I
nte
rna
t
io
na
l J
o
urna
l o
f
E
lect
rica
l a
nd
Co
m
p
ute
r
E
ng
in
ee
ring
(
I
J
E
CE
)
Vo
l.
6
,
No
.
6
,
Dec
em
b
er
201
6
,
p
p
.
281
8
~
2
8
2
7
I
SS
N:
2
0
8
8
-
8708
,
DOI
: 1
0
.
1
1
5
9
1
/
i
j
ec
e
.
v
6i
6
.
1
2
9
7
0
2818
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ia
e
s
jo
u
r
n
a
l.c
o
m/o
n
lin
e/in
d
ex
.
p
h
p
/I
JE
C
E
Face Recog
nition
Ba
sed
o
n Sy
m
m
e
t
rica
l H
a
lf
-
J
o
in Metho
d
u
sing
St
ereo
Visio
n Ca
m
era
E
dy
Wina
rno
1
,
Ag
us
H
a
rj
o
ko
2
,
Ania
t
i
M
urni Ar
y
m
urt
hy
3
,
E
di Wi
na
r
k
o
4
1
F
a
c
u
lt
y
o
f
In
f
o
r
m
a
ti
o
n
T
e
c
h
n
o
l
o
g
y
,
Un
iv
e
rsitas
S
ti
k
u
b
a
n
k
,
S
e
m
a
r
a
n
g
,
In
d
o
n
e
sia
2
,4
D
e
p
ar
t
m
e
n
t
o
f
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
a
n
d
E
l
e
c
t
r
o
n
i
c
s
,
U
n
i
v
e
r
s
i
t
a
s
G
a
d
j
a
h
Ma
d
a
,
Y
o
g
y
a
k
a
r
t
a
,
I
n
d
o
n
e
s
i
a
3
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
s
i
t
as
I
n
d
o
n
e
s
i
a
,
D
e
p
o
k
,
I
n
d
o
n
e
s
i
a
Art
icle
I
nfo
AB
ST
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
J
u
l
14
,
2
0
1
6
R
ev
i
s
ed
No
v
0
6
,
2
0
1
6
A
cc
ep
ted
No
v
1
9
,
2
0
1
6
T
h
e
m
a
in
p
ro
b
lem
in
f
a
c
e
r
e
c
o
g
n
it
io
n
sy
ste
m
b
a
se
d
o
n
h
a
lf
-
f
a
c
e
p
a
tt
e
rn
is
h
o
w
to
a
n
ti
c
i
p
a
te
p
o
se
s
a
n
d
il
lu
m
in
a
n
c
e
v
a
riatio
n
s
to
im
p
ro
v
e
re
c
o
g
n
it
io
n
ra
te.
T
o
so
lv
e
th
is
p
ro
b
lem
,
w
e
c
a
n
u
se
tw
o
len
s
es
o
n
ste
re
o
v
isio
n
c
a
m
e
ra
in
f
a
c
e
re
c
o
g
n
it
io
n
sy
ste
m
.
S
tere
o
v
isio
n
c
a
m
e
ra
h
a
s
le
f
t
a
n
d
rig
h
t
len
se
s
th
a
t
c
a
n
b
e
u
se
d
t
o
p
ro
d
u
c
e
a
2
D
im
a
g
e
o
f
e
a
c
h
len
s.
S
tere
o
v
is
i
o
n
c
a
m
e
ra
in
f
a
c
e
re
c
o
g
n
it
io
n
h
a
s
c
a
p
a
b
il
it
y
to
p
ro
d
u
c
e
tw
o
o
f
2
D
fa
c
e
i
m
a
g
e
s
w
it
h
a
d
iff
e
re
n
t
a
n
g
le.
Bo
th
a
n
g
le
o
f
th
e
f
a
c
e
ima
g
e
w
il
l
p
ro
d
u
c
e
a
d
e
tailed
im
a
g
e
o
f
th
e
f
a
c
e
a
n
d
b
e
tt
e
r
li
g
h
ti
n
g
lev
e
ls
o
n
e
a
c
h
o
f
th
e
lef
t
a
n
d
rig
h
t
len
se
s.
In
th
is
stu
d
y
,
w
e
p
ro
p
o
se
d
a
f
a
c
e
re
c
o
g
n
it
io
n
tec
h
n
iq
u
e
,
u
sin
g
2
len
s
o
n
a
ste
re
o
v
isio
n
c
a
m
e
ra
n
a
m
e
l
y
s
y
m
m
e
tri
c
a
l
h
a
l
f
-
jo
in
.
S
y
m
m
e
tri
c
a
l
h
a
l
f
-
jo
in
is
a
m
e
th
o
d
o
f
n
o
rm
a
li
z
in
g
th
e
im
a
g
e
o
f
th
e
f
a
c
e
d
e
tec
ti
o
n
o
n
e
a
c
h
o
f
th
e
lef
t
a
n
d
r
ig
h
t
len
se
s
in
ste
re
o
v
isio
n
c
a
m
e
ra
,
t
h
e
n
c
ro
p
p
in
g
a
n
d
m
e
rg
in
g
a
t
e
a
c
h
ima
g
e
.
T
e
sts
o
n
f
a
c
e
re
c
o
g
n
it
io
n
ra
te
b
a
se
d
o
n
th
e
v
a
riety
o
f
p
o
se
s
a
n
d
v
a
riatio
n
s
i
n
il
lu
m
in
a
ti
o
n
sh
o
w
s
th
a
t
th
e
s
y
m
m
e
tri
c
a
l
h
a
lf
-
jo
in
m
e
th
o
d
is
a
b
le
to
p
ro
v
id
e
a
h
ig
h
a
c
c
u
ra
c
y
o
f
f
a
c
e
re
c
o
g
n
it
io
n
a
n
d
c
a
n
a
n
ti
c
i
p
a
te
v
a
riatio
n
s
in
g
iv
e
n
p
o
se
a
n
d
il
lu
m
in
a
ti
o
n
v
a
riatio
n
s.
T
h
e
p
ro
p
o
se
d
m
o
d
e
l
is
a
b
le
to
p
ro
d
u
c
e
86%
-
9
7
%
re
c
o
g
n
it
io
n
ra
te
o
n
a
v
a
riet
y
o
f
p
o
se
s
a
n
d
v
a
riatio
n
s
in
a
n
g
les
b
e
tw
e
e
n
0
°
-
2
2
.
5
°
.
T
h
e
v
a
riatio
n
o
f
il
lu
m
in
a
n
c
e
m
e
a
su
re
d
u
sin
g
a
lu
x
m
e
t
e
r
c
a
n
re
su
lt
i
n
9
0
%
-
1
0
0
%
re
c
o
g
n
it
io
n
ra
te
f
o
r
th
e
c
a
teg
o
r
y
o
f
a
t
lea
st
d
im
li
g
h
ti
n
g
lev
e
ls
(a
b
o
v
e
1
0
lu
x
).
K
ey
w
o
r
d
:
Face
r
ec
o
g
n
itio
n
Half
-
j
o
in
Ster
eo
v
is
io
n
S
y
m
m
etr
ical
Co
p
y
rig
h
t
©
2
0
1
6
In
stit
u
te o
f
A
d
v
a
n
c
e
d
E
n
g
i
n
e
e
rin
g
a
n
d
S
c
ien
c
e
.
Al
l
rig
h
ts
re
se
rv
e
d
.
C
o
r
r
e
s
p
o
nd
ing
A
uth
o
r
:
E
d
y
W
i
n
ar
n
o
,
Facu
lt
y
o
f
I
n
f
o
r
m
a
tio
n
T
ec
h
n
o
lo
g
y
,
Un
i
v
er
s
ita
s
Sti
k
u
b
a
n
k
Se
m
ar
a
n
g
,
J
l.
T
r
i L
o
m
b
a
J
u
a
n
g
No
.
1
Sem
ar
an
g
-
5
0
2
4
1
,
J
aw
a
T
en
g
a
h
,
I
n
d
o
n
esia.
E
m
ail: e
d
y
w
in
@
ed
u
.
u
n
is
b
a
n
k
.
ac
.
id
1.
I
NT
RO
D
UCT
I
O
N
Face
d
etec
tio
n
i
s
a
s
y
s
te
m
u
s
ed
to
id
e
n
ti
f
y
a
h
u
m
a
n
f
a
ce
o
n
a
s
tatic
i
m
a
g
e
o
r
v
id
eo
f
r
o
m
a
ca
m
er
a
[
1
]
.
Face
d
etec
tio
n
s
y
s
te
m
is
u
s
ed
i
n
s
ev
er
al
s
t
u
d
ies
an
d
f
u
r
t
h
er
d
ev
e
lo
p
ed
in
to
m
u
ltip
le
f
ac
e
tr
ac
k
in
g
r
esear
ch
to
th
e
s
t
u
d
y
o
f
f
ac
e
r
ec
o
g
n
itio
n
[
2
-
3
]
.
T
h
e
r
esear
ch
f
ield
o
f
f
ac
e
r
ec
o
g
n
itio
n
i
s
a
r
esear
ch
tech
n
iq
u
e
th
at
co
m
b
i
n
es
f
ac
e
d
etec
tio
n
an
d
tr
ac
k
in
g
o
f
f
ac
e
to
k
n
o
w
th
e
id
en
tit
y
o
f
f
ac
es
d
etec
ted
b
y
co
m
p
ar
i
n
g
t
h
e
r
esu
lt
s
o
f
th
e
f
ac
e
i
m
a
g
e
tr
a
in
i
n
g
th
a
t
h
a
s
b
ee
n
s
to
r
ed
in
th
e
d
atab
ase
a
n
d
tes
tin
g
th
e
f
ac
ial
i
m
a
g
e
[
4
]
.
Gen
er
all
y
,
t
h
e
f
ac
e
r
ec
o
g
n
i
tio
n
s
y
s
te
m
h
as
5
-
s
tep
p
r
o
ce
s
s
to
id
en
tify
a
h
u
m
an
f
ac
e:
i
m
a
g
e
ac
q
u
is
itio
n
,
f
ac
e
d
etec
tio
n
,
n
o
r
m
aliza
t
io
n
,
f
ea
tu
r
e
ex
tr
ac
tio
n
an
d
clas
s
i
f
icatio
n
[5
].
I
n
a
f
ac
e
r
ec
o
g
n
itio
n
s
y
s
te
m
,
t
h
e
ar
r
an
g
e
m
e
n
t
o
f
th
e
i
n
ter
n
al
o
r
g
an
s
in
t
h
e
h
u
m
a
n
f
ac
e
i
s
u
s
ed
as
th
e
b
asis
f
o
r
d
etec
tio
n
an
d
f
ac
e
r
e
co
g
n
itio
n
[6
]
.
T
h
e
p
ar
ts
o
f
th
e
f
ac
e
u
s
ed
as t
h
e
b
asi
s
f
o
r
th
e
d
etec
tio
n
o
f
h
u
m
a
n
f
ac
e
i
m
a
g
e
ar
e
th
e
ar
ea
o
f
th
e
e
y
es,
n
o
s
e,
m
o
u
t
h
an
d
c
h
ee
k
s
.
P
ar
ts
o
f
th
e
o
r
g
an
o
n
th
e
f
ac
e
s
y
m
m
e
tr
y
f
o
r
m
a
v
er
tical
p
atter
n
o
n
th
e
f
ac
e.
P
atter
n
-
b
ased
f
ac
ia
l
s
y
m
m
etr
y
v
er
tical
a
x
i
s
w
o
u
ld
p
r
o
d
u
ce
h
al
f
a
f
ac
e
i
m
a
g
e
(
h
alf
-
f
ac
e)
o
f
th
e
le
f
t
a
n
d
r
ig
h
t
h
al
f
f
ac
e
i
m
a
g
es
t
h
at
h
a
v
e
s
i
m
ilar
p
ar
ts
if
t
h
e
m
ir
r
o
r
in
g
p
r
o
ce
s
s
is
d
o
n
e.
Fac
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
J
E
C
E
Vo
l.
6
,
No
.
6
,
Dec
em
b
er
2
0
1
6
: 2
8
1
8
–
2
8
2
7
2819
r
ec
o
g
n
itio
n
u
s
i
n
g
h
al
f
-
f
ac
e
i
m
ag
e
w
o
u
ld
p
r
o
d
u
ce
a
f
ac
ial
r
ec
o
g
n
itio
n
s
y
s
te
m
t
h
at
is
r
ela
tiv
el
y
s
i
m
ilar
to
th
e
u
s
e
o
f
f
u
ll
f
ac
e
i
m
ag
e
(
f
u
ll
-
f
a
ce
)
,
b
ec
au
s
e
th
e
f
ac
e
h
a
s
a
s
y
m
m
etr
ic
n
at
u
r
e
[
7
]
.
Var
iatio
n
s
in
t
h
e
d
ev
elo
p
m
e
n
t
o
f
m
o
d
els
o
f
h
a
lf
-
f
ac
e
i
m
ag
e
w
a
s
u
s
ed
to
an
tic
ip
ate
th
e
li
m
ited
n
u
m
b
er
an
d
v
ar
iet
y
o
f
m
o
d
els
o
f
f
ac
e
i
m
a
g
e
d
u
r
in
g
th
e
tr
ai
n
i
n
g
p
r
o
ce
s
s
[
8
].
T
h
e
s
tu
d
y
o
f
f
ac
e
r
ec
o
g
n
itio
n
b
ased
o
n
h
al
f
-
f
ac
e
i
m
ag
e
w
as
d
o
n
e
b
y
[
9
]
u
s
i
n
g
t
h
e
a
v
er
ag
e
-
h
al
f
-
f
ac
e.
T
h
is
r
esear
ch
r
ec
o
n
s
tr
u
cted
t
h
e
i
m
a
g
e
o
f
th
e
h
alf
-
f
ac
e
to
f
u
ll
-
f
ac
e
i
m
a
g
e
b
y
d
iv
id
i
n
g
th
e
in
itial
i
m
ag
e
i
n
to
t
w
o
eq
u
al
p
ar
ts
b
et
w
ee
n
t
h
e
l
ef
t
an
d
r
i
g
h
t.
T
h
e
p
r
o
ce
s
s
o
f
r
ev
er
s
i
n
g
t
h
e
co
lu
m
n
w
a
s
d
o
n
e
o
n
o
n
e
h
al
f
o
f
t
h
e
f
ac
e
i
m
a
g
e
(
le
f
t/
r
i
g
h
t)
an
d
p
er
f
o
r
m
s
an
a
v
er
ag
e
o
f
t
h
e
m
er
g
e
r
o
f
t
h
e
le
f
t
an
d
r
i
g
h
t i
m
a
g
es.
T
h
e
s
tu
d
y
t
h
en
w
as
co
n
tin
u
ed
i
n
f
ac
e
r
ec
o
g
n
itio
n
r
esear
ch
b
y
t
h
e
s
a
m
e
m
e
th
o
d
,
b
u
t a
p
p
lied
in
th
e
3
D
f
ac
e
i
m
a
g
e
[
10
].
Face
r
ec
o
g
n
it
io
n
r
esear
c
h
w
a
s
also
co
n
d
u
c
ted
b
y
[8
]
u
s
i
n
g
t
h
e
p
atter
n
o
f
h
al
f
-
f
ac
e
i
m
ag
e
i
n
th
e
tr
ain
i
n
g
p
r
o
ce
s
s
.
I
t
is
ar
g
u
ed
th
at
t
h
er
e
ar
e
f
o
u
r
s
tep
s
i
n
th
e
p
r
o
ce
s
s
o
f
f
ac
e
r
ec
o
g
n
it
io
n
.
T
h
e
f
ir
s
t
s
tep
is
to
p
r
o
d
u
ce
a
f
ac
e
i
m
a
g
e
s
y
m
m
etr
y
to
th
e
tr
ai
n
i
n
g
p
r
o
ce
s
s
.
T
h
e
f
ac
e
i
m
a
g
e
is
g
e
n
er
ated
b
y
t
h
e
p
r
o
ce
s
s
o
f
m
ir
r
o
r
in
g
t
h
e
i
m
ag
e
o
f
th
e
le
f
t
h
alf
-
f
ac
e
o
r
th
e
i
m
ag
e
o
f
t
h
e
r
ig
h
t
h
al
f
-
f
ac
e
o
f
th
e
f
r
o
n
tal
f
a
ce
p
o
s
itio
n
an
d
th
e
m
er
g
er
b
et
w
ee
n
t
h
e
m
.
T
h
e
s
ec
o
n
d
s
tep
is
th
e
tr
ai
n
in
g
p
r
o
ce
s
s
u
s
i
n
g
t
h
e
o
r
ig
i
n
al
f
ac
e
i
m
a
g
e
to
b
e
u
s
ed
in
t
h
e
class
i
f
icatio
n
p
r
o
ce
s
s
.
T
h
e
th
ir
d
s
tep
is
t
h
e
tr
ai
n
in
g
p
r
o
ce
s
s
o
f
th
e
f
ac
e
i
m
a
g
e
s
y
m
m
e
tr
y
to
co
n
ti
n
u
e
t
h
e
p
r
o
ce
s
s
o
f
clas
s
i
f
icatio
n
.
T
h
e
f
o
u
r
t
h
s
tep
i
s
to
u
s
e
a
s
co
r
e
v
alu
e
o
f
f
u
s
io
n
o
f
th
e
t
w
o
p
r
ev
io
u
s
clas
s
i
f
icatio
n
p
r
o
ce
s
s
to
b
e
u
s
ed
in
f
ac
e
r
ec
o
g
n
i
tio
n
.
A
f
ac
e
r
ec
o
g
n
itio
n
s
y
s
te
m
th
at
w
a
s
co
n
d
u
cted
b
y
[
8
]
w
a
s
f
u
r
t
h
er
d
ev
elo
p
ed
b
y
[
1
1
]
b
ased
o
n
th
e
p
atter
n
o
f
h
alf
-
f
ac
e
i
m
a
g
e.
T
h
e
d
if
f
er
en
ce
b
et
w
e
en
t
h
e
r
esea
r
ch
co
n
d
u
cted
b
y
[
11
]
an
d
th
at
co
n
d
u
cted
b
y
[
8
]
w
a
s
th
e
ad
d
itio
n
o
f
t
h
e
tr
ai
n
in
g
p
r
o
ce
s
s
o
n
th
e
le
f
t
h
al
f
o
f
th
e
f
ac
e
i
m
a
g
e
a
n
d
th
e
r
ig
h
t
h
alf
o
f
th
e
f
ac
e
i
m
a
g
e.
Fro
m
s
ev
er
al
r
esear
ch
e
s
th
a
t
h
av
e
b
ee
n
d
o
n
e,
th
er
e
ar
e
s
o
m
e
p
r
o
b
lem
s
in
a
f
ac
e
r
ec
o
g
n
itio
n
s
y
s
te
m
b
ased
o
n
a
p
atter
n
o
f
h
alf
-
f
ac
e
i
m
ag
e.
T
h
e
m
ai
n
p
r
o
b
le
m
is
a
v
ar
iatio
n
o
f
t
h
e
f
ac
e
p
o
s
e
t
o
p
r
o
d
u
ce
h
alf
-
f
ac
e
i
m
a
g
e
u
s
ed
f
o
r
f
ac
e
r
ec
o
g
n
itio
n
ac
cu
r
ac
y
.
T
h
e
n
e
x
t
p
r
o
b
le
m
is
t
h
e
v
ar
iat
io
n
o
f
ill
u
m
in
a
n
c
e
o
f
th
e
f
ac
e
i
m
a
g
e
d
ata
s
et
i
n
t
h
e
p
r
o
ce
s
s
o
f
f
a
ce
r
ec
o
g
n
itio
n
ac
c
u
r
ac
y
[
1
2
]
.
R
esi
s
ta
n
t
to
v
ar
iatio
n
s
i
n
f
a
ce
-
i
m
ag
e
p
o
s
e
an
d
illu
m
i
n
atio
n
v
ar
ia
tio
n
s
w
ill
a
f
f
ec
t
t
h
e
ac
c
u
r
ac
y
o
f
f
ac
e
r
ec
o
g
n
i
tio
n
r
esu
lts
.
B
o
th
o
f
th
e
m
ar
e
v
er
y
i
m
p
o
r
tan
t
w
h
e
n
a
f
ac
e
r
ec
o
g
n
itio
n
s
y
s
te
m
w
ill b
e
i
m
p
le
m
e
n
ted
in
t
h
e
r
eal
-
ti
m
e
f
ac
e
r
ec
o
g
n
itio
n
.
T
h
e
v
ar
iatio
n
s
o
f
p
o
s
es
i
n
f
ac
e
r
ec
o
g
n
itio
n
s
y
s
te
m
s
i
n
s
o
m
e
p
r
ev
io
u
s
s
t
u
d
ies
d
ep
en
d
ed
o
n
th
e
p
o
s
itio
n
o
f
th
e
f
r
o
n
tal
f
ac
e.
T
h
e
ac
c
u
r
ac
y
o
f
f
ac
e
r
ec
o
g
n
iti
o
n
w
ill
ar
g
u
ab
l
y
b
e
e
f
f
ec
ti
v
e
if
it
h
a
s
a
f
ac
e
p
o
s
e
p
o
s
itio
n
n
ea
r
0
°
d
ev
iat
io
n
f
r
o
m
t
h
e
f
r
o
n
tal
f
ac
e.
I
m
p
r
o
p
er
p
o
s
itio
n
o
f
th
e
f
ac
e
in
f
r
o
n
t
o
f
th
e
ca
m
er
a
w
ill
af
f
ec
t
th
e
s
h
ap
e
an
d
ch
ar
ac
ter
is
tics
o
f
h
al
f
-
f
ac
e
i
m
ag
e
a
n
d
th
e
i
m
a
g
e
o
f
t
h
e
m
er
g
er
o
n
th
e
tr
ain
i
n
g
p
r
o
ce
s
s
an
d
s
u
b
s
eq
u
e
n
t
te
s
ti
n
g
,
i
t
ca
u
s
es
a
p
o
s
s
ib
le
r
ed
u
ctio
n
in
f
a
ce
r
ec
o
g
n
itio
n
ac
c
u
r
ac
y
.
T
h
es
e
p
r
o
b
lem
s
r
eq
u
ir
e
s
o
lu
tio
n
s
to
an
ticip
ate
t
h
e
v
ar
i
et
y
o
f
p
o
s
es
th
at
o
f
ten
o
cc
u
r
s
o
n
th
e
f
ac
e
i
m
a
g
e
ac
q
u
is
it
io
n
p
r
o
ce
s
s
.
Sh
if
tin
g
o
f
in
co
r
r
ec
tn
es
s
o
f
t
h
e
f
ac
e
p
o
s
e
o
f
f
r
o
n
ta
l f
ac
e
p
o
s
itio
n
0
° at
t
h
e
ti
m
e
o
f
t
h
e
f
ac
e
i
m
ag
e
r
etr
i
ev
al
p
r
o
ce
s
s
ca
n
b
e
ass
is
ted
w
it
h
t
h
e
i
n
s
talla
tio
n
o
f
t
w
o
ca
m
er
as
i
n
p
ar
allel
w
i
th
a
ce
r
tain
a
n
g
le
in
o
r
d
er
to
in
cr
ea
s
e
th
e
f
le
x
ib
ili
t
y
o
f
f
ac
e
p
o
s
es.
Var
iatio
n
s
i
n
li
g
h
tin
g
le
v
els
wh
en
ca
p
t
u
r
in
g
a
f
ac
e
i
m
a
g
e
w
i
ll
also
af
f
ec
t
to
th
e
v
al
u
e
o
f
ac
cu
r
ac
y
o
n
a
f
ac
e
r
ec
o
g
n
i
tio
n
s
y
s
te
m
[
1
2
]
.
L
i
g
h
tin
g
lev
el
s
i
n
s
o
m
e
p
r
ev
io
u
s
s
tu
d
ies
d
ep
en
d
s
o
n
t
h
e
illu
m
i
n
atio
n
o
f
th
e
f
r
o
n
t
s
id
e
o
n
l
y
.
T
h
is
w
ill
af
f
e
ct
th
e
f
ac
ial
i
m
ag
e
an
d
li
g
h
t
in
g
d
etail
i
m
ag
e
o
f
a
f
ac
e,
b
ec
au
s
e
th
e
f
ac
e
h
as
a
n
u
n
e
v
e
n
g
eo
m
etr
ic
s
tr
u
ctu
r
e.
T
o
an
t
icip
ate
th
i
s
p
r
o
b
le
m
s
,
w
e
n
ee
d
a
b
alan
ce
d
li
g
h
ti
n
g
o
f
t
h
e
r
esp
ec
ti
v
e
le
f
t
an
d
r
ig
h
t
s
id
es
o
f
t
h
e
f
ac
e.
T
h
e
lig
h
tin
g
i
n
th
e
r
esp
ec
ti
v
e
lef
t
an
d
r
ig
h
t
s
id
es
o
f
th
e
f
ac
e
ca
n
also
b
e
co
n
d
u
cted
b
y
p
lacin
g
t
h
e
ca
m
er
a
o
n
ea
c
h
o
f
th
e
lef
t
an
d
r
i
g
h
t
s
id
es
o
f
th
e
f
a
ce
.
I
n
s
tallat
io
n
o
f
th
e
d
u
al
ca
m
er
as
o
n
th
e
f
ac
e
i
m
a
g
e
ac
q
u
is
i
tio
n
p
r
o
ce
s
s
w
ill
a
ls
o
p
o
ten
t
iall
y
p
r
o
d
u
ce
d
etailed
i
m
a
g
es
a
n
d
m
a
x
i
m
u
m
ill
u
m
in
a
tio
n
lev
els o
n
ea
ch
s
id
e.
I
n
th
i
s
s
tu
d
y
,
w
e
p
r
o
p
o
s
e
a
s
o
lu
tio
n
to
an
ticip
ate
v
ar
iatio
n
s
i
n
p
o
s
e
an
d
lig
h
tin
g
le
v
el
s
o
f
f
ac
e
i
m
a
g
es
u
s
i
n
g
a
s
ter
eo
v
i
s
io
n
ca
m
er
a.
Face
i
m
ag
e
ac
q
u
is
i
ti
o
n
g
e
n
er
ates
t
w
o
2
D
f
ac
e
i
m
ag
e
s
o
f
t
h
e
f
ac
e
d
etec
tio
n
r
es
u
lts
r
esp
ec
ti
v
el
y
lef
t
a
n
d
r
i
g
h
t
le
n
s
e
s
i
n
s
ter
eo
v
is
io
n
ca
m
er
a.
B
o
th
th
e
2
D
f
ac
e
i
m
a
g
e
r
eq
u
ir
es
th
e
elab
o
r
atio
n
p
r
o
ce
s
s
an
d
p
r
o
d
u
ce
a
f
ac
e
i
m
a
g
e
to
b
e
p
r
o
c
ess
ed
in
t
h
e
s
tep
o
f
n
o
r
m
aliza
t
io
n
,
ex
tr
ac
tio
n
a
n
d
class
i
f
icatio
n
.
T
h
is
s
t
u
d
y
p
r
o
p
o
s
es
a
n
e
w
m
e
th
o
d
to
p
r
o
ce
s
s
an
d
co
m
b
i
n
e
b
o
th
t
h
e
le
f
t
an
d
r
ig
h
t
f
ac
e
i
m
a
g
e
s
in
to
o
n
e
i
m
ag
e
t
h
at
h
as
h
i
g
h
q
u
alit
y
i
n
f
o
r
m
atio
n
o
n
f
ac
e
r
e
co
g
n
itio
n
s
y
s
te
m
.
T
h
e
p
r
o
ce
s
s
o
f
co
m
b
i
n
in
g
th
e
i
m
a
g
es
o
f
le
f
t
h
al
f
-
f
ac
e
a
n
d
r
ig
h
t
h
al
f
-
f
ac
e
is
co
n
d
u
cted
u
s
i
n
g
h
al
f
-
j
o
in
m
eth
o
d
.
Half
-
j
o
in
is
a
m
et
h
o
d
o
f
n
o
r
m
aliza
t
io
n
w
it
h
t
h
e
cr
o
p
p
in
g
p
r
o
ce
s
s
o
n
ea
ch
i
m
a
g
e
o
n
t
h
e
le
f
t le
n
s
a
n
d
th
e
r
i
g
h
t
len
s
.
C
r
o
p
p
in
g
p
r
o
ce
s
s
is
co
n
d
u
c
ted
o
n
th
e
b
asis
o
f
th
e
s
y
m
m
etr
ic
p
atter
n
o
f
th
e
f
ac
e
im
a
g
e
(
s
y
m
m
etr
ical
h
alf
-
j
o
in
)
.
T
h
e
h
alf
o
f
lef
t
f
ac
e
i
m
a
g
e
is
th
e
n
co
m
b
i
n
ed
w
it
h
t
h
e
h
al
f
o
f
r
ig
h
t
f
ac
e
i
m
a
g
e
,
w
h
ich
is
t
h
e
n
u
s
ed
as
th
e
i
n
p
u
t
to
t
h
e
p
r
o
ce
s
s
o
f
ex
tr
ac
tio
n
a
n
d
class
if
icatio
n
.
2.
RE
S
E
ARCH
M
E
T
H
O
D
T
h
e
p
r
o
p
o
s
ed
m
e
th
o
d
i
n
t
h
is
s
tu
d
y
is
a
m
o
d
el
o
f
f
ac
e
r
ec
o
g
n
i
tio
n
u
s
in
g
h
al
f
-
j
o
in
n
o
r
m
a
l
izatio
n
o
n
s
ter
eo
v
is
io
n
ca
m
er
a
as
a
f
ac
e
i
m
ag
e
ac
q
u
is
it
io
n
.
T
h
e
m
et
h
o
d
em
p
lo
y
ed
s
e
v
er
al
p
h
ase
s
o
f
th
e
p
r
o
ce
s
s
o
f
f
ac
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
J
E
C
E
I
SS
N:
2
0
8
8
-
8708
F
a
ce
R
ec
o
g
n
itio
n
B
a
s
ed
o
n
S
y
mme
tr
ica
l H
a
lf
-
Jo
in
Meth
o
d
u
s
in
g
S
tereo
V
is
io
n
C
a
mera
(
E
d
y
W
in
a
r
n
o
)
2820
r
ec
o
g
n
itio
n
:
f
ac
e
d
etec
tio
n
,
n
o
r
m
aliza
tio
n
,
f
ea
tu
r
e
e
x
tr
ac
tio
n
an
d
clas
s
i
f
icatio
n
.
No
r
m
aliza
t
io
n
is
th
e
p
h
ase
o
f
ac
q
u
is
itio
n
i
m
a
g
e
p
r
o
ce
s
s
in
g
t
h
at
w
ill
b
e
u
s
ed
in
v
ar
io
u
s
s
te
p
s
o
f
f
ea
tu
r
e
ex
tr
ac
tio
n
.
Feat
u
r
e
ex
tr
ac
tio
n
is
th
e
p
r
o
ce
s
s
o
f
r
ed
u
ci
n
g
t
h
e
d
i
m
e
n
s
io
n
o
f
th
e
i
m
ag
e.
C
la
s
s
i
f
ic
at
io
n
i
s
t
h
e
p
h
a
s
e
to
m
atc
h
t
h
e
c
h
ar
ac
ter
is
tic
s
o
f
i
m
a
g
e
b
ased
o
n
th
e
ch
ar
ac
ter
i
s
tics
o
f
t
h
e
i
m
ag
e
t
h
at
h
as
b
ee
n
tr
ain
ed
.
I
n
s
h
o
r
t,
th
e
p
r
o
p
o
s
ed
p
r
o
ce
s
s
o
f
f
ac
e
r
ec
o
g
n
itio
n
ca
n
b
e
s
ee
n
i
n
Fi
g
u
r
e
1
.
Fig
u
r
e
1
.
P
r
o
p
o
s
ed
Pr
o
ce
s
s
o
f
Face
R
ec
o
g
n
itio
n
I
n
th
e
cu
r
r
e
n
t
s
t
u
d
y
,
i
m
ag
e
a
cq
u
is
itio
n
w
as
co
n
d
u
c
ted
to
ca
p
tu
r
e
h
u
m
a
n
f
ac
e
i
m
a
g
es
u
s
ed
as
th
e
in
f
o
r
m
atio
n
o
n
t
h
e
f
ac
e
d
etec
t
io
n
p
r
o
ce
s
s
m
a
k
i
n
g
u
s
e
o
f
t
h
e
s
ter
eo
v
is
io
n
ca
m
er
a
w
it
h
t
wo
len
s
es
o
n
t
h
e
le
f
t
an
d
r
ig
h
t.
Face
i
m
ag
e
w
as
ta
k
en
f
r
o
m
f
r
o
n
tal
v
ie
w
w
i
t
h
a
d
ev
iatio
n
to
th
e
ca
m
er
a
ab
o
u
t
1
5
°
o
n
th
e
x
,
y
an
d
z.
Face
d
etec
tio
n
w
as
p
er
f
o
r
m
ed
o
n
ea
c
h
o
f
t
h
e
le
f
t
an
d
r
i
g
h
t
l
en
s
e
s
o
f
a
s
ter
eo
v
i
s
io
n
ca
m
er
a.
T
h
e
i
m
a
g
e
o
f
t
h
e
f
ac
e
o
n
ea
ch
le
n
s
w
a
s
m
ar
k
ed
w
it
h
t
h
e
co
o
r
d
in
ates o
f
t
h
e
r
eg
io
n
o
f
in
ter
est (
R
o
I
)
o
f
f
ac
e
i
m
ag
e.
Face
d
etec
tio
n
p
r
o
ce
s
s
w
as
p
e
r
f
o
r
m
ed
u
s
in
g
t
h
e
Haa
r
C
a
s
ca
d
e
C
lass
i
f
ier
[
1
3
]
to
d
eter
m
i
n
e
th
e
x
an
d
y
co
o
r
d
in
ates
o
f
t
h
e
u
p
p
er
lef
t
co
r
n
er
o
f
th
e
i
m
ag
e
o
f
th
e
o
b
j
ec
t,
th
e
w
id
t
h
(
w
)
a
n
d
h
ei
g
h
t
(
h
)
o
f
i
m
a
g
e
o
f
t
h
e
o
b
j
ec
t.
Du
r
in
g
th
e
tr
ai
n
i
n
g
p
r
o
ce
s
s
,
R
o
I
o
f
i
m
ag
e
w
a
s
n
o
r
m
alize
d
t
h
en
s
to
r
ed
as
tr
ain
i
n
g
d
ata.
At
th
e
te
s
ti
n
g
p
r
o
ce
s
s
,
R
o
I
o
f
th
e
i
m
a
g
e
w
as
n
o
r
m
alize
d
an
d
test
ed
a
n
d
th
e
n
s
to
r
ed
as test d
ata.
R
o
I
o
f
f
ac
e
d
etec
tio
n
is
a
r
ec
t
an
g
u
lar
ar
ea
in
w
h
ich
t
h
e
ca
lc
u
latio
n
o
f
t
h
e
co
o
r
d
in
ates
o
f
f
ac
e
i
m
a
g
e
s
)
,
(
0
0
y
x
,
)
,
(
0
1
y
x
,
)
,
(
1
0
y
x
an
d
)
,
(
1
1
y
x
ca
n
b
e
u
s
ed
to
illu
s
tr
ate
a
f
ac
e
im
a
g
e
to
b
e
p
r
o
ce
s
s
ed
in
th
e
tr
ain
in
g
an
d
test
i
n
g
.
Me
an
w
h
ile,
)
,
(
0
0
y
x
co
o
r
d
in
ate
p
o
in
t
i
s
al
w
a
y
s
p
ar
alle
l
to
o
n
e
co
lu
m
n
w
it
h
t
h
e
co
o
r
d
in
ates
o
f
)
,
(
1
0
y
x
.
C
o
o
r
d
in
ate
p
o
in
t
)
,
(
1
0
y
x
is
al
w
a
y
s
p
ar
allel
to
a
lin
e
w
ith
t
h
e
co
o
r
d
in
ates
o
f
)
,
(
1
1
y
x
.
C
o
o
r
d
in
ate
p
o
in
ts
)
,
(
1
0
y
x
an
d
)
,
(
1
1
y
x
w
il
l
v
ar
y
d
ep
en
d
in
g
o
n
th
e
w
id
th
o
f
th
e
f
ac
e
i
m
a
g
e
(
w
)
w
h
ic
h
d
etec
ted
an
d
ca
lcu
lated
f
r
o
m
p
o
in
t
0
x
.
C
o
o
r
d
in
ate
p
o
in
t
s
)
,
(
1
0
y
x
an
d
)
,
(
1
1
y
x
w
ill
a
ls
o
v
ar
y
d
ep
en
d
in
g
o
n
t
h
e
h
e
i
g
h
t
i
m
a
g
e
o
f
a
f
ac
e
(
h
)
ca
lcu
la
te
d
f
r
o
m
p
o
in
t
0
y
.
Dete
r
m
i
n
atio
n
o
f
R
o
I
o
f
f
ac
e
i
m
a
g
e
o
n
t
h
e
f
ac
e
d
etec
tio
n
p
r
o
ce
s
s
is
s
ee
n
i
n
Fi
g
u
r
e
2
.
Fig
u
r
e
2
.
R
o
I
o
f
Face
I
m
ag
e
B
a
c
k
g
r
o
u
n
d
F
a
c
e
i
mag
e
A
)
,
(
0
0
y
x
B
)
,
(
0
1
y
x
C
)
,
(
1
0
y
x
w
h
D
)
,
(
1
1
y
x
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
J
E
C
E
Vo
l.
6
,
No
.
6
,
Dec
em
b
er
2
0
1
6
: 2
8
1
8
–
2
8
2
7
2821
T
h
e
p
r
o
ce
s
s
o
f
ca
lcu
lati
n
g
t
h
e
R
o
I
co
o
r
d
in
ates
o
f
f
ac
e
i
m
ag
e
w
as
p
er
f
o
r
m
ed
o
n
th
e
co
o
r
d
in
ate
p
o
in
t
)
,
(
0
0
y
x
.
T
h
e
ca
lcu
latio
n
w
as
p
er
f
o
r
m
ed
o
n
co
n
s
id
er
i
n
g
th
e
v
al
u
e
o
f
w
an
d
h
o
f
th
e
d
etec
ted
f
ac
e
i
m
a
g
e.
T
o
d
eter
m
in
e
t
h
e
co
o
r
d
in
ates
o
f
all
R
o
I
o
f
f
ac
e
i
m
ag
e
w
a
s
o
b
tain
ed
f
r
o
m
th
e
ca
lc
u
latio
n
o
f
th
e
m
ai
n
co
o
r
d
in
ate
v
alu
e
s
)
,
(
0
0
y
x
.
T
h
e
ca
lcu
latio
n
to
d
et
er
m
in
e
t
h
e
p
o
in
t
1
x
an
d
1
y
is
s
h
o
w
n
in
E
q
u
atio
n
s
(
1
)
an
d
(
2
)
.
1
x
=
0
x
+(
w
-
1)
(
1
)
1
y
=
0
y
+(
h
-
1)
(
2
)
T
o
d
eter
m
i
n
e
th
e
co
o
r
d
in
ates
R
o
I
o
f
f
ac
e
i
m
a
g
e
s
A
,
B
,
C
an
d
D
ca
n
b
e
d
o
n
e
u
s
in
g
E
q
u
ati
o
n
(
3
)
,
(
4
)
,
(
5
)
an
d
(
6
)
.
A=
)
,
(
0
0
y
x
(
3
)
B
=(
0
x
+(
w
-
1
)
)
,
0
y
)
(
4
)
C
=(
0
x
,
0
y
+(
h
-
1
)
)
(
5
)
D=
(
0
x
+(
w
-
1
)
,
0
y
+(
h
-
1
)
)
(
6
)
Dete
cted
f
ace
i
m
a
g
es
w
er
e
t
h
en
p
r
o
ce
s
s
ed
u
s
in
g
n
o
r
m
aliza
t
io
n
.
T
h
e
n
o
r
m
aliza
tio
n
p
r
o
ce
s
s
co
n
s
i
s
t
s
o
f
t
w
o
s
ta
g
es
i
m
ag
e
p
r
o
ce
s
s
in
g
:
p
r
ep
r
o
ce
s
s
in
g
s
ta
g
e
an
d
h
alf
-
j
o
in
s
tag
e.
Se
v
er
al
m
e
th
o
d
s
u
s
ed
in
t
h
e
p
r
ep
r
o
ce
s
s
in
g
i
n
clu
d
e
cr
o
p
p
in
g
,
R
GB
-
Gr
a
y
,
r
esizi
n
g
,
an
d
c
o
n
tr
ast
-
b
r
ig
h
t
n
e
s
s
ad
j
u
s
t
m
e
n
t
u
s
i
n
g
th
e
h
i
s
to
g
r
a
m
eq
u
aliza
tio
n
[
1
4
]
.
Pre
p
r
o
ce
s
s
in
g
m
et
h
o
d
s
u
s
ed
in
t
h
is
r
esea
r
ch
w
er
e
co
n
d
u
cted
to
i
m
p
r
o
v
e
th
e
s
h
ar
p
n
es
s
o
f
th
e
i
m
ag
e
a
n
d
to
an
ticip
ate
t
h
e
v
ar
iatio
n
s
o
f
ill
u
m
in
a
n
ce
o
n
f
ac
e
i
m
ag
e
ca
p
t
u
r
in
g
p
r
o
ce
s
s
.
T
h
e
n
ex
t
s
tag
e
w
a
s
to
p
er
f
o
r
m
a
f
ac
e
i
m
a
g
e
p
r
o
ce
s
s
in
g
u
s
i
n
g
th
e
s
y
m
m
etr
ica
l
h
al
f
-
j
o
in
.
S
y
m
m
etr
ical
h
al
f
-
j
o
in
i
m
ag
e
w
a
s
f
u
r
t
h
er
p
r
o
ce
s
s
ed
u
s
in
g
f
ea
tu
r
e
e
x
tr
ac
tio
n
an
d
c
lass
if
ic
atio
n
.
W
e
u
s
ed
t
h
e
3
-
lev
el
w
a
v
elet
P
C
A
(
P
r
in
cip
al
C
o
m
p
o
n
en
t
An
al
y
s
i
s
)
an
d
M
ah
alan
o
b
is
d
is
ta
n
ce
m
et
h
o
d
a
s
f
ea
t
u
r
e
e
x
tr
ac
tio
n
an
d
class
i
f
icat
io
n
[
1
5
].
2
.
1
.
Sy
mm
et
rica
l
H
a
lf
-
J
o
in
Half
-
j
o
in
m
et
h
o
d
is
in
s
p
ir
ed
b
y
m
o
d
els
o
f
th
e
h
u
m
a
n
p
er
s
p
ec
tiv
e
u
s
in
g
t
w
o
e
y
es
to
s
ee
an
d
id
en
tify
a
p
er
s
o
n
.
B
o
th
e
y
es
i
n
h
u
m
a
n
s
ar
e
t
h
en
r
ep
r
esen
ted
u
s
in
g
a
d
u
al
ca
m
er
a
v
is
io
n
th
at
h
a
s
l
ef
t
a
n
d
r
i
g
h
t
le
n
s
e
s
.
T
h
e
t
w
o
i
m
a
g
es
i
n
f
ac
e
r
ec
o
g
n
itio
n
is
t
h
e
ap
p
licatio
n
o
f
th
e
c
h
ar
ac
ter
is
tic
s
o
f
th
e
t
w
o
d
if
f
er
en
t
v
ie
ws
s
i
m
u
lt
a
n
eo
u
s
l
y
.
T
h
i
s
m
e
th
o
d
co
m
b
i
n
es
th
e
i
m
a
g
e
o
f
t
h
e
le
f
t
h
alf
-
f
ac
e
a
n
d
th
e
r
i
g
h
t
h
al
f
-
f
ac
e
in
to
a
n
i
m
ag
e
th
at
i
s
r
ea
d
y
to
b
e
e
x
tr
ac
ted
.
T
h
e
cu
r
r
en
t
s
t
u
d
y
u
s
ed
th
e
t
y
p
e
o
f
h
al
f
-
j
o
in
m
et
h
o
d
n
a
m
el
y
s
y
m
m
etr
ical
h
alf
-
j
o
in
.
S
y
m
m
etr
ical
h
a
lf
-
j
o
in
m
eth
o
d
cu
ts
t
h
e
i
m
a
g
e
in
t
h
e
ce
n
ter
o
f
t
h
e
f
ac
e
a
n
d
p
r
o
d
u
ce
s
an
o
th
e
r
i
m
ag
e
o
f
th
e
s
a
m
e
w
id
t
h
.
Hal
f
o
f
th
e
f
a
ce
i
m
a
g
e
o
f
th
e
le
f
t
s
id
e
o
f
t
h
e
lef
t le
n
s
w
as
t
h
en
co
m
b
i
n
ed
w
it
h
h
al
f
o
f
th
e
f
ac
e
i
m
a
g
e
o
f
th
e
r
ig
h
t
s
id
e
o
f
th
e
r
ig
h
t
le
n
s
.
T
h
e
f
ac
e
i
m
a
g
es
wh
en
co
m
b
i
n
ed
r
esu
lted
in
an
o
t
h
er
i
m
a
g
e
w
i
th
t
h
e
w
id
t
h
eq
u
al
to
th
a
t o
f
its
o
r
i
g
i
n
al
i
m
ag
e.
S
y
m
m
etr
ic
h
al
f
-
j
o
in
i
s
a
m
et
h
o
d
o
f
d
iv
id
in
g
th
e
w
id
th
o
f
t
h
e
f
ac
e
i
m
a
g
e
o
n
t
h
e
le
f
t
a
n
d
r
i
g
h
t
le
n
s
es
o
n
a
ca
m
er
a
t
h
at
h
as
th
e
s
a
m
e
w
id
t
h
.
T
h
e
f
ir
s
t
s
tep
i
s
to
d
et
er
m
in
e
th
e
w
id
th
o
f
t
h
e
f
ac
e
i
m
ag
e
(
w
/
w
id
t
h
)
in
p
ix
els.
T
h
e
m
id
p
o
in
t
o
f
t
h
e
f
a
ce
i
m
a
g
e
(
c/
ce
n
ter
)
is
o
b
tai
n
e
d
b
y
d
i
v
id
i
n
g
th
e
w
id
th
o
f
t
h
e
f
ac
e
i
m
a
g
e
i
n
to
2
p
ar
ts
.
T
h
e
im
a
g
e
o
f
th
e
f
ac
e
o
f
th
e
lef
t
le
n
s
is
t
h
e
p
o
in
t
o
f
in
ter
s
ec
tio
n
o
f
t
h
e
f
ar
lef
t
(
x
=
0
)
u
p
to
o
n
e
p
ix
el
b
ef
o
r
e
th
e
m
id
p
o
in
t o
f
th
e
i
m
a
g
e
(
x
=c
-
1
)
.
T
h
e
im
a
g
e
o
f
t
h
e
f
ac
e
o
f
th
e
r
ig
h
t le
n
s
i
s
th
e
p
o
i
n
t o
f
i
n
ter
s
ec
tio
n
o
f
th
e
m
id
p
o
in
t
o
f
t
h
e
i
m
a
g
e
(
x
=
c)
u
p
to
th
e
r
ig
ht
-
m
o
s
t
p
o
in
t
o
f
th
e
i
m
ag
e
(
x
=
w
-
1
)
.
T
h
e
co
m
b
in
ed
f
ac
e
i
m
a
g
e
is
th
e
i
m
ag
e
o
f
th
e
le
f
t
h
alf
o
f
th
e
lef
t le
n
s
o
f
t
h
e
ca
m
er
a
a
n
d
th
e
i
m
a
g
e
o
f
t
h
e
r
ig
h
t
h
al
f
o
f
t
h
e
r
ig
h
t le
n
s
ca
m
er
a.
I
f
s
h
j
l
e
f
t
is
t
h
e
p
r
o
ce
s
s
o
f
cu
t
tin
g
ar
ea
o
f
t
h
e
i
m
a
g
e
o
f
t
h
e
le
f
t
h
alf
o
f
t
h
e
le
f
t
le
n
s
o
f
t
h
e
ca
m
er
a,
s
h
j
r
i
g
h
t
is
t
h
e
p
r
o
ce
s
s
o
f
c
u
tt
in
g
ar
ea
o
f
t
h
e
i
m
ag
e
o
f
t
h
e
r
ig
h
t
h
al
f
o
f
th
e
r
ig
h
t
le
n
s
o
f
t
h
e
ca
m
e
r
a,
C
is
t
h
e
m
id
p
o
in
t
(
ce
n
ter
)
an
d
w
is
th
e
w
id
t
h
r
esu
l
ted
f
r
o
m
d
etec
tio
n
o
f
ea
ch
i
m
a
g
e
o
f
th
e
le
f
t
an
d
r
ig
h
t
len
s
es,
th
e
n
th
e
v
al
u
e
o
f
s
h
j
is
ca
lcu
la
ted
an
d
s
h
o
w
n
in
eq
u
a
tio
n
(
7
)
-
(
1
0
)
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
J
E
C
E
I
SS
N:
2
0
8
8
-
8708
F
a
ce
R
ec
o
g
n
itio
n
B
a
s
ed
o
n
S
y
mme
tr
ica
l H
a
lf
-
Jo
in
Meth
o
d
u
s
in
g
S
tereo
V
is
io
n
C
a
mera
(
E
d
y
W
in
a
r
n
o
)
2822
s
h
j
l
e
f
t
=
1
0
c
x
c
(
7
)
s
h
j
r
i
g
h
t
=
1
w
c
x
w
(
8
)
s
h
j
=
s
h
j
l
e
f
t
+
s
h
j
r
i
g
h
t
(
9
)
s
h
j
=
1
0
c
x
c
+
1
w
c
x
w
(
1
0
)
2
.
2
.
Ro
I
o
f
H
a
lf
-
F
a
ce
I
m
a
g
e
S
y
m
m
etr
ic
h
al
f
-
j
o
in
h
as
t
w
o
ad
d
itio
n
al
R
o
I
d
eter
m
i
n
i
n
g
s
tep
s
to
p
r
o
d
u
ce
h
alf
-
f
ac
e
ar
ea
f
r
o
m
th
e
le
f
t
len
s
a
n
d
h
a
lf
-
f
ac
e
ar
ea
f
r
o
m
t
h
e
r
ig
h
t
le
n
s
o
n
t
h
e
s
ter
eo
v
is
i
o
n
ca
m
er
a.
R
o
I
o
f
h
al
f
-
f
ac
e
is
v
er
tical
r
ec
tan
g
u
lar
ar
ea
.
R
o
I
o
f
le
f
t
h
al
f
-
f
ac
e
a
n
d
r
ig
h
t
h
al
f
-
f
ac
e
h
as
t
h
e
s
a
m
e
w
id
t
h
a
n
d
h
ei
g
h
t.
T
h
e
ca
lcu
lat
io
n
o
f
t
h
e
co
o
r
d
in
ates
o
f
t
h
e
to
p
lef
t,
to
p
r
ig
h
t,
b
o
tto
m
lef
t
an
d
b
o
tto
m
r
i
g
h
t
o
f
h
al
f
-
f
ac
e
i
m
a
g
e
i
s
co
n
d
u
cted
in
th
e
lef
t
an
d
r
ig
h
t
o
f
f
ac
e
i
m
a
g
es
o
n
e
ac
h
le
n
s
o
n
th
e
s
ter
eo
v
i
s
io
n
ca
m
er
a.
T
h
e
co
o
r
d
in
ates
,
w
id
th
an
d
h
ei
g
h
t
o
f
t
h
e
i
m
a
g
e
u
s
ed
in
th
e
ca
lc
u
lat
io
n
ar
e
th
e
co
o
r
d
in
ates
o
f
th
e
R
o
I
d
u
r
in
g
f
ac
e
d
etec
tio
n
p
r
o
ce
s
s
.
T
h
e
w
id
th
o
f
h
al
f
-
f
ac
e
(
c)
is
a
h
al
f
o
f
t
h
e
w
id
th
o
f
th
e
f
ac
e
i
m
ag
e
(
w
)
.
2.
2
.
1
.
Ro
I
o
f
L
ef
t
H
a
lf
-
F
a
ce
I
m
a
g
es
C
o
o
r
d
in
ates
)
,
(
0
0
y
x
an
d
)
,
(
1
0
y
x
ar
e
s
i
m
ilar
t
o
th
e
co
o
r
d
in
ates
in
th
e
p
i
x
el
co
lu
m
n
o
f
R
o
I
i
m
a
g
e
i
n
f
ac
e
d
etec
tio
n
.
C
o
o
r
d
in
ates
)
,
(
0
1
y
x
an
d
)
,
(
1
1
y
x
ar
e
also
s
i
m
ilar
to
t
h
e
p
ix
el
co
lu
m
n
co
o
r
d
in
ates,
b
u
t
t
h
eir
p
o
s
itio
n
s
d
ep
en
d
o
n
th
e
v
a
lu
e
o
f
t
h
e
ce
n
ter
(
c)
an
d
w
id
th
(
w
)
o
f
t
h
e
f
ac
e
i
m
a
g
e.
C
o
o
r
d
in
ates
)
,
(
0
1
y
x
an
d
)
,
(
1
1
y
x
ar
e
d
eter
m
i
n
ed
d
ep
en
d
i
n
g
o
n
t
h
e
m
id
p
o
in
t
(
c)
i
n
t
h
e
i
m
ag
e,
s
o
th
e
p
o
in
t
1
x
ca
n
b
e
ca
lcu
lated
.
Me
an
w
h
ile,
1
y
is
d
eter
m
i
n
ed
d
ep
en
d
in
g
o
n
th
e
v
alu
e
o
f
w
id
t
h
(
h
)
o
f
th
e
i
m
a
g
e.
R
o
I
o
f
lef
t
h
al
f
-
f
ac
e
i
m
ag
e
o
n
th
e
le
f
t le
n
s
is
i
lu
s
tr
ated
in
Fig
u
r
e
3
.
Fig
u
r
e
3
.
R
o
I
o
f
L
e
f
t H
al
f
-
Fac
e
I
m
a
g
e
o
n
t
h
e
L
ef
t
-
L
e
n
s
R
o
I
o
f
lef
t
h
al
f
-
f
ac
e
i
m
a
g
es
o
n
th
e
lef
t
le
n
s
is
ca
lc
u
late
d
b
ased
o
n
th
e
co
o
r
d
in
ates
o
f
p
o
in
t
A
)
,
(
0
0
y
x
,
w
a
n
d
h
.
R
o
I
o
f
le
f
t
h
a
lf
-
f
ac
e
i
m
a
g
e
is
ca
l
cu
lated
u
s
i
n
g
t
h
e
m
id
p
o
in
t
o
f
t
h
e
f
ac
e
i
m
a
g
e
(
c)
u
s
i
n
g
E
q
u
atio
n
(
1
1
)
.
c
=
½
w
(
1
1
)
Nex
t,
1
x
d
an
1
y
is
ca
lcu
lated
u
s
in
g
E
q
u
atio
n
(
1
2
)
d
an
(
1
3
)
.
B
a
c
k
g
r
o
u
n
d
F
a
c
e
i
mag
e
A
)
,
(
0
0
y
x
E
)
,
(
0
1
y
x
C
)
,
(
1
0
y
x
)
F
)
,
(
1
1
y
x
w
h
c
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
J
E
C
E
Vo
l.
6
,
No
.
6
,
Dec
em
b
er
2
0
1
6
: 2
8
1
8
–
2
8
2
7
2823
1
x
=
0
x
+(
c
-
1)
(
1
2
)
1
y
=
0
y
+(
h
-
1)
(
1
3
)
A,
E,
C
a
n
d
F a
s
R
o
I
o
f
lef
t h
a
lf
-
f
ac
e
i
s
ca
lcu
lated
u
s
i
n
g
E
q
u
atio
n
(
1
4
)
-
(
1
7
)
.
A=
)
,
(
0
0
y
x
(
1
4
)
E
=(
0
x
+(
c
-
1
)
)
,
0
y
)
(
1
5
)
C
=(
0
x
,
0
y
+(
h
-
1
)
)
(
1
6
)
F=(
0
x
+(
c
-
1
)
,
0
y
+(
h
-
1
)
)
(
1
7
)
A,
E,
C
a
n
d
F a
r
e
th
en
co
m
b
i
n
ed
w
it
h
a
r
ig
h
t h
al
f
-
f
ac
e
o
n
th
e
r
ig
h
t le
n
s
a
s
s
ymm
et
r
ica
l h
a
lf
-
jo
in
.
2
.
2
.
2
.
Ro
I
o
f
Rig
ht
H
a
lf
-
F
a
c
e
I
m
a
g
es
Si
m
i
lar
to
th
e
d
eter
m
i
n
atio
n
o
f
R
o
I
o
f
lef
t
h
al
f
f
ac
e
o
n
t
h
e
lef
t
le
n
s
,
)
,
(
0
0
y
x
an
d
)
,
(
1
0
y
x
in
R
o
I
o
f
r
ig
h
t
h
al
f
-
f
ac
e
o
n
th
e
r
i
g
h
t
l
en
s
ar
e
s
i
m
ilar
to
th
e
co
o
r
d
in
ates
i
n
t
h
e
p
ix
el
co
l
u
m
n
o
f
R
o
I
i
m
ag
e
i
n
f
ac
e
d
etec
tio
n
.
)
,
(
0
1
y
x
an
d
)
,
(
1
1
y
x
ar
e
th
e
m
id
p
o
in
t
(
c)
f
r
o
m
t
h
e
w
id
th
o
f
i
m
ag
e
(
w
)
.
)
,
(
0
2
y
x
an
d
)
,
(
1
2
y
x
ar
e
also
s
i
m
ilar
w
it
h
th
e
p
i
x
el
co
lu
m
n
co
o
r
d
in
ates,
b
u
t
t
h
eir
p
o
s
itio
n
d
ep
en
d
s
o
n
w
id
t
h
(
w
)
an
d
h
ei
g
h
t
(
h
)
f
r
o
m
th
e
co
o
r
d
in
ates
)
,
(
0
0
y
x
d
an
)
,
(
1
0
y
x
.
R
o
I
o
f
r
i
g
h
t
h
al
f
-
f
ac
e
i
m
a
g
e
o
n
t
h
e
r
ig
h
t le
n
s
is
il
lu
s
tr
ated
in
Fi
g
u
r
e
4
.
Fig
u
r
e
4
.
R
o
I
o
f
R
ig
h
t H
al
f
-
Fa
ce
o
n
th
e
R
i
g
h
t
L
en
s
R
o
I
o
f
r
ig
h
t
h
al
f
-
f
ac
e
i
m
a
g
es
o
n
th
e
r
ig
h
t
le
n
s
w
a
s
ca
lc
u
lat
ed
b
ased
o
n
th
e
co
o
r
d
in
ates
o
f
p
o
in
t
A
)
,
(
0
0
y
x
,
w
,
h
an
d
c.
Fu
r
t
h
er
m
o
r
e,
R
o
I
o
f
r
i
g
h
t
h
alf
-
f
ac
e
i
m
a
g
e
w
as
ca
lcu
lated
u
s
i
n
g
t
h
e
m
id
p
o
in
t
o
f
t
h
e
f
ac
e
i
m
a
g
e
(
c)
u
s
i
n
g
E
q
u
a
tio
n
(
1
8
)
.
c
=
½
w
(
1
8
)
T
h
en
,
1
x
,
1
y
an
d
2
x
w
er
e
ca
lcu
lated
u
s
in
g
E
q
u
atio
n
(
1
9
)
,
(
2
0
)
an
d
(
2
1
)
.
1
x
=
0
x
+c
(
1
9
)
B
a
c
k
g
r
o
u
n
d
F
a
c
e
i
mag
e
H
)
,
(
1
1
y
x
w
h
c
G
)
,
(
0
1
y
x
A
)
,
(
0
0
y
x
B
)
,
(
0
2
y
x
C
)
,
(
1
0
y
x
)
D
)
,
(
1
2
y
x
Evaluation Warning : The document was created with Spire.PDF for Python.
I
J
E
C
E
I
SS
N:
2
0
8
8
-
8708
F
a
ce
R
ec
o
g
n
itio
n
B
a
s
ed
o
n
S
y
mme
tr
ica
l H
a
lf
-
Jo
in
Meth
o
d
u
s
in
g
S
tereo
V
is
io
n
C
a
mera
(
E
d
y
W
in
a
r
n
o
)
2824
1
y
=
0
y
+(
h
-
1)
(
2
0
)
2
x
=
0
x
+(
w
-
1)
(
2
1
)
G,
B
,
H
an
d
D
as
R
o
I
o
f
r
ig
h
t
h
al
f
-
f
ac
e
w
er
e
ca
lcu
lated
u
s
i
n
g
E
q
u
atio
n
(
2
2
)
-
(
2
5
)
.
G=
(
0
x
+c,
0
y
)
(
2
2
)
B
=(
0
x
+(
w
-
1
)
)
,
0
y
)
(
2
3
)
H=
(
0
x
+c,
0
y
+(
h
-
1
)
)
(
2
4
)
D=
(
0
x
+(
w
-
1
)
,
0
y
+(
h
-
1
)
)
(
2
5
)
G,
B
,
H
an
d
D
w
er
e
th
e
n
co
m
b
in
ed
w
it
h
R
o
I
o
f
lef
t
h
a
lf
-
f
ac
e
o
n
th
e
lef
t
len
s
(
A
,
E
,
C
an
d
F)
as
s
ymm
etri
ca
l
h
a
lf
-
jo
in
.
3.
RE
SU
L
T
S
A
ND
AN
AL
Y
SI
S
T
h
is
s
t
u
d
y
u
s
ed
t
h
e
Vio
la
-
J
o
n
es
f
ac
e
d
etec
tio
n
.
E
ac
h
le
f
t
a
n
d
r
ig
h
t
le
n
s
e
s
o
f
a
s
ter
eo
v
is
io
n
ca
m
er
a
w
a
s
u
s
ed
to
tak
e
a
r
a
w
i
m
ag
e
w
it
h
t
h
e
b
ac
k
g
r
o
u
n
d
i
m
a
g
e
s
i
ze
o
f
6
4
0
x
4
8
0
p
ix
els
an
d
th
e
d
etec
ted
f
ac
e
i
m
ag
e.
W
id
th
an
d
h
ei
g
h
t
o
f
t
h
e
f
ac
e
i
m
a
g
e
d
ep
en
d
ed
o
n
th
e
d
is
ta
n
ce
b
et
w
ee
n
t
h
e
f
ac
e
a
n
d
t
h
e
ca
m
er
a.
T
h
e
f
ac
e
i
m
a
g
e
d
etec
ted
o
n
t
h
e
b
ac
k
g
r
o
u
n
d
i
m
a
g
e
w
a
s
m
ar
k
ed
w
it
h
th
e
co
o
r
d
in
ates
o
f
t
h
e
p
o
s
itio
n
o
f
t
h
e
f
ac
e
i
m
a
g
e
R
o
I
.
T
h
e
f
ac
e
i
m
a
g
e
R
o
I
w
as
th
en
s
to
r
ed
i
n
f
ac
e
i
m
a
g
e
s
to
r
ag
e
v
ar
iab
le.
T
h
e
f
ac
e
i
m
ag
e
was t
h
en
n
o
r
m
alize
d
u
s
i
n
g
p
r
ep
r
o
ce
s
s
in
g
a
n
d
h
al
f
-
j
o
in
m
et
h
o
d
.
I
n
p
r
ep
r
o
ce
s
s
in
g
,
t
h
e
cu
tti
n
g
p
r
o
ce
s
s
w
a
s
co
n
d
u
c
ted
to
tak
e
t
h
e
f
ac
e
i
m
a
g
e
a
n
d
it
w
as
co
n
v
er
ted
in
to
g
r
a
y
s
ca
le.
T
h
e
g
r
ay
s
ca
le
i
m
a
g
e
w
a
s
th
e
n
ad
j
u
s
ted
u
s
in
g
t
h
e
r
esizin
g
p
r
o
ce
s
s
to
g
et
a
f
ac
e
i
m
a
g
e
s
ize
o
f
1
6
0
x
1
6
0
p
ix
els.
T
h
e
ad
j
u
s
t
m
e
n
t o
f
co
n
tr
a
s
t a
n
d
b
r
ig
h
t
n
ess
w
a
s
co
n
d
u
cted
u
s
i
n
g
t
h
e
h
is
to
g
r
a
m
eq
u
a
lizatio
n
.
I
t
w
a
s
u
s
ed
to
p
r
o
d
u
ce
th
e
b
etter
f
ac
e
i
m
ag
e
q
u
a
lit
y
.
I
m
a
g
e
r
esu
lt
s
in
ea
c
h
o
f
t
h
e
ca
m
er
a
len
s
w
er
e
t
h
en
p
r
o
ce
s
s
ed
u
s
in
g
t
h
e
m
et
h
o
d
o
f
s
y
m
m
etr
ical
h
al
f
-
j
o
in
.
T
h
e
s
a
m
p
les
o
f
r
es
u
lt
s
o
f
s
o
m
e
R
o
I
o
f
f
ac
e
i
m
ag
e
p
r
ep
r
o
ce
s
s
in
g
o
n
a
s
ter
eo
v
is
io
n
ca
m
er
a
in
s
o
m
e
p
o
s
es
tak
e
n
at
th
e
s
a
m
e
ti
m
e
a
r
e
p
r
esen
ted
in
Fi
g
u
r
e
5
.
Fig
u
r
e
5
.
Face
I
m
a
g
e
P
r
ep
r
o
c
ess
i
n
g
R
es
u
lt
s
o
n
a
Ster
eo
Vis
io
n
C
a
m
er
a
,
(
ab
o
v
e)
th
e
Im
ag
e
o
n
th
e
L
ef
t
L
e
n
s
,
(
b
o
tto
m
)
th
e
I
m
ag
e
o
n
th
e
R
ig
h
t
L
en
s
S
y
m
m
etr
ical
h
a
lf
-
j
o
in
m
e
th
o
d
cu
ts
t
h
e
i
m
ag
e
i
n
t
h
e
ce
n
ter
o
f
th
e
f
ac
e
an
d
p
r
o
d
u
ce
th
e
s
a
m
e
i
m
ag
e
w
id
t
h
.
T
h
e
s
ize
o
f
le
f
t
h
alf
-
f
a
ce
i
m
a
g
es
a
n
d
r
ig
h
t
h
a
f
-
f
ac
e
i
m
ag
e
s
i
s
8
0
x
1
6
0
p
ix
els.
T
h
e
l
ef
t
h
al
f
-
f
ac
e
i
m
a
g
e
an
d
th
e
r
i
g
h
t
h
al
f
-
f
ac
e
i
m
a
g
e
a
r
e
th
en
m
er
g
ed
in
to
o
n
e
o
f
s
y
m
m
etr
ical
h
al
f
-
j
o
in
i
m
a
g
e.
T
h
e
p
r
o
ce
s
s
o
f
m
er
g
er
b
et
w
ee
n
b
o
th
o
f
t
h
e
i
m
a
g
es i
s
illu
s
tr
ated
in
Fi
g
u
r
e
6
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
J
E
C
E
Vo
l.
6
,
No
.
6
,
Dec
em
b
er
2
0
1
6
: 2
8
1
8
–
2
8
2
7
2825
8
0
x
1
6
0
8
0
x
1
6
0
8
0
x
1
6
0
8
0
x
1
6
0
8
0
x
1
6
0
8
0
x
1
6
0
8
0
x
1
6
0
8
0
x
1
6
0
8
0
x
1
6
0
8
0
x
1
6
0
1
6
0
x
1
6
0
1
6
0
x
1
6
0
1
6
0
x
1
6
0
1
6
0
x
1
6
0
1
6
0
x
1
6
0
Fig
u
r
e
6
.
(
A
b
o
v
e)
T
h
e
L
ef
t
H
alf
-
F
ac
e
o
f
Sy
m
m
etr
ical
Hal
f
-
J
o
in
,
(
m
id
d
le)
th
e
R
ig
h
t H
al
f
-
Fa
ce
o
f
Sy
m
m
etr
ical
Hal
f
-
J
o
in
,
(
b
o
tto
m
)
th
e
Me
r
g
er
I
m
ag
e
o
f
S
y
m
m
etr
ical
Hal
f
-
J
o
in
3
.
1
.
F
a
ce
Rec
o
g
nitio
n Ra
t
e
o
n
Sy
mm
et
rica
l H
a
lf
-
J
o
in
Face
r
ec
o
g
n
i
tio
n
r
ate
te
s
ti
n
g
w
a
s
co
n
d
u
cted
to
co
m
p
ar
e
th
e
ac
c
u
r
ac
y
o
f
t
h
e
p
r
o
p
o
s
ed
m
o
d
el
i
n
d
if
f
er
e
n
t
ex
tr
ac
tio
n
a
n
d
class
if
icatio
n
m
e
th
o
d
s
.
W
e
u
s
ed
4
0
0
f
ac
e
i
m
ag
e
s
d
atab
ase
o
f
s
in
g
le
v
is
io
n
an
d
s
y
m
m
etr
ical
h
a
lf
-
j
o
in
.
T
h
e
tes
tin
g
is
co
n
d
u
cted
b
y
co
m
p
ar
in
g
th
e
te
s
t i
m
ag
e
s
t
h
at
h
a
s
b
ee
n
s
to
r
ed
.
T
est i
m
ag
e
d
ata
is
ta
k
en
f
r
o
m
4
0
p
er
s
o
n
s
r
an
d
o
m
l
y
as
th
e
test
d
ata.
T
e
s
t
is
d
o
n
e
4
ti
m
es
u
s
i
n
g
r
a
n
d
o
m
te
s
t
i
m
a
g
es
1
0
0
,
2
0
0
,
3
0
0
an
d
4
0
0
o
f
d
ata.
T
est
i
m
a
g
e
w
as
tak
e
n
r
an
d
o
m
l
y
f
r
o
m
t
h
e
t
w
o
m
et
h
o
d
s
o
f
s
in
g
le
v
is
io
n
an
d
s
y
m
m
etr
ical
h
a
lf
-
j
o
in
.
T
h
e
co
m
p
ar
is
o
n
w
as
co
n
d
u
cted
b
y
co
m
p
ar
i
n
g
th
e
f
ac
e
r
ec
o
g
n
i
tio
n
r
ate
u
s
i
n
g
v
ar
iatio
n
s
o
n
f
ea
t
u
r
e
e
x
tr
ac
tio
n
a
n
d
class
if
icatio
n
.
Su
m
m
ar
y
r
es
u
lts
o
f
f
ac
e
r
ec
o
g
n
it
io
n
r
ate
u
s
in
g
v
ar
iatio
n
s
o
f
ex
tr
ac
tio
n
a
n
d
class
if
icatio
n
ar
e
s
ee
n
i
n
T
ab
le
1
.
T
ab
le
1
.
Su
m
m
ar
y
Re
s
u
lts
o
f
Face
R
ec
o
g
n
itio
n
R
ate
U
s
in
g
Var
iatio
n
s
o
f
E
x
t
r
ac
tio
n
a
n
d
Cl
ass
i
f
icatio
n
I
mag
e
s
R
e
c
o
g
n
i
t
i
o
n
r
a
t
e
(
%)
Eu
c
l
i
d
e
a
n
M
a
h
a
l
a
n
o
b
i
s
P
C
A
3
W
a
v
e
l
e
t
-
P
C
A
P
C
A
3
W
a
v
e
l
e
t
-
P
C
A
1
0
0
9
3
.
0
0
9
2
.
0
0
9
7
.
0
0
9
4
.
0
0
2
0
0
9
3
.
0
0
9
2
.
5
0
9
7
.
5
0
9
4
.
5
0
3
0
0
9
4
.
3
3
9
3
.
0
0
9
7
.
6
7
9
5
.
3
3
4
0
0
9
4
.
7
5
9
2
.
5
0
9
8
.
2
5
9
5
.
7
5
T
h
e
h
ig
h
es
t
ac
cu
r
ac
y
o
f
f
ac
e
r
ec
o
g
n
itio
n
i
s
p
r
o
v
id
ed
u
s
in
g
P
C
A
-
Ma
h
ala
n
o
b
is
.
T
h
e
ac
cu
r
ac
y
o
f
f
ac
e
r
ec
o
g
n
itio
n
u
s
in
g
P
C
A
-
3
W
av
elet
-
Ma
h
ala
n
o
b
is
,
P
C
A
-
E
u
cli
d
ea
n
an
d
3
W
av
elet
-
P
C
A
-
E
u
cl
id
ea
n
ar
e
o
v
er
th
an
9
3
%.
3
.
2
.
Rec
o
g
nitio
n
Ra
t
e
B
a
s
ed
o
n
P
o
s
es Va
ria
t
io
n
T
h
e
m
ea
s
u
r
e
m
en
t
o
f
r
ec
o
g
n
it
i
o
n
r
ate
b
ased
o
n
p
o
s
es
v
ar
iatio
n
is
co
n
d
u
cted
b
y
p
o
s
es
o
f
1
0
p
er
s
o
n
s
w
it
h
7
m
a
in
p
o
s
es
v
ar
iatio
n
a
n
d
2
1
v
ar
iatio
n
s
in
d
if
f
er
e
n
t
an
g
le
s
o
f
t
h
e
f
ac
e
i
m
ag
e.
T
h
e
test
i
n
g
i
s
d
o
n
e
b
y
Evaluation Warning : The document was created with Spire.PDF for Python.
I
J
E
C
E
I
SS
N:
2
0
8
8
-
8708
F
a
ce
R
ec
o
g
n
itio
n
B
a
s
ed
o
n
S
y
mme
tr
ica
l H
a
lf
-
Jo
in
Meth
o
d
u
s
in
g
S
tereo
V
is
io
n
C
a
mera
(
E
d
y
W
in
a
r
n
o
)
2826
co
m
p
ar
i
n
g
t
h
e
le
v
el
o
f
ac
c
u
r
ac
y
o
f
f
ac
e
r
ec
o
g
n
itio
n
.
T
h
e
r
esu
l
ts
o
f
r
ec
o
g
n
itio
n
r
ate
b
ased
o
n
f
ac
e
p
o
s
e
v
ar
iatio
n
ar
e
p
r
esen
ted
i
n
T
ab
l
e
2
.
T
h
is
r
esear
ch
p
r
o
v
id
es
a
lev
el
o
f
f
ac
e
r
ec
o
g
n
it
io
n
r
ate
o
n
s
o
m
e
v
ar
iatio
n
s
o
f
p
o
s
es
w
it
h
d
if
f
er
en
t
r
esu
lt
s
.
Gen
er
all
y
,
t
h
e
test
in
g
o
f
f
ac
e
r
ec
o
g
n
itio
n
r
ate
w
it
h
p
o
s
es
v
ar
iat
io
n
w
ill
p
r
o
d
u
ce
b
etter
f
ac
e
r
ec
o
g
n
itio
n
ac
cu
r
ac
y
i
n
f
r
o
n
ta
l
v
ie
w
p
o
s
itio
n
,
t
u
r
n
,
b
o
w
a
n
d
lo
o
k
ed
u
p
at
a
m
a
x
i
m
u
m
a
n
g
le
o
f
4
5
°.
T
h
e
b
est
ac
cu
r
ac
y
o
f
f
ac
e
r
ec
o
g
n
itio
n
i
s
in
th
e
n
o
r
m
al
p
o
s
itio
n
in
t
h
e
a
n
g
le
o
f
0
°
-
2
2
.
5
°.
T
ab
le
2
.
R
esu
lts
o
f
R
ec
o
g
n
itio
n
R
ate
B
ased
o
n
Face
P
o
s
e
Var
iatio
n
No
P
o
se
A
n
g
l
e
R
e
c
o
g
n
i
t
i
o
n
R
a
t
e
(
%)
No
P
o
se
A
n
g
l
e
R
e
c
o
g
n
i
t
i
o
n
R
a
t
e
(
%)
1
N
o
r
mal
0°
1
0
0
12
C
l
o
c
k
w
i
se
6
7
,
5
°
0
2
F
a
c
i
n
g
l
e
f
t
2
2
,
5
°
97
13
A
n
t
i
c
l
o
c
k
w
i
se
2
2
,
5
°
86
3
F
a
c
i
n
g
l
e
f
t
45°
93
14
A
n
t
i
c
l
o
c
k
w
i
se
45°
81
4
F
a
c
i
n
g
l
e
f
t
6
7
,
5
°
34
15
A
n
t
i
c
l
o
c
k
w
i
se
6
7
,
5
°
0
5
F
a
c
i
n
g
l
e
f
t
90°
0
16
B
o
w
2
2
,
5
°
97
6
F
a
c
i
n
g
r
i
g
h
t
2
2
,
5
°
97
17
B
o
w
45°
90
7
F
a
c
i
n
g
r
i
g
h
t
45°
95
18
B
o
w
6
7
,
5
°
0
8
F
a
c
i
n
g
r
i
g
h
t
6
7
,
5
°
41
19
L
o
o
k
u
p
2
2
,
5
°
96
9
F
a
c
i
n
g
r
i
g
h
t
90°
0
20
L
o
o
k
u
p
45°
86
10
C
l
o
c
k
w
i
se
2
2
,
5
°
91
21
L
o
o
k
u
p
6
7
,
5
°
34
11
C
l
o
c
k
w
i
se
45°
86
3
.
3
.
Rec
o
g
nitio
n
Ra
t
e
B
a
s
ed
o
n
I
llu
m
i
na
nce
Va
ria
t
io
ns
T
esti
n
g
o
f
f
ac
e
r
ec
o
g
n
i
tio
n
r
ate
b
ased
o
n
v
ar
iat
io
n
s
i
n
il
lu
m
i
n
atio
n
w
as
p
er
f
o
r
m
ed
w
it
h
a
lu
x
m
ete
r
to
m
ea
s
u
r
e
li
g
h
t
le
v
el
s
i
n
l
u
x
(
lu
m
e
n
s
/
m
²)
.
T
h
i
s
to
o
l
/d
ev
ic
e
w
a
s
u
s
ed
to
d
eter
m
i
n
e
t
h
e
l
ev
el
o
f
v
ar
ia
tio
n
o
f
lig
h
ti
n
g
in
t
h
e
r
o
o
m
w
h
e
n
t
h
e
test
in
g
p
r
o
ce
s
s
w
a
s
co
n
d
u
cted
.
T
esti
n
g
o
f
f
ac
e
r
ec
o
g
n
itio
n
r
ate
b
ased
o
n
v
ar
iatio
n
s
i
n
t
h
e
le
v
el
o
f
ill
u
m
in
an
ce
w
as
d
o
n
e
in
3
x
4
m
eter
en
clo
s
ed
u
s
i
n
g
v
ar
iat
io
n
s
o
f
l
ig
h
t
as
th
e
lig
h
ti
n
g
s
o
u
r
ce
.
T
esti
n
g
i
s
co
n
d
u
cted
o
n
1
0
p
eo
p
le
w
it
h
1
0
ti
m
e
s
t
ak
en
b
ased
o
n
th
e
v
ar
iatio
n
o
f
ill
u
m
i
n
an
ce
.
T
h
e
r
esu
lt o
f
f
ac
e
r
ec
o
g
n
iti
o
n
r
ate
b
ased
o
n
illu
m
i
n
an
ce
v
ar
iatio
n
s
ca
n
b
e
s
ee
n
i
n
T
ab
le
3
.
T
ab
le
3
.
T
h
e
R
esu
lt o
f
Face
R
ec
o
g
n
itio
n
R
ate
B
ased
o
n
I
ll
u
m
i
n
an
ce
Var
iat
io
n
s
No
I
l
u
mi
n
a
n
c
e
(
l
u
x
)
L
a
mp
s (w
a
t
t
)
L
i
g
h
t
C
a
t
e
g
o
r
y
(
B
a
se
d
o
n
L
u
x
me
t
e
r
)
R
e
c
o
g
n
i
t
i
o
n
R
a
t
e
(
%)
1
8
Off
D
a
r
k
(
0
–
1
0
)
54
2
27
5
G
l
i
mm
e
r
(
1
1
–
80)
90
3
36
7
G
l
i
mm
e
r
(
1
1
–
80)
91
4
85
11
B
r
i
g
h
t
(
8
1
–
4
0
0
)
95
5
1
1
5
14
B
r
i
g
h
t
(
8
1
–
4
0
0
)
97
6
3
5
7
40
B
r
i
g
h
t
(
8
1
–
4
0
0
)
1
0
0
7
4
8
2
60
V
e
r
y
b
r
i
g
h
t
(
4
0
1
–
2
5
0
0
)
1
0
0
8
6
6
9
1
0
0
V
e
r
y
b
r
i
g
h
t
(
4
0
1
–
2
5
0
0
)
1
0
0
9
2
7
5
2
O
u
t
d
o
o
r
S
u
n
n
y
(
2
5
0
0
<
…)
1
0
0
10
5
3
6
1
O
u
t
d
o
o
r
S
u
n
n
y
(
2
5
0
0
<
…)
1
0
0
Gen
er
all
y
,
t
h
e
p
r
o
p
o
s
ed
m
o
d
el
w
ill
g
e
n
er
ate
th
e
ac
cu
r
ac
y
o
f
f
ac
e
r
ec
o
g
n
itio
n
w
h
ic
h
al
w
a
y
s
s
tab
l
e
an
d
r
e
m
ain
s
h
i
g
h
d
esp
ite
th
e
illu
m
in
a
n
ce
le
v
el
is
v
er
y
lo
w
.
Ho
w
ev
er
,
to
p
r
o
d
u
ce
a
h
ig
h
ac
c
u
r
ac
y
o
f
f
ac
e
r
ec
o
g
n
itio
n
ca
n
b
e
p
er
f
o
r
m
ed
u
s
i
n
g
ill
u
m
in
a
n
ce
le
v
els ab
o
v
e
1
1
5
lu
x
s
o
th
at
t
h
e
ac
cu
r
ac
y
ca
n
b
e
m
ax
i
m
ized
.
4.
CO
NCLU
SI
O
N
T
h
e
m
ai
n
p
r
o
b
le
m
in
a
f
ac
e
r
ec
o
g
n
itio
n
s
y
s
te
m
b
ased
o
n
th
e
p
atter
n
o
f
f
ac
e
s
y
m
m
etr
y
is
h
o
w
to
d
ev
elo
p
a
m
et
h
o
d
t
h
at
ca
n
a
n
ticip
ate
v
ar
i
atio
n
s
i
n
f
ac
ial
p
o
s
e
an
d
ill
u
m
in
a
tio
n
v
ar
iati
o
n
s
to
i
m
p
r
o
v
e
t
h
e
ac
cu
r
ac
y
o
f
f
ac
e
r
ec
o
g
n
i
tio
n
.
I
n
t
h
i
s
s
tu
d
y
,
w
e
p
r
o
p
o
s
ed
a
f
ac
e
r
ec
o
g
n
i
tio
n
u
s
es
2
len
s
o
n
a
s
ter
eo
v
is
io
n
ca
m
er
a.
T
h
e
p
r
o
p
o
s
ed
m
et
h
o
d
is
an
i
m
a
g
e
m
er
g
i
n
g
f
r
o
m
l
ef
t
an
d
r
ig
h
t
f
ac
e
i
m
ag
e
ac
q
u
i
s
itio
n
u
s
i
n
g
s
ter
eo
v
is
io
n
ca
m
er
a
n
a
m
el
y
h
al
f
-
j
o
in
m
et
h
o
d
.
Half
-
j
o
in
is
a
m
e
th
o
d
o
f
n
o
r
m
aliza
tio
n
o
f
t
h
e
f
a
ce
d
etec
tio
n
u
s
i
n
g
s
ter
eo
v
is
io
n
ca
m
er
a
to
cu
t
o
n
ea
ch
o
f
t
h
e
i
m
a
g
es
f
r
o
m
e
v
er
y
le
n
s
a
n
d
to
m
er
g
e
it.
T
h
e
m
er
g
er
i
m
a
g
e
o
f
th
e
p
r
o
p
o
s
ed
m
et
h
o
d
(
s
y
m
m
et
r
ica
l
h
al
f
-
j
o
in
)
is
th
e
n
u
s
ed
a
s
i
n
p
u
t
to
th
e
p
r
o
ce
s
s
o
f
ex
tr
ac
tio
n
an
d
cla
s
s
i
f
icatio
n
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
J
E
C
E
Vo
l.
6
,
No
.
6
,
Dec
em
b
er
2
0
1
6
: 2
8
1
8
–
2
8
2
7
2827
in
th
e
f
ac
e
r
ec
o
g
n
itio
n
s
y
s
te
m
.
I
n
co
m
p
ar
is
o
n
o
f
ac
c
u
r
ac
y
o
f
f
ac
e
r
ec
o
g
n
itio
n
o
n
ea
c
h
m
o
d
el
o
f
n
o
r
m
aliza
tio
n
,
s
y
m
m
etr
i
c
h
al
f
-
j
o
in
m
et
h
o
d
p
r
o
d
u
ce
s
f
ac
e
r
ec
o
g
n
i
tio
n
r
ate
u
p
to
9
8
.
2
5
%.
I
n
co
m
p
ar
i
s
o
n
o
f
ac
c
u
r
ac
y
o
f
f
ac
e
r
ec
o
g
n
itio
n
b
ased
o
n
v
ar
iatio
n
s
in
p
o
s
e
an
d
illu
m
i
n
atio
n
v
ar
iatio
n
s
,
th
e
s
y
m
m
et
r
ical
h
alf
-
j
o
in
m
et
h
o
d
is
ab
le
to
p
r
o
d
u
ce
a
h
ig
h
ac
cu
r
ac
y
o
f
f
ac
e
r
ec
o
g
n
itio
n
a
n
d
ab
le
to
an
ticip
ate
v
ar
iatio
n
s
in
p
o
s
e
an
d
illu
m
i
n
atio
n
.
T
h
e
p
r
o
p
o
s
ed
m
o
d
el
is
ab
le
to
p
r
o
d
u
ce
8
6
%
-
9
7
%
ac
cu
r
ac
y
o
n
a
v
ar
iet
y
o
f
p
o
s
es
an
d
v
ar
iatio
n
s
i
n
a
n
g
les
b
et
wee
n
0
°
-
2
2
.
5
°.
O
n
th
e
v
ar
iat
io
n
o
f
ill
u
m
i
n
an
ce
m
ea
s
u
r
ed
u
s
i
n
g
a
lu
x
m
eter
is
ab
le
to
r
esu
lt
i
n
f
ac
e
r
ec
o
g
n
i
tio
n
ac
c
u
r
ac
y
o
f
9
0
%
-
1
0
0
% f
o
r
th
e
ca
teg
o
r
y
o
f
li
g
h
t
in
g
le
v
els ab
o
v
e
1
0
lu
x
.
RE
F
E
R
E
NC
E
S
[1
]
A
.
N.
M
a
rti
n
e
z
-
G
o
n
z
a
lez
,
&
V
.
Ay
a
la
-
Ra
m
irez
,
“
Re
a
l
T
i
m
e
F
a
c
e
De
tec
ti
o
n
Us
in
g
Ne
u
ra
l
Ne
tw
o
rk
s”
,
In
Arti
fi
c
ia
l
In
telli
g
e
n
c
e
(
M
ICAI)
,
2
0
1
1
1
0
th
M
e
x
ica
n
In
ter
n
a
t
io
n
a
l
C
o
n
fer
e
n
c
e
o
n
,
I
EE
E,
1
4
4
-
1
4
9
,
2
0
1
1
.
[2
]
M
.
S
.
De
v
i,
&
P
.
R.
Ba
jaj,
“
A
c
ti
v
e
F
a
c
ial
T
ra
c
k
in
g
”
,
In
Eme
rg
in
g
T
re
n
d
s
in
E
n
g
i
n
e
e
rin
g
a
n
d
T
e
c
h
n
o
lo
g
y
(
ICET
ET
)
,
2
0
1
0
3
rd
I
n
ter
n
a
ti
o
n
a
l
Co
n
fer
e
n
c
e
o
n
,
I
EE
E,
91
-
9
5
,
2
0
1
0
.
[3
]
D.
S
rid
h
a
r
&
I.
M
.
Krish
n
a
,
“
F
a
c
e
Re
c
o
g
n
it
io
n
Us
in
g
T
w
o
Dim
e
n
sio
n
a
l
Disc
re
te
Co
sin
e
T
ra
n
sf
o
rm
,
L
in
e
a
r
Disc
ri
m
in
a
n
t
A
n
a
l
y
sis
A
n
d
K
Ne
a
re
st
Ne
i
g
h
b
o
r
Clas
sif
ier
”
,
IAE
S
In
te
rn
a
ti
o
n
a
l
J
o
u
rn
a
l
o
f
Arti
fi
c
ia
l
I
n
telli
g
e
n
c
e
,
1
(4
),
1
6
1
,
2
0
1
2
.
[4
]
I.
G
.
P
.
S
.
W
ij
a
y
a
,
e
t
a
l
.,
“
F
a
c
e
Re
c
o
g
n
it
io
n
Us
in
g
Ho
l
isti
c
F
e
a
tu
re
s
a
n
d
S
im
p
li
f
ied
L
in
e
a
r
Disc
ri
m
i
n
a
n
t
A
n
a
ly
sis
”
,
In
d
o
n
e
sia
n
J
o
u
rn
a
l
o
f
El
e
c
trica
l
En
g
i
n
e
e
rin
g
a
n
d
C
o
mp
u
ter
S
c
ien
c
e
,
10
(4
)
,
7
7
5
-
7
8
7
,
2
0
1
2
.
[5
]
M
.
E.
W
ib
o
w
o
,
“
T
o
wa
rd
s
p
o
se
-
ro
b
u
st
f
a
c
e
re
c
o
g
n
it
io
n
o
n
v
i
d
e
o
”
,
Ph
D
th
e
sis
,
Qu
e
e
n
sla
n
d
Un
iv
e
rsit
y
o
f
T
e
c
h
n
o
lo
g
y
,
2
0
1
4
.
[6
]
W
.
Ch
e
n
,
e
t
a
l
.
,
“
F
a
c
e
d
e
tec
ti
o
n
b
a
se
d
o
n
h
a
lf
f
a
c
e
-
te
m
p
lat
e
”
,
In
El
e
c
tro
n
ic
M
e
a
su
r
e
me
n
t
&
In
stru
me
n
ts,
ICEM
I'
0
9
,
9
th
In
ter
n
a
ti
o
n
a
l
Co
n
f
e
re
n
c
e
o
n
,
IEE
E
,
4
-
5
4
,
2
0
0
9
.
[7
]
P
.
L
e
k
sh
m
i,
e
t
a
l
.
,
“
A
n
a
l
y
sis
o
f
fa
c
ial
e
x
p
re
ss
io
n
s
u
sin
g
P
CA
o
n
h
a
lf
a
n
d
f
u
ll
f
a
c
e
s”
,
In
Au
d
io
,
L
a
n
g
u
a
g
e
a
n
d
Ima
g
e
Pro
c
e
ss
in
g
,
2
0
0
8
,
IC
AL
IP
2
0
0
8
,
In
ter
n
a
ti
o
n
a
l
Co
n
fer
e
n
c
e
o
n
,
IEE
E,
1
3
7
9
-
1
3
8
3
,
2
0
0
8
.
[8
]
Y.
Xu
,
e
t
a
l
.
,
“
Us
in
g
th
e
o
rig
in
a
l
a
n
d
„sy
m
m
e
tri
c
a
l
fa
c
e
‟
train
in
g
sa
m
p
les
to
p
e
rf
o
rm
re
p
re
se
n
tatio
n
b
a
se
d
tw
o
-
ste
p
f
a
c
e
re
c
o
g
n
it
io
n
,
P
a
tt
e
rn
Rec
o
g
n
i
ti
o
n
,
46
,
4
,
1
1
5
1
-
1
1
5
8
,
2
0
1
3
.
[9
]
J.
Ha
rg
u
e
ss
&
J.
K.
A
g
g
a
r
wa
l,
“
A
c
a
se
f
o
r
th
e
a
v
e
ra
g
e
-
h
a
l
f
-
fa
c
e
in
2
D
a
n
d
3
D
f
o
r
f
a
c
e
re
c
o
g
n
it
io
n
”
,
In
Co
m
p
u
ter
Vi
sio
n
a
n
d
Pa
tt
e
rn
Rec
o
g
n
it
io
n
W
o
rk
sh
o
p
s,
2
0
0
9
,
CVP
R
W
o
rk
sh
o
p
s
2
0
0
9
,
IEE
E
Co
m
p
u
ter
S
o
c
iet
y
Co
n
fer
e
n
c
e
o
n
,
IEE
E,
7
-
1
2
,
2
0
0
9
.
[1
0
]
J.
Ha
rg
u
e
ss
&
J.K.
Ag
g
a
r
wa
l,
“
Is
th
e
re
a
c
o
n
n
e
c
ti
o
n
b
e
tw
e
e
n
f
a
c
e
s
y
m
m
e
tr
y
a
n
d
f
a
c
e
re
c
o
g
n
it
io
n
?
”
,
In
Co
mp
u
te
r
Vi
sio
n
a
n
d
P
a
tt
e
rn
Rec
o
g
n
it
i
o
n
W
o
rk
sh
o
p
s
(
CVP
RW
),
2
0
1
1
IEE
E
Co
mp
u
ter
S
o
c
iety
Co
n
fer
e
n
c
e
o
n
,
IEE
E
,
6
6
-
73
2
0
1
1
.
[1
1
]
M
.
Qiu
,
e
t
a
l
.
,
“
I
n
teg
ra
ti
n
g
th
e
o
rig
in
a
l
f
a
c
e
i
m
a
g
e
s
a
n
d
“
s
y
m
m
e
t
rica
l
f
a
c
e
s
”
to
p
e
rf
o
rm
fa
c
e
re
c
o
g
n
it
io
n
”
,
O
p
ti
k
-
In
ter
n
a
t
io
n
a
l
J
o
u
rn
a
l
f
o
r L
i
g
h
t
a
n
d
E
lec
tro
n
O
p
ti
c
s
,
1
2
5
,
1
1
,
2
6
6
5
-
2
6
7
0
,
2
0
1
4
.
[1
2
]
L
.
L
a
n
,
e
t
a
l
.
,
“
Ill
u
m
in
a
ti
o
n
Co
m
p
e
n
sa
ti
o
n
f
o
r
F
a
c
e
Re
c
o
g
n
it
io
n
U
sin
g
On
ly
On
e
I
m
a
g
e
”
,
Acta
Au
to
ma
ti
c
a
S
in
ic
a
,
39
,
1
2
,
2
0
9
0
-
2
0
9
9
,
2
0
1
3
.
[1
3
]
P
.
V
io
la
&
M
.
J.
Jo
n
e
s,
“
Ro
b
u
st
r
e
a
l
-
ti
m
e
fa
c
e
d
e
tec
ti
o
n
”
,
In
ter
n
a
ti
o
n
a
l
j
o
u
rn
a
l
o
f
c
o
mp
u
ter
v
isio
n
,
57
(
2
),
1
3
7
-
1
5
4
,
2
0
0
4
.
[1
4
]
E.
W
in
a
rn
o
,
e
t
a
l
.
,
“
Im
p
ro
v
e
d
Re
a
l
-
T
i
m
e
F
a
c
e
R
e
c
o
g
n
it
io
n
Ba
se
d
On
T
h
re
e
Lev
e
l
W
a
v
e
let
De
c
o
m
p
o
siti
o
n
-
P
ri
n
c
ip
a
l
Co
m
p
o
n
e
n
t
A
n
a
ly
sis
A
n
d
M
a
h
a
lan
o
b
is
Dista
n
c
e
”
,
J
o
u
rn
a
l
o
f
C
o
mp
u
ter
S
c
ien
c
e
,
1
0
,
5
,
8
4
4
-
8
5
1
,
d
o
i:
1
0
.
3
8
4
4
/j
c
ss
p
.
2
0
1
4
.
8
4
4
.
8
5
1
,
2
0
1
4
.
[1
5
]
E.
W
in
a
rn
o
,
e
t
a
l
.
,
“
De
v
e
lo
p
m
e
n
t
Of
F
a
c
e
Re
c
o
g
n
it
io
n
S
y
ste
m
A
n
d
F
a
c
e
Dista
n
c
e
Esti
m
a
ti
o
n
Us
in
g
S
tere
o
V
isi
o
n
Ca
m
e
ra
”
,
J
o
u
rn
a
l
Of
T
h
e
o
re
ti
c
a
l
An
d
Ap
p
li
e
d
I
n
f
o
rm
a
ti
o
n
T
e
c
h
n
o
l
o
g
y
,
6
7
,
3
,
6
5
2
-
6
5
7
,
2
0
1
4
.
Evaluation Warning : The document was created with Spire.PDF for Python.