Co
m
pu
t
er
Science
a
nd
I
nfo
r
m
a
t
io
n T
ec
hn
o
lo
g
ies
Vo
l.
7
,
No
.
2
,
J
u
ly
20
26
,
p
p
.
2
0
3
~
2
1
3
I
SS
N:
2722
-
3
2
2
1
,
DOI
:
1
0
.
1
1
5
9
1
/csi
t
.
v
7
i
2
.
p
2
0
3
-
2
1
3
203
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ia
e
s
p
r
ime.
co
m/in
d
ex
.
p
h
p
/csi
t
Tra
cking
a
perso
n and
de
terminin
g
t
he loca
tion by
using
c
o
nv
o
lutiona
l
n
eura
l
n
etwork
tec
hn
o
lo
g
y
Z
ina
h
S
hik
er
M
a
k
k
i
,
Ahm
et
Z
eng
in
D
e
p
a
r
t
me
n
t
o
f
C
o
mp
u
t
e
r
E
n
g
i
n
e
e
r
i
n
g
,
S
a
k
a
r
y
a
U
n
i
v
e
r
s
i
t
y
,
S
a
k
a
r
y
a
,
Tü
r
k
i
y
e
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
No
v
1
4
,
2
0
2
5
R
ev
is
ed
Ap
r
1
2
,
2
0
2
6
Acc
ep
ted
Ma
y
1
6
,
2
0
2
6
Trac
k
in
g
in
d
iv
id
u
a
ls
in
re
a
l
-
wo
rld
e
n
v
ir
o
n
m
e
n
ts
re
q
u
ires
r
o
b
u
st,
n
o
n
-
in
tru
si
v
e
m
e
th
o
d
s
t
h
a
t
o
v
e
rc
o
m
e
th
e
li
m
it
a
ti
o
n
s
o
f
d
e
v
ice
-
b
a
se
d
sy
ste
m
s.
Th
is
stu
d
y
p
r
o
p
o
se
s
a
c
o
n
v
o
lu
t
i
o
n
a
l
n
e
u
ra
l
n
e
two
r
k
(CNN
)
-
d
riv
e
n
p
e
rso
n
-
trac
k
in
g
fra
m
e
wo
rk
th
a
t
i
d
e
n
ti
fie
s
targ
e
ted
in
d
iv
i
d
u
a
ls
d
irec
tl
y
fr
o
m
c
a
m
e
r
a
fe
e
d
s,
e
li
m
in
a
ti
n
g
th
e
n
e
e
d
f
o
r
w
e
a
ra
b
le
o
r
g
l
o
b
a
l
p
o
siti
o
n
i
n
g
sy
st
e
m
(G
P
S
)
d
e
v
ice
s
a
n
d
a
d
d
re
ss
in
g
a
m
a
jo
r
d
ra
wb
a
c
k
o
f
trad
it
io
n
a
l
trac
k
in
g
tec
h
n
o
l
o
g
ies
.
Th
e
s
y
ste
m
u
ti
l
ize
s
a
Ten
so
rF
l
o
w
-
train
e
d
CNN
m
o
d
e
l
th
a
t
c
a
n
d
e
tec
t,
re
c
o
g
n
ize
,
a
n
d
l
o
c
a
te
p
e
rso
n
s
o
f
in
tere
st
i
n
re
a
l
-
ti
m
e
,
e
v
e
n
u
n
d
e
r
v
a
ry
in
g
il
l
u
m
in
a
ti
o
n
c
o
n
d
it
i
o
n
s
.
Un
li
k
e
c
o
n
v
e
n
ti
o
n
a
l
a
p
p
ro
a
c
h
e
s,
o
u
r
m
e
th
o
d
in
teg
ra
tes
fa
c
ial
fe
a
t
u
re
e
x
trac
ti
o
n
wit
h
e
n
c
ry
p
te
d
id
e
n
ti
t
y
m
a
n
a
g
e
m
e
n
t,
e
n
a
b
li
n
g
se
c
u
re
m
u
lt
i
-
p
e
rso
n
d
e
tec
ti
o
n
a
n
d
ra
p
i
d
lo
c
a
ti
o
n
re
p
o
rti
n
g
.
Ex
p
e
rime
n
tal
re
su
lt
s
d
e
m
o
n
stra
te
a
9
2
%
a
c
c
u
ra
c
y
in
lo
w
-
li
g
h
t
se
tt
in
g
s
a
n
d
1
0
0
%
a
c
c
u
ra
c
y
u
n
d
e
r
n
o
rm
a
l
li
g
h
t
in
g
,
c
o
n
firmin
g
t
h
e
sy
ste
m
’
s
e
ffe
c
ti
v
e
n
e
ss
fo
r
se
c
u
rit
y
-
or
ien
te
d
a
p
p
li
c
a
ti
o
n
s.
Th
e
fi
n
d
in
g
s
h
i
g
h
li
g
h
t
t
h
e
n
o
v
e
l
ty
o
f
c
o
m
b
i
n
i
n
g
l
ig
h
twe
ig
h
t
CNN
a
rc
h
i
tec
tu
re
,
re
a
l
-
ti
m
e
fa
c
ial
re
c
o
g
n
it
i
o
n
,
a
n
d
h
a
sh
-
b
a
se
d
id
e
n
ti
ty
p
ro
tec
ti
o
n
wit
h
in
a
u
n
ifi
e
d
trac
k
in
g
p
ip
e
li
n
e
.
K
ey
w
o
r
d
s
:
C
o
n
v
o
lu
tio
n
al
n
eu
r
al
n
etwo
r
k
Dee
p
lear
n
in
g
f
o
r
s
u
r
v
eillan
ce
Facial
r
ec
o
g
n
itio
n
s
y
s
tem
s
I
d
en
tity
d
etec
tio
n
R
ea
l
-
tim
e
p
er
s
o
n
tr
ac
k
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
:
Ah
m
et
Z
en
g
in
Dep
ar
tm
en
t o
f
C
o
m
p
u
ter
E
n
g
i
n
ee
r
in
g
,
Sak
a
r
y
a
Un
i
v
er
s
ity
Sak
ar
y
a,
T
u
r
k
ey
E
m
ail:
az
en
g
in
@
s
ak
ar
y
a.
ed
u
.
tr
1.
I
NT
RO
D
UCT
I
O
N
I
t
is
s
im
p
le
to
u
s
e
th
is
tech
n
o
lo
g
y
to
m
o
n
ito
r
p
e
d
estrian
s
an
d
in
s
tan
tly
id
en
tif
y
v
is
ito
r
s
,
p
ar
ticu
lar
ly
in
ca
s
es
o
f
ass
au
lt
o
r
th
ef
t.
Ov
er
th
e
p
r
ev
io
u
s
y
ea
r
s
,
f
ac
i
al
r
ec
o
g
n
itio
n
in
v
ital
m
ea
s
u
r
em
en
ts
an
d
ev
en
ts
,
p
atter
n
r
ec
o
g
n
itio
n
,
an
d
s
ec
u
r
ity
m
o
n
ito
r
in
g
d
ev
ices
h
a
v
e
p
lay
ed
a
c
r
u
cial
r
o
le.
T
o
d
ay
,
it
is
co
m
p
licated
an
d
ch
allen
g
in
g
t
o
id
en
tify
th
e
p
e
r
s
o
n
an
d
p
r
o
p
er
ty
f
r
o
m
t
h
iev
es
an
d
r
ec
en
t
f
r
au
d
in
s
o
m
e
c
o
u
n
tr
ies.
T
h
e
d
aily
n
ews
in
clu
d
es
cr
im
es
s
u
ch
as
cr
ed
it
ca
r
d
f
r
a
u
d
,
c
o
m
p
u
te
r
p
en
etr
atio
n
a
n
d
d
ata
th
ef
t,
as
well
as
in
cr
ea
s
in
g
s
ec
u
r
ity
th
r
ea
ts
f
ac
ed
b
y
co
m
p
an
ies
an
d
g
o
v
er
n
m
e
n
ts
.
T
h
e
p
r
im
ar
y
ca
u
s
e
o
f
th
ese
s
ec
u
r
ity
b
r
ea
ch
es
is
th
at
au
th
en
ticatio
n
is
b
ased
o
n
a
b
a
s
is
,
n
o
t o
u
r
b
i
o
m
etr
ic
id
en
tity
[
1
]
.
C
o
n
d
u
cted
r
esear
ch
to
h
elp
r
ed
u
ce
th
e
s
p
r
ea
d
o
f
t
h
e
v
ir
u
s
,
an
d
m
a
n
y
ad
v
er
tis
in
g
ca
m
p
aig
n
s
h
av
e
id
en
tifie
d
th
e
v
ir
u
s
to
p
eo
p
le
wo
r
ld
wid
e
,
w
h
er
e
th
e
W
o
r
l
d
Hea
lth
Or
g
a
n
izatio
n
(
W
H
O)
p
u
b
lis
h
es
d
is
ea
s
e
p
r
ev
en
tio
n
g
u
id
elin
es,
o
n
e
o
f
th
e
n
ec
ess
ar
y
m
ea
s
u
r
es
to
p
r
ev
en
t
th
e
s
p
r
ea
d
o
f
d
is
ea
s
e
is
th
e
u
s
e
o
f
f
ac
e
m
ask
s
.
T
h
is
r
esear
ch
p
r
o
p
o
s
es
u
tili
zin
g
th
e
f
ac
e
r
ec
o
g
n
itio
n
f
ea
tu
r
e
to
p
r
ev
e
n
t
r
esp
o
n
s
ib
le
in
d
iv
id
u
als
f
r
o
m
v
io
latin
g
th
e
g
en
e
r
al
g
u
id
elin
es
f
o
r
p
r
ev
en
ti
n
g
th
e
s
p
r
ea
d
o
f
th
e
co
r
o
n
av
ir
u
s
.
T
h
e
s
tu
d
y
f
o
cu
s
ed
o
n
f
ac
e
m
ask
s
.
W
e
u
s
ed
an
alg
o
r
ith
m
f
o
r
f
ac
ial
an
aly
s
is
b
y
t
h
r
e
e
ess
en
tial
p
o
in
ts
(
ey
e,
n
o
s
e,
an
d
m
o
u
th
)
.
Per
s
o
n
tr
ac
k
in
g
tec
h
n
o
lo
g
ies
ar
e
p
ar
t
o
f
b
io
m
etr
ics,
an
d
o
t
h
er
e
x
am
p
les
o
f
th
is
f
ea
tu
r
e
a
r
e
f
i
n
g
er
p
r
i
n
t
s
ca
n
n
in
g
tech
n
o
lo
g
ies
an
d
ir
is
s
ca
n
n
in
g
s
y
s
tem
s
,
wh
ich
ar
e
em
er
g
in
g
f
r
o
m
ar
tific
ial
in
tellig
en
ce
[
2
]
.
Peo
p
le
’
s
tr
ac
k
in
g
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
7
2
2
-
3
2
2
1
C
o
m
p
u
t Sci
I
n
f
T
ec
h
n
o
l
,
Vo
l.
7
,
No
.
2
,
J
u
ly
20
26
:
203
-
2
1
3
204
s
y
s
tem
s
u
s
e
d
if
f
er
en
t
m
eth
o
d
s
d
ep
en
d
in
g
o
n
th
e
a
p
p
lica
tio
n
an
d
th
e
m
an
u
f
ac
tu
r
er
.
Ho
wev
er
,
it
u
s
u
ally
co
n
s
is
ts
o
f
s
ev
er
al
s
tag
es
th
a
t
r
ec
o
r
d
,
p
r
o
ce
s
s
,
an
aly
ze
,
a
n
d
m
atch
th
e
f
ac
e
t
o
a
d
ata
b
ase
o
f
im
ag
es.
T
o
id
en
tify
an
d
m
atch
f
ac
ial
im
a
g
e
tr
aits
,
th
ese
s
y
s
tem
s
u
s
e
c
o
m
p
u
ter
v
is
io
n
in
s
ig
h
ts
,
s
tatis
tical
an
aly
s
is
to
o
ls
,
an
d
m
ac
h
i
n
e
lear
n
in
g
[
3
]
,
[
4
]
.
Desp
ite
th
e
wid
e
ad
o
p
tio
n
o
f
f
ac
ial
r
ec
o
g
n
itio
n
tech
n
o
lo
g
i
es,
m
o
s
t
ex
is
t
in
g
tr
ac
k
in
g
s
y
s
tem
s
s
till
s
u
f
f
er
f
r
o
m
c
r
itical
g
ap
s
:
i
)
th
ey
d
ep
en
d
o
n
GPS
o
r
wea
r
ab
l
e
d
ev
ices,
wh
ich
ca
n
n
o
t
b
e
p
l
ac
ed
o
n
in
d
iv
i
d
u
als
wh
o
ar
e
u
n
willin
g
,
u
n
k
n
o
w
n
,
o
r
attem
p
tin
g
to
ev
ad
e
m
o
n
ito
r
in
g
;
ii
)
m
an
y
class
ical
im
ag
e
-
p
r
o
ce
s
s
in
g
ap
p
r
o
ac
h
es
f
ail
u
n
d
er
p
o
o
r
lig
h
tin
g
,
o
cc
lu
s
io
n
,
a
n
d
r
ea
l
-
wo
r
ld
m
o
tio
n
;
an
d
iii
)
s
y
s
tem
s
b
ased
o
n
h
an
d
cr
a
f
ted
f
ea
tu
r
es
—
s
u
ch
as
p
r
in
cip
al
co
m
p
o
n
en
t
an
aly
s
is
(
PC
A)
,
lo
ca
l
b
in
ar
y
p
atter
n
s
(
L
B
P),
o
r
b
asic
s
u
p
p
o
r
t
v
ec
to
r
m
ac
h
in
e
(
SVM)
p
ip
elin
es
—
lack
r
o
b
u
s
tn
ess
an
d
r
ea
l
-
tim
e
ad
ap
ta
b
ilit
y
n
ee
d
ed
f
o
r
m
o
d
er
n
s
u
r
v
eillan
c
e
en
v
ir
o
n
m
en
ts
[
5
]
.
T
h
ese
lim
itatio
n
s
cr
ea
te
a
s
ig
n
if
ican
t
n
e
ed
f
o
r
a
tr
ac
k
in
g
m
eth
o
d
th
at
ca
n
au
to
m
atica
lly
ex
tr
ac
t
d
is
cr
im
in
ativ
e
f
ea
tu
r
e
s
f
r
o
m
r
aw
im
ag
es,
ad
ap
t
to
en
v
ir
o
n
m
en
tal
v
ar
iatio
n
s
,
an
d
p
er
f
o
r
m
ac
cu
r
at
e
r
ec
o
g
n
itio
n
with
o
u
t r
e
q
u
ir
in
g
an
y
p
h
y
s
ical
d
ev
ice
o
n
th
e
tar
g
et.
T
h
e
n
ee
d
f
o
r
Per
s
o
n
t
r
ac
k
in
g
b
ased
o
n
f
ac
e
r
ec
o
g
n
itio
n
te
ch
n
o
lo
g
y
h
as
in
cr
ea
s
ed
in
r
e
ce
n
t
y
ea
r
s
.
Ma
n
y
in
s
titu
tio
n
s
h
av
e
in
co
r
p
o
r
ated
it
in
to
th
e
ir
a
p
p
licatio
n
s
an
d
m
an
u
f
ac
tu
r
e
d
d
ev
ice
s
b
ec
au
s
e
it
s
av
es
h
u
m
an
s
a
s
ig
n
if
ican
t
am
o
u
n
t
o
f
ef
f
o
r
t
an
d
tim
e
in
m
atch
i
n
g
f
ac
es.
T
h
e
Vio
la
-
J
o
n
es
alg
o
r
ith
m
is
u
s
ed
to
d
etec
t
th
e
f
r
o
n
t.
T
h
e
m
o
d
if
ied
PC
A
alg
o
r
ith
m
is
u
s
ed
to
r
ec
o
g
n
ize
f
ac
es
f
r
o
m
im
a
g
es
with
s
o
m
e
d
if
f
er
en
ce
s
,
ac
h
iev
in
g
a
clo
s
e
-
to
-
r
ea
l
-
tim
e
r
ec
o
g
n
itio
n
r
ate.
T
h
e
al
g
o
r
ith
m
an
aly
ze
s
th
e
m
ain
c
o
m
p
o
n
e
n
ts
o
f
th
e
f
ac
e.
T
h
is
tech
n
iq
u
e
r
ec
o
g
n
izes f
ac
ial
p
ictu
r
es b
ef
o
r
e
an
d
a
f
ter
p
last
ic
s
u
r
g
er
y
[
6
]
.
T
o
o
v
er
co
m
e
th
e
lim
itatio
n
s
o
f
th
ese
ea
r
lier
ap
p
r
o
ac
h
es,
t
h
is
r
esear
ch
ad
o
p
ts
co
n
v
o
lu
ti
o
n
al
n
eu
r
al
n
etwo
r
k
s
(
C
NNs),
wh
ich
h
a
v
e
d
em
o
n
s
tr
ated
s
u
p
er
io
r
p
er
f
o
r
m
an
ce
i
n
ex
tr
ac
tin
g
h
ier
a
r
c
h
ical
an
d
in
v
ar
ian
t
f
ac
ial
f
ea
tu
r
es
d
ir
ec
tly
f
r
o
m
r
aw
im
ag
e
d
ata
[
7
]
.
C
NNs
ar
e
s
ig
n
if
ican
tly
m
o
r
e
r
esil
ien
t
to
lig
h
tin
g
v
ar
iatio
n
,
p
o
s
e
d
if
f
er
en
ce
s
,
o
cc
lu
s
io
n
,
a
n
d
im
ag
e
n
o
is
e
—
co
n
d
itio
n
s
th
at
f
r
eq
u
en
tly
d
eg
r
ad
e
th
e
ac
cu
r
ac
y
o
f
h
an
d
cr
af
ted
f
ea
tu
r
e
m
et
h
o
d
s
.
Fu
r
th
er
m
o
r
e
,
C
NN
-
b
ased
s
y
s
tem
s
en
ab
le
r
ea
l
-
tim
e
p
r
o
ce
s
s
in
g
,
s
ca
lab
ilit
y
,
an
d
im
p
r
o
v
ed
g
en
er
aliza
tio
n
,
m
ak
in
g
th
em
i
d
ea
l f
o
r
s
u
r
v
eill
an
ce
s
ce
n
ar
io
s
wh
er
e
s
p
ee
d
an
d
r
eliab
ilit
y
ar
e
ess
en
tial
[
8
]
,
[
9
]
.
T
h
e
p
r
im
ar
y
o
b
jectiv
e
o
f
th
is
r
esear
ch
is
to
tr
ac
k
in
d
iv
id
u
als
u
s
in
g
f
ac
ial
r
ec
o
g
n
itio
n
tech
n
iq
u
es
in
a
s
p
ec
if
ic
ar
ea
,
d
etec
t
v
io
lato
r
s
,
an
d
s
en
d
t
h
eir
lo
ca
ti
o
n
s
.
C
NN
alg
o
r
ith
m
s
wer
e
u
ti
liz
ed
,
with
s
o
m
e
m
o
d
if
icatio
n
s
,
to
s
u
it
an
d
v
er
if
y
th
e
p
r
o
ject
’
s
p
r
im
a
r
y
o
b
j
ec
tiv
e.
So
m
e
p
r
o
p
er
ties
h
a
v
e
b
ee
n
ad
d
ed
t
o
th
e
p
r
o
ject,
f
o
r
ex
am
p
le,
th
e
ab
ili
ty
to
d
is
co
v
er
m
o
r
e
th
an
o
n
e
p
er
s
o
n
s
im
u
ltan
eo
u
s
ly
.
R
ec
en
t
d
ev
elo
p
m
e
n
ts
in
tech
n
o
lo
g
y
an
d
i
n
d
u
s
tr
ial
r
ev
o
lu
tio
n
s
wo
r
ld
wid
e
h
a
v
e
b
ec
o
m
e
ess
en
tial
to
o
u
r
liv
es,
en
ab
lin
g
u
s
to
u
tili
ze
tech
n
o
lo
g
y
m
o
r
e
ef
f
ec
tiv
el
y
to
b
en
ef
it
f
r
o
m
it,
p
ar
ticu
lar
l
y
in
co
m
p
u
ter
v
is
io
n
an
d
ar
t
if
icial
in
tellig
en
ce
,
in
clu
d
in
g
m
ac
h
in
e
lear
n
in
g
an
d
d
ee
p
lear
n
in
g
[
1
0
]
.
T
h
e
f
ea
tu
r
es
o
f
tr
a
c
k
in
g
p
er
s
o
n
s
y
s
tem
s
an
d
f
ac
ial
r
ec
o
g
n
itio
n
i
n
o
u
r
tim
e
em
b
o
d
y
th
e
ess
en
tial
ch
ar
ac
ter
is
tics
th
at
m
u
s
t
b
e
r
elied
u
p
o
n
in
ad
d
r
ess
in
g
v
ar
io
u
s
is
s
u
es,
in
clu
d
in
g
h
ea
lth
,
s
ec
u
r
ity
,
an
d
ed
u
ca
tio
n
.
T
h
is
r
esea
r
ch
p
r
o
p
o
s
es
a
s
y
s
tem
to
m
o
n
ito
r
in
d
iv
id
u
als
in
v
io
lati
o
n
,
g
ath
er
in
f
o
r
m
atio
n
ab
o
u
t
t
h
em
,
a
n
d
tr
an
s
m
it
th
ei
r
lo
ca
tio
n
s
t
o
th
e
r
elev
a
n
t
au
t
h
o
r
ities
r
esp
o
n
s
ib
le
f
o
r
m
a
n
ag
in
g
th
e
s
y
s
tem
.
2.
M
AT
E
R
I
AL
A
ND
M
E
T
H
O
D
2
.
1
.
P
r
o
po
s
ed
s
y
s
t
em
a
rc
hit
ec
t
ure
Per
s
o
n
tr
ac
k
in
g
s
y
s
tem
s
ar
e
am
o
n
g
th
e
s
ig
n
if
ica
n
t
s
y
s
tem
s
f
o
r
id
en
tif
y
in
g
o
r
co
n
f
ir
m
in
g
a
p
er
s
o
n
n
atu
r
ally
f
r
o
m
a
c
o
m
p
u
ter
ize
d
im
ag
e.
T
h
er
e
ar
e
two
s
tr
ateg
ies
f
o
r
co
u
n
ter
in
g
r
ec
o
g
n
it
io
n
.
On
e
is
p
h
o
to
-
b
ased
,
an
d
th
e
o
th
er
is
v
id
eo
-
b
ased
.
T
h
er
e
ar
e
n
o
w
m
o
r
e
class
if
icatio
n
s
f
o
r
it.
I
n
o
u
r
ca
s
e,
a
p
er
s
o
n
tr
ac
k
in
g
s
y
s
tem
b
ased
o
n
f
ac
ial
r
ec
o
g
n
itio
n
tech
n
o
lo
g
y
is
a
s
till
im
ag
e
o
f
a
p
er
s
o
n
.
I
t
ca
n
v
e
r
if
y
an
d
id
en
tify
o
n
e
o
r
m
o
r
e
in
d
iv
id
u
als
b
y
r
ef
er
en
ci
n
g
a
s
to
r
ed
d
atab
ase
o
f
in
d
i
v
id
u
als
to
b
e
tr
ac
k
e
d
.
T
h
e
f
r
a
m
ewo
r
k
f
o
r
f
ac
ial
r
ec
o
g
n
itio
n
co
n
s
is
ts
o
f
s
ev
e
r
al
in
itial
s
tep
s
:
p
r
o
ce
s
s
in
g
th
e
in
f
o
r
m
atio
n
,
e
x
tr
ac
tin
g
k
ey
f
ea
tu
r
es,
an
d
class
if
icatio
n
.
I
n
f
o
r
m
atio
n
p
r
ep
r
o
ce
s
s
in
g
en
co
m
p
ass
es
v
ar
io
u
s
ac
tiv
ities
,
in
clu
d
in
g
s
tan
d
o
f
f
l
o
ca
tio
n
[
1
1
]
,
n
o
is
e
r
ed
u
ctio
n
,
im
ag
e
r
esizin
g
,
an
d
s
ca
lin
g
.
T
h
e
s
tr
u
ctu
r
e
o
f
a
p
er
s
o
n
t
r
ac
k
in
g
s
y
s
tem
b
ase
d
o
n
f
ac
ial
r
ec
o
g
n
itio
n
tech
n
o
lo
g
y
co
n
s
is
ts
o
f
th
r
ee
co
m
p
o
n
en
ts
:
i)
t
r
ain
in
g
a
C
NN
m
o
d
el
b
y
p
lacin
g
a
s
et
o
f
im
ag
es
o
f
th
e
p
eo
p
le
to
b
e
tr
ac
k
ed
;
ii)
i
m
ag
e
p
r
o
ce
s
s
in
g
b
y
r
e
d
u
cin
g
s
o
m
e
o
f
th
e
n
o
is
e
in
th
e
im
a
g
es
;
an
d
iii)
c
o
m
p
ar
in
g
th
e
im
ag
e
ta
k
en
b
y
th
e
ca
m
er
a
an
d
m
atch
in
g
it with
th
e
d
ata
b
ase
with
in
th
e
s
y
s
tem
.
T
h
e
f
u
n
ctio
n
s
o
f
ea
ch
p
ar
t
is
d
escr
ib
e
d
in
Fig
u
r
e
1
.
2
.
2
.
CNN
m
o
del t
ra
ini
ng
On
e
o
f
th
e
cr
itical
s
tep
s
in
o
u
r
wo
r
k
is
t
r
ain
in
g
a
m
o
d
el
u
s
in
g
ad
v
a
n
ce
d
m
ac
h
in
e
lear
n
in
g
tech
n
iq
u
es,
s
u
c
h
as
T
e
n
s
o
r
Flo
w.
I
t
is
an
o
p
en
-
s
o
u
r
ce
lib
r
ar
y
p
r
o
v
id
ed
b
y
Go
o
g
le
to
tr
ain
m
o
d
els.
T
h
e
tr
ain
e
d
m
o
d
el
ca
n
r
ec
o
g
n
ize
f
ac
es
b
y
m
atch
in
g
im
ag
es
s
to
r
ed
in
th
e
b
ase
with
v
id
eo
im
ag
es
ca
p
tu
r
ed
b
y
s
u
r
v
eillan
ce
ca
m
er
as a
n
d
id
e
n
tify
in
g
t
h
e
d
e
s
ir
ed
p
er
s
o
n
.
Evaluation Warning : The document was created with Spire.PDF for Python.
C
o
m
p
u
t Sci
I
n
f
T
ec
h
n
o
l
I
SS
N:
2722
-
3
2
2
1
Tr
a
ck
in
g
a
p
ers
o
n
a
n
d
d
etermin
in
g
th
e
lo
c
a
tio
n
b
y
u
s
in
g
c
o
n
vo
lu
tio
n
a
l n
eu
r
a
l n
etw
o
r
k
…
(
Zin
a
h
Ma
kk
i
)
205
Fig
u
r
e
1
.
Pe
r
s
o
n
tr
ac
k
in
g
s
y
s
tem
ar
ch
itectu
r
e
2
.
3
.
F
a
ce
ex
t
ra
ct
o
r
f
ro
m
ph
o
t
o
s
o
r
v
ideo
s
T
h
e
f
ac
ial
f
ea
tu
r
es
ex
tr
ac
to
r
e
x
tr
ac
ts
cr
itical
f
ac
ial
f
ea
tu
r
es.
W
e
ar
e
d
ev
elo
p
in
g
a
f
e
atu
r
e
ex
tr
ac
tio
n
m
eth
o
d
th
at
wo
r
k
s
o
n
c
o
lo
r
i
m
ag
es
with
o
u
t
r
eq
u
ir
in
g
th
e
u
s
e
o
f
g
lass
es.
T
h
e
f
ac
ial
f
ea
tu
r
es
ex
tr
ac
to
r
co
n
s
is
ts
o
f
th
r
ee
s
eq
u
en
tial
m
o
d
u
les:
f
ac
ial
f
ea
tu
r
es
lo
ca
lizatio
n
,
f
ea
tu
r
e
p
o
in
t
ex
tr
ac
tio
n
,
a
n
d
f
ac
ial
f
ea
tu
r
es
g
en
er
atio
n
.
First,
s
elec
t
th
e
m
ain
p
o
in
ts
-
ey
es,
n
o
s
e,
m
o
u
th
-
f
r
o
m
th
e
im
ag
e
o
f
th
e
f
ac
e.
Seco
n
d
,
th
e
f
ac
e
ex
tr
ac
ti
o
n
u
n
it e
x
tr
ac
ts
th
e
ch
a
r
ac
ter
is
tic
p
o
in
t a
n
d
a
n
aly
ze
s
its
co
m
p
atib
ilit
y
with
th
e
ex
is
tin
g
b
ase.
2
.
4
.
Cla
s
s
if
ica
t
io
n o
f
a
perso
n by
f
a
ce
Face
m
atch
in
g
in
cl
u
d
es
a
ca
t
eg
o
r
ized
a
n
d
d
id
ac
tic
alg
o
r
ith
m
.
T
h
is
c
o
m
p
o
n
en
t
is
r
esp
o
n
s
ib
le
f
o
r
id
en
tify
in
g
a
p
er
s
o
n
.
B
y
p
r
o
v
id
in
g
a
n
ew
p
ictu
r
e,
t
h
e
p
e
r
s
o
n
tr
ac
k
in
g
s
y
s
tem
f
ir
s
t
d
etec
ts
th
e
f
ac
e
an
d
th
en
p
o
s
itio
n
s
it
ac
co
r
d
in
g
ly
[
1
2
]
.
T
h
en
,
th
e
d
etec
ted
f
ac
e
im
a
g
e
is
p
ass
ed
to
th
e
f
ac
ial
f
ea
tu
r
es
ex
tr
ac
to
r
,
wh
er
e
th
e
f
ea
tu
r
es
a
r
e
ex
t
r
ac
ted
u
s
in
g
alg
o
r
ith
m
s
s
im
ilar
to
t
h
o
s
e
em
p
lo
y
ed
in
t
h
e
SVM
alg
o
r
ith
m
to
en
h
a
n
ce
th
e
im
ag
e
r
eso
lu
tio
n
.
T
h
e
im
a
g
e
will b
e
m
atch
ed
,
an
d
it will b
e
d
eter
m
in
ed
w
h
eth
er
it is
in
th
e
d
atab
ase.
2
.
5
.
Resea
rc
h
des
ig
n
W
h
en
r
esear
ch
er
s
wo
r
k
o
n
a
to
p
ic,
th
e
y
h
a
v
e
g
o
als
to
a
ch
iev
e
at
th
e
en
d
o
f
th
e
wo
r
k
.
I
n
th
is
p
r
o
p
o
s
ed
s
y
s
tem
,
th
e
p
r
im
ar
y
g
o
al
is
to
tr
ac
k
s
p
ec
if
ic
i
n
d
iv
i
d
u
als
th
r
o
u
g
h
th
e
ca
m
er
a,
id
e
n
tify
th
e
m
,
tr
a
n
s
m
it
th
eir
lo
ca
tio
n
to
th
e
d
esig
n
ated
p
er
s
o
n
with
in
th
e
s
y
s
tem
,
an
d
v
er
if
y
th
eir
id
en
titi
es.
At
th
e
o
u
ts
et
o
f
o
u
r
wo
r
k
,
we
co
n
d
u
cted
a
l
iter
atu
r
e
r
ev
iew
o
f
all
p
r
e
v
io
u
s
r
ese
ar
ch
in
th
e
f
ield
o
f
tr
ac
k
i
n
g
s
y
s
tem
s
to
d
eter
m
in
e
th
e
s
tr
en
g
th
s
an
d
wea
k
n
ess
e
s
o
f
t
h
e
alg
o
r
ith
m
s
u
s
ed
a
n
d
to
i
d
en
tify
th
e
tech
n
o
lo
g
y
t
h
at
b
est
m
ee
ts
t
h
e
s
y
s
tem
’
s
p
r
im
ar
y
p
u
r
p
o
s
e.
T
h
en
,
th
e
p
r
o
p
o
s
ed
m
eth
o
d
is
d
e
s
ig
n
ed
a
cc
o
r
d
in
g
t
o
th
e
d
ata
a
v
ailab
le
in
th
e
w
o
r
k
en
v
ir
o
n
m
en
t.
Fig
u
r
e
2
illu
s
tr
at
es th
e
m
eth
o
d
o
l
o
g
y
e
m
p
lo
y
e
d
in
th
is
wo
r
k
.
Fig
u
r
e
2
.
Me
th
o
d
o
lo
g
y
o
f
wo
r
k
2
.
6
.
P
r
o
po
s
ed
s
y
s
t
em
W
e
p
r
o
p
o
s
ed
a
s
y
s
tem
to
h
elp
g
o
v
er
n
m
e
n
ts
o
r
p
r
iv
ate
s
ec
u
r
ity
co
m
p
an
ies
tr
ac
k
s
p
ec
if
ic
i
n
d
iv
id
u
als
in
p
u
b
lic
o
r
p
r
iv
ate
p
lace
s
.
On
e
o
f
th
e
p
r
im
a
r
y
f
u
n
ctio
n
s
o
f
th
e
s
y
s
tem
is
to
f
ee
d
it
im
a
g
es
o
f
p
eo
p
le
to
b
e
tr
ac
k
ed
.
T
h
e
s
y
s
tem
also
f
ea
t
u
r
es
an
en
c
r
y
p
tio
n
m
ec
h
a
n
is
m
f
o
r
d
ata
p
r
o
tectio
n
,
u
tili
zin
g
a
h
ash
al
g
o
r
ith
m
to
en
cr
y
p
t
th
e
in
f
o
r
m
atio
n
en
ter
ed
in
to
th
e
s
y
s
tem
an
d
en
s
u
r
e
its
co
n
f
id
en
tiality
.
T
h
e
s
y
s
te
m
co
n
s
is
ts
o
f
th
r
ee
m
ain
p
ar
ts
: in
p
u
t,
p
r
o
ce
s
s
in
g
,
an
d
o
u
t
p
u
t,
as sh
o
wn
i
n
Fig
u
r
e
3
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
7
2
2
-
3
2
2
1
C
o
m
p
u
t Sci
I
n
f
T
ec
h
n
o
l
,
Vo
l.
7
,
No
.
2
,
J
u
ly
20
26
:
203
-
2
1
3
206
Fig
u
r
e
3
.
Pro
p
o
s
ed
s
y
s
tem
s
tep
s
2
.
6
.
1
.
T
he
m
a
in co
m
po
nents
o
f
t
he
pro
po
s
ed
s
y
s
t
em
Peo
p
le
tr
ac
k
in
g
is
a
co
m
p
licated
s
y
s
tem
.
Acc
o
r
d
in
g
to
o
u
r
ca
p
ab
ilit
ies,
we
d
e
s
ig
n
ed
th
e
p
r
o
p
o
s
ed
m
eth
o
d
u
s
in
g
a
c
o
m
p
u
ter
ca
m
er
a
to
r
ea
d
f
ac
es
an
d
s
tar
t
th
e
im
ag
e
an
aly
s
is
p
r
o
ce
s
s
.
Sy
s
te
m
co
m
p
o
n
en
ts
ar
e
d
iv
id
ed
in
to
two
m
ain
p
a
r
ts
:
th
e
p
h
y
s
ical
an
d
s
o
f
twar
e
p
ar
t
s
.
Fig
u
r
e
4
illu
s
tr
ates
th
e
p
r
im
ar
y
co
m
p
o
n
e
n
ts
o
f
th
e
s
y
s
tem
.
W
e
illu
s
tr
ated
th
e
m
ain
s
ch
em
e
o
f
th
e
s
y
s
tem
’
s
co
m
p
o
n
e
n
ts
an
d
d
iv
is
io
n
s
.
W
e
will
ex
p
lain
ea
ch
s
ec
tio
n
with
th
e
b
en
ef
it
o
f
u
s
i
n
g
it in
th
e
s
y
s
tem
.
Fig
u
r
e
4
.
C
o
m
p
o
n
en
ts
o
f
th
e
s
y
s
tem
a.
Har
d
war
e
We
r
ef
er
to
th
e
p
h
y
s
ical
co
m
p
o
n
en
ts
u
s
ed
in
th
e
s
y
s
tem
to
a
ch
iev
e
th
e
d
esire
d
r
esu
lt,
wh
ich
co
n
s
is
ts
o
f
two
p
ar
ts
.
−
C
o
m
p
u
ter
d
e
v
ice
T
h
e
co
m
p
u
ter
is
o
n
e
o
f
t
h
e
m
o
s
t
in
d
is
p
en
s
ab
le
u
n
its
o
f
t
h
e
s
y
s
tem
,
th
r
o
u
g
h
wh
ich
we
co
n
n
ec
t
th
e
ca
m
er
a
p
lace
d
in
a
s
p
ec
if
ic
l
o
ca
tio
n
to
th
e
p
r
o
g
r
am
m
ed
a
p
p
licatio
n
,
wh
ich
d
etec
ts
v
io
lato
r
s
an
d
id
en
tifie
s
th
em
.
Ad
d
itio
n
ally
,
t
h
e
co
m
p
u
ter
ca
n
b
e
d
escr
ib
ed
as
th
e
l
in
k
b
etwe
en
th
e
s
y
s
tem
’
s
h
ar
d
war
e
an
d
s
o
f
twar
e
co
m
p
o
n
en
ts
,
th
r
o
u
g
h
w
h
ich
all
h
ar
d
war
e
an
d
s
o
f
twar
e
co
m
p
o
n
e
n
ts
ar
e
i
n
teg
r
ated
,
in
cl
u
d
in
g
ap
p
licatio
n
s
,
ca
m
er
as,
an
d
d
atab
ases
.
−
C
am
er
a
Mo
s
t
o
f
th
e
s
ec
u
r
it
y
ca
m
er
as
av
ailab
le
o
n
th
e
m
ar
k
et
to
d
a
y
f
ea
tu
r
e
h
i
g
h
r
eso
lu
tio
n
,
as
th
e
ca
m
er
a
is
lin
k
ed
to
th
e
p
r
o
p
o
s
ed
s
y
s
tem
,
wh
ich
allo
ws
u
s
to
cr
ea
t
e
a
d
atab
ase
th
at
in
clu
d
es
th
e
in
d
iv
id
u
als
to
b
e
tr
ac
k
ed
.
W
h
en
t
h
e
ca
m
er
a
o
b
s
er
v
es
a
f
ac
e,
it
d
ete
r
m
in
es
wh
eth
er
it
is
in
s
id
e
a
p
er
s
o
n
in
t
h
e
d
atab
ase.
I
n
o
u
r
p
r
o
p
o
s
ed
s
y
s
tem
,
at
th
e
p
r
im
a
r
y
s
tag
e
o
f
w
o
r
k
,
we
e
v
alu
ated
a
ca
m
er
a
o
n
a
p
e
r
s
o
n
al
lap
to
p
.
b.
So
f
twar
e
T
h
e
co
d
e
a
n
d
lib
r
ar
ies
r
e
p
r
e
s
en
t
th
e
s
o
f
twar
e
co
m
p
o
n
e
n
ts
an
d
al
g
o
r
ith
m
s
c
o
m
b
in
e
d
t
o
f
o
r
m
a
n
in
teg
r
ated
ap
p
licatio
n
with
th
e
ca
m
er
a
to
tr
ac
k
th
e
p
er
s
o
n
’
s
s
y
s
tem
.
T
h
e
p
r
o
g
r
am
m
e
d
ap
p
licatio
n
,
wr
itten
in
th
e
Py
th
o
n
la
n
g
u
a
g
e,
is
a
co
m
b
in
atio
n
o
f
s
ev
er
al
lib
r
ar
ies an
d
alg
o
r
i
th
m
s
th
at
we
will d
is
c
u
s
s
in
d
etail.
Evaluation Warning : The document was created with Spire.PDF for Python.
C
o
m
p
u
t Sci
I
n
f
T
ec
h
n
o
l
I
SS
N:
2722
-
3
2
2
1
Tr
a
ck
in
g
a
p
ers
o
n
a
n
d
d
etermin
in
g
th
e
lo
c
a
tio
n
b
y
u
s
in
g
c
o
n
vo
lu
tio
n
a
l n
eu
r
a
l n
etw
o
r
k
…
(
Zin
a
h
Ma
kk
i
)
207
−
Op
en
C
V
T
h
e
o
p
e
n
co
m
p
u
ter
v
is
io
n
(
O
p
en
C
V)
lib
r
ar
y
is
a
co
llectio
n
o
f
p
r
o
g
r
a
m
m
in
g
i
n
ter
f
ac
es
d
esig
n
ed
to
f
ac
ilit
ate
th
e
d
e
v
elo
p
m
e
n
t
o
f
c
o
m
p
u
ter
v
is
io
n
ap
p
licatio
n
s
.
I
n
tel
co
r
p
o
r
atio
n
h
as
a
f
r
ee
s
o
f
twar
e
lib
r
ar
y
u
n
d
e
r
th
e
o
p
en
-
s
o
u
r
ce
licen
s
e.
W
e
ca
n
b
en
ef
it
f
r
o
m
th
e
Op
en
C
V
lib
r
ar
y
,
p
ar
ticu
lar
ly
in
m
ed
ical,
in
d
u
s
tr
ial,
an
d
ar
tific
ial
in
tellig
en
ce
ap
p
licatio
n
s
[
1
3
]
.
O
p
en
C
V
co
m
p
r
is
e
s
s
ev
er
al
alg
o
r
ith
m
s
an
d
s
o
f
t
war
e
co
d
es,
all
o
f
wh
ich
ar
e
co
m
p
iled
in
to
a
s
in
g
le
f
r
am
ewo
r
k
.
T
h
e
o
b
jectiv
e
o
f
th
is
f
r
am
ewo
r
k
is
to
p
r
o
v
id
e
s
o
lu
tio
n
s
in
co
m
p
u
ter
v
is
io
n
.
I
t c
a
n
b
e
u
s
ed
o
n
all
c
o
m
p
u
ter
s
y
s
tem
s
an
d
f
o
cu
s
es o
n
r
ea
l
-
tim
e
im
ag
e
p
r
o
ce
s
s
in
g
.
L
ib
r
ar
y
c
o
n
ten
ts
(
Op
en
C
V)
:
−
C
o
r
e:
a
b
u
ilt
-
in
m
o
d
u
le
f
o
r
ch
ar
ac
ter
izin
g
f
u
n
d
am
e
n
tal
d
ata
s
tr
u
ctu
r
es
−
I
m
g
Pro
c:
im
ag
e
p
r
o
ce
s
s
in
g
u
n
it
f
o
r
I
m
a
g
e
E
n
h
an
ce
m
en
t.
−
Vid
eo
:
th
e
v
id
e
o
an
aly
s
is
u
n
it
th
at
in
co
r
p
o
r
ates ta
r
g
et
m
o
v
e
m
en
t e
s
tim
ates.
−
C
alib
3
d
: th
e
2
D
im
ag
e
in
f
o
r
m
atio
n
an
aly
s
is
u
n
it.
−
Featu
r
es2
d
:
r
esp
o
n
s
ib
le
f
o
r
d
e
tectin
g
th
e
ta
r
g
et
’
s
k
n
o
wn
p
r
o
p
er
tie
s
an
d
s
ep
ar
atin
g
it
f
r
o
m
th
e
r
est
o
f
th
e
s
u
r
r
o
u
n
d
in
g
e
n
v
ir
o
n
m
en
t,
s
u
c
h
as g
eo
m
etr
ic
s
h
ap
es.
−
Ob
jects d
etec
t f
ac
e
d
etec
tio
n
−
Hig
h
-
u
p
: a
n
ea
s
y
-
to
-
u
s
e
in
ter
f
ac
e
f
o
r
tak
i
n
g
p
ictu
r
es a
n
d
v
id
eo
s
.
−
C
PU
m
o
d
u
le:
Utilize
s
th
e
to
tal
co
m
p
u
tin
g
p
o
wer
o
f
th
e
s
y
s
tem
b
y
u
s
in
g
th
e
p
o
wer
o
f
th
e
v
id
eo
p
r
o
ce
s
s
in
g
ca
r
d
.
Op
en
C
V
L
ib
r
ar
y
Ap
p
licatio
n
s
:
−
Facial
r
ec
o
g
n
itio
n
s
y
s
tem
s
−
Gestu
r
e
r
ec
o
g
n
itio
n
s
y
s
tem
s
−
C
o
m
p
u
ter
-
h
u
m
an
in
te
r
ac
tio
n
−
Mo
b
ile
An
d
r
o
i
d
−
Get
to
k
n
o
w
th
e
ta
r
g
et
−
Mo
tio
n
ch
asin
g
−
T
en
s
o
r
Flo
w:
Af
ter
co
m
p
letin
g
t
h
e
d
ata
e
n
tr
y
,
r
ep
r
o
ce
s
s
in
g
th
e
im
ag
e
s
,
an
d
p
r
e
p
ar
in
g
th
em
f
o
r
t
h
e
im
ag
e
p
r
o
ce
s
s
in
g
s
tag
e
i
n
o
u
r
p
r
o
p
o
s
ed
m
o
d
el,
th
r
ee
im
a
g
e
class
if
icatio
n
s
will
b
e
p
e
r
f
o
r
m
ed
b
ased
o
n
th
e
d
o
c
u
m
en
t
wh
er
e
th
e
m
o
d
el
is
tr
ain
ed
to
id
en
tify
th
ese
ca
s
es
an
d
p
r
o
v
id
e
ac
c
u
r
ate
r
esu
lts
with
th
e
h
elp
o
f
C
NN
tech
n
iq
u
es.
T
h
e
f
o
llo
win
g
m
e
asu
r
es we
r
e
tak
en
to
tr
ain
th
e
C
NN
m
o
d
el
.
Step
1
: U
p
lo
ad
d
ataset
Step
2
: T
h
e
in
p
u
t la
y
er
Step
3
: Co
n
v
o
lu
tio
n
al
lay
er
Fil
ter
th
e
f
ea
tu
r
e
m
ap
u
s
in
g
th
e
s
p
ec
if
ied
n
u
m
b
er
o
f
f
ilte
r
s
.
W
e
m
u
s
t
em
p
lo
y
a
r
ela
y
ac
tiv
atio
n
f
u
n
c
tio
n
af
ter
co
n
v
o
l
u
tio
n
to
ad
d
n
o
n
-
lin
ea
r
ity
to
th
e
n
et
wo
r
k
.
T
h
e
f
ir
s
t
co
n
v
o
lu
tio
n
al
lay
er
co
n
tain
s
1
8
f
ilter
s
,
ea
ch
with
a
7
×
7
k
er
n
el
an
d
e
q
u
al
p
a
d
d
in
g
.
T
h
e
o
u
tp
u
t
an
d
in
p
u
t
ten
s
o
r
s
h
a
v
e
t
h
e
s
am
e
wid
th
a
n
d
h
eig
h
t
with
th
e
s
am
e
p
ad
d
in
g
.
T
o
v
e
r
if
y
t
h
at
th
e
r
o
ws
an
d
c
o
lu
m
n
s
ar
e
th
e
ex
ac
t
s
izes,
th
e
to
p
o
lo
g
y
p
o
ten
tial
an
d
s
p
ec
tr
al
clu
s
ter
in
g
(
T
PS
C
)
m
o
d
el
will a
d
d
ze
r
o
s
to
th
e
m
.
Step
4
: Po
o
lin
g
lay
e
r
:
T
h
e
m
ax
im
u
m
f
ac
ilit
y
will
b
e
d
o
wn
s
am
p
led
f
o
llo
win
g
th
e
co
n
v
en
tio
n
.
T
h
e
g
o
al
is
to
lim
it
th
e
f
ea
tu
r
e
m
ap
’
s
m
o
b
ilit
y
to
p
r
e
v
en
t
o
v
e
r
f
itti
n
g
an
d
im
p
r
o
v
e
co
m
p
u
tatio
n
al
s
p
ee
d
.
T
h
e
tr
ad
itio
n
al
m
ax
p
o
o
lin
g
tech
n
iq
u
e
d
iv
id
es
f
ea
t
u
r
e
m
ap
s
in
to
s
u
b
f
ield
s
an
d
o
n
l
y
k
ee
p
s
th
e
m
ax
im
u
m
v
alu
e
s
.
Fo
llo
win
g
th
e
co
n
v
o
l
u
tio
n
al
s
tag
e,
th
e
p
o
o
l
in
g
co
m
p
u
tatio
n
is
th
e
n
ex
t
s
tag
e.
T
h
e
d
ata
will
b
e
c
o
m
p
r
ess
ed
d
u
e
to
th
e
p
o
o
lin
g
co
m
p
u
tatio
n
.
W
ith
a
d
im
en
s
io
n
o
f
3
×
3
an
d
a
s
tr
id
e
o
f
2
,
we
ca
n
u
tili
ze
th
e
m
ax
p
o
o
lin
g
2
D
m
o
d
u
le.
As in
p
u
t,
we
u
s
e
th
e
o
u
tp
u
t o
f
th
e
p
r
ev
io
u
s
lay
er
.
[
b
atch
s
ize,
1
4
,
an
d
1
5
]
ar
e
o
u
tp
u
t sizes.
Step
5
: Co
n
v
o
lu
tio
n
al
lay
er
a
n
d
p
o
o
li
n
g
lay
er
:
T
h
e
o
u
tp
u
t
s
ize
o
f
th
e
s
ec
o
n
d
C
NN
is
[
b
atch
s
ize,
1
4
,
1
4
,
3
2
]
,
an
d
it
h
as
p
r
ec
is
ely
3
2
f
i
lter
s
.
T
h
e
o
u
tp
u
t sh
ap
e
is
[
b
atch
s
ize,
1
4
,
1
4
,
1
8
]
,
an
d
t
h
e
s
ize
o
f
th
e
p
o
o
lin
g
lay
e
r
r
em
ain
s
u
n
ch
an
g
ed
.
Step
6
: D
en
s
e
lay
er
:
T
h
e
f
u
lly
c
o
n
n
ec
ted
lay
er
m
u
s
t
b
e
d
ef
in
ed
.
I
t
m
u
s
t
b
e
c
o
m
p
r
ess
ed
b
ef
o
r
e
co
m
b
i
n
in
g
i
t
with
th
e
f
ea
tu
r
e
m
a
p
a
n
d
t
h
e
d
e
n
s
e
lay
er
.
T
h
e
d
e
n
s
e
lay
er
will
co
n
n
e
ct
1
7
6
4
n
eu
r
o
n
s
.
W
e
ca
n
ad
d
a
R
eL
U
ac
tiv
atio
n
f
u
n
ctio
n
,
an
d
we
will
b
e
ab
l
e
to
d
o
s
o
.
W
e
a
d
d
a
0
.
3
d
r
o
p
o
u
t
r
eg
u
la
r
izatio
n
ter
m
,
m
ea
n
in
g
3
0
%
o
f
th
e
weig
h
ts
will b
e
0
.
On
ly
d
u
r
in
g
th
e
tr
ain
in
g
p
h
ase
d
o
p
eo
p
le
d
r
o
p
o
u
t.
Step
7
: L
o
g
it
lay
er
W
e
d
ef
in
e
th
e
f
in
al
lay
er
o
f
th
e
m
o
d
el
’
s
f
o
r
ec
ast.
T
h
e
o
u
tp
u
t
s
h
ap
e
is
b
atch
s
ize
1
2
(
th
e
to
tal
n
u
m
b
e
r
o
f
p
h
o
to
s
in
th
e
la
y
er
)
,
t
h
e
s
am
e
as th
e
to
tal
n
u
m
b
er
o
f
im
a
g
es in
th
e
lay
er
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
7
2
2
-
3
2
2
1
C
o
m
p
u
t Sci
I
n
f
T
ec
h
n
o
l
,
Vo
l.
7
,
No
.
2
,
J
u
ly
20
26
:
203
-
2
1
3
208
2
.
6
.
2
.
Used
a
lg
o
rit
hm
s
T
h
e
ca
m
er
a
-
b
ased
p
e
r
s
o
n
-
tr
a
ck
in
g
s
y
s
tem
is
p
r
o
g
r
am
m
e
d
th
r
o
u
g
h
co
m
p
lex
tech
n
o
l
o
g
ies,
ea
ch
p
er
f
o
r
m
in
g
a
d
if
f
er
e
n
t
f
u
n
ctio
n
.
Ho
wev
er
,
t
h
ey
in
teg
r
ate,
s
o
th
e
jo
b
is
n
o
t
d
o
n
e
r
ig
h
t
if
th
e
en
tire
tech
n
o
l
o
g
y
is
n
o
t
wo
r
k
in
g
co
r
r
ec
tly
.
Face
id
en
tific
atio
n
tech
n
o
lo
g
y
is
m
o
r
e
ac
cu
r
ate
an
d
co
m
p
lex
th
an
its
p
r
ed
ec
ess
o
r
s
.
L
ik
ewise,
f
o
r
r
ec
o
g
n
itio
n
tech
n
o
lo
g
y
,
a
p
e
r
s
o
n
’
s
p
h
o
to
is
s
l
o
wly
m
atch
e
d
a
g
ain
s
t
th
e
im
a
g
es
in
t
h
e
d
atab
ase.
Ho
wev
er
,
h
e
d
o
es
n
o
t
f
o
c
u
s
s
o
lely
o
n
th
e
s
h
a
p
e
an
d
f
ea
t
u
r
es
o
f
t
h
e
f
ac
e;
an
y
ch
a
n
g
e
i
n
f
ac
ial
e
x
p
r
ess
io
n
s
b
etwe
en
th
e
p
er
s
o
n
’
s
im
ag
e
an
d
th
e
im
ag
e
in
th
e
d
atab
as
e
in
v
alid
ates
h
is
wo
r
k
.
So
m
e
f
ac
ial
r
ec
o
g
n
itio
n
tech
n
o
lo
g
ies
id
en
tif
y
f
ac
ial
f
ea
tu
r
es
b
y
ex
tr
ac
ti
n
g
f
ea
tu
r
e
s
f
r
o
m
a
p
ictu
r
e
o
f
th
e
f
ac
e;
f
o
r
ex
a
m
p
le,
o
n
e
tech
n
o
lo
g
y
a
n
aly
ze
s
th
e
r
elativ
e
p
o
s
itio
n
an
d
s
ize
o
f
f
ac
ial
o
r
g
an
s
,
s
u
ch
as
t
h
e
n
o
s
e,
d
ete
r
m
in
ed
b
y
th
e
s
ize
an
d
lo
ca
tio
n
o
f
t
h
e
f
ac
e
as
w
ell
as
its
d
is
tan
ce
f
r
o
m
th
e
e
y
es,
an
d
it
d
e
f
in
es
th
e
g
a
p
b
e
twee
n
th
e
ey
es;
I
n
d
eter
m
in
in
g
s
p
ec
if
ic
tech
n
i
q
u
es
to
wo
r
k
,
th
e
r
esear
ch
er
m
u
s
t
d
o
s
er
io
u
s
w
o
r
k
c
o
m
p
a
r
in
g
th
e
m
eth
o
d
s
an
d
s
elec
tin
g
th
e
ap
p
r
o
p
r
iate
tech
n
o
lo
g
y
f
o
r
th
e
wo
r
k
.
I
n
th
is
p
r
o
p
o
s
ed
s
y
s
tem
,
we
d
ef
in
e
elem
en
tar
y
wo
r
k
i
n
g
tech
n
iq
u
es
(
C
NN)
to
id
en
tify
f
ac
es
b
y
tr
ai
n
in
g
a
T
en
s
o
r
Flo
w
m
o
d
el
a
n
d
u
tili
zin
g
a
h
ash
alg
o
r
ith
m
t
o
en
c
r
y
p
t
d
ata
with
in
th
e
s
y
s
tem
,
th
er
eb
y
en
h
a
n
cin
g
th
e
s
y
s
tem
’
s
s
ec
u
r
ity
.
3.
RE
L
AT
E
D
WO
RK
R
ea
l
-
tim
e
v
id
eo
s
u
r
v
eillan
ce
s
y
s
tem
s
ar
e
u
s
ed
in
v
a
r
io
u
s
s
ettin
g
s
,
in
clu
d
in
g
p
u
b
li
c
s
p
ac
es,
co
m
m
er
cial
b
u
ild
in
g
s
,
an
d
p
u
b
lic
in
f
r
astru
ctu
r
e.
Peo
p
le
’
s
d
etec
tio
n
is
cr
itical
to
m
an
y
v
id
eo
s
u
r
v
eillan
ce
s
y
s
tem
s
,
in
clu
d
in
g
p
eo
p
le
id
en
tific
atio
n
,
s
eg
m
en
tatio
n
,
an
d
tr
ac
k
in
g
.
T
o
r
ec
o
g
n
ize
an
d
tr
ac
k
p
eo
p
le,
r
esear
ch
er
s
h
av
e
u
s
ed
a
v
a
r
iety
o
f
im
ag
e
p
r
o
ce
s
s
in
g
an
d
AI
-
b
ased
tech
n
o
lo
g
ies
(
i
n
clu
d
in
g
m
ac
h
in
e
an
d
d
ee
p
lear
n
in
g
)
,
b
u
t th
e
y
m
o
s
tly
u
s
e
a
f
r
o
n
t
-
v
iew
ca
m
er
a
p
er
s
p
ec
tiv
e
[
1
4
]
−
[
1
6
]
.
Ah
m
ed
an
d
J
eo
n
[
1
7
]
d
em
o
n
s
tr
ated
a
m
eth
o
d
th
at
e
m
p
lo
y
s
an
o
v
e
r
h
ea
d
ca
m
er
a
v
iewp
o
in
t.
Siam
Ma
s
k
,
a
d
ee
p
lear
n
in
g
-
b
ased
tech
n
iq
u
e
em
p
lo
y
ed
i
n
th
e
s
y
s
tem
,
is
s
im
p
le,
ad
ap
tab
le,
an
d
f
ast,
o
u
tp
er
f
o
r
m
in
g
o
th
er
r
ea
l
-
tim
e
tr
ac
k
in
g
s
y
s
tem
s
.
T
h
e
m
eth
o
d
in
co
r
p
o
r
ates
th
e
m
ask
b
r
an
ch
in
to
th
e
f
u
ll
co
n
v
o
l
u
tio
n
al
d
u
p
lex
n
eu
r
al
n
etwo
r
k
to
tr
ac
k
th
e
tar
g
et
o
r
p
er
s
o
n
.
T
h
e
p
e
r
s
o
n
’
s
v
i
d
e
o
s
e
q
u
e
n
c
e
is
f
i
r
s
t
g
a
t
h
e
r
e
d
f
r
o
m
a
n
o
v
e
r
h
e
a
d
p
e
r
s
p
e
c
t
i
v
e;
t
h
e
n
,
a
d
d
i
t
i
o
n
a
l
t
r
a
i
n
in
g
i
s
c
o
n
d
u
c
t
e
d
t
h
r
o
u
g
h
l
e
a
r
n
in
g
t
r
a
n
s
f
e
r
.
F
i
n
a
ll
y
,
a
c
o
m
p
a
r
i
s
o
n
is
m
a
d
e
t
o
d
i
f
f
e
r
e
n
t
t
r
a
c
k
i
n
g
a
l
g
o
r
i
t
h
m
s
.
T
h
e
Sia
m
M
a
s
k
a
l
g
o
r
it
h
m
p
r
o
d
u
c
e
s
ex
c
e
l
l
e
n
t
r
es
u
l
ts
.
W
e
h
av
e
co
n
s
tr
u
cted
a
wo
r
k
ab
le
p
eo
p
le
m
o
n
ito
r
in
g
s
y
s
tem
th
at
s
o
lv
es
m
o
s
t
r
ea
l
-
wo
r
ld
co
n
ce
r
n
s
,
ac
co
r
d
in
g
to
J
o
h
n
Kr
u
m
m
.
Du
r
in
g
liv
e
d
em
o
s
in
th
e
liv
in
g
r
o
o
m
,
u
s
e
two
co
l
o
r
ca
m
e
r
as
to
tr
ac
k
v
ar
io
u
s
p
ar
ticip
an
ts
[
1
8
]
.
C
o
lo
r
p
h
o
to
s
p
r
o
tect
p
eo
p
le
’
s
id
en
titi
es,
wh
ile
s
ter
eo
im
ag
es
ar
e
u
s
ed
to
lo
ca
te
th
em
.
T
h
e
s
y
s
tem
is
f
ast en
o
u
g
h
to
g
iv
e
t
h
e
im
p
r
ess
io
n
th
at
th
e
r
o
o
m
is
r
esp
o
n
d
in
g
[
1
9
]
.
Acc
o
r
d
in
g
to
Gó
m
ez
-
Sil
v
a
[
1
9
]
,
h
e
to
o
k
ad
v
a
n
tag
e
o
f
t
h
e
tr
an
s
f
er
o
f
lear
n
i
n
g
f
r
o
m
a
m
u
lti
-
o
b
ject
tr
ac
k
in
g
(
MO
T
)
d
o
m
ai
n
o
f
two
p
h
o
to
g
r
ap
h
s
o
f
a
s
p
ec
if
i
c
p
er
s
o
n
to
b
e
id
e
n
tifie
d
an
d
tr
ac
k
ed
.
I
n
b
o
t
h
d
o
m
ain
s
,
h
e
tau
g
h
t
a
u
n
iq
u
e
d
ee
p
tr
ip
le
s
tr
u
ct
u
r
e.
Six
lev
el
s
o
f
tr
an
s
latio
n
al
lear
n
in
g
we
r
e
im
p
lem
en
ted
an
d
an
aly
ze
d
,
d
em
o
n
s
tr
atin
g
th
at
tr
an
s
f
er
r
in
g
k
n
o
wled
g
e
f
r
o
m
o
n
e
ar
ea
to
an
o
th
er
s
ig
n
if
ican
tly
im
p
r
o
v
es
re
-
id
en
tific
atio
n
p
er
f
o
r
m
a
n
ce
.
T
h
e
ex
p
er
im
en
tal
r
esu
lts
d
em
o
n
s
tr
ate
th
at
th
e
p
r
o
p
o
s
ed
m
et
h
o
d
is
co
m
p
ar
ab
le
to
ex
is
tin
g
s
tate
-
of
-
th
e
-
ar
t
p
r
o
ce
d
u
r
es
in
ter
m
s
o
f
ac
c
u
r
ac
y
an
d
r
o
b
u
s
tn
ess
.
T
h
ese
f
in
d
in
g
s
also
d
em
o
n
s
tr
ate
th
at,
d
esp
ite
th
e
d
ata
is
s
u
e,
d
e
ep
lear
n
in
g
ca
n
b
e
ef
f
ec
tiv
ely
u
tili
ze
d
f
o
r
s
in
g
le
-
s
h
o
t r
ed
ef
i
n
itio
n
.
W
eiy
an
g
ad
d
r
ess
es
th
e
p
r
o
b
l
em
o
f
d
ee
p
f
ac
ial
r
ec
o
g
n
itio
n
(
FR
)
u
n
d
e
r
th
e
O
p
en
Gr
o
u
p
Pro
to
co
l.
W
h
er
e
id
ea
l
f
ac
ial
f
ea
tu
r
es
ar
e
ex
p
ec
ted
to
h
av
e
a
m
ax
im
u
m
d
is
tan
ce
with
in
th
e
lay
e
r
,
e
q
u
al
to
th
e
m
in
im
u
m
d
is
tan
ce
b
etwe
en
ca
te
g
o
r
ies
u
n
d
er
a
s
u
itab
ly
s
elec
ted
m
e
asu
r
em
en
t
ar
ea
,
a
f
ew
e
x
is
tin
g
alg
o
r
ith
m
s
ca
n
ef
f
ec
tiv
ely
ac
h
iev
e
th
is
s
tan
d
ar
d
.
An
an
g
u
la
r
So
f
tMa
x
lo
s
s
(
A
-
So
f
tMa
x
)
en
ab
led
C
NNs
to
lear
n
an
g
u
lar
d
is
cr
im
in
an
t
f
ea
tu
r
es.
W
e
also
co
n
clu
d
e
a
s
p
ec
if
ic
ap
p
r
o
x
im
atio
n
o
f
th
e
id
ea
l
f
ea
tu
r
e
s
tan
d
ar
d
.
I
n
te
n
s
iv
e
an
aly
s
es
an
d
ex
p
er
im
e
n
ts
ar
e
co
n
d
u
cte
d
o
n
t
h
e
s
o
-
ca
lled
wild
in
th
e
wild
(
L
FW
)
an
d
f
ac
es
(
J
T
F).
As
a
r
esu
lt,
a
n
o
v
el
ap
p
r
o
ac
h
to
d
ee
p
r
ec
o
g
n
itio
n
in
f
ac
e
r
ec
o
g
n
itio
n
wa
s
in
tr
o
d
u
ce
d
.
Sp
ec
i
f
ically
,
it
is
b
en
ef
icial
to
lear
n
f
ac
ial
r
ep
r
esen
tatio
n
.
C
o
m
p
et
itiv
e
r
esu
lts
in
m
an
y
co
m
m
o
n
f
ac
ial
s
tan
d
ar
d
s
o
u
tweig
h
o
u
r
ap
p
r
o
ac
h
an
d
its
en
o
r
m
o
u
s
p
o
ten
tial
[
1
6
]
.
Mu
s
taf
a
u
tili
ze
d
k
er
n
el
d
is
cr
i
m
in
an
t
an
aly
s
is
(
KDA)
an
d
S
VM
f
ac
ial
r
ec
o
g
n
itio
n
al
g
o
r
it
h
m
s
,
wh
ich
ap
p
lied
k
er
n
el
an
aly
s
is
to
ex
tr
ac
t
f
ea
tu
r
es
f
r
o
m
th
e
in
p
u
t
im
ag
es.
Ad
d
itio
n
ally
,
th
is
p
r
o
ce
d
u
r
e
was
ap
p
lied
to
b
o
th
th
e
Yale
an
d
OR
L
d
ata
b
ases
to
ass
es
s
th
e
p
er
f
o
r
m
an
ce
o
f
th
e
p
r
o
p
o
s
ed
s
y
s
tem
.
E
x
p
er
im
en
tal
r
esu
lts
s
h
o
wed
th
at
th
e
s
y
s
tem
ac
h
ie
v
ed
a
h
i
g
h
r
ec
o
g
n
itio
n
r
ate,
with
an
ac
cu
r
ac
y
o
f
9
5
.
2
5
%
o
n
th
e
Yale
d
atab
ase
an
d
9
6
%
o
n
th
e
OR
L
[
1
6
]
.
Ku
m
ar
et
a
l.
[
2
0
]
h
as
p
r
o
p
o
s
ed
an
ar
r
an
g
em
e
n
t
f
o
r
m
o
s
t
p
r
o
b
lem
s
f
ac
in
g
th
e
d
az
e
with
in
th
e
ca
teg
o
r
y
o
f
e
x
p
lo
r
in
g
b
o
th
in
n
er
an
d
o
u
ter
s
itu
atio
n
s
,
co
m
p
r
i
s
in
g
d
if
f
e
r
en
t o
b
s
tacle
s
an
d
th
e
ac
k
n
o
wled
g
m
e
n
t
o
f
an
in
d
iv
i
d
u
al
b
ef
o
r
e
th
em
.
Face
s
ca
n
b
e
r
ec
o
g
n
ized
u
tili
zin
g
n
e
u
r
al
lear
n
in
g
m
eth
o
d
s
,
i
n
clu
d
in
g
ex
tr
ac
tio
n
an
d
p
r
ep
a
r
atio
n
m
o
d
u
les.
Pictu
r
es
o
f
co
m
p
a
n
io
n
s
an
d
r
e
lativ
es
ar
e
s
to
r
ed
with
in
th
e
s
m
ar
tp
h
o
n
e
clien
t
d
atab
a
s
e,
an
d
a
n
u
n
u
s
ed
p
i
ctu
r
e
r
ec
o
g
n
itio
n
a
n
d
n
a
v
ig
atio
n
s
y
s
tem
p
r
o
v
i
d
es
p
r
ec
is
e
an
d
q
u
ick
v
o
ice
m
ess
ag
es
to
in
d
iv
id
u
als
with
v
is
u
al
im
p
air
m
en
ts
,
en
ab
lin
g
th
em
to
n
av
i
g
ate
q
u
ick
l
y
.
T
h
e
f
r
am
ewo
r
k
a
p
p
lies
Evaluation Warning : The document was created with Spire.PDF for Python.
C
o
m
p
u
t Sci
I
n
f
T
ec
h
n
o
l
I
SS
N:
2722
-
3
2
2
1
Tr
a
ck
in
g
a
p
ers
o
n
a
n
d
d
etermin
in
g
th
e
lo
c
a
tio
n
b
y
u
s
in
g
c
o
n
vo
lu
tio
n
a
l n
eu
r
a
l n
etw
o
r
k
…
(
Zin
a
h
Ma
kk
i
)
209
to
o
p
e
n
-
air
a
n
d
i
n
d
o
o
r
s
itu
atio
n
s
[2
1
]
,
[
2
2
]
.
An
ex
am
in
ati
o
n
o
f
th
e
ex
ec
u
tio
n
o
f
th
e
ac
tu
alize
d
f
r
am
ewo
r
k
r
ev
ea
ls
th
at
7
5
%
o
f
th
e
p
ar
ticip
an
ts
f
in
d
th
is
f
r
am
ewo
r
k
h
e
lp
f
u
l
an
d
ac
h
ie
v
e
a
9
0
%
ac
c
u
r
ate
r
esu
lt
f
o
r
f
ac
e
r
ec
o
g
n
itio
n
[2
3
]
.
Acc
o
r
d
in
g
to
Fach
r
u
r
r
o
zi
et
a
l.
[2
4
]
,
th
e
r
ea
l
-
tim
e
f
ac
ial
r
e
co
g
n
itio
n
s
y
s
tem
p
r
o
ce
s
s
is
d
i
v
id
ed
in
to
th
r
ee
s
tep
s
:
f
ea
tu
r
e
ex
tr
ac
tio
n
,
clu
s
ter
in
g
,
d
etec
tio
n
,
an
d
r
ec
o
g
n
itio
n
.
E
ac
h
s
tag
e
u
s
es
a
d
if
f
er
en
t
m
eth
o
d
:
th
e
lo
ca
l
b
in
ar
y
p
atter
n
(
L
B
P),
cu
m
u
lativ
e
h
ier
ar
c
h
y
(
AHC),
an
d
E
u
clid
ea
n
d
is
tan
ce
.
C
B
I
R
,
an
im
ag
e
s
ea
r
ch
tech
n
o
lo
g
y
b
ased
o
n
im
ag
e
f
e
atu
r
es,
is
ap
p
lied
as
a
s
ea
r
c
h
m
eth
o
d
.
B
ased
o
n
ex
p
er
im
en
t
s
an
d
test
r
esu
lts
,
th
e
r
ec
all
an
d
ac
cu
r
ac
y
v
al
u
es we
r
e
6
5
.
3
2
% a
n
d
6
4
.
9
3
%,
r
esp
ec
t
iv
ely
.
T
ab
le
1
s
u
m
m
ar
izes th
e
r
elate
d
wo
r
k
.
T
ab
le
1
.
C
o
m
p
a
r
is
o
n
o
f
p
r
e
v
io
u
s
wo
r
k
s
Ti
t
l
e
Y
e
a
r
P
r
o
b
l
e
m
Te
c
h
n
i
q
u
e
u
se
d
R
e
s
u
l
t
“
A
r
e
a
l
-
t
i
me
p
e
r
s
o
n
t
r
a
c
k
i
n
g
sy
st
e
m
b
a
s
e
d
o
n
S
i
a
m
M
a
s
k
n
e
t
w
o
r
k
f
o
r
i
n
t
e
l
l
i
g
e
n
t
v
i
d
e
o
su
r
v
e
i
l
l
a
n
c
e
”
2
0
2
1
V
i
d
e
o
s
u
r
v
e
i
l
l
a
n
c
e
o
f
p
e
o
p
l
e
D
e
e
p
l
e
a
r
n
i
n
g
A
l
g
o
r
i
t
h
m
S
i
a
m
M
a
s
k
Th
e
a
l
g
o
r
i
t
h
m
u
s
e
d
g
a
v
e
e
x
c
e
l
l
e
n
t
r
e
s
u
l
t
s
c
o
m
p
a
r
e
d
t
o
o
t
h
e
r
m
e
t
h
o
d
s
“
M
u
l
t
i
-
c
a
mer
a
m
u
l
t
i
-
p
e
r
s
o
n
t
r
a
c
k
i
n
g
f
o
r
Ea
s
y
L
i
v
i
n
g
”
2
0
0
0
M
o
n
i
t
o
r
p
e
o
p
l
e
i
n
t
h
e
l
i
v
i
n
g
r
o
o
m
U
se
t
h
e
t
e
c
h
n
o
l
o
g
y
o
f
t
w
o
-
c
o
l
o
r
c
a
mer
a
s a
n
d
m
a
t
c
h
p
h
o
t
o
s
S
u
g
g
e
s
t
a
w
o
r
k
a
b
l
e
p
e
o
p
l
e
mo
n
i
t
o
r
i
n
g
sy
s
t
e
m
t
h
a
t
s
o
l
v
e
s
mo
st
r
e
a
l
-
w
o
r
l
d
c
o
n
c
e
r
n
s
.
“
T
r
a
n
s
f
e
r
r
i
n
g
l
e
a
r
n
i
n
g
f
r
o
m
m
u
l
t
i
-
p
e
r
s
o
n
t
r
a
c
k
i
n
g
t
o
p
e
r
s
o
n
re
-
i
d
e
n
t
i
f
i
c
a
t
i
o
n
”
2
0
1
9
Re
-
i
d
e
n
t
i
f
i
c
a
t
i
o
n
o
f
t
h
e
si
n
g
l
e
p
e
r
s
o
n
(re
-
i
d
)
Th
e
t
r
a
n
sf
e
r
o
f
l
e
a
r
n
i
n
g
f
r
o
m
mu
l
t
i
-
o
b
j
e
c
t
t
r
a
c
k
i
n
g
(
M
O
T)
Th
e
p
r
o
p
o
se
d
m
e
t
h
o
d
i
s
c
o
m
p
a
r
a
b
l
e
t
o
e
x
i
st
i
n
g
st
a
t
e
-
of
-
t
h
e
-
a
r
t
p
r
o
c
e
d
u
r
e
s
i
n
t
e
r
ms
o
f
a
c
c
u
r
a
c
y
a
n
d
r
o
b
u
s
t
n
e
ss
“
F
a
c
e
r
e
c
o
g
n
i
t
i
o
n
s
y
s
t
e
m
b
a
s
e
d
o
n
k
e
r
n
e
l
d
i
scri
m
i
n
a
n
t
a
n
a
l
y
si
s
,
k
-
n
e
a
r
e
s
t
n
e
i
g
h
b
o
r
a
n
d
su
p
p
o
r
t
v
e
c
t
o
r
m
a
c
h
i
n
e
”
2
0
1
8
F
a
c
e
R
e
c
o
g
n
i
t
i
o
n
S
V
M
a
n
d
K
N
N
A
c
c
o
r
d
i
n
g
t
o
t
h
e
e
x
p
e
r
i
m
e
n
t
a
l
r
e
su
l
t
s,
t
h
e
s
y
s
t
e
m
a
c
h
i
e
v
e
s
a
h
i
g
h
r
e
c
o
g
n
i
t
i
o
n
r
a
t
e
,
w
i
t
h
a
n
a
c
c
u
r
a
c
y
o
f
9
5
.
2
5
%
i
n
t
h
e
Y
a
l
e
d
a
t
a
b
a
se
a
n
d
9
6
%
o
n
t
h
e
O
R
L
d
a
t
a
b
a
se
.
“
I
n
t
e
l
l
i
g
e
n
t
f
a
c
e
r
e
c
o
g
n
i
t
i
o
n
a
n
d
n
a
v
i
g
a
t
i
o
n
s
y
s
t
e
m
u
s
i
n
g
n
e
u
r
a
l
-
l
e
a
r
n
i
n
g
f
o
r
s
mart
sec
u
r
i
t
y
i
n
t
h
e
i
n
t
e
r
n
e
t
o
f
t
h
i
n
g
s
”
2
0
1
7
S
y
st
e
m fo
r
b
l
i
n
d
p
e
o
p
l
e
G
e
n
e
t
i
c
a
l
g
o
r
i
t
h
m
a
n
d
R
B
F
K
e
r
n
e
l
Th
e
sy
s
t
e
m
i
s
h
e
l
p
f
u
l
a
n
d
p
r
o
v
i
d
e
s
a
9
0
%
a
c
c
u
r
a
t
e
r
e
s
u
l
t
f
o
r
f
a
c
e
r
e
c
o
g
n
i
t
i
o
n
.
“
M
u
l
t
i
-
o
b
j
e
c
t
f
a
c
e
r
e
c
o
g
n
i
t
i
o
n
u
si
n
g
c
o
n
t
e
n
t
B
a
s
e
d
i
ma
g
e
r
e
t
r
i
e
v
a
l
(
C
B
I
R
)
”
2
0
1
7
H
u
ma
n
f
a
c
e
r
e
c
o
g
n
i
t
i
o
n
i
n
a
mu
l
t
i
-
o
b
j
e
c
t
i
ma
g
e
,
LB
P
a
n
d
A
H
C
B
a
se
d
o
n
e
x
p
e
r
i
m
e
n
t
s
a
n
d
t
e
s
t
r
e
su
l
t
s,
t
h
e
r
e
c
a
l
l
a
n
d
a
c
c
u
r
a
c
y
v
a
l
u
e
s
w
e
r
e
6
5
.
3
2
%
a
n
d
6
4
.
9
3
%
,
r
e
sp
e
c
t
i
v
e
l
y
.
4.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
N
T
h
e
im
p
lem
en
tatio
n
p
h
ase
is
t
h
e
f
in
al
p
h
ase
th
at
d
eter
m
in
es
th
e
p
r
o
ject
’
s
s
u
cc
ess
.
I
t
i
s
n
o
t
p
o
s
s
ib
le
to
p
lan
t
ca
m
er
as
in
p
u
b
lic
p
lace
s
b
ec
au
s
e
th
is
r
eq
u
ir
es
a
s
ec
u
r
ity
ap
p
r
o
v
al
m
ac
h
in
e.
W
e
tan
g
ib
l
y
ex
p
er
im
en
ted
with
th
e
s
y
s
tem
b
y
p
r
o
v
i
d
in
g
it
with
an
al
b
u
m
o
f
p
h
o
to
s
tak
en
f
r
o
m
G
o
o
g
le
o
f
ce
leb
r
ities
,
alo
n
g
with
d
ata
id
en
tif
y
in
g
5
0
p
eo
p
le.
T
h
e
d
atab
ase
was su
p
p
lied
with
p
eo
p
le
’
s
in
f
o
r
m
atio
n
,
an
d
th
eir
im
a
g
es
wer
e
ad
d
ed
to
th
e
d
atab
ase.
W
e
r
an
th
e
p
r
o
g
r
am
u
s
in
g
t
h
e
lap
to
p
’
s
ca
m
er
a
an
d
p
ass
ed
p
ictu
r
es
in
f
r
o
n
t
o
f
it.
T
h
is
ex
p
er
im
en
t
was
co
n
d
u
c
ted
in
two
b
atch
es:
th
e
f
ir
s
t
u
s
in
g
d
im
lig
h
t
an
d
th
e
s
ec
o
n
d
u
n
d
e
r
s
tan
d
ar
d
v
iewin
g
co
n
d
itio
n
s
.
T
h
e
r
esu
lts
ar
e
r
ec
o
r
d
ed
as
f
o
llo
ws.
T
h
e
p
r
o
ce
s
s
was
co
m
p
leted
in
lo
w
lig
h
t
wh
en
5
0
ty
p
ical
im
ag
es
wer
e
r
ea
d
in
t
o
th
e
s
y
s
tem
.
T
h
e
ex
p
e
r
im
en
tal
p
h
ase
id
e
n
tifie
d
4
6
im
a
g
e
s
;
at
th
is
s
tag
e,
th
e
s
y
s
tem
ac
h
iev
ed
a
9
2
%
ac
c
u
r
ac
y
in
d
eter
m
i
n
in
g
th
e
wan
te
d
p
e
r
s
o
n
s
.
T
h
e
p
r
o
ce
s
s
was
c
o
n
d
u
cte
d
in
n
o
r
m
al
lig
h
t
with
th
e
ex
ac
t
m
ec
h
an
is
m
f
o
llo
wed
p
r
e
v
io
u
s
ly
,
an
d
5
0
p
h
o
to
s
o
f
p
eo
p
le
wer
e
id
e
n
t
if
ied
o
u
t
o
f
5
0
.
T
h
e
s
y
s
tem
ac
h
iev
ed
1
0
0
%
o
f
th
e
g
o
al
o
f
d
is
tin
g
u
is
h
in
g
th
e
wa
n
ted
p
e
o
p
le
to
b
e
tr
ac
k
ed
b
y
r
eg
u
lar
p
eo
p
le,
th
e
id
ea
l
r
atio
.
T
h
ese
r
esu
lts
in
d
icate
th
at
th
e
C
NN
-
b
ased
ap
p
r
o
ac
h
is
ca
p
ab
le
o
f
m
ain
tain
in
g
h
ig
h
ac
c
u
r
ac
y
ac
r
o
s
s
v
ar
y
in
g
illu
m
i
n
atio
n
c
o
n
d
itio
n
s
,
v
alid
atin
g
th
e
r
o
b
u
s
tn
ess
o
f
f
ea
tu
r
e
ex
tr
ac
tio
n
with
in
th
e
p
r
o
p
o
s
e
d
ar
ch
itectu
r
e.
I
n
p
ar
ticu
lar
,
th
e
m
in
im
al
p
er
f
o
r
m
an
ce
d
r
o
p
i
n
lo
w
lig
h
t
d
em
o
n
s
tr
ates
th
e
ad
v
an
tag
e
o
f
d
ee
p
f
ea
tu
r
e
r
ep
r
esen
tatio
n
o
v
er
tr
ad
itio
n
al
h
an
d
cr
a
f
ted
tech
n
iq
u
es,
wh
ich
ty
p
ically
d
eg
r
ad
e
s
h
ar
p
ly
u
n
d
e
r
th
ese
co
n
d
itio
n
s
.
T
h
e
ac
h
iev
ed
p
r
ec
is
io
n
,
r
ec
all
,
an
d
F1
-
s
co
r
es
as
s
h
o
wn
in
T
ab
le
2
f
u
r
th
er
s
u
p
p
o
r
t
th
e
r
e
liab
ilit
y
o
f
th
e
m
o
d
el,
s
h
o
win
g
th
at
m
is
c
lass
if
icatio
n
s
ar
e
r
ar
e
an
d
p
r
im
ar
ily
ass
o
ciate
d
with
lo
w
-
r
e
s
o
lu
tio
n
o
r
p
ar
tially
o
cc
lu
d
ed
test
im
ag
es.
I
m
p
o
r
ta
n
tly
,
th
e
s
y
s
tem
d
em
o
n
s
tr
ated
s
tr
o
n
g
s
tab
ilit
y
d
u
r
in
g
r
e
al
-
tim
e
ex
ec
u
tio
n
,
with
th
e
m
o
d
el
ca
p
a
b
le
o
f
i
d
en
tify
i
n
g
in
d
iv
i
d
u
als
with
in
f
r
ac
tio
n
s
o
f
a
s
ec
o
n
d
,
m
ak
in
g
it
s
u
ita
b
le
f
o
r
s
u
r
v
eillan
ce
an
d
r
ap
i
d
-
r
esp
o
n
s
e
s
ce
n
ar
io
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
7
2
2
-
3
2
2
1
C
o
m
p
u
t Sci
I
n
f
T
ec
h
n
o
l
,
Vo
l.
7
,
No
.
2
,
J
u
ly
20
26
:
203
-
2
1
3
210
T
ab
le
2
.
Acc
u
r
ac
y
o
f
th
e
r
esu
l
ts
P
r
e
c
i
s
i
o
n
R
e
c
a
l
l
F1
-
s
c
o
r
e
S
u
p
p
o
r
t
0
0
.
9
9
0
.
9
9
0
.
9
9
3
8
0
0
A
c
c
u
r
a
c
y
0
.
9
9
7
6
0
M
a
c
r
o
a
v
g
0
.
9
9
0
.
9
8
0
.
9
9
7
6
0
W
e
i
g
h
t
e
d
a
v
g
0
.
9
9
0
.
9
9
0
.
9
9
7
6
0
Use th
e
f
o
llo
win
g
lo
g
ic
to
ca
l
cu
late
th
e
p
er
ce
n
tag
e
ac
h
iev
e
d
.
Tr
a
ck
in
g
r
ate
=
ℎ
ℎ
ℎ
∗
100
B
ased
o
n
th
e
r
esu
lts
an
d
th
ei
r
co
m
p
ar
is
o
n
,
[
1
6
]
,
[
2
0
]
h
a
v
e
a
ch
iev
ed
r
esu
lts
th
at
ar
e
less
th
an
th
o
s
e
in
o
u
r
p
r
o
p
o
s
ed
s
y
s
tem
,
as
th
e
r
esear
ch
er
s
u
s
ed
d
if
f
e
r
en
t
a
lg
o
r
ith
m
s
th
an
th
e
tec
h
n
o
l
o
g
y
em
p
lo
y
e
d
in
o
u
r
s
y
s
tem
.
T
h
e
f
o
llo
win
g
d
ata
in
T
ab
le
2
s
h
o
ws
th
e
ac
cu
r
ac
y
o
f
th
e
r
esu
lts
o
b
tain
ed
f
r
o
m
th
e
s
y
s
tem
.
C
o
m
p
ar
ed
w
ith
p
r
io
r
s
tu
d
ies
u
s
in
g
PC
A,
L
B
P,
SVM,
an
d
h
y
b
r
id
C
B
I
R
ap
p
r
o
ac
h
es,
th
e
p
r
o
p
o
s
ed
m
o
d
el
o
u
tp
er
f
o
r
m
s
ea
r
lier
s
y
s
tem
s
in
b
o
th
ac
cu
r
ac
y
an
d
co
n
s
is
ten
cy
.
W
h
ile
p
r
io
r
m
eth
o
d
s
r
ep
o
r
t a
cc
u
r
ac
ies r
an
g
in
g
f
r
o
m
6
5
% to
9
6
%,
o
u
r
s
y
s
tem
co
n
s
is
ten
tly
ac
h
iev
es
ac
cu
r
ac
ies
o
f
9
2
%
–
1
0
0
%,
d
ep
en
d
in
g
o
n
lig
h
tin
g
co
n
d
itio
n
s
.
T
h
is
d
em
o
n
s
tr
ates
th
e
b
en
ef
it
o
f
d
e
ep
lear
n
in
g
–
b
ased
h
ier
ar
c
h
ica
l
f
ea
tu
r
e
e
n
co
d
i
n
g
i
n
en
v
ir
o
n
m
en
ts
wh
er
e
s
u
b
tle
v
ar
iatio
n
s
in
p
o
s
e,
ex
p
r
ess
io
n
,
o
r
illu
m
i
n
atio
n
ca
n
s
ig
n
if
ican
tly
r
ed
u
ce
th
e
p
er
f
o
r
m
an
ce
o
f
class
i
ca
l
r
ec
o
g
n
itio
n
alg
o
r
ith
m
s
.
4
.1
.
B
enchm
a
r
k
ing
a
g
a
ins
t
o
t
her
m
o
dels
T
o
f
u
r
th
er
v
alid
ate
th
e
e
f
f
ec
ti
v
en
ess
o
f
t
h
e
p
r
o
p
o
s
ed
C
NN
-
b
ased
tr
ac
k
in
g
s
y
s
tem
,
we
co
n
d
u
cted
a
co
n
ce
p
tu
al
a
n
d
p
e
r
f
o
r
m
an
c
e
-
o
r
ien
ted
b
en
ch
m
ar
k
in
g
co
m
p
ar
is
o
n
ag
ai
n
s
t
s
ev
er
al
wid
ely
u
s
ed
f
ac
ial
r
ec
o
g
n
itio
n
an
d
tr
ac
k
in
g
m
o
d
els,
in
clu
d
in
g
PC
A
-
b
ased
r
ec
o
g
n
itio
n
,
L
B
P
d
escr
ip
to
r
s
,
SVM
class
if
ier
s
,
Siam
Ma
s
k
d
ee
p
tr
ac
k
in
g
,
an
d
C
B
I
R
-
b
ased
m
u
lti
-
o
b
ject
r
ec
o
g
n
itio
n
s
y
s
tem
s
.
T
r
ad
itio
n
al
PC
A
an
d
L
B
P
ap
p
r
o
ac
h
es
d
em
o
n
s
tr
ated
s
tr
o
n
g
p
e
r
f
o
r
m
an
ce
o
n
ly
u
n
d
er
c
o
n
tr
o
lled
lig
h
tin
g
a
n
d
f
r
o
n
tal
-
f
a
ce
co
n
d
itio
n
s
,
with
ac
cu
r
ac
ies
ty
p
ically
r
an
g
in
g
f
r
o
m
6
5
%
to
9
0
%
ac
co
r
d
in
g
to
ea
r
lier
s
tu
d
ies.
Ho
wev
er
,
b
o
th
tech
n
iq
u
es
s
tr
u
g
g
led
i
n
s
ce
n
ar
i
o
s
in
v
o
lv
i
n
g
illu
m
in
atio
n
v
ar
iab
ilit
y
,
o
cc
lu
s
io
n
,
an
d
c
o
m
p
lex
b
ac
k
g
r
o
u
n
d
s
.
T
h
is
alig
n
s
with
o
u
r
f
in
d
in
g
s
,
wh
er
e
C
NN
-
b
ased
r
ec
o
g
n
itio
n
m
ain
tain
ed
h
ig
h
ac
cu
r
ac
y
(
9
2
%
–
1
0
0
%)
ev
en
u
n
d
e
r
lo
w
-
lig
h
t c
o
n
d
itio
n
s
,
u
n
lik
e
h
an
d
c
r
af
ted
f
ea
tu
r
e
m
o
d
els,
wh
ich
ten
d
to
d
eg
r
ad
e
s
ig
n
if
ican
tly
in
s
u
ch
s
itu
atio
n
s
.
Similar
ly
,
SVM
-
b
ased
p
ip
elin
e
m
o
d
els
—
a
lth
o
u
g
h
ef
f
ec
ti
v
e
f
o
r
s
m
all,
well
-
lab
eled
d
atasets
—
r
eq
u
ir
e
m
a
n
u
al
f
ea
tu
r
e
e
x
tr
ac
t
io
n
an
d
d
o
n
o
t
s
ca
le
well
to
lar
g
er
o
r
m
o
r
e
d
i
v
er
s
e
d
atasets
.
I
n
co
n
tr
ast,
C
NNs
in
h
er
en
tly
lear
n
m
u
lti
-
lev
el
f
ea
tu
r
e
h
ier
ar
ch
ies
f
r
o
m
r
aw
im
ag
es,
e
n
ab
lin
g
s
u
p
e
r
io
r
g
en
er
aliza
t
io
n
an
d
r
o
b
u
s
tn
ess
.
T
h
ese
ad
v
an
tag
es
ar
e
clea
r
ly
r
ef
lecte
d
in
th
e
h
i
g
h
er
p
r
ec
is
io
n
,
r
ec
all,
an
d
F1
-
s
co
r
es
o
b
tain
ed
in
o
u
r
ex
p
er
im
en
ts
.
W
h
en
b
en
c
h
m
ar
k
e
d
c
o
n
ce
p
tu
ally
ag
ain
s
t
s
tate
-
of
-
t
h
e
-
ar
t
d
e
ep
lear
n
in
g
m
eth
o
d
s
s
u
c
h
as
Siam
Ma
s
k
,
wh
ich
f
o
cu
s
es
o
n
r
ea
l
-
tim
e
o
b
ject
tr
ac
k
in
g
r
ath
er
th
an
i
d
e
n
tity
r
ec
o
g
n
itio
n
,
o
u
r
s
y
s
tem
p
r
o
v
id
es
an
a
d
d
ed
ad
v
an
tag
e
b
y
co
m
b
in
i
n
g
b
o
t
h
d
etec
tio
n
an
d
id
en
tity
v
er
if
ic
atio
n
.
Siam
Ma
s
k
ex
ce
ls
in
m
o
tio
n
tr
ac
k
in
g
b
u
t
d
o
es
n
o
t
in
h
er
en
tly
p
er
f
o
r
m
f
ac
ial
r
ec
o
g
n
itio
n
.
Ou
r
m
o
d
el
in
teg
r
ates
b
o
t
h
ca
p
a
b
ilit
ies,
m
ak
in
g
it
m
o
r
e
s
u
itab
le
f
o
r
s
ec
u
r
ity
a
p
p
licatio
n
s
r
eq
u
ir
in
g
id
en
tity
-
s
p
ec
if
ic
t
r
ac
k
in
g
r
ath
er
th
a
n
g
en
e
r
ic
o
b
j
ec
t lo
ca
lizatio
n
.
C
o
m
p
ar
ed
with
C
B
I
R
-
b
ased
m
u
lti
-
o
b
ject
r
ec
o
g
n
itio
n
te
ch
n
iq
u
es,
wh
ich
ac
h
iev
ed
r
ec
all
an
d
ac
cu
r
ac
y
v
alu
es
ar
o
u
n
d
6
5
%,
th
e
p
r
o
p
o
s
ed
C
NN
s
y
s
tem
d
em
o
n
s
tr
ates
a
s
ig
n
if
ican
t
p
er
f
o
r
m
a
n
ce
im
p
r
o
v
em
e
n
t.
T
h
e
d
ee
p
co
n
v
o
lu
tio
n
al
ar
ch
itectu
r
e
r
ed
u
ce
s
s
en
s
itiv
ity
to
n
o
is
e
an
d
b
ac
k
g
r
o
u
n
d
clu
tter
—
two
m
ajo
r
lim
itatio
n
s
o
f
C
B
I
R
ap
p
r
o
ac
h
es
th
at
r
ely
h
ea
v
ily
o
n
tex
t
u
r
e
s
im
ilar
ity
r
ath
er
t
h
an
lear
n
ed
f
ea
tu
r
e
ab
s
tr
ac
tio
n
.
Ov
er
all,
th
e
b
en
ch
m
ar
k
in
g
r
esu
lts
co
n
f
i
r
m
t
h
at
C
NN
-
b
ased
m
eth
o
d
s
n
o
t
o
n
ly
o
u
tp
er
f
o
r
m
class
ical
m
ac
h
in
e
lear
n
in
g
tec
h
n
iq
u
es
b
u
t
also
o
f
f
er
ad
v
a
n
tag
es
o
v
er
s
ev
er
al
m
o
d
er
n
d
ee
p
-
lear
n
in
g
tr
ac
k
er
s
b
y
in
te
g
r
atin
g
id
en
tity
r
ec
o
g
n
itio
n
,
m
u
lti
-
p
er
s
o
n
d
etec
tio
n
,
an
d
en
c
r
y
p
te
d
d
ata
h
a
n
d
lin
g
with
in
a
u
n
if
ied
,
p
r
ac
tical
f
r
am
ewo
r
k
.
4
.
2
.
L
im
it
a
t
io
ns
Desp
ite
p
r
o
m
is
in
g
r
esu
lts
,
th
e
p
r
o
p
o
s
ed
s
y
s
tem
h
as
s
ev
er
al
lim
itatio
n
s
th
at
m
u
s
t
b
e
ac
k
n
o
wled
g
ed
.
First,
th
e
m
o
d
el
was
tr
ain
e
d
an
d
e
v
alu
ate
d
u
s
in
g
a
r
elativ
ely
s
m
all
d
ataset,
wh
ich
m
ay
r
estrict
it
i
s
g
en
er
aliza
tio
n
to
b
r
o
ad
er
d
em
o
g
r
ap
h
ic
o
r
en
v
ir
o
n
m
en
tal
v
a
r
iatio
n
s
.
L
ar
g
er
an
d
m
o
r
e
d
iv
er
s
e
d
atasets
wo
u
ld
f
u
r
th
er
s
tr
en
g
th
en
t
h
e
r
eliab
ilit
y
o
f
th
e
s
y
s
tem
.
Seco
n
d
,
th
e
ex
p
er
im
en
ts
wer
e
co
n
d
u
cte
d
u
s
in
g
a
lap
t
o
p
ca
m
er
a
u
n
d
e
r
co
n
t
r
o
lled
in
d
o
o
r
co
n
d
itio
n
s
[2
5
]
.
R
ea
l
-
wo
r
ld
s
u
r
v
eillan
ce
en
v
ir
o
n
m
en
ts
o
f
ten
p
r
esen
t
ad
d
itio
n
al
ch
allen
g
es,
in
clu
d
i
n
g
ex
tr
em
e
lig
h
tin
g
co
n
d
itio
n
s
,
m
o
tio
n
b
lu
r
,
h
ig
h
cr
o
wd
d
en
s
ity
,
p
a
r
tial
o
cc
lu
s
io
n
s
,
an
d
v
a
r
y
in
g
ca
m
er
a
an
g
les
[2
6
]
.
Fu
tu
r
e
wo
r
k
Evaluation Warning : The document was created with Spire.PDF for Python.
C
o
m
p
u
t Sci
I
n
f
T
ec
h
n
o
l
I
SS
N:
2722
-
3
2
2
1
Tr
a
ck
in
g
a
p
ers
o
n
a
n
d
d
etermin
in
g
th
e
lo
c
a
tio
n
b
y
u
s
in
g
c
o
n
vo
lu
tio
n
a
l n
eu
r
a
l n
etw
o
r
k
…
(
Zin
a
h
Ma
kk
i
)
211
s
h
o
u
ld
in
clu
d
e
f
ield
d
e
p
lo
y
m
e
n
t
with
o
u
td
o
o
r
an
d
m
u
lti
-
ca
m
er
a
s
ettin
g
s
to
ass
ess
s
y
s
tem
p
er
f
o
r
m
an
ce
u
n
d
er
o
p
er
atio
n
al
c
o
n
s
tr
ain
ts
.
T
h
ir
d
,
th
e
cu
r
r
e
n
t
s
y
s
tem
r
elies
o
n
f
r
o
n
tal
o
r
n
ea
r
-
f
r
o
n
tal
f
ac
ial
v
is
ib
ilit
y
.
W
h
ile
C
NN
s
im
p
r
o
v
e
ro
b
u
s
tn
ess
,
ex
tr
em
e
s
id
e
p
r
o
f
i
les
o
r
h
ea
v
ily
o
cc
lu
d
e
d
f
ac
es
s
till
p
o
s
e
r
ec
o
g
n
itio
n
ch
allen
g
e
s
.
I
n
teg
r
atin
g
b
o
d
y
p
o
s
e
esti
m
atio
n
o
r
m
u
lti
-
m
o
d
al
b
io
m
etr
ics
co
u
ld
r
e
d
u
ce
th
is
d
ep
en
d
en
cy
.
Fin
ally
,
al
th
o
u
g
h
th
e
s
y
s
tem
in
co
r
p
o
r
ates
h
ash
-
b
ased
en
c
r
y
p
tio
n
f
o
r
d
ata
p
r
o
tectio
n
,
ad
d
itio
n
al
s
ec
u
r
ity
lay
er
s
—
s
u
ch
as
s
ec
u
r
e
co
m
m
u
n
icatio
n
tu
n
n
els,
d
if
f
er
en
tial
p
r
iv
ac
y
,
o
r
GDPR
-
co
m
p
lian
t
an
o
n
y
m
izatio
n
—
s
h
o
u
ld
b
e
co
n
s
id
er
e
d
f
o
r
d
ep
lo
y
m
e
n
t in
g
o
v
er
n
m
en
tal
o
r
co
m
m
er
cial
e
n
v
ir
o
n
m
en
ts
.
4
.
3
.
Nee
d f
o
r
la
rg
er
a
nd
m
o
re
div
er
s
e
da
t
a
s
e
ts
An
o
th
er
i
m
p
o
r
ta
n
t
lim
itatio
n
o
f
th
e
cu
r
r
en
t
s
tu
d
y
is
th
e
r
estricte
d
s
ize
an
d
d
iv
e
r
s
ity
o
f
t
h
e
d
ataset
u
s
ed
f
o
r
m
o
d
el
tr
ain
i
n
g
an
d
v
alid
atio
n
.
T
h
e
s
y
s
tem
was
ev
a
lu
ated
u
s
in
g
a
r
elativ
ely
s
m
all
n
u
m
b
e
r
o
f
im
a
g
es,
m
an
y
o
f
wh
ich
wer
e
s
o
u
r
ce
d
f
r
o
m
p
u
b
licly
a
v
ailab
le
ce
le
b
r
ity
p
h
o
to
s
with
lim
ited
v
ar
i
atio
n
in
p
o
s
e,
ag
e,
eth
n
icity
,
en
v
ir
o
n
m
en
tal
b
a
ck
g
r
o
u
n
d
,
an
d
ca
m
er
a
an
g
le.
T
h
is
lev
el
o
f
h
o
m
o
g
en
eity
m
a
y
n
o
t
f
u
lly
r
e
p
r
esen
t
r
ea
l
-
wo
r
ld
s
u
r
v
eillan
ce
co
n
d
itio
n
s
,
wh
er
e
in
d
i
v
id
u
als
d
if
f
er
s
ig
n
if
ican
tly
i
n
ap
p
ea
r
an
ce
an
d
a
r
e
o
f
te
n
ca
p
tu
r
ed
in
d
y
n
am
ic,
u
n
co
n
tr
o
lled
s
ettin
g
s
.
T
o
en
s
u
r
e
s
tr
o
n
g
er
g
en
er
aliza
t
io
n
an
d
r
ed
u
ce
t
h
e
r
is
k
o
f
o
v
e
r
f
itti
n
g
,
f
u
t
u
r
e
wo
r
k
s
h
o
u
l
d
in
co
r
p
o
r
ate
s
ig
n
if
ican
tly
lar
g
er
an
d
m
o
r
e
d
iv
er
s
e
d
atasets
th
at
in
clu
d
e
m
u
ltip
le
a
g
e
g
r
o
u
p
s
,
eth
n
icities
,
ex
p
r
ess
io
n
v
ar
iatio
n
s
,
o
cc
lu
s
io
n
s
(
s
u
ch
as
m
ask
s
o
r
g
lass
es),
an
d
e
n
v
ir
o
n
m
en
tal
c
o
n
d
itio
n
s
.
E
x
p
an
d
in
g
th
e
d
ataset
in
th
is
m
an
n
er
w
o
u
ld
allo
w
t
h
e
C
NN
m
o
d
el
to
lea
r
n
a
b
r
o
a
d
er
d
is
tr
ib
u
tio
n
o
f
f
ac
ial
f
ea
tu
r
es,
u
ltima
tely
im
p
r
o
v
in
g
ac
cu
r
ac
y
an
d
r
o
b
u
s
tn
ess
wh
en
d
ep
lo
y
e
d
in
r
ea
lis
tic
f
ield
en
v
ir
o
n
m
en
ts
.
Ad
d
itio
n
ally
,
th
e
in
clu
s
io
n
o
f
p
u
b
licly
a
v
ailab
le
lar
g
e
-
s
ca
le
d
atasets
,
s
u
ch
as
L
FW
,
VGGFac
e2
,
o
r
MS
-
C
eleb
-
1
M,
o
r
cu
s
to
m
d
atasets
g
ath
er
ed
f
r
o
m
o
p
er
atio
n
al
s
u
r
v
eillan
ce
ca
m
er
as,
wo
u
ld
p
r
o
v
id
e
m
o
r
e
c
o
m
p
r
e
h
en
s
iv
e
v
al
id
atio
n
an
d
a
m
o
r
e
r
eliab
le
m
ea
s
u
r
e
o
f
r
ea
l
-
wo
r
ld
p
er
f
o
r
m
an
ce.
5.
CO
NCLU
SI
O
N
T
r
ac
k
in
g
is
o
f
ten
r
e
f
er
r
e
d
to
as
a
s
er
v
ice,
an
d
lik
e
a
n
y
o
th
er
m
o
d
er
n
tech
n
o
lo
g
y
o
r
tech
n
ical
ass
is
tan
ce
,
it
h
as
ad
v
an
tag
es
an
d
d
is
ad
v
an
tag
es.
T
r
ac
k
in
g
p
eo
p
le
v
ia
GPS
tech
n
o
lo
g
y
h
as
b
ec
o
m
e
o
u
td
ated
b
ec
au
s
e
it
ca
n
b
e
d
if
f
icu
lt
to
p
lace
a
d
e
v
ice
o
n
a
p
er
s
o
n
if
th
ey
ar
e
wan
te
d
f
o
r
cr
im
es
—
alter
n
ativ
e
tech
n
o
lo
g
y
is
n
ee
d
ed
.
T
h
is
p
r
o
p
o
s
ed
s
y
s
t
em
p
r
esen
ted
a
d
if
f
er
en
t
m
ec
h
an
is
m
f
r
o
m
t
h
e
to
o
ls
p
r
e
v
io
u
s
ly
u
s
ed
in
tr
ac
k
in
g
p
eo
p
le.
T
h
e
m
o
d
el
was
b
u
ilt
u
s
in
g
C
NN
tech
n
iq
u
es
an
d
tr
ain
ed
in
tellig
en
tly
b
y
T
e
n
s
o
r
Flo
w.
T
h
is
tr
ain
ed
f
o
r
m
i
d
en
tifie
s
th
e
wan
ted
p
e
r
s
o
n
s
an
d
s
en
d
s
th
eir
in
f
o
r
m
a
tio
n
an
d
r
ea
lity
to
th
e
s
y
s
tem
ad
m
in
is
tr
ato
r
.
T
h
e
f
o
r
m
m
u
s
t
b
e
c
o
n
n
ec
ted
t
o
its
ca
m
er
a
f
o
r
tr
ac
k
in
g
to
ta
k
e
p
lace
.
T
h
e
s
y
s
tem
test
is
co
n
d
u
cted
in
two
s
tag
es;
th
e
f
ir
s
t is in
ca
s
e
th
e
lig
h
t is n
o
t g
o
o
d
at
th
e
ca
m
er
a.
T
h
e
s
y
s
tem
was tr
ain
ed
with
im
ag
es;
5
0
% o
f
th
e
im
ag
es
wer
e
id
en
tifie
d
,
an
d
th
e
cu
r
r
e
n
t
in
f
o
r
m
atio
n
was
d
eter
m
in
e
d
.
I
n
th
e
o
th
er
p
ar
t,
an
e
x
p
er
i
m
en
t
was
co
n
d
u
cted
o
n
th
e
s
y
s
tem
u
n
d
e
r
n
o
r
m
al
l
ig
h
tin
g
c
o
n
d
itio
n
s
,
u
s
i
n
g
th
e
ex
ac
t
m
ec
h
an
is
m
e
m
p
lo
y
e
d
i
n
th
e
f
i
r
s
t
test
.
T
h
e
s
y
s
tem
ac
h
iev
ed
a
1
0
0
%
ac
cu
r
ac
y
r
ate;
all
im
ag
es
wer
e
r
ec
o
g
n
ized
,
an
d
th
e
co
d
in
g
f
o
lk
was
also
ev
alu
ated
u
s
in
g
th
e
h
ash
al
g
o
r
ith
m
,
with
an
ac
cu
r
ac
y
s
co
r
e
o
f
9
9
,
wh
ic
h
is
co
n
s
id
er
ed
g
o
o
d
.
F
UNDING
I
NF
O
R
M
A
T
I
O
N
T
h
e
au
th
o
r
s
s
tate
n
o
f
u
n
d
in
g
is
in
v
o
lv
ed
.
AUTHO
R
CO
NT
RI
B
UT
I
O
NS ST
A
T
E
M
E
N
T
T
h
is
jo
u
r
n
al
u
s
es
th
e
C
o
n
t
r
ib
u
to
r
R
o
les
T
a
x
o
n
o
m
y
(
C
R
ed
iT)
to
r
ec
o
g
n
ize
in
d
iv
i
d
u
al
au
th
o
r
co
n
tr
ib
u
tio
n
s
,
r
ed
u
ce
au
th
o
r
s
h
ip
d
is
p
u
tes,
an
d
f
ac
ilit
ate
co
llab
o
r
atio
n
.
Na
m
e
o
f
Au
t
ho
r
C
M
So
Va
Fo
I
R
D
O
E
Vi
Su
P
Fu
Z
in
ah
Sh
ik
er
Ma
k
k
i
✓
✓
✓
✓
✓
✓
✓
Ah
m
et
Z
en
g
in
✓
✓
✓
✓
C
:
C
o
n
c
e
p
t
u
a
l
i
z
a
t
i
o
n
M
:
M
e
t
h
o
d
o
l
o
g
y
So
:
So
f
t
w
a
r
e
Va
:
Va
l
i
d
a
t
i
o
n
Fo
:
Fo
r
mal
a
n
a
l
y
s
i
s
I
:
I
n
v
e
s
t
i
g
a
t
i
o
n
R
:
R
e
so
u
r
c
e
s
D
:
D
a
t
a
C
u
r
a
t
i
o
n
O
:
W
r
i
t
i
n
g
-
O
r
i
g
i
n
a
l
D
r
a
f
t
E
:
W
r
i
t
i
n
g
-
R
e
v
i
e
w
&
E
d
i
t
i
n
g
Vi
:
Vi
su
a
l
i
z
a
t
i
o
n
Su
:
Su
p
e
r
v
i
s
i
o
n
P
:
P
r
o
j
e
c
t
a
d
mi
n
i
st
r
a
t
i
o
n
Fu
:
Fu
n
d
i
n
g
a
c
q
u
i
si
t
i
o
n
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
7
2
2
-
3
2
2
1
C
o
m
p
u
t Sci
I
n
f
T
ec
h
n
o
l
,
Vo
l.
7
,
No
.
2
,
J
u
ly
20
26
:
203
-
2
1
3
212
CO
NF
L
I
C
T
O
F
I
N
T
E
R
E
S
T
ST
A
T
E
M
E
NT
T
h
e
au
th
o
r
s
d
ec
lar
e
th
at
th
ey
h
av
e
n
o
k
n
o
w
n
co
m
p
etin
g
f
in
an
cial
in
ter
ests
o
r
p
er
s
o
n
al
r
el
atio
n
s
h
ip
s
th
at
co
u
ld
h
av
e
ap
p
ea
r
ed
t
o
in
f
lu
en
ce
th
e
wo
r
k
r
e
p
o
r
te
d
in
t
h
is
p
ap
er
.
DATA AV
AI
L
AB
I
L
I
T
Y
Data
av
ailab
ilit
y
is
n
o
t
a
p
p
li
ca
b
le
to
th
is
p
ap
er
as
n
o
n
e
w
d
ata
wer
e
c
r
ea
ted
o
r
an
al
y
ze
d
in
t
h
is
s
tu
d
y
.
RE
F
E
R
E
NC
E
S
[
1
]
J.
D
h
a
m
i
j
a
,
T.
C
h
o
u
d
h
u
r
y
,
P
.
K
u
m
a
r
,
a
n
d
Y
.
S
.
R
a
t
h
o
r
e
,
“
A
n
a
d
v
a
n
c
e
m
e
n
t
t
o
w
a
r
d
s
e
f
f
i
c
i
e
n
t
f
a
c
e
r
e
c
o
g
n
i
t
i
o
n
u
si
n
g
l
i
v
e
v
i
d
e
o
f
e
e
d
:
‘
f
o
r
t
h
e
f
u
t
u
r
e
,
’
”
i
n
2
0
1
7
3
r
d
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
a
t
i
o
n
a
l
I
n
t
e
l
l
i
g
e
n
c
e
a
n
d
N
e
t
w
o
r
k
s
(
C
I
N
E)
,
I
EEE,
O
c
t
.
2
0
1
7
,
p
p
.
53
–
56
,
d
o
i
:
1
0
.
1
1
0
9
/
C
I
N
E.
2
0
1
7
.
2
1
.
[
2
]
M
.
H
a
d
a
,
R
.
Y
a
ma
d
a
,
a
n
d
S
.
A
k
a
m
a
t
s
u
,
“
H
o
w
d
o
e
s
t
h
e
t
r
a
n
sf
o
r
ma
t
i
o
n
o
f
a
n
a
v
a
t
a
r
f
a
c
e
g
i
v
i
n
g
a
f
a
v
o
r
a
b
l
e
i
mp
r
e
ssi
o
n
a
f
f
e
c
t
h
u
m
a
n
r
e
c
o
g
n
i
t
i
o
n
o
f
t
h
e
f
a
c
e
?
,
”
i
n
2
0
1
8
I
n
t
e
r
n
a
t
i
o
n
a
l
Wo
r
k
s
h
o
p
o
n
Ad
v
a
n
c
e
d
I
m
a
g
e
T
e
c
h
n
o
l
o
g
y
(
I
W
AI
T
)
,
I
EEE,
Ja
n
.
2
0
1
8
,
p
p
.
1
–
3
,
d
o
i
:
1
0
.
1
1
0
9
/
I
W
A
I
T.
2
0
1
8
.
8
3
6
9
6
5
5
.
[
3
]
H
.
K
w
o
n
,
O
.
K
w
o
n
,
H
.
Y
o
o
n
,
a
n
d
K
.
W
.
P
a
r
k
,
“
F
a
c
e
f
r
i
e
n
d
-
safe
a
d
v
e
r
sari
a
l
e
x
a
m
p
l
e
o
n
f
a
c
e
r
e
c
o
g
n
i
t
i
o
n
sy
s
t
e
m
,
”
i
n
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
U
b
i
q
u
i
t
o
u
s
a
n
d
Fu
t
u
re
N
e
t
w
o
rks
,
I
C
U
F
N
,
2
0
1
9
,
p
p
.
5
4
7
–
5
5
1
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
U
F
N
.
2
0
1
9
.
8
8
0
6
1
2
4
.
[
4
]
S
.
Z
h
a
o
,
Z.
X
u
,
Z.
Z
h
u
,
X
.
L
i
a
n
g
,
Z
.
Zh
a
n
g
,
a
n
d
R
.
J
i
a
n
g
,
“
S
h
o
r
t
a
n
d
l
o
n
g
-
t
e
r
m
r
e
n
e
w
a
b
l
e
e
l
e
c
t
r
i
c
i
t
y
d
e
m
a
n
d
f
o
r
e
c
a
st
i
n
g
b
a
s
e
d
o
n
C
N
N
-
BI
-
G
R
U
m
o
d
e
l
,
”
I
C
C
K
T
r
a
n
sa
c
t
i
o
n
s
o
n
Em
e
r
g
i
n
g
T
o
p
i
c
s
i
n
Art
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
,
v
o
l
.
2
,
n
o
.
1
,
p
p
.
1
–
1
5
,
2
0
2
5
,
d
o
i
:
1
0
.
6
2
7
6
2
/
t
e
t
a
i
.
2
0
2
4
.
5
3
2
2
5
3
.
[
5
]
S
.
J
o
sh
i
,
B
.
L
.
M
a
h
a
n
t
h
i
,
G
.
P
a
v
i
t
h
r
a
,
K
.
S
.
P
o
k
k
u
l
u
r
i
,
S
.
S
.
N
i
n
a
w
e
,
a
n
d
R
.
S
a
h
u
,
“
I
n
t
e
g
r
a
t
i
n
g
LST
M
a
n
d
C
N
N
f
o
r
st
o
c
k
mar
k
e
t
p
r
e
d
i
c
t
i
o
n
:
a
d
y
n
a
mi
c
m
a
c
h
i
n
e
l
e
a
r
n
i
n
g
a
p
p
r
o
a
c
h
,
”
J
o
u
r
n
a
l
o
f
Art
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
a
n
d
T
e
c
h
n
o
l
o
g
y
,
v
o
l
.
5
,
p
p
.
1
6
8
–
1
7
9
,
2
0
2
5
,
d
o
i
:
1
0
.
3
7
9
6
5
/
j
a
i
t
.
2
0
2
5
.
0
6
5
2
.
[
6
]
G
.
G
e
o
r
g
e
,
R
.
B
o
b
e
n
,
B
.
R
a
d
h
a
k
r
i
sh
n
a
n
,
a
n
d
L
.
P
.
S
u
r
e
s
h
,
“
F
a
c
e
r
e
c
o
g
n
i
t
i
o
n
o
n
su
r
g
i
c
a
l
l
y
a
l
t
e
r
e
d
f
a
c
e
s
u
s
i
n
g
p
r
i
n
c
i
p
a
l
c
o
m
p
o
n
e
n
t
a
n
a
l
y
s
i
s,
”
i
n
Pr
o
c
e
e
d
i
n
g
s o
f
I
EE
E
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
C
i
rc
u
i
t
,
Po
w
e
r
a
n
d
C
o
m
p
u
t
i
n
g
T
e
c
h
n
o
l
o
g
i
e
s
,
I
C
C
PC
T
2
0
1
7
,
2
0
1
7
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
C
P
C
T.
2
0
1
7
.
8
0
7
4
3
2
4
.
[
7
]
J.
B
o
u
d
j
a
d
a
r
,
S
.
U
.
I
sl
a
m,
a
n
d
R
.
B
u
y
y
a
,
“
D
y
n
a
m
i
c
f
p
g
a
r
e
c
o
n
f
i
g
u
r
a
t
i
o
n
f
o
r
sca
l
a
b
l
e
e
m
b
e
d
d
e
d
a
r
t
i
f
i
c
i
a
l
i
n
t
e
l
l
i
g
e
n
c
e
(
AI
)
:
a
c
o
-
d
e
s
i
g
n
me
t
h
o
d
o
l
o
g
y
f
o
r
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
(
C
N
N
)
a
c
c
e
l
e
r
a
t
i
o
n
,
”
F
u
t
u
r
e
G
e
n
e
r
a
t
i
o
n
C
o
m
p
u
t
e
r
S
y
st
e
m
s
,
v
o
l
.
1
6
9
,
2
0
2
5
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
f
u
t
u
r
e
.
2
0
2
5
.
1
0
7
7
7
7
.
[
8
]
M
.
A
.
W
a
h
e
d
,
M
.
A
l
q
a
r
a
l
e
h
,
M
.
S
.
A
l
z
b
o
o
n
,
a
n
d
M
.
S
.
Al
-
B
a
t
a
h
,
“
E
v
a
l
u
a
t
i
n
g
AI
a
n
d
m
a
c
h
i
n
e
l
e
a
r
n
i
n
g
mo
d
e
l
s
i
n
b
r
e
a
s
t
c
a
n
c
e
r
d
e
t
e
c
t
i
o
n
:
a
r
e
v
i
e
w
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
s
(
C
N
N
)
a
n
d
g
l
o
b
a
l
r
e
sea
r
c
h
t
r
e
n
d
s
,
”
L
a
t
I
A
,
v
o
l
.
3
,
p
.
1
1
7
,
Ja
n
.
2
0
2
5
,
d
o
i
:
1
0
.
6
2
4
8
6
/
l
a
t
i
a
2
0
2
5
1
1
7
.
[
9
]
M
.
M
.
D
a
s
a
n
d
S
.
S
h
a
s
t
r
i
“
M
a
c
h
i
n
e
l
e
a
r
n
i
n
g
a
p
p
r
o
a
c
h
e
s
f
o
r
e
a
r
l
y
b
r
a
i
n
st
r
o
k
e
d
e
t
e
c
t
i
o
n
u
si
n
g
C
N
N
,
”
J
o
u
rn
a
l
o
f
S
c
i
e
n
t
i
f
i
c
Re
se
a
rc
h
a
n
d
T
e
c
h
n
o
l
o
g
y
,
p
p
.
2
4
3
–
2
5
0
,
2
0
2
5
.
[
1
0
]
B
.
S
a
o
u
d
,
A
.
H
.
H
.
M
.
M
o
h
a
me
d
,
I
.
S
h
a
y
e
a
,
A
.
A
.
El
-
S
a
l
e
h
,
a
n
d
A
.
A
l
a
sh
b
i
,
“
R
e
v
i
e
w
o
f
m
a
sk
e
d
f
a
c
e
r
e
c
o
g
n
i
t
i
o
n
b
a
s
e
d
o
n
d
e
e
p
l
e
a
r
n
i
n
g
,
”
T
e
c
h
n
o
l
o
g
i
es
,
v
o
l
.
1
3
,
n
o
.
7
,
p
.
3
1
0
,
Ju
l
.
2
0
2
5
,
d
o
i
:
1
0
.
3
3
9
0
/
t
e
c
h
n
o
l
o
g
i
e
s
1
3
0
7
0
3
1
0
.
[
1
1
]
S
.
S
a
mb
o
l
e
k
a
n
d
M
.
I
v
a
s
i
c
-
K
o
s
,
“
D
e
t
e
r
mi
n
i
n
g
t
h
e
g
e
o
l
o
c
a
t
i
o
n
o
f
a
p
e
r
s
o
n
d
e
t
e
c
t
e
d
i
n
a
n
i
ma
g
e
t
a
k
e
n
w
i
t
h
a
d
r
o
n
e
,
”
2
0
2
3
,
d
o
i
:
1
0
.
2
1
3
9
/
ssr
n
.
4
3
7
3
9
8
7
.
[
1
2
]
T.
Y
i
n
g
mi
n
g
,
D
.
K
e
n
a
n
,
Q
.
Ji
a
n
f
e
n
g
,
F
.
Li
,
a
n
d
C
.
Y
i
,
“
A
l
e
a
r
n
i
n
g
o
b
s
e
r
v
e
r
-
b
a
se
d
c
o
n
t
r
o
l
st
r
a
t
e
g
y
f
o
r
P
M
S
M
w
i
t
h
p
o
si
t
i
o
n
sen
s
o
r
f
a
u
l
t
i
n
r
a
i
l
w
a
y
,
”
C
o
n
t
r
o
l
E
n
g
i
n
e
e
ri
n
g
Pr
a
c
t
i
c
e
,
v
o
l
.
1
4
2
,
p
.
1
0
5
7
0
5
,
Jan
.
2
0
2
4
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
c
o
n
e
n
g
p
r
a
c
.
2
0
2
3
.
1
0
5
7
0
5
.
[
1
3
]
H
.
T
o
r
k
e
y
,
S
.
H
a
s
h
i
s
h
,
S
.
S
o
u
i
ss
i
,
E.
E.
D
.
H
e
m
d
a
n
,
a
n
d
A
.
S
a
y
e
d
,
“
S
e
i
z
u
r
e
d
e
t
e
c
t
i
o
n
i
n
me
d
i
c
a
l
I
o
T
:
h
y
b
r
i
d
C
N
N
-
LSTM
-
G
R
U
mo
d
e
l
w
i
t
h
d
a
t
a
b
a
l
a
n
c
i
n
g
a
n
d
XAI
i
n
t
e
g
r
a
t
i
o
n
,
”
A
l
g
o
r
i
t
h
m
s
,
v
o
l
.
1
8
,
n
o
.
2
,
2
0
2
5
,
d
o
i
:
1
0
.
3
3
9
0
/
a
1
8
0
2
0
0
7
7
.
[
1
4
]
J.
K
e
r
,
L.
W
a
n
g
,
J.
R
a
o
,
a
n
d
T.
L
i
m,
“
D
e
e
p
l
e
a
r
n
i
n
g
a
p
p
l
i
c
a
t
i
o
n
s
i
n
me
d
i
c
a
l
i
ma
g
e
a
n
a
l
y
s
i
s,
”
I
EEE
Ac
c
e
ss
,
v
o
l
.
6
,
p
p
.
9
3
7
5
–
9
3
8
9
,
2
0
1
8
,
d
o
i
:
1
0
.
1
1
0
9
/
A
C
C
ESS
.
2
0
1
7
.
2
7
8
8
0
4
4
.
[
1
5
]
N
.
O
t
t
e
r
n
e
ss
e
t
a
l
.
,
“
A
n
e
v
a
l
u
a
t
i
o
n
o
f
t
h
e
N
V
I
D
I
A
TX
1
f
o
r
su
p
p
o
r
t
i
n
g
r
e
a
l
-
t
i
m
e
c
o
m
p
u
t
e
r
-
v
i
s
i
o
n
w
o
r
k
l
o
a
d
s
,
”
i
n
2
0
1
7
I
E
EE
Re
a
l
-
T
i
m
e
a
n
d
Em
b
e
d
d
e
d
T
e
c
h
n
o
l
o
g
y
a
n
d
Ap
p
l
i
c
a
t
i
o
n
s
S
y
m
p
o
s
i
u
m
(
RTAS
)
,
I
EEE,
A
p
r
.
2
0
1
7
,
p
p
.
3
5
3
–
3
6
4
,
d
o
i
:
1
0
.
1
1
0
9
/
R
TA
S
.
2
0
1
7
.
3
.
[
1
6
]
M
.
Z.
A
l
-
D
a
b
a
g
h
,
M
.
H
.
M
.
A
l
h
a
b
i
b
,
a
n
d
F
.
H
.
A
L
-
M
u
k
h
t
a
r
,
“
F
a
c
e
r
e
c
o
g
n
i
t
i
o
n
s
y
st
e
m
b
a
se
d
o
n
k
e
r
n
e
l
d
i
s
c
r
i
m
i
n
a
n
t
a
n
a
l
y
si
s,
k
-
n
e
a
r
e
st
n
e
i
g
h
b
o
r
a
n
d
s
u
p
p
o
r
t
v
e
c
t
o
r
mac
h
i
n
e
,
”
I
n
t
e
rn
a
t
i
o
n
a
l
J
o
u
rn
a
l
o
f
R
e
se
a
rc
h
a
n
d
E
n
g
i
n
e
e
r
i
n
g
,
v
o
l
.
5
,
n
o
.
2
,
p
p
.
3
3
5
–
3
3
8
,
2
0
1
8
,
d
o
i
:
1
0
.
2
1
2
7
6
/
i
j
r
e
.
2
0
1
8
.
5
.
3
.
3
.
[
1
7
]
I
.
A
h
me
d
a
n
d
G
.
Je
o
n
,
“
A
r
e
a
l
-
t
i
m
e
p
e
r
so
n
t
r
a
c
k
i
n
g
s
y
s
t
e
m
b
a
s
e
d
o
n
s
i
a
mm
a
s
k
n
e
t
w
o
r
k
f
o
r
i
n
t
e
l
l
i
g
e
n
t
v
i
d
e
o
s
u
r
v
e
i
l
l
a
n
c
e
,
”
J
o
u
rn
a
l
o
f
Re
a
l
-
T
i
m
e
I
m
a
g
e
Pr
o
c
e
ssi
n
g
,
v
o
l
.
1
8
,
n
o
.
5
,
p
p
.
1
8
0
3
–
1
8
1
4
,
O
c
t
.
2
0
2
1
,
d
o
i
:
1
0
.
1
0
0
7
/
s
1
1
5
5
4
-
021
-
0
1
1
4
4
-
5.
[
1
8
]
J.
K
r
u
m
m,
S
.
H
a
r
r
i
s,
B
.
M
e
y
e
r
s
,
B
.
B
r
u
m
i
t
t
,
M
.
H
a
l
e
,
a
n
d
S
.
S
h
a
f
e
r
,
“
M
u
l
t
i
-
c
a
mera
m
u
l
t
i
-
p
e
r
so
n
t
r
a
c
k
i
n
g
f
o
r
e
a
s
y
l
i
v
i
n
g
,
”
i
n
Pro
c
e
e
d
i
n
g
s
T
h
i
r
d
I
EEE
I
n
t
e
r
n
a
t
i
o
n
a
l
W
o
rks
h
o
p
o
n
V
i
su
a
l
S
u
rv
e
i
l
l
a
n
c
e
,
I
EEE
C
o
m
p
u
t
.
S
o
c
,
2
0
0
0
,
p
p
.
3
–
10
,
d
o
i
:
1
0
.
1
1
0
9
/
V
S
.
2
0
0
0
.
8
5
6
8
5
2
.
[
1
9
]
M
.
J.
G
ó
mez
-
S
i
l
v
a
,
E
.
I
z
q
u
i
e
r
d
o
,
A
.
d
e
l
a
Es
c
a
l
e
r
a
,
a
n
d
J.
M
.
A
r
mi
n
g
o
l
,
“
Tr
a
n
sf
e
r
r
i
n
g
l
e
a
r
n
i
n
g
f
r
o
m
m
u
l
t
i
-
p
e
r
so
n
t
r
a
c
k
i
n
g
t
o
p
e
r
s
o
n
r
e
-
i
d
e
n
t
i
f
i
c
a
t
i
o
n
,
”
I
n
t
e
g
r
a
t
e
d
C
o
m
p
u
t
e
r
-
A
i
d
e
d
En
g
i
n
e
e
ri
n
g
,
v
o
l
.
2
6
,
n
o
.
4
,
p
p
.
3
2
9
–
3
4
4
,
S
e
p
.
2
0
1
9
,
d
o
i
:
1
0
.
3
2
3
3
/
I
C
A
-
1
9
0
6
0
3
.
[2
0
]
P
.
M
.
K
u
mar,
U
.
G
a
n
d
h
i
,
R
.
V
a
r
a
t
h
a
r
a
j
a
n
,
G
.
M
a
n
o
g
a
r
a
n
,
J.
R
.
,
a
n
d
T
.
V
a
d
i
v
e
l
,
“
I
n
t
e
l
l
i
g
e
n
t
f
a
c
e
r
e
c
o
g
n
i
t
i
o
n
a
n
d
n
a
v
i
g
a
t
i
o
n
sy
st
e
m
u
si
n
g
n
e
u
r
a
l
l
e
a
r
n
i
n
g
f
o
r
sm
a
r
t
se
c
u
r
i
t
y
i
n
I
n
t
e
r
n
e
t
o
f
T
h
i
n
g
s,
”
C
l
u
st
e
r
C
o
m
p
u
t
i
n
g
,
v
o
l
.
2
2
,
n
o
.
4
,
p
p
.
7
7
3
3
–
7
7
4
4
,
2
0
1
9
,
d
o
i
:
1
0
.
1
0
0
7
/
s
1
0
5
8
6
-
017
-
1
3
2
3
-
4.
[2
1
]
K
.
N
g
a
m
a
k
e
u
r
,
S
.
Y
o
n
g
c
h
a
r
e
o
n
,
J.
Y
u
,
a
n
d
Q
.
Z
.
S
h
e
n
g
,
“
D
e
e
p
C
N
N
-
LSTM
n
e
t
w
o
r
k
f
o
r
i
n
d
o
o
r
l
o
c
a
t
i
o
n
e
s
t
i
m
a
t
i
o
n
u
si
n
g
a
n
a
l
o
g
si
g
n
a
l
s
o
f
p
a
ssi
v
e
i
n
f
r
a
r
e
d
s
e
n
so
r
s,
”
I
EEE
I
n
t
e
rn
e
t
o
f
T
h
i
n
g
s
J
o
u
rn
a
l
,
v
o
l
.
9
,
n
o
.
2
2
,
p
p
.
2
2
5
8
2
–
2
2
5
9
4
,
N
o
v
.
2
0
2
2
,
d
o
i
:
1
0
.
1
1
0
9
/
JI
O
T.
2
0
2
2
.
3
1
8
3
1
4
8
.
[2
2
]
S
.
Y
o
n
g
c
h
a
r
e
o
n
,
J.
Y
u
,
a
n
d
J.
M
a
,
“
E
f
f
i
c
i
e
n
t
d
e
e
p
l
e
a
r
n
i
n
g
-
b
a
s
e
d
d
e
v
i
c
e
-
f
r
e
e
i
n
d
o
o
r
l
o
c
a
l
i
z
a
t
i
o
n
u
s
i
n
g
p
a
ss
i
v
e
i
n
f
r
a
r
e
d
s
e
n
s
o
r
s,
”
S
e
n
so
rs
,
v
o
l
.
2
5
,
n
o
.
5
,
p
.
1
3
6
2
,
F
e
b
.
2
0
2
5
,
d
o
i
:
1
0
.
3
3
9
0
/
s2
5
0
5
1
3
6
2
.
[2
3
]
S
.
P
.
S
h
a
s
h
i
k
u
mar
,
A
.
J.
S
h
a
h
,
Q
.
L
i
,
G
.
D
.
C
l
i
f
f
o
r
d
,
a
n
d
S
.
N
e
ma
t
i
,
“
A
d
e
e
p
l
e
a
r
n
i
n
g
a
p
p
r
o
a
c
h
t
o
m
o
n
i
t
o
r
i
n
g
a
n
d
d
e
t
e
c
t
i
n
g
a
t
r
i
a
l
f
i
b
r
i
l
l
a
t
i
o
n
u
s
i
n
g
w
e
a
r
a
b
l
e
t
e
c
h
n
o
l
o
g
y
,
”
i
n
2
0
1
7
I
EEE
E
MB
S
I
n
t
e
rn
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
Bi
o
m
e
d
i
c
a
l
and
H
e
a
l
t
h
I
n
f
o
rm
a
t
i
c
s
(
BH
I
)
,
I
EEE,
2
0
1
7
,
p
p
.
1
4
1
–
1
4
4
,
d
o
i
:
1
0
.
1
1
0
9
/
B
H
I
.
2
0
1
7
.
7
8
9
7
2
2
5
.
Evaluation Warning : The document was created with Spire.PDF for Python.