I
A
E
S
I
n
t
e
r
n
at
io
n
al
Jou
r
n
al
of
A
r
t
if
ic
ia
l
I
n
t
e
ll
ig
e
n
c
e
(
I
J
-
AI
)
V
ol
.
14
, N
o.
4
,
A
ugus
t
2025
, pp.
3262
~
3273
I
S
S
N
:
2252
-
8938
,
D
O
I
:
10.11591/
ij
a
i.
v
14
.i
4
.pp
3262
-
3273
3262
Jou
r
n
al
h
om
e
page
:
ht
tp
:
//
ij
ai
.
ia
e
s
c
or
e
.c
om
E
n
h
a
n
c
i
n
g t
ou
c
h
l
e
ss
s
m
ar
t
l
oc
k
e
r
sys
t
e
m
s t
h
r
ou
gh
ad
van
c
e
d
f
ac
i
al
r
e
c
ogn
i
t
i
on
t
e
c
h
n
ol
ogy:
a c
on
vol
u
t
i
on
al
n
e
u
r
al
n
e
t
w
or
k
m
od
e
l
ap
p
r
oac
h
A
b
d
u
l
H
ar
is
R
an
gk
u
t
i,
E
vaw
at
y T
an
u
ar
,
F
e
b
r
ia
n
t
Y
ap
s
on
,
F
e
li
x O
c
t
avi
o S
ij
oa
t
m
od
j
o,
V
ar
yl
H
as
b
i
A
t
h
al
a
D
e
pa
r
t
m
e
nt
of
C
om
put
e
r
S
c
i
e
nc
e
, S
c
hool
of
C
om
put
e
r
s
S
c
i
e
nc
e
, B
i
na
N
u
s
a
nt
a
r
a
U
ni
ve
r
s
i
t
y,
B
a
ndung
, 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
M
a
y 7, 2024
R
e
vi
s
e
d
M
a
r
29, 2025
A
c
c
e
pt
e
d
J
un 8, 2025
As
the
world
recovers
from
COVID
-
19,
demand
for
contactless
syst
ems
is
increasin
g,
promising
safety
and
convenien
ce.
Touchless
techn
ology,
particularly
public
locker
security
systems
that
use
facial
recogniti
on
and
hand
detection,
is
advancing
rapidly.
The
system
minimizes
physical
contact,
increasing
user
safety.
It
uses
advanced
models
such
as
multi
-
task
cascaded
convolut
ional
networks
(
MTCNN
)
and
RetinaFace,
Face
N
et512,
ArcFace,
and
visual
geometry
group
(
VGG
)
-
Face
for
face
detecti
on
and
recognition
,
wit
h
a
combination
of
RetinaFace,
ArcFace,
and
L2
n
orm
Euclidean
or
cosine
as
th
e
most
effective
distance
metric
method,
wh
ere
the
accuracy
reaches
96
and
90%.
'
Yourvault'
,
an
application
demonstrati
ng
this
efficient
security
feature,
provides
notificat
ions
for
mask
detection
,
facial
authenti
city
and
locker
status,
offering
a
soluti
on
to
the
probl
em
of
convenience
and
security
of
public
spaces.
Future
research
could
inve
stigat
e
the
impact
of
photo
age
on
facial
recognition
accuracy,
potentially
making
touchles
s
systems
more
efficient.
In
general,
the
application
o
f
this
technology
is
an
important
step
towards
a
safer
and
more
comfortab
le
world
after
the
pandemic.
This
model
approach
can
be
followed
up
with
more
optimal facial recognition.
K
e
y
w
o
r
d
s
:
F
a
c
ia
l
L
oc
ke
r
M
ul
ti
-
ta
s
k c
a
s
c
a
de
d
c
onvolut
io
na
l
ne
twor
ks
R
e
ti
na
F
a
c
e
VGG
-
f
a
c
e
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
bdul
H
a
r
is
R
a
ngkuti
D
e
pa
r
tm
e
nt
of
C
om
put
e
r
S
c
ie
nc
e
, S
c
hool
of
C
om
put
e
r
s
S
c
ie
nc
e
, B
in
a
N
us
a
nt
a
r
a
U
ni
ve
r
s
it
y
P
a
s
ir
ka
li
ki
S
t.
N
o. 25
-
27, P
a
s
ka
l
H
ype
r
S
qua
r
e
B
a
ndung 40181
,
W
e
s
t
J
a
v
a
, I
ndone
s
ia
E
m
a
il
:
r
a
ngku2000@
bi
nus
.a
c
.i
d
1.
I
N
T
R
O
D
U
C
T
I
O
N
T
he
a
dopt
in
g
a
r
ti
f
ic
ia
l
in
te
ll
ig
e
nc
e
a
s
a
s
ys
t
e
m
f
or
c
ont
r
ol
li
ng
in
te
r
ne
t
of
th
in
gs
(
I
oT
)
de
vi
c
e
s
ha
s
be
e
n
a
le
s
s
f
a
vor
e
d
a
lt
e
r
na
ti
ve
due
to
it
s
c
om
pa
r
a
ti
ve
ly
t
im
e
-
in
te
ns
iv
e
na
tu
r
e
w
he
n
ju
xt
a
pos
e
d
w
it
h
bi
om
e
tr
ic
-
ba
s
e
d
s
e
c
ur
it
y
m
e
a
s
ur
e
s
li
ke
f
in
ge
r
pr
in
t
r
e
c
ogni
ti
on
[
1]
.
H
ow
e
ve
r
,
s
in
c
e
th
e
ons
e
t
of
th
e
C
O
V
I
D
-
19
pa
nde
m
ic
,
a
pa
r
a
di
gm
s
hi
f
t
ha
s
pr
om
pt
e
d
a
s
ig
ni
f
ic
a
nt
tr
a
ns
f
or
m
a
ti
on
of
s
ys
te
m
s
th
a
t
tr
a
di
ti
ona
ll
y
r
e
qui
r
e
d
phys
ic
a
l
c
ont
a
c
t
in
to
ope
r
a
ti
ons
de
voi
d
of
s
uc
h
phys
ic
a
l
in
te
r
a
c
ti
on
[
2]
.
V
oi
c
e
a
nd
im
a
ge
-
ba
s
e
d
c
ont
r
ol
a
nd
s
e
c
ur
it
y
s
y
s
te
m
s
ha
ve
e
xpe
r
ie
nc
e
d
r
a
pi
d
pr
ol
if
e
r
a
ti
on
dur
in
g
th
is
pe
r
io
d,
dr
iv
e
n
by
he
ig
ht
e
ne
d
de
m
a
nd.
C
ons
e
que
nt
ly
,
th
e
im
pe
r
a
ti
ve
f
or
de
ve
lo
pi
ng
in
te
gr
a
te
d
s
e
c
ur
it
y
s
ys
te
m
s
w
it
hi
n
th
e
I
oT
f
r
a
m
e
w
or
k,
bol
s
te
r
e
d
by
a
r
ti
f
ic
ia
l
in
te
ll
ig
e
nc
e
,
ha
s
be
c
om
e
pa
r
a
m
ount
,
a
im
in
g
to
s
upe
r
s
e
de
te
c
hnol
ogi
e
s
s
ti
ll
r
e
li
a
nt
on
phys
ic
a
l
to
u
c
h
[
3]
.
O
ne
s
ys
te
m
th
a
t
s
ti
ll
ne
c
e
s
s
it
a
te
s
phys
ic
a
l
to
uc
h
i
s
th
e
lo
c
ki
ng
m
e
c
ha
ni
s
m
e
m
pl
oye
d
in
publ
ic
lo
c
ke
r
s
.
A
n
a
lt
e
r
na
ti
ve
to
r
e
pl
a
c
e
phys
ic
a
l
ke
ys
in
vol
ve
s
us
in
g
pa
s
s
c
od
e
s
f
or
lo
c
ke
r
a
c
c
e
s
s
[
3]
.
H
ow
e
ve
r
,
e
nt
e
r
in
g
a
pa
s
s
c
ode
s
ti
ll
r
e
qui
r
e
s
phy
s
ic
a
l
to
uc
h
a
nd
r
e
m
a
in
s
s
us
c
e
pt
ib
le
to
br
e
a
c
he
s
,
r
e
nde
r
in
g
it
s
ubopti
m
a
l
a
s
a
n
a
lt
e
r
na
ti
ve
.
A
not
he
r
pr
opos
e
d
opt
io
n
is
us
in
g
r
a
di
o
f
r
e
que
nc
y
id
e
nt
if
ic
a
ti
on
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
E
nhanc
in
g t
ouc
hl
e
s
s
s
m
a
r
t
lo
c
k
e
r
s
y
s
te
m
s
t
hr
ough adv
anc
e
d f
ac
ia
l
r
e
c
ogni
ti
on
…
(
A
bdul
H
ar
is
R
angk
ut
i
)
3263
(
R
F
I
D
)
c
a
r
ds
,
s
uc
h
a
s
th
e
e
l
e
c
tr
oni
c
id
e
nt
it
y
c
a
r
d
(E
-
K
T
P
)
,
a
s
a
s
ubs
ti
tu
te
f
or
phys
ic
a
l
ke
y
s
[
4]
.
W
hi
le
th
is
a
lt
e
r
na
ti
ve
e
f
f
ic
ie
nt
ly
r
e
duc
e
s
phys
ic
a
l
c
ont
a
c
t,
th
e
s
e
c
ur
it
y
of
R
F
I
D
c
a
r
ds
is
s
ti
ll
s
us
c
e
pt
ib
le
to
dupl
ic
a
ti
on,
th
us
f
a
il
in
g
to
a
c
hi
e
ve
opt
im
a
l
s
e
c
ur
it
y
le
ve
ls
.
A
not
he
r
a
lt
e
r
na
ti
ve
,
w
hi
c
h
obvi
a
te
s
th
e
ne
e
d
f
or
phys
ic
a
l
to
uc
h
w
hi
le
m
a
in
ta
in
in
g
r
obus
t
s
e
c
ur
it
y,
is
qui
c
k
r
e
s
pons
e
(
QR
)
-
ba
s
e
d
s
e
c
ur
it
y
[
5]
.
Q
R
c
ode
s
m
in
im
iz
e
phys
ic
a
l
c
ont
a
c
t
to
th
e
ut
m
o
s
t
e
xt
e
nt
,
e
ns
ur
in
g
m
a
in
ta
in
e
d
s
e
c
ur
it
y,
a
lb
e
it
ne
c
e
s
s
it
a
ti
ng
a
ddi
ti
ona
l
s
te
ps
f
or
ope
r
a
ti
on.
C
ons
e
que
nt
ly
,
ongoing
de
ve
lo
pm
e
nt
in
publ
ic
lo
c
ke
r
lo
c
ki
ng
s
ys
te
m
s
is
im
pe
r
a
ti
ve
to
e
ns
ur
e
s
e
c
ur
e
,
c
onv
e
ni
e
nt
us
a
g
e
w
it
h
m
in
im
a
l
phys
ic
a
l
to
uc
hpoi
nt
s
.
I
n
th
is
e
r
a
,
te
c
hnol
ogi
c
a
l
pr
ogr
e
s
s
is
s
w
if
tl
y
a
dva
nc
in
g,
pr
ope
ll
e
d
by
th
e
pe
r
va
s
iv
e
in
te
gr
a
ti
on
of
a
r
ti
f
ic
ia
l
in
te
ll
ig
e
nc
e
a
c
r
os
s
m
ul
ti
f
a
r
io
us
dom
a
in
s
,
m
os
t
not
a
bl
y
on
th
e
I
oT
r
e
a
lm
.
T
he
in
f
us
io
n
of
a
r
ti
f
ic
ia
l
in
te
ll
ig
e
n
c
e
in
to
I
oT
s
ys
te
m
s
be
g
e
ts
a
pa
r
a
di
gm
s
hi
f
t,
f
a
c
il
it
a
ti
ng
th
e
s
e
a
m
le
s
s
a
nd
e
f
f
ic
ie
nt
pr
oc
e
s
s
in
g
of
r
e
a
l
-
ti
m
e
da
ta
[
6]
.
B
e
yond
th
is
,
in
c
or
po
r
a
ti
ng
a
r
ti
f
ic
ia
l
in
te
ll
ig
e
nc
e
in
th
e
I
oT
m
il
ie
u
a
ugm
e
nt
s
da
ta
s
e
c
ur
it
y
w
it
hi
n
th
e
s
ys
te
m
a
nd
or
c
he
s
tr
a
te
s
a
ju
di
c
io
us
opt
im
iz
a
ti
on
of
pow
e
r
ut
il
iz
a
ti
on
[
7]
.
C
ons
e
que
nt
ly
,
th
e
or
c
he
s
tr
a
ti
on
of
I
oT
s
ys
te
m
s
is
poi
s
e
d
f
or
r
e
f
in
e
m
e
nt
,
c
e
nt
e
r
in
g
it
s
f
oc
us
on
da
ta
a
c
qui
s
it
io
n
by
ut
il
iz
in
g
in
te
gr
a
te
d,
e
ne
r
gy
-
e
f
f
ic
ie
nt
pl
a
tf
o
r
m
s
s
uc
h
a
s
A
r
dui
no,
e
s
pr
e
s
s
if
s
y
s
te
m
s
(
E
S
P
)
, a
nd R
a
s
pbe
r
r
y
[
8]
.
T
he
pr
opos
e
d
m
e
th
odol
ogi
c
a
l
a
ppr
oa
c
h
in
th
i
s
r
e
s
e
a
r
c
h
in
vo
lv
e
s
de
ve
lo
pi
ng
a
f
a
c
e
r
e
c
ogni
ti
on
-
ba
s
e
d
lo
c
ki
ng
s
y
s
te
m
w
it
h
ha
nd
de
te
c
ti
on
a
s
it
s
na
vi
ga
ti
ona
l
s
ys
te
m
.
I
n
a
pr
e
vi
ous
s
tu
dy
on
c
r
e
a
ti
ng
a
f
a
c
e
r
e
c
ogni
ti
on
-
ba
s
e
d
lo
c
ke
r
s
y
s
te
m
[
9]
,
th
e
na
vi
ga
ti
on
s
ys
t
e
m
a
n
d
us
e
r
in
te
r
f
a
c
e
(
U
I
)
s
ti
ll
ne
c
e
s
s
it
a
te
d
phys
ic
a
l
c
ont
a
c
t.
A
s
a
n a
ddi
ti
ona
l
s
e
c
ur
it
y m
e
a
s
ur
e
, l
iv
e
li
ne
s
s
de
te
c
ti
on
r
e
s
e
a
r
c
h
[
10]
ha
s
be
e
n e
xpl
or
e
d t
o pr
e
ve
nt
t
he
us
e
of
phot
ogr
a
phs
to
a
c
c
e
s
s
lo
c
k
e
r
s
.
H
ow
e
v
e
r
,
th
is
s
tu
dy
r
e
li
e
s
s
ol
e
ly
on
bl
in
k
de
te
c
ti
on,
le
a
vi
ng
it
vul
ne
r
a
bl
e
to
e
xpl
oi
ta
ti
on
th
r
ough
vi
de
o
-
ba
s
e
d
a
ppr
oa
c
he
s
.
I
nc
or
por
a
ti
ng
a
ddi
ti
ona
l
s
e
c
ur
it
y
f
e
a
tu
r
e
s
is
c
r
uc
ia
l,
e
s
pe
c
ia
ll
y
c
on
s
id
e
r
in
g
r
e
por
ts
w
he
r
e
in
a
n
e
le
m
e
nt
a
r
y
s
c
hool
s
tu
de
nt
br
e
a
c
he
d
a
f
a
c
e
r
e
c
ogni
ti
on
-
ba
s
e
d
pa
c
ka
ge
lo
c
k
e
r
us
in
g
onl
y
a
phot
ogr
a
ph
[
11]
.
T
h
e
r
e
f
or
e
,
th
e
m
e
th
odol
ogi
c
a
l
a
ppr
oa
c
h
unde
r
de
ve
lo
pm
e
nt
in
th
is
r
e
s
e
a
r
c
h
is
im
pe
r
a
ti
ve
to
e
nha
nc
e
lo
c
ke
r
s
e
c
ur
it
y
a
nd
c
ons
tr
uc
t
a
s
ys
te
m
th
a
t
e
li
m
in
a
te
s
th
e
ne
e
d f
or
phys
ic
a
l
to
uc
h i
n i
ts
l
oc
ki
ng me
c
ha
ni
s
m
.
T
he
l
oc
k
e
r
s
ys
te
m
de
ve
lo
pe
d i
n t
hi
s
r
e
s
e
a
r
c
h c
om
pr
is
e
s
two
m
a
in
c
om
pone
nt
s
:
th
e
lo
c
ke
r
s
y
s
te
m
a
nd
th
e
Y
our
va
ul
t
bo
ot
h
f
or
lo
c
ke
r
a
c
c
e
s
s
.
T
he
Y
our
va
ul
t
boot
h
is
c
ons
tr
uc
te
d
us
in
g
f
a
c
e
r
e
c
ogni
ti
on
f
r
om
th
e
L
ig
ht
F
a
c
e
f
r
a
m
e
w
or
k,
M
a
s
k
a
nd
F
a
ke
f
a
c
e
de
te
c
ti
on
ut
il
iz
in
g
you
onl
y
lo
ok
onc
e
ve
r
s
io
n
8
(
Y
O
L
O
v8
)
,
a
nd
ha
nd
de
te
c
ti
on
us
in
g
M
e
di
a
pi
pe
.
T
he
lo
c
ke
r
s
ys
te
m
is
bui
lt
w
it
h
E
S
P
8266,
s
e
r
vi
ng
a
s
th
e
c
e
nt
r
a
l
c
ont
r
ol
uni
t
f
or
lo
c
ke
r
ke
ys
.
T
he
boot
h
s
ys
t
e
m
c
a
n
c
onne
c
t
w
it
h
m
ul
ti
pl
e
lo
c
ke
r
s
,
pr
ovi
de
d
th
e
lo
c
ke
r
s
a
nd
boot
hs
a
r
e
in
th
e
e
xa
c
t
s
a
m
e
r
e
gi
s
te
r
e
d
lo
c
a
ti
on.
F
ir
e
ba
s
e
f
ir
e
s
to
r
e
a
nd f
ir
e
ba
s
e
r
e
a
l
-
ti
m
e
da
ta
ba
s
e
gove
r
n t
he
e
c
o
s
ys
te
m
of
Y
our
v
a
ul
t.
2.
L
I
T
E
R
A
T
U
R
E
R
E
V
I
E
W
A
c
om
pr
e
he
ns
iv
e
in
ve
s
ti
ga
ti
on
in
to
im
a
ge
pr
oc
e
s
s
in
g,
c
ha
r
a
c
te
r
iz
e
d
by
pr
e
c
is
io
n
a
nd
r
e
le
va
nc
e
,
be
c
om
e
s
im
pe
r
a
ti
ve
to
im
pl
e
m
e
nt
a
to
uc
hl
e
s
s
pa
r
a
di
gm
w
i
th
in
lo
c
ke
r
s
e
c
ur
it
y
s
y
s
te
m
s
.
T
he
to
uc
hl
e
s
s
na
vi
ga
ti
on f
r
a
m
e
w
or
k unde
r
e
xa
m
in
a
ti
on i
n t
hi
s
r
e
s
e
a
r
c
h i
s
gr
o
unde
d i
n t
he
nua
nc
e
d r
e
a
lm
of
ha
nd de
te
c
ti
on.
T
hi
s
s
e
le
c
ti
on
s
te
m
s
f
r
om
it
s
in
he
r
e
nt
a
tt
r
ib
ut
e
s
of
e
f
f
ic
ie
nc
y,
us
e
r
in
tu
it
iv
e
ne
s
s
,
a
nd
a
r
e
m
a
r
ka
bl
e
a
bi
li
ty
t
o
c
a
te
r
t
o i
ndi
vi
dua
ls
f
a
c
in
g c
om
m
uni
c
a
ti
ve
c
ha
ll
e
nge
s
[
12]
.
T
o
a
ugm
e
nt
ha
nd de
te
c
ti
on c
a
pa
bi
li
ti
e
s
, t
hi
s
s
tu
dy
in
te
gr
a
te
s
M
e
di
a
pi
pe
te
c
hnol
ogy,
r
e
now
ne
d
f
or
it
s
c
om
m
e
nd
a
bl
e
a
ve
r
a
ge
a
c
c
ur
a
c
y
of
up
to
99%
,
th
e
r
e
by
f
or
ti
f
yi
ng
th
e
f
ounda
ti
on
o
f
r
e
li
a
bi
li
ty
[
13]
.
F
ur
th
e
r
m
o
r
e
,
th
is
r
e
s
e
a
r
c
h
e
nde
a
vor
s
to
e
le
va
te
th
e
to
uc
hl
e
s
s
lo
c
ke
r
e
xpe
r
ie
nc
e
by
in
tr
oduc
in
g
a
s
ophi
s
ti
c
a
te
d
f
a
c
e
r
e
c
ogni
t
io
n
s
ys
te
m
a
s
th
e
li
nc
hpi
n
f
or
s
e
c
ur
in
g
th
e
s
e
s
to
r
a
ge
uni
ts
.
T
he
f
r
a
m
e
w
or
k
of
c
hoi
c
e
f
or
th
is
pu
r
pos
e
is
th
e
il
lu
s
tr
io
us
L
ig
ht
F
a
c
e
,
m
e
ti
c
ul
ous
ly
c
ur
a
te
d
f
o
r
it
s
ni
m
bl
e
c
ha
r
a
c
te
r
,
ve
r
s
a
ti
le
s
uppor
t
f
or
f
a
c
e
r
e
c
ogni
ti
on
m
o
de
ls
,
pr
of
ic
ie
nt
m
e
tr
ic
d
is
ta
nc
e
c
om
put
a
ti
ons
,
a
nd a
de
pt
f
a
c
e
de
te
c
ti
on pr
ow
e
s
s
[
14]
.
A
s
a
n
a
ddi
ti
ona
l
la
ye
r
of
s
e
c
ur
it
y,
th
is
r
e
s
e
a
r
c
h
in
te
gr
a
te
s
a
m
a
s
k
de
te
c
ti
on
a
nd
f
a
c
i
a
l
a
ut
he
nt
ic
it
y
s
ys
te
m
us
in
g
Y
O
L
O
v8.
T
he
ut
il
iz
a
ti
on
of
Y
O
L
O
v8
in
di
s
c
e
r
ni
ng
m
a
s
k
us
a
ge
e
xhi
bi
t
s
th
e
c
a
pa
bi
li
ty
to
a
c
c
ur
a
te
ly
c
la
s
s
if
y
in
di
vi
dua
ls
c
or
r
e
c
tl
y
w
e
a
r
in
g
m
a
s
ks
,
th
os
e
w
it
h
im
pr
ope
r
m
a
s
k
pl
a
c
e
m
e
nt
,
a
nd
th
os
e
not
w
e
a
r
in
g
m
a
s
ks
,
a
c
hi
e
vi
ng
hi
gh
pr
e
c
is
io
n
a
nd
a
c
c
ur
a
c
y
[
15]
.
M
a
s
k
us
a
ge
de
te
c
ti
on
e
nha
nc
e
s
f
a
c
e
r
e
c
ogni
ti
on
pe
r
f
or
m
a
nc
e
by
c
om
pr
e
he
ns
iv
e
ly
r
e
m
ovi
ng
th
e
us
e
r
'
s
m
a
s
k
w
he
n
ne
c
e
s
s
a
r
y.
E
m
pl
oyi
ng
Y
O
L
O
v8
f
o
r
f
a
c
ia
l
a
ut
he
nt
ic
it
y
de
te
c
ti
on
a
im
s
to
m
it
ig
a
te
th
e
r
is
k
of
una
ut
hor
iz
e
d
da
ta
us
a
ge
th
r
ough
m
e
di
a
s
uc
h
a
s
phot
os
a
nd
vi
de
os
.
M
a
s
k
us
a
g
e
de
te
c
ti
on
e
nha
n
c
e
s
f
a
c
e
r
e
c
ogni
ti
on
pe
r
f
or
m
a
nc
e
by
c
om
pr
e
he
n
s
iv
e
ly
r
e
m
ovi
ng
th
e
us
e
r
'
s
m
a
s
k w
he
n ne
c
e
s
s
a
r
y. E
m
pl
oyi
ng Y
O
L
O
v8 f
or
f
a
c
ia
l
a
ut
he
nt
ic
it
y de
te
c
ti
on a
im
s
t
o m
it
ig
a
te
t
he
r
is
k of
una
ut
hor
iz
e
d
da
ta
us
a
g
e
th
r
ough
m
e
di
a
s
uc
h
a
s
phot
o
s
a
nd
vi
d
e
os
.
A
n
ove
r
vi
e
w
of
r
e
s
e
a
r
c
h
r
e
la
te
d
to
f
a
c
ia
l
r
e
c
ogni
ti
on t
e
c
hni
que
s
, w
hi
c
h i
s
a
s
e
c
ur
it
y ke
y t
e
c
hni
que
f
or
doi
ng s
om
e
th
in
g c
a
n be
s
e
e
n i
n T
a
bl
e
1.
T
he
m
ode
l
s
ut
il
i
z
e
d
in
th
i
s
r
e
s
e
a
r
c
h
e
nc
om
p
a
s
s
F
a
c
e
N
e
t5
12,
A
r
c
F
a
c
e
,
a
n
d
V
G
G
-
F
a
c
e
,
c
o
upl
e
d
w
it
h
va
r
io
u
s
di
s
t
a
nc
e
m
e
tr
i
c
s
a
va
il
a
bl
e
in
th
e
L
ig
ht
F
a
c
e
f
r
a
m
e
w
or
k,
s
uc
h
a
s
c
o
s
in
e
,
E
u
c
li
d
e
a
n
,
a
nd
E
uc
li
de
a
n
L
2
nor
m
.
D
r
a
w
in
g
on
pr
e
vi
o
us
r
e
s
e
a
r
c
h
f
in
di
ng
s
,
F
a
c
e
N
e
t
512
e
x
hi
bi
ts
a
n
a
c
c
ur
a
c
y
r
a
te
of
up
to
97%
,
w
he
r
e
a
s
A
r
c
F
a
c
e
de
m
ons
tr
a
te
s
a
c
c
ur
a
c
y
le
ve
l
s
r
e
a
c
hi
ng
8
8%
[
16]
.
H
ow
e
v
e
r
,
a
r
e
por
t
f
r
om
th
e
S
in
ga
p
or
e
de
f
e
ns
e
s
c
i
e
nc
e
a
nd t
e
c
hn
ol
ogy a
ge
n
c
y
r
e
g
a
r
di
ng t
h
e
r
e
s
il
i
e
n
c
e
of
f
a
c
e
r
e
c
ogn
it
io
n m
e
t
hod
s
a
g
a
in
s
t
noi
s
e
s
u
gge
s
t
s
t
ha
t
V
G
G
-
F
a
c
e
boa
s
ts
t
h
e
hi
g
he
s
t
a
c
c
ur
a
c
y r
a
te
, r
e
a
c
hi
ng 82%
c
om
p
a
r
e
d t
o F
a
c
e
N
e
t
a
nd A
r
c
F
a
c
e
[
17]
.
H
e
n
c
e
, t
hi
s
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
,
V
ol
. 14, No. 4, A
ugus
t
2025
:
3262
-
3273
3264
s
tu
dy
s
tr
a
te
gi
c
a
ll
y
in
c
or
por
a
te
s
a
ll
th
r
e
e
m
od
e
l
s
to
a
s
c
e
r
t
a
in
t
h
e
m
e
th
odol
o
gy
th
a
t
a
tt
a
in
s
h
ig
h
a
c
c
ur
a
c
y
a
nd
de
m
on
s
tr
a
t
e
s
r
e
s
il
ie
n
c
e
a
g
a
in
s
t
noi
s
e
, a
li
gni
ng
w
it
h t
he
c
ha
r
a
c
t
e
r
is
ti
c
s
of
t
he
c
a
m
e
r
a
d
e
vi
c
e
s
i
n
us
e
.
T
a
bl
e
1. C
om
pa
r
is
on
a
na
ly
s
i
s
of
pr
e
vi
ous
m
e
th
ods
R
e
f
e
r
e
nc
e
s
C
ondi
t
i
on
M
e
t
hods
A
c
c
ur
a
c
y
[
16]
N
or
m
a
l
F
a
c
e
N
e
t
512
A
r
c
F
a
c
e
97.4%
87.8%
[
17]
N
oi
s
e
0%
, 25%
, 50%
, 75%
a
nd 100%
F
a
c
e
N
e
t
512
A
r
c
F
a
c
e
VGG
-
F
a
c
e
< 60%
80.4%
81.6%
[
18]
L
a
be
l
e
d f
a
c
e
s
i
n t
he
w
i
l
d (
L
F
W
)
a
nd c
e
l
e
br
i
t
i
e
s
i
n f
r
ont
a
l
-
pr
of
i
l
e
i
n t
he
w
i
l
d (
C
F
P
-
F
P
)
M
T
C
N
N
+A
r
c
F
a
c
e
R
e
t
i
na
F
a
c
e
+A
r
c
F
a
c
e
L
F
W
:
99.83%
C
F
P
-
F
P
:
98.37%
L
F
W
:
99.86%
C
F
P
-
F
P
:
99.49%
[
19]
L
F
W
M
T
C
N
N
R
e
t
i
na
F
a
c
e
E
a
s
y:
85.1%
H
a
r
d:
60.7%
E
a
s
y:
87.8%
H
a
r
d:
47.3%
[
20]
T
e
s
t
i
ng us
i
ng 5 di
f
f
e
r
e
nt
da
t
a
s
e
t
s
A
r
c
F
a
c
e
F
a
c
e
N
e
t
512
VGG
-
F
a
c
e
L
ow
e
s
t
va
l
u
e
:
97.4%
85.45%
93.95%
[
21]
T
e
s
t
i
ng on 200 f
a
c
e
s
a
m
pl
e
s
M
T
C
N
N
+A
r
c
F
a
c
e
R
e
t
i
na
F
a
c
e
+A
r
c
F
a
c
e
70
-
90%
80
-
100%
O
th
e
r
s
tu
di
e
s
de
s
c
r
ib
e
th
e
po
s
it
io
n
of
th
e
c
la
s
s
r
oom
s
ur
ve
il
la
nc
e
c
a
m
e
r
a
a
s
not
f
ix
e
d.
S
o
th
a
t
th
e
a
ngl
e
of
th
e
f
a
c
e
r
e
c
or
de
d
th
r
ough
th
e
s
ur
ve
il
la
nc
e
vi
de
o
is
d
if
f
e
r
e
nt
.
T
hus
,
th
e
de
e
p
le
a
r
ni
ng
-
ba
s
e
d
f
a
c
ia
l
ve
r
if
ic
a
ti
on
m
ode
l
ha
s
a
c
hi
e
ve
d
s
uf
f
ic
ie
nt
a
nd
c
ont
r
ol
le
d
r
e
s
ul
ts
[
22]
.
I
n
ge
ne
r
a
l,
th
e
r
e
a
r
e
a
bout
30
-
40%
o
f
ge
ne
ti
c
di
s
or
de
r
s
a
s
s
oc
ia
te
d w
it
h c
e
r
ta
in
f
a
c
ia
l
c
ha
r
a
c
te
r
is
ti
c
s
c
a
ll
e
d dys
m
or
phi
c
f
e
a
tu
r
e
s
. I
n a
not
he
r
s
tu
dy by
a
na
ly
z
in
g
th
e
pe
r
f
or
m
a
nc
e
of
c
l
a
s
s
if
ie
r
s
b
a
s
e
d
on
de
e
p
le
a
r
ni
ng
f
a
c
ia
l
r
e
c
ogni
ti
on
m
ode
ls
th
r
ough
dys
m
or
phi
c
f
e
a
tu
r
e
de
te
c
ti
on
[
23]
.
T
he
f
a
c
e
de
te
c
ti
on
a
li
gnm
e
nt
m
e
th
ods
ut
il
iz
e
d
in
th
is
s
tu
dy
a
r
e
m
ul
ti
-
ta
s
k
c
a
s
c
a
d
e
d
c
onvolut
io
na
l
ne
twor
ks
(
M
T
C
N
N
)
a
nd
R
e
ti
na
F
a
c
e
.
T
he
s
e
m
e
th
ods
di
f
f
e
r
in
th
e
f
a
c
e
de
t
e
c
ti
on
s
ta
ge
,
w
it
h
M
T
C
N
N
ut
il
iz
in
g
a
th
r
e
e
-
s
ta
ge
f
a
c
e
de
te
c
ti
on
a
pp
r
oa
c
h,
w
hi
le
R
e
ti
na
F
a
c
e
a
dopt
s
a
s
in
gl
e
-
s
ta
ge
a
ppr
oa
c
h
e
m
pl
oyi
ng
a
s
in
gl
e
c
onvolut
io
na
l
ne
twor
k
[
24]
.
T
hi
s
di
s
s
im
il
a
r
it
y
ha
s
im
pl
ic
a
ti
ons
f
or
th
e
r
e
s
ul
ts
,
a
s
R
e
ti
na
F
a
c
e
de
m
ons
tr
a
te
s
s
upe
r
io
r
s
pe
e
d
a
nd
de
te
c
ti
on
c
a
p
a
bi
li
ti
e
s
,
pa
r
ti
c
ul
a
r
ly
in
lo
w
-
li
ght
c
ondi
ti
ons
,
c
om
pa
r
e
d
to
M
T
C
N
N
[
18]
.
A
ddi
ti
ona
ll
y,
R
e
ti
na
F
a
c
e
e
x
c
e
ls
i
n
de
te
c
ti
ng
f
a
c
e
s
in
va
r
io
us
pos
e
s
,
in
c
lu
di
ng
ti
lt
e
d
pos
e
s
,
a
nd
e
xhi
bi
ts
r
obus
t
p
e
r
f
or
m
a
nc
e
a
c
r
os
s
di
f
f
e
r
e
nt
s
c
a
le
s
[
25]
.
H
ow
e
v
e
r
,
it
is
not
e
w
or
th
y
th
a
t
M
T
C
N
N
a
ls
o
pe
r
f
or
m
s
a
dm
ir
a
bl
y,
of
f
e
r
in
g
m
or
e
a
c
c
ur
a
te
a
nd
s
ta
bl
e
f
a
c
e
de
te
c
ti
on
th
a
n
R
e
ti
na
F
a
c
e
[
19]
.
C
ons
e
que
nt
ly
,
th
is
r
e
s
e
a
r
c
h
in
c
or
por
a
te
s
M
T
C
N
N
a
nd
R
e
ti
na
F
a
c
e
f
or
f
a
c
e
de
te
c
ti
on
a
li
gnm
e
nt
to
de
te
r
m
in
e
th
e
m
e
th
od t
ha
t
a
c
hi
e
ve
s
hi
gh
a
c
c
ur
a
c
y w
hi
le
op
e
r
a
ti
ng e
f
f
ic
ie
nt
ly
a
nd s
w
if
tl
y.
3.
P
R
O
P
O
S
E
D
M
E
T
H
O
D
T
hi
s
r
e
s
e
a
r
c
h
T
hi
s
r
e
s
e
a
r
c
h
c
e
nt
e
r
s
a
r
ound
th
e
a
ppl
ic
a
ti
on
of
f
a
c
ia
l
da
ta
ha
il
in
g
f
r
om
a
d
a
ta
ba
s
e
e
nc
om
pa
s
s
in
g
50
s
tu
de
nt
s
.
T
he
a
s
s
e
m
bl
y
of
th
is
f
a
c
ia
l
da
ta
w
a
s
a
pi
vot
a
l
e
le
m
e
nt
of
th
e
r
e
s
e
a
r
c
h,
r
e
a
li
z
e
d
th
r
ough
th
e
us
e
of
di
ve
r
s
e
de
vi
c
e
s
th
r
oughout
th
e
s
tu
dy'
s
pr
ogr
e
s
s
io
n.
T
he
S
a
m
s
ung
G
a
la
xy
J
5
P
r
im
e
s
m
a
r
tp
hone
c
a
m
e
r
a
,
te
th
e
r
e
d
to
a
la
pt
op
vi
a
th
e
D
r
oi
dC
a
m
a
ppl
ic
a
ti
on,
f
unc
ti
one
d
a
s
th
e
pr
in
c
ip
a
l
in
s
tr
um
e
nt
f
or
da
ta
ga
th
e
r
in
g. T
he
e
xpe
r
im
e
nt
a
l
pr
oc
e
dur
e
w
a
s
bi
f
ur
c
a
te
d i
nt
o t
w
o di
s
ti
nc
t
s
e
gm
e
nt
s
.
T
he
i
ni
ti
a
l
s
e
gm
e
nt
e
nt
a
il
e
d
th
e
r
e
c
ogni
ti
on
of
30
pr
e
-
r
e
gi
s
te
r
e
d
f
a
c
e
s
.
S
ubs
e
qu
e
nt
ly
,
th
e
s
e
c
ond
pha
s
e
z
e
r
oe
d
in
on
th
e
de
te
c
ti
on
of
ni
ne
uni
que
f
a
c
e
s
,
e
a
c
h
e
xhi
bi
te
d
in
f
iv
e
di
ve
r
s
e
f
a
c
ia
l
pos
it
io
ns
.
T
he
s
ys
te
m
'
s
pe
r
f
or
m
a
nc
e
w
a
s
pr
im
a
r
il
y
e
va
lu
a
te
d
by
m
e
a
s
ur
in
g
th
e
f
a
ls
e
ne
ga
ti
ve
r
a
te
.
T
hi
s
m
e
tr
ic
r
e
f
le
c
ts
in
s
ta
nc
e
s
w
he
r
e
in
di
vi
dua
ls
w
e
r
e
c
or
r
e
c
tl
y
di
s
c
e
r
ne
d,
ye
t
th
e
qua
nt
if
ie
d
di
s
ta
nc
e
va
lu
e
s
ur
pa
s
s
e
d
th
e
e
s
ta
bl
is
he
d
m
in
im
um
th
r
e
s
hol
d.
T
hr
e
s
hol
d
is
de
r
iv
e
d
by
pi
npoi
nt
in
g
th
e
s
m
a
ll
e
s
t
di
s
ta
nc
e
b
e
twe
e
n
e
r
r
one
ous
ly
d
e
te
c
te
d
in
di
vi
dua
l
s
.
I
n
c
onduc
ti
ng
m
ul
ti
pl
e
e
xpe
r
im
e
nt
s
w
it
h
va
r
io
us
f
a
c
e
s
in
nu
m
e
r
ous
pos
it
io
ns
,
w
e
hop
e
to
f
a
c
il
it
a
te
a
c
om
pr
e
he
ns
iv
e
a
na
ly
s
i
s
of
th
e
s
ys
te
m
'
s
r
obus
tn
e
s
s
.
C
onc
u
r
r
e
nt
ly
,
th
e
s
ys
te
m
'
s
e
f
f
e
c
ti
ve
ne
s
s
w
il
l
be
r
e
in
f
or
c
e
d
ba
s
e
d
on
it
s
pr
of
ic
ie
nc
y
in
a
c
c
ur
a
te
ly
id
e
nt
if
yi
ng
in
di
vi
dua
ls
,
a
c
c
ount
in
g
f
or
va
r
ia
ti
ons
in
di
s
ti
nc
t
f
a
c
ia
l
pos
e
s
.
A
n
a
ddi
ti
ona
l
a
im
i
s
to
ve
r
if
y
th
e
r
e
li
a
bi
li
ty
of
th
e
di
s
ta
nc
e
a
c
c
ur
a
c
y
m
e
tr
ic
s
e
m
pl
oye
d
th
r
oughout
th
e
e
va
lu
a
ti
on
pr
oc
e
s
s
.
T
he
r
e
f
or
e
,
to
bol
s
te
r
a
c
c
ur
a
c
y
pe
r
f
or
m
a
nc
e
,
w
e
im
pl
e
m
e
nt
s
e
ve
r
a
l
s
im
il
a
r
it
y
m
e
th
ods
a
nd
a
c
onvolut
io
na
l
ne
ur
a
l
ne
twor
k
(
C
N
N
)
m
ode
l
s
ys
te
m
.
T
h
e
s
e
to
ol
s
a
r
e
goi
ng
to
a
id
in
a
tt
a
in
in
g
a
c
le
a
r
unde
r
s
ta
ndi
ng
of
opt
im
a
l
pe
r
f
o
r
m
a
nc
e
in
th
e
de
te
c
ti
on
of
hum
a
n
f
a
c
e
s
.
F
or
th
is
r
e
a
s
on,
a
n
ove
r
vi
e
w
of
t
he
s
ta
ge
s
of
t
he
t
r
a
in
in
g da
ta
r
e
tr
ie
va
l
pr
oc
e
s
s
i
n t
he
l
oc
ke
r
l
e
ndi
ng s
ys
te
m
i
s
pr
ovi
de
d.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
E
nhanc
in
g t
ouc
hl
e
s
s
s
m
a
r
t
lo
c
k
e
r
s
y
s
te
m
s
t
hr
ough adv
anc
e
d f
ac
ia
l
r
e
c
ogni
ti
on
…
(
A
bdul
H
ar
is
R
angk
ut
i
)
3265
F
ig
ur
e
1
pr
ovi
de
s
a
de
pi
c
ti
on
of
th
e
ope
r
a
ti
ona
l
m
e
c
ha
ni
s
m
of
th
e
lo
c
ke
r
le
ndi
ng
s
ys
te
m
.
I
t
de
m
ons
tr
a
te
s
th
e
a
s
yn
c
hr
onous
s
yn
c
hr
oni
z
a
ti
on
a
nd
tr
a
in
in
g
pr
oc
e
s
s
of
s
e
ve
r
a
l
m
ode
ls
d
e
di
c
a
te
d
to
th
e
de
te
c
ti
on
a
nd
r
e
c
ogni
ti
on
of
f
a
c
ia
l
im
a
g
e
s
s
to
r
e
d
in
th
e
da
ta
ba
s
e
.
T
he
s
ys
te
m
onl
y
in
it
ia
te
s
da
ta
s
ync
hr
oni
z
a
ti
on
w
he
n
m
odi
f
ic
a
ti
ons
a
r
e
m
a
de
a
t
th
e
ti
m
e
of
t
he
la
s
t
s
ync
hr
oni
z
a
ti
on
of
th
e
f
a
c
ia
l
da
ta
ba
s
e
.
C
ha
nge
s
to
th
e
s
ync
hr
oni
z
a
ti
on
ti
m
e
c
a
n
tr
a
ns
pi
r
e
due
to
v
a
r
io
us
r
e
a
s
on
s
,
s
uc
h
a
s
th
e
a
ddi
ti
on
or
in
put
of
a
ne
w
us
e
r
'
s
f
a
c
e
,
th
e
pr
oc
e
s
s
of
de
le
ti
ng
a
us
e
r
,
or
a
lt
e
r
a
ti
ons
to
a
u
s
e
r
'
s
im
a
ge
th
a
t
i
s
c
on
s
id
e
r
e
d
unc
le
a
r
.
I
f
a
m
odi
f
ic
a
ti
on
is
e
xe
c
ut
e
d
a
t
th
e
ti
m
e
of
th
e
la
s
t
s
ync
hr
oni
z
a
t
io
n,
th
e
s
ys
t
e
m
w
il
l
e
xt
r
a
c
t
phot
os
f
r
om
th
e
da
ta
ba
s
e
a
nd
tr
a
in
th
e
m
od
e
l
us
in
g
th
e
s
e
im
a
g
e
s
.
G
e
ne
r
a
ll
y,
th
e
L
ig
ht
F
a
c
e
f
r
a
m
e
w
or
k
c
onduc
ts
th
e
tr
a
in
in
g
pr
oc
e
s
s
in
de
p
e
nde
nt
ly
a
nd
p
e
r
f
or
m
s
it
e
xc
lu
s
iv
e
ly
dur
in
g
th
e
s
ync
hr
oni
z
a
ti
on
pr
oc
e
s
s
.
F
ol
lo
w
in
g
th
e
m
ode
l'
s
f
or
m
a
ti
on
th
r
ough
th
is
tr
a
in
in
g
pr
oc
e
s
s
,
th
e
s
ys
te
m
w
il
l
in
it
ia
te
a
tr
a
ns
it
io
n
pr
oc
e
s
s
to
th
e
pr
im
a
r
y
di
s
pl
a
y
m
e
nu. I
n t
he
f
a
c
e
de
te
c
ti
on s
ys
te
m
c
a
r
r
ie
d out by t
he
l
oc
ke
r
, t
h
e
s
ys
te
m
w
il
l
c
ont
in
ue
t
o c
a
r
r
y
out
t
he
pr
oc
e
s
s
opt
im
a
ll
y a
s
t
he
numbe
r
of
us
e
r
s
i
nvol
ve
d i
n s
to
r
in
g goods
or
s
e
c
ur
it
ie
s
i
n t
he
l
oc
ke
r
i
nc
r
e
a
s
e
s
.
F
ig
ur
e
1. T
he
s
ync
hr
oni
z
a
ti
on me
c
h
a
ni
s
m
of
t
he
l
oc
ke
r
bor
r
ow
in
g s
ys
te
m
F
ig
ur
e
2
vi
s
ua
ll
y
de
pi
c
ts
th
e
ope
r
a
ti
ona
l
in
te
r
f
a
c
e
of
th
e
Y
ou
r
va
ul
t
lo
c
ke
r
r
e
nt
a
l
a
ppl
ic
a
ti
on.
T
he
s
ys
te
m
is
in
g
e
ni
ous
ly
e
ngi
ne
e
r
e
d
to
a
ut
om
a
ti
c
a
ll
y
s
yn
c
hr
oni
z
e
da
ta
w
he
ne
v
e
r
m
odi
f
ic
a
ti
ons
oc
c
ur
,
a
lwa
y
s
pr
ovi
di
ng us
e
r
s
w
it
h t
he
m
os
t
c
ur
r
e
nt
i
nf
or
m
a
ti
on. T
he
a
ppl
ic
a
ti
on'
s
na
vi
ga
ti
on i
s
s
e
a
m
le
s
s
ly
f
a
c
il
it
a
te
d us
in
g
ha
nd
de
te
c
ti
on
te
c
hnol
ogy,
of
f
e
r
in
g
a
us
e
r
-
f
r
ie
ndl
y
e
xpe
r
ie
nc
e
.
A
s
pa
r
t
of
th
e
pr
oc
e
s
s
,
u
s
e
r
s
c
a
n
a
c
c
e
s
s
di
f
f
e
r
e
nt
m
e
nu
opt
io
ns
by
s
im
pl
y
r
a
is
in
g
th
e
ir
ha
nd.
W
he
n
a
t
r
a
ns
a
c
ti
on
is
in
it
ia
te
d
a
t
a
Y
our
va
ul
t
ki
os
k,
a
f
a
c
ia
l
ve
r
if
ic
a
ti
on
pr
oc
e
s
s
is
put
in
to
a
c
ti
on
to
va
li
da
te
th
e
us
e
r
'
s
id
e
nt
it
y.
P
r
ovi
de
d
th
e
us
e
r
'
s
f
a
c
e
is
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
,
V
ol
. 14, No. 4, A
ugus
t
2025
:
3262
-
3273
3266
a
c
c
ur
a
te
ly
r
e
c
ogni
z
e
d
a
nd
th
e
y
pos
s
e
s
s
a
de
qua
t
e
f
unds
f
or
lo
c
ke
r
r
e
nt
a
l,
th
e
s
ys
te
m
pr
oc
e
e
ds
to
r
e
ve
a
l
th
e
lo
c
ke
r
th
a
t
is
e
it
he
r
c
ur
r
e
nt
ly
bor
r
ow
e
d
or
in
us
e
.
A
f
te
r
th
is
,
th
e
s
ys
te
m
unl
oc
ks
th
e
de
s
ig
na
t
e
d
lo
c
ke
r
w
hi
le
a
ppr
opr
ia
te
ly
de
duc
ti
ng
th
e
ne
c
e
s
s
a
r
y
f
unds
f
r
om
th
e
us
e
r
'
s
a
c
c
ount
.
H
ow
e
v
e
r
,
s
houl
d
th
e
s
ys
te
m
f
a
il
to
c
onf
ir
m
th
e
us
e
r
'
s
id
e
nt
it
y
dur
in
g
th
e
f
a
c
ia
l
ve
r
if
ic
a
ti
on
s
ta
g
e
,
th
e
tr
a
ns
a
c
ti
on
is
in
s
ta
nt
ly
a
nnul
le
d.
T
he
in
te
r
f
a
c
e
th
e
n
r
e
ve
r
ts
to
it
s
m
a
in
m
e
nu.
T
hi
s
pr
oc
e
dur
e
is
im
pl
e
m
e
nt
e
d
to
bol
s
te
r
tr
a
ns
a
c
ti
ona
l
s
e
c
ur
it
y
a
nd
pr
e
ve
nt
una
ut
hor
iz
e
d l
oc
ke
r
a
c
c
e
s
s
.
F
ig
ur
e
3
de
pi
c
ts
th
e
f
unc
ti
ona
li
ty
of
a
na
vi
ga
ti
on
s
ys
te
m
e
m
pl
oyi
ng
ha
nd
de
te
c
ti
on
te
c
hnol
ogy
f
or
f
a
c
ia
l
r
e
c
ogni
ti
on.
W
it
hi
n
th
e
m
a
in
m
e
nu,
us
e
r
s
a
r
e
pr
om
pt
e
d
to
r
a
is
e
e
it
he
r
th
e
ir
r
ig
ht
or
le
f
t
ha
nd,
w
it
h
di
s
ti
nc
t
f
unc
ti
ons
a
tt
r
ib
ut
e
d
to
e
a
c
h
ha
nd.
S
pe
c
if
ic
a
ll
y,
th
e
le
f
t
ha
nd
s
ig
ni
f
ie
s
th
e
pr
oc
e
s
s
of
bor
r
ow
in
g
a
lo
c
ke
r
,
w
hi
le
th
e
r
ig
ht
ha
nd
de
not
e
s
a
c
c
e
s
s
in
g
th
e
lo
c
ke
r
.
U
po
n
r
e
a
c
hi
ng
th
e
lo
c
k
e
r
ope
ni
ng
m
e
nu,
u
s
e
r
s
a
r
e
pr
e
s
e
nt
e
d
w
it
h
two
c
le
a
r
opt
io
ns
:
to
e
it
he
r
pr
oc
e
e
d
w
it
h
th
e
lo
c
ke
r
r
e
nt
a
l
pr
oc
e
s
s
or
to
te
r
m
in
a
te
th
e
lo
c
ke
r
r
e
nt
a
l.
U
s
e
r
s
c
a
n
m
a
ke
th
e
ir
c
hoi
c
e
by
a
li
gni
ng
th
e
ir
ha
nd
s
e
le
c
ti
on
(
r
ig
ht
or
le
f
t)
w
it
h
th
e
c
or
r
e
s
ponding
in
s
tr
uc
ti
ons
di
s
pl
a
ye
d
in
th
e
a
ppl
ic
a
ti
on.
T
hi
s
us
e
r
-
f
r
ie
ndl
y
opt
io
n
s
ys
te
m
s
im
pl
if
ie
s
th
e
de
c
is
io
n
-
m
a
ki
ng
pr
oc
e
s
s
f
or
us
e
r
s
,
w
he
th
e
r
th
e
y
w
is
h
to
c
ont
in
ue
r
e
nt
in
g
a
lo
c
ke
r
or
c
e
a
s
e
th
e
r
e
nt
a
l
pr
oc
e
s
s
.
A
vi
s
ua
l
r
e
pr
e
s
e
nt
a
ti
on of
t
hi
s
pr
oc
e
s
s
i
s
pr
ovi
de
d i
n t
he
a
c
c
om
pa
nyi
ng
c
ha
r
t.
F
ig
ur
e
2. T
he
ope
r
a
ti
ona
l
in
te
r
f
a
c
e
of
t
he
Y
our
va
ul
t
boot
h a
ppl
ic
a
ti
on
F
ig
ur
e
3. T
he
f
unc
ti
ona
li
ty
of
t
he
na
vi
ga
ti
on s
ys
te
m
ut
il
iz
in
g ha
nd de
te
c
ti
on
F
ig
ur
e
4
s
how
c
a
s
e
s
th
e
pr
e
-
f
a
c
e
r
e
c
ogni
ti
on
pr
oc
e
s
s
,
w
hi
c
h
is
m
e
ti
c
ul
ous
ly
de
s
ig
ne
d
to
e
ns
ur
e
th
a
t
onl
y
one
f
a
c
e
is
de
te
c
te
d
w
it
hi
n
th
e
c
a
m
e
r
a
f
r
a
m
e
.
F
o
ll
ow
in
g
th
is
,
th
e
s
ys
te
m
pr
oc
e
e
ds
to
ve
r
if
y
th
a
t
th
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
E
nhanc
in
g t
ouc
hl
e
s
s
s
m
a
r
t
lo
c
k
e
r
s
y
s
te
m
s
t
hr
ough adv
anc
e
d f
ac
ia
l
r
e
c
ogni
ti
on
…
(
A
bdul
H
ar
is
R
angk
ut
i
)
3267
us
e
r
'
s
f
a
c
e
i
s
no
t
ob
s
tr
uc
te
d
by
a
ny
f
or
m
of
c
ove
r
in
g
or
m
a
s
k
.
T
hi
s
c
r
uc
ia
l
s
te
p
s
e
r
ve
s
to
e
nh
a
nc
e
s
e
c
ur
it
y
m
e
a
s
ur
e
s
,
e
f
f
e
c
ti
ve
ly
pr
e
ve
nt
in
g
th
e
us
e
of
s
poof
in
g
m
e
th
ods
.
F
ur
th
e
r
m
or
e
,
th
e
p
r
oc
e
s
s
in
c
lu
de
s
a
va
li
da
ti
on
m
e
c
ha
ni
s
m
to
a
ut
he
nt
ic
a
te
th
e
ge
nui
ne
ne
s
s
of
th
e
de
te
c
te
d
f
a
c
e
,
di
s
ti
ngui
s
hi
ng
it
f
r
om
pr
in
te
d
or
di
s
pl
a
ye
d
im
a
ge
s
.
T
hi
s
s
tr
in
ge
nt
va
li
da
ti
on
e
ns
ur
e
s
th
a
t
th
e
f
a
c
ia
l
im
a
ge
c
a
pt
ur
e
d
by
th
e
c
a
m
e
r
a
is
a
ut
he
nt
ic
a
nd
not
a
r
e
pr
oduc
ti
on
f
r
o
m
a
di
f
f
e
r
e
nt
m
e
di
um
.
I
n
th
e
f
in
a
l
s
ta
ge
,
th
e
s
ys
te
m
va
li
da
te
s
th
e
us
e
r
'
s
f
a
c
e
,
e
n
s
ur
in
g
it
is
a
c
c
ur
a
te
ly
de
te
c
te
d w
it
hi
n a
s
pe
c
if
ie
d di
s
ta
nc
e
t
hr
e
s
hol
d. U
pon
m
e
e
ti
ng a
ll
t
he
s
e
s
tr
in
ge
nt
c
r
it
e
r
ia
, t
he
s
ys
te
m
s
uc
c
e
s
s
f
ul
ly
de
te
c
ts
th
e
us
e
r
'
s
f
a
c
e
a
nd
pr
oc
e
e
d
s
to
r
e
tu
r
n
t
he
us
e
r
I
D
f
or
f
u
r
th
e
r
pr
oc
e
s
s
in
g,
e
na
bl
in
g
r
e
c
ogni
ti
on of
t
he
t
e
na
nt
'
s
f
a
c
e
s
to
r
e
d w
it
hi
n t
he
us
e
r
f
a
c
e
d
a
ta
b
a
s
e
.
F
ig
ur
e
5
pr
ovi
de
s
a
n
ove
r
vi
e
w
of
th
e
ope
r
a
ti
on
of
th
e
s
m
a
r
t
l
oc
ke
r
s
ys
te
m
,
de
s
ig
n
e
d
f
or
us
e
r
a
nd
c
us
to
m
e
r
r
e
nt
a
l
us
e
. T
ypi
c
a
ll
y, s
m
a
r
t
lo
c
ke
r
s
di
s
pl
a
y a
gr
e
e
n l
ig
ht
w
he
n a
va
il
a
bl
e
a
nd a
r
e
d l
ig
ht
w
he
n i
n us
e
.
H
ow
e
ve
r
,
if
th
e
lo
c
ke
r
li
ght
is
of
f
,
it
s
ig
ni
f
ie
s
th
a
t
th
e
lo
c
ke
r
i
s
unl
oc
ke
d
a
nd
ha
s
be
e
n
ope
n
e
d
by
th
e
us
e
r
to
de
pos
it
or
r
e
tr
ie
ve
it
e
m
s
.
S
ubs
e
que
nt
ly
,
th
e
li
ght
s
w
il
l
r
e
-
il
lu
m
in
a
te
onc
e
th
e
lo
c
ke
r
i
s
c
lo
s
e
d,
a
ut
om
a
ti
c
a
ll
y
s
e
c
ur
in
g
th
e
lo
c
ke
r
.
T
hi
s
s
e
a
m
le
s
s
pr
oc
e
s
s
of
ope
ni
ng
a
nd
c
lo
s
in
g
th
e
lo
c
k
e
r
f
ur
ni
s
he
s
us
e
r
s
w
it
h
pe
r
ti
ne
nt
in
f
or
m
a
ti
on
if
th
e
y
w
is
h
to
a
va
il
th
e
m
s
e
lv
e
s
of
a
s
m
a
r
t
lo
c
ke
r
.
T
he
in
te
ll
ig
e
nt
lo
c
ke
r
s
ys
te
m
a
de
pt
ly
r
e
c
ogni
z
e
s
th
e
f
a
c
e
s
of
it
s
te
na
nt
s
w
it
h
pr
e
c
is
io
n,
s
w
if
tl
y
r
e
la
yi
ng
in
f
or
m
a
ti
on
to
th
e
lo
c
ke
r
s
ys
te
m
to
f
a
c
il
it
a
te
th
e
ope
ni
ng
a
nd
c
lo
s
in
g
pr
oc
e
s
s
e
s
.
A
n
ove
r
vi
e
w
of
th
e
da
ta
s
e
t
us
e
d
in
th
e
e
xpe
r
im
e
nt
c
a
n
be
f
ound
a
t:
ht
tp
s
:/
/ww
w
.ka
ggl
e
.c
om
/d
a
ta
s
e
ts
/y
oha
n
e
s
07/
your
va
ul
t
-
f
a
c
e
-
r
e
c
ogni
ti
on
-
da
ta
s
e
ts
.
I
n
e
s
s
e
nc
e
,
th
is
r
e
s
e
a
r
c
h
pr
e
s
e
nt
s
a
n i
nnova
ti
ve
a
ppr
oa
c
h t
o ut
il
iz
in
g l
oc
ke
r
s
ba
s
e
d on f
a
c
ia
l
r
e
c
ogni
ti
on a
nd a
s
s
o
c
ia
te
d pr
ot
oc
ol
s
.
F
ig
ur
e
4. T
he
pr
e
-
f
a
c
ia
l
r
e
c
ogni
ti
on pr
oc
e
s
s
F
ig
ur
e
5. T
he
ope
r
a
ti
on of
t
he
Y
our
va
ul
t
lo
c
ke
r
s
ys
te
m
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
,
V
ol
. 14, No. 4, A
ugus
t
2025
:
3262
-
3273
3268
4.
R
E
S
U
L
T
S
A
N
D
D
I
S
C
U
S
S
I
O
N
T
he
r
e
s
e
a
r
c
h
e
nc
om
pa
s
s
e
d
two
di
s
ti
nc
t
e
xpe
r
im
e
nt
s
a
im
e
d
a
t
id
e
nt
if
yi
ng
th
e
m
e
th
od
w
it
h
th
e
hi
ghe
s
t
a
nd
m
os
t
c
ons
is
t
e
nt
a
c
c
ur
a
c
y.
T
he
f
ir
s
t
e
xpe
r
im
e
nt
e
nt
a
il
e
d
th
e
r
e
c
ogni
ti
on
of
30
out
o
f
50
pr
e
-
r
e
gi
s
te
r
e
d
f
a
c
e
s
,
a
ll
po
s
it
io
ne
d
f
a
c
in
g
th
e
f
r
ont
.
S
ubs
e
que
nt
ly
,
th
e
s
e
c
ond
e
xpe
r
im
e
nt
f
oc
us
e
d
on
di
s
c
e
r
ni
ng
th
e
f
a
c
ia
l
f
e
a
tu
r
e
s
of
9
in
di
vi
dua
ls
a
c
r
os
s
f
iv
e
di
f
f
e
r
e
nt
f
a
c
ia
l
or
ie
nt
a
ti
ons
:
f
r
ont
,
to
p,
bot
to
m
,
a
nd
s
la
nt
in
g
le
f
t
a
nd
r
ig
ht
a
t
45
de
gr
e
e
s
.
F
ol
lo
w
in
g
th
e
c
om
pl
e
ti
on
of
bot
h
e
xpe
r
im
e
nt
s
,
a
th
or
ough
e
va
lu
a
ti
on
pr
oc
e
s
s
w
a
s
in
it
ia
te
d,
dur
in
g
w
hi
c
h
th
e
to
p
5
m
e
th
ods
,
di
s
t
in
gui
s
he
d
by
th
e
hi
ghe
s
t
num
be
r
of
c
or
r
e
c
t
r
e
c
ogni
ti
ons
,
unde
r
w
e
nt
a
m
e
ti
c
ul
ou
s
th
r
e
s
hol
di
ng
e
li
m
in
a
ti
on
pr
oc
e
dur
e
.
T
hi
s
pr
oc
e
s
s
m
e
ti
c
ul
ous
ly
de
f
in
e
d
th
e
pa
r
a
m
e
te
r
s
f
or
di
f
f
e
r
e
nt
ia
ti
ng
be
twe
e
n
tr
ue
pos
it
iv
e
a
nd
f
a
l
s
e
po
s
it
iv
e
de
te
c
ti
ons
,
e
f
f
e
c
ti
ve
ly
e
li
m
in
a
ti
ng
a
ny
in
s
ta
nc
e
s
of
f
a
ls
e
po
s
it
iv
e
s
.
C
on
s
e
que
nt
ly
,
a
c
om
pr
e
he
ns
iv
e
a
na
ly
s
is
w
a
s
c
onduc
te
d
to
de
te
r
m
in
e
th
e
f
a
ls
e
ne
ga
ti
ve
r
a
te
a
nd
a
c
c
ur
a
c
y
m
e
tr
ic
s
f
or
t
he
s
e
a
f
or
e
m
e
nt
io
ne
d t
op 5 me
th
ods
.
B
a
s
e
d
on
th
e
f
in
di
ngs
pr
e
s
e
nt
e
d
in
T
a
bl
e
2,
th
e
e
xp
e
r
im
e
nt
'
s
pr
im
a
r
y
a
im
w
a
s
to
r
e
c
ogni
z
e
30
out
of
50
r
e
gi
s
te
r
e
d
f
a
c
e
s
w
it
h
a
f
r
ont
a
l
or
ie
nt
a
ti
on,
ut
il
iz
in
g
th
e
M
T
C
N
N
a
s
th
e
f
a
c
e
d
e
te
c
ti
on
a
lg
or
it
hm
.
T
he
r
e
s
ul
ts
in
di
c
a
te
th
a
t
th
e
A
r
c
F
a
c
e
m
ode
l
d
e
m
ons
tr
a
te
s
opt
im
a
l
pe
r
f
or
m
a
nc
e
,
pa
r
ti
c
ul
a
r
ly
w
he
n
e
m
pl
oyi
ng
c
os
in
e
a
nd
E
uc
li
de
a
n
L
2
nor
m
,
a
c
hi
e
vi
ng
a
n
a
c
c
ur
a
c
y
of
m
or
e
th
a
n
96%
.
C
onve
r
s
e
ly
,
w
he
n
ut
il
iz
in
g
F
a
c
e
N
e
t5
12
pe
r
f
or
m
a
nc
e
,
th
e
m
os
t
e
f
f
e
c
ti
ve
pe
r
f
or
m
a
nc
e
is
a
c
hi
e
ve
d
w
it
h
E
uc
li
de
a
n
a
nd
E
uc
li
de
a
n
L
2
nor
m
s
,
yi
e
ld
in
g
a
n
a
c
c
ur
a
c
y
of
70%
.
N
ot
a
bl
y,
th
e
pe
r
f
o
r
m
a
nc
e
of
V
G
G
-
F
a
c
e
is
in
f
e
r
io
r
to
bot
h
A
r
c
F
a
c
e
a
nd
F
a
c
e
N
e
t5
12
.
F
ur
th
e
r
m
or
e
,
it
is
e
vi
de
nt
th
a
t
th
e
E
uc
li
de
a
n
L
2
nor
m
c
ons
is
te
nt
ly
de
li
ve
r
s
a
m
or
e
s
ta
bl
e
pe
r
f
or
m
a
nc
e
c
om
pa
r
e
d t
o both
c
os
in
e
a
nd E
uc
li
de
a
n L
2
nor
m
s
w
he
n e
va
lu
a
ti
ng t
he
i
de
nt
if
ie
d f
a
c
ia
l
f
e
a
tu
r
e
s
.
T
a
bl
e
2.
F
a
c
ia
l
r
e
c
ogni
ti
on e
xp
e
r
im
e
nt
s
f
or
30 pe
opl
e
us
in
g M
T
C
N
N
M
ode
l
D
i
s
t
a
nc
e
m
e
t
r
i
c
P
e
r
c
e
nt
a
ge
C
or
r
e
c
t
(
%
)
W
r
ong (
%
)
I
nva
l
i
d (
%
)
A
r
c
F
a
c
e
C
os
i
ne
96.7
3.3
0
E
uc
l
i
de
a
n
63.3
6.7
30
E
uc
l
i
de
a
n L
2
96.7
3.3
0
F
a
c
e
N
e
t
512
C
os
i
ne
66.7
13.3
20
E
uc
l
i
de
a
n
70
30
0
E
uc
l
i
de
a
n L
2
70
30
0
VGG
-
F
a
c
e
C
os
i
ne
56.7
43.3
0
E
uc
l
i
de
a
n
53.3
46.7
0
E
uc
l
i
de
a
n L
2
56.7
43.3
0
E
xpe
r
im
e
nt
f
in
d
in
gs
ba
s
e
d
on
T
a
bl
e
3
a
im
e
d
to
r
e
c
ogni
z
e
30
f
a
c
e
s
out
of
50
r
e
gi
s
te
r
e
d
f
a
c
e
s
w
it
h
a
f
r
ont
a
l
or
ie
nt
a
ti
on,
e
m
pl
oyi
ng
R
e
ti
na
F
a
c
e
a
s
th
e
f
a
c
e
de
te
c
ti
on
a
lg
or
it
hm
.
T
he
ta
bl
e
il
lu
s
tr
a
te
s
th
a
t
th
e
A
r
c
F
a
c
e
m
ode
l
yi
e
ld
s
opt
im
a
l
pe
r
f
or
m
a
nc
e
w
he
n
ut
il
iz
in
g
bot
h
c
os
in
e
a
nd
E
uc
li
de
a
n
L
2
nor
m
,
a
c
hi
e
vi
ng
a
n
opt
im
a
l
a
c
c
ur
a
c
y
of
90%
,
a
lb
e
it
w
it
h
f
e
w
e
r
c
or
r
e
c
t
r
e
c
ogni
ti
o
ns
th
a
n
w
he
n
M
T
C
N
N
is
u
s
e
d.
I
n
th
e
c
a
s
e
of
F
a
c
e
N
e
t5
12
pe
r
f
or
m
a
nc
e
,
th
e
be
s
t
r
e
s
ul
t
is
a
c
hi
e
ve
d
w
it
h
E
uc
li
de
a
n
a
nd
E
uc
li
de
a
n
L
2
no
r
m
,
w
it
h
a
n
a
c
c
ur
a
c
y of
m
or
e
t
ha
n 76%
, a
c
c
om
pa
ni
e
d by a
hi
ghe
r
numbe
r
o
f
c
or
r
e
c
t
r
e
c
ogni
ti
ons
t
ha
n M
T
C
N
N
.
N
ot
a
bl
y,
VGG
-
F
a
c
e
e
xhi
bi
ts
in
f
e
r
io
r
pe
r
f
or
m
a
nc
e
c
om
pa
r
e
d
to
bot
h
A
r
c
F
a
c
e
a
nd
F
a
c
e
N
e
t5
12
.
S
ta
bi
li
ty
of
E
uc
li
de
a
n
L
2
nor
m
w
it
h
R
e
ti
na
F
a
c
e
,
F
ur
th
e
r
m
or
e
,
it
is
e
vi
de
nt
th
a
t
th
e
E
uc
li
de
a
n
L
2
N
or
m
w
it
h
R
e
ti
na
F
a
c
e
c
ons
is
te
nt
ly
pr
ovi
de
s
a
m
or
e
s
ta
bl
e
p
e
r
f
or
m
a
nc
e
c
om
pa
r
e
d
t
o
bot
h
c
os
in
e
a
nd
E
uc
li
de
a
n
L
2
nor
m
s
w
he
n
a
s
s
e
s
s
in
g t
he
i
de
nt
if
ie
d f
a
c
ia
l
f
e
a
tu
r
e
s
.
T
a
bl
e
3.
T
he
r
e
s
ul
t
of
f
a
c
ia
l
r
e
c
ogni
ti
on e
xpe
r
im
e
nt
s
f
or
30 pe
o
pl
e
us
in
g R
e
ti
na
F
a
c
e
M
ode
l
D
i
s
t
a
nc
e
m
e
t
r
i
c
P
e
r
c
e
nt
a
ge
C
or
r
e
c
t
(
%
)
W
r
ong (
%
)
I
nva
l
i
d (
%
)
A
r
c
F
a
c
e
C
os
i
ne
90
10
0
E
uc
l
i
de
a
n
60
3.3
36.7
E
uc
l
i
de
a
n L
2
90
10
0
F
a
c
e
N
e
t
512
C
os
i
ne
73.3
16.7
10
E
uc
l
i
de
a
n
76.7
23.3
0
E
uc
l
i
de
a
n L
2
76.7
23.3
0
VGG
-
F
a
c
e
C
os
i
ne
56.7
43.3
0
E
uc
l
i
de
a
n
56.7
43.3
0
E
uc
l
i
de
a
n L
2
56.7
43.3
0
B
a
s
e
d
on
th
e
da
ta
pr
e
s
e
nt
e
d
in
T
a
bl
e
4,
th
e
e
xpe
r
im
e
nt
f
oc
us
e
d
on
r
e
c
ogni
z
in
g
th
e
f
a
c
e
s
of
9
in
di
vi
dua
ls
a
c
r
os
s
f
iv
e
di
f
f
e
r
e
nt
f
a
c
ia
l
pos
it
io
ns
,
ut
il
iz
in
g
M
T
C
N
N
a
s
th
e
f
a
c
e
de
te
c
ti
on
a
lg
or
it
hm
.
T
he
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
E
nhanc
in
g t
ouc
hl
e
s
s
s
m
a
r
t
lo
c
k
e
r
s
y
s
te
m
s
t
hr
ough adv
anc
e
d f
ac
ia
l
r
e
c
ogni
ti
on
…
(
A
bdul
H
ar
is
R
angk
ut
i
)
3269
r
e
s
ul
ts
in
di
c
a
te
th
a
t
th
e
A
r
c
F
a
c
e
m
ode
l
e
xhi
bi
ts
s
upe
r
io
r
pe
r
f
or
m
a
nc
e
w
he
n
e
m
pl
oyi
ng
bot
h
c
os
in
e
a
nd
E
uc
li
de
a
n
L
2
n
or
m
w
it
h
opt
im
a
l
a
c
c
ur
a
c
y
of
m
or
e
th
a
n
73%
.
C
onve
r
s
e
ly
,
f
or
F
a
c
e
N
e
t5
12
,
th
e
be
s
t
pe
r
f
or
m
a
nc
e
is
a
c
hi
e
ve
d
w
it
h
E
uc
li
de
a
n,
w
hi
le
th
e
le
a
s
t
e
f
f
e
c
t
iv
e
pe
r
f
or
m
a
nc
e
,
r
e
la
ti
ve
to
ot
h
e
r
m
e
th
ods
,
i
s
obs
e
r
ve
d
w
he
n
us
in
g
c
os
in
e
a
nd
E
uc
li
de
a
n
L
2
nor
m
.
S
ur
pr
is
in
gl
y,
in
th
is
e
xpe
r
im
e
nt
,
V
G
G
-
F
a
c
e
de
m
ons
tr
a
te
s
a
pe
r
f
or
m
a
nc
e
le
ve
l
e
qui
va
le
nt
to
F
a
c
e
N
e
t5
12
.
T
he
s
e
f
in
di
ngs
unde
r
s
c
or
e
th
e
s
upe
r
io
r
pe
r
f
or
m
a
nc
e
of
A
r
c
F
a
c
e
,
pa
r
ti
c
ul
a
r
ly
w
he
n
e
m
pl
oyi
ng
c
os
in
e
a
nd
E
uc
li
de
a
n
L
2
nor
m
in
th
e
c
ont
e
xt
of
r
e
c
ogni
z
in
g
f
a
c
e
s
a
c
r
os
s
va
r
ie
d
pos
it
io
ns
,
w
hi
le
a
ls
o
hi
ghl
ig
ht
in
g
th
e
c
om
pe
ti
ti
ve
pe
r
f
or
m
a
nc
e
of
V
G
G
-
F
a
c
e
c
om
pa
r
e
d t
o
F
a
c
e
N
e
t5
12
.
T
a
bl
e
4.
T
he
r
e
s
ul
t
of
f
a
c
ia
l
r
e
c
ogni
ti
on e
xpe
r
im
e
nt
s
f
or
5 di
f
f
e
r
e
nt
pos
e
us
in
g M
T
C
N
N
M
ode
l
D
i
s
t
a
nc
e
m
e
t
r
i
c
P
e
r
c
e
nt
a
ge
C
or
r
e
c
t
(
%
)
W
r
ong (
%
)
I
nva
l
i
d (
%
)
A
r
c
F
a
c
e
C
os
i
ne
73.3
20
6.7
E
uc
l
i
de
a
n
33.3
22.2
44.4
E
uc
l
i
de
a
n L
2
73.3
17.8
8.9
F
a
c
e
N
e
t
512
C
os
i
ne
40
26.7
33.3
E
uc
l
i
de
a
n
57.8
42.2
0
E
uc
l
i
de
a
n L
2
48.9
48.9
2.2
VGG
-
F
a
c
e
C
os
i
ne
57.8
37.8
4.4
E
uc
l
i
de
a
n
55.6
44.4
0
E
uc
l
i
de
a
n L
2
57.8
37.8
4.4
B
a
s
e
d
on
th
e
da
ta
pr
e
s
e
nt
e
d
in
T
a
bl
e
5,
th
e
e
xpe
r
im
e
nt
a
im
e
d
to
r
e
c
ogni
z
e
th
e
f
a
c
e
s
of
9
in
di
vi
dua
ls
a
c
r
os
s
f
iv
e
di
f
f
e
r
e
nt
f
a
c
ia
l
po
s
it
io
ns
,
ut
il
iz
in
g
R
e
ti
na
F
a
c
e
a
s
th
e
f
a
c
e
d
e
te
c
ti
on a
lg
or
it
hm
.
T
he
r
e
s
ul
ts
in
di
c
a
te
th
a
t
th
e
A
r
c
F
a
c
e
m
ode
l
de
m
ons
tr
a
te
s
opt
im
a
l
pe
r
f
or
m
a
nc
e
w
he
n
ut
il
iz
in
g
bot
h
c
os
in
e
a
nd
E
uc
li
de
a
n
L
2
nor
m
w
it
h
a
c
c
ur
a
c
y
m
or
e
th
a
n
66%
,
a
lb
e
it
w
it
h
f
e
w
e
r
c
or
r
e
c
t
r
e
c
ogni
ti
ons
th
a
n
w
h
e
n
M
T
C
N
N
i
s
us
e
d.
F
or
F
a
c
e
N
e
t5
12
,
th
e
be
s
t
pe
r
f
or
m
a
nc
e
is
a
c
hi
e
v
e
d
w
it
h
E
uc
li
de
a
n L
2
N
or
m
,
a
c
c
om
pa
ni
e
d
by
a
hi
ghe
r
num
be
r
of
c
or
r
e
c
t
r
e
c
ogni
ti
ons
th
a
n
M
T
C
N
N
.
H
ow
e
v
e
r
,
it
is
not
e
w
or
th
y
th
a
t
V
G
G
-
F
a
c
e
e
xhi
bi
ts
th
e
poor
e
s
t
pe
r
f
or
m
a
nc
e
c
om
pa
r
e
d
to
A
r
c
F
a
c
e
a
nd
F
a
c
e
N
e
t5
12
a
nd
f
a
ll
s
s
hor
t
of
M
T
C
N
N
.
T
h
e
s
e
r
e
s
ul
ts
e
m
pha
s
iz
e
th
e
nua
nc
e
d
pe
r
f
or
m
a
nc
e
tr
a
de
-
of
f
s
a
c
r
os
s
di
f
f
e
r
e
nt
a
lg
or
it
hm
s
,
w
it
h
A
r
c
F
a
c
e
e
xc
e
ll
in
g
in
s
pe
c
if
ic
m
e
tr
ic
s
but
pot
e
nt
ia
ll
y
e
xhi
bi
ti
ng
lo
w
e
r
a
b
s
ol
ut
e
r
e
c
ogni
ti
on
num
b
e
r
s
.
M
e
a
nw
hi
le
,
F
a
c
e
N
e
t5
12
di
s
pl
a
y
s
c
om
pe
ti
ti
ve
m
e
tr
ic
s
, pa
r
ti
c
ul
a
r
ly
c
onc
e
r
ni
ng
c
os
in
e
a
nd E
uc
li
de
a
n,
a
nd V
G
G
-
F
a
c
e
l
a
gs
be
hi
nd i
ts
c
ount
e
r
pa
r
ts
.
T
a
bl
e
5.
T
he
r
e
s
ul
t
of
f
a
c
ia
l
r
e
c
ogni
ti
on e
xpe
r
im
e
nt
s
f
or
5 di
f
f
e
r
e
nt
pos
e
us
in
g R
e
ti
na
F
a
c
e
M
ode
l
D
i
s
t
a
nc
e
m
e
t
r
i
c
P
e
r
c
e
nt
a
ge
C
or
r
e
c
t
(
%
)
W
r
ong (
%
)
I
nva
l
i
d (
%
)
A
r
c
F
a
c
e
C
os
i
ne
68.9
28.9
2.2
E
uc
l
i
de
a
n
37.8
20
42.2
E
uc
l
i
de
a
n L
2
66.7
24.4
8.9
F
a
c
e
N
e
t
512
C
os
i
ne
57.8
26.7
15.6
E
uc
l
i
de
a
n
57.8
42.2
0
E
uc
l
i
de
a
n L
2
66.7
33.3
0
VGG
-
F
a
c
e
C
os
i
ne
51.1
48.9
0
E
uc
l
i
de
a
n
53.3
46.7
0
E
uc
l
i
de
a
n L
2
51.1
48.9
0
B
a
s
e
d
on
T
a
bl
e
6,
th
e
5
m
ode
ls
w
it
h
th
e
hi
ghe
s
t
a
c
c
ur
a
c
y
a
r
e
de
te
r
m
in
e
d
ba
s
e
d
on
th
e
to
ta
l
num
be
r
of
c
or
r
e
c
t
f
a
c
ia
l
im
a
ge
r
e
c
ogni
ti
ons
f
r
om
s
e
v
e
r
a
l
e
xpe
r
im
e
n
ts
.
A
n
in
va
li
d
va
lu
e
,
c
ons
id
e
r
e
d
a
f
a
il
ur
e
in
c
a
pt
ur
in
g
a
nd
r
e
c
ogni
z
in
g
a
f
a
c
e
,
is
e
xc
lu
de
d
f
r
om
th
e
r
e
c
ogni
ti
on
e
r
r
or
c
a
lc
ul
a
ti
on.
D
e
s
pi
te
th
e
pr
oc
e
s
s
of
r
e
c
ogni
z
in
g
f
a
c
ia
l
im
a
ge
s
ba
s
e
d
on
it
s
a
c
c
ur
a
c
y
in
s
e
ve
r
a
l
e
x
pe
r
im
e
nt
s
,
e
r
r
or
s
in
f
a
c
ia
l
r
e
c
ogni
ti
on
pe
r
s
is
t.
T
he
r
e
f
or
e
,
th
e
e
li
m
in
a
ti
on
of
r
e
c
ogni
ti
on
e
r
r
or
s
ne
c
e
s
s
it
a
te
s
th
e
e
s
ta
bl
is
hm
e
nt
of
a
th
r
e
s
hol
d
f
or
th
e
r
e
s
ul
ti
ng
di
s
ta
nc
e
va
lu
e
.
T
hi
s
pr
oc
e
s
s
is
c
r
uc
ia
l
to
pr
e
ve
nt
e
r
r
or
s
in
f
a
c
ia
l
r
e
c
ogni
ti
on.
T
he
m
e
th
od
ut
il
iz
in
g
th
e
E
uc
li
de
a
n
L
2
nor
m
a
nd
th
e
c
os
in
e
m
e
th
od
f
o
r
im
a
ge
m
e
a
s
ur
e
m
e
nt
s
de
m
ons
tr
a
te
s
th
a
t
a
s
th
e
di
s
ta
nc
e
a
ppr
oa
c
he
s
z
e
r
o,
th
e
a
c
c
ur
a
c
y
of
f
a
c
e
r
e
c
ogni
ti
on
in
c
r
e
a
s
e
s
.
C
ons
e
que
nt
ly
,
th
e
th
r
e
s
hol
d
de
te
r
m
in
a
ti
on
is
ba
s
e
d
on
th
e
s
m
a
ll
e
s
t
va
lu
e
a
m
ong
th
e
f
a
ls
e
r
e
c
ogni
ti
on
di
s
ta
n
c
e
s
,
r
ounde
d
to
two
d
e
c
im
a
l
pl
a
c
e
s
,
e
n
s
ur
in
g
a
z
e
r
o
num
be
r
of
f
a
ls
e
po
s
it
iv
e
s
.
S
ub
s
e
que
nt
ly
,
tr
ue
ne
ga
ti
ve
s
a
r
e
m
e
a
s
ur
e
d
by
d
e
te
r
m
in
in
g
w
he
th
e
r
th
e
di
s
ta
nc
e
va
lu
e
f
or
c
or
r
e
c
t
r
e
c
ogni
ti
on
e
xc
e
e
ds
a
s
e
t
th
r
e
s
hol
d.
A
va
lu
e
is
c
ons
id
e
r
e
d
tr
ue
pos
it
iv
e
if
th
e
c
or
r
e
c
t
r
e
c
ogni
ti
on
di
s
ta
nc
e
is
le
s
s
th
a
n
th
e
s
pe
c
if
ie
d
th
r
e
s
hol
d.
F
ol
lo
w
in
g
th
is
,
th
e
f
a
ls
e
ne
ga
ti
ve
r
a
te
a
nd
a
c
c
ur
a
c
y a
r
e
c
a
lc
ul
a
te
d, a
id
in
g i
n i
de
nt
if
yi
ng t
he
m
os
t
e
f
f
e
c
ti
ve
m
e
th
od. T
he
s
e
f
in
di
ngs
hi
ghl
ig
ht
t
he
i
nt
r
ic
a
te
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
,
V
ol
. 14, No. 4, A
ugus
t
2025
:
3262
-
3273
3270
tr
a
de
-
of
f
s
in
pe
r
f
or
m
a
nc
e
a
c
r
os
s
di
f
f
e
r
e
nt
a
lg
or
it
hm
s
,
w
it
h
e
a
c
h
m
e
th
od
e
xhi
bi
ti
ng
di
s
ti
nc
t
s
tr
e
ngt
hs
a
nd
w
e
a
kne
s
s
e
s
i
n t
he
c
ont
e
xt
of
f
a
c
ia
l
r
e
c
ogni
ti
on a
c
c
ur
a
c
y.
T
a
bl
e
6.
T
op 5 me
th
ods
w
it
h t
he
hi
ghe
s
t
num
be
r
of
c
or
r
e
c
t
r
e
c
ogni
ti
ons
A
l
i
gnm
e
nt
M
ode
l
D
i
s
t
a
nc
e
m
e
t
r
i
c
C
or
r
e
c
t
W
r
ong
M
T
C
N
N
A
r
c
F
a
c
e
E
uc
l
i
de
a
n L
2
62
9
M
T
C
N
N
A
r
c
F
a
c
e
C
os
i
ne
62
10
R
e
t
i
na
F
a
c
e
A
r
c
F
a
c
e
C
os
i
ne
58
16
R
e
t
i
na
F
a
c
e
A
r
c
F
a
c
e
E
uc
l
i
de
a
n L
2
57
14
R
e
t
i
na
F
a
c
e
F
a
c
e
ne
t
512
E
uc
l
i
de
a
n L
2
49
26
T
a
bl
e
7
s
how
s
th
a
t
R
e
ti
na
F
a
c
e
,
ut
il
iz
in
g
th
e
A
r
c
F
a
c
e
m
od
e
l
w
it
h
th
e
c
o
s
in
e
di
s
ta
n
c
e
m
e
tr
ic
,
a
c
hi
e
ve
s
a
s
ig
ni
f
ic
a
nt
ly
hi
ghe
r
le
ve
l
of
a
c
c
ur
a
c
y
w
it
h
a
r
e
l
a
ti
ve
ly
f
a
s
t
a
v
e
r
a
ge
r
e
c
ogni
ti
on
ti
m
e
.
A
s
a
n
a
lt
e
r
na
ti
ve
,
w
it
h
a
s
li
ght
va
r
ia
nc
e
in
a
c
c
ur
a
c
y,
R
e
ti
na
F
a
c
e
,
e
m
pl
oyi
ng
th
e
A
r
c
F
a
c
e
m
ode
l
w
it
h
th
e
E
uc
li
de
a
n
L
2
nor
m
di
s
ta
nc
e
m
e
tr
ic
,
s
u
c
c
e
e
d
s
in
r
e
duc
in
g
th
e
num
be
r
o
f
in
c
or
r
e
c
t
r
e
c
ogni
ti
ons
,
w
hi
le
m
a
in
ta
in
in
g
a
c
om
pa
r
a
bl
e
a
c
c
ur
a
c
y
le
v
e
l
a
nd
a
r
e
la
ti
ve
ly
s
im
il
a
r
a
ve
r
a
ge
r
e
c
ogni
ti
on
ti
m
e
.
I
n
c
onc
lu
s
io
n,
w
he
n
th
e
E
uc
li
de
a
n L
2
no
r
m
is
us
e
d a
s
t
he
di
s
ta
nc
e
m
e
tr
ic
, t
hi
s
e
xpe
r
im
e
nt
de
m
ons
tr
a
te
s
be
tt
e
r
s
ta
bi
li
ty
due
t
o a
l
ow
e
r
num
be
r
of
f
a
ls
e
ne
ga
ti
ve
s
c
om
pa
r
e
d
to
us
in
g
c
os
in
e
.
C
o
ns
e
que
nt
ly
,
th
is
e
xpe
r
im
e
nt
de
te
r
m
in
e
s
th
a
t
R
e
ti
na
F
a
c
e
w
it
h
th
e
A
r
c
F
a
c
e
m
ode
l
ut
il
iz
in
g
th
e
E
uc
li
de
a
n
L
2
nor
m
di
s
ta
nc
e
m
e
tr
ic
i
s
th
e
opt
im
a
l
c
hoi
c
e
.
N
ot
onl
y
doe
s
it
pr
ov
id
e
s
uf
f
ic
ie
nt
a
c
c
ur
a
c
y,
but
it
a
ls
o
r
e
s
ul
ts
in
lo
w
e
r
f
a
ls
e
ne
ga
ti
ve
s
,
c
ont
r
ib
ut
in
g
t
o
e
nha
nc
e
d s
t
a
bi
li
ty
a
nd c
ons
is
te
n
c
y i
n f
a
c
ia
l
r
e
c
ogni
ti
on.
T
a
bl
e
7.
T
op 5 me
th
ods
w
it
h t
he
l
ow
e
s
t
f
a
ls
e
ne
g
a
ti
ve
r
a
te
va
lu
e
s
a
nd highe
s
t
a
c
c
ur
a
c
y
a
f
te
r
l
im
it
in
g on the
di
s
ta
nc
e
unt
il
no f
a
ls
e
po
s
it
iv
e
s
A
l
i
gnm
e
nt
M
ode
l
D
i
s
t
a
nc
e
m
e
t
r
i
c
D
i
s
t
a
nc
e
l
i
m
i
t
e
r
T
r
ue
pos
i
t
i
ve
T
r
ue
ne
ga
t
i
ve
F
a
l
s
e
ne
ga
t
i
ve
F
a
l
s
e
ne
ga
t
i
ve
r
a
t
e
A
c
c
ur
a
c
y
T
i
m
e
(
s
e
c
onds
)
R
e
t
i
na
F
a
c
e
A
r
c
F
a
c
e
C
os
i
ne
0.40
34
24
16
0.41379
0.676
1.473
R
e
t
i
na
F
a
c
e
A
r
c
F
a
c
e
E
uc
l
i
de
a
n L
2
0.89
33
24
14
0.42105
0.662
1.482
M
T
C
N
N
A
r
c
F
a
c
e
E
uc
l
i
de
a
n L
2
0.91
30
32
9
0.51613
0.549
2.494
M
T
C
N
N
A
r
c
F
a
c
e
C
os
i
ne
0.41
29
33
10
0.53226
0.542
2.518
R
e
t
i
na
F
a
c
e
F
a
c
e
N
e
t
512
E
uc
l
i
de
a
n L
2
0.56
5
44
26
0.89796
0.413
1.430
A
n
ove
r
vi
e
w
of
th
e
U
I
di
s
pl
a
y
f
or
th
e
box
bo
r
r
ow
in
g
us
in
g
th
e
r
e
c
ogni
ti
on
m
e
th
od
in
th
is
r
e
s
e
a
r
c
h
c
a
n
be
s
e
e
n
in
F
ig
ur
e
6.
T
h
e
U
I
d
a
s
hboa
r
d
in
F
ig
ur
e
6
pr
e
s
e
n
ts
th
e
pr
im
a
r
y
in
te
r
f
a
c
e
f
or
lo
c
ke
r
bor
r
ow
in
g
w
it
hi
n
th
e
Y
our
va
ul
t
a
ppl
ic
a
ti
on.
I
n
th
e
f
ir
s
t
m
e
nu,
us
e
r
s
a
r
e
p
r
om
pt
e
d
to
s
e
le
c
t
by
r
a
is
in
g
th
e
ir
le
f
t
ha
nd
to
bor
r
ow
a
lo
c
ke
r
or
th
e
ir
r
ig
ht
ha
nd
to
op
e
n
a
bor
r
ow
e
d
lo
c
ke
r
.
A
s
u
s
e
r
s
r
a
is
e
th
e
ir
ha
nd
s
,
a
c
ount
dow
n
a
ni
m
a
ti
on
c
onf
ir
m
s
th
e
m
e
nu
s
e
le
c
ti
on.
U
pon
c
hoos
in
g
to
op
e
n
a
lo
c
ke
r
,
us
e
r
s
a
r
e
di
r
e
c
te
d
to
th
e
s
e
c
ond
m
e
nu,
w
hi
c
h
of
f
e
r
s
th
e
c
hoi
c
e
to
e
it
he
r
ope
n
a
c
u
r
r
e
nt
ly
b
or
r
ow
e
d
lo
c
ke
r
or
e
nd
th
e
lo
c
ke
r
bor
r
ow
in
g
s
e
s
s
io
n.
T
h
e
na
vi
ga
ti
on
s
ys
te
m
in
th
e
s
e
c
ond
m
e
nu
a
ls
o
in
vo
lv
e
s
ha
nd
g
e
s
tu
r
e
s
.
O
nc
e
us
e
r
s
ha
ve
s
e
le
c
te
d
e
it
he
r
bor
r
ow
in
g
o
r
ope
ni
ng
a
lo
c
ke
r
,
th
e
a
ppl
ic
a
ti
on
d
is
pl
a
ys
a
U
I
gui
di
ng
us
e
r
s
to
pos
it
io
n
th
e
ir
f
a
c
e
s
f
or
e
f
f
e
c
ti
ve
r
e
c
ogni
ti
on.
A
t
th
e
bot
to
m
o
f
th
e
in
te
r
f
a
c
e
,
not
if
ic
a
ti
ons
in
f
or
m
us
e
r
s
a
bout
us
in
g
m
a
s
ks
,
non
-
a
ut
he
nt
ic
f
a
c
ia
l
f
e
a
tu
r
e
s
,
d
e
te
c
ti
on
of
m
or
e
th
a
n
one
f
a
c
e
,
a
nd
gui
da
nc
e
to
m
ove
c
lo
s
e
r
or
f
a
r
th
e
r
f
r
om
th
e
c
a
m
e
r
a
.
U
pon
s
uc
c
e
s
s
f
ul
f
a
c
ia
l
de
te
c
ti
on,
th
e
s
ys
te
m
c
a
pt
ur
e
s
a
n
im
a
ge
a
nd
id
e
nt
if
ie
s
th
e
us
e
r
'
s
f
a
c
e
I
D
.
I
f
th
e
us
e
r
is
r
e
c
ogni
z
e
d
a
nd
pos
s
e
s
s
e
s
s
uf
f
ic
ie
nt
to
ke
ns
,
th
e
a
ppl
ic
a
ti
on
di
s
pl
a
ys
th
e
us
e
r
'
s
na
m
e
a
nd
th
e
num
be
r
of
th
e
bor
r
ow
e
d
lo
c
ke
r
.
T
hi
s
in
te
r
f
a
c
e
a
ut
om
a
ti
c
a
ll
y
c
ha
nge
s
to
th
e
f
ir
s
t
m
e
nu
onc
e
u
s
e
r
s
s
e
c
ur
e
ly
c
lo
s
e
t
he
bor
r
ow
e
d l
oc
ke
r
door
.
T
he
U
I
di
s
pl
a
y
of
th
e
Y
our
va
ul
t
a
ppl
ic
a
ti
on
in
F
ig
ur
e
7
r
e
pr
e
s
e
nt
s
not
if
ic
a
ti
on
m
e
s
s
a
ge
s
to
th
e
us
e
r
.
I
f
th
e
us
e
r
la
c
ks
to
ke
ns
,
th
e
s
uc
c
e
s
s
f
ul
ly
r
e
c
ogni
z
e
d
u
s
e
r
'
s
na
m
e
a
nd
a
not
if
ic
a
ti
on
pr
om
pt
in
g
th
e
m
to
r
e
pl
e
ni
s
h
th
e
ir
to
ke
ns
a
r
e
s
how
n.
I
n
c
a
s
e
of
a
n
e
r
r
or
in
f
a
c
ia
l
r
e
c
ogni
ti
on,
th
e
s
y
s
te
m
i
s
s
ue
s
a
not
if
ic
a
ti
on
a
dvi
s
in
g
th
e
us
e
r
to
r
e
c
he
c
k
th
e
ir
f
a
c
ia
l
pos
it
io
ni
ng.
U
s
e
r
s
a
r
e
in
f
or
m
e
d
if
th
e
y
ha
ve
ye
t
to
bor
r
ow
a
lo
c
ke
r
w
he
n
a
tt
e
m
pt
in
g
to
ope
n
one
or
i
f
th
e
y
ha
ve
a
lr
e
a
dy
bor
r
ow
e
d
a
lo
c
ke
r
w
he
n
a
tt
e
m
pt
in
g
to
bor
r
ow
a
not
he
r
,
a
s
e
a
c
h
u
s
e
r
is
li
m
it
e
d
to
bor
r
ow
in
g
onl
y
one
lo
c
ke
r
a
t
a
gi
ve
n
lo
c
a
ti
on.
A
ddi
ti
ona
ll
y,
th
e
s
ys
te
m
a
le
r
ts
th
e
us
e
r
if
th
e
lo
c
ke
r
s
a
r
e
oc
c
upi
e
d.
D
ur
in
g
da
ta
ba
s
e
s
ync
hr
oni
z
a
ti
on,
th
e
s
ys
te
m
is
te
m
por
a
r
il
y
una
va
il
a
bl
e
,
a
nd
a
not
if
ic
a
ti
on
is
di
s
pl
a
ye
d,
in
f
or
m
in
g
us
e
r
s
th
a
t
th
e
s
ys
te
m
is
unde
r
goi
ng
da
ta
s
ync
hr
oni
z
a
ti
on.
T
hi
s
not
if
ic
a
ti
on
m
a
na
ge
s
us
e
r
e
xpe
c
ta
ti
ons
a
nd
pr
ovi
de
s
tr
a
ns
pa
r
e
nc
y
r
e
ga
r
di
ng
th
e
a
ppl
ic
a
ti
on'
s
s
ta
tu
s
dur
in
g
s
ync
hr
oni
z
a
ti
on pr
oc
e
s
s
e
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
E
nhanc
in
g t
ouc
hl
e
s
s
s
m
a
r
t
lo
c
k
e
r
s
y
s
te
m
s
t
hr
ough adv
anc
e
d f
ac
ia
l
r
e
c
ogni
ti
on
…
(
A
bdul
H
ar
is
R
angk
ut
i
)
3271
F
ig
ur
e
6. U
I
di
s
pl
a
y of
Y
our
va
ul
t
boot
h
a
ppl
ic
a
ti
on
F
ig
ur
e
7. A
le
r
t
or
not
if
ic
a
ti
on dis
pl
a
y U
I
of
Y
our
va
ul
t
boot
h a
ppl
ic
a
ti
on
5.
C
O
N
C
L
U
S
I
O
N
B
a
s
e
d on the
f
in
di
ngs
of
t
hi
s
r
e
s
e
a
r
c
h, t
he
i
m
pl
e
m
e
nt
a
ti
on of
f
a
c
e
r
e
c
ogni
ti
on w
it
h ha
nd de
te
c
ti
on
a
s
it
s
na
vi
ga
ti
on
s
ys
t
e
m
in
th
e
lo
c
ke
r
bor
r
ow
in
g
s
ys
te
m
ha
s
pr
ov
e
n
to
be
s
uc
c
e
s
s
f
ul
in
pr
ovi
di
ng
e
f
f
ic
ie
nt
a
nd
r
a
pi
d
s
e
c
ur
it
y
w
hi
le
m
in
im
iz
in
g
phys
ic
a
l
to
uc
h
dur
in
g
th
e
lo
c
ke
r
lo
c
ki
ng
pr
oc
e
s
s
.
T
he
e
xpe
r
im
e
nt
a
l
r
e
s
ul
ts
de
m
ons
tr
a
te
th
a
t
th
e
ut
il
iz
a
ti
on
o
f
R
e
ti
na
F
a
c
e
w
it
h
th
e
A
r
c
F
a
c
e
m
ode
l
a
nd
a
ppl
yi
ng
th
e
E
uc
li
de
a
n
L
2
nor
m
di
s
ta
nc
e
m
e
tr
ic
r
e
s
ul
t
in
hi
gh
a
nd
s
ta
bl
e
a
c
c
ur
a
c
y
le
ve
ls
.
T
he
Y
our
va
ul
t
a
ppl
ic
a
ti
on
e
xhi
bi
ts
c
om
m
e
nda
bl
e
f
a
c
ia
l
r
e
c
ogni
ti
on
c
a
p
a
bi
li
ti
e
s
,
e
ve
n
a
m
id
s
t
va
r
ia
ti
ons
in
f
a
c
ia
l
pos
e
s
.
F
ur
th
e
r
m
or
e
,
in
c
or
por
a
ti
ng
m
a
s
k
de
te
c
ti
on
a
nd
f
a
c
ia
l
a
ut
h
e
nt
ic
it
y
ve
r
if
ic
a
ti
on
us
in
g
Y
O
L
O
v8
a
dds
a
la
y
e
r
of
s
e
c
ur
it
y.
C
on
s
e
que
nt
ly
,
de
ve
lo
pi
ng
Y
our
va
ul
t
to
uc
hl
e
s
s
lo
c
ke
r
s
e
c
ur
it
y
s
ys
te
m
is
e
xpe
c
te
d
to
be
a
pr
a
c
ti
c
a
l
s
ol
ut
io
n,
r
e
pl
a
c
in
g
c
onve
nt
io
na
l
m
e
th
ods
in
vol
vi
ng
phys
ic
a
l
to
uc
h.
T
hi
s
s
ys
te
m
hol
ds
pot
e
nt
ia
l
f
or
im
pl
e
m
e
nt
a
ti
on
in
publ
ic
lo
c
ke
r
s
e
tt
in
gs
,
of
f
e
r
in
g
he
ig
ht
e
ne
d
s
e
c
ur
it
y
le
ve
ls
a
nd
opt
im
a
l
us
e
r
c
onve
ni
e
nc
e
.
F
or
f
ur
th
e
r
r
e
s
e
a
r
c
h,
th
e
r
e
is
a
n
e
xpe
c
ta
ti
on
f
or
s
ys
te
m
opt
im
iz
a
ti
on
to
a
c
hi
e
ve
f
a
s
te
r
s
pe
e
ds
w
hi
le
m
a
in
ta
in
in
g
r
obus
t
us
e
r
f
a
c
ia
l
da
ta
s
e
c
ur
it
y.
A
ddi
ti
ona
ll
y,
in
-
de
pt
h
r
e
s
e
a
r
c
h
is
w
a
r
r
a
nt
e
d,
in
vol
vi
ng
c
ol
le
c
ti
ng
m
or
e
va
r
ie
d
da
ta
,
in
c
lu
di
ng
di
ve
r
s
e
a
ge
gr
oups
, e
th
ni
c
it
ie
s
, a
nd othe
r
uni
que
c
ha
r
a
c
te
r
is
ti
c
s
.
F
U
N
D
I
N
G
I
N
F
O
R
M
A
T
I
O
N
T
h
i
s
r
e
s
e
a
r
c
h
r
e
c
e
i
v
e
d
no
s
p
e
c
if
i
c
g
r
a
n
t
f
r
o
m
a
n
y f
u
nd
in
g
a
g
e
n
c
y,
c
o
m
m
e
r
c
i
a
l,
or
n
ot
-
f
o
r
-
p
r
o
f
i
t
s
e
c
t
or
s
.
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.
Evaluation Warning : The document was created with Spire.PDF for Python.