I
nte
rna
t
io
na
l J
o
urna
l o
f
I
nfo
r
m
a
t
ics a
nd
Co
mm
u
n
ica
t
io
n T
ec
hn
o
lo
g
y
(
I
J
-
I
CT
)
Vo
l.
7
,
No
.
1
,
A
p
r
il
201
8
,
p
p
.
16
~
23
I
SS
N:
2252
-
8776
,
DOI
: 1
0
.
1
1
5
9
1
/i
j
i
ct.
v
7
i1
.
p
p
1
6
-
23
16
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ia
e
s
co
r
e.
co
m/jo
u
r
n
a
ls
/in
d
ex
.
p
h
p
/
I
JI
C
T
Face
Reco
g
nition
in t
he
Scra
m
bled
Do
m
a
in
Using
MK
-
R
DA and
A
NN
K
a
v
it
a
K
a
da
m
*
,
C.
R
.
J
a
dh
a
v
De
p
a
rt
m
e
n
t
o
f
Co
m
p
u
ter E
n
g
in
e
e
rin
g
,
P
u
n
e
Un
iv
e
rsity
,
In
d
ia
Art
icle
I
nfo
AB
ST
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
J
an
1
6
,
2
0
1
8
R
ev
i
s
ed
Mar
1
,
2
0
1
8
A
cc
ep
ted
Ma
r
1
0
,
2
0
1
8
F
a
c
ial
lo
o
k
id
e
n
ti
ty
is
a
v
it
a
l
m
issi
o
n
b
y
m
e
a
n
s
o
f
h
u
m
a
n
-
in
tera
c
ti
n
g
stru
c
tu
re
s
th
a
t
g
o
a
l
t
o
b
e
a
w
a
re
o
f
v
e
rsio
n
s
w
it
h
in
th
e
h
u
m
a
n
’s
e
m
o
ti
o
n
a
l
sta
te.
th
e
p
rin
c
i
p
le
c
h
a
ll
e
n
g
e
o
r
th
e
c
ru
c
ial
p
a
rt
in
su
rv
e
il
lan
c
e
s
o
c
iety
is
th
e
p
riv
a
c
y
-
sh
ield
in
g
e
ra
.
b
e
c
a
u
se
t
h
e
ra
p
id
i
m
p
ro
v
e
m
e
n
t
in
th
e
in
tern
e
t
in
ter
n
a
ti
o
n
a
l
it
tu
rn
s
i
n
t
o
v
e
r
y
e
ss
e
n
ti
a
l
to
sc
ra
m
b
le
th
e
p
ics
in
t
h
e
v
id
e
o
o
r
f
il
e
s
f
o
r
th
e
d
u
ra
ti
o
n
o
f
tran
sm
is
sio
n
.
i
n
th
is
t
h
e
b
i
o
m
e
tri
c
id
e
n
ti
ty
o
f
p
h
o
t
o
g
ra
p
h
s
o
r
f
a
c
e
s
f
ro
m
sc
ra
m
b
led
p
ictu
re
s
p
lay
s
a
c
o
m
p
lete
l
y
to
u
g
h
m
issio
n
.
Nu
m
b
e
rs
o
f
v
a
rio
u
s
tec
h
n
o
l
o
g
y
a
re
c
a
rried
o
u
t
to
p
ro
v
id
e
p
riv
a
ten
e
ss
f
o
r
th
e
d
u
ra
ti
o
n
o
f
su
rv
e
il
lan
c
e
o
r
d
u
rin
g
tran
sm
issio
n
o
f
v
id
e
o
h
o
w
e
v
e
r
th
e
y
'
re
lac
k
o
f
e
ss
e
n
ti
a
l
traits,
li
k
e
re
v
e
rsib
il
it
y
o
r
v
isib
le
f
in
e
m
a
in
ten
a
n
c
e
.
in
lo
ts
o
f
sc
ra
m
b
li
n
g
m
e
th
o
d
s
t
h
e
f
a
c
e
s
a
r
e
c
o
v
e
re
d
b
y
a
f
e
w
a
n
i
m
a
ti
o
n
w
h
ich
m
a
y
a
d
d
it
io
n
a
ll
y
o
r
m
a
y
n
o
t
c
o
v
e
r
a
ll
f
a
c
e
s o
r
it
re
c
e
i
v
e
s h
a
rd
to
re
c
o
v
e
r
p
ics
f
ro
m
th
is
tec
h
n
iq
u
e
.
M
a
n
y
g
u
id
e
m
e
th
o
d
a
lso
a
re
u
s
u
se
d
b
y
w
h
ich
we
w
il
l
u
n
sc
ra
m
b
le
a
n
p
h
o
to
b
u
t
t
h
e
y
a
r
e
n
o
lo
n
g
e
r
p
o
w
e
r
f
u
l
th
a
t
a
g
o
o
d
d
e
a
l.
t
o
o
v
e
rc
o
m
e
a
ll
t
h
is
m
a
tt
e
rs
w
e
p
ro
p
o
se
d
a
n
o
v
e
l
a
p
p
r
o
a
c
h
-
M
a
n
y
-
Ke
rn
e
l
Ra
n
d
o
m
Disc
ri
m
in
a
te
a
n
a
l
y
sis
(M
K
-
RDA
)
to
f
in
d
o
u
t
d
isc
ri
m
in
a
ti
v
e
p
a
tt
e
rn
s
f
ro
m
c
h
a
o
ti
c
in
d
ica
to
rs.
stru
c
tu
re
s
g
e
t
b
e
tt
e
r
a
c
c
u
ra
c
y
b
rin
g
a
b
o
u
t
b
e
st
p
h
o
t
o
s.
T
o
P
I
E
a
n
d
ORL
d
a
tas
e
ts h
a
s g
e
tt
in
g
a
b
o
v
e
n
in
e
ty
%
a
c
c
u
ra
c
y
.
K
ey
w
o
r
d
:
Face
s
cr
a
m
b
li
n
g
Facial
b
io
m
etr
ics
Ma
n
y
m
a
n
i
f
o
ld
s
Mo
b
ile
b
io
m
etr
ics
R
an
d
o
m
d
i
s
cr
i
m
i
n
a
n
t a
n
a
l
y
s
is
Co
p
y
rig
h
t
©
2
0
1
8
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
:
Kav
ita
Kad
a
m
,
Dep
ar
t
m
en
t o
f
C
o
m
p
u
ter
E
n
g
i
n
ee
r
in
g
,
P
u
n
e
Un
iv
er
s
it
y
,
I
n
d
ia
.
1.
I
NT
RO
D
UCT
I
O
N
As t
h
e
s
p
ee
d
y
i
m
p
r
o
v
e
m
en
t
with
i
n
t
h
e
n
e
t i
n
ter
n
atio
n
al
it
w
i
ll b
ec
o
m
e
v
er
y
es
s
en
t
ial
to
s
cr
a
m
b
le
t
h
e
i
m
a
g
es i
n
t
h
e
v
id
eo
o
r
f
iles
a
ll
th
r
o
u
g
h
tr
an
s
m
i
s
s
io
n
.
A
s
a
en
d
r
esu
lt,
f
ac
e
s
cr
a
m
b
lin
g
A
.
M
elle,
J
.
-
L
.
D
u
g
ela
y
[
1
3
]
is
em
er
g
in
g
as
a
p
r
ac
ti
ca
l
ap
p
r
o
ac
h
to
d
ef
en
d
th
e
co
n
s
u
m
er
’
s
b
o
d
il
y
id
en
ti
f
icat
io
n
.
T
h
er
ef
o
r
e
b
y
s
cr
a
m
b
li
n
g
f
ac
e
d
etec
ted
in
n
o
n
-
p
u
b
lic
v
id
eo
is
ca
r
r
ied
o
u
t.
Scr
a
m
b
li
n
g
co
u
ld
b
e
v
er
y
ea
s
y
an
d
p
r
o
p
er
m
eth
o
d
as
co
m
p
ar
e
to
o
th
er
tech
n
iq
u
e
as
i
n
s
cr
a
m
b
lin
g
n
o
t
all
th
e
m
atter
s
ar
e
h
id
in
g
.
T
h
e
u
n
s
cr
a
m
b
li
n
g
also
ca
n
b
e
ac
co
m
p
lis
h
ed
th
r
u
m
a
n
u
al
tech
n
iq
u
e.
Scr
a
m
b
lin
g
h
a
s
ac
q
u
ir
e
m
o
r
e
r
ec
o
g
n
itio
n
w
it
h
in
th
e
ar
ea
o
f
s
u
r
v
eilla
n
ce
as
i
t
h
o
ld
th
e
p
r
iv
aten
es
s
r
u
les.
An
y
o
th
er
b
en
e
f
it
o
f
th
e
s
cr
a
m
b
l
in
g
ap
p
r
o
ac
h
is
th
at
it
les
s
d
if
f
ic
u
lt
th
a
n
t
h
e
s
tan
d
ar
d
en
cr
y
p
tio
n
m
et
h
o
d
s
.
T
h
e
s
cr
am
b
li
n
g
i
s
ex
ec
u
ted
th
r
o
u
g
h
m
a
n
y
tec
h
n
i
q
u
es
s
i
m
p
l
y
b
y
w
a
y
o
f
p
r
o
tectin
g
o
r
b
y
w
a
y
o
f
p
u
tti
n
g
ca
r
icatu
r
e
in
th
e
f
r
o
n
t
o
f
f
ac
e
f
o
r
p
r
iv
aten
e
s
s
r
ea
s
o
n
.
H
o
w
e
v
er
f
r
o
m
th
is
k
i
n
d
o
f
s
cr
am
b
lin
g
t
h
e
r
ec
o
g
n
it
io
n
o
r
id
e
n
tit
y
t
u
r
n
s
i
n
to
v
er
y
h
ar
d
s
p
ec
if
icall
y
in
t
h
e
ar
ea
o
f
p
r
o
tectio
n
h
ad
b
ee
n
er
as
in
g
t
h
e
f
ac
e
t
u
r
n
s
in
to
d
if
f
ic
u
lt
to
p
ick
o
u
t
th
e
m
an
o
r
w
o
m
a
n
.
On
t
h
is
w
e
'
v
e
u
s
ed
A
r
n
o
ld
tr
a
n
s
f
o
r
m
atio
n
al
g
o
r
ith
m
i
s
u
s
ed
to
s
cr
a
m
b
le
t
h
e
i
m
a
g
es
Y.
W
an
g
,
T
.
L
i
[
1
5
]
.
T
h
is
tech
n
iq
u
e
s
cr
a
m
b
le
d
th
e
p
h
o
to
g
r
ap
h
s
b
y
p
i
x
el
en
l
ar
g
e
m
e
n
t te
c
h
n
iq
u
e.
W
ith
in
th
e
d
is
cip
li
n
e
o
f
f
ac
e
r
e
co
g
n
itio
n
i
n
s
id
e
t
h
e
r
ec
e
n
t
y
ea
r
s
it
ac
h
iev
e
s
p
len
t
y
o
f
f
ac
to
r
s
to
b
ec
o
m
e
a
w
ar
e
o
f
th
e
f
ac
e
s
.
T
h
is
tec
h
n
iq
u
e
co
m
m
o
n
l
y
co
n
s
id
er
s
t
h
e
s
e
m
a
n
tic
f
ac
e
f
a
s
h
io
n
s
w
h
i
ch
is
in
teg
r
atio
n
o
f
d
if
f
er
e
n
t
s
e
m
a
n
tic
m
a
tter
s
to
g
eth
er
w
it
h
n
o
s
tr
il,
e
y
e,
a
n
d
m
o
u
th
.
b
y
u
s
i
n
g
th
e
u
s
ag
e
o
f
t
h
is
s
e
m
a
n
tic
n
ea
r
b
y
Evaluation Warning : The document was created with Spire.PDF for Python.
IJ
-
I
C
T
I
SS
N:
2252
-
8776
F
a
ce
R
ec
o
g
n
itio
n
in
t
h
e
S
cra
mb
led
Do
m
MK
-
R
DA
a
n
d
A
N
N
(
K
a
vita
K
a
d
a
m
)
17
b
in
ar
y
s
t
y
les
(
L
B
P
)
m
a
y
b
e
b
u
ilt
to
r
ec
o
v
er
th
e
p
h
o
to
g
r
ap
h
.
I
n
s
tead
o
f
2
n
d
a
3
D
m
o
d
el
is
u
s
ed
f
o
r
b
etter
im
p
ac
t
o
n
r
ec
o
v
er
ed
p
h
o
to
g
r
ap
h
.
T
h
e
d
is
tr
ict
o
f
in
ter
est
i
s
d
ec
o
d
ed
b
y
m
ea
n
s
o
f
r
estrict
d
ec
r
y
p
tio
n
w
it
h
co
m
p
lete
o
r
in
co
m
p
lete
r
ec
o
r
d
s
o
f
t
h
e
n
a
m
e
o
f
t
h
e
g
a
m
e
k
e
y
,
t
h
er
ef
o
r
e
m
ax
i
m
u
m
cr
it
ical
to
o
n
e
o
f
a
k
i
n
d
tier
s
o
f
s
cr
a
m
b
li
n
g
a
m
en
d
m
e
n
t.
I
n
f
o
r
m
a
tio
n
o
f
th
e
en
tire
s
ec
r
et
k
e
y
g
i
v
es
leg
i
ti
m
ate
cu
s
to
m
er
s
i
m
ag
e
r
ev
i
v
al
at
a
ex
ce
llen
ce
s
tag
e
v
er
y
n
ea
r
th
e
p
ar
ticu
lar
.
d
esp
it
e
th
e
in
f
as
h
io
n
ab
le
m
e
th
o
d
is
lo
s
s
y
,
w
e
d
is
p
la
y
t
h
at
d
u
r
in
g
s
er
ies
lo
s
s
,
p
lan
n
ed
th
r
u
r
ese
m
b
la
n
ce
m
etr
ic
s
,
is
e
q
u
al
o
r
s
m
aller
t
h
an
t
h
e
o
n
l
y
o
cc
u
r
b
y
J
P
E
G
d
en
s
it
y
.
B
u
t
th
e
s
cr
a
m
b
led
i
m
ag
e
co
u
l
d
b
e
v
er
y
o
n
e
o
f
a
k
in
d
as
ex
am
i
n
e
to
th
e
au
th
e
n
tic
f
ac
ial
p
h
o
to
g
r
ap
h
.
I
t
b
ec
o
m
es
d
if
f
ic
u
lt
to
co
m
p
a
r
e
th
e
3
D
m
o
d
el
P
er
ak
is
,
P
.
;
P
ass
alis
,
G.
;
T
h
eo
h
ar
is
,
T
.
;
Kak
ad
iar
is
,
I
.
A
.
[
1
]
w
it
h
s
cr
a
m
b
led
p
h
o
to
g
r
ap
h
s
b
ec
au
s
e
t
h
e
s
e
m
a
n
tic
m
o
d
el
g
r
o
w
to
b
e
ch
ao
tic
p
atter
n
.
T
o
a
v
o
id
th
o
s
e
r
ec
o
r
d
s
-
d
r
iv
en
tactic
s
is
u
s
ed
;
in
t
h
is
t
ec
h
n
iq
u
e
ch
ao
tic
s
i
g
n
als
ar
e
s
u
r
el
y
ta
k
en
i
n
to
co
n
s
id
er
atio
n
as
a
s
et
o
f
r
ec
o
r
d
s
p
o
in
ts
s
p
r
ea
d
o
v
er
m
an
if
o
ld
s
.
P
len
t
y
o
f
r
ec
o
v
er
y
m
et
h
o
d
o
r
tech
n
iq
u
e
is
g
iv
e
n
an
d
th
e
y
ef
f
icien
tl
y
u
s
ed
f
o
r
i
n
f
o
r
m
atio
n
-
d
r
iv
e
n
f
ac
e
r
ep
u
t
atio
n
.
B
u
t
,
f
o
r
th
e
s
cr
a
m
b
led
p
h
o
to
s
w
e
n
ee
d
a
s
tr
o
n
g
ap
p
r
o
ac
h
to
ad
d
r
ess
th
e
s
cr
a
m
b
led
s
n
ap
s
h
o
ts
.
i
n
t
h
is
p
ap
er
w
e
p
r
o
p
o
s
ed
a
n
e
w
m
eth
o
d
k
n
o
w
n
as M
a
n
y
-
Ker
n
el
R
a
n
d
o
m
Di
s
cr
i
m
in
at
e
ev
alu
a
tio
n
(
MK
-
R
D
A
)
to
d
e
al
w
it
h
ch
ao
tic
i
m
a
g
es
m
o
r
e
ef
f
icac
io
u
s
l
y
w
i
th
in
th
e
s
cr
a
m
b
led
d
o
m
ai
n
.
W
e
ad
d
itio
n
all
y
g
iv
e
s
a
w
a
y
s
alie
n
ce
v
er
s
io
n
u
s
ed
in
MK
-
R
D
A
f
o
r
p
atter
n
d
is
co
v
er
y
f
r
o
m
c
h
a
o
tic
f
ac
ial
aler
ts
.
1
.
1
.
F
a
ce
Scra
m
bli
ng
Scr
a
m
b
li
n
g
ca
p
tu
r
ed
p
er
s
o
n
al
p
h
o
to
g
r
ap
h
ca
n
b
e
an
s
w
er
to
s
i
m
p
li
f
y
i
n
g
a
s
ch
e
m
e.
W
e
p
r
o
p
o
s
e
an
i
m
a
g
e
-
s
cr
a
m
b
li
n
g
m
et
h
o
d
f
o
r
n
u
m
er
o
u
s
f
o
r
m
atted
(
b
it
m
ap
an
d
J
PEG)
im
a
g
es
to
n
o
n
-
p
u
b
li
c
in
f
o
r
m
at
io
n
.
T
h
e
n
o
r
m
al
p
h
o
to
s
ar
e
tr
an
s
f
o
r
m
ed
in
to
p
ec
u
liar
la
y
o
u
t
o
r
en
cr
y
p
ted
lay
o
u
t.
th
o
s
e
s
cr
a
m
b
led
p
h
o
to
g
r
ap
h
s
ar
e
h
id
i
n
g
th
e
f
ac
ts
o
f
p
h
o
to
s
.
u
s
in
g
A
r
n
o
ld
r
ew
o
r
k
.
t
h
at
i
s
r
e
m
o
d
el
p
ix
el
o
r
co
lo
r
atio
n
.
as
th
e
ch
ao
t
ic
s
ch
e
m
e
m
a
y
b
e
v
er
y
d
if
f
u
s
ed
to
s
ch
e
m
e
p
ar
am
eter
s
an
d
i
n
itial
v
alu
e
s
,
th
e
ch
ao
tic
s
er
ies
th
at
's
m
ad
e
h
a
s
th
e
ch
a
r
ac
ter
s
o
f
s
o
p
h
is
ticat
io
n
n
o
i
s
e,
w
id
e
b
an
d
,
co
r
r
ec
t r
en
e
w
al
a
n
d
co
m
p
li
ca
ted
to
p
r
e
d
ictio
n
len
g
th
y
-
ter
m
.
1
.
2
.
Sem
a
ntic
F
a
cia
l C
o
m
po
nents
A
ll
f
ac
e
is
d
ep
en
d
in
g
u
p
o
n
f
ac
ial
f
ea
t
u
r
es
w
h
ic
h
h
a
s
d
etec
ted
f
ac
e
in
n
at
u
r
al
p
ictu
r
es
t
o
d
is
co
v
er
ex
p
r
ess
io
n
o
f
p
ix
t
h
e
u
s
a
g
e
o
f
s
p
ec
ial
p
atch
es.
I
n
lap
to
p
im
ag
in
at
iv
e
a
n
d
p
r
escien
t
to
d
etec
tin
g
p
atch
es
h
as
u
s
ed
to
s
p
ec
if
ic
s
t
y
les o
f
ap
p
r
o
ac
h
o
r
s
et
o
f
r
u
les.
P
C
A
(
P
r
in
cip
al
C
o
m
p
o
n
e
n
t
An
al
y
s
i
s
)
is
p
h
o
to
g
r
ap
h
co
m
p
r
e
s
s
io
n
a
n
d
r
ep
u
tatio
n
w
h
ich
h
a
s
ex
tr
ac
ted
th
e
is
s
u
e
o
r
ite
m
.
P
C
A
i
s
a
p
r
o
ce
d
u
r
e
th
at
m
a
k
es
u
s
e
o
f
a
n
o
r
th
o
g
o
n
al
c
h
an
g
e
to
tr
an
s
f
o
r
m
a
f
i
x
ed
o
f
r
ea
s
o
n
s
o
f
lik
el
y
co
n
n
ec
ted
v
ar
iab
les
o
r
m
o
v
in
g
o
b
j
ec
t
in
to
a
f
ix
ed
o
f
v
alu
es
o
f
li
n
ea
r
l
y
u
n
co
r
r
elate
d
v
ar
iab
les
r
ef
er
r
ed
to
as
i
m
p
o
r
tan
t
m
ec
h
an
i
s
m
s
.
T
h
e
p
r
ec
is
e
m
u
tab
le
p
r
i
ce
ex
tr
a
th
an
o
r
eq
u
al
to
q
u
a
n
ti
t
y
o
f
p
r
ed
o
m
in
a
n
t
t
h
i
n
g
.
th
is
m
o
d
i
f
icatio
n
is
d
is
tin
ct
in
s
u
ch
a
m
eth
o
d
th
at
th
e
lead
er
f
u
n
d
a
m
e
n
tal
ch
ar
ac
ter
is
tic
h
a
s
th
e
p
r
in
cip
le
an
d
all
f
o
llo
w
i
n
g
th
i
n
g
i
n
m
o
v
e
h
as
t
h
e
h
ig
h
e
s
t
alter
atio
n
v
iab
le
u
n
d
er
t
h
e
r
estra
in
t
t
h
at
it's
f
ar
o
r
th
o
g
o
n
al
ch
ar
ac
ter
is
tic.
T
h
e
r
es
u
lti
n
g
v
ec
to
r
s
ar
e
an
u
n
co
r
r
elate
d
o
r
th
o
g
o
n
al
f
o
u
n
d
atio
n
s
e
t.
P
C
A
is
ex
p
r
es
s
i
v
e
to
t
h
e
co
m
p
ar
ati
v
e
s
ca
li
n
g
o
f
t
h
e
s
p
e
cif
ic
v
ar
iab
les.
FL
D
A
(
Fi
s
h
er
L
i
n
ea
r
Dis
cr
i
m
in
an
t
An
al
y
s
i
s
)
ex
ce
r
p
t
th
e
att
r
ib
u
te
f
r
o
m
s
n
ap
s
h
o
ts
w
h
ic
h
p
r
eser
v
es
th
e
d
is
cr
i
m
i
n
ati
v
e
is
s
u
e
o
f
i
m
ag
es
w
h
il
s
t
p
lu
m
m
eti
n
g
m
ea
s
u
r
e
m
e
n
t
at
t
h
e
p
h
o
to
g
r
ap
h
ar
ea
.
FL
D
A
g
et
s
th
e
m
o
d
i
f
icat
io
n
m
atr
ix
b
y
m
ea
n
s
o
f
ex
p
lo
iti
n
g
t
h
e
a
m
o
n
g
-
e
leg
a
n
ce
s
ca
tter
m
ed
iu
m
F
L
D
A
ca
n
n
o
t
r
eser
v
at
io
n
th
e
s
p
ec
if
ic
co
v
er
s
o
f
t
h
at
p
ar
ticu
l
ar
m
a
g
n
if
icen
ce
.
L
FD
A
h
as
b
ee
n
d
esti
n
y
to
tr
iu
m
p
h
o
v
er
d
r
a
w
b
ac
k
s
o
f
F
L
D
A
Y.
R
ah
u
la
m
at
h
av
a
n
,
R
.
C
.
-
W
.
P
h
an
,
J
.
A
.
C
h
a
m
b
er
s
,
D.
J
.
Par
is
h
[
5
]
.
L
FD
A
(
L
o
ca
l
Fis
h
er
Dis
cr
i
m
i
n
an
t
An
al
y
s
i
s
)
d
iv
is
io
n
s
p
ictu
r
e
ex
a
m
p
les
in
e
v
e
r
y
ele
g
an
ce
i
n
to
m
an
if
o
ld
lo
ca
l
class
es
w
it
h
i
n
th
e
b
etter
d
im
e
n
s
io
n
a
l
p
h
o
to
s
p
ac
e
b
y
u
s
i
n
g
A
i;j
;
8
i;
j
.
I
t
th
en
s
ch
e
m
e
s
p
h
o
to
g
r
ap
h
s
f
itti
n
g
to
a
n
ei
g
h
b
o
r
h
o
o
d
m
a
g
n
if
icen
ce
i
n
ad
v
a
n
ce
to
ev
er
y
o
th
er
w
h
ile
m
ai
n
tain
i
n
g
p
r
ed
ictab
le
i
m
a
g
er
ies
o
f
o
t
h
er
n
ei
g
h
b
o
r
h
o
o
d
tr
ain
in
g
s
ep
ar
atel
y
Y.
R
a
h
u
la
m
at
h
a
v
an
,
R
.
C
.
-
W
.
P
h
an
,
J
.
A
.
C
h
a
m
b
er
s
,
D.
J
.
P
ar
is
h
[
5
]
.
2.
SYST
E
M
ARCH
I
T
E
CT
U
R
E
Sch
e
m
atic
m
ac
h
i
n
e
ar
ch
itec
tu
r
e
is
as
s
h
o
w
n
i
n
F
ig
u
r
e
1
.
T
h
e
i
m
ag
e
th
a
t
is
to
b
e
s
cr
a
m
b
led
is
f
ir
s
t
g
iv
e
n
t
o
th
e
d
e
v
ice
b
y
u
s
i
n
g
th
e
u
s
er
.
Fro
m
t
h
is
i
m
a
g
e,
f
a
ce
is
d
etec
ted
v
ia
t
h
e
Vio
la
-
j
o
h
n
s
s
et
o
f
r
u
les
to
s
cr
a
m
b
le
t
h
e
f
ac
e
o
r
to
en
cr
y
p
t
th
e
i
m
a
g
e
t
h
at
i
s
to
b
e
s
e
n
d
in
g
to
d
i
f
f
er
en
t
s
id
e.
Af
ter
t
h
e
f
ac
e
d
etec
ted
b
y
m
ea
n
s
o
f
Vio
la
-
J
o
h
n
s
alg
o
r
ith
m
p
h
o
to
g
r
ap
h
is
o
r
f
ac
e
is
s
cr
a
m
b
led
b
y
m
ea
n
s
o
f
u
s
i
n
g
A
r
n
o
ld
T
r
an
s
f
o
r
m
atio
n
alg
o
r
ith
m
.
T
h
en
th
is
s
cr
a
m
b
le
d
im
a
g
e
is
s
en
d
to
d
if
f
er
en
t
asp
ec
t.
T
o
p
e
r
ce
iv
e
th
e
in
d
i
v
id
u
a
l
o
r
f
ac
e
in
s
cr
a
m
b
le
p
h
o
to
at
g
i
v
e
u
p
asp
ec
t
a
M
K
-
R
D
A
i.e
.
Ma
n
y
-
Ker
n
el
R
a
n
d
o
m
Di
s
cr
i
m
in
a
te
e
v
a
lu
at
io
n
tec
h
n
iq
u
e
i
s
u
s
ed
.
T
h
is
p
h
o
to
is
th
e
n
ex
a
m
i
n
e
w
i
th
d
if
f
er
en
t
p
ix
s
a
v
es
in
d
atas
et
an
d
ac
co
r
d
in
g
to
th
eir
m
atc
h
r
atin
g
is
g
i
v
e
n
to
ev
er
y
p
ict
u
r
es
t
h
at
'
s
s
a
m
e
to
t
h
at
r
ec
o
v
er
p
h
o
to
.
T
h
e
d
ataset
u
s
ed
i
n
t
h
is
g
ad
g
e
t
f
o
r
m
atc
h
in
g
f
ac
es
ar
e
O
R
L
Data
s
et,
P
I
E
Data
s
et,
an
d
P
UB
FIG
d
ataset.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8776
IJ
-
I
C
T
Vo
l.
7
,
No
.
1
,
A
p
r
il
20
1
8
:
16
–
23
18
Fig
u
r
e
1
.
S
y
s
te
m
A
r
c
h
itect
u
r
e
2
.
1
.
User
M
o
du
le:
User
ca
n
u
p
lo
ad
th
e
s
cr
a
m
b
le
d
im
a
g
e
to
th
e
s
y
s
te
m
o
r
u
p
lo
ad
th
e
d
atasets
to
th
e
s
y
s
te
m
.
2
.
2
.
P
re
-
pro
ce
s
s
ing
:
I
n
i
m
a
g
e
p
r
o
ce
s
s
in
g
o
p
er
atio
n
i
m
p
le
m
en
t
to
t
h
e
p
r
e
-
p
r
o
ce
s
s
i
n
g
.
I
n
t
h
e
s
y
s
te
m
i
m
a
g
e
co
n
v
e
r
t
in
to
g
r
e
y
s
ca
le
i
m
a
g
e
m
ea
n
s
i
m
ag
e
co
n
v
er
t
in
to
b
lack
a
n
d
w
h
i
te
i
m
a
g
e
w
h
ich
r
e
m
o
v
e
b
r
ig
h
t
n
ess
o
f
i
m
a
g
e.
I
n
f
ac
ia
l
i
m
a
g
e
f
i
n
d
f
ac
e
i
s
f
ac
ia
l o
r
n
o
n
f
ac
ia
l
w
h
ic
h
is
f
i
n
d
o
r
i
m
p
lem
en
t in
to
f
ac
e
r
eo
r
g
an
izatio
n
alg
o
r
ith
m
.
2
.
3
.
Vio
la
J
o
nes
:
Usi
n
g
v
io
la
j
o
n
es
alg
o
r
ith
m
d
etec
tin
g
h
u
m
a
n
f
ac
e
s
w
it
h
f
ac
es
p
atter
n
lik
e
e
y
es,
n
o
is
e
an
d
lip
s
.
T
h
e
p
r
o
b
lem
to
b
e
s
o
l
v
ed
is
d
etec
tio
n
o
f
f
ac
e
s
i
n
a
n
i
m
ag
e.
A
h
u
m
a
n
ca
n
d
o
th
i
s
ea
s
il
y
,
b
u
t
a
co
m
p
u
ter
n
ee
d
s
p
r
ec
is
e
in
s
tr
u
ctio
n
s
a
n
d
co
n
s
tr
ain
ts
.
T
o
m
a
k
e
t
h
e
ta
s
k
m
o
r
e
m
an
a
g
ea
b
le,
Vio
la
–
J
o
n
es
r
eq
u
ir
es
f
u
ll
v
ie
w
f
r
o
n
tal
u
p
r
ig
h
t
f
ac
es.
T
h
u
s
in
o
r
d
er
t
o
b
e
d
etec
ted
,
th
e
en
tire
f
ac
e
m
u
s
t
p
o
in
t
to
w
ar
d
s
th
e
ca
m
er
a
an
d
s
h
o
u
ld
n
o
t
b
e
tilt
ed
to
eith
er
s
id
e.
W
h
ile
it s
ee
m
s
t
h
ese
co
n
s
tr
ain
ts
co
u
ld
d
i
m
i
n
is
h
t
h
e
alg
o
r
it
h
m
’
s
u
til
it
y
s
o
m
e
w
h
at,
b
ec
au
s
e
th
e
d
etec
tio
n
s
tep
is
m
o
s
t
o
f
te
n
f
o
llo
w
ed
b
y
a
r
ec
o
g
n
iti
o
n
s
tep
,
i
n
p
r
ac
tice
t
h
ese
li
m
it
s
o
n
p
o
s
e
ar
e
q
u
ite
ac
ce
p
tab
le.
2
.
4
.
MK
-
RDA
Ma
n
y
k
er
n
els
R
a
n
d
o
m
Dec
r
e
m
en
t
An
al
y
s
is
f
o
r
r
an
d
o
m
f
ac
e
r
ec
o
g
n
itio
n
tec
h
n
iq
u
es.
W
h
i
ch
h
a
s
u
s
ed
to
d
if
f
er
en
t
t
y
p
es
o
f
k
er
n
e
l
u
s
ed
lik
e
b
lu
r
i
m
a
g
e
s
h
ar
p
n
es
s
,
in
cr
ea
s
i
n
g
b
r
ig
h
t
n
es
s
to
in
cr
ea
s
e
i
m
a
g
e
q
u
alit
y
.
Af
ter
in
cr
ea
s
e
q
u
alit
y
i
m
ag
e
w
il
l b
e
r
ec
o
g
n
ized
u
s
in
g
ch
ao
t
ic
s
ig
n
als.
2
.
5
.
F
ee
d
-
F
o
r
w
a
rd
Neura
l
Net
w
o
rk
s
A
co
llectio
n
o
f
n
e
u
r
o
n
s
co
n
n
ec
ted
to
g
eth
er
i
n
a
n
e
t
w
o
r
k
ca
n
b
e
r
ep
r
e
s
en
ted
b
y
ad
ir
ec
ted
g
r
ap
h
:.
No
d
es
r
ep
r
esen
t th
e
n
eu
r
o
n
s
,
an
d
ar
r
o
w
s
r
ep
r
esen
t th
e
li
n
k
s
b
et
w
ee
n
th
e
m
.
E
ac
h
n
o
d
e
h
as
it
s
n
u
m
b
er
,
an
d
a
lin
k
co
n
n
ec
ti
n
g
t
w
o
n
o
d
es
w
ill
h
av
e
a
p
air
o
f
n
u
m
b
er
s
(
e.
g
.
(
1
,
4
)
co
n
n
ec
ti
n
g
n
o
d
es
1
an
d
4
)
.
Net
w
o
r
k
s
w
it
h
o
u
t
c
y
cle
s
(
f
ee
d
b
ac
k
lo
o
p
s
)
ar
e
ca
lled
a
f
ee
d
-
f
o
r
w
ar
d
n
et
w
o
r
k
s
(
o
r
p
er
ce
p
tr
o
n
)
.
Fig
u
r
e
2
.
Feed
-
fo
r
w
ar
d
Neu
r
al
Net
w
o
r
k
s
a.
I
n
p
u
t
n
o
d
es
o
f
t
h
e
n
e
t
w
o
r
k
(
n
o
d
es
1
,
2
an
d
3
)
ar
e
ass
o
ciate
d
w
it
h
th
e
i
n
p
u
tv
ar
iab
les
(
x
1
,
.
.
.
,
x
m
)
.
T
h
ey
d
o
n
o
t c
o
m
p
u
te
an
y
t
h
in
g
,
b
u
t si
m
p
l
y
p
ass
t
h
e
v
a
lu
e
s
to
th
e
p
r
o
ce
s
s
in
g
n
o
d
es.
b.
Ou
tp
u
t n
o
d
es (
4
an
d
5
)
ar
e
ass
o
ciate
d
w
it
h
th
e
o
u
tp
u
t
v
ar
iab
les (
y
1
,
.
.
.
,
y
n
)
.
Evaluation Warning : The document was created with Spire.PDF for Python.
IJ
-
I
C
T
I
SS
N:
2252
-
8776
F
a
ce
R
ec
o
g
n
itio
n
in
t
h
e
S
cra
mb
led
Do
m
MK
-
R
DA
a
n
d
A
N
N
(
K
a
vita
K
a
d
a
m
)
19
3.
M
AT
H
E
M
AT
I
CAL
M
O
DE
L
S={s,
e,
X,
Y,
}
W
h
er
e,
s
=
Star
t o
f
th
e
p
r
o
g
r
a
m
.
1.
L
o
g
i
n
w
it
h
S
y
s
te
m
.
2.
Up
lo
ad
s
cr
am
b
led
i
m
a
g
e.
e
=
E
n
d
o
f
th
e
p
r
o
g
r
a
m
.
Fin
d
m
a
tch
i
n
g
o
f
s
cr
a
m
b
led
im
ag
e
i
n
r
an
k
i
n
g
.
X
=
I
n
p
u
t o
f
t
h
e
p
r
o
g
r
a
m
.
I
n
p
u
t o
f
th
i
s
s
y
s
te
m
is
d
ata
o
f
d
if
f
er
en
t scr
a
m
b
led
i
m
a
g
e
u
p
lo
ad
o
r
u
p
lo
ad
d
if
f
er
en
t
d
atasets
w
h
ich
co
n
v
er
t i
n
to
s
c
r
a
m
b
led
.
Y
=
Ou
tp
u
t
o
f
th
e
p
r
o
g
r
a
m
.
F
ir
s
t
w
e
ar
e
g
o
i
n
g
to
ex
tr
ac
t
t
h
e
d
is
tin
ct
k
i
n
d
f
ac
e
r
eo
r
g
an
iza
tio
n
al
g
o
r
ith
m
t
h
at
ar
e
co
m
e
ac
r
o
s
s
to
e
y
e,
lip
s
a
n
d
n
o
s
tr
il.
T
h
o
s
e
ev
er
y
p
ar
t p
u
t in
f
o
r
ce
to
MK
-
R
D
A
t
h
r
o
u
g
h
ch
ao
tic
s
i
g
n
a
l.
X,
Y
∈
U
L
et
U
b
e
th
e
Se
t o
f
S
y
s
te
m
.
U=
{Sc,
Dsc}
W
h
er
e
Sc,
Dsc
ar
e
th
e
ele
m
en
ts
o
f
t
h
e
s
et.
Sc
=
Scr
a
m
b
led
i
m
a
g
e.
Dsc=
Scr
a
m
b
led
i
m
ag
e
d
atase
ts
.
A
.
)
E
q
u
atio
n
I
n
s
cr
a
m
b
led
d
ev
ice
h
a
s
d
o
n
e
o
p
er
atio
n
in
ex
cl
u
s
iv
e
d
ata
s
et
s
.
ev
er
y
d
ata
s
ets
h
as
o
n
e
-
of
-
a
-
k
in
d
s
ize
f
o
r
m
atted
p
h
o
to
s
ar
e
to
b
e
h
ad
.
I
n
g
ad
g
et
d
atab
ase
s
to
r
ed
an
d
m
atc
h
i
n
g
lo
ca
ted
th
e
p
ictu
r
es i
n
t
h
ese
d
atasets
.
D=
Data
s
ets.
(
1
)
W
h
er
e,
P
i=
n
o
o
f
f
ac
ial
p
atch
e
s
I
m
g
=
I
m
ag
e.
I
n
eq
(
1
)
h
as
ca
lc
u
lated
p
atch
e
s
o
f
f
ac
ial
i
m
a
g
es.
I
t
u
s
i
n
g
v
io
la
J
o
n
es
alg
o
r
it
h
m
f
o
r
d
etec
tin
g
f
ac
ial
at
tr
ib
u
te
li
k
e
e
y
es,
n
o
s
e,
an
d
lip
s
co
r
n
er
.
T
h
ese
p
atch
es a
r
e
ex
tr
ac
ted
to
i
m
ag
es
u
s
i
n
g
alg
o
r
it
h
m
.
(
2
)
W
h
er
e,
SI=
Scr
a
m
b
led
i
m
a
g
e.
Fo
r
g
en
er
ati
n
g
s
cr
a
m
b
led
i
m
a
g
es
ap
p
l
y
A
r
n
o
ld
tr
an
s
f
o
r
m
f
o
r
f
ac
ial
i
m
a
g
es.
I
n
tr
an
s
f
o
r
m
atio
n
h
as
d
i
f
f
er
en
t
ex
p
an
s
io
n
w
h
ich
h
as c
o
n
v
er
t i
m
ag
e
i
n
to
s
cr
a
m
b
led
d
o
m
ai
n
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8776
IJ
-
I
C
T
Vo
l.
7
,
No
.
1
,
A
p
r
il
20
1
8
:
16
–
23
20
(
3
)
W
h
er
e
i
m
g
=to
p
k
m
atc
h
i
n
g
im
ag
e
s
.
Ker
=M
K
-
R
D
A
k
er
n
el
I
n
eq
u
atio
n
(
3
)
ca
lcu
lati
n
g
M
K
-
R
D
A
k
er
n
el
f
o
r
s
cr
a
m
b
led
i
m
ag
e
s
w
h
ich
h
as
d
if
f
er
e
n
t
t
y
p
es
k
er
n
el
u
s
ed
.
I
t
h
as
ca
lcu
lati
n
g
s
i
m
ilar
i
m
ag
e
s
w
it
h
u
s
in
g
f
ac
ial
p
atch
e
s
.
4.
DATAS
E
T
E
VA
L
U
T
I
O
N
On
t
h
is
d
ev
ice
w
e
'
v
e
g
o
t
tak
e
n
o
n
atte
n
tio
n
a
d
atab
ase
o
r
d
ataset
f
o
r
co
m
p
ar
is
o
n
o
f
o
b
tain
ed
i
m
a
g
e
s
w
it
h
t
h
e
d
ataset
s
n
ap
s
h
o
t
s
f
o
r
r
ec
o
g
n
itio
n
o
f
p
ict
u
r
es
f
r
o
m
t
h
e
s
cr
a
m
b
led
p
ix
.
W
e
u
s
ed
t
h
r
ee
f
o
r
m
o
f
d
ata
s
et
f
o
r
ev
alu
a
tio
n
p
u
r
p
o
s
e
w
h
ich
ca
n
b
e
as o
b
s
er
v
e:
4
.
1
.
O
RL
Da
t
a
s
et
:
T
h
e
Ou
r
Data
b
ase
o
f
Face
s
,
f
o
r
m
all
y
k
n
o
w
n
a
s
as
“
T
h
e
OR
L
Data
b
ase
o
f
Face
s
”.
I
n
clu
d
e
f
ac
e
p
ictu
r
es
tak
en
f
r
o
m
A
p
r
il
1
9
9
2
t
o
A
p
r
il
1
9
9
4
o
n
th
e
lab
w
i
th
te
n
s
p
ec
ial
p
ix
o
f
ea
ch
o
f
f
o
r
t
y
d
is
tin
ct
to
p
ics.
Fo
r
a
f
e
w
s
u
b
j
ec
ts
,
p
h
o
to
s
h
a
v
e
b
ee
n
ex
cited
ab
o
u
t v
ar
y
i
n
g
m
ild
,
ex
tr
ao
r
d
in
ar
y
f
ac
ial
e
x
p
r
ess
io
n
a
n
d
f
ac
ial
in
f
o
r
m
atio
n
.
4.
2
.
P
I
E
Da
t
a
s
et
:
P
I
E
Data
s
et
k
n
o
w
n
as
it
as
C
MU
P
o
s
e,
I
llu
m
i
n
atio
n
,
an
d
E
x
p
r
ess
io
n
(
P
I
E
)
d
atab
ase.
T
h
is
d
ata
s
et
co
n
tain
4
1
,
3
6
8
p
ictu
r
es
o
f
6
8
h
u
m
a
n
b
ein
g
s
u
n
d
er
1
3
s
p
ec
if
ic
an
g
le
s
,
f
o
r
t
y
th
r
ee
o
n
e
-
of
-
a
-
k
i
n
d
li
g
h
t
s
co
n
d
itio
n
s
,
a
n
d
w
i
th
f
o
u
r
s
p
ec
if
ic
e
x
p
r
ess
io
n
s
.
4.
3
.
PU
B
F
I
G
Da
t
a
s
et
:
P
UB
FIG
d
ataset
is
Stan
d
s
f
o
r
P
u
b
lic
Fig
u
r
es
Face
Data
s
e
t
is
th
e
b
i
g
g
e
s
t
d
ataset
o
f
r
e
al
p
h
r
ase
en
co
m
p
as
s
5
8
,
7
9
7
p
ictu
r
es
o
f
t
w
o
h
u
n
d
r
ed
p
e
o
p
le
g
ath
er
ed
f
r
o
m
th
e
i
n
ter
n
et.
T
h
ese
d
ataset
p
h
o
to
s
ar
e
tak
en
in
ex
cl
u
s
iv
e
o
u
t
o
f
co
n
tr
o
l
an
d
n
o
n
co
o
p
er
ativ
e
s
itu
atio
n
.
as
a
r
esu
lt
t
h
er
e's
lar
g
e
v
ar
ia
n
t
is
ex
p
r
ess
io
n
,
lig
h
t
s
,
p
o
s
e,
ca
m
er
a,
s
ce
n
e,
p
ar
a
m
ete
r
s
an
d
i
m
a
g
i
n
g
s
it
u
atio
n
s
,
an
d
s
o
o
n
.
5.
E
XP
E
R
I
M
E
NT
A
L
SE
T
UP
AND
RE
SUL
T
ANA
L
YS
I
S
5
.
1
E
x
peri
m
ent
a
l Set
up
P
r
o
p
o
s
ed
s
cr
am
b
led
i
m
ag
e
m
atc
h
in
g
s
y
s
te
m
b
ec
a
m
e
ap
p
lied
in
J
av
a.
It
ca
n
b
e
r
u
n
o
n
w
i
n
d
o
w
s
XP
/
w
i
n
d
o
w
s
Vis
ta
o
r
o
n
w
i
n
d
o
w
s
7
w
o
r
k
i
n
g
s
y
s
te
m
.
Fo
r
s
to
r
in
g
r
ec
o
r
d
s
d
atab
ase
u
s
ed
is
M
y
SQ
L
5
.
2
Resul
t
Ana
ly
s
is
5
.
2
.
1
I
m
a
g
e
P
ro
ce
s
s
ing
:
Up
lo
ad
i
m
ag
e
to
f
ee
d
f
o
r
w
ar
d
alg
o
r
ith
m
to
i
m
a
g
e
p
r
o
ce
s
s
in
g
w
h
ich
h
a
s
g
et
i
m
ag
e
k
e
y
p
o
in
ts
to
m
atc
h
in
g
r
es
u
lt.
Fig
u
r
e
3
.
I
m
a
g
e
P
r
o
ce
s
s
in
g
Evaluation Warning : The document was created with Spire.PDF for Python.
IJ
-
I
C
T
I
SS
N:
2252
-
8776
F
a
ce
R
ec
o
g
n
itio
n
in
t
h
e
S
cra
mb
led
Do
m
MK
-
R
DA
a
n
d
A
N
N
(
K
a
vita
K
a
d
a
m
)
21
5
.
2
.
2
.
E
rr
o
r
G
ra
ph
:
I
t
is
d
is
p
lay
to
p
r
o
ce
s
s
in
g
im
ag
e
g
r
ap
h
h
o
w
m
a
n
y
i
m
a
g
es
ar
e
tr
ain
,
f
ac
es
ar
e
d
etec
te
d
ar
e
s
o
m
e
co
m
b
i
n
atio
n
al
v
al
u
e
s
.
Fig
u
r
e
4.
E
r
r
o
r
Gr
ap
h
5
.
2
.
3
Rec
o
g
nitio
n:
I
t
h
as
f
i
n
d
to
m
atc
h
i
n
g
i
m
a
g
es
to
d
ataset
b
y
u
s
in
g
f
ac
ial
k
e
y
p
o
in
t’
s
v
alu
e
s
.
I
t
d
is
p
la
y
t
h
e
all
m
atc
h
in
g
i
m
a
g
es i
n
d
ataset.
Fig
u
r
e
5.
R
ec
o
g
n
i
tio
n
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8776
IJ
-
I
C
T
Vo
l.
7
,
No
.
1
,
A
p
r
il
20
1
8
:
16
–
23
22
5
.
2
.
4
M
a
t
ching
Acc
ura
cy
:
I
t D
is
p
la
y
t
h
e
ac
cu
r
ac
y
o
f
m
a
t
ch
in
g
r
es
u
lt.
Fig
u
r
e
6
.
Ma
tch
in
g
A
cc
u
r
ac
y
Fo
r
ev
alu
atio
n
ca
u
s
e,
w
e
k
ee
p
in
m
i
n
d
th
e
s
tate
o
f
a
f
f
air
s
o
f
f
ac
e
is
o
latio
n
s
af
et
y
.
in
s
p
ite
o
f
o
u
r
ap
p
r
o
ac
h
m
a
y
b
e
i
m
p
le
m
e
n
te
d
to
a
d
iv
er
s
it
y
o
f
p
h
o
to
g
r
ap
h
g
lad
,
h
u
m
a
n
f
ac
es
ar
e
n
o
tab
l
y
u
s
ed
f
o
r
d
if
f
ic
u
lt
y
r
ep
u
tatio
n
.
I
n
o
u
r
i
m
itatio
n
test
,
w
e
ca
r
r
ied
o
u
t
th
e
p
r
o
j
ec
te
d
f
ilter
o
u
t
to
a
s
ep
ar
atio
n
o
f
ap
p
r
o
x
im
a
tel
y
2
0
0
f
ac
es
f
r
o
m
t
h
e
s
p
ec
ial
o
f
Face
s
an
d
o
n
e
o
f
a
k
in
d
p
r
in
cip
les
f
o
r
th
e
s
cr
a
m
b
l
in
g
s
tr
e
n
g
th
r
es
tr
ictio
n
λ
.
I
n
s
cr
a
m
b
led
i
m
a
g
es
ar
ea
h
as
o
n
e
o
f
a
k
in
d
f
o
r
m
s
o
f
d
ataset
s
ar
e
u
s
ed
.
I
n
all
d
atasets
h
av
e
e
x
ce
p
tio
n
al
s
ize
ca
p
ab
le
p
h
o
to
g
r
ap
h
s
ar
e
av
ailab
le.
I
n
s
cr
a
m
b
led
p
ictu
r
es
ac
cu
r
ac
y
an
d
o
v
er
all
p
er
f
o
r
m
an
ce
to
b
e
h
ad
in
b
est o
f
i
m
a
g
es a
n
d
len
g
t
h
.
I
f
i
m
ag
e
e
x
ce
lle
n
t is lo
w
t
h
e
n
n
o
w
n
o
t c
o
m
e
ac
r
o
s
s
to
p
r
o
p
er
f
ac
ial
attr
ib
u
te
i
n
g
ad
g
et.
I
n
c
u
r
r
en
t
s
y
s
te
m
h
a
s
d
etec
ted
to
th
e
f
ac
e
in
h
er
b
al
p
ix
.
Ho
w
ev
er
i
t
h
as
lo
w
ac
cu
r
ac
y
a
n
d
ef
f
icien
c
y
f
ee
in
o
u
r
s
et
o
f
r
u
le
s
.
T
ab
le
1
.
Me
th
o
d
R
esu
lt
M
e
t
h
o
d
P
C
A
K
P
C
A
L
D
A
K
L
D
A
LLP
MK
-
R
D
A
A
c
c
u
r
a
c
y
7
6
.
0
7
6
.
0
8
0
.
0
8
1
.
5
8
3
.
1
9
0
.
1
I
n
ab
o
v
e
tab
le
h
as
s
h
o
w
th
e
d
if
f
er
e
n
t
s
t
y
les
o
f
ap
p
r
o
ac
h
a
n
d
its
ac
cu
r
ac
y
d
eg
r
ee
.
I
t
c
o
m
p
ar
e
s
th
e
d
is
tin
ct
iv
e
ap
p
r
o
ac
h
w
h
ic
h
h
a
s
u
s
ed
to
I
n
ex
clu
s
i
v
e
d
atasets
h
as
ex
tr
ao
r
d
in
ar
y
ac
cu
r
ac
y
d
is
co
v
er
.
d
u
e
to
th
e
f
ac
t
ea
c
h
d
atasets
p
ict
u
r
e
le
n
g
th
an
d
n
ice
ar
e
m
o
d
i
f
ied
.
p
ictu
r
e
s
cr
a
m
b
led
d
o
m
ai
n
ar
e
d
ep
en
d
ab
le
in
to
i
m
ag
e
n
ice
d
u
e
to
th
e
f
ac
t
m
a
n
y
ti
m
e
n
o
is
e
p
ict
u
r
es
ar
e
av
a
ilab
le
f
o
r
d
etec
tio
n
s
o
f
ac
ial
c
h
ar
ac
ter
is
tic
ar
e
n
o
t
d
etec
ted
w
ell.
T
o
d
is
co
v
er
f
ac
ia
l
attr
ib
u
te
w
e
u
s
ed
to
f
ac
ial
al
g
o
r
ith
m
w
h
ic
h
h
as
lo
ca
te
p
r
e
cisel
y
all
s
e
n
s
iti
v
e
f
ac
ial
attr
ib
u
te.
T
ab
le
2.
R
esu
lt f
o
r
OR
L
Da
t
as
et
M
e
t
h
o
d
P
C
A
K
P
C
A
L
D
A
K
L
D
A
LLP
MK
-
R
D
A
A
c
c
u
r
a
c
y
75
75
77
80
85
92
A
b
o
v
e
en
d
r
esu
lt f
o
r
OR
L
d
atasets
w
h
ich
p
ictu
r
es a
r
e
b
etter
s
atis
f
ac
to
r
y
f
o
r
P
I
E
d
atasets
s
o
r
o
u
tin
el
y
elev
ated
ac
cu
r
ac
y
o
f
d
ev
ice.
S
y
s
te
m
i
s
i
m
p
le
m
en
ted
u
p
to
First,
in
th
e
s
y
s
te
m
i
m
a
g
es
ar
e
s
to
r
ed
to
d
ir
ec
to
r
y
f
o
r
th
e
m
a
tch
i
n
g
p
u
r
p
o
s
e
w
h
i
ch
is
ca
lled
d
at
aset
an
d
p
ath
o
f
th
e
i
m
a
g
e
s
to
r
ed
in
to
th
e
d
at
ab
ase.
Fro
m
th
e
f
il
e
b
r
o
w
s
er
i
m
a
g
e
i
s
to
g
i
v
e
n
as
in
p
u
t
f
o
r
th
e
f
u
r
th
er
p
r
o
ce
s
s
in
g
.
Af
ter
g
etti
n
g
i
m
ag
e
to
t
h
e
s
y
s
te
m
,
it
f
ir
s
t
p
r
ep
r
o
ce
s
s
es th
e
i
m
a
g
e.
A
f
ter
P
r
ep
r
o
ce
s
s
ed
i
m
ag
e,
i
m
ag
e
p
r
o
ce
s
s
in
g
o
p
er
atio
n
s
to
b
e
d
o
n
e
o
n
th
e
i
m
ag
e.
Evaluation Warning : The document was created with Spire.PDF for Python.
IJ
-
I
C
T
I
SS
N:
2252
-
8776
F
a
ce
R
ec
o
g
n
itio
n
in
t
h
e
S
cra
mb
led
Do
m
MK
-
R
DA
a
n
d
A
N
N
(
K
a
vita
K
a
d
a
m
)
23
6.
CO
NCLU
SI
O
N
W
e
h
av
e
g
o
t
lo
ca
ted
th
e
is
s
u
es
r
elate
d
to
th
e
s
cr
a
m
b
led
p
i
ctu
r
es
f
o
r
th
e
d
u
r
atio
n
o
f
t
h
e
b
io
m
etr
ic
h
ea
li
n
g
o
f
p
ict
u
r
es.
Fo
r
t
h
at
w
e
ad
v
an
ce
d
a
b
r
an
d
n
e
w
tech
n
iq
u
e
–
Ma
n
y
-
Ker
n
el
R
an
d
o
m
Di
s
cr
i
m
i
n
a
n
t
a
n
al
y
s
i
s
(
MK
-
R
D
A
)
f
o
r
s
cr
a
m
b
led
f
a
ce
r
ep
u
tatio
n
.
W
e
ex
te
n
s
i
v
el
y
u
t
ilized
a
s
alie
n
ce
-
co
n
s
cio
u
s
f
ac
e
r
ec
o
g
n
itio
n
s
ch
e
m
e
w
h
ic
h
ca
n
w
o
r
k
w
it
h
c
h
ao
tic
p
atter
n
s
i
n
t
h
e
s
cr
a
m
b
le
d
d
o
m
ai
n
.
W
h
ich
ex
tr
ac
t
u
n
iq
u
e
p
atter
n
an
d
u
s
ed
to
m
o
r
e
t
h
an
o
n
e
k
er
n
el
to
f
i
n
d
s
i
m
ilar
r
an
k
i
n
g
o
f
d
atasets
f
ac
ial
p
ictu
r
es
RE
F
E
R
E
NC
E
S
[1
]
Rich
a
rd
Jia
n
g
,
S
o
m
a
y
a
A
l
-
M
a
a
d
e
e
d
,
A
h
m
e
d
Bo
u
ri
d
a
n
e
,
Da
n
n
y
Cro
o
k
e
s.
“
F
a
c
e
Re
c
o
g
n
it
io
n
in
th
e
S
c
ra
m
b
led
Do
m
a
in
v
ia S
a
li
e
n
c
e
-
Aw
a
r
e
En
se
m
b
les
o
f
M
a
n
y
Ke
rn
e
ls
”.
IEE
E
T
ra
n
s o
f
In
f
o
rm
a
ti
o
n
v
o
l
1
1
,
n
o
p
8
,
a
u
g
2
0
1
6
.
[2
]
P
e
ra
k
is,
P
.
;
P
a
ss
a
li
s
,
G
.;
T
h
e
o
h
a
ris,
T
.;
Ka
k
a
d
iaris,
I.
A
.
"
3
D
F
a
c
ial
L
a
n
d
m
a
rk
De
t
e
c
ti
o
n
u
n
d
e
r
L
a
rg
e
Ya
w
a
n
d
Ex
p
re
ss
io
n
V
a
riati
o
n
s"
,
IEE
E
T
ra
n
s
P
a
tt
e
rn
A
n
a
lys
is
a
n
d
M
a
c
h
in
e
In
telli
g
e
n
c
e
,
2
0
1
3
;
35
(
7
)
:
1
5
5
2
–
1
5
6
4
.
[3
]
T
a
h
e
ri,
S
.
,
P
a
tel,
V
.
M
.
,
Ch
e
ll
a
p
p
a
,
R.
"
Co
m
p
o
n
e
n
t
-
Ba
se
d
Re
c
o
g
n
it
i
o
n
o
f
F
a
c
e
s
a
n
d
F
a
c
ial
Ex
p
re
ss
io
n
s"
,
IEE
E
T
ra
n
s
Af
fec
ti
v
e
Co
mp
u
t
in
g
,
2
0
1
3
;
4
(4
)
:
3
6
0
-
3
7
1
.
[4
]
S
in
g
h
,
A
.
;
Ka
ra
n
a
m
,
S
.;
K
u
m
a
r,
D.
"
Co
n
str
u
c
ti
v
e
L
e
a
rn
in
g
f
o
r
Hu
m
a
n
-
Ro
b
o
t
In
tera
c
ti
o
n
"
,
IE
EE
Po
ten
ti
a
ls
,
2
0
1
3
;
32(
4
)
:
1
3
–
1
9
.
[5
]
M
c
Du
ff
,
D.
Ka
li
o
u
b
y
,
R.
E.
P
ica
rd
,
R.
W
.
"
Cro
w
d
so
u
rc
in
g
F
a
c
ial
Re
sp
o
n
se
s
t
o
O
n
li
n
e
Vid
e
o
s.”
IEE
E
T
ra
n
s
Af
fec
ti
v
e
Co
mp
u
t
in
g
,
2
0
1
2
;
3
(
4
)
:
4
5
6
–
4
6
8
.
[6
]
Y.
Ra
h
u
lam
a
th
a
v
a
n
,
R.
C.
-
W
.
P
h
a
n
,
J.
A
.
Ch
a
m
b
e
rs,
D.
J.
P
a
rish
.
“
F
a
c
ial
Ex
p
re
ss
io
n
Re
c
o
g
n
it
io
n
i
n
th
e
En
c
ry
p
ted
Do
m
a
in
Ba
se
d
o
n
L
o
c
a
l
F
ish
e
r
D
isc
rim
in
a
n
t
A
n
a
l
y
sis”
,
IEE
E
T
ra
n
.
Af
fec
ti
v
e
Co
mp
u
ti
n
g
,
2
0
1
3
;
4(
1
)
:
83
-
9
2
.
[7
]
P
i
n
g
L
iu
,
S
h
izh
o
n
g
Ha
n
,
Zi
b
o
M
e
n
g
,
Ya
n
T
o
n
g
.
"
Fa
c
i
a
l
Exp
re
ss
io
n
Rec
o
g
n
it
i
o
n
v
ia
a
B
o
o
ste
d
De
e
p
B
e
li
e
f
Ne
two
rk
"
,
CVP
R
2
0
1
4
.
[8
]
Y.
Ra
h
u
lam
a
th
a
v
a
n
,
R.
C.
-
W
.
P
h
a
n
,
J.
A
.
Ch
a
m
b
e
rs,
D.
J.
P
a
rish
.
“
F
a
c
ial
Ex
p
re
ss
io
n
Re
c
o
g
n
it
io
n
i
n
th
e
En
c
ry
p
ted
Do
m
a
in
Ba
se
d
o
n
L
o
c
a
l
F
ish
e
r
D
isc
rim
in
a
n
t
A
n
a
l
y
sis”
,
IEE
E
T
ra
n
.
Af
fec
ti
v
e
Co
mp
u
ti
n
g
,
2
0
1
3
;
4
(
1
)
,
8
3
-
9
2
.
[9
]
T
.
Ho
n
d
a
,
Y.
M
u
ra
k
a
m
i,
Y.
Y
a
n
a
g
ih
a
ra
,
T
.
Ku
m
a
k
i,
T
.
F
u
ji
n
o
.
“
Hie
ra
rc
h
ica
l
ima
g
e
-
sc
ra
mb
li
n
g
me
th
o
d
wi
t
h
sc
ra
mb
le
-
lev
e
l
c
o
n
tro
l
la
b
il
it
y
fo
r
p
riv
a
c
y
p
r
o
tec
ti
o
n
,
”
P
r
o
c
.
I
EE
E
5
6
th
In
ter
n
a
ti
o
n
a
l
M
id
w
e
st
S
y
m
p
o
siu
m
o
n
C
ircu
it
s
a
n
d
S
y
ste
m
s (M
W
S
C
A
S
),
2
0
1
3
,
p
p
.
1
3
7
1
-
1
3
7
4
.
[1
0
]
L
in
Y
Y,
L
iu
T
L
,
F
u
h
C
S
.
“
M
u
lt
ip
le
k
e
rn
e
l
lea
rn
in
g
f
o
r
d
im
e
n
si
o
n
a
li
ty
re
d
u
c
ti
o
n
.
IEE
E
T
r
a
n
sa
c
t
io
n
s
o
n
P
a
tt
e
r
n
An
a
lys
is
a
n
d
M
a
c
h
i
n
e
In
tell
ig
e
n
c
e
,
2
0
1
1
,
3
3
(6
):
1
1
4
7
-
1
1
6
0
.
[1
1
]
Z.
Erk
in
,
M
.
F
ra
n
z
,
J.
G
u
a
jard
o
,
S
.
Ka
tze
n
b
e
isse
r,
I.
L
a
g
e
n
d
ij
k
,
T.
T
o
f
t,
“
Priva
c
y
-
Pre
se
rv
in
g
Fa
c
e
Rec
o
g
n
it
io
n
,
”
P
r
o
c
.
Ni
n
th
I
n
t’l
S
y
m
p
.
P
riv
a
c
y
E
n
h
a
n
c
i
n
g
T
e
c
h
n
o
lo
g
ies
(
P
ET
S
’0
9
),
2
0
0
9
,
p
p
.
2
3
5
-
2
5
3
.
[1
2
]
Ja
y
a
ti
la
k
e
,
D.;
Ise
z
a
k
i
,
T
.;
T
e
r
a
m
o
to
,
Y
.;
Eg
u
c
h
i,
K
.;
S
u
z
u
k
i,
K.
"
Ro
b
o
t
A
ss
isted
P
h
y
sio
th
e
ra
p
y
to
S
u
p
p
o
r
t
Re
h
a
b
il
it
a
ti
o
n
o
f
F
a
c
ial
P
a
ra
ly
si
s"
,
IEE
E
T
r
a
n
s
Ne
u
ra
l
S
y
ste
ms
a
n
d
Reh
a
b
il
it
a
t
io
n
E
n
g
i
n
e
e
rin
g
,
2
0
1
4
;
2
2
(
3
)
:
6
4
4
-
6
5
3
.
[1
3
]
J.
W
rig
h
t,
A
.
Y
a
n
g
,
A
.
Ga
n
e
sh
,
S
.
S
a
str
y
,
Y.
M
a
.
“
Ro
b
u
st
F
a
c
e
Re
c
o
g
n
it
io
n
v
ia
S
p
a
rse
Re
p
re
se
n
tatio
n
,
”
IEE
E
T
ra
n
s
.
Pa
tt
e
rn
A
n
a
lys
is a
n
d
M
a
c
h
i
n
e
In
t
e
ll
ig
e
n
c
e
,
2
0
0
9
;
3
1
(
2
)
:
2
1
0
-
2
2
7
.
A
lso
se
e
CV
P
R
2
0
1
4
.
[1
4
]
A
.
M
e
ll
e
,
J.
-
L
.
Du
g
e
la
y
.
“
S
c
ra
m
b
li
n
g
f
a
c
e
s
fo
r
p
riv
a
c
y
p
ro
tec
ti
o
n
u
si
n
g
b
a
c
k
g
ro
u
n
d
se
lf
-
simila
rit
ies
,
”
P
r
o
c
.
2
0
1
4
IEE
E
In
tern
a
ti
o
n
a
l
Co
n
f
e
re
n
c
e
o
n
I
m
a
g
e
P
ro
c
e
ss
in
g
(ICI
P
),
2
0
1
4
:
6
0
4
6
-
6
0
5
0
.
[1
5
]
F
lec
k
,
S
.
;
S
tras
se
r,
W
.
"
S
ma
rt
Ca
me
ra
Ba
se
d
M
o
n
it
o
ri
n
g
S
y
ste
m
a
n
d
Its
A
p
p
li
c
a
ti
o
n
to
Assiste
d
L
ivi
n
g
"
,
P
r
o
c
e
e
d
in
g
s
o
f
th
e
IEE
E,
2
0
0
8
;
9
6
(
10
)
Oc
t:
1
6
9
8
–
1
7
1
4
.
[1
6
]
Y.
W
a
n
g
,
T
.
L
i,
“
S
tu
d
y
o
n
Ima
g
e
En
c
ry
p
ti
o
n
Al
g
o
rith
m
B
a
se
d
o
n
Arn
o
ld
T
ra
n
sf
o
rm
a
ti
o
n
a
n
d
Ch
a
o
ti
c
S
y
ste
m
.
”
P
ro
c
.
2
0
1
0
I
n
tern
a
ti
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
In
telli
g
e
n
t
S
y
ste
m
De
sig
n
&
En
g
in
e
e
rin
g
A
p
p
li
c
a
ti
o
n
,
2
0
1
0
:
4
4
9
-
4
5
1
.
Evaluation Warning : The document was created with Spire.PDF for Python.