I
AE
S In
t
er
na
t
io
na
l J
o
urna
l o
f
Art
if
icia
l In
t
ellig
ence
(
I
J
-
AI
)
Vo
l.
1
5
,
No
.
4
,
A
u
g
u
s
t 2
0
2
6
,
p
p
.
3
9
0
3
~
3
9
1
8
I
SS
N:
2
2
5
2
-
8
9
3
8
,
DOI
: 1
0
.
1
1
5
9
1
/ijai.v
15
.i
4
.
p
p
3
9
0
3
-
3
9
1
8
3903
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
a
i
.
ia
esco
r
e.
co
m
A nov
el embed
de
d app
ro
a
ch t
o
face recog
nition us
i
n
g
multi
-
threaded
contro
ller bas
ed
on
weig
htless ne
ura
l
network
Ahm
a
d Z
a
rk
a
s
i
1
,
H
a
dip
urna
wa
n Sa
t
ria
2
,
Ang
g
ina
P
ri
ma
nita
2
,
Der
is
Stia
wa
n
1
1
D
e
p
a
r
t
me
n
t
o
f
C
o
m
p
u
t
e
r
En
g
i
n
e
e
r
i
n
g
,
F
a
c
u
l
t
y
o
f
C
o
mp
u
t
e
r
S
c
i
e
n
c
e
,
U
n
i
v
e
r
si
t
a
s Sr
i
w
i
j
a
y
a
,
P
a
l
e
mb
a
n
g
,
I
n
d
o
n
e
s
i
a
2
D
e
p
a
r
t
me
n
t
o
f
I
n
f
o
r
mat
i
c
s
E
n
g
i
n
e
e
r
i
n
g
,
F
a
c
u
l
t
y
o
f
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
,
U
n
i
v
e
r
si
t
a
s
S
r
i
w
i
j
a
y
a
,
P
a
l
e
m
b
a
n
g
,
I
n
d
o
n
e
si
a
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
No
v
1
,
2
0
2
5
R
ev
is
ed
J
u
l 1
3
,
2
0
2
6
Acc
ep
ted
J
u
l 2
1
,
2
0
2
6
Th
is
re
se
a
rc
h
p
re
se
n
ts
a
h
i
g
h
e
ff
icie
n
c
y
e
m
b
e
d
d
e
d
fa
c
e
re
c
o
g
n
it
i
o
n
sy
ste
m
b
a
se
d
o
n
t
h
e
we
ig
h
tl
e
ss
n
e
u
ra
l
n
e
two
rk
-
fa
c
e
re
c
o
g
n
it
i
o
n
a
lg
o
rit
h
m
(W
NN
-
F
RA)
in
teg
ra
ted
wit
h
a
m
u
lt
i
-
t
h
re
a
d
c
o
n
tro
l
ler
to
e
n
h
a
n
c
e
e
x
e
c
u
ti
o
n
ti
m
e
a
n
d
re
c
o
g
n
it
i
o
n
a
c
c
u
ra
c
y
u
n
d
e
r
li
m
it
e
d
h
a
rd
wa
re
re
so
u
rc
e
s.
T
h
e
sy
ste
m
imp
lem
e
n
ts
a
c
e
n
ter
-
sc
a
n
fe
a
tu
re
a
li
g
n
m
e
n
t
m
o
d
e
l
to
a
d
d
re
ss
re
so
lu
ti
o
n
d
isc
re
p
a
n
c
ies
b
e
twe
e
n
re
fe
re
n
c
e
a
n
d
i
n
p
u
t
fa
c
ial
ima
g
e
s.
T
h
e
m
u
lt
i
-
th
re
a
d
e
d
a
rc
h
it
e
c
tu
re
d
i
v
id
e
s
p
ro
c
e
ss
in
g
i
n
to
th
re
e
c
o
n
c
u
rre
n
t
t
h
re
a
d
s
fro
n
t,
left,
a
n
d
rig
h
t
fa
c
ial
o
r
ien
tatio
n
s
e
a
c
h
h
a
n
d
li
n
g
a
p
p
r
o
x
ima
tely
2
0
fa
c
ial
p
a
tt
e
rn
s.
Ex
p
e
rime
n
tal
e
v
a
l
u
a
ti
o
n
o
n
a
d
a
tas
e
t
o
f
6
0
fa
c
ial
ima
g
e
s
d
e
m
o
n
stra
ted
a
m
a
x
imu
m
re
c
o
g
n
it
io
n
a
c
c
u
ra
c
y
o
f
9
6
.
8
3
%
a
n
d
a
n
a
v
e
ra
g
e
e
x
e
c
u
ti
o
n
ti
m
e
ra
n
g
in
g
fr
o
m
1
6
to
3
4
m
il
li
s
e
c
o
n
d
s
p
e
r
d
a
tas
e
t,
c
o
n
firmi
n
g
re
a
l
-
ti
m
e
p
e
rfo
rm
a
n
c
e
.
Co
m
p
a
ra
ti
v
e
a
n
a
ly
sis
sh
o
ws
th
a
t
th
e
m
u
l
ti
-
th
re
a
d
e
d
a
p
p
ro
a
c
h
re
d
u
c
e
d
t
h
e
e
x
e
c
u
t
io
n
t
ime
b
y
o
v
e
r
6
7
%
c
o
m
p
a
re
d
to
sin
g
le
-
th
re
a
d
p
ro
c
e
ss
in
g
0
.
0
9
se
c
o
n
d
v
s.
0
.
2
7
1
se
c
o
n
d
,
wh
il
e
m
a
in
tain
in
g
b
a
lan
c
e
d
wo
rk
l
o
a
d
d
istri
b
u
ti
o
n
a
c
ro
ss
t
h
r
e
a
d
s.
M
e
m
o
ry
a
n
a
l
y
sis
re
v
e
a
led
th
a
t
th
e
e
n
ti
re
sy
ste
m
re
q
u
ire
d
o
n
l
y
2
4
K
B
fro
m
th
e
a
v
a
il
a
b
le 5
1
2
KB
flas
h
c
a
p
a
c
it
y
,
in
d
ica
ti
n
g
e
fficie
n
t
re
so
u
rc
e
u
ti
li
z
a
ti
o
n
.
Th
e
re
su
lt
s
c
o
n
firm
th
a
t
i
n
teg
ra
ti
n
g
WNN
-
F
RA
with
m
u
l
ti
-
th
re
a
d
in
g
p
r
o
v
i
d
e
s
a
ro
b
u
st,
l
o
w
-
c
o
st,
a
n
d
sc
a
lab
le
so
lu
ti
o
n
fo
r
re
a
l
-
ti
m
e
fa
c
ial
p
a
tt
e
r
n
re
c
o
g
n
it
i
o
n
in
e
m
b
e
d
d
e
d
e
n
v
ir
o
n
m
e
n
ts.
K
ey
w
o
r
d
s
:
C
en
ter
s
ca
n
r
ec
o
g
n
itio
n
E
m
b
ed
d
e
d
s
y
s
tem
Face
r
ec
o
g
n
itio
n
Mu
lti
-
th
r
ea
d
co
n
tr
o
ller
R
ea
l
-
tim
e
p
r
o
ce
s
s
in
g
W
eig
h
tles
s
n
eu
r
al
n
etwo
r
k
T
h
is i
s
a
n
o
p
e
n
a
c
c
e
ss
a
rticle
u
n
d
e
r th
e
CC B
Y
-
SA
li
c
e
n
se
.
C
o
r
r
e
s
p
o
nd
ing
A
uth
o
r
:
Ah
m
ad
Z
ar
k
asi
Dep
ar
tm
en
t o
f
C
o
m
p
u
ter
E
n
g
i
n
ee
r
in
g
,
Facu
lty
o
f
C
o
m
p
u
ter
Scien
ce
,
Un
iv
er
s
itas
Sriwijay
a
Palem
b
an
g
,
I
n
d
o
n
esia
E
m
ail: z
ar
k
asi9
8
@
g
m
ail.
co
m
1.
I
NT
RO
D
UCT
I
O
N
T
h
e
p
r
im
ar
y
c
h
allen
g
e
in
i
m
ag
e
p
atter
n
p
r
o
ce
s
s
in
g
o
n
em
b
ed
d
e
d
s
y
s
tem
s
lies
in
th
e
in
h
er
en
t
lim
itatio
n
s
o
f
th
eir
co
m
p
u
tatio
n
al
r
eso
u
r
ce
s
.
T
h
is
ch
allen
g
e
b
ec
o
m
es
in
cr
ea
s
in
g
ly
co
m
p
le
x
wh
en
th
e
s
y
s
tem
is
r
eq
u
ir
ed
to
co
m
b
in
e
m
u
ltip
le
an
aly
tical
m
eth
o
d
s
f
o
r
p
r
o
c
ess
in
g
th
e
s
am
e
o
b
ject
[
1
]
–
[
3
]
.
Su
ch
co
n
s
tr
ain
ts
o
f
ten
h
in
d
er
th
e
d
ev
el
o
p
m
en
t a
n
d
im
p
lem
e
n
tatio
n
o
f
f
ac
ial
p
atter
n
r
ec
o
g
n
itio
n
s
y
s
tem
s
o
n
d
ev
ices
em
p
l
o
y
in
g
s
in
g
le
-
ch
ip
co
n
tr
o
ller
ar
c
h
itectu
r
es
[
4
]
–
[
6
]
.
T
o
en
s
u
r
e
t
h
e
ef
f
ec
tiv
en
ess
o
f
f
ac
e
-
p
att
er
n
p
r
o
ce
s
s
in
g
o
n
em
b
ed
d
e
d
p
latf
o
r
m
s
,
two
k
e
y
cr
iter
ia
m
u
s
t
b
e
m
et:
f
ast
ex
ec
u
tio
n
tim
e
an
d
f
lex
i
b
ilit
y
in
th
e
class
if
icatio
n
p
r
o
ce
s
s
[
7
]
.
Sev
er
al
p
r
e
v
io
u
s
s
tu
d
ies
h
av
e
d
em
o
n
s
tr
ated
t
h
at
th
e
weig
h
tles
s
n
eu
r
al
n
etwo
r
k
(
W
NN)
ap
p
r
o
ac
h
,
p
ar
ticu
lar
ly
th
e
W
ilk
ie,
Sto
n
h
am
,
an
d
Alek
s
an
d
er
r
ec
o
g
n
iti
o
n
d
ev
ice
(
W
iSAR
D
)
m
o
d
el
[
8
]
–
[
1
0
]
,
p
r
o
v
id
es
an
ef
f
icien
t
s
o
lu
tio
n
to
th
ese
c
h
a
llen
g
es.
T
h
e
W
iSAR
D
m
eth
o
d
u
tili
ze
s
a
s
to
r
ag
e
an
d
class
if
icatio
n
m
ec
h
a
n
is
m
in
wh
ich
f
ac
ial
p
atter
n
d
ata
a
r
e
d
is
tr
ib
u
ted
ac
r
o
s
s
m
u
ltip
le
r
an
d
o
m
-
ac
ce
s
s
m
em
o
r
y
(
R
AM
)
n
o
d
es
o
r
g
a
n
ized
with
in
a
s
et
o
f
R
AM
d
is
cr
im
in
ato
r
s
,
ea
ch
r
esp
o
n
s
ib
le
f
o
r
id
en
tify
in
g
a
s
p
ec
if
ic
p
atter
n
c
lass
.
T
h
e
W
iS
AR
D
ar
ch
itectu
r
e
co
n
s
is
ts
o
f
s
ev
er
al
d
is
cr
im
in
ato
r
s
o
p
er
atin
g
i
n
p
ar
allel
to
ev
alu
ate
in
p
u
t
im
ag
es
an
d
d
eter
m
in
e
th
eir
co
r
r
esp
o
n
d
en
ce
to
th
e
r
ep
r
esen
ted
class
.
I
ts
g
en
er
ativ
e
p
r
o
p
er
ty
m
a
k
es
W
iSA
R
D
p
ar
ticu
lar
ly
r
elev
an
t
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
1
5
,
No
.
4
,
Au
g
u
s
t 2
0
2
6
:
3
9
0
3
-
3
9
1
8
3904
f
o
r
ap
p
licatio
n
s
in
o
p
en
s
et
r
e
co
g
n
itio
n
,
wh
er
e
th
e
s
y
s
tem
m
u
s
t
id
en
tify
p
r
ev
io
u
s
ly
u
n
s
e
en
d
ata
th
at
ar
e
n
o
t
p
ar
t
o
f
th
e
tr
ain
in
g
s
et
[
1
1
]
–
[
1
3
]
.
R
esear
ch
o
n
o
p
en
s
et
r
ec
o
g
n
itio
n
h
as
s
h
o
w
n
th
at
W
iSA
R
D
ex
h
ib
its
a
s
u
p
er
io
r
a
b
ilit
y
to
r
ep
r
esen
t
th
e
d
is
tin
ct
ch
ar
ac
ter
is
tics
o
f
d
ata
d
is
tr
ib
u
tio
n
s
co
m
p
ar
ed
to
co
n
v
en
tio
n
al
class
if
icatio
n
ap
p
r
o
ac
h
es.
T
h
e
p
r
in
cip
al
ad
v
a
n
tag
e
o
f
th
e
W
iSA
R
D
m
eth
o
d
lies
in
its
ab
ilit
y
to
p
r
o
d
u
ce
s
m
o
o
th
er
m
atch
in
g
co
n
to
u
r
s
,
m
o
r
e
e
v
en
ly
d
is
tr
ib
u
ted
m
ea
s
u
r
em
en
t
d
ata,
a
n
d
m
o
r
e
r
e
p
r
e
s
en
tativ
e
s
im
ilar
ity
v
alu
es
b
etwe
en
in
p
u
t
an
d
r
ef
e
r
en
ce
p
atter
n
s
[
1
4
]
,
[
1
5
]
.
T
h
e
s
e
ch
ar
ac
ter
is
tics
m
ak
e
W
iSA
R
D
h
ig
h
ly
s
u
itab
le
f
o
r
im
p
lem
e
n
tatio
n
o
n
em
b
ed
d
ed
s
y
s
tem
s
,
as
all
co
m
p
u
tatio
n
al
p
r
o
ce
s
s
es
ar
e
ex
ec
u
ted
d
ir
ec
tly
with
in
th
e
R
AM
u
n
its
,
s
ig
n
if
ican
tly
r
ed
u
cin
g
p
r
o
ce
s
s
in
g
lo
ad
an
d
ex
ec
u
tio
n
tim
e
[
1
6
]
,
[
1
7
]
.
E
m
b
ed
d
e
d
-
b
ased
f
ac
e
p
atter
n
p
r
o
ce
s
s
in
g
is
a
tech
n
ical
a
p
p
r
o
ac
h
th
at
i
n
teg
r
ates
m
u
ltip
le
p
r
o
ce
s
s
o
r
co
m
p
o
n
en
ts
d
esig
n
ed
to
s
u
p
p
o
r
t
ea
ch
o
th
er
,
b
o
th
in
ter
m
s
o
f
s
o
f
twar
e
an
d
h
ar
d
war
e
[
1
8
]
–
[
2
0
]
.
T
h
e
m
ain
o
b
jectiv
e
o
f
th
is
in
teg
r
atio
n
i
s
to
en
ab
le
th
e
s
y
s
tem
to
o
p
e
r
ate
in
p
ar
allel
a
n
d
s
im
u
ltan
e
o
u
s
ly
,
as
well
as
to
d
is
tr
ib
u
te
th
e
co
m
p
u
tatio
n
al
wo
r
k
lo
ad
b
etwe
en
h
ar
d
wa
r
e
a
n
d
s
o
f
twar
e
co
m
p
o
n
e
n
ts
to
ac
h
iev
e
m
o
r
e
e
f
f
icien
t
f
ac
ial
r
ec
o
g
n
itio
n
p
r
o
ce
s
s
in
g
.
Sev
er
al
p
r
e
v
io
u
s
s
tu
d
ies
[
2
1
]
–
[
2
3
]
h
av
e
s
h
o
wn
th
at
f
ac
ial
p
atter
n
s
ca
n
b
e
g
r
o
u
p
ed
in
to
m
u
ltip
le
d
is
cr
i
m
in
ato
r
s
,
ea
ch
s
er
v
in
g
as
a
p
r
o
ce
s
s
in
g
u
n
it
f
o
r
th
e
f
ac
ial
r
ec
o
g
n
itio
n
task
.
T
o
ac
ce
ler
ate
th
e
id
en
tific
atio
n
p
r
o
ce
s
s
,
ea
ch
f
ac
ial
g
r
o
u
p
is
ass
ig
n
ed
a
s
p
ec
if
ic
lab
el
o
r
b
in
ar
y
co
d
e
f
o
r
in
s
tan
ce
,
th
e
f
r
o
n
t
f
ac
e
(
1
0
)
,
r
ig
h
t
f
ac
e
(
0
1
)
,
an
d
le
f
t
f
ac
e
(
1
1
)
.
W
h
en
th
e
in
p
u
t
d
ata
co
r
r
esp
o
n
d
s
to
a
f
r
o
n
t
-
f
ac
e
lab
el
(
1
0
)
,
th
e
alg
o
r
ith
m
p
er
f
o
r
m
s
r
ec
o
g
n
itio
n
o
n
ly
with
in
th
e
f
r
o
n
t
-
f
ac
e
lib
r
ar
y
,
wh
ile
th
e
r
ig
h
t
an
d
lef
t
-
f
ac
e
lib
r
ar
ies
ar
e
b
y
p
ass
ed
[
2
4
]
.
Al
th
o
u
g
h
th
is
s
tr
ateg
y
e
f
f
ec
tiv
el
y
r
ed
u
ce
s
ex
ec
u
tio
n
tim
e,
t
h
e
alg
o
r
ith
m
p
r
esen
ts
ce
r
tain
lim
itatio
n
s
,
as
n
o
t
all
f
ac
ial
d
ata
ar
e
p
r
o
ce
s
s
ed
s
i
m
u
ltan
eo
u
s
ly
.
I
f
all
f
ac
e
p
att
er
n
s
wer
e
p
r
o
ce
s
s
ed
co
n
cu
r
r
en
tly
,
th
e
e
x
ec
u
tio
n
ti
m
e
wo
u
ld
in
cr
ea
s
e
u
p
to
th
r
ee
f
o
ld
.
Mo
r
eo
v
er
,
th
e
r
ec
o
g
n
itio
n
ac
cu
r
ac
y
f
o
r
r
ig
h
t
an
d
lef
t
f
ac
es
is
n
o
t
n
ec
ess
ar
ily
lo
wer
th
an
th
at
o
f
t
h
e
f
r
o
n
t
f
ac
e.
T
h
is
lim
itatio
n
o
f
ten
ar
is
es
f
r
o
m
ch
allen
g
es
in
ac
cu
r
ately
d
eter
m
i
n
in
g
th
e
r
eg
io
n
o
f
in
ter
est
(
R
OI
)
,
p
a
r
ticu
lar
ly
f
o
r
f
ac
es
o
r
ie
n
ted
t
o
th
e
r
ig
h
t
o
r
lef
t.
C
o
n
s
eq
u
en
tly
,
th
e
s
y
s
tem
m
ay
s
o
m
etim
es
r
ec
o
g
n
ize
r
ig
h
t
o
r
lef
t
f
ac
es
ac
cu
r
ately
ev
en
wh
en
th
e
in
p
u
t
co
r
r
esp
o
n
d
s
to
a
f
r
o
n
t
-
f
ac
e
p
a
tter
n
,
an
d
v
ice
v
er
s
a,
r
esu
ltin
g
in
a
ce
r
tain
d
e
g
r
ee
o
f
u
n
ce
r
ta
in
ty
in
r
ec
o
g
n
itio
n
o
u
tco
m
es.
T
o
a
d
d
r
ess
th
ese
c
h
allen
g
es,
a
d
ata
-
p
r
o
ce
s
s
in
g
m
ec
h
an
is
m
ca
p
a
b
le
o
f
o
p
er
at
in
g
in
p
ar
allel
a
n
d
s
im
u
ltan
eo
u
s
ly
with
in
a
s
in
g
l
e
-
ch
ip
co
n
tr
o
ller
is
r
e
q
u
ir
e
d
.
T
h
e
m
o
s
t
ap
p
r
o
p
r
iate
ap
p
r
o
ac
h
f
o
r
th
is
p
u
r
p
o
s
e
is
th
e
im
p
lem
en
tatio
n
o
f
a
m
u
lt
ith
r
ea
d
in
g
m
ec
h
an
is
m
[
2
5
]
,
[
2
6
]
,
wh
ich
e
n
ab
les
th
e
f
ac
ial
p
atter
n
r
ec
o
g
n
itio
n
p
r
o
ce
s
s
to
b
e
ex
ec
u
ted
co
n
c
u
r
r
en
tly
ac
r
o
s
s
m
u
ltip
le
in
d
e
p
en
d
en
t
e
x
ec
u
tio
n
th
r
ea
d
s
,
t
h
er
eb
y
m
in
im
izin
g
p
r
o
ce
s
s
in
g
tim
e
with
o
u
t c
o
m
p
r
o
m
is
in
g
s
y
s
tem
ac
cu
r
ac
y
.
L
iter
ally
,
a
th
r
ea
d
is
d
ef
in
ed
as
th
e
s
m
allest
u
n
it
o
f
a
p
r
o
ce
s
s
o
r
th
at
ca
n
b
e
ex
ec
u
ted
with
in
a
p
r
o
ce
s
s
.
A
th
r
ea
d
en
c
o
m
p
ass
es
s
ev
er
al
p
r
o
ce
s
s
o
r
r
eso
u
r
ce
s
,
in
clu
d
in
g
a
p
r
o
g
r
am
c
o
u
n
ter
,
a
s
tack
p
o
in
ter
,
a
n
d
a
d
ed
icate
d
d
ata
a
r
ea
f
o
r
it
s
o
wn
s
tack
,
allo
win
g
th
e
p
r
o
ce
s
s
o
r
to
in
d
e
p
en
d
e
n
tly
ex
ec
u
te
b
r
an
ch
in
g
in
s
tr
u
ctio
n
s
[
2
7
]
–
[
2
9
]
.
Op
er
at
io
n
ally
,
a
th
r
ea
d
ca
n
f
u
n
ctio
n
in
th
r
ee
p
r
im
a
r
y
m
o
d
es:
i
)
s
eq
u
en
tial
ex
ec
u
tio
n
,
ii)
in
ter
r
u
p
tio
n
allo
win
g
th
e
p
r
o
ce
s
s
o
r
to
s
witch
to
an
o
th
er
th
r
ea
d
,
an
d
iii)
co
n
c
u
r
r
e
n
t
ex
ec
u
tio
n
with
o
th
er
th
r
ea
d
s
with
in
th
e
s
am
e
p
r
o
ce
s
s
,
a
m
ec
h
an
is
m
co
m
m
o
n
l
y
k
n
o
wn
as
m
u
ltit
h
r
ea
d
in
g
.
Mu
ltit
h
r
ea
d
in
g
is
a
p
r
o
ce
s
s
in
g
tech
n
iq
u
e
in
wh
ic
h
a
s
in
g
le
p
r
o
ce
s
s
with
in
an
ap
p
licatio
n
is
d
iv
id
ed
in
to
m
u
lti
p
le
th
r
ea
d
s
th
at
ca
n
r
u
n
in
p
ar
allel
[
3
0
]
–
[
3
2
]
.
T
h
i
s
ap
p
r
o
ac
h
en
h
an
ce
s
co
m
p
u
t
atio
n
al
ef
f
icien
cy
s
in
ce
ea
ch
th
r
ea
d
ca
n
h
an
d
le
d
if
f
er
en
t
p
ar
ts
o
f
a
task
s
im
u
ltan
eo
u
s
ly
.
C
o
n
s
eq
u
en
tly
,
th
e
im
p
lem
en
tatio
n
o
f
m
u
lti
-
th
r
ea
d
in
g
s
er
v
es
as
a
p
o
ten
tial
s
o
lu
tio
n
to
o
v
er
c
o
m
e
ex
ec
u
tio
n
tim
e
c
o
n
s
tr
ain
ts
in
em
b
ed
d
e
d
-
b
ased
s
y
s
tem
s
.
I
n
th
is
r
esear
ch
,
a
W
NN
m
eth
o
d
co
m
b
in
ed
with
th
e
ce
n
ter
s
ca
n
alg
o
r
ith
m
is
p
r
o
p
o
s
ed
as
th
e
p
r
im
ar
y
a
p
p
r
o
a
ch
f
o
r
f
ac
e
-
p
atter
n
r
ec
o
g
n
itio
n
.
T
h
e
e
n
tire
r
ec
o
g
n
itio
n
p
r
o
ce
s
s
is
im
p
lem
en
ted
with
in
a
m
u
lti
-
th
r
ea
d
co
n
tr
o
ll
er
to
en
s
u
r
e
p
ar
allel
an
d
ef
f
icien
t
e
x
ec
u
tio
n
.
T
h
e
m
ain
o
b
jectiv
e
o
f
th
is
r
esear
c
h
is
to
d
ev
elo
p
a
n
o
v
el
alg
o
r
ith
m
th
at
p
er
f
o
r
m
s
f
ac
ial
d
ata
m
atch
in
g
b
ased
o
n
th
e
ce
n
tr
al
p
o
s
itio
n
o
f
an
im
ag
e
b
y
f
ir
s
t
d
eter
m
in
in
g
its
o
p
tim
al
ce
n
ter
p
o
i
n
t.
T
h
is
ap
p
r
o
ac
h
is
e
x
p
ec
ted
to
p
r
o
d
u
ce
a
s
im
p
le,
f
ast,
an
d
ac
cu
r
ate
co
m
p
u
tatio
n
al
p
r
o
ce
s
s
f
o
r
f
ac
e
-
p
atter
n
r
ec
o
g
n
itio
n
,
en
a
b
lin
g
th
e
co
r
e
d
ata
to
b
e
p
r
o
ce
s
s
ed
c
o
m
p
lete
ly
an
d
c
o
r
r
ec
tly
.
Fu
r
th
er
m
o
r
e,
th
e
p
r
o
ce
s
s
ed
d
ata
ar
e
g
r
o
u
p
ed
in
to
m
em
o
r
y
ce
ll
s
f
o
r
s
u
b
s
eq
u
en
t
an
aly
s
is
,
g
en
er
atin
g
h
ig
h
-
q
u
ality
in
f
o
r
m
a
tio
n
th
at
ac
cu
r
ately
r
ep
r
esen
ts
th
e
d
eg
r
ee
o
f
s
im
ilar
ity
b
etwe
en
t
h
e
in
p
u
t
p
atte
r
n
an
d
th
e
r
ef
e
r
en
ce
p
atter
n
.
Ov
er
all,
th
is
s
tu
d
y
aim
s
to
d
ev
elo
p
a
f
ac
e
p
atter
n
d
etec
tio
n
s
y
s
tem
ch
a
r
ac
ter
ized
b
y
s
h
o
r
ter
ex
ec
u
tio
n
tim
e,
h
ig
h
er
r
ec
o
g
n
itio
n
ac
cu
r
ac
y
,
an
d
a
h
ar
d
war
e
p
r
o
t
o
ty
p
e
d
esig
n
th
at
is
s
im
p
le,
co
s
t
-
ef
f
ec
tiv
e
,
an
d
e
f
f
icien
t
f
o
r
em
b
ed
d
e
d
im
p
lem
en
tatio
n
.
2.
M
AT
E
R
I
AL
A
ND
CO
NT
R
O
L
L
E
R
2
.
1
.
M
ulti
-
t
hrea
ded c
o
ntr
o
ller
As
a
co
m
p
o
n
e
n
t
o
f
c
o
m
p
u
tat
io
n
,
a
th
r
ea
d
ca
r
r
ies
th
e
ess
en
tial
p
r
o
ce
s
s
in
g
r
eso
u
r
ce
s
r
e
q
u
ir
ed
f
o
r
p
r
o
g
r
a
m
ex
ec
u
tio
n
.
T
h
ese
in
clu
d
e
th
e
p
r
o
g
r
am
co
u
n
ter
,
wh
i
ch
s
to
r
es
th
e
ad
d
r
ess
o
f
th
e
n
ex
t
in
s
tr
u
ctio
n
to
b
e
ex
ec
u
ted
;
th
e
s
tack
p
o
in
ter
,
w
h
ich
p
o
in
ts
to
th
e
m
em
o
r
y
lo
c
atio
n
with
in
th
e
s
tack
;
an
d
a
d
ed
icate
d
d
ata
ar
ea
f
o
r
th
e
s
tack
,
wh
ich
en
ab
les
th
e
ex
ec
u
tio
n
o
f
b
r
a
n
ch
in
g
in
s
tr
u
ctio
n
s
an
d
s
u
b
r
o
u
tin
e
ca
lls
.
W
i
th
th
ese
attr
ib
u
tes,
a
t
h
r
ea
d
is
ca
p
a
b
le
o
f
p
er
f
o
r
m
in
g
p
r
o
g
r
a
m
e
x
ec
u
t
io
n
in
a
s
tr
u
ctu
r
ed
an
d
f
lex
ib
l
e
m
an
n
er
.
B
ased
o
n
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
A
n
o
ve
l e
mb
e
d
d
ed
a
p
p
r
o
a
ch
t
o
fa
ce
r
ec
o
g
n
itio
n
u
s
in
g
m
u
lti
-
th
r
ea
d
ed
c
o
n
tr
o
ller
b
a
s
ed
…
(
A
h
ma
d
Za
r
ka
s
i
)
3905
its
o
p
er
atio
n
al
p
r
in
cip
les,
a
th
r
ea
d
ca
n
f
u
n
ctio
n
in
th
r
ee
p
r
im
ar
y
co
n
d
itio
n
s
:
f
ir
s
t,
it
ca
n
b
e
ex
ec
u
ted
s
eq
u
en
tially
f
o
llo
win
g
th
e
in
s
tr
u
ctio
n
f
lo
w;
s
ec
o
n
d
,
it
ca
n
b
e
tem
p
o
r
ar
ily
p
a
u
s
ed
o
r
in
te
r
r
u
p
ted
to
allo
w
th
e
p
r
o
ce
s
s
o
r
to
s
witch
ex
ec
u
tio
n
to
an
o
th
er
th
r
ea
d
;
an
d
th
ir
d
,
it
ca
n
r
u
n
c
o
n
cu
r
r
en
tly
with
o
th
er
th
r
ea
d
s
with
in
th
e
s
am
e
p
r
o
ce
s
s
,
a
c
o
n
ce
p
t
k
n
o
wn
as
m
u
ltit
h
r
ea
d
in
g
[
3
3
]
,
[
3
4
]
.
Mu
ltit
h
r
ea
d
i
n
g
its
elf
is
a
p
r
o
g
r
am
m
i
n
g
a
n
d
p
r
o
ce
s
s
m
an
ag
em
en
t
tech
n
iq
u
e
in
wh
ich
a
s
in
g
le
p
r
o
ce
s
s
ex
ec
u
tin
g
an
ap
p
licatio
n
is
d
iv
id
ed
in
to
m
u
ltip
le
th
r
ea
d
s
th
at
ca
n
o
p
e
r
ate
in
p
ar
allel
o
r
s
im
u
ltan
eo
u
s
ly
.
T
h
is
ap
p
r
o
ac
h
en
h
an
ce
s
p
r
o
ce
s
s
in
g
ef
f
icien
cy
b
y
d
is
tr
ib
u
tin
g
wo
r
k
lo
ad
s
ac
r
o
s
s
s
ev
er
al
ex
ec
u
tio
n
u
n
its
.
C
o
n
s
eq
u
en
tly
,
th
e
im
p
lem
en
tatio
n
o
f
m
u
ltit
h
r
ea
d
i
n
g
is
o
f
ten
r
e
g
ar
d
e
d
as
o
n
e
o
f
th
e
m
o
s
t
ef
f
ec
tiv
e
s
o
lu
tio
n
s
f
o
r
im
p
r
o
v
i
n
g
a
p
p
licatio
n
p
er
f
o
r
m
an
ce
,
ac
ce
ler
atin
g
s
y
s
tem
r
esp
o
n
s
iv
en
ess
,
an
d
o
p
tim
izin
g
p
r
o
ce
s
s
o
r
r
eso
u
r
ce
u
t
ilizatio
n
.
Mu
lti
-
th
r
ea
d
in
g
f
u
n
d
a
m
en
tally
r
ef
er
s
to
a
p
r
o
ce
s
s
m
an
ag
e
m
en
t
ca
p
ab
ilit
y
th
at
en
ab
les
th
e
p
r
o
v
is
io
n
an
d
co
n
tr
o
l
o
f
m
u
ltip
le
co
n
c
u
r
r
en
t
ex
ec
u
tio
n
p
ath
s
with
in
a
s
in
g
le
p
r
o
ce
s
s
.
I
n
o
th
e
r
wo
r
d
s
,
a
s
in
g
le
p
r
o
ce
s
s
ca
n
co
n
tain
s
ev
er
al
t
h
r
ea
d
s
r
u
n
n
i
n
g
in
p
ar
allel,
allo
win
g
th
e
s
y
s
tem
to
en
h
an
ce
p
r
o
ce
s
s
o
r
u
tili
za
tio
n
ef
f
icien
cy
.
Un
lik
e
tr
ad
itio
n
al
ap
p
r
o
ac
h
es
wh
er
e
ea
ch
p
r
o
c
ess
co
n
tain
s
o
n
ly
o
n
e
ex
ec
u
t
io
n
th
r
ea
d
an
d
th
e
co
n
ce
p
t
o
f
th
r
ea
d
in
g
is
n
o
t
y
e
t
im
p
lem
en
ted
,
th
is
co
n
v
e
n
tio
n
al
m
o
d
el
is
r
ef
er
r
ed
to
as
s
in
g
le
th
r
ea
d
ed
.
B
o
th
p
r
o
ce
s
s
es a
r
e
illu
s
tr
ated
in
Fig
u
r
e
1
,
wh
ich
r
ep
r
esen
ts
th
e
th
r
ea
d
ed
p
r
o
ce
s
s
m
o
d
el.
T
h
e
lef
t
s
id
e
o
f
Fig
u
r
e
1
(
a
)
illu
s
tr
ates
th
e
s
in
g
le
th
r
ea
d
ed
m
o
d
el,
in
wh
ich
p
r
o
ce
s
s
m
an
ag
em
en
t
s
u
p
p
o
r
ts
o
n
ly
o
n
e
u
s
er
p
r
o
ce
s
s
with
a
s
in
g
le
th
r
ea
d
,
as
e
x
em
p
lifie
d
b
y
Mic
r
o
s
o
f
t
-
d
is
k
o
p
er
atin
g
s
y
s
tem
(
MS
-
DOS
)
.
C
o
n
v
e
r
s
ely
,
th
e
m
u
lti
-
th
r
ea
d
ed
ap
p
r
o
ac
h
,
illu
s
tr
ated
o
n
th
e
r
ig
h
t
s
id
e
o
f
Fig
u
r
e
1
(
b
)
,
h
as
b
ee
n
wid
ely
ad
o
p
ted
b
y
m
o
d
er
n
p
r
o
ce
s
s
m
an
ag
em
en
t
s
y
s
tem
s
th
at
in
teg
r
ate
m
u
ltit
h
r
ea
d
i
n
g
ca
p
ab
i
liti
es
to
im
p
r
o
v
e
p
er
f
o
r
m
an
c
e,
ef
f
icien
cy
,
an
d
o
v
er
all
s
y
s
tem
r
esp
o
n
s
iv
en
ess
[
3
5
]
–
[
3
7
]
.
I
n
th
e
m
u
lti
-
th
r
ea
d
ed
m
o
d
el,
a
p
r
o
ce
s
s
is
r
eg
ar
d
ed
as
th
e
p
r
im
ar
y
u
n
it o
f
r
eso
u
r
ce
s
r
esp
o
n
s
ib
le
f
o
r
m
an
a
g
in
g
a
n
d
co
o
r
d
i
n
atin
g
th
e
ex
ec
u
tio
n
o
f
m
u
ltip
le
th
r
ea
d
s
with
in
it.
(
a)
(
b
)
Fig
u
r
e
1
.
T
h
r
ea
d
e
d
p
r
o
ce
s
s
m
o
d
el
f
o
r
(
a)
s
in
g
le
t
h
r
ea
d
e
d
an
d
(
b
)
m
u
lti
-
th
r
ea
d
ed
I
n
th
is
r
esear
ch
,
th
r
ea
d
s
ar
e
u
tili
ze
d
to
ex
ec
u
te
th
e
f
ac
e
-
p
atter
n
r
ec
o
g
n
itio
n
p
r
o
ce
s
s
in
p
ar
allel.
I
n
p
r
ev
io
u
s
r
esear
ch
,
t
h
e
f
ac
ial
p
atter
n
r
ec
o
g
n
itio
n
p
r
o
ce
s
s
em
p
lo
y
ed
an
en
co
d
in
g
m
ec
h
a
n
is
m
,
wh
er
e
f
ac
ial
p
atter
n
s
co
n
s
is
tin
g
o
f
m
u
ltip
le
o
r
ien
tatio
n
s
s
u
ch
as
f
r
o
n
t,
r
ig
h
t,
lef
t,
to
p
,
an
d
b
o
tto
m
we
r
e
ass
ig
n
ed
s
p
ec
if
i
c
r
ec
o
g
n
itio
n
co
d
es.
T
h
e
p
u
r
p
o
s
e
o
f
th
is
m
ec
h
an
is
m
i
s
to
e
n
ab
le
th
e
f
ac
e
r
ec
o
g
n
itio
n
alg
o
r
ith
m
(
FR
A)
to
d
eter
m
in
e
wh
ic
h
f
ac
ial
o
r
ien
t
atio
n
s
er
v
es
as
th
e
in
p
u
t.
T
h
r
o
u
g
h
th
is
ap
p
r
o
ac
h
,
th
e
d
etec
ted
co
d
e
d
ir
ec
ts
th
e
s
y
s
tem
to
id
en
tify
th
e
in
p
u
t
d
ata
ac
co
r
d
in
g
to
its
co
r
r
esp
o
n
d
in
g
f
a
cial
-
p
atter
n
g
r
o
u
p
.
T
h
e
ad
v
an
ta
g
e
o
f
th
is
m
ec
h
an
is
m
lies
in
its
ab
ilit
y
t
o
r
ed
u
ce
th
e
ex
ec
u
tio
n
tim
e
o
f
th
e
r
ec
o
g
n
itio
n
p
r
o
ce
s
s
,
as
o
n
ly
a
s
p
ec
if
ic
s
u
b
s
et
o
f
f
ac
ial
p
atter
n
s
is
p
r
o
ce
s
s
ed
r
ath
er
t
h
an
th
e
en
tire
d
ataset
f
r
o
m
in
d
ex
1
t
o
.
Ho
wev
er
,
a
m
ajo
r
c
h
allen
g
e
ar
is
es
wh
en
th
e
s
y
s
tem
f
ails
to
ac
cu
r
ately
d
eter
m
in
e
t
h
e
p
r
ec
is
e
f
ac
ial
r
eg
io
n
,
lead
i
n
g
to
o
v
e
r
lap
p
in
g
r
ec
o
g
n
itio
n
b
etwe
en
ad
jace
n
t
f
ac
ial
o
r
ien
tatio
n
s
.
Fo
r
in
s
tan
c
e,
a
f
r
o
n
t
f
ac
e
o
r
ie
n
tatio
n
with
an
an
g
le
b
etwe
en
30
°
an
d
4
5
°
m
ay
s
till
b
e
r
ec
o
g
n
ized
as a
f
r
o
n
t f
ac
e,
as illu
s
tr
ated
in
Fig
u
r
e
2
.
Fig
u
r
e
2
illu
s
tr
ates
two
f
ac
ial
im
ag
e
o
r
ien
tatio
n
s
,
n
am
ely
th
e
f
r
o
n
t
a
n
d
lef
t
f
ac
e
p
o
s
itio
n
s
.
T
h
e
f
r
o
n
t
o
r
ien
tatio
n
is
p
r
esen
ted
in
Fig
u
r
es
2
(
a)
an
d
2
(
c)
with
an
an
g
le
o
f
0
°
,
wh
ile
th
e
lef
t
f
ac
e
o
r
ien
tatio
n
is
s
h
o
wn
in
Fig
u
r
es
2
(
b
)
an
d
2
(
d
)
with
an
ap
p
r
o
x
im
ate
an
g
le
o
f
4
0
°
.
Fig
u
r
es
2
(
a)
an
d
2
(
b
)
r
ep
r
esen
t
th
e
o
r
ig
in
al
f
a
cial
im
ag
es,
wh
er
ea
s
Fig
u
r
es
2
(
c)
an
d
2
(
d
)
d
is
p
lay
th
e
f
ac
ial
im
ag
es
ex
tr
ac
ted
f
r
o
m
th
e
r
e
g
io
n
d
esig
n
ated
as
th
e
R
OI
.
I
d
ea
lly
,
in
ca
s
e
Fig
u
r
e
2
(
d
)
,
th
e
f
ac
e
s
h
o
u
ld
alr
ea
d
y
b
e
ca
teg
o
r
ize
d
with
in
th
e
lef
t
f
ac
e
r
eg
io
n
.
Nev
er
th
eless
,
in
p
r
ac
tice,
th
e
f
r
o
n
t
-
f
ac
e
class
if
ier
is
s
till
c
ap
ab
le
o
f
d
etec
tin
g
a
n
d
r
ec
o
g
n
izin
g
th
e
p
atter
n
ac
cu
r
ately
with
in
th
e
f
r
o
n
tal
ca
teg
o
r
y
.
T
h
is
in
d
icate
s
th
e
p
r
esen
ce
o
f
o
v
er
la
p
p
in
g
ar
ea
s
b
etwe
en
f
ac
ial
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
1
5
,
No
.
4
,
Au
g
u
s
t 2
0
2
6
:
3
9
0
3
-
3
9
1
8
3906
o
r
ien
tatio
n
s
.
T
h
e
m
o
s
t
ef
f
ec
ti
v
e
ap
p
r
o
ac
h
to
en
s
u
r
e
co
m
p
r
e
h
en
s
iv
e
r
ec
o
g
n
itio
n
o
f
all
f
ac
i
al
o
r
ien
tatio
n
s
is
to
p
er
f
o
r
m
a
c
o
m
p
lete
s
ca
n
o
f
th
e
f
ac
ial
a
r
ea
.
Ho
wev
e
r
,
t
h
is
wo
u
ld
ca
u
s
e
th
e
s
y
s
tem
to
lo
s
e
r
ea
l
-
tim
e
p
er
f
o
r
m
an
ce
a
n
d
s
ig
n
if
ican
tl
y
in
cr
ea
s
e
r
ec
o
g
n
itio
n
tim
e,
th
er
eb
y
af
f
ec
tin
g
th
e
o
v
er
al
l
ef
f
icien
cy
o
f
th
e
h
ar
d
war
e
in
p
er
f
o
r
m
in
g
f
ac
e
-
p
atter
n
r
ec
o
g
n
itio
n
.
Fig
u
r
e
3
p
r
esen
ts
th
e
p
r
o
p
o
s
ed
a
r
ch
itectu
r
e
o
f
th
e
m
u
lti
-
th
r
ea
d
ed
co
n
tr
o
ller
d
ev
e
lo
p
ed
to
a
d
d
r
ess
th
is
is
s
u
e.
(
a)
(
b
)
(
c)
(
d
)
Fig
u
r
e
2
.
Ov
e
r
lap
p
in
g
f
ac
ial
i
m
ag
e
r
ec
o
g
n
itio
n
b
etwe
en
f
r
o
n
t a
n
d
lef
t
v
iew
:
(
a)
o
r
ig
in
al
i
m
ag
e
(0
°
)
,
(
b
)
o
r
ig
in
al
im
ag
e
(
4
0
°
)
,
(
c)
R
OI
(
0
°
)
,
an
d
(
d
)
R
OI
(
4
0
°
)
I
n
p
u
t
fa
c
e
p
e
t
t
e
r
n
E
v
e
n
t
L
o
o
p
O
u
t
p
u
t
R
e
s
u
lt
(
W
in
n
e
r
C
la
s
s
)
T
h
r
e
a
d
s
U
n
b
lo
c
k
in
g
T
a
s
k
Q
u
e
u
e
M
a
in
T
h
r
e
a
d
Fig
u
r
e
3
.
Pro
p
o
s
ed
m
u
lti
-
th
r
ea
d
ed
co
n
tr
o
ller
ar
c
h
itectu
r
e
T
h
is
r
esear
ch
d
elin
ea
tes
th
e
f
u
n
ctio
n
al
ar
ch
itectu
r
e
o
f
th
e
m
u
lti
-
th
r
ea
d
co
n
t
r
o
ller
in
to
th
r
ee
p
r
im
ar
y
co
m
p
o
n
en
ts
:
d
ata
ex
tr
ac
tio
n
,
d
ata
an
aly
s
is
,
an
d
d
ata
in
teg
r
atio
n
,
as
illu
s
tr
ated
in
Fig
u
r
e
3
.
I
n
t
h
e
d
ata
ex
tr
ac
tio
n
s
tag
e,
th
e
m
ain
th
r
ea
d
is
r
esp
o
n
s
ib
le
f
o
r
m
an
ag
i
n
g
d
u
al
d
ata
s
tr
ea
m
s
o
r
i
g
in
at
in
g
f
r
o
m
th
e
i
n
p
u
t
f
ac
e
p
atter
n
s
,
wh
ich
a
r
e
s
u
b
s
e
q
u
en
tly
p
lace
d
in
to
a
task
q
u
e
u
e.
T
h
ese
d
ata
s
tr
ea
m
s
co
n
s
is
t
o
f
ex
tr
ac
ted
f
ac
ial
f
ea
tu
r
es
th
at
s
er
v
e
as
id
en
tifi
er
s
,
wh
er
e
f
ac
ial
ar
ea
in
f
o
r
m
atio
n
is
r
ep
r
esen
te
d
in
a
b
in
a
r
y
f
o
r
m
at
to
en
ab
l
e
f
u
r
th
er
p
r
o
ce
s
s
in
g
.
B
y
im
p
le
m
en
tin
g
a
m
u
lti
-
th
r
ea
d
ed
c
o
n
tr
o
ller
ar
ch
itectu
r
e
,
th
e
d
ata
ex
ec
u
tio
n
p
r
o
ce
s
s
b
ec
o
m
es
m
o
r
e
ef
f
icien
t,
as
co
m
p
u
tatio
n
al
w
o
r
k
lo
a
d
s
ca
n
b
e
d
is
tr
ib
u
te
d
an
d
ex
ec
u
te
d
in
p
ar
allel.
T
h
is
ap
p
r
o
ac
h
co
n
tr
asts
with
th
e
s
in
g
le
-
th
r
ea
d
m
o
d
el,
wh
ich
r
e
q
u
ir
es
lo
n
g
er
ex
ec
u
tio
n
tim
e
d
u
e
to
its
in
h
er
en
t
lim
itatio
n
s
in
h
an
d
lin
g
m
u
ltip
l
e
d
ata
s
tr
ea
m
s
s
im
u
ltan
eo
u
s
ly
.
I
n
th
e
d
ata
an
aly
s
is
s
tag
e,
ea
c
h
th
r
ea
d
f
u
n
ctio
n
s
as
a
R
AM
d
is
cr
im
in
ato
r
th
at
ac
tiv
ely
r
etr
iev
es
d
ata
f
r
o
m
th
e
task
q
u
eu
e
f
o
r
an
aly
s
is
.
T
h
e
task
q
u
eu
e
s
er
v
es
t
o
o
r
g
an
ize
an
d
d
is
p
lay
th
e
s
eq
u
en
ce
o
f
r
esu
lts
o
b
tain
ed
f
r
o
m
t
h
e
p
atter
n
r
ec
o
g
n
itio
n
p
r
o
ce
s
s
with
in
th
e
R
AM
d
is
cr
im
in
ato
r
s
an
d
to
d
et
er
m
in
e
th
e
win
n
er
class
p
atter
n
.
On
ce
th
e
s
im
ilar
ity
ac
cu
r
ac
y
o
f
th
e
r
ec
o
g
n
ized
p
atter
n
is
o
b
tain
e
d
,
th
e
r
esu
lt is
tr
an
s
m
itted
b
ac
k
to
th
e
r
esu
lt
q
u
e
u
e
as
th
e
win
n
er
class
.
E
ac
h
th
r
ea
d
r
ep
r
esen
ts
a
p
atter
n
f
u
n
ctio
n
th
at
o
p
e
r
ates
in
d
ep
en
d
en
tly
wh
en
ac
tiv
ated
,
th
e
r
eb
y
e
n
h
an
cin
g
th
e
co
n
tr
o
ller
’
s
p
e
r
f
o
r
m
a
n
ce
in
r
ea
d
i
n
g
an
d
p
r
o
ce
s
s
in
g
d
ata
s
tr
ea
m
s
.
W
ith
in
th
e
m
u
lti
-
th
r
ea
d
ed
ar
ch
itectu
r
e
,
th
e
u
s
e
o
f
th
e
W
NN
m
eth
o
d
,
wh
ich
u
tili
ze
s
R
AM
as
th
e
p
r
im
ar
y
d
ata
-
p
r
o
ce
s
s
in
g
m
ed
i
u
m
,
s
im
p
lifie
s
th
e
co
m
p
u
tatio
n
al
p
r
o
ce
s
s
o
f
f
ac
e
-
p
atter
n
r
e
co
g
n
itio
n
.
T
h
e
d
ata
in
teg
r
atio
n
s
tag
e
f
o
c
u
s
es
o
n
t
h
e
r
o
le
o
f
th
e
m
ain
th
r
ea
d
,
w
h
ich
is
r
esp
o
n
s
ib
le
f
o
r
m
an
ag
in
g
,
p
r
o
ce
s
s
in
g
,
an
d
in
teg
r
atin
g
d
ata
o
r
ig
in
atin
g
f
r
o
m
th
e
v
ar
io
u
s
th
r
ea
d
s
with
in
th
e
s
y
s
tem
.
T
h
e
in
teg
r
atio
n
p
r
o
ce
s
s
is
ca
r
r
ied
o
u
t
th
r
o
u
g
h
a
b
lo
ck
i
n
g
f
u
n
ctio
n
m
ec
h
an
is
m
,
wh
ich
allo
ws
th
e
s
y
s
tem
to
wait
u
n
til
all
r
e
q
u
ir
ed
d
ata
a
r
e
av
ailab
le
b
ef
o
r
e
m
e
r
g
in
g
th
em
.
T
h
is
ap
p
r
o
ac
h
en
h
a
n
ce
s
th
e
q
u
ali
ty
o
f
d
ata
p
r
o
ce
s
s
in
g
r
esu
lts
b
y
en
s
u
r
i
n
g
th
a
t
in
teg
r
atio
n
o
cc
u
r
s
in
a
c
o
n
tr
o
lled
an
d
ac
cu
r
ate
m
an
n
e
r
,
th
e
r
eb
y
p
r
o
d
u
cin
g
c
o
n
s
is
ten
t
an
d
co
m
p
lete
o
u
tp
u
ts
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
A
n
o
ve
l e
mb
e
d
d
ed
a
p
p
r
o
a
ch
t
o
fa
ce
r
ec
o
g
n
itio
n
u
s
in
g
m
u
lti
-
th
r
ea
d
ed
c
o
n
tr
o
ller
b
a
s
ed
…
(
A
h
ma
d
Za
r
ka
s
i
)
3907
Fu
r
th
er
m
o
r
e
,
it
p
r
o
v
i
d
es
g
r
ea
t
er
f
lex
ib
ilit
y
i
n
d
eter
m
in
in
g
t
h
e
f
o
r
m
at,
s
tr
u
ctu
r
e,
a
n
d
p
r
esen
tatio
n
s
tr
ateg
y
o
f
th
e
d
ata,
wh
eth
er
f
o
r
v
is
u
aliza
tio
n
p
u
r
p
o
s
es o
r
f
o
r
tr
a
n
s
m
is
s
i
o
n
to
o
t
h
er
s
y
s
tem
s
.
2
.
2
.
Weig
htle
s
s
neura
l net
wo
rk
f
a
ce
re
co
g
nitio
n a
lg
o
rit
h
m
des
ig
n
T
h
e
lear
n
i
n
g
p
r
o
ce
s
s
in
th
is
s
tu
d
y
was
co
n
d
u
cted
b
y
d
e
v
e
lo
p
in
g
a
W
NN
-
FR
A
th
at
im
p
lem
en
ts
a
ce
n
ter
-
s
ca
n
p
atter
n
r
ec
o
g
n
itio
n
m
o
d
el.
T
h
e
W
NN
-
FR
A
m
et
h
o
d
is
d
esig
n
ed
to
p
r
o
ce
s
s
f
ea
tu
r
e
-
ex
tr
ac
tio
n
d
ata
o
b
tain
ed
f
r
o
m
th
e
f
ac
e
d
etec
tio
n
s
tag
e.
T
h
ese
d
ata
a
r
e
r
e
p
r
esen
ted
as
f
a
cial
p
atter
n
s
c
ap
tu
r
ed
f
r
o
m
th
r
ee
im
ag
e
o
r
ien
tatio
n
s
:
f
r
o
n
t,
lef
t,
an
d
r
ig
h
t.
I
n
th
e
W
NN
-
FR
A
m
eth
o
d
,
th
e
lea
r
n
in
g
p
r
o
ce
s
s
is
p
er
f
o
r
m
e
d
u
s
in
g
R
AM
,
wh
er
e
th
e
f
ac
ial
p
atter
n
d
ata
ar
e
r
ep
r
esen
ted
in
a
b
in
ar
y
f
o
r
m
at.
T
h
is
b
in
ar
y
r
ep
r
ese
n
tatio
n
is
ch
o
s
en
to
alig
n
with
th
e
f
u
n
d
am
e
n
tal
p
r
in
cip
les
o
f
t
h
e
W
NN,
wh
ich
r
elies
o
n
a
s
im
p
le
m
em
o
r
y
s
tr
u
ctu
r
e
y
et
p
o
s
s
ess
es
th
e
ca
p
ab
ilit
y
to
p
er
f
o
r
m
e
f
f
icien
t
p
atter
n
class
if
icatio
n
.
C
o
n
s
eq
u
en
tly
,
ea
ch
f
ac
ial
p
atter
n
o
b
tain
e
d
f
r
o
m
f
ea
tu
r
e
e
x
tr
ac
tio
n
ca
n
b
e
s
y
s
tem
atica
lly
m
ap
p
e
d
in
to
th
e
W
NN
m
em
o
r
y
f
o
r
b
o
th
t
r
ain
in
g
an
d
test
in
g
p
u
r
p
o
s
es.
I
n
th
e
tr
ain
in
g
p
h
ase
,
ea
c
h
R
AM
d
is
cr
im
in
ato
r
m
u
s
t
f
ir
s
t
b
e
in
itialized
b
y
s
ettin
g
all
m
em
o
r
y
lo
ca
tio
n
s
in
th
e
R
AM
to
a
lo
g
ical
s
tate
o
f
0
.
T
h
is
in
itializatio
n
p
r
o
ce
s
s
en
s
u
r
es
th
at
n
o
d
ef
au
lt
v
alu
es
r
em
ai
n
in
m
em
o
r
y
th
at
co
u
ld
af
f
ec
t
th
e
lear
n
in
g
o
u
tc
o
m
e.
On
ce
th
e
in
itializatio
n
s
tag
e
is
co
m
p
lete,
th
e
s
y
s
tem
s
elec
ts
a
tr
ain
in
g
p
atter
n
r
ep
r
esen
ted
b
y
a
b
i
n
ar
y
v
ec
to
r
o
f
(
·
)
b
its
.
T
h
is
b
in
ar
y
r
e
p
r
esen
tatio
n
s
er
v
es
as
th
e
in
p
u
t
ad
d
r
ess
u
s
ed
to
ac
tiv
ate
s
p
ec
if
ic
m
em
o
r
y
lo
ca
ti
o
n
s
with
in
th
e
R
AM
.
Fo
r
ea
ch
tr
ain
in
g
p
atter
n
p
r
o
v
id
e
d
,
a
lo
g
ical
v
alu
e
o
f
1
i
s
s
to
r
ed
in
th
e
R
AM
m
em
o
r
y
lo
ca
tio
n
co
r
r
esp
o
n
d
in
g
to
th
e
a
d
d
r
ess
d
eter
m
in
ed
b
y
th
e
in
p
u
t
p
atter
n
.
T
h
is
p
r
o
ce
d
u
r
e
is
r
e
p
ea
ted
iter
ativ
el
y
f
o
r
all
a
v
ailab
le
tr
ain
in
g
p
a
tter
n
s
,
allo
win
g
th
e
R
AM
d
is
cr
im
in
ato
r
to
g
r
a
d
u
ally
co
n
s
tr
u
ct
a
m
em
o
r
y
m
a
p
th
at
r
e
p
r
esen
ts
th
e
r
elatio
n
s
h
ip
b
etwe
en
in
p
u
t
p
atter
n
s
an
d
th
eir
co
r
r
esp
o
n
d
i
n
g
m
em
o
r
y
a
d
d
r
ess
es.
Af
ter
th
e
en
tire
tr
ain
in
g
p
r
o
ce
s
s
is
co
m
p
leted
,
th
e
c
o
n
te
n
ts
o
f
th
e
R
AM
m
e
m
o
r
y
co
n
s
is
t
o
f
a
co
m
b
in
atio
n
o
f
l
o
g
ical
v
alu
es
0
an
d
1
.
A
lo
g
ical
v
alu
e
o
f
1
i
n
d
icate
s
th
at
th
e
co
r
r
esp
o
n
d
in
g
m
em
o
r
y
lo
ca
tio
n
h
as
b
ee
n
ad
d
r
ess
ed
b
y
o
n
e
o
f
th
e
tr
ain
in
g
p
atter
n
s
,
wh
er
ea
s
a
lo
g
ical
v
alu
e
o
f
0
s
ig
n
if
ie
s
th
at
th
e
lo
ca
tio
n
r
em
ain
s
u
n
u
s
ed
.
At
th
is
s
tag
e
,
th
e
R
AM
m
em
o
r
y
ca
n
b
e
r
e
g
ar
d
ed
as
h
a
v
in
g
s
to
r
e
d
th
e
i
n
itial
d
ata
p
atter
n
s
r
eq
u
ir
ed
f
o
r
th
e
s
u
b
s
eq
u
en
t
p
a
tter
n
r
ec
o
g
n
itio
n
p
h
ase.
Nex
t,
th
e
o
u
tp
u
ts
f
r
o
m
ea
ch
R
AM
n
o
d
e
ar
e
ag
g
r
eg
ated
b
y
a
s
u
m
m
atio
n
d
ev
ice
(
)
.
T
h
is
co
m
p
o
n
en
t
f
u
n
ctio
n
s
to
s
u
m
all
o
u
tp
u
ts
f
r
o
m
th
e
R
AM
n
o
d
es
with
in
a
s
in
g
le
R
AM
d
is
cr
im
in
ato
r
,
an
d
th
e
r
esu
ltin
g
to
tal
is
th
e
n
d
i
v
id
ed
b
y
th
e
o
v
er
all
n
u
m
b
er
o
f
R
AM
n
o
d
es
(
K)
.
C
o
n
s
eq
u
en
tly
,
t
h
e
o
b
tain
ed
o
u
tp
u
t
v
alu
e
r
ef
lects
th
e
p
r
o
p
o
r
tio
n
o
f
R
AM
n
o
d
es
th
at
p
r
o
d
u
ce
a
lo
g
ical
1
f
o
r
a
g
iv
en
in
p
u
t
v
ec
to
r
X.
T
h
is
d
at
a
s
er
v
es
as
a
m
ea
s
u
r
e
o
f
th
e
s
im
ilar
ity
b
etwe
en
t
h
e
test
ed
in
p
u
t
p
atter
n
an
d
th
e
r
ef
er
en
ce
p
atter
n
s
s
to
r
ed
in
m
em
o
r
y
d
u
r
in
g
th
e
tr
ai
n
in
g
p
r
o
ce
s
s
.
T
h
r
o
u
g
h
th
is
m
e
ch
an
is
m
,
th
e
R
AM
d
is
cr
im
in
ato
r
is
ca
p
a
b
le
o
f
d
is
tin
g
u
is
h
in
g
in
p
u
t
p
atter
n
s
b
a
s
ed
o
n
th
e
d
is
tr
ib
u
tio
n
o
f
lo
g
i
ca
l
1
v
alu
e
f
o
r
m
e
d
d
u
r
in
g
tr
ain
in
g
.
T
h
e
co
m
p
lete
ar
ch
itectu
r
e
o
f
th
e
W
NN
-
FR
A
m
eth
o
d
d
ev
elo
p
ed
in
th
is
r
esear
ch
is
illu
s
tr
ated
in
Fig
u
r
e
4
.
T
h
e
ar
c
h
itectu
r
e
d
em
o
n
s
tr
ates
th
e
f
lo
w
o
f
f
ac
i
al
p
atter
n
d
ata
f
r
o
m
th
e
f
ea
t
u
r
e
ex
tr
ac
tio
n
s
tag
e
th
r
o
u
g
h
t
h
e
tr
ai
n
in
g
p
h
ase,
d
ata
s
to
r
ag
e
in
m
em
o
r
y
,
an
d
f
in
ally
to
th
e
f
ac
e
r
ec
o
g
n
itio
n
p
r
o
ce
s
s
.
T
h
e
s
eq
u
en
tial
s
tag
es
o
f
th
e
W
N
N
-
FR
A
lear
n
in
g
p
r
o
ce
s
s
ar
e
as
f
o
llo
ws
:
i)
d
eter
m
in
e
th
e
n
u
m
b
er
o
f
in
p
u
t
v
ec
to
r
s
o
b
tain
ed
f
r
o
m
f
ac
ial
f
ea
tu
r
e
e
x
tr
ac
tio
n
,
b
o
t
h
th
o
s
e
s
to
r
ed
in
th
e
d
atab
ase
an
d
th
o
s
e
u
s
ed
as
p
r
im
ar
y
in
p
u
ts
;
ii)
d
iv
id
e
th
e
d
atab
ase
in
p
u
t
v
ec
to
r
s
in
to
th
r
ee
m
ain
g
r
o
u
p
s
r
ig
h
t,
f
r
o
n
t,
an
d
lef
t
;
iii)
p
r
o
ce
s
s
th
ese
d
ata
with
in
s
ev
er
al
d
is
cr
im
in
ato
r
class
es
u
n
d
er
th
e
m
u
ltit
h
r
ea
d
in
g
m
ec
h
an
is
m
;
iv
)
p
er
f
o
r
m
s
im
u
ltan
eo
u
s
r
ec
o
g
n
itio
n
o
f
th
e
d
ata
ac
r
o
s
s
ea
ch
d
is
cr
im
in
ato
r
class
u
s
in
g
th
e
th
r
ea
d
-
b
ased
m
ec
h
an
is
m
;
an
d
v
)
d
eter
m
in
e
th
e
o
p
t
im
al
v
alu
e
f
r
o
m
th
e
d
is
cr
im
in
ato
r
class
es
to
id
en
tify
th
e
win
n
er
class
.
T
h
is
ar
ch
itectu
r
e
p
r
o
v
id
es
a
clea
r
o
p
er
atio
n
al
f
r
am
ewo
r
k
illu
s
tr
atin
g
h
o
w
f
ac
ial
d
ata
ar
e
p
r
o
ce
s
s
ed
an
d
class
if
ied
ef
f
icien
tly
th
r
o
u
g
h
p
ar
allelize
d
lear
n
in
g
an
d
r
ec
o
g
n
itio
n
m
ec
h
an
is
m
s
b
ased
o
n
th
e
W
NN
-
FR
A.
Fig
u
r
e
4
illu
s
tr
ates
th
at
th
e
l
ea
r
n
in
g
p
r
o
ce
s
s
co
n
s
is
ts
o
f
t
wo
m
ain
co
m
p
o
n
en
ts
:
r
ef
er
e
n
ce
p
atter
n
s
an
d
in
p
u
t
p
atter
n
s
.
T
h
e
r
ef
e
r
en
ce
p
atter
n
s
r
ep
r
esen
t
f
ac
ial
im
ag
es
th
at
h
av
e
u
n
d
e
r
g
o
n
e
f
ea
tu
r
e
ex
tr
ac
tio
n
,
b
eg
in
n
in
g
with
th
e
ac
q
u
is
itio
n
o
f
r
ed
,
g
r
ee
n
,
b
lu
e
(
R
GB
)
f
ac
ial
d
ata
an
d
f
o
llo
wed
b
y
th
e
f
a
ce
d
etec
tio
n
s
tag
e.
On
ce
th
e
f
ac
e
is
d
etec
ted
,
a
f
e
atu
r
e
d
etec
tio
n
p
r
o
ce
s
s
is
ca
r
r
ied
o
u
t
to
id
en
tif
y
k
ey
f
ac
ial
c
o
m
p
o
n
en
ts
s
u
ch
as
th
e
ey
es,
n
o
s
e,
a
n
d
m
o
u
th
.
T
h
is
p
r
o
ce
s
s
p
r
o
d
u
ce
s
a
n
ew
im
ag
e
th
at
ca
p
tu
r
es
o
n
ly
a
s
p
ec
if
ic
f
ac
ial
p
ix
e
l
r
eg
io
n
,
f
o
r
i
n
s
tan
ce
,
o
f
s
ize
n
×
m
p
ix
els.
T
h
e
r
esu
ltin
g
f
ea
t
u
r
e
im
ag
e
is
th
en
co
n
v
er
ted
in
to
a
g
r
ay
s
ca
le
im
ag
e
an
d
s
u
b
s
eq
u
e
n
tly
in
to
a
b
in
ar
y
im
ag
e.
T
h
e
b
in
ar
y
im
ag
e
is
n
o
r
m
alize
d
in
t
o
d
ata
v
al
u
es
o
f
0
an
d
1
.
Af
ter
t
h
e
f
ac
ial
f
ea
tu
r
e
p
atter
n
s
ar
e
f
o
r
m
ed
,
th
ese
b
in
ar
y
p
atter
n
s
s
er
v
e
as
in
p
u
t
v
ec
to
r
s
f
o
r
t
h
e
W
NN
m
eth
o
d
an
d
ar
e
s
to
r
ed
ac
r
o
s
s
2
0
R
AM
n
o
d
es
d
is
tr
ib
u
ted
with
in
th
r
ee
R
AM
d
is
cr
im
in
ato
r
s
.
T
o
en
ab
le
s
im
u
ltan
eo
u
s
p
r
o
ce
s
s
in
g
o
f
in
p
u
t
-
p
atter
n
r
ec
o
g
n
itio
n
,
t
h
e
th
r
ee
R
AM
d
is
cr
im
in
ato
r
s
ar
e
co
n
f
i
g
u
r
e
d
to
f
u
n
cti
o
n
as
a
m
u
lti
-
th
r
ea
d
c
o
n
tr
o
ller
(
T
h
r
ea
d
1
,
T
h
r
ea
d
2
,
an
d
T
h
r
ea
d
3
)
u
n
d
er
th
e
m
ain
th
r
ea
d
b
lo
c
k
.
I
n
th
e
in
p
u
t
p
atter
n
s
tag
e,
th
e
p
r
o
ce
s
s
is
s
im
ilar
to
th
at
o
f
th
e
r
ef
er
en
ce
p
atter
n
th
e
k
ey
d
if
f
er
en
ce
lies
in
th
e
f
ac
t
th
at
th
e
r
ef
er
e
n
ce
p
atter
n
s
ar
e
s
to
r
ed
with
in
th
e
d
ataset,
wh
er
ea
s
th
e
in
p
u
t
p
atter
n
s
er
v
es
as
th
e
p
atter
n
to
b
e
r
ec
o
g
n
ize
d
d
u
r
i
n
g
th
e
lear
n
in
g
p
r
o
ce
s
s
.
T
h
e
in
p
u
t
R
GB
p
atter
n
,
r
ep
r
esen
tin
g
f
ac
ial
f
ea
tu
r
es,
u
n
d
er
g
o
es
a
s
er
ies o
f
e
x
tr
ac
tio
n
s
tag
es
u
n
til
it
is
c
o
n
v
e
r
ted
in
to
b
in
ar
y
d
ata.
T
h
e
r
esu
ltin
g
b
in
ar
y
d
ata
tak
e
th
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
1
5
,
No
.
4
,
Au
g
u
s
t 2
0
2
6
:
3
9
0
3
-
3
9
1
8
3908
f
o
r
m
o
f
a
m
atr
i
x
co
r
r
esp
o
n
d
i
n
g
to
th
e
p
ix
el
d
im
e
n
s
io
n
s
o
f
th
e
in
p
u
t
f
ac
ial
p
atter
n
.
On
ce
th
e
in
p
u
t
p
atter
n
is
p
r
ep
ar
e
d
,
th
e
d
ata
ar
e
f
o
r
war
d
ed
to
th
e
r
ec
o
g
n
itio
n
p
r
o
ce
s
s
with
in
th
e
c
o
m
p
ar
ato
r
b
lo
ck
,
wh
er
e
th
e
p
atter
n
is
co
m
p
ar
ed
s
im
u
ltan
eo
u
s
ly
ac
r
o
s
s
all
th
r
ea
d
s
.
E
ac
h
th
r
ea
d
p
er
f
o
r
m
s
2
0
r
ec
o
g
n
itio
n
iter
a
tio
n
s
,
allo
win
g
th
e
s
y
s
tem
to
o
b
tain
b
o
th
th
e
p
atter
n
ac
cu
r
ac
y
lev
el
an
d
th
e
e
x
ec
u
tio
n
tim
e
p
er
f
o
r
m
a
n
ce
f
o
r
ea
ch
f
ac
ial
p
atter
n
.
Sin
ce
th
e
in
p
u
t
p
atter
n
im
ag
e
d
ata
ar
e
ca
p
tu
r
ed
with
in
a
d
y
n
am
ically
m
o
v
in
g
s
y
s
tem
,
th
e
n
u
m
b
er
o
f
p
ix
els
o
b
tain
ed
v
ar
ies
ac
r
o
s
s
s
am
p
les.
T
h
is
v
ar
iatio
n
af
f
ec
t
s
th
e
p
atter
n
r
ec
o
g
n
itio
n
ac
c
u
r
ac
y
,
as
th
e
p
ix
el
d
im
en
s
io
n
s
o
f
th
e
r
e
f
er
en
ce
p
atter
n
d
if
f
er
f
r
o
m
th
o
s
e
o
f
th
e
in
p
u
t
p
atter
n
.
T
o
ad
d
r
ess
th
is
is
s
u
e,
th
e
p
r
esen
t
s
tu
d
y
em
p
lo
y
s
a
ce
n
ter
-
s
ca
n
r
ec
o
g
n
itio
n
m
o
d
el
.
Fig
u
r
e
5
ill
u
s
tr
ates
th
e
d
esig
n
o
f
th
e
ce
n
ter
-
s
ca
n
m
o
d
el.
T
h
e
ce
n
ter
-
s
ca
n
r
ec
o
g
n
itio
n
al
g
o
r
ith
m
is
d
esig
n
ed
to
im
p
r
o
v
e
th
e
ac
cu
r
ac
y
lev
el
ac
h
iev
e
d
b
y
t
h
e
p
r
e
v
io
u
s
alg
o
r
ith
m
.
T
h
is
ap
p
r
o
ac
h
e
m
p
h
asizes
ac
h
iev
in
g
b
alan
ce
b
et
wee
n
th
e
r
ef
e
r
en
ce
p
atter
n
an
d
th
e
in
p
u
t
p
atter
n
,
p
ar
ticu
lar
ly
wh
e
n
d
is
cr
ep
an
ci
es
ar
is
e
d
u
e
to
d
if
f
er
en
ce
s
in
m
em
o
r
y
ca
p
ac
ity
o
r
im
ag
e
r
eso
lu
tio
n
b
etwe
en
th
e
r
ef
er
en
ce
p
atter
n
(
d
ataset)
an
d
th
e
in
p
u
t
p
atter
n
.
Su
ch
d
is
cr
ep
an
cies
o
cc
u
r
b
ec
au
s
e
th
e
in
p
u
t
p
atter
n
is
ac
q
u
ir
ed
f
r
o
m
d
y
n
a
m
ic
f
ac
ia
l
p
o
s
itio
n
s
,
wh
ic
h
m
a
y
r
esu
lt
in
s
m
aller
o
r
lar
g
er
im
ag
e
d
im
en
s
io
n
s
.
T
h
e
ce
n
ter
-
s
ca
n
m
et
h
o
d
m
in
im
iz
es
th
e
d
if
f
er
en
ce
in
m
em
o
r
y
ca
p
ac
ity
o
b
s
er
v
ed
in
p
r
ev
io
u
s
alg
o
r
ith
m
s
b
y
lo
ca
tin
g
th
e
ce
n
t
r
al
p
o
in
t
d
u
r
i
n
g
th
e
in
itial c
h
ec
k
in
g
p
r
o
ce
s
s
.
R
e
f
e
r
e
n
c
e
P
e
tt
e
r
n
s
(3
0
s
a
m
p
l
e
s
)
P
r
i
m
a
r
y
f
a
c
i
a
l
e
x
t
r
a
c
ti
o
n
d
a
t
a
(I
n
p
u
t
P
e
t
te
r
n
)
R
A
M
D
i
s
cr
i
m
i
n
a
t
or
Fr
o
n
t
(
1
0
s
a
m
p
l
e
s
)
RA
M
N
o
d
e
D
a
ta
E
x
tr
a
c
ti
o
n
0
n
.
.
.
0
1
0
1
0
0
1
..
9
n
.
.
.
0
1
0
1
0
1
1
R
A
M
D
i
s
cr
i
m
i
n
a
t
or
R
i
g
h
t
(
1
0
s
a
m
p
l
e
s
)
RA
M
N
o
d
e
D
a
ta
E
x
tr
a
c
ti
o
n
1
0
n
.
.
.
1
1
1
1
1
1
0
..
1
9
n
.
.
.
1
0
1
0
1
0
1
1
R
A
M
D
i
s
cr
i
m
i
n
a
t
or
L
ef
t
(
1
0
s
a
m
p
l
e
s
)
RA
M
N
o
d
e
D
a
ta
E
x
tr
a
c
ti
o
n
2
0
n
.
.
.
1
1
1
1
1
1
0
..
2
9
n
.
.
.
0
1
0
1
0
1
1
Ma
i
n
T
h
r
e
a
d
C
o
mp
a
r
a
t
o
r
W
i
n
n
e
r
C
l
a
s
s
n
.
.
.
1
0
1
0
1
0
1
1
F
r
o
n
t
n
.
.
.
1
0
1
0
1
0
1
1
L
ea
r
n
i
n
g
Pr
o
c
e
s
s
o
f
WNN
-
F
RA
T
h
r
e
a
d
1
T
h
r
e
a
d
2
T
h
r
e
a
d
3
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
0
0
0
0
0
0
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
0
0
0
0
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
0
0
0
0
0
0
0
0
0
0
0
0
0
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
0
0
0
0
0
0
0
0
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
1
1
1
1
1
1
1
1
1
0
0
0
0
0
0
0
0
0
0
0
0
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
1
1
1
1
1
1
1
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
1
1
0
0
1
1
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
1
1
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
0
0
0
1
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
0
0
0
0
1
1
1
1
1
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
1
1
1
1
1
1
1
1
1
1
1
1
0
0
0
0
0
1
1
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
1
1
1
1
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
1
0
0
0
0
1
1
1
1
1
1
1
1
1
1
1
1
0
1
0
0
0
1
1
1
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
1
1
1
1
1
1
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
1
1
1
0
0
0
0
0
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
0
0
1
1
1
0
0
0
0
0
0
1
0
0
0
0
1
1
1
1
1
1
1
1
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
1
1
1
1
1
0
1
0
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
0
1
1
1
1
1
1
0
0
0
1
1
1
1
1
1
1
1
1
1
0
0
0
0
1
1
0
0
0
0
0
1
0
0
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
0
0
0
0
1
1
1
1
1
1
0
0
0
1
1
1
1
1
1
1
1
1
1
1
0
0
0
0
1
1
1
0
0
0
0
0
0
0
0
0
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
0
0
0
1
1
1
1
1
1
1
1
1
1
1
1
0
0
0
0
1
1
1
1
0
0
0
0
0
0
0
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
0
0
0
1
1
1
1
1
1
1
1
1
1
1
1
1
1
0
0
0
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
0
0
0
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
0
0
0
0
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
0
0
0
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
0
0
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
0
0
0
0
0
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
0
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
0
1
1
1
1
0
0
0
0
0
0
0
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
0
0
0
0
0
1
1
1
0
0
0
0
0
0
0
0
0
0
0
0
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
0
0
0
0
1
1
1
1
1
1
1
1
1
1
0
0
0
0
0
0
0
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
0
0
0
0
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
B
i
n
a
r
y
Da
t
a
M
a
t
r
i
x
R
G
B
Bi
n
a
r
y
F
ea
tu
r
e
E
x
tr
a
c
ti
o
n
E
u
cli
d
e
a
n
F
e
a
t
u
r
e
E
x
t
r
a
c
t
i
o
n
6
5
x
3
5
p
x
Fig
u
r
e
4
.
W
NN
-
FR
A
ar
ch
itectu
r
e
d
esig
n
Fig
u
r
e
5
.
C
en
ter
s
ca
n
m
o
d
el
d
esig
n
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
A
n
o
ve
l e
mb
e
d
d
ed
a
p
p
r
o
a
ch
t
o
fa
ce
r
ec
o
g
n
itio
n
u
s
in
g
m
u
lti
-
th
r
ea
d
ed
c
o
n
tr
o
ller
b
a
s
ed
…
(
A
h
ma
d
Za
r
ka
s
i
)
3909
As
illu
s
tr
ated
in
Fig
u
r
e
5
,
t
h
e
in
p
u
t
p
atter
n
h
as
a
s
m
aller
d
ata
ca
p
ac
ity
th
an
th
e
r
ef
er
e
n
c
e
p
atter
n
,
s
p
ec
if
ically
with
two
f
ewe
r
c
o
lu
m
n
s
in
th
e
i
n
p
u
t
m
atr
ix
.
B
ef
o
r
e
t
h
e
ch
ec
k
in
g
p
r
o
ce
s
s
b
e
g
in
s
,
th
e
al
g
o
r
ith
m
f
ir
s
t
d
eter
m
in
es
th
e
in
itial
an
d
f
in
al
p
o
s
itio
n
s
f
o
r
v
er
if
icati
o
n
.
Acc
o
r
d
in
g
ly
,
th
e
in
itial
c
h
ec
k
in
g
p
o
s
itio
n
is
d
ef
in
ed
at
R
AM
n
o
d
e
p
o
in
ts
R
(
1
,
n
-
6)
to
R
(
m
,
n
-
6)
,
wh
ic
h
ar
e
eq
u
iv
alen
t
to
R
(
1
,
1
)
’
to
R
(
m
,
1
)
’
.
T
h
e
f
i
n
al
ch
ec
k
in
g
p
o
s
itio
n
is
s
et
at
R
AM
n
o
d
e
p
o
i
n
ts
R
(
1
,
n
-
1)
t
o
R
(
m
,
n
-
1)
,
wh
ich
c
o
r
r
esp
o
n
d
t
o
R
(
1
,
n
)
’
to
R
(
m
,
n
)
’
.
T
h
is
r
u
le
also
a
p
p
li
es
in
v
er
s
ely
wh
en
th
e
o
p
p
o
s
ite
co
n
d
itio
n
o
cc
u
r
s
.
T
h
e
s
eq
u
e
n
ce
o
f
o
p
er
atio
n
s
in
th
e
ce
n
ter
-
s
ca
n
r
ec
o
g
n
itio
n
alg
o
r
ith
m
is
as
f
o
llo
ws;
i)
if
th
e
m
em
o
r
y
ca
p
ac
ity
(
r
eso
lu
tio
n
)
o
f
th
e
in
p
u
t
p
atter
n
d
ata
is
eq
u
al
to
th
at
o
f
th
e
r
ef
er
en
ce
p
atter
n
,
th
e
c
o
m
p
ar
at
o
r
d
ir
ec
tly
p
er
f
o
r
m
s
th
e
v
er
if
ic
atio
n
p
r
o
ce
s
s
b
ased
o
n
th
e
p
r
ed
ef
in
e
d
i
n
itial
an
d
f
i
n
al
p
o
s
itio
n
s
,
ii)
if
th
e
m
em
o
r
y
ca
p
ac
ity
o
f
th
e
in
p
u
t
p
atter
n
is
s
m
aller
th
an
th
at
o
f
th
e
r
ef
er
e
n
ce
p
atter
n
,
th
e
al
g
o
r
ith
m
f
ir
s
t id
en
tifie
s
th
e
in
itial a
n
d
f
in
al
p
o
in
ts
o
f
th
e
r
ef
er
en
ce
p
atter
n
b
ef
o
r
e
p
er
f
o
r
m
in
g
th
e
ch
ec
k
in
g
p
r
o
ce
s
s
.
T
h
e
co
m
p
ar
is
o
n
is
th
e
n
co
n
d
u
cted
ac
c
o
r
d
in
g
to
th
e
d
ef
in
ed
m
em
o
r
y
ca
p
ac
ity
;
an
d
iii)
c
o
n
v
er
s
ely
,
i
f
th
e
i
n
p
u
t
p
atter
n
h
as
a
lar
g
er
m
em
o
r
y
ca
p
ac
ity
th
a
n
th
e
r
ef
er
en
ce
p
atter
n
,
t
h
e
alg
o
r
ith
m
f
ir
s
t
d
eter
m
in
es
th
e
in
itial
an
d
f
in
al
p
o
in
ts
o
f
th
e
in
p
u
t
p
atter
n
,
af
ter
wh
ich
th
e
d
ata
co
m
p
ar
is
o
n
is
ex
ec
u
ted
b
ased
o
n
th
e
co
r
r
esp
o
n
d
in
g
m
em
o
r
y
ca
p
ac
ity
.
3.
RE
SU
L
T
AND
DI
SCUS
SI
O
N
3
.
1
.
F
e
a
t
ure
ex
t
r
a
ct
io
n r
esu
lt
T
h
e
f
ea
tu
r
e
ex
tr
ac
tio
n
p
r
o
ce
s
s
co
n
s
is
ts
o
f
s
ev
er
al
s
tag
es,
b
eg
in
n
in
g
with
f
ac
e
d
etec
tio
n
.
T
h
is
s
tag
e
s
tar
ts
b
y
ca
p
tu
r
in
g
f
ac
ial
im
a
g
es
as
in
p
u
t
d
ata
[
1
]
.
T
h
e
im
ag
es
ac
q
u
ir
ed
f
r
o
m
th
e
ca
m
er
a
ar
e
n
o
t
p
r
o
ce
s
s
ed
d
ir
ec
tly
b
u
t
f
ir
s
t
u
n
d
er
g
o
a
p
r
ep
r
o
ce
s
s
in
g
s
tag
e.
I
n
th
is
s
tag
e,
th
e
co
lo
r
im
ag
e
R
GB
is
co
n
v
er
ted
in
to
a
g
r
ay
s
ca
le
im
ag
e.
T
h
e
p
u
r
p
o
s
e
o
f
th
is
co
n
v
er
s
io
n
is
to
r
ed
u
ce
co
m
p
lex
ity
,
s
im
p
l
if
y
an
d
ac
ce
ler
ate
co
m
p
u
tatio
n
,
wh
ile
s
till
p
r
eser
v
in
g
t
h
e
ess
en
tial
in
f
o
r
m
atio
n
r
eq
u
ir
ed
f
o
r
f
ac
e
d
etec
tio
n
.
Su
b
s
eq
u
en
tly
,
th
e
s
y
s
tem
id
en
tifie
s
r
eg
io
n
s
th
at
p
o
ten
tially
co
n
tain
f
ac
ial
o
b
je
cts.
A
s
ca
n
n
in
g
p
r
o
ce
s
s
is
p
er
f
o
r
m
ed
ac
r
o
s
s
th
e
en
tire
im
ag
e
to
lo
ca
te
s
u
b
-
win
d
o
ws
th
at
ar
e
lik
ely
to
in
clu
d
e
a
f
ac
e.
E
ac
h
s
u
b
-
win
d
o
w
is
th
en
an
aly
ze
d
b
ased
o
n
p
r
ev
io
u
s
ly
ex
tr
ac
ted
f
ea
t
u
r
es.
T
h
is
ap
p
r
o
ac
h
allo
ws
co
m
p
u
tatio
n
al
r
eso
u
r
ce
s
t
o
b
e
f
o
cu
s
ed
o
n
l
y
o
n
r
elev
an
t
ar
ea
s
,
th
er
eb
y
im
p
r
o
v
in
g
b
o
th
p
r
o
ce
s
s
in
g
ef
f
icien
cy
an
d
d
etec
tio
n
ac
cu
r
ac
y
.
T
h
e
f
in
al
s
tag
e
em
p
lo
y
s
th
e
ca
s
ca
d
e
class
if
ier
alg
o
r
ith
m
,
wh
ich
o
p
er
ates
th
r
o
u
g
h
m
u
ltip
le
test
in
g
s
tag
es
th
at
p
r
o
g
r
ess
iv
ely
elim
in
ate
im
ag
e
r
eg
io
n
s
d
is
s
im
ilar
to
f
ac
ial
p
atter
n
s
.
I
n
th
e
in
itial
s
tag
e,
th
e
test
in
g
p
ar
am
eter
s
ar
e
k
ep
t
s
im
p
le
to
en
s
u
r
e
th
at
p
o
ten
tial
f
ac
e
ca
n
d
id
ates
ar
e
n
o
t
m
is
s
ed
.
I
n
s
u
b
s
eq
u
en
t
s
tag
es,
th
e
cr
iter
ia
b
ec
o
m
e
in
cr
ea
s
in
g
ly
s
elec
tiv
e,
r
etain
in
g
o
n
ly
r
e
g
io
n
s
th
at
s
tr
o
n
g
ly
m
atch
f
a
cial
ch
ar
ac
ter
is
tics
.
T
h
is
h
ie
r
ar
ch
ical
f
ilter
in
g
m
ec
h
an
is
m
ef
f
ec
tiv
ely
r
ed
u
c
es
f
alse
d
etec
tio
n
s
an
d
y
ield
s
m
o
r
e
ac
cu
r
ate
p
r
ed
ictio
n
s
o
f
f
ac
ial
p
o
s
itio
n
s
with
in
th
e
im
ag
e
[
2
]
,
[
3
]
.
Af
ter
th
e
f
ac
e
is
s
u
cc
ess
f
u
lly
d
etec
ted
in
th
e
ca
m
er
a
im
a
g
e,
th
e
s
y
s
tem
d
eter
m
in
es
th
e
R
OI
,
th
e
s
p
ec
if
ic
p
o
r
tio
n
o
f
th
e
im
a
g
e
th
at
co
n
tain
s
th
e
u
s
er
’
s
f
ac
ial
ar
ea
.
T
h
e
R
OI
is
d
er
iv
ed
f
r
o
m
th
e
co
o
r
d
in
ates
o
b
tain
ed
d
u
r
in
g
th
e
p
r
e
v
io
u
s
Haa
r
ca
s
ca
d
e
d
etec
tio
n
p
r
o
ce
s
s
,
co
n
s
is
tin
g
o
f
th
e
an
d
p
o
s
iti
o
n
s
as we
ll
as th
e
wid
th
an
d
h
eig
h
t
v
al
u
es
o
f
th
e
d
etec
ted
f
ac
ial
ar
ea
in
ea
c
h
f
r
am
e.
T
h
e
s
ize
o
f
th
e
R
OI
m
ay
v
ar
y
d
e
p
en
d
in
g
o
n
th
e
p
r
o
x
im
ity
o
f
th
e
f
ac
e
to
th
e
ca
m
er
a.
T
o
m
ain
tai
n
d
ata
c
o
n
s
is
ten
cy
d
u
r
in
g
th
e
r
ec
o
g
n
itio
n
s
tag
e,
th
e
R
OI
is
r
esized
to
88
×
8
8
p
ix
els
,
r
e
p
r
esen
tin
g
8
8
p
ix
els
h
o
r
izo
n
tally
(
f
r
o
m
t
h
e
lef
t
t
o
th
e
r
ig
h
t
s
id
e
o
f
th
e
f
ac
e)
a
n
d
8
8
p
ix
els
v
er
tically
(
f
r
o
m
th
e
f
o
r
e
h
ea
d
to
t
h
e
c
h
in
)
.
T
h
e
n
e
x
t
s
tep
in
v
o
lv
es
c
o
n
v
e
r
tin
g
t
h
e
R
OI
im
ag
e
f
r
o
m
R
GB
f
o
r
m
at
to
g
r
a
y
s
ca
le
.
Fo
llo
win
g
th
is
,
th
e
s
y
s
tem
ap
p
lies
a
th
r
esh
o
ld
in
g
p
r
o
ce
s
s
u
s
in
g
an
ad
ap
tiv
e
th
r
esh
o
ld
v
alu
e
th
at
ad
ju
s
ts
ac
co
r
d
in
g
to
th
e
a
v
er
ag
e
l
ig
h
t
in
ten
s
ity
with
in
th
e
R
OI
,
en
s
u
r
in
g
s
tab
le
co
n
v
er
s
io
n
r
esu
lts
d
esp
ite
lig
h
tin
g
v
ar
iatio
n
s
.
T
h
e
th
r
esh
o
l
d
in
g
p
r
o
ce
s
s
p
r
o
d
u
ce
s
a
b
in
a
r
y
im
ag
e
,
wh
ich
is
th
en
n
o
r
m
alize
d
u
s
in
g
a
r
esize
m
eth
o
d
to
a
f
ix
ed
s
ize
o
f
60
×
6
0
p
ix
els
.
T
h
is
n
o
r
m
aliza
tio
n
aim
s
to
s
tan
d
ar
d
ize
th
e
f
ac
ial
im
ag
e
d
im
en
s
io
n
s
u
s
ed
as
in
p
u
t,
th
e
r
e
b
y
s
im
p
lify
in
g
s
u
b
s
eq
u
en
t
s
ta
g
es
o
f
d
ata
s
to
r
ag
e
,
p
r
o
ce
s
s
in
g
,
an
d
r
ec
o
g
n
itio
n
.
T
ab
le
1
p
r
esen
ts
th
e
e
x
p
er
im
e
n
tal
r
esu
lts
o
f
th
e
f
ac
e
d
etec
tio
n
s
tag
e.
T
ab
le
1
.
Face
d
etec
tio
n
r
esu
lt
F
a
c
e
d
e
t
e
c
t
i
o
n
F
a
c
e
_
1
F
a
c
e
_
2
F
a
c
e
_
3
F
a
c
e
_
4
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
1
5
,
No
.
4
,
Au
g
u
s
t 2
0
2
6
:
3
9
0
3
-
3
9
1
8
3910
T
ab
le
1
p
r
esen
ts
th
e
r
esu
lts
o
f
test
in
g
co
n
d
u
cte
d
u
s
in
g
f
o
u
r
d
if
f
er
en
t
f
ac
ial
s
am
p
les.
T
h
er
e
ar
e
f
o
u
r
f
ac
ial
p
atter
n
s
:
Face
_
1
,
Face
_
2
,
Face
_
3
,
an
d
Face
_
4
.
Face
_
1
r
e
p
r
esen
ts
th
e
u
s
er
’
s
f
ac
ial
p
atter
n
d
ata
,
wh
ich
is
later
u
s
ed
as
th
e
d
ataset
in
th
e
lear
n
in
g
p
r
o
ce
s
s
,
wh
er
ea
s
Face
_
2
,
Face
_
3
,
an
d
Face
_
4
ar
e
f
ac
ial
p
atter
n
s
em
p
lo
y
e
d
f
o
r
test
in
g
d
u
r
in
g
th
e
r
ec
o
g
n
itio
n
p
h
as
e.
I
m
ag
e
ac
q
u
is
itio
n
was
ca
r
r
ied
o
u
t
at
s
ev
er
al
d
is
tan
ce
s
—
1
0
cm
,
2
0
cm
,
an
d
3
0
cm
alth
o
u
g
h
th
e
tab
le
o
n
ly
d
is
p
lay
s
th
e
r
esu
lts
o
b
tai
n
ed
at
a
d
is
tan
ce
o
f
1
0
cm
.
T
h
e
p
u
r
p
o
s
e
o
f
th
is
f
a
ce
d
etec
tio
n
test
is
to
en
s
u
r
e
th
at
th
e
s
y
s
tem
f
u
n
ctio
n
s
ac
co
r
d
in
g
to
its
d
esig
n
an
d
to
ev
alu
ate
th
e
ac
c
u
r
ac
y
l
ev
el
in
r
ec
o
g
n
izin
g
f
ac
ial
o
b
je
cts
u
n
d
er
v
ar
y
i
n
g
co
n
d
itio
n
s
.
T
h
e
test
in
g
p
r
o
ce
s
s
was
p
er
f
o
r
m
ed
u
s
in
g
b
o
th
s
till
im
ag
es
an
d
v
id
e
o
r
ec
o
r
d
ed
d
ir
ec
tly
th
r
o
u
g
h
th
e
ca
m
er
a.
Fu
r
th
er
m
o
r
e
,
test
in
g
was
co
n
d
u
cted
with
v
a
r
iatio
n
s
in
f
ac
ial
o
r
ien
tatio
n
,
in
cl
u
d
i
n
g
f
r
o
n
tal
,
lef
t
-
s
id
e
,
an
d
r
ig
h
t
-
s
id
e
v
iews.
T
h
ese
v
ar
iatio
n
s
wer
e
c
h
o
s
en
t
o
as
s
ess
th
e
s
y
s
tem
’
s
p
er
f
o
r
m
an
c
e
wh
en
e
n
co
u
n
ter
in
g
ch
an
g
e
s
in
en
v
ir
o
n
m
e
n
tal
co
n
d
itio
n
s
an
d
o
b
ject
p
o
s
itio
n
in
g
.
On
ce
th
e
f
ac
ial
p
atter
n
w
as
s
u
cc
ess
f
u
lly
d
etec
ted
,
th
e
s
y
s
tem
d
is
p
lay
ed
a
b
lu
e
b
o
u
n
d
in
g
b
o
x
ar
o
u
n
d
th
e
id
en
tifie
d
f
ac
e
with
in
t
h
e
f
r
am
e.
T
h
is
v
is
u
aliza
tio
n
,
as
s
h
o
wn
in
T
ab
le
1
,
s
er
v
es a
s
an
in
d
icatio
n
o
f
th
e
s
u
cc
ess
f
u
l e
x
ec
u
tio
n
o
f
th
e
f
ac
e
d
etec
tio
n
p
r
o
ce
s
s
.
T
h
e
n
e
x
t
s
tag
e
is
th
e
f
ac
ial
f
ea
tu
r
e
e
x
tr
ac
tio
n
p
r
o
ce
s
s
,
w
h
ich
s
er
v
es
as
a
cr
u
cial
s
tep
in
a
f
ac
e
r
ec
o
g
n
itio
n
s
y
s
tem
,
as it a
im
s
to
o
b
tain
in
f
o
r
m
atio
n
r
ep
r
esen
tin
g
th
e
u
n
i
q
u
e
ch
a
r
ac
ter
is
tics
o
f
ea
ch
in
d
iv
id
u
al.
I
n
th
is
s
tu
d
y
,
th
e
Haa
r
ca
s
ca
d
e
class
if
i
er
alg
o
r
ith
m
is
em
p
lo
y
ed
to
d
etec
t
f
ac
ial
f
ea
tu
r
es
with
in
th
e
im
ag
e.
I
n
ad
d
itio
n
to
id
en
tify
in
g
th
e
p
r
e
s
en
ce
o
f
a
f
ac
e
,
th
e
alg
o
r
ith
m
also
p
r
o
d
u
ce
s
co
o
r
d
in
ate
in
f
o
r
m
atio
n
in
th
e
f
o
r
m
o
f
a
b
o
u
n
d
in
g
b
o
x
,
a
r
ec
ta
n
g
u
lar
b
o
u
n
d
ar
y
i
n
d
icatin
g
th
e
e
x
ac
t
lo
ca
tio
n
o
f
th
e
f
ac
e
in
ea
c
h
ca
m
er
a
f
r
a
m
e.
As
illu
s
tr
ated
in
Fig
u
r
e
6
,
wh
ic
h
d
e
p
icts
th
e
f
ea
tu
r
e
ex
tr
ac
ti
o
n
p
r
o
ce
s
s
,
o
n
ce
t
h
e
f
ac
ial
p
o
s
itio
n
is
d
etec
ted
th
r
o
u
g
h
t
h
e
b
o
u
n
d
in
g
b
o
x
i
n
t
h
e
r
aw
im
ag
e,
th
e
s
y
s
tem
d
ete
r
m
in
es
th
e
R
OI
,
th
e
s
p
ec
if
ic
p
o
r
tio
n
o
f
th
e
im
ag
e
co
n
tain
in
g
th
e
r
elev
an
t
f
ac
ial
f
ea
tu
r
es
f
o
r
f
u
r
th
e
r
a
n
aly
s
is
,
with
a
p
i
x
el
d
im
e
n
s
io
n
o
f
88
×
88
.
Dete
r
m
in
i
n
g
th
e
R
OI
is
ess
en
tial,
a
s
it
a
llo
ws
co
m
p
u
tatio
n
al
r
eso
u
r
ce
s
to
b
e
f
o
cu
s
ed
o
n
t
h
e
ar
ea
s
co
n
tain
in
g
h
i
g
h
in
f
o
r
m
atio
n
al
v
alu
e
wh
ile
r
ed
u
cin
g
u
n
n
ec
ess
ar
y
p
r
o
ce
s
s
in
g
o
n
ir
r
elev
a
n
t
im
ag
e
r
eg
i
o
n
s
.
I
n
th
is
r
esear
ch
,
th
e
R
OI
s
er
v
es
as
th
e
p
r
im
ar
y
in
p
u
t
f
o
r
tr
a
n
s
f
o
r
m
in
g
v
is
u
al
d
ata
in
to
a
d
ig
ital
r
e
p
r
esen
tatio
n
th
at
ca
n
b
e
p
r
o
ce
s
s
ed
b
y
th
e
r
ec
o
g
n
itio
n
s
y
s
tem
.
I
n
a
d
d
itio
n
to
f
u
n
ctio
n
in
g
as
th
e
i
n
p
u
t
f
o
r
f
ea
t
u
r
e
ex
tr
ac
tio
n
,
th
e
R
OI
s
ize
is
also
u
tili
ze
d
to
e
s
tim
ate
th
e
r
elativ
e
d
is
tan
ce
b
etwe
en
th
e
f
ac
e
an
d
th
e
ca
m
er
a.
Gen
e
r
ally
,
a
lar
g
er
R
OI
in
d
icate
s
th
at
th
e
f
ac
e
is
cl
o
s
er
to
th
e
ca
m
er
a,
an
d
c
o
n
v
er
s
ely
,
a
s
m
aller
R
OI
co
r
r
es
p
o
n
d
s
t
o
a
g
r
ea
ter
d
is
tan
ce
.
On
ce
th
e
d
ata
a
r
e
o
b
t
ain
ed
,
th
e
n
ex
t step
is
to
co
n
v
er
t th
em
in
to
b
in
ar
y
f
o
r
m
at
.
B
in
ar
y
d
ata
m
atr
ix
Fig
u
r
e
6
.
Featu
r
e
ex
tr
ac
tio
n
r
e
s
u
lt
T
h
e
co
n
v
er
s
io
n
o
f
im
ag
es
in
t
o
b
in
a
r
y
f
o
r
m
is
a
cr
u
cial
s
tag
e,
as
it
s
im
p
lifie
s
d
ata
p
r
o
ce
s
s
in
g
with
in
th
e
m
icr
o
co
n
tr
o
ller
s
y
s
tem
.
T
h
is
is
ac
h
iev
ed
b
y
tr
an
s
f
o
r
m
in
g
th
e
v
is
u
al
r
ep
r
esen
tatio
n
in
to
a
f
u
n
d
am
e
n
tal
d
ig
ital
f
o
r
m
at
b
in
a
r
y
.
T
h
e
co
n
v
er
s
io
n
p
r
o
ce
s
s
b
eg
i
n
s
b
y
tr
a
n
s
f
o
r
m
in
g
th
e
R
OI
im
ag
e
f
r
o
m
R
GB
f
o
r
m
at
in
t
o
g
r
ay
s
ca
le
.
Su
b
s
eq
u
en
tly
,
th
e
s
y
s
tem
ap
p
lies
a
th
r
esh
o
ld
in
g
m
eth
o
d
with
an
ad
ap
tiv
e
t
h
r
esh
o
ld
v
alu
e
th
at
ad
ju
s
ts
ac
co
r
d
in
g
to
th
e
av
er
a
g
e
lig
h
t
in
ten
s
ity
with
in
th
e
R
OI
,
en
s
u
r
in
g
th
at
th
e
co
n
v
e
r
s
io
n
p
r
o
ce
s
s
r
em
ain
s
s
tab
le
d
esp
ite
v
ar
iatio
n
s
in
illu
m
in
atio
n
co
n
d
itio
n
s
.
T
h
e
r
esu
lt
o
f
th
e
th
r
esh
o
l
d
in
g
p
r
o
ce
s
s
is
a
b
in
a
r
y
im
ag
e
,
wh
ich
is
th
en
n
o
r
m
alize
d
in
s
ize
th
r
o
u
g
h
a
r
esize
o
p
er
a
tio
n
to
a
f
ix
ed
r
eso
lu
tio
n
o
f
60
×
6
0
p
i
x
els
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
A
n
o
ve
l e
mb
e
d
d
ed
a
p
p
r
o
a
ch
t
o
fa
ce
r
ec
o
g
n
itio
n
u
s
in
g
m
u
lti
-
th
r
ea
d
ed
c
o
n
tr
o
ller
b
a
s
ed
…
(
A
h
ma
d
Za
r
ka
s
i
)
3911
T
h
is
n
o
r
m
aliza
tio
n
en
s
u
r
es
co
n
s
is
ten
t
f
ac
ial
d
ata
d
im
en
s
io
n
s
u
s
ed
as
in
p
u
t,
t
h
er
eb
y
f
ac
ilit
atin
g
th
e
p
r
o
ce
s
s
es
of
s
to
r
ag
e
,
tr
a
n
s
m
is
s
io
n
,
an
d
r
ec
o
g
n
itio
n
.
T
h
e
f
in
al
o
u
tco
m
e
o
f
th
e
f
ea
tu
r
e
e
x
tr
ac
tio
n
s
tag
e
in
th
is
s
tu
d
y
is
a
f
ac
ial
p
atter
n
r
ep
r
esen
ted
as
a
b
in
ar
y
im
ag
e
.
T
h
is
b
i
n
ar
y
d
ata
also
s
er
v
es
as
th
e
d
ataset
f
o
r
t
h
e
s
u
b
s
eq
u
en
t
p
atter
n
r
ec
o
g
n
itio
n
p
r
o
ce
s
s
.
3
.
2
.
M
emo
ry
re
qu
irem
ent
a
na
ly
s
is
B
ef
o
r
e
d
esig
n
in
g
th
e
f
ac
ial
p
atter
n
d
ataset,
a
m
em
o
r
y
r
e
q
u
ir
em
en
t
a
n
aly
s
is
was
co
n
d
u
cted
o
n
th
e
s
in
g
le
-
ch
ip
m
icr
o
co
n
t
r
o
ller
s
y
s
tem
[
7
]
,
[
8
]
.
T
h
is
a
n
aly
s
is
aim
ed
to
d
eter
m
in
e
th
e
m
em
o
r
y
ca
p
ac
ity
n
ee
d
ed
to
ef
f
icien
tly
s
to
r
e
an
d
m
an
ag
e
f
ac
ial
p
atter
n
d
ata
[
1
0
]
.
T
h
is
s
tep
is
ess
en
tial b
ec
au
s
e
th
e
av
ailab
le
r
eso
u
r
ce
s
in
a
m
icr
o
co
n
tr
o
ller
a
r
e
in
h
er
en
tly
lim
ited
;
th
er
e
f
o
r
e,
p
r
ec
is
e
ca
lcu
latio
n
s
ar
e
r
eq
u
ir
e
d
t
o
e
n
s
u
r
e
th
at
th
e
s
y
s
tem
o
p
er
ates
o
p
tim
ally
with
o
u
t
e
x
ce
ed
in
g
th
e
ca
p
ac
ity
o
f
th
e
h
a
r
d
war
e
u
s
ed
.
T
h
e
m
icr
o
co
n
tr
o
ller
u
tili
ze
d
in
th
is
s
tu
d
y
is
th
e
AR
M®
3
2
-
b
it
C
o
r
tex
®
-
M4
,
wh
ich
f
ea
tu
r
es
a
f
lo
atin
g
-
p
o
in
t
u
n
it
(
FP
U)
,
a
m
ax
im
u
m
C
PU
f
r
eq
u
e
n
cy
o
f
8
4
MH
z
,
5
1
2
KB
o
f
Flas
h
m
em
o
r
y
,
an
d
9
6
KB
o
f
SR
AM
.
T
h
e
lim
ite
d
m
em
o
r
y
ca
p
ac
it
y
p
r
esen
ts
o
n
e
o
f
th
e
p
r
im
ar
y
c
h
allen
g
es
in
im
p
lem
e
n
tin
g
a
n
em
b
ed
d
ed
-
b
ased
f
ac
e
r
ec
o
g
n
itio
n
s
y
s
tem
,
as
th
e
s
to
r
ag
e
an
d
p
r
o
ce
s
s
in
g
o
f
d
ig
ital
im
ag
es
d
em
an
d
r
elativ
ely
lar
g
e
m
em
o
r
y
s
p
ac
e.
C
o
n
s
eq
u
en
tly
,
an
o
p
tim
izatio
n
s
tr
ateg
y
is
r
eq
u
ir
ed
,
o
n
e
th
at
in
v
o
l
v
es
m
o
d
if
y
in
g
an
d
m
an
ag
i
n
g
d
ata
allo
ca
tio
n
in
m
em
o
r
y
to
en
s
u
r
e
ef
f
icien
t r
eso
u
r
ce
u
tili
za
tio
n
.
I
n
th
is
r
esear
ch
,
th
e
f
ac
i
al
p
atter
n
d
ataset
is
im
p
lem
en
ted
with
in
th
e
m
ain
th
r
ea
d
,
wh
ic
h
o
p
er
ates
with
th
r
ee
ac
tiv
e
th
r
ea
d
s
[
2
5
]
,
[
2
6
]
.
E
ac
h
th
r
ea
d
is
r
esp
o
n
s
ib
le
f
o
r
p
r
o
ce
s
s
in
g
2
0
f
ac
ial
p
atter
n
s
s
to
r
ed
in
m
em
o
r
y
.
Sp
ec
if
ically
,
T
h
r
ea
d
1
h
an
d
les
f
r
o
n
t
-
f
ac
e
p
atter
n
s
,
T
h
r
ea
d
2
p
r
o
ce
s
s
es
lef
t
-
f
ac
e
p
atter
n
s
,
an
d
T
h
r
ea
d
3
is
d
e
d
icate
d
to
r
ig
h
t
-
f
ac
e
p
atter
n
s
.
W
ith
th
is
s
tr
u
ctu
r
e,
t
h
e
r
ec
o
g
n
itio
n
p
r
o
ce
s
s
ca
n
b
e
ex
ec
u
ted
in
p
ar
allel
,
th
er
eb
y
i
m
p
r
o
v
i
n
g
ex
ec
u
tio
n
ef
f
icien
c
y
wh
ile
m
ain
tain
in
g
s
y
s
tem
s
tab
ilit
y
.
T
h
e
r
esu
lts
o
f
th
e
s
y
s
tem
m
em
o
r
y
r
eq
u
ir
em
en
t
ca
lcu
latio
n
s
ar
e
p
r
esen
ted
in
T
ab
le
2
,
wh
ich
s
er
v
es
as
th
e
b
as
is
f
o
r
ev
alu
atin
g
m
em
o
r
y
e
f
f
icien
cy
an
d
r
eso
u
r
ce
o
p
tim
izatio
n
in
t
h
is
em
b
ed
d
ed
f
ac
e
r
ec
o
g
n
itio
n
s
y
s
tem
.
T
ab
le
2
.
Me
m
o
r
y
r
eq
u
ir
em
e
n
t
Th
r
e
a
d
s
N
u
mb
e
r
o
f
f
a
c
e
p
a
t
t
e
r
n
s
B
i
n
a
r
y
d
a
t
a
mat
r
i
x
N
u
mb
e
r
o
f
p
i
x
e
l
s
N
u
mb
e
r
o
f
memo
r
i
e
s
To
t
a
l
mem
o
r
y
r
e
q
u
i
r
e
m
e
n
t
(
a
=
2
0
)
(
p
i
x
e
l
)
(
P
=
a
×
p
i
x
e
l
)
(
n
=
p
/
8
)
b
y
t
e
(
K
B
)
Th
r
e
a
d
1
F
r
o
n
t
60
×
60
72
,
000
9
,
0
0
0
9
Th
r
e
a
d
2
Le
f
t
60
×
50
60
,
000
7
,
5
0
0
7
.
5
Th
r
e
a
d
3
R
i
g
h
t
60
×
50
60
,
000
7
,
5
0
0
7
.
5
Af
ter
d
eter
m
in
i
n
g
th
e
to
tal
n
u
m
b
er
o
f
f
ac
ial
p
atter
n
s
t
o
b
e
u
s
ed
in
th
e
d
ataset,
a
co
m
p
r
eh
en
s
iv
e
m
em
o
r
y
r
e
q
u
ir
em
e
n
t
ca
lcu
latio
n
was
ca
r
r
ied
o
u
t
to
en
s
u
r
e
th
at
th
e
allo
ca
ted
ca
p
ac
ity
d
o
es
n
o
t
ex
ce
ed
th
e
m
ax
im
u
m
lim
it
a
v
ailab
le
o
n
t
h
e
m
icr
o
c
o
n
tr
o
ller
.
T
h
is
ca
lc
u
latio
n
p
r
o
ce
s
s
in
v
o
lv
e
d
s
ev
e
r
al
k
ey
p
ar
am
eter
s
,
in
clu
d
in
g
th
e
s
ize
o
f
ea
ch
f
ac
ial
p
atter
n
,
th
e
n
u
m
b
er
o
f
ac
ti
v
e
th
r
ea
d
s
,
an
d
th
e
d
is
tr
ib
u
tio
n
o
f
d
ata
b
etwe
en
f
lash
m
em
o
r
y
an
d
SR
AM
.
B
ased
o
n
th
e
an
aly
s
is
r
esu
lts
,
th
e
to
tal
m
em
o
r
y
r
eq
u
ir
e
d
to
s
to
r
e
6
0
f
ac
ial
p
atter
n
s
in
th
e
d
ataset
is
2
4
KB
,
c
o
n
s
is
tin
g
o
f
9
KB
f
o
r
f
r
o
n
t
-
f
ac
e
p
at
ter
n
s
,
7
.
5
KB
f
o
r
r
ig
h
t
-
f
ac
e
p
a
tter
n
s
,
an
d
7
.
5
KB
f
o
r
lef
t
-
f
ac
e
p
atter
n
s
.
T
h
ese
f
in
d
in
g
s
in
d
icate
th
at
th
e
R
OM
m
em
o
r
y
ca
p
ac
ity
o
f
th
e
m
icr
o
co
n
tr
o
ller
s
till
h
as
4
8
8
KB
av
ailab
le.
T
h
e
n
ex
t
s
t
ag
e
in
v
o
lv
es
co
n
s
tr
u
ctin
g
th
e
f
ac
ial
p
atter
n
d
ataset
b
ased
o
n
th
e
ac
q
u
ir
ed
im
a
g
e
d
ata.
I
m
a
g
e
ac
q
u
is
itio
n
is
p
er
f
o
r
m
ed
u
s
in
g
a
R
asp
b
er
r
y
Pi
ca
m
er
a
(
R
asp
iC
am
)
m
o
u
n
ted
o
n
th
e
r
o
b
o
t’
s
n
ec
k
s
ec
tio
n
.
T
h
e
p
lace
m
en
t
o
f
th
e
ca
m
er
a
in
t
h
is
p
o
s
itio
n
e
n
ab
le
s
d
ir
ec
t
im
ag
e
ca
p
tu
r
e
o
f
t
h
e
u
s
er
’
s
f
ac
e
with
a
n
alig
n
ed
v
iewin
g
an
g
le
—
co
v
e
r
in
g
th
e
f
r
o
n
tal
,
lef
t
-
s
id
e
,
an
d
r
ig
h
t
-
s
id
e
p
er
s
p
ec
tiv
es.
E
ac
h
ca
p
tu
r
ed
f
ac
ia
l
im
ag
e
r
ep
r
esen
ts
a
d
is
tin
ct
in
d
iv
id
u
al
f
ac
ial
p
atter
n
an
d
s
er
v
es
as
a
d
ata
s
o
u
r
ce
f
o
r
b
u
ild
i
n
g
th
e
f
ac
ial
p
atter
n
d
ataset.
T
h
is
im
ag
e
ac
q
u
is
itio
n
p
r
o
ce
s
s
is
d
esig
n
ed
t
o
b
e
f
lex
ib
le
,
allo
win
g
th
e
g
en
e
r
atio
n
o
f
v
is
u
al
d
ata
d
ir
ec
tly
in
d
ig
ital f
o
r
m
at
.
C
o
n
s
eq
u
en
tly
,
th
e
d
ataset
co
n
s
tr
u
c
tio
n
p
r
o
ce
s
s
ca
n
b
e
p
er
f
o
r
m
e
d
ef
f
icien
tly
with
o
u
t
r
eq
u
ir
in
g
ad
d
itio
n
al
co
n
v
er
s
io
n
s
tep
s
.
T
h
e
d
ata
ca
p
tu
r
ed
b
y
th
e
R
asp
iC
am
,
af
ter
b
ein
g
co
n
v
er
ted
in
to
b
in
ar
y
im
ag
es,
ar
e
s
u
b
s
eq
u
en
tly
u
s
ed
as
th
e
r
ep
r
esen
tatio
n
o
f
th
e
ex
tr
ac
ted
f
ac
ial
p
atter
n
s
.
T
h
e
im
ag
e
r
ep
r
esen
tatio
n
in
b
in
a
r
y
f
o
r
m
at
is
illu
s
tr
ated
in
Fig
u
r
e
6
,
wh
ich
s
h
o
ws
th
e
s
tr
u
ctu
r
e
o
f
th
e
b
in
ar
y
d
ata
m
atr
ix
as
th
e
f
in
al
o
u
tp
u
t
o
f
th
e
im
ag
e
co
n
v
e
r
s
io
n
an
d
v
is
u
al
d
ata
s
to
r
a
g
e
p
r
o
ce
s
s
es
with
in
th
e
s
y
s
tem
.
T
h
e
r
esu
ltin
g
b
in
ar
y
d
ata
ar
e
th
en
s
av
ed
in
a
tex
t
f
ile
with
a
.
tx
t
ex
ten
s
io
n
t
o
allo
w
f
u
r
th
er
p
r
o
ce
s
s
in
g
in
th
e
s
u
b
s
eq
u
en
t
s
tag
es.
T
h
e
te
x
t
f
ile
co
n
tain
in
g
th
e
b
in
ar
y
m
atr
ix
is
p
r
o
ce
s
s
ed
u
s
in
g
a
C
-
b
ased
p
r
o
g
r
am
,
wh
ich
co
n
v
e
r
ts
it
in
to
a
d
ataset.
cp
p
f
ile
f
u
n
ctio
n
s
as
a
lib
r
ar
y
with
in
th
e
s
y
s
tem
.
I
n
th
e
d
ata
s
et.
cp
p
f
ile,
th
e
b
in
ar
y
d
ata
ar
e
d
ec
lar
ed
an
d
s
to
r
e
d
u
s
in
g
th
e
p
r
o
g
r
am
m
em
o
r
y
(
PR
OGM
E
M
)
d
ir
ec
tiv
e
,
en
s
u
r
in
g
th
at
th
e
d
ata
ar
e
p
lace
d
in
th
e
f
lash
m
em
o
r
y
o
f
th
e
m
i
cr
o
co
n
tr
o
ller
.
T
h
is
s
to
r
ag
e
s
tr
ateg
y
is
em
p
lo
y
ed
to
o
p
tim
ize
r
eso
u
r
ce
u
tili
za
tio
n
b
y
en
ab
lin
g
th
e
d
ata
to
b
e
s
to
r
ed
p
er
m
an
e
n
tly
with
o
u
t
co
n
s
u
m
in
g
th
e
lim
ite
d
R
AM
ca
p
ac
ity
.
T
h
e
d
ataset
u
s
ed
in
th
is
s
tu
d
y
c
o
n
s
is
ts
o
f
6
0
f
ac
ial
im
ag
es
,
r
ep
r
esen
tin
g
a
s
in
g
le
f
ac
ial
id
en
tity
.
E
ac
h
id
e
n
tity
is
ca
p
tu
r
ed
at
th
r
ee
d
if
f
er
e
n
t
d
is
tan
ce
s
1
0
cm
,
2
0
cm
,
an
d
3
0
cm
with
f
iv
e
im
a
g
es
tak
en
p
er
d
is
tan
ce
.
All
d
ata
s
to
r
ed
in
th
e
d
ataset.
cp
p
f
ile
is
s
u
b
s
e
q
u
en
tly
u
p
lo
a
d
ed
to
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
1
5
,
No
.
4
,
Au
g
u
s
t 2
0
2
6
:
3
9
0
3
-
3
9
1
8
3912
th
e
m
icr
o
co
n
tr
o
ller
an
d
allo
ca
ted
with
in
th
e
R
AM
n
o
d
es
lo
ca
ted
in
th
e
R
AM
d
is
cr
im
in
ato
r
s
.
T
h
is
s
tr
u
ctu
r
e
allo
ws
th
e
s
y
s
tem
to
p
er
f
o
r
m
f
ac
ial
id
en
tific
atio
n
p
r
o
ce
s
s
es
ef
f
icien
tly
wh
ile
m
ain
tain
in
g
o
p
tim
al
m
em
o
r
y
allo
ca
tio
n
an
d
m
in
im
izin
g
c
o
m
p
u
tatio
n
al
o
v
er
h
ea
d
.
3
.
3
.
Weig
htle
s
s
neura
l net
wo
rk
f
a
ce
re
co
g
nitio
n a
lg
o
rit
h
m
T
h
e
in
itial
s
tag
e
o
f
s
y
s
tem
test
in
g
an
d
an
aly
s
is
in
th
is
s
tu
d
y
is
th
e
d
ataset
v
alid
atio
n
p
r
o
ce
s
s
.
T
h
is
s
tep
aim
s
to
en
s
u
r
e
th
at
th
e
f
a
cial
p
atter
n
r
ec
o
g
n
itio
n
s
y
s
tem
f
u
n
ctio
n
s
co
r
r
ec
tly
an
d
is
c
ap
ab
le
o
f
id
en
tif
y
in
g
f
ac
ial
d
ata
ac
co
r
d
in
g
t
o
th
e
s
to
r
ed
id
en
titi
es.
Valid
atio
n
is
p
er
f
o
r
m
ed
b
y
s
elec
tin
g
a
n
u
m
b
er
o
f
f
ac
ial
p
atter
n
s
am
p
les
f
r
o
m
t
h
e
d
ataset
a
n
d
t
esti
n
g
th
eir
r
ec
o
g
n
itio
n
ac
cu
r
a
cy
.
T
h
e
d
ataset
v
alid
atio
n
p
r
o
ce
s
s
p
lay
s
a
cr
u
cial
r
o
le
in
g
u
ar
a
n
teein
g
s
y
s
tem
r
eliab
ilit
y
b
y
en
s
u
r
in
g
t
h
at
ea
c
h
f
ac
ial
d
ata
en
tr
y
s
to
r
ed
in
t
h
e
d
atab
ase
ca
n
b
e
ac
cu
r
ately
r
ec
o
g
n
ized
b
ased
o
n
th
e
in
d
iv
id
u
al
’
s
id
en
tity
a
n
d
im
ag
e
ca
p
tu
r
e
d
is
tan
ce
.
I
n
th
is
r
esear
ch
,
th
e
v
alid
ated
d
ata
i
n
clu
d
e
t
h
r
ee
f
ac
ial
class
es
ca
p
tu
r
ed
at
th
r
e
e
d
if
f
er
e
n
t
d
is
tan
ce
s
:
1
0
c
m
,
2
0
cm
,
an
d
3
0
cm
.
E
ac
h
co
m
b
in
atio
n
o
f
f
ac
ial
c
lass
an
d
d
is
tan
ce
is
test
ed
th
r
o
u
g
h
d
ir
ec
t
m
atch
in
g
ag
ain
s
t
th
e
r
e
f
er
en
ce
d
ata
s
to
r
ed
in
th
e
d
ataset,
allo
win
g
f
o
r
th
e
ev
al
u
atio
n
o
f
th
e
s
y
s
tem
’
s
id
en
tific
atio
n
ac
cu
r
a
cy
.
T
h
e
v
alid
atio
n
r
esu
lts
ar
e
ex
p
ec
ted
to
y
ield
a
1
0
0
%
r
ec
o
g
n
itio
n
ac
cu
r
ac
y
,
s
in
ce
th
e
test
d
ata
ar
e
d
ir
ec
tl
y
d
er
iv
e
d
f
r
o
m
th
e
s
am
e
d
ataset
u
s
ed
f
o
r
tr
ain
in
g
.
I
f
th
e
ac
h
iev
e
d
ac
cu
r
ac
y
is
u
n
d
er
1
0
0
%,
it
in
d
icate
s
th
at
th
e
alg
o
r
ith
m
h
as
n
o
t
y
et
f
u
n
ctio
n
ed
o
p
tim
ally
in
r
e
co
g
n
izin
g
its
in
ter
n
al
f
ac
ial
p
atter
n
s
.
T
h
e
d
etailed
r
esu
lts
o
f
th
e
f
ac
ial
d
ataset
v
alid
atio
n
test
ar
e
p
r
esen
ted
in
Fig
u
r
e
7
.
Fig
u
r
e
7
(
a)
p
r
esen
ts
th
e
test
r
esu
lts
f
o
r
T
h
r
ea
d
1
,
wh
ich
r
e
p
r
esen
ts
th
e
f
r
o
n
t
-
f
ac
e
p
atter
n
.
B
ased
o
n
th
e
v
alid
atio
n
r
esu
lts
,
Face
1
ac
h
iev
ed
t
h
e
h
ig
h
est
r
ec
o
g
n
iti
o
n
ac
c
u
r
ac
y
o
f
1
0
0
%
,
with
an
ex
ec
u
tio
n
tim
e
o
f
0
.
0
3
3
s
ec
o
n
d
s
.
T
h
is
in
d
icate
s
th
at
th
e
s
y
s
tem
co
n
s
i
s
ten
tl
y
r
ec
o
g
n
ize
d
Face
1
as
th
e
co
r
r
ec
t
o
u
tp
u
t.
Fo
r
Face
2
,
th
e
ac
cu
r
ac
y
r
ea
c
h
ed
4
5
.
2
8
%
with
a
n
ex
ec
u
tio
n
tim
e
o
f
0
.
0
1
9
s
ec
o
n
d
s
,
w
h
ile
Face
3
s
h
o
wed
an
ac
cu
r
ac
y
o
f
3
8
.
1
4
%
an
d
an
e
x
ec
u
tio
n
tim
e
o
f
0
.
0
1
6
s
ec
o
n
d
s
.
Me
an
wh
ile,
th
e
test
f
o
r
F
ac
e
4
r
esu
lted
in
a
n
ac
cu
r
ac
y
o
f
6
5
.
8
1
%
,
with
a
r
elativ
ely
lo
n
g
er
ex
ec
u
tio
n
tim
e
o
f
0
.
0
2
5
s
ec
o
n
d
s
co
m
p
ar
e
d
to
th
e
o
th
er
f
ac
es.
A
s
im
ilar
p
atter
n
ca
n
b
e
o
b
s
er
v
ed
in
Fig
u
r
es
7
(
b
)
an
d
7
(
c)
,
wh
ich
r
ep
r
esen
t
T
h
r
ea
d
2
(
le
f
t
-
f
ac
e
p
atter
n
)
an
d
T
h
r
ea
d
3
(
r
ig
h
t
-
f
ac
e
p
atter
n
)
,
r
esp
ec
tiv
ely
.
O
v
er
all,
th
e
v
alid
atio
n
r
esu
lts
s
h
o
w
th
at
Face
4
in
T
h
r
ea
d
3
ac
h
iev
ed
t
h
e
h
ig
h
est
r
ec
o
g
n
i
tio
n
ac
cu
r
ac
y
o
f
8
0
.
9
2
%
,
w
h
ile
th
e
f
astes
t
ex
ec
u
tio
n
tim
e
was
r
ec
o
r
d
ed
b
y
T
h
r
ea
d
1
o
n
Face
3
,
at
0
.
0
1
6
s
ec
o
n
d
s
.
T
h
u
s
,
it
ca
n
b
e
c
o
n
clu
d
e
d
th
at
th
e
d
ataset
r
ec
o
g
n
itio
n
v
alid
atio
n
p
r
o
ce
s
s
o
p
er
ated
in
ac
c
o
r
d
a
n
c
e
with
th
e
s
y
s
tem
’
s
d
esig
n
,
wh
er
e
ea
ch
th
r
ea
d
d
em
o
n
s
tr
ated
a
p
atter
n
-
m
atch
in
g
ac
cu
r
ac
y
o
f
u
p
to
1
0
0
%
with
r
esp
ec
t to
its
co
r
r
esp
o
n
d
i
n
g
r
e
f
er
en
ce
p
atter
n
s
in
t
h
e
d
ataset.
T
h
e
n
e
x
t
s
tag
e
in
v
o
l
v
es
an
aly
zin
g
th
e
d
i
f
f
er
en
ce
in
e
x
ec
u
tio
n
tim
e
b
etwe
en
f
ac
i
al
p
atter
n
r
ec
o
g
n
itio
n
p
r
o
ce
s
s
es
th
at
u
til
ize
th
e
th
r
ea
d
in
g
m
ec
h
an
is
m
an
d
th
o
s
e
th
at
d
o
n
o
t.
T
h
e
n
o
n
-
th
r
ea
d
e
d
a
p
p
r
o
ac
h
is
d
iv
id
ed
in
to
two
m
eth
o
d
s
:
i
)
s
ca
n
n
in
g
all
f
ac
ial
p
atter
n
s
e
x
h
au
s
tiv
ely
an
d
ii)
s
elec
tin
g
p
r
ev
io
u
s
ly
d
etec
ted
f
ac
ial
p
atter
n
p
o
s
itio
n
s
.
T
h
e
f
ir
s
t
m
eth
o
d
h
as
a
d
r
awb
ac
k
o
f
r
elativ
ely
lo
n
g
er
e
x
ec
u
tio
n
t
im
e,
as
th
e
s
y
s
tem
m
u
s
t
r
ec
o
g
n
ize
p
atter
n
s
s
eq
u
en
tially
f
r
o
m
s
tar
t
to
f
i
n
is
h
,
r
esu
ltin
g
i
n
laten
cy
a
n
d
r
ed
u
ce
d
r
ea
l
-
tim
e
p
er
f
o
r
m
an
ce
.
C
o
n
v
er
s
ely
,
th
e
s
ec
o
n
d
m
eth
o
d
is
m
o
r
e
tim
e
-
ef
f
icien
t
b
u
t
r
is
k
s
s
k
ip
p
in
g
ce
r
tain
f
ac
ial
p
atter
n
g
r
o
u
p
s
,
p
o
ten
tially
lo
wer
in
g
s
y
s
tem
ac
cu
r
ac
y
.
I
n
co
n
tr
ast,
th
e
u
s
e
o
f
th
e
th
r
ea
d
co
n
t
r
o
ller
m
ec
h
an
is
m
in
th
is
s
tu
d
y
s
ig
n
if
ican
tly
im
p
r
o
v
es
s
y
s
tem
ef
f
icien
cy
.
T
h
r
o
u
g
h
th
i
s
m
ec
h
an
is
m
,
m
u
ltip
le
f
ac
ial
p
atter
n
g
r
o
u
p
s
ca
n
b
e
ex
ec
u
ted
alm
o
s
t
s
im
u
ltan
e
o
u
s
ly
,
th
e
r
eb
y
ac
ce
ler
atin
g
co
m
p
u
tatio
n
with
o
u
t
co
m
p
r
o
m
is
in
g
ac
cu
r
ac
y
.
B
ased
o
n
th
e
r
esu
lts
p
r
esen
ted
in
Fig
u
r
e
8
,
t
h
e
n
o
n
-
th
r
ea
d
ed
ap
p
r
o
ac
h
(
a
)
p
r
o
d
u
ce
d
an
e
x
ec
u
ti
o
n
tim
e
o
f
0
.
2
7
1
s
,
wh
ile
ap
p
r
o
ac
h
(
b
)
r
eq
u
ir
ed
0
.
0
9
s
.
T
h
e
th
r
ea
d
ed
ap
p
r
o
a
ch
also
ac
h
iev
ed
an
ex
ec
u
tio
n
tim
e
o
f
0
.
0
9
s
.
Alth
o
u
g
h
th
e
tim
es
ap
p
ea
r
s
i
m
ilar
,
ap
p
r
o
ac
h
(
b
)
ac
tu
ally
p
r
o
ce
s
s
es
o
n
ly
o
n
e
-
th
ir
d
o
f
th
e
t
o
tal
d
ata
h
a
n
d
led
b
y
th
e
th
r
ea
d
ed
ap
p
r
o
ac
h
.
T
h
er
ef
o
r
e,
th
e
d
if
f
er
e
n
ce
in
ex
ec
u
tio
n
tim
e
b
etwe
en
th
e
two
ap
p
r
o
ac
h
es
d
em
o
n
s
tr
ates
a
p
r
o
ce
s
s
in
g
ef
f
icien
cy
im
p
r
o
v
em
en
t
o
f
0
.
6
7
%
,
co
n
f
ir
m
i
n
g
th
at
th
e
u
s
e
o
f
a
th
r
ea
d
i
n
g
m
ec
h
an
is
m
ca
n
s
ig
n
if
ican
tly
en
h
a
n
ce
th
e
p
e
r
f
o
r
m
an
ce
o
f
th
e
f
ac
ial
p
atter
n
r
ec
o
g
n
itio
n
s
y
s
tem
.
Af
ter
th
e
d
ataset
v
alid
atio
n
s
tag
e
was
co
m
p
leted
,
th
e
n
e
x
t
s
tep
was
to
test
th
e
f
ac
ial
p
atter
n
r
ec
o
g
n
itio
n
p
r
o
ce
s
s
u
s
in
g
th
e
ce
n
ter
-
s
ca
n
m
o
d
el
ap
p
r
o
ac
h
.
I
n
th
is
p
h
ase,
th
e
s
y
s
tem
was
ev
alu
ated
b
y
ap
p
ly
in
g
th
e
d
e
v
elo
p
e
d
m
eth
o
d
,
n
a
m
ely
th
e
W
NN
-
FR
A
[
2
3
]
,
[
2
4
]
,
wh
ich
o
p
e
r
ates
th
r
o
u
g
h
a
m
u
ltit
h
r
ea
d
co
n
tr
o
ller
m
ec
h
a
n
is
m
.
T
h
is
m
ec
h
an
is
m
lev
er
ag
es
th
e
p
r
o
ce
s
s
in
g
ca
p
ab
ilit
ies
o
f
an
AR
M
-
b
ased
m
icr
o
co
n
tr
o
ller
,
wh
ic
h
s
u
p
p
o
r
ts
th
e
ac
tiv
atio
n
o
f
m
u
ltip
le
th
r
ea
d
s
in
p
ar
allel
to
i
m
p
r
o
v
e
r
ec
o
g
n
itio
n
ef
f
icien
cy
.
T
h
e
test
in
g
p
r
o
ce
s
s
f
o
cu
s
ed
o
n
an
aly
zin
g
th
e
s
y
s
tem
’
s
p
er
f
o
r
m
an
ce
i
n
r
ec
o
g
n
izin
g
f
ac
ial
p
atter
n
s
f
r
o
m
th
e
d
ataset
ag
ain
s
t
p
r
im
ar
y
f
ac
ial
p
atter
n
s
o
b
tain
e
d
i
n
r
ea
l
tim
e.
T
h
e
test
d
ataset
co
n
s
is
ted
o
f
6
0
f
ac
ial
p
atter
n
s
f
r
o
m
th
e
f
ir
s
t
u
s
er
,
wh
ile
th
e
in
p
u
t
d
ata
co
m
p
r
is
ed
6
0
p
r
im
a
r
y
f
ac
ial
p
atter
n
s
co
llected
f
r
o
m
f
o
u
r
f
ac
ial
s
am
p
le
v
ar
iatio
n
s
,
as
p
r
esen
ted
in
T
ab
le
1
.
T
h
e
p
r
im
a
r
y
f
ac
ial
p
atter
n
s
s
er
v
ed
as
test
in
g
d
ata,
with
ea
ch
in
d
iv
id
u
al
h
a
v
in
g
1
5
f
ac
ial
p
atter
n
v
ar
iatio
n
s
.
T
h
e
ex
p
er
i
m
en
ts
wer
e
co
n
d
u
cted
u
s
in
g
th
r
ee
im
ag
e
-
ca
p
tu
r
e
d
is
tan
ce
s
1
0
cm
,
2
0
cm
,
a
n
d
3
0
cm
.
E
ac
h
d
ata
ca
p
tu
r
e
was
p
er
f
o
r
m
ed
r
an
d
o
m
ly
t
o
ass
es
s
th
e
s
y
s
tem
’
s
co
n
s
is
ten
cy
an
d
r
eliab
ilit
y
u
n
d
er
v
ar
y
in
g
c
o
n
d
itio
n
s
.
T
h
is
ap
p
r
o
ac
h
aim
ed
to
p
r
o
v
id
e
an
in
-
d
ep
th
ev
alu
atio
n
o
f
th
e
s
y
s
tem
’
s
ca
p
ab
ilit
y
to
ac
cu
r
ately
r
ec
o
g
n
ize
tar
g
et
f
ac
es
wh
ile
co
n
s
id
er
in
g
ch
an
g
es
in
d
is
tan
ce
an
d
Evaluation Warning : The document was created with Spire.PDF for Python.