in
in
g
da
t
a
s
e
t
c
ont
a
in
in
g
a
m
i
x
of
‘
r
e
a
l’
a
nd
‘
a
lt
e
r
e
d’
im
a
g
e
s
,
th
e
n
e
va
lu
a
te
d
on
a
te
s
t
da
ta
s
e
t
w
it
h
a
s
im
il
a
r
di
s
tr
ib
ut
io
n.
T
he
e
va
lu
a
ti
on
r
e
s
ul
ts
on
th
e
c
om
bi
ne
d
te
s
t
da
ta
s
e
t
a
r
e
s
how
n
in
T
a
bl
e
1.
O
ur
m
ode
l
a
c
hi
e
ve
d
a
n a
c
c
ur
a
c
y
of
97
.
39%
,
de
m
ons
tr
a
ti
ng
th
a
t
tr
a
in
in
g
w
it
h
di
ve
r
s
e
da
ta
,
in
c
lu
di
ng
de
gr
a
de
d
im
a
ge
s
,
e
na
bl
e
s
th
e
C
N
N
to
l
e
a
r
n
r
obus
t
f
e
a
tu
r
e
r
e
pr
e
s
e
nt
a
ti
ons
th
a
t
a
r
e
r
e
s
il
ie
nt
to
va
r
io
us
pe
r
tu
r
ba
ti
ons
. T
he
hi
gh r
e
s
ul
ts
on t
hi
s
c
om
bi
ne
d
da
ta
s
e
t
a
r
e
a
ke
y f
in
di
ng of
t
he
s
tu
dy, de
m
ons
tr
a
ti
n
g
th
a
t
e
xpos
ur
e
t
o i
m
pe
r
f
e
c
t
da
ta
dur
in
g t
r
a
in
in
g i
s
a
n e
f
f
e
c
ti
ve
s
t
r
a
te
gy f
or
bui
ld
in
g r
obus
t
bi
om
e
tr
ic
m
ode
ls
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8814
I
nt
J
A
dv A
ppl
S
c
i
,
V
ol
.
14
, N
o.
4
,
D
e
c
e
m
be
r
20
25
:
1350
-
1358
1354
T
a
bl
e
1.
P
e
r
f
or
m
a
nc
e
m
e
tr
ic
s
on t
he
c
om
bi
ne
d d
a
ta
s
e
t
:
r
e
a
l
a
nd
a
lt
e
r
e
d
im
a
ge
s
A
c
c
ur
a
c
y
P
r
e
c
i
s
i
on
R
e
c
a
l
l
F1
-
S
c
or
e
0.97
39
0.9
437
0.9
227
0.9
331
4.2
.
T
r
ai
n
in
g
c
u
r
ve
an
al
ys
is
T
o
ga
in
a
d
e
e
pe
r
unde
r
s
t
a
ndi
ng
of
th
e
m
ode
l'
s
le
a
r
ni
ng
pr
oc
e
s
s
,
w
e
a
n
a
ly
z
e
d
th
e
tr
a
in
in
g
a
nd
va
li
da
ti
on
a
c
c
ur
a
c
y
c
ur
ve
s
ove
r
50
e
poc
hs
,
a
s
s
how
n
in
F
ig
ur
e
2.
T
he
c
ur
ve
s
s
how
th
e
dyna
m
ic
s
of
m
ode
l
a
c
c
ur
a
c
y
im
pr
ove
m
e
nt
on
bot
h
th
e
tr
a
in
in
g
da
ta
(
s
how
n
by
th
e
bl
ue
li
ne
)
a
nd
th
e
va
li
da
ti
on
da
ta
(
s
how
n
by
th
e
or
a
nge
l
in
e
)
a
s
t
he
numbe
r
of
e
poc
hs
i
nc
r
e
a
s
e
s
. I
n t
he
e
a
r
ly
t
r
a
in
in
g pha
s
e
, s
pe
c
if
ic
a
ll
y a
r
ound the
f
ir
s
t
10
to
15
e
poc
hs
,
th
e
r
e
is
a
ve
r
y
s
h
a
r
p
in
c
r
e
a
s
e
in
a
c
c
ur
a
c
y,
f
r
om
a
r
ound
80%
to
ove
r
95%
.
T
hi
s
in
di
c
a
te
s
th
a
t
th
e
m
ode
l
c
a
n quic
kl
y r
e
c
ogni
z
e
a
nd l
e
a
r
n i
m
por
ta
nt
pa
tt
e
r
ns
f
r
om
f
in
ge
r
pr
in
t
im
a
ge
s
.
A
f
te
r
pa
s
s
in
g
th
is
in
it
ia
l
pha
s
e
,
th
e
a
c
c
ur
a
c
y
c
ur
ve
s
be
ga
n
to
f
l
a
tt
e
n
a
nd
s
how
e
d a
te
nde
nc
y
to
w
a
r
ds
c
onve
r
ge
nc
e
.
B
ot
h
tr
a
in
in
g
a
nd
va
li
da
ti
on
a
c
c
ur
a
c
ie
s
s
ta
bi
li
z
e
d
a
t
a
r
ound
97
–
99%
,
in
di
c
a
ti
ng
th
a
t
th
e
m
ode
l
ha
d
r
e
a
c
he
d
it
s
opt
im
a
l
pe
r
f
or
m
a
nc
e
a
nd
th
a
t
a
ddi
ti
ona
l
tr
a
in
in
g
no
lo
nge
r
pr
ovi
de
d
s
ig
ni
f
ic
a
nt
im
pr
ove
m
e
nt
s
.
F
ur
th
e
r
m
or
e
,
th
e
r
e
w
a
s
a
s
m
a
ll
di
f
f
e
r
e
nc
e
be
t
w
e
e
n
th
e
tr
a
in
in
g
a
nd
va
li
da
ti
on
c
ur
ve
s
,
w
it
h
tr
a
in
in
g
a
c
c
ur
a
c
y
s
li
ght
ly
hi
ghe
r
.
T
he
di
f
f
e
r
e
nc
e
be
twe
e
n
tr
a
i
ni
ng
a
nd
va
li
da
ti
on
a
c
c
ur
a
c
y
r
e
m
a
in
s
w
it
hi
n
r
e
a
s
ona
bl
e
li
m
it
s
,
th
u
s
not
in
di
c
a
ti
ng
ov
e
r
f
it
ti
ng.
T
hi
s
is
e
vi
de
n
t
f
r
om
th
e
va
li
da
ti
on
a
c
c
ur
a
c
y,
w
hi
c
h
r
e
m
a
in
s
hi
gh
a
nd
s
ta
bl
e
,
w
it
h
no
s
ig
ns
of
de
c
li
ne
.
T
hi
s
in
di
c
a
te
s
th
a
t
th
e
m
ode
l
c
a
n
p
e
r
f
or
m
w
e
ll
on
ne
w
,
pr
e
vi
ous
ly
uns
e
e
n
da
ta
.
A
f
te
r
pa
s
s
in
g
th
e
in
it
ia
l
pha
s
e
,
th
e
a
c
c
ur
a
c
y
c
ur
ve
te
nds
to
f
la
tt
e
n
a
nd
r
e
a
c
h
a
poi
nt
o
f
c
onve
r
ge
nc
e
.
B
ot
h
tr
a
in
in
g
a
nd
va
li
da
ti
on
a
c
c
ur
a
c
y
c
ons
is
t
e
n
tl
y
hove
r
in
th
e
r
a
nge
of
97
–
99%
,
in
di
c
a
ti
n
g
th
a
t
th
e
m
ode
l
ha
s
a
c
hi
e
ve
d
opt
im
a
l
pe
r
f
or
m
a
nc
e
,
s
o
f
ur
th
e
r
tr
a
in
in
g
no
lo
nge
r
p
r
ovi
de
s
s
ig
ni
f
ic
a
nt
im
pr
ove
m
e
nt
.
T
hi
s
a
c
c
ur
a
c
y
c
ur
ve
a
na
ly
s
i
s
is
in
li
ne
w
it
h
th
e
f
in
a
l
e
va
lu
a
ti
on
r
e
s
ul
ts
,
w
hi
c
h
s
how
th
a
t
th
e
m
ode
l
c
a
n be
t
r
a
in
e
d c
ons
i
s
te
nt
ly
, s
ta
bl
y, a
nd
c
a
n a
c
hi
e
v
e
r
e
li
a
bl
e
hi
gh pe
r
f
or
m
a
nc
e
on t
e
s
t
da
ta
.
F
ig
ur
e
2
.
T
r
a
in
in
g a
nd va
li
da
ti
on a
c
c
ur
a
c
y c
ur
ve
s
4.3
.
C
on
f
u
s
io
n
m
at
r
ix
an
d
R
O
C
c
u
r
ve
an
al
ys
is
F
or
a
m
or
e
c
om
pr
e
he
ns
iv
e
e
va
lu
a
ti
on
in
F
ig
ur
e
3
,
a
n
a
na
ly
s
is
w
a
s
pe
r
f
or
m
e
d
us
in
g
a
c
onf
us
io
n
m
a
tr
ix
a
nd
a
r
e
c
e
iv
e
r
ope
r
a
ti
ng
c
ha
r
a
c
te
r
is
ti
c
(
R
O
C
)
c
ur
ve
,
a
s
s
how
n
in
F
ig
ur
e
3
(
a
)
.
T
he
c
onf
us
io
n
m
a
tr
ix
pr
ovi
de
s
a
de
ta
il
e
d
ove
r
vi
e
w
of
th
e
di
s
tr
ib
ut
io
n
o
f
m
ode
l
p
r
e
di
c
ti
ons
.
I
f
c
la
s
s
0
r
e
pr
e
s
e
nt
s
one
ge
nde
r
a
nd
c
la
s
s
1
r
e
pr
e
s
e
nt
s
th
e
ot
he
r
,
3,794
c
la
s
s
0
s
a
m
pl
e
s
a
nd
87
2
c
la
s
s
1
s
a
m
pl
e
s
w
e
r
e
c
or
r
e
c
tl
y
c
la
s
s
if
ie
d.
C
onve
r
s
e
ly
,
onl
y
52
c
la
s
s
0
s
a
m
pl
e
s
w
e
r
e
in
c
or
r
e
c
tl
y
de
te
c
t
e
d
a
s
c
la
s
s
1,
a
nd
73
c
la
s
s
1
s
a
m
pl
e
s
w
e
r
e
in
c
or
r
e
c
tl
y pr
e
di
c
te
d a
s
c
la
s
s
0. T
he
s
e
r
e
s
ul
ts
de
m
on
s
tr
a
te
t
ha
t
th
e
numbe
r
of
c
or
r
e
c
t
pr
e
di
c
ti
ons
s
ig
ni
f
ic
a
nt
l
y
out
num
be
r
s
t
he
numbe
r
of
e
r
r
or
s
, i
ndi
c
a
ti
ng a
ba
la
nc
e
d m
ode
l
pe
r
f
or
m
a
nc
e
i
n r
e
c
ogni
z
in
g both c
la
s
s
e
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
dv A
ppl
S
c
i
I
S
S
N
:
2252
-
8814
C
onv
ol
ut
io
nal
ne
ur
al
ne
tw
o
r
k
m
ode
l
fo
r
f
in
ge
r
pr
in
t
-
bas
e
d g
e
n
de
r
c
la
s
s
if
ic
at
io
n
…
(
R
is
qy
Si
w
i
P
r
adi
ni
)
1355
F
ur
th
e
r
m
or
e
,
a
na
ly
s
is
u
s
in
g
th
e
R
O
C
c
ur
ve
a
nd
th
e
a
r
e
a
und
e
r
th
e
c
ur
ve
(
A
U
C
)
va
lu
e
s
how
n
in
F
ig
ur
e
3
(
b
)
f
ur
th
e
r
c
or
r
obo
r
a
te
s
th
e
s
e
r
e
s
ul
ts
.
T
he
R
O
C
c
ur
ve
,
in
di
c
a
te
d
by
th
e
or
a
nge
li
ne
,
a
ppe
a
r
s
to
c
ur
ve
s
ha
r
pl
y
to
w
a
r
d
th
e
uppe
r
le
f
t
c
or
ne
r
,
a
w
a
y
f
r
om
th
e
r
a
ndo
m
di
a
gona
l
li
ne
,
in
di
c
a
ti
ng
e
xc
e
ll
e
nt
m
ode
l
di
s
c
r
im
in
a
ti
on.
T
he
a
c
hi
e
ve
d
A
U
C
va
lu
e
w
a
s
0.98,
c
lo
s
e
to
th
e
id
e
a
l
va
lu
e
of
1.0.
T
hi
s
m
e
a
ns
th
e
m
ode
l
ha
s
a
98%
pr
oba
bi
li
ty
o
f
s
c
or
in
g
a
r
a
ndom
pos
it
iv
e
s
a
m
pl
e
hi
ghe
r
th
a
n
a
r
a
ndom
ne
ga
ti
ve
s
a
m
pl
e
.
O
ve
r
a
ll
,
th
e
c
om
bi
na
ti
on
of
th
e
c
onf
u
s
io
n
m
a
tr
ix
a
nd
R
O
C
c
ur
ve
r
e
s
ul
ts
d
e
m
ons
tr
a
te
s
th
a
t
th
e
de
v
e
lo
pe
d
m
ode
l
not
onl
y
a
c
hi
e
ve
s
a
hi
gh
le
ve
l
of
a
c
c
ur
a
c
y
but
a
l
s
o
e
xhi
bi
ts
s
tr
ong
c
ons
is
te
nc
y
a
nd
r
e
li
a
bi
li
ty
in
di
s
ti
ngui
s
hi
ng
be
twe
e
n t
he
t
w
o ge
nde
r
c
la
s
s
e
s
.
4.4
.
A
d
van
t
age
s
of
t
r
ai
n
in
g on
a
ggr
e
gat
e
d
d
at
a
T
he
a
c
hi
e
ve
d
a
c
c
u
r
a
c
y
o
f
97
.
39
%
o
n
a
te
s
t
da
ta
s
e
t
c
ons
is
t
in
g
of
a
c
om
bi
n
a
t
io
n
of
c
le
a
n
a
nd
de
g
r
a
de
d
i
m
a
g
e
s
de
m
o
ns
t
r
a
t
e
s
h
ig
h
ly
s
ig
ni
f
i
c
a
n
t
r
e
s
u
lt
s
.
T
hi
s
f
i
nd
in
g
di
r
e
c
t
ly
s
u
pp
or
ts
t
he
m
a
i
n
r
e
s
e
a
r
c
h
hyp
ot
he
s
is
,
na
m
e
ly
th
a
t
t
r
a
in
i
ng
C
N
N
s
on
d
iv
e
r
s
e
a
nd
c
h
a
l
le
ng
in
g
da
ta
s
e
ts
is
a
n
e
f
f
e
c
t
iv
e
s
tr
a
te
g
y
f
o
r
pr
oduc
in
g
r
obus
t
m
ode
ls
.
R
a
t
he
r
th
a
n
t
r
a
i
ni
ng
t
he
m
ode
l
on
ly
o
n
da
ta
of
p
e
r
f
e
c
t
qua
li
ty
a
nd
h
opi
ng
i
t
w
il
l
ge
ne
r
a
l
iz
e
to
l
e
s
s
-
th
a
n
-
id
e
a
l
da
ta
,
th
is
a
p
pr
oa
c
h
e
xp
li
c
it
ly
te
a
c
he
s
th
e
m
o
de
l
to
r
e
c
o
gn
iz
e
e
s
s
e
n
ti
a
l
f
e
a
t
u
r
e
s
th
a
t
r
e
m
a
i
n
c
ons
is
te
nt
e
ve
n
w
he
n
t
he
i
m
a
ge
is
s
m
e
a
r
e
d,
d
i
s
to
r
te
d,
o
r
da
m
a
ge
d
.
I
n
t
hi
s
c
o
nt
e
xt
,
a
lt
e
r
e
d
im
a
ge
s
c
a
n
be
vi
e
w
e
d
a
s
a
f
o
r
m
o
f
d
a
ta
a
u
gm
e
nt
a
ti
on
s
pe
c
i
f
ic
t
o
th
e
b
io
m
e
t
r
i
c
d
om
a
in
.
W
hi
le
c
om
m
o
n
a
ugm
e
n
ta
t
io
n
te
c
hn
iq
u
e
s
,
s
uc
h
a
s
r
a
nd
om
r
o
ta
ti
on
o
r
f
l
ip
pi
n
g,
a
r
e
o
f
te
n
us
e
d
to
i
m
p
r
o
ve
a
m
ode
l
'
s
ge
ne
r
a
l
iz
a
bi
li
ty
,
t
he
us
e
of
a
lt
e
r
a
ti
ons
th
a
t
m
i
m
ic
r
e
a
l
-
w
o
r
ld
f
o
r
e
ns
ic
de
g
r
a
da
t
io
n,
s
uc
h
a
s
th
os
e
f
o
und
i
n
th
e
S
O
C
O
F
in
g
d
a
ta
s
e
t,
a
l
lo
w
s
t
he
m
o
de
l
to
le
a
r
n
a
m
or
e
i
nv
a
r
i
a
nt
a
n
d
pr
a
c
t
ic
a
l
ly
r
e
le
va
n
t
r
e
p
r
e
s
e
nt
a
ti
on
.
T
he
hi
gh
a
c
c
u
r
a
c
y
a
c
hi
e
ve
d
de
m
ons
tr
a
te
s
th
a
t
th
e
m
o
de
l
is
n
ot
s
im
pl
y
m
e
m
o
r
iz
i
ng
a
r
ti
f
a
c
ts
f
r
om
a
lt
e
r
e
d
da
ta
b
ut
is
s
uc
c
e
s
s
f
u
ll
y
e
x
t
r
a
c
t
in
g
un
de
r
l
yi
ng
p
a
t
te
r
ns
th
a
t
a
r
e
c
o
ns
is
te
n
t
a
c
r
os
s
di
f
f
e
r
e
nt
im
a
ge
qua
l
it
y
le
v
e
ls
.
(
a
)
(
b)
F
ig
ur
e
3. C
om
pr
e
he
ns
iv
e
e
va
lu
a
ti
on i
n (
a
)
c
onf
us
io
n m
a
tr
ix
a
nd (
b)
R
O
C
c
ur
ve
4.5
.
C
om
p
ar
at
iv
e
an
al
ys
is
w
it
h
r
e
le
van
t
p
r
e
vi
ou
s
s
t
u
d
y
T
o
pl
a
c
e
our
r
e
s
ul
ts
in
th
e
c
ont
e
xt
of
th
e
e
xi
s
ti
ng
li
te
r
a
tu
r
e
,
w
e
c
om
pa
r
e
our
m
ode
l'
s
p
e
r
f
or
m
a
nc
e
w
it
h
r
e
s
ul
ts
r
e
por
te
d
by
ot
he
r
s
tu
di
e
s
us
in
g
th
e
S
O
C
O
F
in
g
da
t
a
s
e
t.
T
hi
s
c
om
pa
r
is
on
is
pr
e
s
e
nt
e
d
in
T
a
bl
e
2.
O
ur
m
ode
l
de
m
ons
tr
a
te
s
ve
r
y
c
om
pe
ti
ti
ve
pe
r
f
or
m
a
nc
e
,
s
ur
pa
s
s
in
g
s
e
ve
r
a
l
pr
e
vi
ous
s
tu
di
e
s
th
a
t
us
e
d
onl
y
“
r
e
a
l”
da
ta
.
W
hi
le
s
li
ght
ly
be
lo
w
th
e
99%
a
c
c
ur
a
c
y
r
e
por
te
d
by
T
hongli
m
e
t
al
.
[
23]
a
nd
G
us
ti
s
ya
f
a
nd
S
in
a
ga
[
28]
w
it
h
99.9667%
a
c
c
ur
a
c
y,
it
is
im
por
ta
nt
to
not
e
t
ha
t
our
m
ode
l
w
a
s
e
va
lu
a
te
d
on
a
n
in
he
r
e
nt
ly
m
or
e
c
ha
ll
e
ngi
ng
da
ta
s
e
t,
a
s
it
in
c
lu
de
s
bot
h
pr
is
ti
ne
a
nd
de
g
r
a
de
d
im
a
ge
s
.
T
hi
s
s
tr
ongl
y
s
ugg
e
s
ts
th
a
t
our
c
om
bi
ne
d t
r
a
in
in
g a
ppr
oa
c
h pr
oduc
e
s
a
s
upe
r
io
r
a
nd mor
e
r
e
li
a
bl
e
m
ode
l
f
or
pr
a
c
ti
c
a
l
a
ppl
ic
a
ti
ons
.
I
t
is
im
por
ta
nt
to
not
e
th
a
t
th
e
hi
gh
pe
r
f
or
m
a
nc
e
a
c
hi
e
ve
d
in
th
is
s
tu
dy
w
a
s
a
c
hi
e
ve
d
us
in
g
a
r
e
la
ti
ve
ly
s
im
pl
e
a
nd
c
om
put
a
ti
ona
ll
y
e
f
f
ic
ie
nt
a
r
c
hi
te
c
tu
r
e
.
T
hi
s
f
in
di
ng
c
ha
ll
e
nge
s
th
e
c
om
m
on
a
s
s
um
pt
io
n
th
a
t
onl
y
ve
r
y
de
e
p
a
nd
c
om
pl
e
x
m
ode
ls
c
a
n
a
c
hi
e
ve
s
ta
te
-
of
-
th
e
-
a
r
t
r
e
s
ul
ts
.
O
ur
r
e
s
ul
ts
c
onf
ir
m
th
a
t
da
ta
-
r
e
la
te
d
s
tr
a
te
gi
e
s
,
pa
r
ti
c
ul
a
r
ly
th
e
qua
li
ty
a
nd
di
ve
r
s
it
y
o
f
tr
a
in
in
g
da
ta
,
c
a
n
be
a
s
im
po
r
ta
nt
,
or
e
ve
n
m
or
e
s
o, t
ha
n t
he
c
om
pl
e
xi
ty
of
t
he
ne
twor
k a
r
c
hi
te
c
tu
r
e
i
ts
e
lf
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8814
I
nt
J
A
dv A
ppl
S
c
i
,
V
ol
.
14
, N
o.
4
,
D
e
c
e
m
be
r
20
25
:
1350
-
1358
1356
T
he
pr
a
c
ti
c
a
l
im
pl
ic
a
ti
ons
of
th
e
s
e
f
in
di
ngs
a
r
e
s
ig
ni
f
ic
a
nt
,
pa
r
ti
c
ul
a
r
ly
in
f
or
e
ns
ic
c
ont
e
xt
s
,
w
he
r
e
la
te
nt
f
in
ge
r
pr
in
t
qua
li
ty
is
r
a
r
e
ly
id
e
a
l.
R
e
ly
in
g
on
m
ode
ls
tr
a
in
e
d
s
ol
e
ly
on
hi
gh
-
qua
li
ty
la
bor
a
to
r
y
da
ta
pos
e
s
s
ig
ni
f
ic
a
nt
r
is
ks
a
nd
r
e
s
ul
t
s
in
le
s
s
r
e
li
a
bl
e
s
ys
t
e
m
s
.
T
h
is
r
e
s
e
a
r
c
h
de
m
ons
tr
a
te
s
th
a
t
it
is
pos
s
ib
le
to
bui
ld
a
s
in
gl
e
m
ode
l
th
a
t
m
a
in
ta
in
s
hi
gh
pe
r
f
o
r
m
a
nc
e
de
s
pi
te
va
r
yi
ng
im
a
ge
qua
li
ty
.
W
it
h
th
e
pr
opos
e
d
tr
a
in
in
g
m
e
th
odol
ogy,
th
e
de
ve
lo
pe
d
s
ys
te
m
be
c
om
e
s
m
or
e
r
e
li
a
bl
e
w
he
n
a
ppl
ie
d
to
r
e
a
l
-
w
or
ld
e
vi
de
nc
e
f
r
om
c
r
im
e
s
c
e
ne
s
.
I
ts
a
bi
li
ty
to
pr
oc
e
s
s
pa
r
ti
a
ll
y
di
s
to
r
te
d
im
a
ge
s
a
ll
ow
s
th
e
m
ode
l
to
s
ti
ll
pr
ovi
de
us
e
f
u
l
in
f
or
m
a
ti
on
f
or
th
e
in
ve
s
ti
ga
ti
on
pr
oc
e
s
s
.
T
hi
s
h
a
s
th
e
pot
e
nt
ia
l
to
im
pr
ove
id
e
nt
if
ic
a
ti
on
a
c
c
ur
a
c
y
a
nd
ove
r
a
ll
e
f
f
ic
ie
nc
y i
n t
he
f
or
e
ns
ic
pr
oc
e
s
s
,
e
ve
n w
he
n t
he
a
va
il
a
bl
e
e
vi
de
nc
e
i
s
f
a
r
f
r
om
i
de
a
l.
T
a
bl
e
2
.
C
om
pa
r
is
on of
m
ode
l
pe
r
f
or
m
a
nc
e
on t
he
S
O
C
O
F
in
g
da
ta
s
e
t
in
pr
e
vi
ous
s
tu
dy
R
e
f
e
r
e
nc
e
s
D
a
t
a
s
e
t
I
nput
M
ode
l
a
r
c
hi
t
e
c
t
ur
e
A
c
c
ur
a
c
y
O
l
a
de
l
e
e
t
al
. [
21]
R
e
a
l
O
nl
y i
m
a
ge
7
-
l
a
ye
r
C
N
N
(
2 c
onv, 2 pool
, 2 F
C
)
72%
I
l
oa
nus
i
a
nd
E
j
i
ogu
[
27]
R
e
a
l
O
nl
y i
m
a
ge
20
-
l
a
ye
r
C
N
N
91
.
3%
T
hongl
i
m
e
t
al
.
[
23]
R
e
a
l
O
nl
y i
m
a
ge
C
N
N
(
2
d
onv, 2
p
ool
, 2
d
e
ns
e
)
99%
G
us
t
i
s
ya
f
a
nd
S
i
na
ga
[
28]
A
l
t
e
r
e
d
O
nl
y i
m
a
ge
C
N
N
w
i
t
h e
ns
e
m
bl
e
a
nd ba
t
c
h nor
m
99
.
9667%
O
ur
m
ode
l
R
e
a
l
+
a
l
t
e
r
e
d
I
m
a
ge
+ha
nd t
ype
+f
i
nge
r
t
ype
7
-
l
a
ye
r
C
N
N
(
3 c
onv a
nd M
a
xP
ool
i
ng,
4 de
ns
e
)
97
.
39%
5.
C
O
N
C
L
U
S
I
O
N
T
hi
s
r
e
s
e
a
r
c
h
f
oc
us
e
s
on
a
c
r
it
ic
a
l
c
ha
ll
e
nge
in
f
or
e
ns
ic
bi
o
m
e
tr
ic
s
,
na
m
e
ly
th
e
de
ve
lo
pm
e
nt
of
a
c
c
ur
a
te
a
nd
r
e
li
a
bl
e
f
in
ge
r
pr
in
t
-
ba
s
e
d
ge
nde
r
c
la
s
s
if
ic
a
ti
on
s
ys
te
m
s
f
or
r
e
a
l
-
w
or
ld
a
ppl
ic
a
ti
ons
.
G
iv
e
n
th
a
t
f
or
e
ns
ic
e
vi
de
nc
e
is
r
a
r
e
ly
f
ound
in
pe
r
f
e
c
t
c
ondi
ti
on,
th
is
r
e
s
e
a
r
c
h
pr
opos
e
s
a
nd
va
li
da
te
s
a
tr
a
in
in
g
m
e
th
odol
ogy
th
a
t
di
r
e
c
tl
y
a
ddr
e
s
s
e
s
th
e
is
s
ue
of
d
a
ta
de
gr
a
d
a
ti
on.
T
he
pr
opos
e
d
m
e
th
odol
ogy
c
om
bi
ne
s
r
a
w
f
in
ge
r
pr
in
t
im
a
ge
s
a
nd
s
ynt
he
ti
c
a
ll
y
m
odi
f
ie
d
im
a
ge
s
f
r
om
t
he
S
O
C
O
F
in
g
da
ta
s
e
t
in
to
a
s
in
gl
e
,
uni
f
ie
d
tr
a
in
in
g
s
e
t.
T
he
C
N
N
m
ode
l
u
s
e
d
i
s
li
ght
w
e
ig
ht
a
nd
c
om
p
ut
a
ti
ona
ll
y
e
f
f
ic
ie
nt
ye
t
a
c
hi
e
ve
s
a
ve
r
y
hi
gh
a
c
c
ur
a
c
y
of
97
.
39%
on
a
te
s
t
da
ta
s
e
t
th
a
t
a
ls
o
in
c
lu
de
s
a
c
o
m
bi
na
ti
on
of
c
le
a
n
a
nd
de
gr
a
de
d
im
a
ge
s
.
T
hi
s
pe
r
f
or
m
a
nc
e
not
onl
y
s
ur
pa
s
s
e
s
m
a
ny
pr
e
vi
ous
s
tu
di
e
s
but
is
a
l
s
o
a
c
hi
e
ve
d
und
e
r
m
or
e
r
e
a
li
s
ti
c
a
nd
c
ha
ll
e
ngi
ng t
e
s
ti
ng c
ondi
ti
ons
. T
he
s
e
f
in
di
ngs
e
m
pha
s
i
z
e
t
ha
t
d
a
ta
s
tr
a
te
gy pla
ys
a
ke
y r
ol
e
i
n buil
di
ng r
obus
t
m
ode
ls
.
B
y
in
c
or
por
a
ti
ng
im
pe
r
f
e
c
t
da
ta
in
to
th
e
tr
a
in
in
g
pr
oc
e
s
s
,
m
ode
l
s
c
a
n
be
de
v
e
lo
pe
d
to
be
in
he
r
e
nt
ly
m
or
e
r
obus
t
a
nd
m
a
in
ta
in
hi
gh
pe
r
f
or
m
a
nc
e
.
T
hi
s
r
e
s
e
a
r
c
h
a
l
s
o
s
e
ts
a
ne
w
be
nc
hm
a
r
k
in
f
in
ge
r
pr
in
t
-
ba
s
e
d
ge
nde
r
c
la
s
s
if
ic
a
ti
on
a
nd
pr
ovi
de
s
a
c
l
e
a
r
f
ounda
ti
on
f
or
th
e
de
ve
lo
pm
e
nt
of
m
or
e
r
e
li
a
bl
e
bi
om
e
tr
ic
s
ys
te
m
s
r
e
a
dy f
or
us
e
i
n r
e
a
l
-
w
or
ld
s
c
e
na
r
io
s
.
F
U
N
D
I
N
G
I
N
F
O
R
M
A
T
I
O
N
T
he
im
pl
e
m
e
nt
a
ti
on
of
th
is
r
e
s
e
a
r
c
h
w
a
s
f
unde
d
by
L
P
P
M
I
ns
ti
tu
t
T
e
knol
ogi
S
a
in
s
da
n
K
e
s
e
ha
ta
n
R
S
dr
.
S
oe
pr
a
oe
n
K
e
s
da
m
V
/B
R
W
to
s
uppor
t
th
e
tr
id
ha
r
m
a
of
hi
ghe
r
e
duc
a
ti
on,
e
s
pe
c
ia
ll
y
in
th
e
a
s
pe
c
ts
of
r
e
s
e
a
r
c
h a
nd s
c
ie
nt
if
ic
publi
c
a
ti
on.
A
U
T
H
O
R
C
O
N
T
R
I
B
U
T
I
O
N
S
S
T
A
T
E
M
E
N
T
T
hi
s
jo
ur
na
l
us
e
s
th
e
C
ont
r
ib
ut
or
R
ol
e
s
T
a
xonomy
(
C
R
e
di
T
)
to
r
e
c
ogni
z
e
in
di
vi
dua
l
a
ut
hor
c
ont
r
ib
ut
io
ns
, r
e
duc
e
a
ut
hor
s
hi
p di
s
put
e
s
,
a
nd f
a
c
il
it
a
te
c
ol
la
bo
r
a
ti
on.
N
am
e
o
f
A
u
t
h
or
C
M
So
Va
Fo
I
R
D
O
E
Vi
Su
P
Fu
R
is
qy S
iwi
P
r
a
di
ni
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
W
a
hyu T
e
ja
K
us
um
a
✓
✓
✓
✓
✓
A
gung S
e
ti
a
B
udi
✓
✓
✓
✓
C
:
C
onc
e
pt
ua
l
i
z
a
t
i
on
M
:
M
e
t
hodol
ogy
So
:
So
f
t
w
a
r
e
Va
:
Va
l
i
da
t
i
on
Fo
:
Fo
r
m
a
l
a
na
l
ys
i
s
I
:
I
nve
s
t
i
ga
t
i
on
R
:
R
e
s
our
c
e
s
D
:
D
a
t
a
C
ur
a
t
i
on
O
:
W
r
i
t
i
ng
-
O
r
i
gi
na
l
D
r
a
f
t
E
:
W
r
i
t
i
ng
-
R
e
vi
e
w
&
E
di
t
i
ng
Vi
:
Vi
s
ua
l
i
z
a
t
i
on
Su
:
Su
pe
r
vi
s
i
on
P
:
P
r
oj
e
c
t
a
dm
i
ni
s
t
r
a
t
i
on
Fu
:
Fu
ndi
ng a
c
qui
s
i
t
i
on
C
O
N
F
L
I
C
T
O
F
I
N
T
E
R
E
S
T
S
T
A
T
E
M
E
N
T
T
he
a
ut
hor
de
c
la
r
e
s
no c
onf
li
c
t
of
i
nt
e
r
e
s
t.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
dv A
ppl
S
c
i
I
S
S
N
:
2252
-
8814
C
onv
ol
ut
io
nal
ne
ur
al
ne
tw
o
r
k
m
ode
l
fo
r
f
in
ge
r
pr
in
t
-
bas
e
d g
e
n
de
r
c
la
s
s
if
ic
at
io
n
…
(
R
is
qy
Si
w
i
P
r
adi
ni
)
1357
D
A
T
A
A
V
A
I
L
A
B
I
L
I
T
Y
T
he
da
ta
s
e
t
us
e
d i
n t
hi
s
s
tu
dy w
a
s
t
a
ke
n f
r
om
pa
pe
r
[
22]
. T
he
da
ta
s
e
t
is
t
he
S
oko F
in
ge
r
pr
in
t
D
a
ta
s
e
t
(
S
O
C
O
F
in
g)
,
w
hi
c
h
c
ont
a
in
s
f
in
ge
r
pr
in
t
im
a
ge
s
in
va
r
io
us
c
on
di
ti
ons
,
in
c
lu
di
ng
bot
h
r
e
a
l
im
a
ge
s
a
nd
a
lt
e
r
e
d
im
a
ge
s
w
it
h
f
e
a
tu
r
e
s
s
uc
h
a
s
obl
it
e
r
a
ti
on,
c
e
nt
r
a
l
r
ot
a
ti
on,
a
nd
z
-
c
ut
.
T
hi
s
da
ta
s
e
t
is
w
id
e
ly
us
e
d
in
f
or
e
ns
ic
bi
om
e
tr
ic
s
r
e
s
e
a
r
c
h be
c
a
us
e
i
t
pr
e
s
e
nt
s
r
e
a
l
c
h
a
ll
e
nge
s
i
n f
in
ge
r
pr
in
t
id
e
nt
if
ic
a
ti
on a
nd c
la
s
s
if
ic
a
ti
on.
R
E
F
E
R
E
N
C
E
S
[
1]
S
.
A
.
A
bdul
r
a
hm
a
n
a
nd
B
.
A
l
ha
ya
ni
,
“
A
c
om
pr
e
he
n
s
i
ve
s
ur
ve
y
on
t
he
bi
om
e
t
r
i
c
s
ys
t
e
m
s
b
a
s
e
d
on
phy
s
i
ol
ogi
c
a
l
a
nd
be
ha
vi
our
a
l
c
ha
r
a
c
t
e
r
i
s
t
i
c
s
,”
M
at
e
r
i
al
s
T
oday
:
P
r
oc
e
e
di
ngs
, vol
. 80, pp. 2642
–
2646, 2023,
doi
:
10.1016/
j
.m
a
t
pr
.2021.07.005.
[
2]
G
.
S
i
ngh,
G
.
B
ha
r
dw
a
j
,
S
.
V
.
S
i
ngh,
a
nd
V
.
G
a
r
g,
“
B
i
om
e
t
r
i
c
i
d
e
nt
i
f
i
c
a
t
i
on
s
ys
t
e
m
:
s
e
c
ur
i
t
y
a
nd
pr
i
va
c
y
c
onc
e
r
n,”
i
n
A
r
t
i
f
i
c
i
al
I
nt
e
l
l
i
ge
nc
e
f
or
a
Sus
t
ai
nabl
e
I
ndus
t
r
y
4.0
,
C
ha
m
:
S
pr
i
nge
r
I
nt
e
r
na
t
i
ona
l
P
ubl
i
s
hi
ng,
2021,
pp.
245
–
264
,
doi
:
10.1007
/
978
-
3
-
030
-
77070
-
9_15.
[
3]
B
.
H
a
s
s
a
n,
E
.
I
z
qui
e
r
do,
a
nd
T
.
P
i
a
t
r
i
k,
“
S
of
t
bi
om
e
t
r
i
c
s
:
a
s
ur
ve
y,”
M
ul
t
i
m
e
di
a
T
ool
s
and
A
ppl
i
c
at
i
ons
,
vol
.
83,
no.
5,
pp
.
15151
–
15194, M
a
r
. 2021, doi
:
10.1007/
s
11042
-
021
-
10622
-
8.
[
4]
R
.
T
ha
kur
,
S
.
K
um
a
r
,
S
.
K
.
S
i
ngh,
K
.
S
i
ngl
a
,
S
.
K
.
S
ha
r
m
a
,
a
nd
V
.
A
r
ya
,
“
C
ybe
r
s
yne
r
gy:
unl
oc
ki
ng
t
he
pot
e
nt
i
a
l
u
s
e
of
bi
om
e
t
r
i
c
s
ys
t
e
m
s
a
nd
m
ul
t
i
m
e
di
a
f
or
e
ns
i
c
s
i
n
c
ybe
r
c
r
i
m
e
i
nve
s
t
i
ga
t
i
ons
,”
D
i
gi
t
al
F
or
e
ns
i
c
s
and
C
y
be
r
C
r
i
m
e
I
nv
e
s
t
i
gat
i
on:
R
e
c
e
nt
A
dv
anc
e
s
and F
ut
ur
e
D
i
r
e
c
t
i
ons
, pp. 241
–
267, 2024, doi
:
10.1201/
9781
003207573
-
12.
[
5]
F
.
B
.
I
bi
t
a
yo,
O
.
A
.
O
l
a
nr
e
w
a
j
u,
a
nd
M
.
B
.
O
ye
l
a
dun,
“
A
f
i
nge
r
pr
i
nt
ba
s
e
d
ge
nde
r
de
t
e
c
t
or
s
ys
t
e
m
us
i
ng
f
i
nge
r
pr
i
nt
pa
t
t
e
r
n
a
na
l
ys
i
s
,”
I
nt
e
r
nat
i
onal
J
our
nal
of
A
dv
anc
e
d
R
e
s
e
ar
c
h
i
n
C
om
put
e
r
Sc
i
e
n
c
e
,
vol
.
13,
no.
4,
pp.
35
–
47,
A
ug.
2022,
doi
:
10.26483/
i
j
a
r
c
s
.v13i
4.6885.
[
6]
S
. H
e
m
a
l
a
t
ha
,
“
A
s
ys
t
e
m
a
t
i
c
r
e
v
i
e
w
o
n f
i
nge
r
p
r
i
n
t
ba
s
e
d bi
om
e
t
r
i
c
a
u
t
he
nt
i
c
a
t
i
o
n
s
ys
t
e
m
,”
i
n
I
nt
e
r
na
t
i
o
na
l
C
onf
e
r
e
nc
e
o
n E
m
e
r
gi
ng
T
r
e
n
ds
i
n
I
n
f
or
m
at
i
o
n
T
e
c
hno
l
o
gy
and
E
ng
i
n
e
e
r
i
ng,
i
c
-
E
T
I
T
E
20
20
,
F
e
b.
2
020
,
p
p.
1
–
4
,
do
i
:
10
.11
09
/
i
c
-
E
T
I
T
E
47
903
.2
020
.3
42.
[
7]
R
.
K
.
V
e
r
m
a
e
t
al
.
,
“
Z
i
nc
oxi
de
(
Z
nO
)
na
nopa
r
t
i
c
l
e
s
:
s
ynt
he
s
i
s
pr
ope
r
t
i
e
s
a
n
d
t
he
i
r
f
or
e
ns
i
c
a
ppl
i
c
a
t
i
ons
i
n
l
a
t
e
nt
f
i
nge
r
pr
i
nt
s
de
ve
l
opm
e
nt
,”
M
at
e
r
i
al
s
T
oday
:
P
r
oc
e
e
di
ng
s
, vol
. 69, pp. 36
–
41, 2022, doi
:
10.1016/
j
.m
a
t
pr
.2022.08.074.
[
8]
S
.
S
ha
r
m
a
,
R
.
S
hr
e
s
t
ha
,
K
.
K
r
i
s
ha
n,
a
nd
T
.
K
a
nc
ha
n,
“
S
e
x
e
s
t
i
m
a
t
i
on
f
r
om
f
i
nge
r
pr
i
nt
r
i
dge
de
ns
i
t
y:
a
r
e
vi
e
w
of
l
i
t
e
r
a
t
ur
e
,”
A
c
t
a
B
i
om
e
di
c
a
, vol
. 92, no. 5, 2021, doi
:
10.23750/
a
bm
.v92i
5.11471.
[
9]
C
.
D
.
A
nj
a
na
,
C
.
V
.
P
r
i
ya
t
ha
,
a
nd
M
.
S
.
S
i
va
P
r
a
s
a
d,
“
A
c
om
pa
r
a
t
i
ve
s
t
udy
on f
r
i
c
t
i
on
r
i
dge
po
r
e
f
e
a
t
ur
e
s
of
m
a
l
e
s
a
nd
f
e
m
a
l
e
s
,”
I
nt
e
r
nat
i
onal
J
our
nal
of
B
i
om
e
t
r
i
c
s
, vol
. 16, no. 2, pp. 158
–
175, 2024, doi
:
10.
1504/
I
J
B
M
.2024.137089.
[
10]
V
. A
. C
ha
va
n a
nd
R
. K
um
a
r
, “
E
xpl
or
i
ng t
he
pot
e
nt
i
a
l
of
r
i
dge
de
ns
i
t
y a
s
a
m
e
a
s
ur
e
of
s
e
x i
de
nt
i
f
i
c
a
t
i
on,”
2020.
[
11]
D
.
D
a
s
e
t
al
.
,
“
S
e
xua
l
di
m
or
phi
s
m
a
nd
t
opol
ogi
c
a
l
v
a
r
i
a
bi
l
i
t
y
i
n
f
i
nge
r
pr
i
nt
r
i
d
ge
de
ns
i
t
y
i
n
a
nor
t
h
-
w
e
s
t
I
ndi
a
n
popul
a
t
i
on,”
T
he
Sc
i
e
nc
e
of
N
at
ur
e
, vol
. 111, no. 3, p. 23, J
un. 2024, doi
:
10.1007/
s
00114
-
024
-
01911
-
x.
[
12]
M
.
E
.
N
a
ndi
,
O
.
O
l
a
bi
yi
,
a
nd
O
.
C
l
e
t
us
,
“
V
a
r
i
a
t
i
on
i
n
t
hum
bp
r
i
nt
pa
t
t
e
r
ns
a
nd
r
i
dge
de
ns
i
t
y
c
ount
s
be
t
w
e
e
n
t
w
o
m
a
j
or
e
t
hni
c
gr
oups
i
n N
i
ge
r
i
a
,”
A
r
t
i
c
l
e
i
n J
our
nal
of
E
x
pe
r
i
m
e
nt
al
M
e
di
c
al
Sc
i
e
nc
e
s
, 2021.
[
13]
L
.
A
kt
e
r
,
M
.
A
.
B
a
s
e
d,
A
.
B
.
M
.
T
.
U
.
I
s
l
a
m
,
a
nd
E
.
U
.
R
a
hm
a
n,
“
C
om
pa
r
a
t
i
ve
a
na
l
ys
i
s
of
f
i
nge
r
pr
i
nt
-
ba
s
e
d
ge
nde
r
c
l
a
s
s
i
f
i
c
a
t
i
on
e
m
pl
oyi
ng c
onvol
ut
i
ona
l
ne
ur
a
l
ne
t
w
or
k,”
i
n
2024 P
ar
ul
I
nt
e
r
nat
i
onal
C
onf
e
r
e
nc
e
on E
ngi
ne
e
r
i
ng and T
e
c
hnol
ogy
, P
I
C
E
T
2024
,
M
a
y 2024, pp. 1
–
5
,
doi
:
10.1109/
P
I
C
E
T
60765.2024.10716057.
[
14]
L
.
B
e
r
r
i
c
he
,
“
C
om
pa
r
a
t
i
ve
s
t
udy
of
f
i
nge
r
pr
i
nt
-
ba
s
e
d
ge
nde
r
i
de
nt
i
f
i
c
a
t
i
on,”
G
e
ne
t
i
c
s
R
e
s
e
a
r
c
h
,
vol
.
2022,
pp.
1
–
9,
D
e
c
.
2022,
doi
:
10.1155/
2022/
1626953.
[
15]
J
.
G
upt
a
,
S
.
P
a
t
ha
k,
a
nd
G
.
K
um
a
r
,
“
D
e
e
p
l
e
a
r
ni
ng
(
C
N
N
)
a
nd
t
r
a
ns
f
e
r
l
e
a
r
ni
ng:
a
r
e
vi
e
w
,”
J
our
nal
of
P
hy
s
i
c
s
:
C
onf
e
r
e
nc
e
Se
r
i
e
s
, vol
. 2273, no. 1, p. 012029,
M
a
y 2022, doi
:
10.1088/
1742
-
6596/
2273/
1/
012029.
[
16]
P
ur
w
ono, A
. M
a
’
a
r
i
f
, W
. R
a
hm
a
ni
a
r
, H
. I
. K
. F
a
t
hur
r
a
hm
a
n, A
. Z
. K
.
F
r
i
s
ky, a
nd Q
. M
.
U
. H
a
q,
“
U
nde
r
s
t
a
ndi
ng of
c
onvol
ut
i
ona
l
ne
ur
a
l
ne
t
w
or
k
(
C
N
N
)
:
a
r
e
vi
e
w
,”
I
nt
e
r
nat
i
onal
J
our
nal
of
R
obot
i
c
s
and
C
ont
r
ol
Sy
s
t
e
m
s
,
vol
.
2,
no.
4,
pp.
739
–
748,
J
a
n.
2022,
doi
:
10.31763/
i
j
r
c
s
.v2i
4.888.
[
17]
D
.
A
r
i
va
l
a
ga
n,
K
.
B
hoop
a
t
hy
B
e
ga
n,
S
.
E
w
i
ns
P
on
P
us
hpa
,
a
nd
K
.
R
a
j
e
ndr
a
n,
“
A
nove
l
i
nt
e
l
l
i
ge
nt
12
-
l
a
ye
r
c
onvol
ut
i
ona
l
ne
ur
a
l
ne
t
w
or
k
m
ode
l
f
or
ge
nde
r
c
l
a
s
s
i
f
i
c
a
t
i
on
us
i
ng
f
i
nge
r
pr
i
nt
i
m
a
ge
s
,”
J
our
nal
of
I
nt
e
l
l
i
ge
nt
and
F
uz
z
y
Sy
s
t
e
m
s
,
vol
.
45,
no.
2,
pp.
2685
–
2706, A
ug. 2023, doi
:
10.3233/
J
I
F
S
-
224284.
[
18]
G
.
J
a
ya
ka
l
a
,
L
.
S
udha
,
a
nd
A
.
P
r
of
e
s
s
or
,
“
G
e
nde
r
c
l
a
s
s
i
f
i
c
a
t
i
on
b
a
s
e
d
on
f
i
nge
r
pr
i
nt
a
na
l
ys
i
s
,”
T
ur
k
i
s
h
J
ou
r
nal
of
C
om
put
e
r
and
M
at
he
m
at
i
c
s
E
duc
at
i
on
, vol
. 12, no. 10, pp. 1249
–
1256, 2021.
[
19]
O
.
S
.
O
l
uf
uns
o,
A
.
E
.
E
vw
i
e
kpa
e
f
e
,
a
nd
M
.
E
.
I
r
he
bhude
,
“
D
e
t
e
r
m
i
na
t
i
on
of
ge
nde
r
f
r
om
f
i
nge
r
pr
i
n
t
s
us
i
ng
dyna
m
i
c
hor
i
z
ont
a
l
vot
i
ng e
ns
e
m
bl
e
de
e
p l
e
a
r
ni
ng a
ppr
oa
c
h,”
I
nt
e
r
nat
i
onal
J
our
nal
of
A
dv
anc
e
s
i
n I
nt
e
l
l
i
ge
nt
I
nf
or
m
at
i
c
s
, vol
. 8, no. 3,
pp. 324
–
336
,
N
ov. 2022, doi
:
10.26555/
i
j
a
i
n.v8i
3.927.
[
20]
R
.
S
i
ngh
,
N
.
S
ha
r
m
a
,
R
.
C
h
a
uh
a
n,
A
.
C
ho
ud
ha
r
y,
a
nd
R
.
G
up
t
a
,
“
E
n
ha
n
c
e
d
f
i
n
ge
r
pr
i
n
t
a
l
t
e
r
a
t
i
o
n
de
t
e
c
t
i
o
n
us
i
ng
l
i
g
ht
w
e
i
gh
t
C
N
N
m
od
e
l
t
r
a
i
ne
d
on
S
O
C
O
F
i
ng
da
t
a
s
e
t
,”
i
n
20
23
3r
d
I
nt
e
r
na
t
i
ona
l
C
o
nf
e
r
e
nc
e
o
n
Sm
ar
t
G
e
ne
r
at
i
o
n
C
om
pu
t
i
n
g,
C
om
m
uni
c
a
t
i
o
n
a
nd
N
e
t
w
or
k
i
n
g,
S
M
A
R
T
G
E
N
C
O
N
20
23
,
D
e
c
.
202
3,
pp
.
1
–
6
,
do
i
:
10
.11
09
/
S
M
A
R
T
G
E
N
C
O
N
6
07
55.
20
23.
10
441
98
0.
[
21]
M
.
O
.
O
l
a
de
l
e
,
T
.
M
.
A
de
poj
u,
O
.
A
.
O
l
a
t
oke
,
O
.
A
.
O
j
o,
a
nd
O
r
i
m
ogunj
e
,
“
C
onvol
ut
i
ona
l
ne
ur
a
l
ne
t
w
or
k
f
or
f
i
nge
r
pr
i
nt
-
ba
s
e
d
ge
nde
r
c
l
a
s
s
i
f
i
c
a
t
i
on,”
Sc
i
e
n
c
e
s
, E
ngi
ne
e
r
i
ng &
E
nv
i
r
on
m
e
nt
al
T
e
c
hnol
ogy
(
I
C
O
N
SE
E
T
)
, vol
. 7, no. 14, pp. 112
–
117, 2022
.
[
22]
Y
.
I
.
S
he
hu,
A
.
R
ui
z
-
G
a
r
c
i
a
,
V
.
P
a
l
a
de
,
a
nd
A
.
J
a
m
e
s
,
“
S
okot
o
C
ove
nt
r
y
F
i
nge
r
pr
i
nt
D
a
t
a
s
e
t
,”
K
a
ggl
e
,
J
ul
.
2018.
[
O
nl
i
ne
]
.
A
va
i
l
a
bl
e
:
ht
t
ps
:
/
/
w
w
w
.ka
ggl
e
.c
om
/
r
ui
z
ga
r
a
/
s
oc
of
i
ng
.
[
23]
P
.
T
hongl
i
m
,
S
.
T
hongs
uw
a
n,
a
nd
P
.
B
ur
a
na
s
i
r
i
,
“
G
e
nde
r
c
l
a
s
s
i
f
i
c
a
t
i
on
us
i
ng
c
onvol
ut
i
ona
l
ne
ur
a
l
ne
t
w
or
ks
ba
s
e
d
on
f
i
nge
r
pr
i
n
t
a
na
l
ys
i
s
w
i
t
h i
n
-
l
i
ne
di
gi
t
a
l
hol
ogr
a
phy,”
i
n
Q
uant
i
t
at
i
v
e
P
has
e
I
m
agi
ng X
, M
a
r
. 2024, p. 100
,
doi
:
10.1117/
12.3010005.
[
24]
M
. A
.
H
a
be
e
b,
Y
.
L
. K
ha
l
e
e
l
, R
.
D
. I
s
m
a
i
l
,
Z
. T
. A
l
-
Q
a
ys
i
,
a
nd F
. N
.
A
m
e
e
n,
“
D
e
e
p l
e
a
r
ni
ng a
ppr
oa
c
h
e
s
f
or
ge
nde
r
c
l
a
s
s
i
f
i
c
a
t
i
on
f
r
om
f
a
c
i
a
l
i
m
a
ge
s
,”
M
e
s
opot
am
i
an J
our
nal
of
B
i
g D
at
a
, vol
. 2024, pp. 185
–
198, O
c
t
. 2024, doi
:
10.58496/
M
J
B
D
/
2024/
013.
[
25]
S
.
K
um
a
r
e
t
al
.
,
“
F
a
c
e
s
poof
i
ng,
a
ge
,
g
e
nde
r
a
nd
f
a
c
i
a
l
e
xpr
e
s
s
i
on
r
e
c
ogni
t
i
o
n
us
i
ng
a
dv
a
nc
e
n
e
ur
a
l
ne
t
w
or
k
a
r
c
hi
t
e
c
t
ur
e
-
ba
s
e
d
bi
om
e
t
r
i
c
s
ys
t
e
m
,”
Se
ns
o
r
s
, vol
. 22, no. 14, p. 5160, J
ul
. 2022, doi
:
10.3390/
s
2
2145160.
[
26]
S
.
A
r
or
a
,
R
.
M
i
t
t
a
l
,
H
.
K
uk
r
e
j
a
,
a
nd
M
.
P
.
S
.
B
ha
t
i
a
,
“
A
n
e
va
l
ua
t
i
on
of
de
noi
s
i
ng
t
e
c
hni
que
s
a
nd
c
l
a
s
s
i
f
i
c
a
t
i
on
of
bi
om
e
t
r
i
c
i
m
a
ge
s
ba
s
e
d
on
de
e
p
l
e
a
r
ni
ng,”
M
ul
t
i
m
e
di
a
T
ool
s
and
A
ppl
i
c
at
i
ons
,
vol
.
82,
no.
6,
pp.
8287
–
8302,
M
a
r
.
2023,
doi
:
10.1007/
s
11042
-
021
-
11573
-
w.
[
27]
O
. N
. I
l
oa
nus
i
a
nd U
.
C
. E
j
i
ogu, “
G
e
nde
r
c
l
a
s
s
i
f
i
c
a
t
i
on f
r
om
f
us
e
d m
ul
t
i
-
f
i
nge
r
pr
i
nt
t
ype
s
,”
I
nf
or
m
at
i
on Se
c
u
r
i
t
y
J
our
nal
, vol
.
29,
no. 5, pp. 209
–
219, S
e
p. 2020, doi
:
10.1080/
19393555.2020.1741742.
[
28]
A
.
I
.
G
us
t
i
s
ya
f
a
nd
A
.
S
i
na
ga
,
“
I
m
pl
e
m
e
nt
a
t
i
on
of
c
onvol
ut
i
ona
l
ne
ur
a
l
ne
t
w
or
k
t
o
c
l
a
s
s
i
f
i
c
a
t
i
on
ge
nde
r
ba
s
e
d
on
f
i
nge
r
pr
i
nt
,
”
I
nt
e
r
nat
i
onal
J
our
nal
of
M
ode
r
n
E
duc
at
i
on
and
C
om
put
e
r
Sc
i
e
nc
e
,
vol
.
13,
no.
4,
pp.
55
–
67,
A
ug.
2021,
doi
:
10.5815/
I
J
M
E
C
S
.2021.04.05.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8814
I
nt
J
A
dv A
ppl
S
c
i
,
V
ol
.
14
, N
o.
4
,
D
e
c
e
m
be
r
20
25
:
1350
-
1358
1358
B
I
O
G
R
A
P
H
I
E
S
O
F
A
U
T
H
O
R
S
Risqy
Siwi
Pradini
is
a
lecturer
in
the
Informatics
Study
Progr
am,
Faculty
of
Scienc
e
and
Techn
ology,
ITSK
Soepra
oen.
She
earn
ed
a
Bache
lor
of
Applied
Scienc
e
degre
e
from
Politeknik
Negeri
Malang
and
a
Master
of
Computer
Science
degree
from
Universitas
Brawijay
a.
She
is
cu
rrently
pursuing
r
esear
ch
in
Informa
tion
Systems
,
Machine
Learn
ing,
and
Deep Learning. She can be contacted
at email: risqy
pradini@
itsk
-
soepraoen.a
c.id.
Wahyu
Teja
Kusuma
is
a
lecturer
at
ITSK
Soepraoen.
He
currentl
y
serves
as
the
Head
of
the
LPPM
ITSK
Soepraoen.
He
earned
his
bachelor'
s
a
nd
master'
s
degrees
in
computer
science from
Universi
tas Brawi
jaya. He i
s currentl
y compl
eting h
is doct
oral prog
ram
in
computer
science
at
Universitas
Brawijaya.
He
can
be
contacted
at
email:
wtkusuma@
itsk
-
soepraoen.a
c.id.
Agung
Setia
Budi
is
a
Bachelor
of
Electrical
Engineering
graduate
from
Brawijay
a
University
in
2009.
He
graduated
from
the
Graduate
School
of
Engineering,
University
of
Miyazaki
in
2012
and
Master
of
Electrical
Engineering,
University
of
Brawijaya
in
2013.
He
then
continu
ed
his
doctoral
studies
and
graduated
fr
om
the
Interdisciplinary
Graduate
School
of
Agriculture
and
Engineering,
University
of
Miya
zaki
in
2019.
Currently,
he
is
active
as
a
lecturer
at
the
Faculty
of
Computer
Science,
Brawija
ya
University.
He
can
be
contacted
at email
: agungs
etiabudi
@
ub.ac.id.
Evaluation Warning : The document was created with Spire.PDF for Python.