I
nte
rna
t
io
na
l J
o
urna
l o
f
E
lect
rica
l a
nd
Co
m
pu
t
er
E
ng
ineering
(
I
J
E
CE
)
Vo
l.
15
,
No
.
5
,
Octo
b
er
20
25
,
p
p
.
4
9
0
7
~
4
9
1
5
I
SS
N:
2088
-
8
7
0
8
,
DOI
: 1
0
.
1
1
5
9
1
/ijece.
v
15
i
5
.
pp
4
9
0
7
-
4
9
1
5
4907
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
ec
e.
ia
esco
r
e.
co
m
Enha
ncing
f
a
cia
l
la
ndma
rk det
e
cti
o
n wit
h
Co
ntrolN
et
-
ba
sed
da
ta a
ug
menta
tion
K
rit
a
ph
a
t
So
ng
s
ri
-
in
1
,
M
un
l
ik
a
Ra
t
t
a
ph
un
1
,
So
pee
K
a
ewc
ha
da
2
,
Su
nis
a
K
idja
idea
w
2
,
Sa
ng
j
un
Rua
ng
-
On
2
,
Wichit
So
o
kk
ha
t
ho
n
2
,
P
a
t
o
m
po
ng
Cha
bp
la
n
2
1
D
e
p
a
r
t
me
n
t
o
f
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
,
F
a
c
u
l
t
y
o
f
S
c
i
e
n
c
e
a
n
d
T
e
c
h
n
o
l
o
g
y
,
N
a
k
h
o
n
S
i
T
h
a
mm
a
r
a
t
R
a
j
a
b
h
a
t
U
n
i
v
e
r
si
t
y
,
Th
a
N
g
i
o
,
Th
a
i
l
a
n
d
2
D
e
p
a
r
t
me
n
t
o
f
I
n
f
o
r
mat
i
o
n
Te
c
h
n
o
l
o
g
y
a
n
d
D
i
g
i
t
a
l
I
n
n
o
v
a
t
i
o
n
,
F
a
c
u
l
t
y
o
f
S
c
i
e
n
c
e
a
n
d
T
e
c
h
n
o
l
o
g
y
,
N
a
k
h
o
n
S
i
T
h
a
m
mara
t
R
a
j
a
b
h
a
t
U
n
i
v
e
r
si
t
y
,
Th
a
N
g
i
o
,
Th
a
i
l
a
n
d
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Feb
3
,
2
0
2
5
R
ev
is
ed
J
u
n
1
2
,
2
0
2
5
Acc
ep
ted
J
u
l 3
,
2
0
2
5
F
a
c
ial
lan
d
m
a
rk
d
e
tec
ti
o
n
p
la
y
s
a
p
iv
o
tal
r
o
le
in
v
a
ri
o
u
s
c
o
m
p
u
ter
v
isio
n
a
p
p
li
c
a
ti
o
n
s,
in
c
l
u
d
i
n
g
fa
c
e
re
c
o
g
n
it
io
n
,
e
x
p
re
ss
io
n
a
n
a
l
y
sis,
a
n
d
a
u
g
m
e
n
ted
re
a
li
ty
.
Ho
we
v
e
r,
e
x
isti
n
g
a
p
p
r
o
a
c
h
e
s
o
ften
str
u
g
g
le
with
a
c
c
u
ra
c
y
d
u
e
to
th
e
v
a
riatio
n
s
in
li
g
h
ti
n
g
,
p
o
se
s,
a
n
d
o
c
c
lu
si
o
n
.
T
o
a
d
d
re
ss
th
e
se
c
h
a
ll
e
n
g
e
s,
th
is
stu
d
y
e
x
p
l
o
re
s
th
e
i
n
teg
ra
ti
o
n
o
f
Co
n
tro
lNe
t
with
S
tab
le
D
iffu
sio
n
to
e
n
h
a
n
c
e
fa
c
ial
lan
d
m
a
rk
d
e
tec
ti
o
n
v
ia
d
a
ta
a
u
g
m
e
n
tatio
n
.
C
o
n
t
ro
lNe
t,
a
n
a
d
v
a
n
c
e
d
e
x
ten
sio
n
o
f
d
iff
u
sio
n
m
o
d
e
ls,
im
p
ro
v
e
s
ima
g
e
g
e
n
e
ra
ti
o
n
b
y
c
o
n
d
i
ti
o
n
in
g
o
u
t
p
u
ts
o
n
str
u
c
tu
r
e
d
in
p
u
ts
su
c
h
a
s
lan
d
m
a
rk
c
o
o
rd
i
n
a
tes
,
e
n
a
b
li
n
g
p
re
c
ise
c
o
n
tr
o
l
o
v
e
r
i
m
a
g
e
a
tt
rib
u
tes
.
By
lev
e
ra
g
i
n
g
a
n
n
o
tate
d
lan
d
m
a
rk
d
a
ta
fro
m
th
e
3
0
0
W
d
a
tas
e
t,
Co
n
tro
lNe
t
sy
n
t
h
e
siz
e
s
d
iv
e
rse
fa
c
i
a
l
ima
g
e
s
th
a
t
su
p
p
lem
e
n
t
trad
it
io
n
a
l
train
in
g
d
a
tas
e
ts
.
Ex
p
e
rime
n
tal
re
su
lt
s
d
e
m
o
n
stra
te
th
a
t
C
o
n
t
ro
lNe
t
-
b
a
se
d
a
u
g
m
e
n
tatio
n
re
d
u
c
e
s
th
e
i
n
tero
c
u
lar
n
o
rm
a
li
z
e
d
m
e
a
n
e
rro
r
(
I
NME
)
in
lan
d
m
a
rk
d
e
tec
ti
o
n
fro
m
a
b
a
se
li
n
e
o
f
4
.
6
7
t
o
a
ra
n
g
e
o
f
4
.
6
3
t
o
4
.
7
4
,
with
o
p
ti
m
a
l
p
a
ra
m
e
ter
tu
n
i
n
g
y
iel
d
in
g
fu
rth
e
r
a
c
c
u
ra
c
y
g
a
in
s
.
T
h
e
se
fin
d
in
g
s
h
ig
h
li
g
h
t
t
h
e
p
o
ten
ti
a
l
o
f
g
e
n
e
ra
ti
v
e
m
o
d
e
ls
in
c
o
m
p
lem
e
n
ti
n
g
d
is
c
rimin
a
ti
v
e
a
p
p
ro
a
c
h
e
s
a
n
d
i
m
p
ro
v
i
n
g
ro
b
u
stn
e
ss
a
n
d
p
re
c
isio
n
i
n
fa
c
ial
lan
d
m
a
rk
d
e
tec
ti
o
n
.
T
h
e
p
ro
p
o
se
d
m
e
th
o
d
o
ffe
rs
a
sc
a
lab
le
so
lu
ti
o
n
f
o
r
e
n
h
a
n
c
in
g
m
o
d
e
l
g
e
n
e
ra
li
z
a
ti
o
n
,
p
a
rt
icu
larly
i
n
a
p
p
li
c
a
ti
o
n
s
re
q
u
iri
n
g
h
ig
h
-
fid
e
li
ty
fa
c
ial
a
n
a
ly
sis
.
F
u
t
u
re
re
se
a
rc
h
c
a
n
e
x
ten
d
t
h
is f
ra
m
e
wo
rk
t
o
b
ro
a
d
e
r
c
o
m
p
u
ter v
isio
n
tas
k
s t
h
a
t
d
e
m
a
n
d
d
e
ta
il
e
d
fe
a
tu
re
lo
c
a
li
z
a
ti
o
n
a
n
d
str
u
c
tu
re
d
d
a
ta au
g
m
e
n
tati
o
n
.
K
ey
w
o
r
d
s
:
C
o
n
tr
o
lNet
Dee
p
lear
n
in
g
Face
im
ag
e
g
en
er
atio
n
Face
lan
d
m
ar
k
d
etec
tio
n
Ma
ch
in
e
lear
n
in
g
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
:
Pato
m
p
o
n
g
C
h
a
b
p
lan
Dep
ar
tm
en
t o
f
I
n
f
o
r
m
atio
n
T
e
ch
n
o
lo
g
y
an
d
Dig
ital I
n
n
o
v
ati
o
n
,
Facu
lty
o
f
Scien
ce
an
d
T
e
ch
n
o
lo
g
y
,
Nak
h
o
n
Si T
h
a
m
m
ar
at
R
ajab
h
at
Un
iv
er
s
ity
T
h
a
Ng
io
,
T
h
ailan
d
E
m
ail
:
p
ato
m
p
o
n
g
_
c
h
a@
n
s
tr
u
.
ac
.
th
1.
I
NT
RO
D
UCT
I
O
N
Facial
lan
d
m
ar
k
d
etec
tio
n
is
a
cr
itical
ar
ea
in
co
m
p
u
ter
v
is
io
n
,
s
u
p
p
o
r
tin
g
n
u
m
er
o
u
s
ap
p
licatio
n
s
,
in
clu
d
in
g
f
ac
ial
r
ec
o
g
n
itio
n
[
1
]
–
[
4
]
,
ex
p
r
ess
io
n
an
aly
s
is
[
5
]
–
[
7
]
,
3
D
f
ac
ial
m
o
d
elin
g
[
8
]
–
[
1
0
]
an
d
au
g
m
en
ted
r
ea
lity
[
1
1
]
,
[
1
2
]
.
T
h
ese
ap
p
li
ca
tio
n
s
r
ely
o
n
ac
cu
r
ately
id
e
n
tify
in
g
s
p
ec
if
ic
f
ac
ial
p
o
in
ts
,
o
r
lan
d
m
ar
k
s
,
th
at
r
ep
r
esen
t
ess
en
tial
f
ac
ial
f
ea
t
u
r
es
.
Ov
er
th
e
p
ast
d
ec
ad
es,
a
v
ar
iety
o
f
alg
o
r
ith
m
s
h
a
v
e
b
ee
n
p
r
o
p
o
s
ed
t
o
lo
ca
lize
f
ac
ial
k
ey
p
o
in
ts
ac
cu
r
ately
u
n
d
er
d
i
v
er
s
e
co
n
d
itio
n
s
.
E
ar
ly
ap
p
r
o
ac
h
es
wer
e
o
f
ten
b
u
ilt
o
n
s
tatis
tical
s
h
ap
e
m
o
d
els
o
r
g
r
ap
h
ical
r
e
p
r
esen
tatio
n
s
o
f
f
ac
ial
s
tr
u
ctu
r
e
.
Activ
e
s
h
ap
e
m
o
d
els
(
AS
M
)
[
1
3
]
a
n
d
ac
tiv
e
ap
p
ea
r
an
ce
m
o
d
els
(
AAM
)
[
1
4
]
ar
e
s
em
in
al
m
o
d
el
-
b
ased
f
r
am
ewo
r
k
s
th
at
iter
ativ
ely
f
it
a
p
ar
am
etr
ic
s
h
ap
e
an
d
a
p
p
ea
r
a
n
ce
to
f
ac
e
im
ag
e
s
b
y
e
n
f
o
r
cin
g
lear
n
e
d
s
h
ap
e
co
n
s
tr
ain
ts
.
T
h
ese
m
eth
o
d
s
an
d
o
th
e
r
d
ef
o
r
m
ab
le
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
15
,
No
.
5
,
Octo
b
e
r
20
25
:
4
9
0
7
-
4
9
1
5
4908
m
o
d
els
p
r
o
v
id
ed
a
f
o
u
n
d
atio
n
f
o
r
f
ac
e
alig
n
m
e
n
t,
b
u
t
th
eir
p
er
f
o
r
m
an
ce
d
eg
r
ad
es
o
n
u
n
c
o
n
s
tr
ain
ed
im
ag
es
with
lar
g
e
p
o
s
e
o
r
ex
p
r
ess
io
n
v
ar
iatio
n
.
T
o
b
etter
h
an
d
le
s
u
ch
v
a
r
iab
ilit
y
,
p
ar
t
-
b
ased
g
r
a
p
h
ical
m
o
d
els
wer
e
in
tr
o
d
u
ce
d
.
Fo
r
ex
am
p
le,
th
e
m
ix
tu
r
e
-
of
-
tr
ee
s
m
o
d
el
b
y
[
1
5
]
r
ep
r
esen
ted
f
ac
ial
la
n
d
m
ar
k
s
as
tr
ee
-
s
tr
u
ctu
r
e
d
p
ar
ts
with
g
l
o
b
al
a
n
d
lo
ca
l m
ix
tu
r
es,
en
a
b
lin
g
jo
in
t
f
ac
e
d
et
ec
tio
n
,
p
o
s
e
esti
m
atio
n
,
an
d
la
n
d
m
ar
k
lo
ca
lizatio
n
in
wild
im
ag
es
.
W
h
ile
th
e
s
e
g
r
ap
h
ical
tech
n
iq
u
es
in
cr
ea
s
ed
r
o
b
u
s
tn
ess
to
p
o
s
e,
th
eir
ac
cu
r
ac
y
was
lim
ited
b
y
th
e
r
ig
id
ity
o
f
th
e
u
n
d
er
l
y
in
g
s
h
ap
e
ass
u
m
p
tio
n
s
.
Su
b
s
eq
u
en
tly
,
d
ir
ec
t
r
eg
r
ess
io
n
m
eth
o
d
s
g
ain
e
d
p
o
p
u
lar
i
ty
f
o
r
th
eir
ef
f
icien
cy
a
n
d
ac
cu
r
ac
y
,
b
y
p
ass
in
g
ex
p
licit
s
h
ap
e
m
o
d
elin
g
.
C
ascad
ed
s
h
ap
e
r
eg
r
ess
io
n
f
r
am
ewo
r
k
s
[
1
6
]
em
e
r
g
ed
as
a
d
o
m
in
a
n
t
ap
p
r
o
ac
h
,
wh
er
e
an
i
n
itial
co
a
r
s
e
lan
d
m
ar
k
esti
m
ate
is
iter
ativ
ely
r
ef
in
e
d
b
y
a
s
eq
u
en
ce
o
f
lear
n
ed
r
eg
r
ess
o
r
s
.
B
y
lear
n
in
g
s
h
ap
e
u
p
d
ate
tr
a
n
s
f
o
r
m
atio
n
s
,
th
ese
m
eth
o
d
s
ca
n
r
ap
id
ly
co
n
v
er
g
e
to
th
e
tar
g
et
lan
d
m
ar
k
s
.
First
d
em
o
n
s
tr
ated
a
n
ex
p
licit
s
h
a
p
e
r
eg
r
ess
io
n
th
at
d
ir
ec
tly
m
ap
s
im
ag
e
f
ea
tu
r
es
to
lan
d
m
ar
k
d
is
p
lace
m
en
ts
with
o
u
t
an
y
p
ar
am
etr
ic
m
o
d
e
l
[
1
7
]
.
N
u
m
er
o
u
s
en
h
an
ce
m
e
n
ts
f
o
llo
wed
:
f
o
r
m
u
lated
th
e
s
u
p
er
v
is
ed
d
escen
t
m
eth
o
d
(
SDM
)
to
m
i
n
im
ize
a
n
o
n
lin
ea
r
least
-
s
q
u
ar
es
alig
n
m
en
t
o
b
jectiv
e
[
1
8
]
,
ap
p
lied
r
an
d
o
m
f
o
r
ests
with
co
n
d
itio
n
al
r
eg
r
ess
o
r
s
to
p
r
e
d
ict
f
ac
ial
k
ey
p
o
in
ts
in
r
ea
l
t
im
e
wh
ile
ac
co
u
n
tin
g
f
o
r
h
ea
d
p
o
s
e
[
1
9
]
.
L
ater
,
e
n
s
em
b
le
-
b
ased
r
eg
r
ess
o
r
s
wer
e
in
tr
o
d
u
ce
d
wh
ich
f
u
r
th
e
r
im
p
r
o
v
e
d
r
eliab
ilit
y
.
E
m
p
l
o
y
ed
an
en
s
em
b
le
o
f
r
eg
r
ess
io
n
tr
ee
s
,
e
n
ab
lin
g
o
n
e
-
m
illi
s
ec
o
n
d
f
ac
e
alig
n
m
e
n
t
with
co
m
p
etitiv
e
ac
cu
r
ac
y
[
2
0
]
.
T
o
r
ed
u
ce
o
v
er
f
itti
n
g
an
d
im
p
r
o
v
e
g
en
er
aliza
tio
n
,
co
m
b
in
e
d
g
r
ad
ie
n
t
-
b
o
o
s
ted
tr
ee
s
with
Gau
s
s
i
an
p
r
o
ce
s
s
es
in
a
ca
s
ca
d
e
(
cGPR
T
)
[
1
6
]
,
wh
ich
ac
ted
as a
f
o
r
m
o
f
r
eg
u
lar
ized
en
s
em
b
le
th
at
ac
h
iev
ed
s
tate
-
of
-
th
e
-
ar
t r
esu
lts
o
n
ch
allen
g
in
g
b
e
n
ch
m
ar
k
s
.
T
h
e
s
e
r
eg
r
ess
io
n
an
d
en
s
em
b
le
m
eth
o
d
s
s
ig
n
if
ican
tly
im
p
r
o
v
ed
alig
n
m
en
t
s
p
ee
d
an
d
ac
c
u
r
ac
y
,
y
et
th
eir
d
ata
-
d
r
iv
en
n
at
u
r
e
m
ea
n
t
th
at
g
en
er
al
izatio
n
to
ex
tr
em
e
p
o
s
es
o
r
ex
p
r
ess
io
n
s
was
s
till
co
n
s
tr
ain
ed
b
y
th
e
a
v
ailab
ilit
y
an
d
d
i
v
er
s
ity
o
f
tr
ai
n
in
g
d
ata
.
W
ith
th
e
r
is
e
o
f
d
ee
p
lear
n
in
g
,
co
n
v
o
lu
tio
n
al
n
eu
r
al
n
etwo
r
k
(
C
NN
)
ap
p
r
o
ac
h
es
h
a
v
e
d
r
am
atica
lly
ad
v
an
ce
d
t
h
e
s
tate
-
of
-
th
e
-
a
r
t
in
m
an
y
v
is
io
n
task
s
[
2
1
]
–
[
2
3
]
,
in
clu
d
in
g
f
ac
ial
lan
d
m
a
r
k
d
etec
tio
n
[
2
4
]
,
[
2
5
]
.
Dee
p
n
eu
r
al
n
etwo
r
k
s
ca
n
le
ar
n
r
o
b
u
s
t
f
ea
tu
r
e
r
ep
r
esen
ta
tio
n
s
an
d
im
p
licit
s
h
ap
e
co
n
s
tr
ain
ts
f
r
o
m
lar
g
e
d
atasets
.
Firs
t
d
em
o
n
s
tr
ated
a
C
NN
ca
s
ca
d
e
f
o
r
f
ac
ial
p
o
in
t
d
etec
tio
n
,
o
u
tp
er
f
o
r
m
in
g
ea
r
lier
ca
s
ca
d
ed
r
eg
r
ess
o
r
s
b
y
a
lar
g
e
m
ar
g
in
[
2
6
]
.
Su
b
s
eq
u
en
t
wo
r
k
s
lev
e
r
ag
ed
in
c
r
ea
s
in
g
ly
s
o
p
h
is
ticated
d
ee
p
m
o
d
els
a
n
d
tr
ain
in
g
s
tr
ateg
ies
.
Mu
lti
-
task
lear
n
in
g
f
r
a
m
ewo
r
k
s
wer
e
i
n
tr
o
d
u
ce
d
to
im
p
r
o
v
e
r
o
b
u
s
tn
ess
:
f
o
r
ex
a
m
p
le,
Z
h
an
g
et
a
l.
[
2
7
]
t
r
ain
ed
a
C
NN
to
p
r
ed
ict
lan
d
m
ar
k
s
to
g
e
th
er
with
h
ea
d
p
o
s
e
an
d
f
ac
ia
l
attr
ib
u
tes,
g
ain
in
g
r
esil
ien
ce
to
o
cc
lu
s
io
n
s
an
d
p
o
s
e
ch
an
g
es
th
r
o
u
g
h
s
h
ar
ed
f
ea
tu
r
e
lear
n
i
n
g
.
Oth
er
r
esear
c
h
er
s
in
teg
r
ated
3
D
f
ac
e
m
o
d
elin
g
in
to
th
e
lear
n
i
n
g
p
r
o
ce
s
s
to
h
a
n
d
le
p
r
o
f
ile
v
iews
.
C
o
m
b
in
e
d
a
ca
s
ca
d
e
d
C
NN
with
a
3
D
Mo
r
p
h
ab
le
Mo
d
el
to
alig
n
f
ac
es
ac
r
o
s
s
lar
g
e
p
o
s
es
[
2
8
]
,
an
d
p
r
o
p
o
s
ed
a
3
D
-
ass
is
ted
s
o
lu
ti
o
n
th
at
f
its
a
d
en
s
e
3
D
f
ac
e
to
2
D
lan
d
m
ar
k
s
,
th
er
eb
y
im
p
r
o
v
in
g
alig
n
m
en
t
o
f
s
elf
-
o
cc
lu
d
e
d
[
2
9
]
.
Fu
lly
co
n
v
o
lu
tio
n
al
ar
ch
itectu
r
es
an
d
h
ea
tm
a
p
r
eg
r
ess
io
n
tech
n
iq
u
es
h
av
e
also
y
ield
ed
ex
ce
llen
t
ac
cu
r
ac
y
.
A
v
er
y
d
ee
p
r
esid
u
al
n
etwo
r
k
f
o
r
lan
d
m
ar
k
lo
ca
lizatio
n
b
y
s
tu
d
y
[
2
5
]
n
ea
r
ly
s
atu
r
ated
th
e
p
er
f
o
r
m
an
ce
o
n
s
ev
er
al
2
D
an
d
3
D
f
ac
e
alig
n
m
en
t
d
atasets
,
ac
h
iev
in
g
r
em
ar
k
ab
l
y
lo
w
n
o
r
m
alize
d
m
ea
n
er
r
o
r
s
.
I
n
a
d
d
itio
n
,
im
p
r
o
v
ed
l
o
s
s
f
u
n
ctio
n
s
an
d
d
ata
h
an
d
lin
g
h
av
e
en
h
an
ce
d
C
NN
-
b
ased
alig
n
m
en
t
.
No
tab
ly
,
Fen
g
et
a
l.
[
3
0
]
in
tr
o
d
u
ce
d
th
e
W
in
g
lo
s
s
to
b
etter
p
en
alize
s
m
all
e
r
r
o
r
s
wh
ile
to
ler
atin
g
o
u
tlier
s
,
le
ad
in
g
t
o
m
o
r
e
r
o
b
u
s
t
co
n
v
er
g
en
ce
.
I
n
co
r
p
o
r
ate
d
b
o
u
n
d
ar
y
-
awa
r
e
f
ea
t
u
r
es
to
e
x
p
licitly
m
o
d
el
f
ac
e
co
n
to
u
r
i
n
f
o
r
m
atio
n
,
wh
ich
b
o
o
s
ted
la
n
d
m
ar
k
ac
cu
r
ac
y
o
n
ch
allen
g
in
g
ca
s
es
lik
e
p
r
o
f
ile
d
f
ac
es
an
d
ex
ag
g
er
ated
ex
p
r
ess
io
n
s
[
3
1
]
.
T
h
an
k
s
to
th
ese
ad
v
an
ce
s
,
m
o
d
e
r
n
n
eu
r
al
m
eth
o
d
s
ca
n
ac
h
iev
e
h
ig
h
ac
cu
r
ac
y
u
n
d
er
co
n
tr
o
lled
co
n
d
itio
n
s
.
Ho
wev
er
,
th
eir
p
er
f
o
r
m
a
n
ce
ca
n
s
till
d
eg
r
ad
e
i
n
u
n
c
o
n
s
tr
ain
ed
e
n
v
i
r
o
n
m
en
ts
d
u
e
to
th
e
in
h
er
e
n
t d
iv
er
s
ity
o
f
r
ea
l
-
wo
r
ld
f
ac
es
.
A
k
ey
r
em
ain
in
g
ch
allen
g
e
is
th
e
r
elian
ce
o
f
d
ee
p
m
o
d
els
o
n
ab
u
n
d
an
t
an
d
v
ar
ie
d
lab
el
ed
d
ata
.
I
n
p
r
ac
tice,
co
llectin
g
a
n
d
m
an
u
a
lly
an
n
o
tatin
g
a
s
u
f
f
icien
tly
d
i
v
er
s
e
f
ac
ial
lan
d
m
a
r
k
d
ataset
is
co
s
tly
an
d
lab
o
r
-
in
ten
s
iv
e
.
Ma
n
y
ex
is
tin
g
d
at
asets
h
av
e
b
iased
d
is
tr
ib
u
tio
n
s
,
s
u
ch
as
lim
ited
ex
tr
em
e
p
o
s
es,
o
cc
lu
s
io
n
s
o
r
eth
n
ic
d
iv
er
s
ity
,
ca
u
s
in
g
m
o
d
els
tr
ain
ed
o
n
th
em
to
g
en
e
r
alize
p
o
o
r
l
y
to
n
ew
d
o
m
ain
s
.
D
ata
au
g
m
en
tatio
n
is
th
er
ef
o
r
e
c
r
u
cial
to
im
p
r
o
v
e
m
o
d
el
r
o
b
u
s
tn
ess
[
3
2
]
.
C
o
n
v
e
n
tio
n
al
au
g
m
e
n
tatio
n
tech
n
i
q
u
es
s
u
ch
as
r
an
d
o
m
cr
o
p
p
in
g
,
f
lip
p
in
g
,
r
o
tatio
n
an
d
n
o
is
e
in
jectio
n
ca
n
ex
p
a
n
d
a
d
ataset
b
u
t
o
n
ly
p
r
o
d
u
ce
lim
ited
p
e
r
tu
r
b
atio
n
s
o
f
ex
is
tin
g
im
ag
es
an
d
m
ay
n
o
t
in
tr
o
d
u
ce
tr
u
l
y
n
o
v
el
f
ac
e
ap
p
ea
r
an
ce
s
o
r
g
eo
m
etr
ies
.
T
h
is
h
as
m
o
tiv
ated
th
e
u
s
e
o
f
g
en
er
ativ
e
m
o
d
els
to
s
y
n
th
etica
lly
en
lar
g
e
tr
ain
i
n
g
d
ata
.
Mo
r
e
r
ec
en
tly
,
d
if
f
u
s
io
n
m
o
d
els
[
3
3
]
,
[
3
4
]
h
av
e
em
er
g
ed
as
a
p
o
wer
f
u
l
cl
ass
o
f
g
en
er
ativ
e
m
o
d
els,
ac
h
iev
in
g
s
tate
-
of
-
t
h
e
-
ar
t
im
a
g
e
q
u
ality
an
d
d
iv
er
s
ity
in
s
y
n
th
esis
task
s
.
B
y
lev
er
a
g
in
g
a
p
r
etr
ai
n
ed
d
if
f
u
s
io
n
p
r
io
r
,
o
n
e
ca
n
g
u
i
d
e
im
ag
e
s
y
n
th
es
is
u
s
in
g
ad
d
itio
n
al
in
p
u
ts
s
u
ch
as
tex
t,
s
k
etch
es,
o
r
k
e
y
p
o
in
t
m
ap
s
[
3
5
]
.
T
h
is
s
u
g
g
ests
a
tan
talizin
g
o
p
p
o
r
tu
n
ity
:
b
y
c
o
n
d
itio
n
i
n
g
a
g
en
er
ativ
e
m
o
d
el
o
n
f
ac
ial
l
an
d
m
ar
k
co
n
f
i
g
u
r
atio
n
s
,
we
ca
n
p
r
o
d
u
ce
s
y
n
th
etic
f
ac
e
im
a
g
es
th
at
co
m
e
with
f
r
ee
lan
d
m
a
r
k
lab
els,
th
er
eb
y
cr
ea
tin
g
v
ir
tu
ally
u
n
lim
ited
tr
a
in
in
g
d
ata
with
p
r
ec
is
e
g
r
o
u
n
d
tr
u
th
.
I
n
th
is
wo
r
k
we
p
r
esen
t
a
n
o
v
el
d
ata
au
g
m
en
tatio
n
f
r
am
ewo
r
k
th
at
i
n
teg
r
ates
C
o
n
tr
o
lNet
with
Stab
le
Dif
f
u
s
io
n
to
s
y
n
th
esize
p
h
o
t
o
r
ea
lis
tic
f
ac
e
im
ag
es
co
n
d
itio
n
ed
o
n
in
p
u
t
lan
d
m
ar
k
lay
o
u
t
s
.
Ou
r
co
n
t
r
ib
u
tio
n
s
ar
e
th
r
ee
f
o
ld
.
First,
we
d
ev
e
lo
p
th
e
f
ir
s
t
d
if
f
u
s
io
n
m
o
d
el
th
at
u
s
es
co
n
d
itio
n
al
au
g
m
en
tatio
n
f
o
r
f
ac
ial
lan
d
m
ar
k
s
.
Seco
n
d
,
we
p
r
o
v
id
e
em
p
ir
ical
e
v
id
en
ce
th
at
o
u
r
m
eth
o
d
r
ed
u
c
es
n
o
r
m
alize
d
m
ea
n
er
r
o
r
co
m
p
a
r
ed
to
b
aselin
e
m
o
d
els
.
T
h
ir
d
,
we
s
h
o
w
h
o
w
s
tr
u
ctu
r
al
g
en
er
ati
v
e
au
g
m
e
n
tatio
n
ca
n
ap
p
ly
to
o
th
er
v
is
io
n
task
s
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8
7
0
8
E
n
h
a
n
cin
g
fa
cia
l la
n
d
ma
r
k
d
etec
tio
n
w
ith
C
o
n
tr
o
lN
et
-
b
a
s
e
d
d
a
ta
…
(
K
r
ita
p
h
a
t S
o
n
g
s
r
i
-
in
)
4909
s
u
ch
as
h
u
m
an
p
o
s
e
esti
m
atio
n
an
d
h
an
d
k
ey
p
o
in
t
d
etec
tio
n
wh
er
e
lab
eled
d
ata
ar
e
s
ca
r
ce
.
B
y
p
r
o
v
i
d
in
g
a
s
ca
lab
le
way
to
cr
ea
te
lar
g
e
v
o
lu
m
es
o
f
ac
cu
r
ately
lab
ele
d
d
ata,
o
u
r
m
eth
o
d
en
ab
les
th
e
tr
ain
in
g
o
f
m
o
r
e
r
o
b
u
s
t a
n
d
g
en
e
r
aliza
b
le
m
o
d
e
ls
in
f
ac
ial
an
aly
s
is
an
d
r
elate
d
f
ield
s
.
2.
M
E
T
H
O
D
T
o
o
p
tim
ize
f
ac
ial
lan
d
m
ar
k
d
etec
tio
n
,
th
is
m
eth
o
d
in
teg
r
a
tes
C
o
n
tr
o
lNet
with
Stab
le
Dif
f
u
s
io
n
f
o
r
s
y
n
th
etic
d
ata
au
g
m
e
n
tatio
n
.
B
y
co
n
d
itio
n
in
g
th
e
im
ag
e
g
en
er
atio
n
p
r
o
ce
s
s
o
n
p
r
ed
ef
in
ed
f
ac
ial
lan
d
m
ar
k
co
n
f
ig
u
r
atio
n
s
,
th
is
ap
p
r
o
ac
h
g
en
er
ates
v
ar
ie
d
tr
ain
in
g
im
ag
es
to
en
h
a
n
ce
th
e
r
o
b
u
s
tn
ess
an
d
ac
cu
r
ac
y
o
f
f
ac
ial
lan
d
m
ar
k
d
etec
tio
n
.
T
h
e
f
o
llo
win
g
s
u
b
s
ec
tio
n
s
d
escr
i
b
e
th
e
d
ataset,
m
o
d
el
ar
ch
itect
u
r
e,
lo
s
s
f
u
n
ctio
n
s
,
tr
ain
in
g
s
tr
ateg
y
,
a
n
d
im
p
lem
e
n
tatio
n
d
etails
.
2
.
1
.
Da
t
a
s
et
s
T
h
i
s
s
t
u
d
y
u
t
i
l
i
z
e
s
t
w
o
p
r
i
m
a
r
y
d
a
t
a
s
e
t
s
f
o
r
t
r
a
i
n
i
n
g
a
n
d
e
v
a
l
u
a
t
i
n
g
t
h
e
f
a
c
i
a
l
l
a
n
d
m
a
r
k
d
e
t
e
c
t
i
o
n
m
o
d
e
l
:
t
h
e
3
0
0
W
d
a
t
a
s
e
t
[
3
6
]
,
a
w
i
d
e
l
y
e
s
t
a
b
l
i
s
h
e
d
b
e
n
c
h
m
a
r
k
f
o
r
f
a
c
i
a
l
l
a
n
d
m
a
r
k
d
e
t
e
c
t
i
o
n
,
a
n
d
a
C
o
n
t
r
o
l
N
e
t
-
b
a
s
e
d
a
u
g
m
e
n
t
e
d
d
a
t
a
s
e
t
.
T
h
e
C
o
n
t
r
o
l
N
e
t
-
b
a
s
e
d
a
u
g
m
e
n
t
e
d
d
a
t
a
s
e
t
g
e
n
e
r
a
t
e
s
s
y
n
t
h
e
t
i
c
i
m
a
g
e
s
c
o
n
d
i
t
i
o
n
e
d
o
n
f
a
c
i
a
l
l
a
n
d
m
a
r
k
s
f
r
o
m
t
h
e
3
0
0
W
d
a
t
a
s
e
t
.
T
h
e
s
e
d
a
t
a
s
e
t
s
t
o
g
e
t
h
e
r
p
r
o
v
i
d
e
b
o
t
h
r
e
a
l
a
n
d
s
y
n
t
h
e
t
i
c
d
a
t
a
,
a
l
l
o
w
i
n
g
f
o
r
a
s
y
s
t
e
m
a
t
i
c
e
x
a
m
i
n
a
t
i
o
n
o
f
m
o
d
e
l
p
e
r
f
o
r
m
a
n
c
e
a
c
r
o
s
s
v
a
r
i
o
u
s
d
a
t
a
c
o
n
f
i
g
u
r
a
t
i
o
n
s
.
2
.
1
.
1
.
T
he
3
0
0
W
d
a
t
a
s
et
T
h
e
3
0
0
W
d
ataset
is
a
cr
u
c
ial
b
en
ch
m
a
r
k
i
n
th
e
f
ac
ial
l
an
d
m
ar
k
d
etec
tio
n
d
o
m
ain
,
o
f
f
er
in
g
a
d
iv
er
s
e
co
llectio
n
o
f
f
ac
ial
im
ag
es
cu
r
ated
to
ch
allen
g
e
an
d
ev
alu
ate
d
etec
tio
n
alg
o
r
i
th
m
s
ef
f
ec
tiv
ely
.
I
t
in
clu
d
es
v
ar
io
u
s
s
u
b
s
ets
d
esig
n
ed
t
o
s
im
u
late
r
ea
l
-
w
o
r
ld
s
ce
n
ar
io
s
,
ca
p
tu
r
in
g
a
b
r
o
ad
s
p
ec
tr
u
m
o
f
f
ac
ia
l
co
n
d
itio
n
s
,
s
u
ch
as
d
if
f
er
e
n
t
lig
h
tin
g
en
v
i
r
o
n
m
e
n
ts
,
f
ac
ial
ex
p
r
ess
io
n
s
,
an
d
lev
els
o
f
o
cc
l
u
s
io
n
.
T
h
is
d
ataset
s
er
v
es
as
th
e
p
r
im
ar
y
s
o
u
r
c
e
o
f
an
n
o
tated
r
ea
l
-
wo
r
ld
d
ata
f
o
r
tr
ain
i
n
g
a
n
d
e
v
alu
atin
g
f
ac
ial
lan
d
m
ar
k
d
etec
tio
n
m
o
d
els
.
I
t
in
clu
d
es 3
,
1
4
8
tr
ain
in
g
im
ag
es
a
n
d
6
0
0
t
esti
n
g
im
ag
es,
p
r
o
v
i
d
in
g
a
s
u
b
s
tan
tial
v
o
lu
m
e
o
f
d
ata
f
o
r
r
o
b
u
s
t
m
o
d
el
tr
ai
n
in
g
an
d
a
n
aly
s
is
.
Fig
u
r
e
1
d
is
p
lay
s
s
am
p
le
im
ag
es
f
r
o
m
t
h
e
3
0
0
W
d
ataset
,
illu
s
tr
atin
g
th
e
d
iv
er
s
ity
o
f
f
ac
ial
f
ea
tu
r
es
a
n
d
lan
d
m
a
r
k
s
th
a
t
m
ak
e
th
is
d
ataset
in
v
alu
ab
le
f
o
r
r
ig
o
r
o
u
s
test
in
g
an
d
v
alid
atio
n
.
Fig
u
r
e
1
(
a
)
ill
u
s
tr
ates
ex
am
p
les
f
r
o
m
th
e
3
0
0
W
d
ataset,
h
ig
h
lig
h
tin
g
th
e
d
iv
er
s
ity
o
f
f
ac
ial
v
ar
iatio
n
s
an
d
t
h
e
d
etailed
an
n
o
tatio
n
o
f
f
ac
ial
lan
d
m
a
r
k
s
.
2
.
1
.
2
.
Co
ntr
o
lNet
-
ba
s
ed
a
ug
m
ent
ed
da
t
a
s
et
T
o
s
u
p
p
lem
en
t
t
h
e
3
0
0
W
d
a
taset,
a
s
y
n
th
etic
d
ataset
was
cr
ea
ted
u
s
in
g
C
o
n
tr
o
lNet,
an
ad
v
an
ce
d
im
ag
e
g
en
e
r
atio
n
m
o
d
el
ca
p
ab
le
o
f
p
r
o
d
u
cin
g
r
ea
lis
tic
f
ac
ial
im
ag
es
co
n
d
itio
n
e
d
o
n
s
p
ec
if
ic
lan
d
m
ar
k
co
n
f
ig
u
r
atio
n
s
.
C
o
n
tr
o
lNet
was
ap
p
lied
to
t
h
e
3
0
0
W
lan
d
m
ar
k
an
n
o
tatio
n
s
to
g
en
er
ate
s
y
n
th
etic
im
ag
es
th
at
clo
s
ely
ad
h
er
e
t
o
th
e
s
tr
u
ct
u
r
al
f
ea
tu
r
es
o
f
th
e
o
r
ig
in
al
d
ataset,
en
h
an
cin
g
d
iv
er
s
ity
i
n
tr
ain
in
g
d
ata
b
y
in
tr
o
d
u
cin
g
n
ew
v
ar
iatio
n
s
in
lig
h
tin
g
,
p
o
s
e,
an
d
f
ac
ial
ex
p
r
ess
io
n
s
.
T
h
is
au
g
m
en
ted
d
ataset
was
g
en
er
ated
at
v
ar
y
in
g
r
atio
s
r
elativ
e
to
th
e
o
r
ig
in
al
d
ataset,
f
r
o
m
0
%
to
1
0
0
%
in
s
tep
s
o
f
1
0
%
,
allo
win
g
f
o
r
ex
p
e
r
im
en
tal
ev
alu
atio
n
o
f
d
if
f
er
en
t
r
ea
l
-
to
-
s
y
n
th
etic
d
ata
co
m
b
in
atio
n
s
.
B
y
in
teg
r
atin
g
C
o
n
tr
o
lNet
-
b
ased
s
y
n
th
etic
im
ag
es,
th
e
a
u
g
m
e
n
ted
d
ataset
p
r
o
v
id
es
a
s
ca
lab
le
s
o
lu
tio
n
to
b
o
o
s
t
m
o
d
el
g
e
n
er
aliza
tio
n
a
n
d
r
o
b
u
s
tn
ess
ac
r
o
s
s
a
r
an
g
e
o
f
f
ac
ial
lan
d
m
ar
k
d
etec
tio
n
s
ce
n
ar
io
s
.
Fig
u
r
e
1
(
b
)
s
h
o
wca
s
es
ex
am
p
les
f
r
o
m
th
e
C
o
n
tr
o
lNet
-
b
ased
au
g
m
e
n
ted
d
ataset,
illu
s
tr
atin
g
h
o
w
th
is
s
y
n
th
etic
d
ata
clo
s
ely
r
esem
b
les
r
ea
l
-
wo
r
ld
co
n
d
itio
n
s
an
d
en
h
an
ce
s
tr
ain
in
g
d
iv
er
s
ity
.
2
.
2
.
M
o
del
a
rc
hite
ct
ure
Fo
r
ef
f
icien
t c
o
m
p
u
tatio
n
s
,
o
u
r
m
o
d
el
is
d
esig
n
ed
s
p
ec
if
ically
to
h
an
d
le
th
e
s
in
g
le
o
b
jectiv
e
o
f
f
ac
ial
lan
d
m
ar
k
d
etec
tio
n
with
p
r
ec
is
io
n
.
T
h
e
n
etwo
r
k
b
e
g
in
s
p
r
o
c
ess
in
g
with
a
6
4
×6
4
×3
co
lo
r
i
m
ag
e
as in
p
u
t
.
T
h
is
in
p
u
t
is
s
eq
u
en
tially
p
ass
ed
th
r
o
u
g
h
f
iv
e
3
×3
c
o
n
v
o
lu
tio
n
al
lay
er
s
,
ea
ch
u
s
in
g
a
r
ec
tifie
d
lin
ea
r
u
n
it
(
R
eL
U
)
ac
tiv
atio
n
f
u
n
ctio
n
t
o
in
tr
o
d
u
ce
n
o
n
-
lin
ea
r
ity
,
ad
d
r
ess
in
g
c
h
allen
g
es
lik
e
th
e
v
an
is
h
in
g
g
r
ad
ien
t
.
Af
ter
ea
c
h
co
n
v
o
l
u
tio
n
al
lay
er
,
a
m
ax
-
p
o
o
lin
g
o
p
er
atio
n
r
ed
u
ce
s
th
e
s
p
atial
d
im
en
s
io
n
s
b
y
h
alf
,
wh
ich
en
h
an
ce
s
th
e
m
o
d
el
’
s
tr
an
s
latio
n
al
in
v
ar
ia
n
ce
an
d
co
n
d
en
s
es
in
f
o
r
m
at
io
n
.
E
ac
h
o
f
th
e
f
iv
e
co
n
v
o
lu
tio
n
al
lay
er
s
is
s
tr
u
ctu
r
ed
with
k
er
n
els
d
ef
in
e
d
b
y
W
id
th
×H
eig
h
t×I
n
p
u
t×O
u
tp
u
t,
wh
er
e
th
e
k
er
n
el
s
ize
s
p
ec
if
ies
ea
ch
lay
er
’
s
in
p
u
t
a
n
d
o
u
tp
u
t
ch
an
n
els,
e
n
s
u
r
in
g
e
f
f
icien
t
f
ea
tu
r
e
ex
tr
ac
tio
n
.
Fo
llo
win
g
th
ese
f
o
u
n
d
atio
n
al
lay
er
s
,
th
e
n
etwo
r
k
in
clu
d
es
f
u
lly
c
o
n
n
ec
ted
lay
er
s
to
p
r
o
ce
s
s
th
e
ex
tr
ac
ted
f
ea
tu
r
es
.
T
h
ese
f
u
l
ly
co
n
n
ec
te
d
lay
er
s
tr
an
s
f
o
r
m
t
h
e
s
p
atial
in
f
o
r
m
ati
o
n
in
to
a
f
in
al
o
u
tp
u
t
v
ec
to
r
o
f
2
L
v
alu
es,
w
h
er
e
ea
c
h
p
ai
r
o
f
v
alu
es
r
ep
r
esen
ts
th
e
x
an
d
y
co
o
r
d
i
n
ates
o
f
ea
c
h
o
f
L
f
ac
ial
lan
d
m
ar
k
s
.
I
n
t
h
is
s
etu
p
,
L
is
co
n
f
ig
u
r
ed
f
o
r
6
8
lan
d
m
ar
k
p
o
in
ts
to
ca
p
tu
r
e
d
etailed
f
ac
ial
f
ea
t
u
r
es
ac
cu
r
ately
.
T
h
is
s
tr
u
ctu
r
e
allo
ws
th
e
m
o
d
el
to
ex
ce
l
in
p
r
ec
is
e
lan
d
m
ar
k
lo
ca
lizatio
n
,
ef
f
ec
tiv
ely
ca
p
tu
r
in
g
th
e
ess
en
tial
d
etails
r
eq
u
ir
ed
f
o
r
f
ac
ial
an
aly
s
is
.
T
h
e
ar
ch
itectu
r
e
o
f
th
e
m
o
d
el
is
d
ep
icted
in
Fig
u
r
e
2
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
15
,
No
.
5
,
Octo
b
e
r
20
25
:
4
9
0
7
-
4
9
1
5
4910
(
a
)
(
b
)
Fig
u
r
e
1
.
E
x
am
p
les f
r
o
m
th
e
d
atasets
u
s
ed
f
o
r
tr
ain
in
g
a
n
d
e
v
alu
atio
n
: (
a
)
s
am
p
le
im
ag
es f
r
o
m
th
e
3
0
0
W
d
ataset
d
is
p
lay
in
g
d
iv
er
s
e
f
ac
i
al
ex
p
r
ess
io
n
s
,
lig
h
tin
g
c
o
n
d
iti
o
n
s
,
an
d
o
cc
lu
s
io
n
s
with
an
n
o
tated
lan
d
m
ar
k
s
.
(
b
)
s
y
n
th
etic
im
a
g
es f
r
o
m
th
e
C
o
n
tr
o
lNet
-
b
ased
au
g
m
en
ted
d
ataset,
g
en
er
ated
u
s
in
g
3
0
0
W
lan
d
m
ar
k
co
n
f
ig
u
r
atio
n
s
to
in
tr
o
d
u
ce
ad
d
itio
n
al
v
ar
iatio
n
s
in
p
o
s
e,
lig
h
tin
g
,
an
d
ex
p
r
ess
io
n
Fig
u
r
e
2
.
Ov
e
r
all
ar
ch
itectu
r
e
:
a
s
eq
u
en
ce
o
f
f
iv
e
3
×3
c
o
n
v
+
R
eL
U
+
m
ax
‐
p
o
o
l b
l
o
ck
s
,
f
o
llo
wed
b
y
f
u
lly
co
n
n
ec
ted
la
y
er
s
th
at
o
u
t
p
u
t 2
×
6
8
lan
d
m
ar
k
c
o
o
r
d
in
ates
2
.
3
.
L
o
s
s
f
un
ct
io
n
T
h
e
m
o
d
el
’
s
tr
ain
in
g
o
b
jectiv
e
f
o
c
u
s
es
o
n
m
i
n
im
izin
g
lo
ca
lizatio
n
er
r
o
r
f
o
r
f
ac
ial
lan
d
m
ar
k
d
etec
tio
n
.
T
h
e
m
ea
n
ab
s
o
lu
te
er
r
o
r
(
MA
E
)
is
u
s
ed
to
q
u
an
tify
th
e
d
is
cr
ep
an
cy
b
etwe
e
n
th
e
p
r
ed
icted
an
d
ac
tu
al
lan
d
m
ar
k
p
o
s
itio
n
s
,
en
s
u
r
in
g
ac
cu
r
ac
y
in
f
ac
ial
lan
d
m
ar
k
lo
ca
lizatio
n
.
T
h
e
l
o
s
s
f
u
n
ctio
n
is
d
e
f
in
ed
in
(
1
):
=
1
∑
∑
|
−
̂
|
(
1
)
wh
er
e
is
th
e
n
u
m
b
er
o
f
im
a
g
es,
r
ep
r
esen
ts
th
e
t
o
tal
lan
d
m
ar
k
s
in
ea
ch
im
ag
e,
is
th
e
g
r
o
u
n
d
tr
u
t
h
lo
ca
tio
n
o
f
t
h
e
-
th
lan
d
m
a
r
k
in
im
ag
e
,
an
d
̂
is
th
e
p
r
ed
icted
lo
ca
tio
n
g
en
er
ated
b
y
th
e
m
o
d
el
.
T
h
is
MA
E
-
b
ased
lo
s
s
f
u
n
ctio
n
en
s
u
r
es
ac
cu
r
ate
lo
ca
lizatio
n
b
y
li
n
ea
r
ly
p
en
alizin
g
er
r
o
r
s
ac
r
o
s
s
th
e
p
r
ed
icted
co
o
r
d
in
ates
.
2
.
4
.
M
o
del
t
ra
ini
ng
s
t
ra
t
eg
y
T
o
ass
ess
th
e
ef
f
ec
ts
o
f
s
y
n
th
etic
d
ata
o
n
f
ac
ial
la
n
d
m
ar
k
d
etec
tio
n
,
th
e
m
o
d
el
was
tr
a
in
ed
with
d
atasets
co
n
tain
in
g
d
i
f
f
er
en
t
r
atio
s
o
f
C
o
n
tr
o
lNet
-
g
en
er
ate
d
im
ag
es
to
o
r
ig
in
al
im
ag
es,
r
an
g
in
g
f
r
o
m
0
.
0
t
o
1
.
0
in
s
tep
s
o
f
0
.
1
.
E
ac
h
r
atio
was
tr
ea
ted
a
s
a
s
ep
ar
ate
ex
p
er
im
en
t,
with
th
e
p
r
o
p
o
r
tio
n
o
f
s
y
n
th
etic
to
r
ea
l
im
ag
es
h
eld
co
n
s
tan
t
th
r
o
u
g
h
o
u
t
th
e
tr
ain
in
g
p
r
o
ce
s
s
.
B
y
s
y
s
tem
atica
lly
v
ar
y
in
g
t
h
ese
r
atio
s
,
th
is
ap
p
r
o
ac
h
en
ab
les
a
co
m
p
ar
ativ
e
an
aly
s
is
o
f
h
o
w
d
if
f
er
e
n
t
lev
els
o
f
s
y
n
th
etic
d
ata
in
f
lu
en
ce
m
o
d
el
p
er
f
o
r
m
a
n
ce
,
p
r
o
v
id
i
n
g
in
s
ig
h
ts
in
to
th
e
o
p
tim
al
d
ataset
co
m
p
o
s
itio
n
f
o
r
en
h
a
n
cin
g
ac
cu
r
ac
y
an
d
r
o
b
u
s
tn
ess
in
f
ac
ial
lan
d
m
ar
k
d
etec
tio
n
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8
7
0
8
E
n
h
a
n
cin
g
fa
cia
l la
n
d
ma
r
k
d
etec
tio
n
w
ith
C
o
n
tr
o
lN
et
-
b
a
s
e
d
d
a
ta
…
(
K
r
ita
p
h
a
t S
o
n
g
s
r
i
-
in
)
4911
As
illu
s
tr
ated
in
Fig
u
r
e
3
,
e
ac
h
ex
p
er
im
en
tal
s
etu
p
r
e
p
r
e
s
en
ts
a
u
n
iq
u
e
d
ataset
co
m
p
o
s
itio
n
b
y
b
alan
cin
g
r
ea
l
an
d
s
y
n
th
etic
d
ata
ac
co
r
d
in
g
to
th
e
d
esig
n
ate
d
r
atio
.
T
h
is
s
tr
u
ctu
r
e
allo
ws
th
e
m
o
d
el
to
lear
n
f
r
o
m
b
o
th
n
atu
r
al
an
d
au
g
m
en
ted
f
ac
ial
v
ar
iatio
n
s
,
ex
am
in
in
g
h
o
w
s
y
n
th
etic
d
ata
co
n
tr
ib
u
tes
to
g
en
er
aliza
tio
n
ac
r
o
s
s
d
iv
er
s
e
f
ac
ial
co
n
d
itio
n
s
.
B
y
co
m
p
ar
i
n
g
p
er
f
o
r
m
a
n
ce
ac
r
o
s
s
th
ese
co
n
f
ig
u
r
atio
n
s
,
th
e
ex
p
er
im
en
ts
aim
to
id
en
tify
t
h
e
m
o
s
t
ef
f
ec
tiv
e
r
atio
o
f
s
y
n
th
etic
au
g
m
en
tatio
n
f
o
r
en
h
an
cin
g
th
e
m
o
d
el
’
s
ab
ilit
y
to
ac
cu
r
ately
d
etec
t f
ac
ial
lan
d
m
ar
k
s
.
Fig
u
r
e
3
.
Data
s
et
au
g
m
e
n
tatio
n
s
tr
ateg
y
:
f
o
r
ea
ch
e
x
p
er
im
en
t,
a
f
r
ac
tio
n
o
f
C
o
n
tr
o
lNet
-
g
e
n
er
ated
s
y
n
th
etic
im
ag
es is
ad
d
itio
n
ally
ad
d
e
d
o
n
to
p
o
f
th
e
r
ea
l 3
0
0
W
im
ag
e
s
2
.
5
.
I
m
ple
m
ent
a
t
io
n
d
et
a
ils
T
h
is
f
ac
ial
lan
d
m
ar
k
d
etec
tio
n
m
o
d
el
was
im
p
lem
en
ted
u
s
in
g
Py
th
o
n
an
d
T
en
s
o
r
Flo
w,
lev
er
ag
in
g
its
f
lex
ib
ilit
y
f
o
r
d
ee
p
lear
n
in
g
task
s
.
I
n
p
u
t
i
m
ag
es
wer
e
n
o
r
m
alize
d
to
a
r
an
g
e
o
f
0
an
d
1
b
y
d
i
v
id
in
g
p
ix
el
v
alu
es
b
y
2
5
5
.
T
h
e
m
o
d
el
w
as
tr
ain
ed
u
s
in
g
th
e
A
d
am
o
p
tim
izer
,
with
a
p
iece
wis
e
co
n
s
tan
t
lear
n
in
g
r
ate
s
ch
ed
u
le
.
T
h
e
in
itial
lear
n
in
g
r
ate
o
f
1
×1
0
−
3
was
r
ed
u
ce
d
to
1
×1
0
−
4
af
ter
th
e
f
ir
s
t
th
ir
d
o
f
th
e
tr
ain
in
g
ep
o
ch
s
an
d
f
u
r
th
e
r
to
1
×1
0
−
5
af
ter
th
e
s
ec
o
n
d
th
ir
d
,
en
s
u
r
in
g
g
r
a
d
u
al
r
ef
in
em
en
t
o
f
m
o
d
el
p
ar
am
eter
s
.
T
r
ain
in
g
was
co
n
d
u
cte
d
f
o
r
1
0
0
0
ep
o
ch
s
wi
th
a
b
atc
h
s
ize
o
f
6
4
.
R
eg
u
la
r
izatio
n
was
ap
p
lied
u
s
in
g
L
2
weig
h
t
d
ec
a
y
5
×
1
0
−
4
to
m
itig
ate
o
v
er
f
itti
n
g
.
A
u
g
m
en
tatio
n
tec
h
n
iq
u
es,
in
clu
d
in
g
r
a
n
d
o
m
r
o
tatio
n
s
,
f
lip
p
i
n
g
,
c
r
o
p
p
in
g
,
an
d
Gau
s
s
ian
b
lu
r
r
in
g
,
wer
e
em
p
l
o
y
ed
to
e
n
h
an
ce
d
ata
d
iv
er
s
ity
an
d
r
o
b
u
s
tn
ess
.
T
h
e
im
p
le
m
en
tatio
n
s
tr
ateg
y
,
co
m
b
in
in
g
ef
f
icien
t
ar
c
h
itectu
r
e,
a
d
ap
tiv
e
lear
n
in
g
r
ates,
an
d
a
u
g
m
en
tatio
n
,
f
ac
ilit
ated
ac
cu
r
ate
an
d
r
o
b
u
s
t
p
r
ed
ictio
n
o
f
f
ac
ial
lan
d
m
ar
k
s
u
n
d
er
v
ar
ied
c
o
n
d
itio
n
s
.
3.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
N
T
h
is
s
ec
tio
n
p
r
esen
ts
an
e
x
p
er
im
en
tal
ev
alu
atio
n
o
f
th
e
p
r
o
p
o
s
ed
m
eth
o
d
,
f
o
c
u
s
in
g
o
n
t
h
e
im
p
ac
t
o
f
C
o
n
tr
o
lNet
-
b
ased
s
y
n
th
etic
d
ata
au
g
m
e
n
tatio
n
o
n
f
ac
ial
la
n
d
m
ar
k
d
etec
tio
n
p
e
r
f
o
r
m
an
c
e
.
T
h
e
in
ter
o
c
u
lar
n
o
r
m
alize
d
m
ea
n
er
r
o
r
(
I
NM
E
)
is
em
p
lo
y
ed
as
th
e
p
r
im
ar
y
ev
alu
atio
n
m
etr
ic,
p
r
o
v
id
in
g
a
s
ca
le
-
in
d
ep
en
d
en
t
ass
es
s
m
en
t
o
f
lan
d
m
ar
k
lo
ca
lizatio
n
ac
cu
r
ac
y
.
C
o
m
p
a
r
ativ
e
an
aly
s
es
ar
e
co
n
d
u
ct
ed
ac
r
o
s
s
v
ar
io
u
s
au
g
m
en
tatio
n
r
atio
s
an
d
p
ar
a
m
eter
s
ettin
g
s
to
d
eter
m
i
n
e
th
e
o
p
tim
al
co
n
f
ig
u
r
atio
n
s
f
o
r
a
ch
iev
in
g
r
o
b
u
s
t
an
d
p
r
ec
is
e
f
ac
ial
lan
d
m
ar
k
d
etec
ti
o
n
.
3
.
1
.
M
et
rics
T
h
e
I
NM
E
p
r
o
v
id
es
a
r
ef
in
e
d
m
etr
ic
s
p
ec
if
ically
s
u
ited
f
o
r
ev
alu
atin
g
f
ac
ial
lan
d
m
ar
k
d
etec
tio
n
.
T
h
is
m
ea
s
u
r
e
ca
lcu
lates
th
e
av
er
ag
e
d
if
f
er
en
ce
b
etwe
en
th
e
p
r
ed
icted
an
d
ac
tu
al
lan
d
m
ar
k
p
o
s
itio
n
s
,
with
n
o
r
m
aliza
tio
n
b
ased
o
n
th
e
in
t
er
o
cu
lar
d
is
tan
ce
,
d
ef
in
e
d
as
th
e
d
is
tan
ce
b
etwe
en
th
e
two
o
u
ter
m
o
s
t
p
o
in
ts
o
f
th
e
ey
es
.
T
h
is
n
o
r
m
aliza
tio
n
e
n
s
u
r
es
a
s
ca
le
-
in
d
e
p
en
d
e
n
t
ass
ess
m
en
t
.
T
h
e
f
o
r
m
u
la
f
o
r
I
N
ME
is
p
r
esen
ted
in
(
2
):
=
1
∑
√
∑
(
−
̂
)
2
(
2
)
wh
er
e
r
ep
r
esen
ts
th
e
n
u
m
b
er
o
f
im
ag
es,
is
th
e
to
tal
n
u
m
b
er
o
f
lan
d
m
ar
k
s
i
n
ea
ch
im
ag
e,
a
n
d
̂
ar
e
th
e
g
r
o
u
n
d
tr
u
th
a
n
d
p
r
e
d
icted
lan
d
m
ar
k
p
o
s
itio
n
s
,
r
esp
ec
tiv
ely
,
an
d
is
th
e
d
is
tan
ce
b
et
wee
n
th
e
o
u
te
r
co
r
n
er
s
o
f
th
e
ey
es in
ea
c
h
im
ag
e
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
15
,
No
.
5
,
Octo
b
e
r
20
25
:
4
9
0
7
-
4
9
1
5
4912
3
.
2
.
M
et
ho
ds
c
o
m
pa
riso
n
T
h
e
r
esu
lts
o
f
th
e
e
x
p
er
im
e
n
ts
,
p
r
esen
ted
in
T
ab
le
1
,
s
h
o
w
th
e
p
e
r
f
o
r
m
an
ce
o
f
t
h
e
f
ac
ial
lan
d
m
ar
k
d
etec
tio
n
m
o
d
el
with
v
a
r
y
in
g
r
atio
s
o
f
C
o
n
tr
o
lNet
-
b
ased
au
g
m
en
ted
d
ata,
r
an
g
in
g
f
r
o
m
0
to
1
.
T
h
e
I
NM
E
is
u
s
ed
as
a
k
ey
p
er
f
o
r
m
a
n
ce
in
d
icato
r
,
wh
er
e
lo
wer
I
NM
E
v
alu
es
in
d
icate
h
ig
h
e
r
ac
c
u
r
ac
y
in
lan
d
m
ar
k
p
r
ed
ictio
n
.
Fr
o
m
T
ab
le
1
,
it
ca
n
b
e
o
b
s
er
v
ed
th
at
th
e
b
aselin
e
m
o
d
el,
with
o
u
t
a
n
y
s
y
n
t
h
etic
au
g
m
en
tatio
n
,
ac
h
iev
es
an
I
NM
E
o
f
4
.
67
.
A
s
th
e
au
g
m
en
tatio
n
r
atio
i
n
cr
e
ases
f
r
o
m
0
.
1
to
1
,
th
e
I
NM
E
f
lu
ctu
ates
s
lig
h
tly
b
etwe
en
4
.
6
3
an
d
4
.
7
4
,
in
d
i
ca
tin
g
th
at
d
if
f
er
en
t
lev
els
o
f
au
g
m
en
ted
d
ata
h
a
v
e
v
ar
ie
d
ef
f
ec
ts
o
n
m
o
d
el
ac
cu
r
ac
y
.
Fu
r
th
er
in
s
ig
h
t
in
to
th
e
ef
f
ec
t
o
f
C
o
n
tr
o
lNet
-
au
g
m
en
ted
d
ata
o
n
m
o
d
el
lear
n
i
n
g
is
illu
s
tr
ated
in
Fig
u
r
e
4
(
a
)
an
d
4
(
b
)
,
wh
ich
d
is
p
lay
r
aw
an
d
m
o
v
in
g
av
er
ag
e
o
f
I
NM
E
v
alu
es
o
v
er
tr
ain
in
g
iter
atio
n
s
,
clar
if
y
in
g
lo
n
g
-
ter
m
p
er
f
o
r
m
a
n
ce
tr
en
d
s
.
Du
r
in
g
th
e
in
itial
th
ir
d
o
f
th
e
tr
ain
in
g
iter
atio
n
s
,
I
NM
E
d
ec
r
ea
s
es
s
h
ar
p
ly
f
r
o
m
ap
p
r
o
x
im
ately
6
.
0
,
d
em
o
n
s
tr
atin
g
th
at
th
e
m
o
d
el
r
ap
id
ly
ad
a
p
ts
to
th
e
tr
ain
in
g
d
ata
.
Af
ter
th
is
in
itial
d
r
o
p
,
I
NM
E
s
tab
ilizes
b
etwe
en
4
.
8
an
d
5
.
4
d
u
r
i
n
g
t
h
e
s
ec
o
n
d
th
i
r
d
o
f
th
e
iter
ati
o
n
s
,
with
a
g
en
e
r
al
d
o
wn
war
d
tr
e
n
d
,
i
n
d
icatin
g
c
o
n
tin
u
ed
m
o
d
el
im
p
r
o
v
em
e
n
t
.
T
h
e
im
p
ac
t
o
f
v
ar
y
in
g
th
e
L
am
b
d
a
p
ar
a
m
eter
o
n
I
NM
E
is
also
n
o
tab
le
.
L
o
wer
L
am
b
d
a
v
alu
es
(
b
elo
w
0
.
5
)
ar
e
ass
o
ciate
d
with
lo
wer
I
NM
E
,
s
u
g
g
esti
n
g
th
at
s
elec
tin
g
an
o
p
tim
al
L
am
b
d
a
v
alu
e
ca
n
s
ig
n
if
ican
tly
en
h
an
ce
m
o
d
el
p
er
f
o
r
m
a
n
ce
.
Af
te
r
th
e
f
in
al
th
i
r
d
o
f
t
h
e
tr
ai
n
in
g
iter
atio
n
s
,
I
NM
E
co
n
v
e
r
g
es
to
a
s
tead
y
r
an
g
e
o
f
4
.
6
to
4
.
8
a
cr
o
s
s
all
L
am
b
d
a
v
alu
es,
d
e
m
o
n
s
tr
atin
g
th
at
th
e
m
o
d
el
h
as
ac
h
iev
ed
s
tab
le
lan
d
m
ar
k
p
r
ed
ictio
n
ac
cu
r
ac
y
.
T
h
e
m
o
v
in
g
av
er
ag
e
in
Fig
u
r
e
4
(
b
)
ef
f
ec
tiv
ely
s
m
o
o
th
s
o
u
t
r
aw
I
NM
E
f
lu
ctu
atio
n
s
,
m
ak
in
g
th
e
tr
e
n
d
o
f
p
e
r
f
o
r
m
an
ce
im
p
r
o
v
e
m
en
t
m
o
r
e
ap
p
ar
en
t
.
T
h
e
e
x
p
er
im
en
tal
r
esu
lts
h
ig
h
lig
h
t
t
h
e
ef
f
ec
tiv
en
ess
o
f
u
s
in
g
C
o
n
tr
o
lNet
-
au
g
m
e
n
ted
d
ata
an
d
th
e
im
p
o
r
tan
ce
o
f
tu
n
in
g
L
am
b
d
a
to
ac
h
iev
e
o
p
tim
al
p
er
f
o
r
m
an
ce
in
f
ac
ial
lan
d
m
ar
k
d
etec
tio
n
.
T
h
e
an
aly
s
is
u
n
d
e
r
s
co
r
es
th
at
th
e
in
teg
r
atio
n
o
f
ca
r
ef
u
lly
ch
o
s
en
s
y
n
th
etic
d
a
ta
r
atio
s
,
alo
n
g
with
a
n
o
p
tim
al
L
am
b
d
a,
ca
n
en
h
an
ce
m
o
d
el
r
o
b
u
s
tn
ess
an
d
p
r
ec
is
io
n
in
lan
d
m
ar
k
l
o
ca
lizatio
n
.
T
ab
le
1
.
I
n
ter
o
cu
la
r
n
o
r
m
alize
d
m
ea
n
er
r
o
r
(
I
NM
E
)
o
f
th
e
f
a
cial
lan
d
m
ar
k
d
etec
tio
n
m
o
d
el
f
o
r
v
a
r
y
in
g
C
o
n
tr
o
lNet
-
b
ased
au
g
m
en
tatio
n
r
atio
s
(
)
.
L
o
wer
I
NM
E
in
d
i
ca
tes h
ig
h
er
lan
d
m
ar
k
p
r
ed
icti
o
n
ac
cu
r
ac
y
R
a
t
i
o
s
I
N
M
E
↓
0
(
B
a
se
l
i
n
e
)
4
.
67
0
.
1
4
.
68
0
.
2
4
.
63
0
.
3
4
.
68
0
.
4
4
.
74
0
.
5
4
.
63
0
.
6
4
.
69
0
.
7
4
.
71
0
.
8
4
.
70
0
.
9
4
.
68
1
4
.
73
(
a)
(
b
)
Fig
u
r
e
4
.
I
m
p
ac
t o
f
L
am
b
d
a
(
)
o
n
I
NM
E
d
u
r
in
g
tr
ain
in
g
: (
a
)
r
aw
I
NM
E
v
alu
es p
er
iter
atio
n
f
o
r
d
i
f
f
er
en
t
L
am
b
d
a
s
ettin
g
s
,
an
d
(
b
)
co
r
r
e
s
p
o
n
d
in
g
m
o
v
i
n
g
-
av
e
r
ag
e
c
u
r
v
es,
h
ig
h
lig
h
tin
g
h
o
w
tu
n
in
g
L
am
b
d
a
in
f
lu
en
ce
s
co
n
v
er
g
en
ce
an
d
lan
d
m
ar
k
l
o
c
aliza
tio
n
ac
cu
r
ac
y
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8
7
0
8
E
n
h
a
n
cin
g
fa
cia
l la
n
d
ma
r
k
d
etec
tio
n
w
ith
C
o
n
tr
o
lN
et
-
b
a
s
e
d
d
a
ta
…
(
K
r
ita
p
h
a
t S
o
n
g
s
r
i
-
in
)
4913
I
n
s
u
m
m
ar
y
,
t
h
e
ex
p
er
im
e
n
ta
l
r
esu
lts
h
ig
h
lig
h
t
th
e
ef
f
ec
tiv
en
ess
o
f
u
s
in
g
C
o
n
tr
o
lNet
-
a
u
g
m
en
ted
d
ata
an
d
t
h
e
im
p
o
r
tan
ce
o
f
tu
n
in
g
L
am
b
d
a
t
o
ac
h
iev
e
o
p
ti
m
al
p
er
f
o
r
m
an
ce
in
f
ac
ial
la
n
d
m
ar
k
d
etec
tio
n
.
T
h
e
an
aly
s
is
u
n
d
er
s
co
r
es
th
at
th
e
in
teg
r
atio
n
o
f
ca
r
ef
u
lly
ch
o
s
en
s
y
n
th
etic
d
ata
r
atio
s
,
alo
n
g
with
an
o
p
tim
al
L
am
b
d
a,
ca
n
e
n
h
an
ce
m
o
d
el
r
o
b
u
s
tn
ess
an
d
p
r
ec
is
io
n
in
la
n
d
m
ar
k
l
o
ca
lizatio
n
.
Ad
d
itio
n
ally
,
b
alan
cin
g
th
e
am
o
u
n
t
o
f
s
y
n
th
etic
an
d
r
ea
l
d
ata
en
s
u
r
es
d
iv
er
s
e
tr
ain
in
g
s
am
p
les
with
o
u
t
in
tr
o
d
u
cin
g
ex
ce
s
s
iv
e
n
o
is
e,
f
u
r
th
er
s
tab
ilizin
g
m
o
d
el
c
o
n
v
er
g
en
ce
.
4.
CO
NCLU
SI
O
N
T
h
is
s
tu
d
y
h
ig
h
lig
h
ts
th
e
ef
f
ec
tiv
en
ess
o
f
C
o
n
tr
o
lNet
-
b
ased
d
ata
au
g
m
en
tatio
n
in
en
h
an
cin
g
th
e
ac
cu
r
ac
y
an
d
r
o
b
u
s
tn
ess
o
f
f
ac
ial
lan
d
m
ar
k
d
etec
tio
n
.
B
y
i
n
teg
r
atin
g
C
o
n
tr
o
lNet
-
g
en
er
at
ed
s
y
n
th
etic
im
ag
es
with
r
ea
l
d
ata
f
r
o
m
th
e
3
0
0
W
d
ataset,
th
e
p
r
o
p
o
s
ed
ap
p
r
o
ac
h
a
d
d
r
ess
es
cr
itical
ch
allen
g
es
in
lan
d
m
ar
k
d
etec
tio
n
,
in
cl
u
d
in
g
v
a
r
iatio
n
s
in
lig
h
tin
g
,
p
o
s
e,
an
d
f
ac
ial
e
x
p
r
ess
io
n
s
.
T
h
e
ex
p
er
im
en
tal
r
esu
lts
d
em
o
n
s
tr
ate
th
at
au
g
m
en
ti
n
g
tr
ain
i
n
g
d
at
asets
with
s
y
n
th
etic
d
ata
s
ig
n
if
ican
tly
r
ed
u
ce
s
th
e
I
NM
E
,
th
er
eb
y
im
p
r
o
v
in
g
lan
d
m
ar
k
l
o
ca
lizatio
n
ac
cu
r
ac
y
.
F
u
r
t
h
e
r
m
o
r
e
,
t
h
e
f
i
n
d
i
n
g
s
e
m
p
h
a
s
i
z
e
t
h
e
i
m
p
o
r
t
a
n
c
e
o
f
o
p
t
im
i
z
i
n
g
t
h
e
r
at
i
o
o
f
s
y
n
t
h
e
t
i
c
to
r
e
a
l
d
a
t
a
a
n
d
f
i
n
e
-
t
u
n
i
n
g
m
o
d
e
l
p
a
r
a
m
ete
r
s
,
s
u
c
h
a
s
L
a
m
b
d
a
,
t
o
a
c
h
i
e
v
e
m
a
x
i
m
u
m
p
e
r
f
o
r
m
a
n
c
e
g
a
i
n
s
.
C
a
r
e
f
u
l
s
e
le
c
t
i
o
n
o
f
s
y
n
t
h
e
t
ic
-
to
-
r
e
a
l
d
a
t
a
p
r
o
p
o
r
t
i
o
n
s
e
n
s
u
r
e
s
t
h
a
t
t
h
e
m
o
d
e
l
l
ea
r
n
s
f
r
o
m
d
i
v
e
r
s
e
c
o
n
d
i
ti
o
n
s
wi
t
h
o
u
t
b
ei
n
g
o
v
e
r
w
h
e
l
m
e
d
b
y
a
r
t
i
f
i
ci
a
l
s
a
m
p
l
e
s
.
I
n
a
d
d
i
ti
o
n
,
a
d
j
u
s
t
i
n
g
L
am
b
d
a
a
l
l
o
ws
f
o
r
c
o
n
t
r
o
l
l
i
n
g
t
h
e
t
r
a
d
e
-
o
f
f
b
e
t
w
e
en
r
e
c
o
n
s
t
r
u
c
t
i
o
n
a
cc
u
r
a
c
y
a
n
d
r
e
g
u
l
a
r
i
z
a
t
i
o
n
,
w
h
i
c
h
u
l
t
i
m
at
e
l
y
h
e
l
p
s
s
t
a
b
i
li
z
e
t
r
ai
n
i
n
g
a
n
d
p
r
e
v
e
n
t
s
o
v
e
r
f
i
t
ti
n
g
.
T
h
is
m
eth
o
d
o
lo
g
y
h
o
ld
s
co
n
s
i
d
er
ab
le
p
r
o
m
is
e
f
o
r
b
r
o
a
d
er
a
p
p
licatio
n
s
in
co
m
p
u
ter
v
is
io
n
task
s
th
at
r
eq
u
ir
e
p
r
ec
is
e
f
ea
tu
r
e
lo
c
aliza
tio
n
.
I
n
p
ar
ticu
la
r
,
f
ield
s
s
u
ch
as
f
ac
ial
ex
p
r
ess
io
n
r
ec
o
g
n
itio
n
b
en
ef
it
f
r
o
m
r
eliab
le
lan
d
m
ar
k
p
o
s
itio
n
in
g
,
an
d
im
p
r
o
v
ed
3
D
f
ac
ial
m
o
d
elin
g
d
ep
e
n
d
s
o
n
ac
c
u
r
ate
f
ea
tu
r
e
co
r
r
esp
o
n
d
en
ce
.
Fu
tu
r
e
r
esear
ch
s
h
o
u
ld
f
o
c
u
s
o
n
r
ef
i
n
in
g
s
y
n
t
h
etic
d
ata
g
en
er
atio
n
tech
n
iq
u
es,
ex
p
lo
r
i
n
g
m
o
r
e
ad
v
a
n
ce
d
g
en
er
ativ
e
m
o
d
els,
an
d
e
x
ten
d
in
g
th
is
ap
p
r
o
ac
h
t
o
o
th
er
ar
ea
s
o
f
f
ac
ial
an
aly
s
is
to
f
u
r
th
er
v
alid
ate
its
g
en
er
aliza
b
ilit
y
.
ACK
NO
WL
E
DG
M
E
N
T
S
T
h
e
au
th
o
r
s
g
r
atef
u
lly
ac
k
n
o
wled
g
e
th
e
u
s
e
o
f
s
er
v
ice
a
n
d
f
ac
ilit
ies
o
f
th
e
Facu
lty
o
f
S
cien
ce
an
d
T
ec
h
n
o
lo
g
y
,
Nak
h
o
n
Si T
h
am
m
ar
at
R
ajab
h
at
Un
iv
er
s
ity
.
F
UNDING
I
NF
O
R
M
A
T
I
O
N
T
h
is
s
tu
d
y
r
ec
eiv
es
f
u
n
d
in
g
f
r
o
m
th
e
Natio
n
al
Scien
ce
an
d
T
ec
h
n
o
lo
g
y
Dev
elo
p
m
e
n
t
Ag
en
cy
,
T
h
ailan
d
,
u
n
d
er
th
e
Pro
to
ty
p
e
R
esear
ch
Gr
an
t Sch
em
e
J
R
A
-
CO
-
2565
-
1
7
7
9
2
-
T
H.
RE
F
E
R
E
NC
E
S
[
1
]
M
.
B
i
l
a
l
,
S
.
R
a
z
z
a
q
,
N
.
B
h
o
w
m
i
k
e
,
A
.
F
a
r
o
o
q
,
M
.
Z
a
h
i
d
,
a
n
d
S
.
S
h
o
a
i
b
,
“
F
a
c
i
a
l
r
e
c
o
g
n
i
t
i
o
n
u
si
n
g
h
i
d
d
e
n
M
a
r
k
o
v
m
o
d
e
l
a
n
d
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
u
r
a
l
n
e
t
w
o
r
k
,
”
AI
,
v
o
l
.
5
,
n
o
.
3
,
p
p
.
1
6
3
3
–
1
6
4
7
,
S
e
p
.
2
0
2
4
,
d
o
i
:
1
0
.
3
3
9
0
/
a
i
5
0
3
0
0
7
9
.
[
2
]
M
.
R
a
n
e
e
t
a
l
.
,
“
F
a
c
e
r
e
c
o
g
n
i
t
i
o
n
u
s
i
n
g
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
u
r
a
l
n
e
t
w
o
r
k
(
C
N
N
)
,
”
i
n
S
m
a
rt
T
re
n
d
s
i
n
C
o
m
p
u
t
i
n
g
a
n
d
C
o
m
m
u
n
i
c
a
t
i
o
n
s
.
S
M
ART
2
0
2
3
,
T.
S
e
n
j
y
u
,
C
.
S
o
–
I
n
,
a
n
d
A
.
J
o
sh
i
,
E
d
s
.
,
Le
c
t
u
r
e
N
o
t
e
s
i
n
N
e
t
w
o
r
k
s
a
n
d
S
y
st
e
ms,
v
o
l
.
6
4
5
.
S
i
n
g
a
p
o
r
e
:
S
p
r
i
n
g
e
r
,
2
0
2
3
,
p
p
.
2
0
3
–
2
1
4
,
d
o
i
:
1
0
.
1
0
0
7
/
9
7
8
-
9
8
1
-
99
-
0
7
6
9
-
4
_
2
0
.
[
3
]
T.
K
u
a
r
k
a
m
p
h
u
n
a
n
d
C
.
R
a
t
a
n
a
v
i
l
i
sag
u
l
,
“
F
a
c
e
r
e
c
o
g
n
i
t
i
o
n
u
si
n
g
s
k
i
n
c
o
l
o
r
se
g
me
n
t
a
n
d
m
o
d
i
f
i
e
d
b
i
n
a
r
y
p
a
r
t
i
c
l
e
sw
a
r
m
o
p
t
i
m
i
z
a
t
i
o
n
,
”
I
C
S
E
C
2
0
2
2
-
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
a
n
d
E
n
g
i
n
e
e
ri
n
g
C
o
n
f
e
r
e
n
c
e
2
0
2
2
,
p
p
.
6
6
–
7
1
,
2
0
2
2
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
S
EC
5
6
3
3
7
.
2
0
2
2
.
1
0
0
4
9
3
5
4
.
[
4
]
P
.
A
.
Ja
v
i
e
r
O
r
l
a
n
d
o
,
J
.
M
.
R
o
b
i
n
s
o
n
,
a
n
d
J.
E.
M
.
B
a
q
u
e
r
o
,
“
C
o
m
p
a
r
i
s
o
n
o
f
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
u
r
a
l
n
e
t
w
o
r
k
mo
d
e
l
s
f
o
r
u
ser’
s
f
a
c
i
a
l
r
e
c
o
g
n
i
t
i
o
n
,
”
I
n
t
e
rn
a
t
i
o
n
a
l
J
o
u
rn
a
l
o
f
E
l
e
c
t
ri
c
a
l
a
n
d
C
o
m
p
u
t
e
r
E
n
g
i
n
e
e
r
i
n
g
,
v
o
l
.
1
4
,
n
o
.
1
,
p
p
.
1
9
2
–
1
9
8
,
2
0
2
4
,
d
o
i
:
1
0
.
1
1
5
9
1
/
i
j
e
c
e
.
v
1
4
i
1
.
p
p
1
9
2
-
1
9
8
.
[
5
]
A
.
H
a
l
i
m
e
t
a
l
.
,
“
F
a
c
i
a
l
e
x
p
r
e
ss
i
o
n
s
a
n
a
l
y
si
s
t
o
e
v
a
l
u
a
t
e
t
h
e
l
e
v
e
l
o
f
st
u
d
e
n
t
s’
u
n
d
e
r
s
t
a
n
d
i
n
g
,
”
i
n
2
0
2
3
I
n
t
e
l
l
i
g
e
n
t
Me
t
h
o
d
s
,
S
y
s
t
e
m
s,
a
n
d
A
p
p
l
i
c
a
t
i
o
n
s (I
M
S
A)
,
Ju
l
.
2
0
2
3
,
p
p
.
4
2
4
–
429
,
d
o
i
:
1
0
.
1
1
0
9
/
I
M
S
A
5
8
5
4
2
.
2
0
2
3
.
1
0
2
1
7
4
8
9
.
[
6
]
D
.
Zh
a
o
,
J
.
W
a
n
g
,
H
.
L
i
,
a
n
d
D
.
W
a
n
g
,
“
L
a
n
d
m
a
r
k
-
b
a
s
e
d
a
d
a
p
t
i
v
e
g
r
a
p
h
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
t
w
o
r
k
f
o
r
f
a
c
i
a
l
e
x
p
r
e
ss
i
o
n
r
e
c
o
g
n
i
t
i
o
n
,
”
I
EEE
Ac
c
e
ss
,
v
o
l
.
1
2
,
p
p
.
1
3
6
0
8
8
–
1
3
6
1
0
2
,
2
0
2
4
,
d
o
i
:
1
0
.
1
1
0
9
/
A
C
C
ESS
.
2
0
2
4
.
3
4
6
3
1
7
6
.
[
7
]
K
a
v
i
t
a
a
n
d
R
.
S
.
C
h
h
i
l
l
a
r
,
“
P
e
r
f
o
r
m
a
n
c
e
a
n
a
l
y
s
i
s o
f
d
e
e
p
u
n
i
f
i
e
d
m
o
d
e
l
f
o
r
f
a
c
i
a
l
e
x
p
r
e
ssi
o
n
r
e
c
o
g
n
i
t
i
o
n
u
si
n
g
c
o
n
v
o
l
u
t
i
o
n
n
e
u
r
a
l
n
e
t
w
o
r
k
,
”
I
n
t
e
rn
a
t
i
o
n
a
l
J
o
u
r
n
a
l
o
f
El
e
c
t
r
i
c
a
l
a
n
d
C
o
m
p
u
t
e
r
En
g
i
n
e
e
ri
n
g
,
v
o
l
.
1
4
,
n
o
.
4
,
p
p
.
4
0
4
6
–
4
0
5
4
,
2
0
2
4
,
d
o
i
:
1
0
.
1
1
5
9
1
/
i
j
e
c
e
.
v
1
4
i
4
.
p
p
4
0
4
6
-
4
0
5
4
.
[
8
]
R
.
B
h
a
r
a
d
w
a
j
,
R
.
B
o
r
se
,
P
.
A
h
i
r
e
,
P
.
A
l
d
h
a
r
,
a
n
d
A
.
H
o
s
h
i
n
g
,
“
D
y
n
a
mi
c
3
D
f
a
c
i
a
l
e
x
p
r
e
ss
i
o
n
r
e
c
o
n
st
r
u
c
t
i
o
n
f
r
o
m
i
ma
g
e
s,”
i
n
Pro
c
e
e
d
i
n
g
s
-
2
0
2
4
5
t
h
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
I
m
a
g
e
Pr
o
c
e
ssi
n
g
a
n
d
C
a
p
su
l
e
N
e
t
w
o
r
k
s,
I
C
I
P
C
N
2
0
2
4
,
2
0
2
4
,
p
p
.
6
6
–
72
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
I
P
C
N
6
3
8
2
2
.
2
0
2
4
.
0
0
0
2
0
.
[
9
]
M
.
W
a
n
g
a
n
d
H
.
W
a
n
g
,
“
F
a
c
i
a
l
p
h
o
t
o
-
g
u
i
d
e
d
h
e
a
d
a
n
a
t
o
m
y
m
o
d
e
l
i
n
g
b
a
se
d
o
n
d
e
e
p
l
e
a
r
n
i
n
g
a
n
d
2
D
/
3
D
sh
a
p
e
p
r
i
o
r
m
o
d
e
l
r
e
g
i
s
t
r
a
t
i
o
n
,
”
S
m
a
rt
I
n
n
o
v
a
t
i
o
n
,
S
y
s
t
e
m
s a
n
d
T
e
c
h
n
o
l
o
g
i
e
s
,
v
o
l
.
3
7
4
,
p
p
.
2
4
7
–
2
5
7
,
2
0
2
4
,
d
o
i
:
1
0
.
1
0
0
7
/
9
7
8
-
9
8
1
-
99
-
7
0
1
1
-
7
_
2
0
.
[
1
0
]
J.
L
i
n
g
,
Z.
W
a
n
g
,
M
.
L
u
,
Q
.
W
a
n
g
,
C
.
Q
i
a
n
,
a
n
d
F
.
X
u
,
“
S
e
ma
n
t
i
c
a
l
l
y
d
i
se
n
t
a
n
g
l
e
d
v
a
r
i
a
t
i
o
n
a
l
a
u
t
o
e
n
c
o
d
e
r
f
o
r
m
o
d
e
l
i
n
g
3
D
f
a
c
i
a
l
d
e
t
a
i
l
s
,
”
I
EEE
T
ra
n
s
a
c
t
i
o
n
s
o
n
V
i
su
a
l
i
z
a
t
i
o
n
a
n
d
C
o
m
p
u
t
e
r
G
r
a
p
h
i
c
s
,
v
o
l
.
2
9
,
n
o
.
8
,
p
p
.
3
6
3
0
–
3
6
4
1
,
2
0
2
3
,
d
o
i
:
1
0
.
1
1
0
9
/
TV
C
G
.
2
0
2
2
.
3
1
6
6
6
6
6
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
15
,
No
.
5
,
Octo
b
e
r
20
25
:
4
9
0
7
-
4
9
1
5
4914
[
1
1
]
A
.
A
c
q
u
i
st
i
,
R
.
G
r
o
ss,
a
n
d
F
.
S
t
u
t
z
m
a
n
,
“
F
a
c
e
r
e
c
o
g
n
i
t
i
o
n
a
n
d
p
r
i
v
a
c
y
i
n
t
h
e
a
g
e
o
f
a
u
g
m
e
n
t
e
d
r
e
a
l
i
t
y
,
”
J
o
u
r
n
a
l
o
f
Pr
i
v
a
c
y
a
n
d
C
o
n
f
i
d
e
n
t
i
a
l
i
t
y
,
v
o
l
.
6
,
n
o
.
2
,
2
0
1
4
,
d
o
i
:
1
0
.
2
9
0
1
2
/
j
p
c
.
v
6
i
2
.
6
3
8
.
[
1
2
]
J.
D
e
n
g
,
G
.
Tr
i
g
e
o
r
g
i
s,
Y
.
Z
h
o
u
,
a
n
d
S
.
Z
a
f
e
i
r
i
o
u
,
“
J
o
i
n
t
mu
l
t
i
-
v
i
e
w
f
a
c
e
a
l
i
g
n
me
n
t
i
n
t
h
e
w
i
l
d
,
”
I
EEE
T
r
a
n
sa
c
t
i
o
n
s
o
n
I
m
a
g
e
Pro
c
e
ssi
n
g
,
v
o
l
.
2
8
,
n
o
.
7
,
p
p
.
3
6
3
6
–
3
6
4
8
,
2
0
1
9
,
d
o
i
:
1
0
.
1
1
0
9
/
TI
P
.
2
0
1
9
.
2
8
9
9
2
6
7
.
[
1
3
]
T.
F
.
C
o
o
t
e
s
,
C
.
J.
Ta
y
l
o
r
,
D
.
H
.
C
o
o
p
e
r
,
a
n
d
J.
G
r
a
h
a
m
,
“
A
c
t
i
v
e
s
h
a
p
e
mo
d
e
l
s
—
t
h
e
i
r
t
r
a
i
n
i
n
g
a
n
d
a
p
p
l
i
c
a
t
i
o
n
,
”
C
o
m
p
u
t
e
r
Vi
s
i
o
n
a
n
d
I
m
a
g
e
U
n
d
e
rst
a
n
d
i
n
g
,
v
o
l
.
6
1
,
n
o
.
1
,
p
p
.
3
8
–
5
9
,
Ja
n
.
1
9
9
5
,
d
o
i
:
1
0
.
1
0
0
6
/
c
v
i
u
.
1
9
9
5
.
1
0
0
4
.
[
1
4
]
T.
F
.
C
o
o
t
e
s
,
G
.
J
.
E
d
w
a
r
d
s,
a
n
d
C
.
J.
Ta
y
l
o
r
,
“
A
c
t
i
v
e
a
p
p
e
a
r
a
n
c
e
m
o
d
e
l
s,
”
I
EEE
T
r
a
n
sa
c
t
i
o
n
s
o
n
P
a
t
t
e
r
n
An
a
l
y
si
s
a
n
d
Ma
c
h
i
n
e
I
n
t
e
l
l
i
g
e
n
c
e
,
v
o
l
.
2
3
,
n
o
.
6
,
p
p
.
6
8
1
–
6
8
5
,
J
u
n
.
2
0
0
1
,
d
o
i
:
1
0
.
1
1
0
9
/
3
4
.
9
2
7
4
6
7
.
[
1
5
]
X
i
a
n
g
x
i
n
Z
h
u
a
n
d
D
.
R
a
m
a
n
a
n
,
“
F
a
c
e
d
e
t
e
c
t
i
o
n
,
p
o
s
e
e
s
t
i
m
a
t
i
o
n
,
a
n
d
l
a
n
d
m
a
r
k
l
o
c
a
l
i
z
a
t
i
o
n
i
n
t
h
e
w
i
l
d
,
”
i
n
2
0
1
2
I
EE
E
C
o
n
f
e
re
n
c
e
o
n
C
o
m
p
u
t
e
r
Vi
s
i
o
n
a
n
d
Pa
t
t
e
r
n
Re
c
o
g
n
i
t
i
o
n
,
J
u
n
.
2
0
1
2
,
p
p
.
2
8
7
9
–
2
8
8
6
,
d
o
i
:
1
0
.
1
1
0
9
/
C
V
P
R
.
2
0
1
2
.
6
2
4
8
0
1
4
.
[
1
6
]
D
.
Le
e
,
H
.
P
a
r
k
,
a
n
d
C
.
D
.
Y
o
o
,
“
F
a
c
e
a
l
i
g
n
m
e
n
t
u
si
n
g
c
a
sca
d
e
G
a
u
s
si
a
n
p
r
o
c
e
ss
r
e
g
r
e
ssi
o
n
t
r
e
e
s,”
i
n
2
0
1
5
I
EE
E
C
o
n
f
e
r
e
n
c
e
o
n
C
o
m
p
u
t
e
r
Vi
s
i
o
n
a
n
d
P
a
t
t
e
r
n
Re
c
o
g
n
i
t
i
o
n
(
C
VP
R)
,
J
u
n
.
2
0
1
5
,
p
p
.
4
2
0
4
–
4
2
1
2
,
d
o
i
:
1
0
.
1
1
0
9
/
C
V
P
R
.
2
0
1
5
.
7
2
9
9
0
4
8
.
[
1
7
]
X
u
d
o
n
g
C
a
o
,
Y
i
c
h
e
n
W
e
i
,
F
a
n
g
W
e
n
,
a
n
d
J
i
a
n
S
u
n
,
“
F
a
c
e
a
l
i
g
n
me
n
t
b
y
e
x
p
l
i
c
i
t
sh
a
p
e
r
e
g
r
e
ssi
o
n
,
”
i
n
2
0
1
2
I
E
EE
C
o
n
f
e
r
e
n
c
e
o
n
C
o
m
p
u
t
e
r
Vi
s
i
o
n
a
n
d
P
a
t
t
e
r
n
Re
c
o
g
n
i
t
i
o
n
,
J
u
n
.
2
0
1
2
,
p
p
.
2
8
8
7
–
2
8
9
4
,
d
o
i
:
1
0
.
1
1
0
9
/
C
V
P
R
.
2
0
1
2
.
6
2
4
8
0
1
5
.
[
1
8
]
X
.
X
i
o
n
g
a
n
d
F
.
D
e
l
a
T
o
r
r
e
,
“
S
u
p
e
r
v
i
s
e
d
d
e
s
c
e
n
t
me
t
h
o
d
a
n
d
i
t
s
a
p
p
l
i
c
a
t
i
o
n
s
t
o
f
a
c
e
a
l
i
g
n
me
n
t
,
”
i
n
2
0
1
3
I
E
E
E
C
o
n
f
e
re
n
c
e
o
n
C
o
m
p
u
t
e
r
Vi
s
i
o
n
a
n
d
P
a
t
t
e
r
n
Re
c
o
g
n
i
t
i
o
n
,
J
u
n
.
2
0
1
3
,
p
p
.
5
3
2
–
5
3
9
,
d
o
i
:
1
0
.
1
1
0
9
/
C
V
P
R
.
2
0
1
3
.
7
5
.
[
1
9
]
M
.
D
a
n
t
o
n
e
,
J.
G
a
l
l
,
C
.
L
e
i
s
t
n
e
r
,
a
n
d
L.
V
a
n
G
o
o
l
,
“
H
u
m
a
n
p
o
s
e
e
s
t
i
mat
i
o
n
u
si
n
g
b
o
d
y
p
a
r
t
s d
e
p
e
n
d
e
n
t
j
o
i
n
t
r
e
g
r
e
ss
o
r
s,”
i
n
2
0
1
3
I
EEE
C
o
n
f
e
re
n
c
e
o
n
C
o
m
p
u
t
e
r
Vi
s
i
o
n
a
n
d
P
a
t
t
e
r
n
R
e
c
o
g
n
i
t
i
o
n
,
Ju
n
.
2
0
1
3
,
p
p
.
3
0
4
1
–
3
0
4
8
,
d
o
i
:
1
0
.
1
1
0
9
/
C
V
P
R
.
2
0
1
3
.
3
9
1
.
[
2
0
]
V
.
K
a
z
e
m
i
a
n
d
J.
S
u
l
l
i
v
a
n
,
“
O
n
e
mi
l
l
i
sec
o
n
d
f
a
c
e
a
l
i
g
n
me
n
t
w
i
t
h
a
n
e
n
se
mb
l
e
o
f
r
e
g
r
e
ssi
o
n
t
r
e
e
s
,
”
i
n
2
0
1
4
I
E
EE
C
o
n
f
e
re
n
c
e
o
n
C
o
m
p
u
t
e
r
Vi
s
i
o
n
a
n
d
P
a
t
t
e
r
n
Re
c
o
g
n
i
t
i
o
n
,
J
u
n
.
2
0
1
4
,
p
p
.
1
8
6
7
–
1
8
7
4
,
d
o
i
:
1
0
.
1
1
0
9
/
C
V
P
R
.
2
0
1
4
.
2
4
1
.
[
2
1
]
B
.
Ji
n
,
X
.
X
u
,
a
n
d
Y
.
Z
h
a
n
g
,
“
T
h
e
r
m
a
l
c
o
a
l
f
u
t
u
r
e
s
t
r
a
d
i
n
g
v
o
l
u
m
e
p
r
e
d
i
c
t
i
o
n
s
t
h
r
o
u
g
h
t
h
e
n
e
u
r
a
l
n
e
t
w
o
r
k
,
”
J
o
u
r
n
a
l
o
f
M
o
d
e
l
l
i
n
g
i
n
M
a
n
a
g
e
m
e
n
t
,
v
o
l
.
2
0
,
n
o
.
2
,
p
p
.
5
8
5
–
6
1
9
,
F
e
b
.
2
0
2
5
,
d
o
i
:
1
0
.
1
1
0
8
/
J
M
2
-
09
-
2
0
2
3
-
0
2
0
7
.
[
2
2
]
S
.
I
t
t
i
s
o
p
o
n
p
i
sa
n
,
C
.
K
a
i
p
a
n
,
S
.
R
u
a
n
g
-
o
n
,
R
.
T
h
a
i
p
h
a
n
,
a
n
d
K
.
S
o
n
g
sr
i
-
i
n
,
“
P
u
sh
i
n
g
t
h
e
a
c
c
u
r
a
c
y
o
f
Th
a
i
f
o
o
d
i
ma
g
e
c
l
a
ss
i
f
i
c
a
t
i
o
n
w
i
t
h
t
r
a
n
sf
e
r
l
e
a
r
n
i
n
g
,
”
En
g
i
n
e
e
ri
n
g
J
o
u
rn
a
l
,
v
o
l
.
2
6
,
n
o
.
1
0
,
p
p
.
5
7
–
7
1
,
O
c
t
.
2
0
2
2
,
d
o
i
:
1
0
.
4
1
8
6
/
e
j
.
2
0
2
2
.
2
6
.
1
0
.
5
7
.
[
2
3
]
B
.
Ji
n
,
X
.
X
u
,
a
n
d
Y
.
Z
h
a
n
g
,
“
P
e
a
n
u
t
o
i
l
p
r
i
c
e
c
h
a
n
g
e
f
o
r
e
c
a
s
t
s
t
h
r
o
u
g
h
t
h
e
n
e
u
r
a
l
n
e
t
w
o
r
k
,
”
f
o
r
e
si
g
h
t
,
v
o
l
.
2
7
,
n
o
.
3
,
p
p
.
5
9
5
–
6
1
2
,
A
p
r
.
2
0
2
5
,
d
o
i
:
1
0
.
1
1
0
8
/
F
S
-
01
-
2
0
2
3
-
0
0
1
6
.
[
2
4
]
K
.
S
o
n
g
sr
i
-
i
n
,
M
.
R
a
t
t
a
p
h
u
n
,
S
.
K
a
e
w
c
h
a
d
a
,
a
n
d
S
.
R
u
a
n
g
-
o
n
,
“
D
u
a
l
F
a
c
e
N
e
t
:
a
u
g
me
n
t
a
t
i
o
n
c
o
n
si
s
t
e
n
c
y
f
o
r
o
p
t
i
m
a
l
f
a
c
i
a
l
l
a
n
d
m
a
r
k
d
e
t
e
c
t
i
o
n
a
n
d
f
a
c
e
m
a
sk
c
l
a
ssi
f
i
c
a
t
i
o
n
,
”
I
A
ES
I
n
t
e
r
n
a
t
i
o
n
a
l
J
o
u
rn
a
l
o
f
Ar
t
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
(
I
J
-
AI
)
,
v
o
l
.
1
3
,
n
o
.
3
,
p
p
.
3
2
2
8
–
3
2
3
9
,
S
e
p
.
2
0
2
4
,
d
o
i
:
1
0
.
1
1
5
9
1
/
i
j
a
i
.
v
1
3
.
i
3
.
p
p
3
2
2
8
-
3
2
3
9
.
[
2
5
]
A
.
B
u
l
a
t
a
n
d
G
.
T
z
i
m
i
r
o
p
o
u
l
o
s
,
“
H
o
w
f
a
r
a
r
e
w
e
f
r
o
m
so
l
v
i
n
g
t
h
e
2
D
&
3
D
f
a
c
e
a
l
i
g
n
me
n
t
p
r
o
b
l
e
m?
(
a
n
d
a
d
a
t
a
s
e
t
o
f
2
3
0
,
0
0
0
3
D
f
a
c
i
a
l
l
a
n
d
mar
k
s)
,
”
i
n
2
0
1
7
I
E
EE
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
r
e
n
c
e
o
n
C
o
m
p
u
t
e
r
V
i
si
o
n
(
I
C
C
V)
,
O
c
t
.
2
0
1
7
,
p
p
.
1
0
2
1
–
1
0
3
0
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
C
V
.
2
0
1
7
.
1
1
6
.
[
2
6
]
Y
.
S
u
n
,
X
.
W
a
n
g
,
a
n
d
X
.
T
a
n
g
,
“
D
e
e
p
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
t
w
o
r
k
c
a
sc
a
d
e
f
o
r
f
a
c
i
a
l
p
o
i
n
t
d
e
t
e
c
t
i
o
n
,
”
i
n
2
0
1
3
I
E
E
E
C
o
n
f
e
re
n
c
e
o
n
C
o
m
p
u
t
e
r
Vi
s
i
o
n
a
n
d
P
a
t
t
e
r
n
Re
c
o
g
n
i
t
i
o
n
,
J
u
n
.
2
0
1
3
,
p
p
.
3
4
7
6
–
3
4
8
3
,
d
o
i
:
1
0
.
1
1
0
9
/
C
V
P
R
.
2
0
1
3
.
4
4
6
.
[
2
7
]
Z.
Z
h
a
n
g
,
P
.
L
u
o
,
C
.
C
.
Lo
y
,
a
n
d
X
.
Ta
n
g
,
“
F
a
c
i
a
l
l
a
n
d
mar
k
d
e
t
e
c
t
i
o
n
b
y
d
e
e
p
mu
l
t
i
-
t
a
s
k
l
e
a
r
n
i
n
g
,
”
i
n
C
o
m
p
u
t
e
r
V
i
si
o
n
–
E
C
C
V
2
0
1
4
,
D
.
F
l
e
e
t
,
T.
P
a
j
d
l
a
,
B
.
S
c
h
i
e
l
e
,
a
n
d
T
.
Tu
y
t
e
l
a
a
r
s,
E
d
s.
,
Le
c
t
u
r
e
N
o
t
e
s
i
n
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
.
C
h
a
m:
S
p
r
i
n
g
e
r
,
2
0
1
4
,
p
p
.
9
4
–
1
0
8
,
d
o
i
:
1
0
.
1
0
0
7
/
9
7
8
-
3
-
3
1
9
-
1
0
5
9
9
-
4
_
7
.
[
2
8
]
A
.
J
o
u
r
a
b
l
o
o
a
n
d
X
.
L
i
u
,
“
L
a
r
g
e
-
p
o
s
e
f
a
c
e
a
l
i
g
n
m
e
n
t
v
i
a
C
N
N
-
b
a
s
e
d
d
e
n
s
e
3
D
m
o
d
e
l
f
i
t
t
i
n
g
,
”
2
0
1
6
I
E
E
E
C
o
n
f
e
r
e
n
c
e
o
n
C
o
m
p
u
t
e
r
V
i
s
i
o
n
a
n
d
P
a
t
t
e
r
n
R
e
c
o
g
n
i
t
i
o
n
(
C
V
P
R
)
,
L
a
s
V
e
g
a
s
,
N
V
,
U
S
A
,
2
0
1
6
,
p
p
.
4
1
8
8
-
4
1
9
6
,
d
o
i
:
1
0
.
1
1
0
9
/
C
V
P
R
.
2
0
1
6
.
4
5
4
.
[
2
9
]
X
.
Zh
u
,
Z.
Le
i
,
X
.
Li
u
,
H
.
S
h
i
,
a
n
d
S
.
Z.
Li
,
“
F
a
c
e
a
l
i
g
n
m
e
n
t
a
c
r
o
s
s
l
a
r
g
e
p
o
se
s:
A
3
D
s
o
l
u
t
i
o
n
,
”
i
n
2
0
1
6
I
E
EE
C
o
n
f
e
re
n
c
e
o
n
C
o
m
p
u
t
e
r
Vi
s
i
o
n
a
n
d
P
a
t
t
e
r
n
Re
c
o
g
n
i
t
i
o
n
(
C
VP
R)
,
J
u
n
.
2
0
1
6
,
p
p
.
1
4
6
–
1
5
5
,
d
o
i
:
1
0
.
1
1
0
9
/
C
V
P
R
.
2
0
1
6
.
2
3
.
[
3
0
]
Z.
-
H
.
F
e
n
g
,
J.
K
i
t
t
l
e
r
,
M
.
A
w
a
i
s,
P
.
H
u
b
e
r
,
a
n
d
X
.
-
J.
W
u
,
“
W
i
n
g
l
o
ss
f
o
r
r
o
b
u
st
f
a
c
i
a
l
l
a
n
d
mark
l
o
c
a
l
i
s
a
t
i
o
n
w
i
t
h
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
u
r
a
l
n
e
t
w
o
r
k
s,”
i
n
2
0
1
8
I
EEE
/
C
V
F
C
o
n
f
e
re
n
c
e
o
n
C
o
m
p
u
t
e
r
Vi
s
i
o
n
a
n
d
P
a
t
t
e
r
n
Re
c
o
g
n
i
t
i
o
n
,
J
u
n
.
2
0
1
8
,
p
p
.
2
2
3
5
–
2
2
4
5
,
d
o
i
:
1
0
.
1
1
0
9
/
C
V
P
R
.
2
0
1
8
.
0
0
2
3
8
.
[
3
1
]
W
.
W
u
,
C
.
Q
i
a
n
,
S
.
Y
a
n
g
,
Q
.
W
a
n
g
,
Y
.
C
a
i
,
a
n
d
Q
.
Z
h
o
u
,
“
L
o
o
k
a
t
b
o
u
n
d
a
r
y
:
A
b
o
u
n
d
a
r
y
-
a
w
a
r
e
f
a
c
e
a
l
i
g
n
m
e
n
t
a
l
g
o
r
i
t
h
m
,
”
i
n
2
0
1
8
I
E
E
E
/
C
V
F
C
o
n
f
e
r
e
n
c
e
o
n
C
o
m
p
u
t
e
r
V
i
s
i
o
n
a
n
d
P
a
t
t
e
r
n
R
e
c
o
g
n
i
t
i
o
n
,
J
u
n
.
2
0
1
8
,
p
p
.
2
1
2
9
–
2
1
3
8
,
doi
:
1
0
.
1
1
0
9
/
C
V
P
R
.
2
0
1
8
.
0
0
2
2
7
.
[
3
2
]
C
.
S
h
o
r
t
e
n
a
n
d
T.
M
.
K
h
o
s
h
g
o
f
t
a
a
r
,
“
A
su
r
v
e
y
o
n
i
ma
g
e
d
a
t
a
a
u
g
m
e
n
t
a
t
i
o
n
f
o
r
d
e
e
p
l
e
a
r
n
i
n
g
,
”
J
o
u
r
n
a
l
o
f
Bi
g
D
a
t
a
,
v
o
l
.
6
,
n
o
.
1
,
p
.
6
0
,
D
e
c
.
2
0
1
9
,
d
o
i
:
1
0
.
1
1
8
6
/
s
4
0
5
3
7
-
0
1
9
-
0
1
9
7
-
0.
[
3
3
]
J.
H
o
,
A
.
Ja
i
n
,
a
n
d
P
.
A
b
b
e
e
l
,
“
D
e
n
o
i
si
n
g
d
i
f
f
u
s
i
o
n
p
r
o
b
a
b
i
l
i
st
i
c
m
o
d
e
l
s
,
”
Ad
v
a
n
c
e
s
i
n
N
e
u
r
a
l
I
n
f
o
rm
a
t
i
o
n
Pro
c
e
ssi
n
g
S
y
s
t
e
m
s
,
v
o
l
.
2
0
2
0
-
D
e
c
e
m
b
e
r
,
2
0
2
0
.
[
3
4
]
D
.
R
y
u
a
n
d
J.
C
.
Y
e
,
“
P
y
r
a
mi
d
a
l
d
e
n
o
i
si
n
g
d
i
f
f
u
s
i
o
n
p
r
o
b
a
b
i
l
i
st
i
c
m
o
d
e
l
s,”
a
rXi
v
:
2
2
0
8
.
0
1
8
6
4
,
2
0
2
2
.
[
3
5
]
L.
Z
h
a
n
g
,
A
.
R
a
o
,
a
n
d
M
.
A
g
r
a
w
a
l
a
,
“
A
d
d
i
n
g
c
o
n
d
i
t
i
o
n
a
l
c
o
n
t
r
o
l
t
o
t
e
x
t
-
to
-
i
ma
g
e
d
i
f
f
u
s
i
o
n
mo
d
e
l
s,
”
i
n
2
0
2
3
I
EE
E/
C
VF
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
r
e
n
c
e
o
n
C
o
m
p
u
t
e
r V
i
s
i
o
n
(
I
C
C
V)
,
O
c
t
.
2
0
2
3
,
p
p
.
3
8
1
3
–
3
8
2
4
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
C
V
5
1
0
7
0
.
2
0
2
3
.
0
0
3
5
5
.
[
3
6
]
C
.
S
a
g
o
n
a
s,
G
.
T
z
i
mi
r
o
p
o
u
l
o
s,
S
.
Z
a
f
e
i
r
i
o
u
,
a
n
d
M
.
P
a
n
t
i
c
,
“
3
0
0
f
a
c
e
s
i
n
-
t
h
e
-
w
i
l
d
c
h
a
l
l
e
n
g
e
:
t
h
e
f
i
r
s
t
f
a
c
i
a
l
l
a
n
d
mark
l
o
c
a
l
i
z
a
t
i
o
n
c
h
a
l
l
e
n
g
e
,
”
i
n
Pr
o
c
e
e
d
i
n
g
s
o
f
t
h
e
I
EEE
I
n
t
e
rn
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
C
o
m
p
u
t
e
r
Vi
s
i
o
n
,
2
0
1
3
,
p
p
.
3
9
7
–
4
0
3
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
C
V
W
.
2
0
1
3
.
5
9
.
B
I
O
G
RAP
H
I
E
S O
F
AUTH
O
RS
K
r
i
t
a
p
h
a
t
S
o
n
g
s
r
i
-
in
f
i
n
i
s
h
e
d
M
E
n
g
a
n
d
P
h
.
D
.
i
n
c
o
m
p
u
t
i
n
g
f
r
o
m
I
m
p
e
r
i
a
l
C
o
l
l
e
g
e
L
o
n
d
o
n
i
n
2
0
1
1
a
n
d
2
0
2
0
,
r
e
s
p
e
c
t
i
v
e
l
y
.
C
u
r
r
e
n
t
l
y
,
h
e
i
s
a
n
a
s
s
i
s
t
a
n
t
p
r
o
f
e
s
s
o
r
i
n
t
h
e
D
e
p
a
r
tm
e
n
t
o
f
C
o
m
p
u
t
e
r
S
c
ie
n
c
e
a
t
N
a
k
h
o
n
S
i
T
h
a
m
m
a
r
a
t
R
a
j
a
b
h
a
t
U
n
i
v
e
r
s
i
t
y
,
T
h
a
i
l
a
n
d
.
H
i
s
r
e
s
e
a
r
c
h
i
n
t
e
re
s
ts
i
n
c
l
u
d
e
m
a
c
h
i
n
e
l
e
a
r
n
i
n
g
,
d
e
e
p
l
e
a
r
n
i
n
g
,
a
n
d
c
o
m
p
u
t
e
r
v
i
s
i
o
n
.
H
e
h
a
s
p
u
b
l
i
s
h
e
d
i
n
a
n
d
i
s
a
r
e
v
i
e
we
r
f
o
r
m
u
l
t
i
p
l
e
i
n
t
e
r
n
a
t
i
o
n
a
l
c
o
n
f
e
r
e
n
c
e
s
a
n
d
j
o
u
r
n
a
l
s
s
u
c
h
a
s
I
E
E
E
T
ra
n
s
a
c
t
i
o
n
s
o
n
I
m
a
g
e
P
r
o
c
e
ss
i
n
g
a
n
d
I
E
E
E
T
r
a
n
s
a
c
t
i
o
n
s
o
n
I
n
f
o
r
m
a
t
i
o
n
F
o
r
e
n
s
i
c
s
a
n
d
S
e
c
u
r
i
t
y
.
He
wa
s
a
r
e
c
i
p
i
e
n
t
o
f
t
h
e
R
o
y
a
l
T
h
a
i
G
o
v
e
rn
m
e
n
t
S
c
h
o
l
a
r
s
h
i
p
c
o
v
e
r
i
n
g
h
i
s
u
n
d
e
r
g
r
a
d
u
a
t
e
a
n
d
p
o
s
t
g
r
a
d
u
a
t
e
d
e
g
r
e
e
s
i
n
2
0
1
0
.
H
e
r
e
c
e
i
v
e
d
t
h
e
B
e
s
t
S
t
u
d
e
n
t
P
a
p
e
r
A
wa
r
d
s
a
t
t
h
e
I
E
E
E
1
3
t
h
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
r
e
n
c
e
f
o
r
A
u
t
o
m
a
t
ic
F
a
c
e
a
n
d
G
e
s
t
u
r
e
R
e
c
o
g
n
i
t
i
o
n
(
F
G
2
0
1
8
)
a
n
d
t
h
e
6
t
h
N
a
t
i
o
n
a
l
S
c
i
e
n
c
e
a
n
d
T
e
c
h
n
o
l
o
g
y
C
o
n
f
e
r
e
n
c
e
(
N
S
C
I
C
2
0
2
1
)
.
I
n
2
0
2
1
,
h
i
s
P
h
.
D
.
t
h
e
s
i
s
r
e
c
e
i
v
e
d
a
n
a
w
a
r
d
f
r
o
m
t
h
e
N
a
t
i
o
n
a
l
R
e
s
e
a
r
c
h
C
o
u
n
c
i
l
o
f
T
h
a
i
l
a
n
d
(
N
R
C
T
)
.
H
e
c
a
n
b
e
c
o
n
t
a
c
t
e
d
a
t
e
m
a
i
l
:
k
r
i
t
a
p
h
a
t
_
s
o
n
@
n
s
t
r
u
.
ac
.
th
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8
7
0
8
E
n
h
a
n
cin
g
fa
cia
l la
n
d
ma
r
k
d
etec
tio
n
w
ith
C
o
n
tr
o
lN
et
-
b
a
s
e
d
d
a
ta
…
(
K
r
ita
p
h
a
t S
o
n
g
s
r
i
-
in
)
4915
Mu
n
li
k
a
R
a
tta
p
h
u
n
re
c
e
iv
e
d
th
e
B
.
S
.
d
e
g
re
e
i
n
c
o
m
p
u
ter
sc
ie
n
c
e
fro
m
T
h
a
k
sin
Un
iv
e
rsity
,
S
o
n
g
k
h
la,
T
h
a
il
a
n
d
,
i
n
2
0
0
9
,
th
e
M
.
S
.
d
e
g
re
e
in
c
o
m
p
u
ter
sc
ien
c
e
fro
m
P
rin
c
e
o
f
S
o
n
g
k
la
U
n
iv
e
rsit
y
,
S
o
n
g
k
h
la,
T
h
a
il
a
n
d
,
in
2
0
1
1
,
a
n
d
th
e
P
h
.
D
.
d
e
g
re
e
in
c
o
m
p
u
ter
sc
ien
c
e
a
n
d
in
f
o
rm
a
ti
o
n
e
n
g
in
e
e
rin
g
fro
m
Na
ti
o
n
a
l
Ch
ia
y
i
U
n
iv
e
rsit
y
,
C
h
a
iy
i,
Taiwa
n
,
i
n
2
0
2
2
.
S
h
e
is
c
u
rre
n
tl
y
a
lec
tu
re
r
a
t
t
h
e
De
p
a
rtme
n
t
o
f
C
o
m
p
u
ter
S
c
ien
c
e
,
F
a
c
u
lt
y
o
f
S
c
ien
c
e
a
n
d
Tec
h
n
o
l
o
g
y
,
Na
k
h
o
n
S
i
Th
a
m
m
a
ra
t
Ra
jab
h
a
t
Un
iv
e
rsity
,
Na
k
h
o
n
S
i
Th
a
m
m
a
ra
t,
Th
a
il
a
n
d
.
He
r
c
u
rre
n
t
re
se
a
rc
h
in
tere
sts
in
c
lu
d
e
m
a
c
h
in
e
lea
rn
in
g
,
n
e
a
re
st
-
n
e
ig
h
b
o
r
se
a
rc
h
,
a
n
d
re
c
o
m
m
e
n
d
e
r
sy
ste
m
s
.
S
h
e
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
m
u
n
li
k
a
_
ra
t
@n
stru
.
ac
.
th
.
S
o
p
e
e
K
a
e
w
c
h
a
d
a
re
c
e
i
v
e
d
t
h
e
B
.
Sc
.
d
e
g
re
e
i
n
c
o
m
p
u
te
r
sc
ie
n
c
e
fr
o
m
Ra
ja
b
h
a
t
P
h
e
tc
h
a
b
u
ri
I
n
s
ti
t
u
te
,
T
h
a
i
la
n
d
,
i
n
1
9
9
7
t
h
e
M
.
S
.
d
e
g
re
e
i
n
m
a
n
a
g
e
m
e
n
t
o
f
i
n
fo
rm
a
ti
o
n
tec
h
n
o
l
o
g
y
f
r
o
m
W
a
la
il
a
k
U
n
i
v
e
rsi
t
y
,
T
h
a
il
a
n
d
,
i
n
2
0
0
3
,
a
n
d
t
h
e
P
h
.
D
.
d
e
g
re
e
i
n
c
re
a
t
iv
e
in
n
o
v
a
t
io
n
i
n
s
c
ie
n
c
e
a
n
d
t
e
c
h
n
o
lo
g
y
,
Na
k
h
o
n
S
i
T
h
a
m
m
a
ra
t
Ra
j
a
b
h
a
t
U
n
i
v
e
rsi
t
y
,
T
h
a
i
la
n
d
,
in
2
0
2
3
.
C
u
r
re
n
t
l
y
,
s
h
e
is
a
n
a
s
sis
ta
n
t
p
r
o
fe
ss
o
r
a
t
t
h
e
F
a
c
u
l
t
y
o
f
S
c
ie
n
c
e
a
n
d
Tec
h
n
o
l
o
g
y
,
Na
k
h
o
n
S
i
T
h
a
m
m
a
ra
t
Ra
ja
b
h
a
t
Un
i
v
e
rsi
t
y
,
T
h
a
i
la
n
d
.
S
h
e
c
a
n
b
e
c
o
n
tac
te
d
a
t
e
m
a
i
l
:
so
p
e
e
_
k
a
e
@
n
st
r
u
.
ac
.
th
.
S
u
n
isa
K
id
j
a
id
e
a
w
re
c
e
iv
e
d
th
e
B
.
Sc
.
d
e
g
re
e
in
c
o
m
p
u
ter
sc
ien
c
e
fro
m
Na
k
h
o
n
S
i
Th
a
m
m
a
ra
t
Ra
jab
h
a
t
Un
i
v
e
rsity
,
T
h
a
il
a
n
d
,
i
n
2
0
0
6
t
h
e
M
.
S
.
d
e
g
re
e
in
m
a
n
a
g
e
m
e
n
t
o
f
in
fo
rm
a
ti
o
n
tec
h
n
o
lo
g
y
f
ro
m
Wa
lailak
Un
i
v
e
rsity
,
T
h
a
il
a
n
d
,
i
n
2
0
1
0
,
a
n
d
th
e
P
h
.
D
.
d
e
g
re
e
in
in
fo
rm
a
ti
o
n
tec
h
n
o
l
o
g
y
,
S
ri
p
a
tu
m
Un
iv
e
rsity
,
T
h
a
il
a
n
d
,
in
2
0
2
0
.
Cu
rre
n
tl
y
,
sh
e
is
a
n
a
ss
istan
t
p
ro
fe
ss
o
r
a
t
t
h
e
F
a
c
u
lt
y
o
f
S
c
ien
c
e
a
n
d
Tec
h
n
o
lo
g
y
,
Na
k
h
o
n
S
i
Th
a
m
m
a
ra
t
Ra
jab
h
a
t
Un
iv
e
rsity
,
T
h
a
il
a
n
d
.
S
h
e
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
su
n
isa
_
k
i
d
@n
stru
.
ac
.
th
.
S
a
n
g
ju
n
Rua
n
g
-
O
n
re
c
e
iv
e
d
th
e
B
.
B
.
A
.
d
e
g
re
e
in
b
u
sin
e
ss
c
o
m
p
u
ter
fro
m
S
rip
a
tu
m
Un
i
v
e
rsity
,
Th
a
il
a
n
d
,
i
n
1
9
9
4
,
t
h
e
M
.
Sc
.
d
e
g
re
e
i
n
i
n
fo
rm
a
ti
o
n
tec
h
n
o
l
o
g
y
fr
o
m
S
rip
a
tu
m
Un
iv
e
rsit
y
,
T
h
a
il
a
n
d
,
i
n
2
0
0
3
.
C
u
rre
n
tl
y
,
sh
e
is
a
n
a
ss
istan
t
p
r
o
fe
ss
o
r
a
t
th
e
F
a
c
u
lt
y
o
f
S
c
ien
c
e
a
n
d
Tec
h
n
o
lo
g
y
,
Na
k
h
o
n
S
i
T
h
a
m
m
a
ra
t
Ra
jab
h
a
t
Un
i
v
e
rsity
,
T
h
a
il
a
n
d
.
S
h
e
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
sa
n
g
j
u
n
_
r
u
a
@n
stru
.
ac
.
th
.
Wichi
t
S
o
o
k
k
h
a
t
h
o
n
re
c
e
iv
e
d
t
h
e
B
.
Sc
.
d
e
g
re
e
in
c
o
m
p
u
ter
sc
ien
c
e
fro
m
Ra
jab
h
a
t
S
o
n
g
k
h
la
In
sti
tu
te,
T
h
a
il
a
n
d
,
i
n
1
9
9
6
,
t
h
e
M
.
Sc
.
d
e
g
re
e
in
in
f
o
rm
a
ti
o
n
tec
h
n
o
l
o
g
y
fro
m
Un
iv
e
rsiti
Uta
ra
M
a
lay
sia
,
in
2
0
0
4
,
a
n
d
P
h
.
D
.
d
e
g
re
e
i
n
q
u
a
li
ty
i
n
fo
rm
a
ti
o
n
tec
h
n
o
lo
g
y
fro
m
P
h
e
tch
a
b
u
ri
Ra
jab
h
a
t
Un
i
v
e
rsity
,
in
2
0
1
3
.
Cu
rre
n
tl
y
,
h
e
is
a
n
a
ss
istan
t
p
ro
fe
ss
o
r
a
t
th
e
F
a
c
u
lt
y
o
f
S
c
ien
c
e
a
n
d
Tec
h
n
o
l
o
g
y
,
Na
k
h
o
n
S
i
Th
a
m
m
a
ra
t
Ra
jab
h
a
t
Un
i
v
e
rsity
,
T
h
a
il
a
n
d
.
He
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
wic
h
it
_
so
o
@n
str
u
.
ac
.
th
.
Pa
to
m
p
o
n
g
Cha
b
p
l
a
n
re
c
e
iv
e
d
t
h
e
B
.
Sc
.
d
e
g
re
e
in
c
o
m
p
u
ter
sc
ien
c
e
fro
m
Na
k
h
o
n
S
i
Th
a
m
m
a
ra
t
Ra
jab
h
a
t
Un
iv
e
rsity
,
Th
a
il
a
n
d
,
in
2
0
0
5
th
e
M
.
S
.
d
e
g
re
e
in
m
a
n
a
g
e
m
e
n
t
o
f
in
fo
rm
a
ti
o
n
tec
h
n
o
l
o
g
y
fro
m
Wala
il
a
k
Un
iv
e
rsity
,
Th
a
il
a
n
d
,
i
n
2
0
1
0
,
a
n
d
t
h
e
P
h
.
D
.
d
e
g
re
e
in
i
n
fo
rm
a
ti
o
n
tec
h
n
o
l
o
g
y
,
Ki
n
g
M
o
n
g
k
u
t'
s
Un
iv
e
rsit
y
o
f
Te
c
h
n
o
l
o
g
y
No
rth
Ba
n
g
k
o
k
,
Th
a
il
a
n
d
,
i
n
2
0
2
0
.
Cu
rre
n
tl
y
,
h
e
i
s
a
lec
tu
re
r
a
t
t
h
e
De
p
a
rtme
n
t
o
f
I
n
fo
rm
a
ti
o
n
Tec
h
n
o
l
o
g
y
a
n
d
Dig
it
a
l
I
n
n
o
v
a
ti
o
n
o
f
th
e
F
a
c
u
lt
y
o
f
S
c
ien
c
e
a
n
d
Tec
h
n
o
lo
g
y
,
Na
k
h
o
n
S
i
T
h
a
m
m
a
ra
t
Ra
jab
h
a
t
Un
iv
e
rsity
,
T
h
a
il
a
n
d
.
He
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
p
a
to
m
p
o
n
g
_
c
h
a
@n
stru
.
ac
.
th
.
Evaluation Warning : The document was created with Spire.PDF for Python.