I
AE
S In
t
er
na
t
io
na
l J
o
urna
l o
f
Art
if
icia
l In
t
ellig
ence
(
I
J
-
AI
)
Vo
l.
1
4
,
No
.
6
,
Dec
em
b
er
2
0
2
5
,
p
p
.
4
4
1
5
~
4
4
2
6
I
SS
N:
2
2
5
2
-
8
9
3
8
,
DOI
: 1
0
.
1
1
5
9
1
/ijai.v
14
.i
6
.
p
p
4
4
1
5
-
4
4
2
6
4415
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
a
i
.
ia
esco
r
e.
co
m
A review
of drive
r distract
io
n det
e
ction whil
e drivin
g
bas
ed on
co
nv
o
lutiona
l neural networks
G
ha
dy
Alha
m
a
d
,
M
o
ha
m
a
d
-
B
a
s
s
a
m
K
urdy
D
e
p
a
r
t
me
n
t
o
f
I
n
f
o
r
m
a
t
i
o
n
Te
c
h
n
o
l
o
g
y
,
S
y
r
i
a
n
V
i
r
t
u
a
l
U
n
i
v
e
r
si
t
y
,
D
a
m
a
s
c
u
s,
S
y
r
i
a
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Sep
1
,
2
0
2
4
R
ev
is
ed
J
u
l 5
,
2
0
2
5
Acc
ep
ted
Oct
1
6
,
2
0
2
5
Driv
e
r
d
istrac
ti
o
n
re
p
re
se
n
ts
a
m
a
jo
r
c
a
u
se
o
f
traffic
a
c
c
id
e
n
ts
,
p
o
si
n
g
a
se
rio
u
s
t
h
re
a
t
to
h
u
m
a
n
l
ife.
I
n
th
is
re
v
iew
,
we
p
re
se
n
t
th
e
late
st
re
se
a
rc
h
fin
d
i
n
g
s
o
f
d
ri
v
e
r
d
istrac
ti
o
n
d
e
tec
ti
o
n
b
a
se
d
o
n
c
o
n
v
o
l
u
ti
o
n
a
l
n
e
u
ra
l
n
e
two
rk
s
(CNN
s).
I
n
g
e
n
e
ra
l,
th
e
a
n
a
ly
sis o
f
d
riv
e
r
b
e
h
a
v
io
r
w
h
il
e
d
riv
i
n
g
is
re
p
re
se
n
ted
b
y
e
it
h
e
r
d
e
tec
ti
n
g
d
r
iv
e
r
d
r
o
ws
in
e
ss
o
r
a
tt
e
n
t
io
n
d
i
v
e
r
sio
n
fr
o
m
d
riv
i
n
g
b
y
o
t
h
e
r
a
c
ti
v
it
ies
,
a
ll
o
f
wh
ich
fa
ll
u
n
d
e
r
th
e
d
e
fin
it
io
n
o
f
d
ri
v
e
r
d
istrac
ti
o
n
.
F
a
c
ial
fe
a
tu
re
s
a
re
o
ften
th
e
b
a
sis
fo
r
d
e
tec
ti
n
g
d
r
iv
e
r
d
ro
ws
in
e
ss
.
I
n
m
o
st
p
a
p
e
rs,
it
is
ty
p
ica
ll
y
d
o
n
e
b
y
e
y
e
b
li
n
k
in
g
,
y
a
wn
in
g
,
a
n
d
h
e
a
d
m
o
v
e
m
e
n
t.
As
fo
r
t
h
e
d
riv
e
r
a
tt
e
n
ti
o
n
d
i
v
e
rsio
n
,
it
is
th
ro
u
g
h
t
h
e
p
o
siti
o
n
o
f
th
e
h
a
n
d
a
n
d
fa
c
e
.
I
t
in
v
o
lv
e
s
m
a
n
y
a
c
ti
v
it
ies
,
tex
t
m
e
ss
a
g
e
s,
m
a
k
in
g
p
h
o
n
e
c
a
ll
s,
a
d
j
u
stin
g
th
e
ra
d
io
,
c
o
n
su
m
i
n
g
b
e
v
e
ra
g
e
s,
re
a
c
h
in
g
fo
r
o
b
jec
ts
b
e
h
in
d
t
h
e
d
riv
e
r,
a
p
p
l
y
i
n
g
m
a
k
e
u
p
,
i
n
tera
c
ti
n
g
wit
h
p
a
ss
e
n
g
e
rs,
a
n
d
o
th
e
r
sim
il
a
r
d
istrac
ti
o
n
s
.
Ho
we
v
e
r,
su
g
g
e
stin
g
n
e
w
m
e
th
o
d
o
lo
g
ie
s
in
d
r
iv
e
r
d
istrac
ti
o
n
d
e
tec
ti
o
n
a
n
d
c
h
o
o
si
n
g
a
p
p
r
o
p
riate
CNN
-
b
a
se
d
tec
h
n
iq
u
e
s
is
a
b
ig
c
h
a
l
len
g
e
g
i
v
e
n
th
e
wi
d
e
v
a
riety
e
x
p
e
rime
n
ts
a
n
d
stu
d
ies
i
n
th
is
fiel
d
.
Th
e
re
fo
re
,
p
re
v
i
o
u
s
p
a
p
e
r
s
s
h
o
u
l
d
b
e
re
v
isit
e
d
t
o
p
ro
d
u
c
e
n
e
w
m
e
th
o
d
s
b
y
tak
in
g
a
d
v
a
n
tag
e
o
f
th
e
tec
h
n
iq
u
e
s
u
se
d
.
As
a
re
su
lt
,
th
is
p
a
p
e
r
re
v
iew
s
re
se
a
rc
h
a
p
p
ro
a
c
h
e
s a
n
d
re
v
e
a
l
s
th
e
e
ffe
c
ti
v
e
n
e
ss
o
f
CNN
in
d
e
tec
ti
n
g
d
r
iv
e
r
d
istrac
ti
o
n
.
F
in
a
ll
y
,
t
h
e
a
rti
c
le
li
sts
tec
h
n
iq
u
e
s
th
a
t
c
a
n
b
e
u
se
d
a
s
b
e
n
c
h
m
a
rk
s
i
n
th
is co
n
tex
t
.
K
ey
w
o
r
d
s
:
C
o
n
v
o
lu
tio
n
al
n
eu
r
al
n
etwo
r
k
C
o
g
n
itiv
e
d
is
tr
ac
tio
n
D
is
tr
ac
tio
n
Dr
iv
er
b
eh
a
v
io
r
F
ac
ial
f
ea
tu
r
es
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
:
Gh
ad
y
Alh
am
ad
Dep
ar
tm
en
t o
f
I
n
f
o
r
m
atio
n
T
e
ch
n
o
lo
g
y
,
Sy
r
ia
n
Vir
tu
al
Un
iv
er
s
ity
Dam
ascu
s
,
Sy
r
ia
E
m
ail:
g
h
ad
y
_
2
1
3
4
5
2
@
s
v
u
o
n
lin
e.
o
r
g
1.
I
NT
RO
D
UCT
I
O
N
Dis
tr
ac
ted
d
r
iv
in
g
h
as
b
ec
o
m
e
a
d
o
m
in
an
t
ca
u
s
e
o
f
tr
a
f
f
ic
ac
cid
en
ts
[
1
]
.
Acc
o
r
d
in
g
to
th
e
Natio
n
al
Hig
h
way
T
r
af
f
ic
Saf
ety
Ad
m
in
is
tr
atio
n
(
NHT
SA)
,
d
is
tr
ac
ted
d
r
iv
in
g
r
e
f
er
s
to
an
y
a
ctio
n
th
at
d
iv
er
ts
a
d
r
iv
er
’
s
f
o
cu
s
awa
y
f
r
o
m
s
af
ely
o
p
e
r
atin
g
th
e
v
eh
icle,
a
n
d
in
clu
d
es
a
n
y
th
in
g
th
at
s
h
if
ts
atten
tio
n
f
r
o
m
th
e
task
o
f
d
r
iv
in
g
[
2
]
.
B
ased
o
n
its
r
ep
o
r
t
f
r
o
m
th
e
Un
ited
States
,
b
etwe
en
2
0
1
1
an
d
2
0
2
0
,
a
p
p
r
o
x
im
ately
3
2
,
4
8
3
p
e
o
p
le
lo
s
t
th
eir
liv
es
in
cr
ash
es
in
f
lu
en
ce
d
b
y
d
r
iv
er
d
is
tr
ac
tio
n
.
I
n
2
0
2
0
alo
n
e
,
d
is
tr
ac
tio
n
-
r
elate
d
f
atalities
r
ea
ch
ed
3
,
1
4
2
n
atio
n
wid
e,
ac
co
u
n
tin
g
f
o
r
8
%
o
f
all
m
o
to
r
v
eh
icle
d
ea
th
s
,
m
ar
k
in
g
an
in
cr
ea
s
e
o
f
2
3
co
m
p
ar
ed
to
2
0
1
9
.
C
r
ash
es
in
v
o
lv
in
g
d
is
tr
ac
ted
d
r
i
v
in
g
r
e
p
r
esen
ted
1
4
%
o
f
in
j
u
r
y
cr
ash
es
an
d
1
3
%
o
f
all
p
o
lice
-
r
ep
o
r
ted
tr
af
f
ic
ac
cid
en
ts
th
at
y
ea
r
.
Am
o
n
g
d
r
iv
er
s
ag
ed
1
5
to
2
0
in
v
o
lv
ed
in
f
atal
cr
ash
es,
7
%
wer
e
r
ep
o
r
ted
as
d
is
tr
ac
ted
,
m
ak
i
n
g
th
is
ag
e
g
r
o
u
p
th
e
m
o
s
t
a
f
f
ec
ted
b
y
d
is
tr
ac
tio
n
d
u
r
in
g
d
ea
d
ly
ac
cid
e
n
ts
.
Ad
d
itio
n
ally
,
3
9
6
f
atalities
wer
e
lin
k
ed
s
p
ec
if
ically
to
ce
ll
p
h
o
n
e
-
r
elate
d
d
is
tr
ac
tio
n
s
,
co
m
p
r
is
in
g
1
3
%
o
f
all
d
ea
th
s
in
v
o
lv
i
n
g
a
d
is
tr
ac
ted
d
r
iv
er
.
I
n
2
0
2
0
,
d
is
tr
ac
ted
d
r
i
v
in
g
was
r
esp
o
n
s
ib
le
f
o
r
th
e
d
ea
th
s
o
f
5
8
7
n
o
n
-
o
cc
u
p
an
ts
,
in
clu
d
in
g
p
ed
estria
n
s
,
cy
clis
ts
,
an
d
o
th
er
i
n
d
iv
id
u
als
[
3
]
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
1
4
,
No
.
6
,
Dec
em
b
er
2
0
2
5
:
4
4
1
5
-
4
4
2
6
4416
T
h
er
e
ar
e
m
a
n
y
r
ea
s
o
n
s
f
o
r
d
is
tr
ac
tio
n
th
at
ca
n
lead
to
im
p
air
ed
d
r
iv
in
g
an
d
ac
cid
e
n
ts
.
T
h
e
u
s
e
o
f
m
o
b
ile
p
h
o
n
es
is
o
n
e
o
f
th
e
m
ain
ca
u
s
es
o
f
d
is
tr
ac
tio
n
.
I
n
f
a
ct,
u
s
in
g
m
o
b
ile
d
ev
ices
is
n
o
t
th
e
o
n
ly
ca
u
s
e
o
f
d
is
tr
ac
tio
n
,
b
u
t
ac
co
r
d
in
g
to
NHT
SA,
it
also
in
clu
d
es,
talk
in
g
o
r
tex
tin
g
o
n
y
o
u
r
p
h
o
n
e,
ea
tin
g
an
d
d
r
in
k
in
g
,
talk
in
g
to
p
eo
p
le
in
y
o
u
r
v
eh
ic
le,
f
id
d
lin
g
with
s
ter
eo
,
en
ter
tain
m
en
t o
r
n
av
ig
atio
n
s
y
s
tem
[
2
]
.
T
h
e
ce
n
ter
s
f
o
r
d
is
ea
s
e
co
n
tr
o
l
a
n
d
p
r
ev
e
n
tio
n
(
C
DC
)
id
en
tifie
s
th
r
ee
m
ain
c
ateg
o
r
ies
o
f
d
is
tr
ac
ted
d
r
iv
in
g
:
co
g
n
itiv
e,
v
is
u
al,
an
d
m
a
n
u
al.
C
o
g
n
itiv
e
d
is
tr
ac
tio
n
s
o
cc
u
r
wh
en
a
d
r
iv
e
r
’
s
m
in
d
wan
d
e
r
s
awa
y
f
r
o
m
t
h
e
task
o
f
d
r
iv
in
g
,
m
ea
n
in
g
th
at
e
v
en
if
th
eir
b
o
d
y
r
em
ain
s
in
a
p
r
o
p
er
d
r
iv
in
g
p
o
s
tu
r
e,
t
h
eir
f
o
c
u
s
is
m
en
tally
d
iv
er
ted
.
Vis
u
al
d
is
tr
ac
tio
n
s
h
ap
p
en
w
h
en
a
d
r
iv
er
’
s
ey
es
ar
e
tak
en
o
f
f
t
h
e
r
o
ad
d
u
e
to
f
atig
u
e,
d
r
o
wsi
n
ess
,
in
atten
tio
n
,
o
r
t
h
e
u
s
e
o
f
s
m
ar
tp
h
o
n
es.
Ma
n
u
al
d
is
tr
ac
tio
n
s
in
v
o
lv
e
tem
p
o
r
a
r
ily
tak
in
g
th
e
h
an
d
s
o
f
f
th
e
s
teer
in
g
wh
ee
l
to
p
er
f
o
r
m
task
s
s
u
ch
as
u
s
in
g
a
p
h
o
n
e,
ea
tin
g
o
r
d
r
in
k
i
n
g
,
g
r
o
o
m
i
n
g
,
o
r
in
ter
ac
tin
g
with
p
ass
en
g
er
s
[
4
]
.
Pre
v
io
u
s
wo
r
k
h
as
aim
ed
to
a
d
d
r
ess
th
e
is
s
u
es
r
elate
d
to
d
e
tectin
g
d
r
iv
er
f
atig
u
e
a
n
d
d
r
o
wsi
n
ess
[
5
]
–
[
7]
as
v
is
u
al
d
is
tr
ac
tio
n
,
an
d
d
etec
t
in
g
atten
tio
n
d
i
v
er
s
io
n
f
r
o
m
d
r
iv
i
n
g
b
y
o
th
er
ac
tiv
ities
[
8
]
–
[
10
]
as
m
an
u
al
d
is
tr
ac
tio
n
.
Ho
wev
er
,
th
e
s
tu
d
ies
s
till
lack
atten
tio
n
to
co
g
n
i
tiv
e
d
is
tr
ac
tio
n
.
E
v
e
n
th
e
d
ef
i
n
itio
n
s
o
f
c
o
g
n
itiv
e
d
is
tr
ac
tio
n
ar
e
n
o
t
co
m
p
letel
y
ag
r
ee
d
u
p
o
n
in
th
e
f
ield
o
f
d
r
iv
in
g
s
af
ety
.
So
m
e
r
esear
ch
er
s
h
av
e
d
ef
in
ed
co
g
n
itiv
e
d
is
tr
ac
tio
n
in
way
s
th
at
o
v
er
la
p
with
v
is
u
al
d
is
tr
a
ctio
n
[
1
1
]
,
o
r
with
th
e
co
n
ce
p
t
o
f
d
r
iv
e
r
m
e
n
tal
wo
r
k
lo
ad
[
1
2
]
.
Oth
er
s
tu
d
ies
d
ef
in
e
it
as
s
h
if
tin
g
atten
tio
n
to
s
ec
o
n
d
ar
y
task
s
th
at
ar
e
n
o
t
r
elate
d
to
th
e
d
r
iv
in
g
task
[
1
]
.
I
n
f
ac
t,
f
r
o
m
m
y
p
o
in
t
o
f
v
iew,
cu
r
r
e
n
t
s
tu
d
ies
h
av
e
n
o
t
p
aid
en
o
u
g
h
atten
tio
n
to
th
e
d
r
iv
er
’
s
f
ee
lin
g
s
an
d
p
s
y
ch
o
lo
g
ical
s
tate
as
co
g
n
itiv
e
d
is
tr
ac
tio
n
,
f
o
r
ex
am
p
le.
Alth
o
u
g
h
C
h
a
n
a
n
d
S
i
n
g
h
a
l
[
1
3
]
i
n
o
l
d
s
t
u
d
y
a
n
a
l
y
s
e
d
t
h
e
r
el
a
ti
o
n
s
h
i
p
b
e
t
w
e
e
n
e
m
o
t
i
o
n
a
l
s
i
d
e
a
n
d
co
g
n
i
t
i
v
e
d
i
s
t
r
a
c
ti
o
n
.
T
o
a
c
h
i
e
v
e
t
h
i
s
g
o
a
l
,
a
d
r
i
v
i
n
g
s
i
m
u
l
at
o
r
w
as
u
s
e
d
,
a
n
d
t
h
e
em
o
t
i
o
n
a
l
w
o
r
d
s
w
e
r
e
d
i
v
i
d
e
d
in
t
o
t
h
r
e
e
ca
t
e
g
o
r
i
es
:
n
e
u
t
r
a
l
,
n
e
g
a
t
i
v
e
,
a
n
d
p
o
s
i
ti
v
e
.
I
n
a
d
d
i
t
i
o
n
,
C
h
a
n
d
a
n
d
K
a
r
t
h
i
k
e
y
a
n
[
1
4
]
p
r
o
p
o
s
e
d
a
m
o
d
e
l
c
o
m
p
o
s
e
d
o
f
t
w
o
m
a
i
n
c
o
m
p
o
n
en
t
s
:
d
et
e
c
ti
n
g
d
r
i
v
e
r
f
a
t
i
g
u
e
a
n
d
a
n
a
l
y
z
i
n
g
t
h
e
d
r
i
v
e
r
’
s
e
m
o
t
i
o
n
a
l
s
ta
t
e
t
o
p
r
ev
e
n
t
r
e
c
k
l
e
s
s
d
r
i
v
i
n
g
.
T
h
e
y
c
o
m
b
i
n
e
d
f
a
t
i
g
u
e
a
s
s
e
s
s
m
e
n
t
w
i
t
h
e
m
o
t
i
o
n
a
n
a
ly
s
i
s
,
o
b
s
e
r
v
i
n
g
t
h
at
d
r
i
v
e
r
b
eh
a
v
i
o
r
c
a
n
v
a
r
y
a
c
r
o
s
s
m
u
l
ti
p
l
e
s
t
a
t
es
,
i
n
cl
u
d
i
n
g
n
o
r
m
a
l
,
f
a
t
i
g
u
e
d
,
a
g
g
r
e
s
s
i
v
e
,
d
i
s
t
u
r
b
e
d
,
a
n
d
u
n
d
e
r
t
h
e
i
n
f
l
u
e
n
c
e
o
f
a
l
c
o
h
o
l
.
R
ec
en
tly
,
a
v
ar
iety
o
f
m
eth
o
d
s
h
av
e
b
ee
n
p
r
o
p
o
s
ed
f
o
r
d
r
iv
er
d
is
tr
ac
tio
n
d
etec
tio
n
.
E
u
cli
d
ea
n
asp
ec
t
r
atio
(
E
AR
)
[
15
]
,
[1
6
]
,
p
er
ce
n
tag
e
o
f
ey
elid
clo
s
u
r
e
(
PERC
L
OS)
,
f
r
eq
u
en
cy
o
f
o
p
en
m
o
u
th
(
FOM)
[1
7
]
,
s
u
p
p
o
r
t
v
ec
to
r
m
ac
h
in
e
(
S
VM
)
[
14
]
,
[
1
7
]
–
[
1
9
]
m
o
d
el
,
lo
n
g
s
h
o
r
t
-
ter
m
m
em
o
r
y
(
L
STM
)
[
2
0
]
,
[
2
1
]
,
co
n
v
o
l
u
tio
n
al
n
eu
r
al
n
etwo
r
k
s
(
C
NN)
[
14
]
,
[
19
]
–
[
2
3
]
.
E
A
R
wh
ich
m
ea
s
u
r
es
th
e
d
i
s
tan
ce
b
etwe
en
v
er
tical
an
d
h
o
r
izo
n
tal
ey
e
lan
d
m
ar
k
p
o
in
ts
b
y
u
s
in
g
an
E
u
clid
ea
n
d
is
tan
ce
m
eth
o
d
to
d
etec
t
th
e
ey
e
s
tate
[
1
5
]
.
PERC
L
OS
r
ep
r
esen
ts
th
e
p
r
o
p
o
r
tio
n
o
f
f
r
am
es
in
wh
ich
t
h
e
ey
es
ar
e
clo
s
ed
r
elativ
e
to
th
e
to
tal
f
r
am
es
with
in
a
g
iv
en
tim
e
p
er
i
o
d
,
wh
ile
F
OM
r
ef
er
s
to
th
e
p
r
o
p
o
r
tio
n
o
f
f
r
am
es
in
wh
ich
th
e
m
o
u
th
i
s
o
p
en
co
m
p
a
r
ed
to
th
e
to
tal
f
r
am
es
o
v
er
th
e
s
am
e
tim
e
in
ter
v
al
[
1
7
]
.
T
h
e
SV
M
m
o
d
el
is
esp
ec
ially
s
u
itab
l
e
f
o
r
class
if
icatio
n
task
s
in
v
o
lv
in
g
s
m
all
s
am
p
le
s
izes
[
2
0
]
.
L
STM
is
a
s
p
ec
ialized
v
ar
ian
t
o
f
th
e
r
ec
u
r
r
e
n
t
n
eu
r
al
n
etwo
r
k
(
R
NN)
,
ty
p
ically
u
s
ed
with
C
NN,
an
d
ca
n
ef
f
ec
tiv
ely
ca
p
tu
r
e
th
e
tim
e
in
f
o
r
m
atio
n
o
f
th
e
i
n
p
u
t
im
a
g
e
s
eq
u
en
ce
[
24
]
.
R
NNs
ar
e
well
-
s
u
ited
f
o
r
an
aly
zin
g
tim
e
s
er
ies
d
ata;
h
o
wev
er
,
th
ey
ar
e
g
en
er
ally
n
o
t
r
eg
ar
d
e
d
as
ef
f
ec
tiv
e
f
o
r
im
ag
e
p
r
o
ce
s
s
in
g
task
s
.
Ho
wev
e
r
,
C
NN
is
t
h
e
m
o
s
t
p
r
o
m
is
in
g
way
in
co
m
p
u
ter
v
is
io
n
-
b
ased
.
I
t
is
th
e
m
o
s
t
estab
lis
h
ed
alg
o
r
ith
m
am
o
n
g
v
a
r
io
u
s
d
ee
p
lear
n
in
g
m
o
d
els
[
25
]
.
C
NNs
r
esem
b
le
s
tan
d
ar
d
n
eu
r
al
n
etwo
r
k
s
in
th
at
t
h
ey
ar
e
co
m
p
o
s
ed
o
f
n
eu
r
o
n
s
wit
h
lear
n
ab
le
weig
h
ts
an
d
b
iases
[
26
]
.
I
t
is
v
er
y
wid
ely
u
s
ed
to
p
er
f
o
r
m
im
ag
e
class
if
icatio
n
,
o
b
ject
d
etec
tio
n
,
im
ag
e
r
ec
o
g
n
itio
n
,
f
ac
e
r
ec
o
g
n
itio
n
an
d
s
ev
er
al
o
th
er
task
s
r
elate
d
to
im
a
g
e
p
r
o
ce
s
s
in
g
[
27
]
.
Sin
ce
th
e
s
o
lu
tio
n
is
b
ased
o
n
C
NN,
th
is
will
lead
r
esear
ch
er
s
to
s
elec
t
th
e
b
est
am
o
n
g
a
s
et
o
f
tech
n
iq
u
es
to
ac
h
iev
e
th
eir
g
o
al
,
s
u
ch
as
p
r
ep
r
o
ce
s
s
in
g
tech
n
iq
u
es,
m
o
d
el
ar
ch
it
ec
tu
r
e
an
d
C
NN’
s
ap
p
r
o
ac
h
es,
d
atasets
,
an
d
m
eth
o
d
o
lo
g
ies.
T
h
e
d
ea
lin
g
o
f
e
ac
h
o
f
th
ese
tech
n
o
lo
g
ies
m
a
y
s
o
m
etim
es
d
if
f
er
d
ep
en
d
i
n
g
o
n
t
h
e
ty
p
e
o
f
d
i
s
tr
ac
tio
n
.
Dete
ct
in
g
d
r
iv
er
d
is
tr
ac
tio
n
r
eq
u
ir
es
r
ea
l
-
tim
e
d
r
iv
in
g
m
o
n
ito
r
in
g
.
T
h
er
ef
o
r
e,
th
e
v
id
eo
is
ca
p
tu
r
ed
b
y
eith
er
a
s
m
ar
t
p
h
o
n
e
ca
m
er
a
o
r
an
attac
h
e
d
ca
m
er
a.
B
i
r
r
ell
an
d
Fo
wk
es
[
28
]
r
ec
o
r
d
ed
v
id
e
o
s
with
f
o
u
r
ca
m
er
as
m
o
u
n
te
d
in
s
id
e
th
e
v
eh
icle
.
T
h
e
f
ir
s
t
ca
m
er
a,
attac
h
ed
to
a
s
m
ar
tp
h
o
n
e,
r
ec
o
r
d
ed
h
ig
h
-
d
e
f
in
itio
n
v
i
d
eo
f
o
cu
s
ed
o
n
th
e
d
r
iv
er
’
s
f
ac
e.
T
h
e
o
th
er
th
r
ee
ca
m
er
as
r
ec
o
r
d
ed
i
n
s
tan
d
ar
d
d
e
f
in
itio
n
:
two
m
o
n
ito
r
ed
th
e
f
o
r
war
d
an
d
r
ea
r
w
ar
d
d
r
iv
in
g
s
ce
n
es,
wh
ile
th
e
th
ir
d
ca
p
tu
r
ed
th
e
ac
tiv
ity
o
n
th
e
s
m
ar
tp
h
o
n
e
.
As f
o
r
Ma
li
et
a
l.
[
29
]
u
s
ed
th
e
f
r
o
n
t c
am
er
a
o
f
a
s
m
ar
tp
h
o
n
e
t
o
ca
p
tu
r
e
im
ag
es o
f
th
e
d
r
iv
er
,
an
d
th
en
f
ee
d
t
h
e
i
m
ag
es
to
th
e
s
m
ar
tp
h
o
n
e
f
o
r
i
m
ag
e
p
r
o
ce
s
s
in
g
.
Fu
r
th
er
m
o
r
e
,
m
an
y
o
th
er
th
in
g
s
s
h
o
u
ld
b
e
co
n
s
id
er
e
d
f
o
r
th
o
s
e
r
ec
o
r
d
in
g
s
.
I
t
will
b
e
d
is
c
u
s
s
ed
in
s
ec
tio
n
2
.
A
t
y
p
ical
C
NN
ar
ch
itectu
r
e
co
n
s
is
ts
o
f
th
r
ee
m
ai
n
lay
er
s
:
a
co
n
v
o
lu
tio
n
al
lay
er
,
a
p
o
o
l
in
g
lay
er
,
an
d
a
f
u
lly
c
o
n
n
ec
t
ed
lay
er
[
26
]
.
T
h
e
ar
r
an
g
em
e
n
t
a
n
d
co
n
f
i
g
u
r
atio
n
o
f
th
ese
lay
er
s
d
ef
in
e
th
e
m
o
d
el,
wh
ic
h
is
s
u
b
s
eq
u
en
tly
d
esig
n
ed
an
d
tr
ai
n
ed
o
n
a
d
ataset
to
ad
d
r
ess
a
s
p
ec
if
ic
p
r
o
b
lem
.
Ho
wev
e
r
,
a
p
r
e
-
tr
ain
ed
m
o
d
el
r
ef
er
s
to
a
n
etwo
r
k
cr
ea
ted
an
d
tr
ain
ed
b
y
o
t
h
er
s
o
n
a
la
r
g
e
d
ataset,
in
ten
d
ed
t
o
ad
d
r
ess
a
p
r
o
b
lem
s
im
ilar
to
th
e
o
n
e
at
h
an
d
.
Mo
s
t
s
tu
d
ies
r
eu
s
e
well
-
k
n
o
wn
p
r
e
-
tr
ain
ed
m
o
d
els,
in
cl
u
d
in
g
VGG
-
16
an
d
VGG
-
1
9
,
wh
ich
ar
e
a
d
o
p
ted
b
y
[
3
0
]
,
[
3
1
]
,
R
esNe
t
b
y
[
32
]
,
[
33
]
,
a
n
d
Go
o
g
leNe
t
(
I
n
ce
p
tio
n
v
1
)
by
[
34
]
.
T
r
ain
in
g
th
e
m
o
d
els
f
r
o
m
s
cr
atch
wo
u
ld
s
p
en
d
a
s
er
io
u
s
am
o
u
n
t
o
f
tim
e
as
well
as
co
n
s
u
m
e
h
u
g
e
d
ata.
Sev
er
al
r
ef
er
en
ce
s
h
av
e
s
h
o
wn
r
ev
i
ews
o
n
p
r
e
-
tr
ain
e
d
m
o
d
els,
s
u
ch
as
[
35
]
,
[
36
]
.
I
n
f
ac
t,
th
e
p
r
e
-
t
r
ain
ed
m
o
d
el
is
o
f
ten
u
s
ed
in
co
m
b
in
atio
n
with
o
n
e
o
f
th
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
A
r
ev
iew
o
f d
r
iver d
is
t
r
a
ctio
n
d
etec
tio
n
w
h
ile
d
r
ivin
g
b
a
s
ed
o
n
co
n
vo
lu
tio
n
a
l n
eu
r
a
l
(
Gh
a
d
y
A
lh
a
ma
d
)
4417
“T
r
an
s
f
er
lear
n
in
g
”
o
r
“Fin
e
tu
n
in
g
”
a
p
p
r
o
ac
h
es.
T
h
e
m
aj
o
r
ch
allen
g
e
id
en
tifie
d
th
at
e
ac
h
o
f
th
e
r
e
v
iewe
d
s
tu
d
ies
u
s
ed
d
if
f
er
en
t
d
atasets
to
ac
h
iev
e
th
eir
g
o
als.
I
n
ad
d
it
io
n
,
th
ese
d
atasets
ar
e
o
f
ten
lim
ited
,
an
d
s
o
m
e
o
f
th
em
ar
e
n
o
t
p
u
b
lic.
I
m
ag
eN
et
is
o
n
e
o
f
th
e
m
o
s
t
p
o
p
u
lar
p
u
b
lic
d
atasets
.
T
h
e
I
m
ag
eN
et
p
r
o
ject
is
a
lar
g
e
v
is
u
al
d
atab
ase
d
esig
n
ed
f
o
r
u
s
e
in
v
is
u
al
o
b
ject
r
ec
o
g
n
itio
n
s
o
f
twar
e
r
esear
ch
.
Mo
r
e
th
an
1
4
m
illi
o
n
im
ag
es
h
av
e
b
ee
n
h
an
d
-
an
n
o
tated
b
y
th
e
p
r
o
ject
to
id
en
tif
y
th
e
o
b
jects
p
r
esen
t
in
ea
ch
im
a
g
e,
an
d
f
o
r
o
v
er
o
n
e
m
illi
o
n
im
ag
es,
b
o
u
n
d
in
g
b
o
x
es
ar
e
also
in
clu
d
ed
.
I
m
ag
eNe
t
en
co
m
p
ass
es
m
o
r
e
th
an
2
0
,
0
0
0
ca
teg
o
r
ies
[
37
]
.
T
h
er
e
a
r
e
also
d
atasets
th
at
a
r
e
m
o
r
e
r
ele
v
an
t
t
o
a
s
p
ec
if
i
c
to
p
ic,
s
u
ch
as
StateFar
m
f
o
r
d
is
tr
ac
ted
d
r
i
v
er
d
etec
tio
n
th
at
is
u
s
ed
b
y
Om
e
r
u
s
tao
g
lu
et
a
l.
[
38
]
.
Du
e
to
t
h
e
lim
ited
d
atasets
,
s
o
m
e
r
esear
ch
er
s
ten
d
to
tr
ain
th
eir
m
o
d
els o
n
th
eir
o
w
n
d
ata
s
et
s
u
ch
as
[
39
]
.
T
h
is
p
ap
er
will
p
r
o
v
id
e
a
r
e
v
i
ew
o
f
d
r
iv
e
r
d
is
tr
ac
tio
n
d
etec
t
io
n
,
co
n
s
id
er
i
n
g
th
e
r
elian
ce
o
n
C
NNs
at
all
s
tag
es,
lev
er
ag
in
g
ap
p
r
o
p
r
iate
tech
n
iq
u
es,
a
n
d
p
r
o
p
o
s
in
g
n
ew
a
p
p
r
o
ac
h
es
in
th
is
f
iel
d
.
T
h
is
r
e
v
iew
will
d
is
cu
s
s
d
r
iv
er
d
is
tr
ac
tio
n
b
ased
o
n
C
NN
with
f
o
cu
s
in
g
o
n
n
etwo
r
k
in
p
u
t,
p
r
e
-
tr
ain
ed
m
o
d
els,
d
atasets
,
an
d
m
eth
o
d
o
l
o
g
ies.
Fin
ally
,
r
ec
o
m
m
en
d
atio
n
s
will
b
e
p
r
esen
ted
in
co
n
clu
s
io
n
to
d
er
iv
e
n
ew
m
eth
o
d
s
f
o
r
d
etec
tin
g
d
r
iv
e
r
d
is
tr
ac
tio
n
u
s
in
g
C
NN
tech
n
iq
u
es.
2.
NE
T
WO
RK
I
NP
U
T
C
NN
is
s
p
ec
if
ically
d
esig
n
ed
to
p
r
o
ce
s
s
in
p
u
t
im
ag
es.
T
h
ese
im
ag
es
ar
e
ex
tr
ac
ted
f
r
o
m
v
id
eo
f
r
am
es
o
f
d
r
iv
er
s
wh
ile
d
r
iv
in
g
.
Vid
e
o
s
ar
e
ca
p
tu
r
e
d
b
y
ca
m
er
as
m
o
u
n
te
d
in
th
e
ca
r
.
T
h
e
r
esear
ch
e
r
s
ty
p
ically
p
r
ep
ar
e
t
h
e
in
p
u
t
im
ag
es
d
u
r
in
g
a
p
r
e
-
p
r
o
ce
s
s
in
g
s
tep
,
th
en
f
ee
d
th
em
in
to
th
e
n
etwo
r
k
.
T
h
e
y
u
s
e
v
ar
io
u
s
tech
n
iq
u
es
to
ac
h
iev
e
th
at.
Pre
p
r
o
ce
s
s
in
g
th
e
i
n
p
u
t
im
ag
es
o
cc
u
r
s
m
o
r
e
o
r
less
in
t
h
e
f
o
llo
win
g
s
et
o
f
co
m
m
o
n
s
tag
es,
as sh
o
wn
in
Fig
u
r
e
1
.
Fig
u
r
e
1
.
I
n
p
u
t i
m
ag
es p
r
e
-
p
r
o
c
e
s
s
in
g
2
.
1
.
Video
a
cquis
it
io
n
/s
pecif
ica
t
io
n
D
u
r
i
n
g
t
h
i
s
s
t
a
g
e
,
t
h
e
v
i
d
e
o
o
b
t
a
i
n
e
d
f
r
o
m
a
f
i
x
e
d
c
a
m
e
r
a
o
r
s
m
a
r
t
p
h
o
n
e
i
s
d
i
v
i
d
e
d
i
n
t
o
i
n
d
i
v
i
d
u
a
l
f
r
a
m
e
s
f
o
r
f
u
r
t
h
e
r
a
n
a
l
y
s
i
s
[
4
0
]
.
V
i
d
e
o
i
s
c
a
p
t
u
r
e
d
i
n
t
w
o
w
a
y
s
.
T
h
e
f
i
r
s
t
i
s
s
u
it
a
b
l
e
f
o
r
v
i
s
u
a
l
d
i
s
t
r
a
c
ti
o
n
s
s
u
c
h
a
s
d
r
o
ws
i
n
es
s
d
et
e
c
ti
o
n
s
y
s
t
em
s
,
w
h
e
r
e
o
n
l
y
t
h
e
d
r
i
v
e
r
’
s
f
a
c
e
is
c
a
p
t
u
r
e
d
.
T
h
e
s
e
c
o
n
d
is
s
u
i
t
a
b
l
e
f
o
r
o
t
h
e
r
d
i
s
t
r
a
ct
i
o
n
d
e
t
e
c
t
i
o
n
s
y
s
t
e
m
s
,
w
h
i
c
h
c
a
n
u
s
e
m
u
l
t
i
p
l
e
c
a
m
e
r
a
s
i
n
d
i
f
f
e
r
e
n
t
d
i
r
e
c
t
i
o
n
s
.
T
h
e
t
a
r
g
e
t
i
s
m
o
s
t
l
y
t
h
e
h
a
n
d
s
a
n
d
h
e
a
d
.
T
r
a
n
e
t
a
l
.
[
4
1
]
p
r
o
p
o
s
e
s
a
d
e
e
p
l
e
a
r
n
i
n
g
ap
p
r
o
a
c
h
t
o
d
e
t
e
ct
i
n
g
d
is
t
r
a
ct
ed
d
r
i
v
i
n
g
b
e
h
a
v
i
o
r
s
.
T
h
e
i
m
a
g
e
s
c
a
p
t
u
r
e
d
u
s
i
n
g
t
wo
c
a
m
e
r
a
s
,
b
y
w
h
i
c
h
t
h
e
b
o
d
y
m
o
v
e
m
e
n
t
s
a
n
d
f
a
c
e
o
f
t
h
e
d
r
i
v
e
r
a
r
e
m
o
n
i
t
o
r
e
d
r
e
s
p
e
c
t
i
v
el
y
.
C
h
e
n
et
a
l
.
[
2
4
]
p
r
e
s
e
n
t
e
d
a
d
r
i
v
e
r
d
r
o
ws
i
n
es
s
est
i
m
a
t
i
o
n
s
y
s
te
m
.
T
h
e
y
u
s
e
a
m
o
b
i
l
e
p
h
o
n
e
c
a
m
e
r
a
p
o
s
i
t
i
o
n
e
d
i
n
f
r
o
n
t
o
f
t
h
e
d
r
i
v
e
r
'
s
f
a
c
e
t
o
ca
p
t
u
r
e
v
a
r
i
o
u
s
s
h
o
t
s
o
f
f
a
c
i
a
l
e
x
p
r
e
s
s
i
o
n
s
,
i
n
c
l
u
d
i
n
g
e
y
e
s
a
n
d
m
o
u
t
h
.
T
h
e
v
id
e
o
s
p
ec
if
icatio
n
is
r
e
lated
to
v
id
e
o
s
p
ee
d
(
f
r
a
m
es
p
er
s
ec
o
n
d
(
FP
S
)
)
n
u
m
b
e
r
o
f
f
r
a
m
es,
in
f
r
ar
ed
ca
m
er
a
(
I
R
)
,
g
r
a
y
s
ca
le
o
r
R
GB
,
an
d
im
a
g
e
r
eso
lu
tio
n
.
All
o
f
th
em
co
u
ld
af
f
ec
t
d
ir
ec
tly
th
e
r
esu
lt
o
f
m
o
d
el.
Acc
o
r
d
in
g
to
Ma
g
án
et
a
l.
[
42
]
,
it
is
cr
u
cial
t
o
d
e
ter
m
in
e
th
e
a
p
p
r
o
p
r
iate
f
r
am
e
r
ate
at
wh
ic
h
th
e
ca
m
er
a
co
m
m
u
n
icate
s
with
th
e
s
y
s
tem
.
A
h
ig
h
f
r
am
e
r
ate
m
ay
in
cr
ea
s
e
s
y
s
tem
lo
ad
d
u
e
to
th
e
lar
g
e
n
u
m
b
e
r
o
f
FP
S
th
at
r
eq
u
ir
e
p
r
o
ce
s
s
in
g
,
wh
er
ea
s
a
lo
w
FP
S
ca
n
ad
v
er
s
ely
af
f
ec
t
s
y
s
tem
p
er
f
o
r
m
an
ce
.
Fo
r
in
s
tan
ce
,
g
iv
en
th
at
th
e
av
er
ag
e
b
lin
k
d
u
r
atio
n
r
an
g
es
f
r
o
m
1
0
0
to
4
0
0
m
s
[
42
]
,
a
f
r
am
e
r
ate
o
f
1
0
FP
S
m
ay
s
u
f
f
ice
to
d
etec
t
b
lin
k
s
wh
ile
p
r
ev
e
n
tin
g
s
y
s
tem
o
v
er
lo
ad
.
T
h
e
to
tal
n
u
m
b
e
r
o
f
f
r
am
es
d
ep
e
n
d
s
o
n
b
o
t
h
th
e
v
id
e
o
d
u
r
atio
n
an
d
th
e
FP
S.
Ma
g
á
n
et
a
l.
[
42
]
ev
al
u
ate
6
0
0
f
r
am
es
ea
ch
tim
e
a
n
ew
f
r
a
m
e
is
ca
p
tu
r
ed
b
y
th
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
1
4
,
No
.
6
,
Dec
em
b
er
2
0
2
5
:
4
4
1
5
-
4
4
2
6
4418
ca
m
er
a.
T
h
ey
e
v
alu
ated
th
e
d
r
iv
er
’
s
d
r
o
wsi
n
ess
lev
el
at
a
g
iv
en
m
o
m
en
t
u
s
in
g
d
ata
co
llected
f
r
o
m
th
e
p
r
ec
ed
in
g
60
s
ec
o
n
d
s
.
So
m
e
r
esear
ch
er
s
ten
d
to
u
s
e
an
i
n
f
r
ar
ed
ca
m
er
a
as
an
I
R
ca
m
er
a
allo
ws
o
n
e
t
o
c
o
n
s
tr
u
ct
v
is
io
n
-
b
ased
s
y
s
tem
s
th
at
ca
n
wo
r
k
i
n
d
if
f
e
r
en
t
li
g
h
t
co
n
d
itio
n
s
,
f
o
r
ex
am
p
le
d
u
r
in
g
th
e
n
ig
h
t
o
r
d
u
r
in
g
b
a
d
wea
th
er
ev
e
n
ts
[
43
]
.
J
ab
b
ar
et
a
l.
[
44
]
r
ec
o
r
d
with
in
f
r
ar
ed
ca
m
er
a
to
o
b
tain
n
i
g
h
t
-
tim
e
v
id
eo
.
T
h
e
r
esu
ltin
g
co
n
ten
t
is
9
.
5
h
o
u
r
s
o
f
v
id
e
o
s
th
at
h
av
e
a
r
eso
lu
t
io
n
o
f
6
4
0
×
4
8
0
at
3
0
FPS
.
W
h
ile
s
o
m
e
r
ely
o
n
R
GB
ca
m
er
as.
I
n
ad
d
itio
n
,
o
th
er
s
co
n
v
er
t
f
r
am
es
to
g
r
ay
s
c
ale,
s
in
ce
m
o
s
t
lib
r
ar
ies
th
at
u
s
ed
to
d
etec
t
o
b
jects
tak
e
g
r
a
y
s
ca
le
im
ag
es,
s
u
ch
a
s
Op
en
C
V.
I
m
ag
e
r
eso
lu
tio
n
s
ar
e
o
f
ten
d
eter
m
in
e
d
o
n
th
e
b
asis
o
f
m
o
d
els
an
d
p
r
o
p
o
s
ed
m
eth
o
d
s
.
T
ab
le
1
s
h
o
ws
h
o
w
th
e
p
a
p
er
s
p
r
ep
o
ce
s
s
ed
th
e
in
p
u
t im
ag
es.
T
ab
le
1
.
Pre
p
r
o
ce
s
s
in
g
th
e
in
p
u
t im
ag
es
-
v
id
eo
s
p
ec
if
icatio
n
A
u
t
h
o
r
s
C
a
mer
a
FPS
F
r
a
mes
n
u
m
b
e
r
I
mag
e
r
e
so
l
u
t
i
o
n
I
mag
e
c
o
l
o
r
B
e
h
a
v
i
o
r
D
i
st
r
a
c
t
i
o
n
t
y
p
e
A
c
c
u
r
a
c
y
(
%)
F
l
o
r
e
z
e
t
a
l
.
[
4
5
]
-
25
7
,
5
0
0
-
3
,
000
1
1
2
×
1
1
2
G
r
a
y
-
sca
l
e
F
a
t
i
g
u
e
V
i
su
a
l
9
9
.
7
1
Jab
b
a
r
e
t
a
l
.
[
4
4
]
M
o
b
i
l
e
c
a
mera
/
o
n
d
a
sh
b
o
a
r
d
30
6
0
0
,
0
0
0
6
4
0
×
4
8
0
I
n
f
r
a
r
e
d
c
a
mer
a
F
a
t
i
g
u
e
V
i
su
a
l
88
G
h
a
z
a
l
e
t
a
l
.
[
4
6
]
D
a
sh
b
o
a
r
d
c
a
mer
a
1
4
.
9
-
1
0
0
×
1
0
0
G
r
a
y
-
sca
l
e
F
a
t
i
g
u
e
V
i
su
a
l
95
Tr
a
n
e
t
a
l
.
[
4
1
]
D
u
a
l
c
a
meras
:
i
n
f
r
o
n
t
o
f
t
h
e
d
r
i
v
e
r
/
r
i
g
h
t
si
d
e
o
f
t
h
e
d
r
i
v
e
r
8
-
64
×
64
R
G
B
F
a
t
i
g
u
e
,
t
e
x
t
i
n
g
,
p
h
o
n
e
,
r
a
d
i
o
,
r
e
a
c
h
i
n
g
b
a
c
k
w
a
r
d
,
a
d
j
u
s
t
i
n
g
h
a
i
r
o
r
m
a
k
e
u
p
,
a
n
d
i
n
t
e
r
a
c
t
i
n
g
w
i
t
h
p
a
s
s
e
n
g
e
r
s
V
i
su
a
l
M
a
n
u
a
l
9
6
.
7
M
a
g
á
n
e
t
a
l
.
[
4
2
]
C
a
mer
a
m
o
u
n
t
e
d
u
n
d
e
r
t
h
e
f
r
o
n
t
mi
r
r
o
r
10
6
0
0
64
×
64
-
F
a
t
i
g
u
e
V
i
su
a
l
Tr
a
i
n
i
n
g
d
a
t
a
:
6
0
Te
st
i
n
g
d
a
t
a
:
6
0
F
u
z
z
y
:
9
3
2
.
2
.
Da
t
a
a
ug
m
ent
a
t
io
n
Data
a
u
g
m
en
tat
io
n
r
ef
er
s
t
o
t
h
e
p
r
o
c
ess
o
f
a
r
t
if
i
ci
all
y
g
e
n
er
ati
n
g
n
ew
d
at
a
f
r
o
m
ex
is
ti
n
g
t
r
ai
n
i
n
g
d
at
ase
ts
.
C
o
m
m
o
n
te
ch
n
i
q
u
es
in
c
lu
d
e
r
esiz
in
g
,
f
li
p
p
in
g
,
r
o
t
a
tin
g
,
c
r
o
p
p
i
n
g
,
an
d
p
a
d
d
i
n
g
.
T
h
is
a
p
p
r
o
ac
h
h
el
p
s
m
iti
g
ate
is
s
u
es
s
u
c
h
as
o
v
er
f
i
tt
in
g
a
n
d
l
im
ite
d
d
at
a
av
ail
a
b
ili
t
y
,
w
h
i
le
e
n
h
a
n
c
in
g
m
o
d
e
l
r
o
b
u
s
t
n
ess
a
n
d
o
v
e
r
all
p
e
r
f
o
r
m
a
n
c
e
[
47
]
.
F
o
r
e
x
a
m
p
l
e,
J
a
b
b
a
r
et
a
l.
[
4
4
]
u
tili
ze
d
C
o
d
eBo
x
t
o
c
r
e
ate
a
d
d
it
io
n
al
i
m
a
g
es
b
y
ap
p
l
y
i
n
g
a
s
er
i
es
o
f
a
u
g
m
e
n
ta
ti
o
n
o
p
er
ati
o
n
s
t
o
t
h
o
s
e
e
x
t
r
a
cte
d
f
r
o
m
v
i
d
e
o
f
r
a
m
es
,
as
il
lu
s
tr
ate
d
in
Fi
g
u
r
e
2
i
n
a
d
d
itio
n
,
T
ab
le
2
s
h
o
ws
o
th
er
d
ata
au
g
m
en
tatio
n
ar
ticles ex
am
p
les.
Fig
u
r
e
2
.
Data
au
g
m
en
ta
tio
n
[
4
4
]
T
ab
le
2
.
Pre
p
r
o
ce
s
s
in
g
th
e
in
p
u
t im
ag
e
-
d
ata
a
u
g
m
e
n
tatio
n
A
u
t
h
o
r
s
D
a
t
a
a
u
g
m
e
n
t
a
t
i
o
n
B
e
h
a
v
i
o
r
D
i
st
r
a
c
t
i
o
n
t
y
p
e
A
c
c
u
r
a
c
y
F
l
o
r
e
z
e
t
a
l
.
[
4
5
]
R
o
t
a
t
i
o
n
r
a
n
g
e
:
2
0
%
H
o
r
i
z
o
n
t
a
l
:
t
r
u
e
F
i
l
l
mo
d
e
:
n
e
a
r
e
s
t
R
e
s
u
l
t
:
5
i
m
a
g
e
s
o
f
e
a
c
h
i
m
a
g
e
F
a
t
i
g
u
e
V
i
su
a
l
9
9
.
7
1
%
C
h
a
n
d
a
n
d
K
a
r
t
h
i
k
e
y
a
n
[
1
4
]
B
r
i
g
h
t
n
e
ss ra
n
g
e
:
7
5
%
R
o
t
a
t
i
o
n
i
n
t
e
r
v
a
l
:
±
2
d
e
g
r
e
e
S
h
e
e
r
r
a
n
g
e
:
±
2
%
Zo
o
m
t
r
a
n
sf
o
r
ma
t
i
o
n
i
n
t
e
r
v
a
l
:
±
2
%
F
a
t
i
g
u
e
,
d
r
u
n
k
e
n
,
r
e
c
k
l
e
ss
V
i
su
a
l
C
o
g
n
i
t
i
v
e
9
3
%
C
h
e
n
e
t
a
l
.
[
4
]
S
p
a
t
i
a
l
s
t
r
e
a
m
:
c
r
o
p
p
i
n
g
,
r
o
t
a
t
i
n
g
,
h
o
r
i
z
o
n
t
a
l
f
l
i
p
p
i
n
g
,
a
n
d
sh
i
f
t
i
n
g
.
Te
mp
o
r
a
l
s
t
r
e
a
m
:
r
a
n
d
o
m
c
r
o
p
p
i
n
g
a
n
d
h
o
r
i
z
o
n
t
a
l
f
l
i
p
p
i
n
g
.
T
e
x
t
i
n
g
,
p
h
o
n
e
,
r
a
d
i
o
,
r
e
a
c
h
i
n
g
,
r
e
a
c
h
i
n
g
b
a
c
k
w
a
r
d
,
a
d
j
u
s
t
i
n
g
h
a
i
r
o
r
m
a
k
e
u
p
,
a
n
d
i
n
t
e
r
a
c
t
i
n
g
w
i
t
h
p
a
s
s
e
n
g
e
r
s
M
a
n
u
a
l
En
h
a
n
c
e
t
h
e
a
c
c
u
r
a
c
y
b
y
a
p
p
r
o
x
i
m
a
t
e
l
y
3
0
%
A
y
t
e
k
i
n
a
n
d
M
e
n
ç
i
k
[
3
0
]
R
a
n
d
o
m
r
o
t
a
t
i
o
n
F
a
t
i
g
u
e
V
i
su
a
l
9
1
%
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
A
r
ev
iew
o
f d
r
iver d
is
t
r
a
ctio
n
d
etec
tio
n
w
h
ile
d
r
ivin
g
b
a
s
ed
o
n
co
n
vo
lu
tio
n
a
l n
eu
r
a
l
(
Gh
a
d
y
A
lh
a
ma
d
)
4419
2
.
3
.
F
e
a
t
ure
ex
t
r
a
ct
io
n
I
t
i
s
a
m
e
t
h
o
d
o
f
C
NN
t
h
a
t
r
ec
o
g
n
i
z
e
s
k
e
y
p
a
t
te
r
n
s
i
n
a
n
im
a
g
e
f
o
r
c
l
a
s
s
i
f
i
ca
t
i
o
n
p
u
r
p
o
s
e
s
.
F
ea
t
u
r
e
e
x
t
r
a
c
t
i
o
n
i
n
d
r
o
w
s
i
n
ess
d
e
t
ec
t
i
o
n
s
y
s
t
e
m
s
d
e
p
e
n
d
s
m
a
i
n
l
y
o
n
e
x
t
r
a
c
t
i
n
g
t
h
e
c
o
o
r
d
i
n
a
t
es
o
f
f
a
c
i
a
l
l
a
n
d
m
a
r
k
s
f
r
o
m
t
h
e
i
m
a
g
e
,
s
u
c
h
a
s
e
y
es
a
n
d
m
o
u
t
h
.
T
h
is
a
i
m
s
t
o
cl
a
s
s
i
f
y
d
r
i
v
e
r
f
at
i
g
u
e
t
h
r
o
u
g
h
y
a
w
n
i
n
g
a
n
d
e
y
e
s
b
l
i
n
k
i
n
g
,
s
u
c
h
a
s
[
2
4
]
,
[
4
2
]
,
[
4
6
]
.
A
s
f
o
r
o
t
h
e
r
d
r
i
v
e
r
d
is
t
r
a
c
t
i
o
n
d
e
te
c
t
i
o
n
s
y
s
t
e
m
s
,
t
h
e
y
a
r
e
n
o
t
l
i
m
it
e
d
to
e
x
t
r
a
c
t
i
n
g
t
h
e
f
a
ci
a
l
l
a
n
d
m
a
r
k
s
.
Z
h
a
n
g
et
a
l
.
[
4
8
]
e
x
t
r
a
ct
ed
f
a
c
i
a
l
,
m
o
u
t
h
,
a
n
d
h
a
n
d
f
ea
t
u
r
e
s
f
r
o
m
i
m
a
g
es
c
a
p
t
u
r
e
d
b
y
a
ca
m
e
r
a
m
o
u
n
t
ed
o
n
t
h
e
v
e
h
i
cl
e
’
s
d
a
s
h
b
o
a
r
d
.
T
r
a
n
e
t
a
l
.
[
4
1
]
e
x
t
r
a
ct
e
d
f
a
cia
l
a
n
d
b
o
d
y
f
e
a
t
u
r
es
u
s
i
n
g
t
w
o
ca
m
e
r
a
s
,
o
n
e
m
o
u
n
t
e
d
i
n
f
r
o
n
t
o
f
t
h
e
d
r
i
v
e
r
a
n
d
t
h
e
o
t
h
e
r
o
n
t
h
e
d
r
i
v
e
r
'
s
r
i
g
h
t
s
id
e
.
F
i
g
u
r
e
3
s
h
o
ws
f
a
c
e
a
n
d
e
y
e
s
f
e
a
t
u
r
e
e
x
t
r
ac
t
io
n
a
n
d
F
i
g
u
r
e
4
s
h
o
w
s
b
o
d
y
f
e
a
t
u
r
e
e
x
t
r
a
c
t
i
o
n
e
x
a
m
p
le
s
.
S
t
u
d
i
es
u
s
e
v
a
r
i
o
u
s
t
e
c
h
n
i
q
u
e
s
t
o
e
x
t
r
a
ct
t
h
e
s
e
f
ea
t
u
r
e
s
,
Y
a
n
e
t
a
l
.
[
2
1
]
i
n
v
e
s
ti
g
a
t
e
t
h
e
c
o
n
t
e
x
t
u
a
l
c
u
es
t
h
a
t
m
o
s
t
s
i
g
n
i
f
ic
a
n
t
ly
i
n
f
l
u
e
n
c
e
s
p
e
ci
f
i
c
d
r
i
v
e
r
a
ct
i
o
n
s
b
y
a
n
a
l
y
z
i
n
g
s
k
i
n
-
l
i
k
e
r
e
g
io
n
s
.
S
k
i
n
r
e
g
i
o
n
s
a
r
e
i
n
i
ti
a
ll
y
e
x
t
r
a
c
t
e
d
u
s
i
n
g
a
G
a
u
s
s
i
a
n
m
i
x
t
u
r
e
m
o
d
e
l
(
G
MM
)
t
r
a
i
n
e
d
o
n
s
k
i
n
i
m
a
g
e
s
,
a
n
d
s
u
b
s
e
q
u
e
n
t
l
y
p
r
o
c
e
s
s
e
d
b
y
R
*
C
N
N
t
o
c
l
a
s
s
i
f
y
t
h
e
d
r
i
v
e
r
’
s
a
c
t
i
o
n
s
.
Fig
u
r
e
3
.
Face
o
r
ey
es f
ea
tu
r
e
ex
tr
ac
tio
n
[
4
6
]
Fig
u
r
e
4
.
B
o
d
y
f
ea
tu
r
e
ex
tr
ac
t
io
n
[
2
0
]
S
e
s
h
a
d
r
i
e
t
a
l
.
[
4
9
]
e
m
p
l
o
y
ed
a
h
i
s
t
o
g
r
a
m
o
f
o
r
i
e
n
t
e
d
g
r
a
d
i
e
n
t
s
(
H
OG
)
i
n
c
o
m
b
i
n
at
i
o
n
w
i
t
h
a
n
A
d
a
B
o
o
s
t
c
l
as
s
i
f
i
e
r
,
t
r
a
i
n
e
d
s
ep
a
r
a
t
e
l
y
f
o
r
e
a
c
h
s
i
d
e
o
f
t
h
e
f
a
c
i
a
l
r
e
g
i
o
n
s
,
t
o
c
a
t
e
g
o
r
i
ze
m
o
b
i
l
e
p
h
o
n
e
u
s
a
g
e
a
s
r
i
g
h
t
h
a
n
d
,
l
e
f
t
h
a
n
d
,
o
r
n
o
n
e
.
D
a
s
e
t
a
l
.
[
5
0
]
i
n
t
r
o
d
u
c
e
d
a
v
i
d
e
o
-
b
a
s
e
d
h
a
n
d
d
e
t
e
c
t
i
o
n
d
a
t
as
e
t
i
n
a
n
a
u
t
o
m
o
t
i
v
e
e
n
v
i
r
o
n
m
e
n
t
a
n
d
u
t
i
li
z
e
d
a
n
ag
g
r
e
g
a
t
e
c
h
a
n
n
e
l
f
e
at
u
r
e
s
(
AC
F
)
o
b
j
e
c
t
d
e
t
e
ct
o
r
.
G
h
a
za
l
et
a
l
.
[
4
6
]
e
m
p
l
o
y
e
d
a
l
e
a
r
n
i
n
g
-
b
a
s
e
d
a
p
p
r
o
a
c
h
f
o
r
f
a
c
e
d
e
t
e
c
ti
o
n
t
h
a
t
r
e
l
ie
s
o
n
t
h
e
H
a
a
r
-
l
i
k
e
f
e
at
u
r
e
s
p
r
o
p
o
s
e
d
b
y
V
i
o
l
a
a
n
d
J
o
n
e
s
,
c
o
m
b
i
n
e
d
w
i
t
h
c
as
c
a
d
e
cl
a
s
s
i
f
ie
r
s
.
T
o
o
p
t
i
m
i
ze
p
r
o
c
e
s
s
i
n
g
e
f
f
i
c
i
e
n
c
y
,
a
n
i
n
t
e
g
r
al
i
m
a
g
e
i
s
u
t
i
l
i
ze
d
t
o
m
i
n
i
m
iz
e
t
h
e
c
o
m
p
u
t
a
t
i
o
n
a
l
c
o
s
t
o
f
i
n
i
t
ia
l
i
m
a
g
e
p
r
o
c
e
s
s
i
n
g
a
n
d
t
o
e
n
a
b
l
e
f
a
s
t
c
a
l
c
u
la
t
i
o
n
o
f
r
e
c
t
a
n
g
u
l
a
r
f
e
a
t
u
r
e
s
.
T
h
e
v
a
l
u
e
o
f
t
h
e
i
n
t
e
g
r
al
i
m
a
g
e
at
a
n
y
p
o
i
n
t
(
x
,
y
)
is
c
o
m
p
u
t
ed
i
n
a
s
i
n
g
le
p
as
s
o
v
e
r
t
h
e
o
r
i
g
i
n
a
l
i
m
a
g
e
a
n
d
r
e
p
r
e
s
e
n
t
s
t
h
e
s
u
m
o
f
a
ll
p
i
x
e
l
in
t
e
n
s
i
ti
e
s
l
o
c
a
t
e
d
a
b
o
v
e
a
n
d
t
o
t
h
e
l
e
f
t
o
f
(
x
,
y
)
,
a
s
e
x
p
r
es
s
e
d
in
(
1
)
.
(
,
)
=
(
,
)
+
(
−
1
,
)
+
(
,
−
1
)
+
(
−
1
,
−
1
)
,
(
1
)
W
h
er
e
th
e
in
teg
r
al
im
ag
e
is
∑
(
′
,
′
)
′
≤
,
′
≤
,
a
n
d
i (
x
,
y
)
,
is
th
e
o
r
ig
i
n
al
im
ag
e.
T
o
en
ab
le
r
ap
id
a
n
d
ac
cu
r
ate
f
ac
e
d
etec
tio
n
,
th
e
au
th
o
r
s
ap
p
l
ied
th
e
Ad
aBo
o
s
t a
lg
o
r
ith
m
to
co
n
s
tr
u
ct
a
lig
h
tweig
h
t
y
et
ef
f
ec
ti
v
e
f
a
ce
d
etec
to
r
d
er
iv
e
d
f
r
o
m
th
e
ca
lcu
lated
im
ag
e
f
ea
tu
r
e
v
alu
es
o
b
tain
ed
th
r
o
u
g
h
th
e
in
teg
r
al
im
ag
e
m
eth
o
d
.
Ha
ar
-
lik
e
f
ea
tu
r
es
a
n
d
a
ca
s
ca
d
e
o
f
class
if
ier
s
wer
e
u
tili
ze
d
to
i
d
en
tify
b
o
th
f
ac
ial
r
eg
io
n
s
an
d
s
p
ec
if
ic
s
u
b
r
eg
i
o
n
s
s
u
ch
as
th
e
ey
es
an
d
f
r
o
n
t
al
f
ac
e.
T
h
is
s
ec
tio
n
p
r
o
v
id
es
a
b
r
ief
o
v
er
v
iew
o
f
s
ev
er
al
f
ea
tu
r
e
r
ec
o
g
n
itio
n
tec
h
n
iq
u
es,
with
ad
d
itio
n
al
ex
am
p
les s
u
m
m
ar
ized
in
T
a
b
le
3
.
T
ab
le
3
.
Pre
p
r
o
ce
s
s
in
g
th
e
in
p
u
t im
ag
e
-
f
ea
tu
r
e
ex
tr
ac
tio
n
A
u
t
h
o
r
s
M
e
a
su
r
e
B
e
h
a
v
i
o
r
D
i
st
r
a
c
t
i
o
n
t
y
p
e
A
c
c
u
r
a
c
y
F
l
o
r
e
z
e
t
a
l
.
[
4
5
]
Ey
e
F
a
t
i
g
u
e
V
i
su
a
l
9
9
.
7
1
%
C
h
a
n
d
a
n
d
K
a
r
t
h
i
k
e
y
a
n
[
1
4
]
Ey
e
u
n
d
e
r
c
o
n
d
i
t
i
o
n
s
l
i
k
e
:
f
a
t
i
g
u
e
,
d
r
u
n
k
e
n
n
e
ss
a
n
d
a
g
g
r
e
ss
i
o
n
F
a
t
i
g
u
e
,
d
r
u
n
k
e
n
,
r
e
c
k
l
e
ss
V
i
su
a
l
C
o
g
n
i
t
i
v
e
93
%
C
h
e
n
e
t
a
l
.
[
4
]
D
i
st
r
a
c
t
i
o
n
b
e
h
a
v
i
o
r
Te
x
t
i
n
g
,
p
h
o
n
e
,
r
a
d
i
o
,
r
e
a
c
h
i
n
g
b
a
c
k
w
a
r
d
,
a
d
j
u
st
i
n
g
h
a
i
r
o
r
m
a
k
e
u
p
,
a
n
d
i
n
t
e
r
a
c
t
i
n
g
w
i
t
h
p
a
ss
e
n
g
e
r
s
M
a
n
u
a
l
I
n
c
r
e
a
se
t
h
e
a
c
c
u
r
a
c
y
r
a
t
e
b
y
n
e
a
r
l
y
3
0
%
Jab
b
a
r
e
t
a
l
.
[
4
4
]
Y
a
w
n
i
n
g
,
r
e
d
u
c
e
d
b
l
i
n
k
r
a
t
e
,
h
e
a
d
n
o
d
d
i
n
g
,
c
o
n
v
e
r
sa
t
i
o
n
s
,
a
n
d
d
r
o
w
s
y
e
y
e
s
F
a
t
i
g
u
e
V
i
su
a
l
88
%
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
1
4
,
No
.
6
,
Dec
em
b
er
2
0
2
5
:
4
4
1
5
-
4
4
2
6
4420
2.
4
.
Reg
i
o
n o
f
inte
r
est
T
h
e
r
e
g
i
o
n
o
f
i
n
t
er
est
(
R
OI
)
t
y
p
ic
all
y
r
e
f
e
r
s
t
o
t
h
e
s
ig
n
i
f
i
ca
n
t
an
d
r
el
e
v
a
n
t
p
o
r
ti
o
n
s
o
f
a
n
im
a
g
e
.
B
y
f
o
c
u
s
i
n
g
o
n
th
e
R
O
I
,
u
n
n
ec
es
s
ar
y
p
r
o
c
ess
i
n
g
o
f
n
o
n
-
ess
e
n
ti
al
im
ag
e
a
r
e
as
ca
n
b
e
a
v
o
id
e
d
,
th
er
e
b
y
i
m
p
r
o
v
i
n
g
co
m
p
u
tati
o
n
a
l
e
f
f
ic
ie
n
c
y
[
51
]
.
Fi
g
u
r
e
5
i
ll
u
s
tr
at
es
e
x
a
m
p
l
es
o
f
R
O
I
ex
tr
ac
ti
o
n
u
s
ed
i
n
d
i
f
f
er
e
n
t
s
t
u
d
i
es.
Flo
r
e
z
et
a
l.
[
4
5
]
p
r
o
p
o
s
e
d
a
m
et
h
o
d
f
o
r
d
e
tec
ti
n
g
d
r
i
v
e
r
d
r
o
wsi
n
ess
b
y
co
n
c
e
n
t
r
at
in
g
o
n
t
h
e
e
y
e
r
e
g
i
o
n
,
as
ill
u
s
tr
at
ed
i
n
F
ig
u
r
e
5
(
a)
.
Ou
t
o
f
t
h
e
4
6
8
la
n
d
m
ar
k
s
i
d
e
n
ti
f
i
e
d
d
u
r
in
g
t
h
e
f
ac
i
al
la
n
d
m
ar
k
d
ete
cti
o
n
p
h
ase
,
o
n
l
y
f
o
u
r
p
o
i
n
ts
w
e
r
e
s
e
lec
te
d
t
o
d
ef
i
n
e
th
e
R
OI
.
I
n
t
h
e
Me
d
i
aPi
p
e
f
a
ce
m
es
h
m
o
d
el
,
t
h
e
s
e
lec
ted
p
o
i
n
ts
w
er
e
6
3
,
1
1
7
,
2
9
3
,
an
d
3
4
6
,
wh
ic
h
,
w
h
en
c
o
n
n
ec
t
ed
,
f
o
r
m
an
i
r
r
eg
u
l
ar
q
u
a
d
r
il
ate
r
a
l
r
e
p
r
ese
n
t
in
g
t
h
e
R
O
I
.
T
h
en
,
th
ey
p
r
o
p
o
s
e
d
a
m
e
th
o
d
f
o
r
R
O
I
co
r
r
ec
ti
o
n
to
c
o
r
r
e
ct
th
e
s
el
ec
ted
ir
r
eg
u
l
ar
a
r
ea
.
Ses
h
ad
r
i
et
a
l.
[
4
9
]
ex
tr
ac
ted
an
o
th
er
R
OI
,
with
d
if
f
er
en
t
m
eth
o
d
.
T
h
e
Vio
la
an
d
J
o
n
es
f
ac
e
d
e
tectio
n
alg
o
r
ith
m
was
em
p
lo
y
ed
to
id
e
n
tify
th
e
s
u
b
ject’
s
f
ac
e
in
th
e
in
itial
f
r
am
e
o
f
a
v
id
e
o
.
T
h
e
f
ac
ial
lan
d
m
ar
k
s
lo
ca
lized
in
th
is
f
ir
s
t
f
r
am
e
wer
e
th
en
u
s
ed
as
in
itializatio
n
p
o
in
ts
f
o
r
tr
ac
k
in
g
in
s
u
b
s
eq
u
en
t
f
r
am
es.
A
to
tal
o
f
4
9
f
a
cial
lan
d
m
ar
k
s
wer
e
d
etec
ted
b
y
th
e
alg
o
r
ith
m
.
Du
r
in
g
th
e
tr
ain
in
g
p
h
ase,
r
e
g
io
n
s
o
f
i
n
ter
est
wer
e
cr
o
p
p
e
d
f
o
r
b
o
th
p
o
s
itiv
e
an
d
n
e
g
a
tiv
e
class
s
am
p
les
b
ased
o
n
th
e
f
ac
ial
alig
n
m
e
n
t
r
esu
lts
.
As
illu
s
tr
ated
in
Fi
g
u
r
e
5
(
b
)
,
r
ec
tan
g
u
lar
cr
o
p
s
o
f
s
ize
5
0
×8
0
wer
e
g
en
er
ated
,
u
s
in
g
lan
d
m
a
r
k
1
8
as
th
e
to
p
-
r
ig
h
t
c
o
r
n
er
a
n
d
l
an
d
m
ar
k
2
3
as
th
e
to
p
-
lef
t
co
r
n
er
o
f
t
h
e
cr
o
p
p
ed
r
eg
io
n
.
T
h
ese
r
e
g
io
n
s
wer
e
u
t
ilized
to
cr
ea
te
p
o
s
itiv
e
an
d
n
eg
ativ
e
class
s
am
p
les
co
r
r
esp
o
n
d
in
g
to
ca
s
es
in
wh
ich
s
u
b
jects
wer
e
eith
er
h
o
ld
in
g
o
r
n
o
t
h
o
ld
in
g
a
ce
ll
p
h
o
n
e
in
th
eir
lef
t
o
r
r
ig
h
t
h
an
d
.
T
ab
le
4
p
r
esen
ts
a
lis
t o
f
p
ap
er
s
with
r
eg
io
n
s
o
f
i
n
ter
est an
d
f
ac
ial
m
eth
o
d
u
s
ed
.
(
a)
(
b
)
Fig
u
r
e
5
.
R
OI
ex
tr
ac
tio
n
o
f
(
a)
ey
es o
p
en
a
n
d
e
y
es c
lo
s
ed
an
d
(
b
)
r
ig
h
t
h
an
d
a
n
d
lef
t
h
an
d
[
4
5
]
,
[
4
9
]
T
ab
le
4
.
Pre
p
r
o
ce
s
s
in
g
th
e
in
p
u
t im
ag
e
-
R
OI
ex
tr
ac
tio
n
A
u
t
h
o
r
s
F
a
c
i
a
l
met
h
o
d
R
O
I
B
e
h
a
v
i
o
r
D
i
st
r
a
c
t
i
o
n
t
y
p
e
A
c
c
u
r
a
c
y
(
%)
C
h
e
n
e
t
a
l
.
[
2
4
]
C
N
N
-
LSTM
F
a
c
e
/
e
y
e
s
a
n
d
mo
u
t
h
F
a
t
i
g
u
e
V
i
su
a
l
75
C
h
i
r
r
a
e
t
a
l
.
[
5
2
]
H
a
a
r
c
a
sca
d
e
Ey
e
s
F
a
t
i
g
u
e
V
i
su
a
l
9
6
.
4
2
P
a
r
k
e
t
a
l
.
[
5
3
]
VGG
-
F
a
c
e
N
e
t
F
a
c
e
F
a
t
i
g
u
e
V
i
su
a
l
7
3
.
0
6
P
h
a
n
e
t
a
l
.
[
5
4
]
D
l
i
b
F
a
c
e
/
e
y
e
s
a
n
d
mo
u
t
h
F
a
t
i
g
u
e
V
i
su
a
l
97
H
a
sh
e
mi
e
t
a
l
.
[
3
1
]
V
i
o
l
a
a
n
d
J
o
n
e
s
F
a
c
e
/
e
y
e
s
F
a
t
i
g
u
e
V
i
su
a
l
A
v
g
.
9
6
.
4
2
S
e
sh
a
d
r
i
e
t
a
l
.
[
4
9
]
S
u
p
e
r
v
i
se
d
d
e
s
c
e
n
t
met
h
o
d
(
S
D
M
)
a
l
g
o
r
i
t
h
m
H
a
n
d
s
U
si
n
g
c
e
l
l
p
h
o
n
e
M
a
n
u
a
l
-
3.
P
RE
-
T
RA
I
N
E
D
M
O
D
E
L
S
Pre
-
tr
ain
ed
C
NNs
r
ef
er
to
m
o
d
els
th
at
h
av
e
b
ee
n
p
r
e
v
io
u
s
ly
tr
ain
e
d
o
n
lar
g
e
-
s
ca
le
d
atasets
(
e.
g
.
,
I
m
ag
eNe
t)
an
d
s
u
b
s
eq
u
en
tly
ad
a
p
ted
to
p
er
f
o
r
m
d
i
v
er
s
e
task
s
s
u
ch
as
im
ag
e
class
if
icatio
n
,
o
b
ject
d
etec
tio
n
,
a
n
d
f
ac
ial
r
ec
o
g
n
iti
o
n
[
55
]
.
Fo
r
ex
a
m
p
le,
T
h
e
VGG
-
1
6
ar
c
h
itectu
r
e
is
a
p
r
e
-
tr
ain
ed
C
NN
d
esig
n
e
d
f
o
r
im
ag
e
class
if
icatio
n
.
T
r
ai
n
ed
o
n
a
lar
g
e
-
s
ca
le
an
d
d
iv
er
s
e
d
ataset,
it
ca
n
b
e
ef
f
icien
tly
f
in
e
-
tu
n
e
d
to
p
er
f
o
r
m
ac
cu
r
ately
o
n
d
o
m
ain
-
s
p
ec
if
ic
im
ag
e
class
if
icatio
n
p
r
o
b
lem
s
[
5
5
]
,
[
5
6
]
.
Pre
-
tr
ain
e
d
m
o
d
els ca
n
s
er
v
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
A
r
ev
iew
o
f d
r
iver d
is
t
r
a
ctio
n
d
etec
tio
n
w
h
ile
d
r
ivin
g
b
a
s
ed
o
n
co
n
vo
lu
tio
n
a
l n
eu
r
a
l
(
Gh
a
d
y
A
lh
a
ma
d
)
4421
as
a
f
o
u
n
d
atio
n
f
o
r
v
a
r
io
u
s
task
s
,
ev
en
wh
en
th
ese
task
s
d
if
f
er
s
ig
n
if
ican
tly
f
r
o
m
th
e
o
r
ig
in
al
tr
ain
in
g
o
b
jectiv
e.
T
h
is
ap
p
r
o
ac
h
is
r
ef
er
r
ed
to
as
tr
a
n
s
f
er
lear
n
i
n
g
.
L
e
v
er
ag
in
g
p
r
e
-
tr
ain
e
d
m
o
d
els
allo
ws
f
o
r
co
n
s
id
er
ab
le
s
av
in
g
s
in
b
o
t
h
tim
e
an
d
co
m
p
u
tatio
n
al
r
eso
u
r
ce
s
th
at
wo
u
ld
o
th
er
wis
e
b
e
n
ee
d
ed
to
tr
ain
a
n
eu
r
al
n
etwo
r
k
f
r
o
m
s
cr
atch
.
Mo
r
eo
v
er
,
b
y
f
in
e
-
tu
n
in
g
th
e
p
r
e
-
tr
ain
ed
m
o
d
el
o
n
a
s
p
ec
if
i
c
d
ataset,
it
ca
n
b
e
ad
ap
ted
t
o
a
p
ar
ticu
lar
p
r
o
b
l
em
,
o
f
te
n
r
e
q
u
ir
in
g
s
u
b
s
tan
ti
ally
less
d
ata
[
55
]
.
T
ab
le
5
p
r
esen
ts
co
m
p
ar
is
o
n
b
etwe
en
th
e
m
o
s
t
f
am
o
u
s
p
r
e
-
tr
ain
ed
m
o
d
els.
T
h
e
s
tu
d
ie
s
r
eu
s
ed
p
r
e
-
tr
ain
ed
m
o
d
els
in
th
eir
p
u
r
p
o
s
ed
m
eth
o
d
s
.
T
h
e
y
s
o
m
etim
es
ten
d
to
u
s
e
m
o
r
e
t
h
an
o
n
e
m
o
d
e
l
at
th
e
s
am
e
tim
e
in
co
m
p
a
r
i
n
g
th
e
ac
c
u
r
ac
y
o
f
th
eir
r
esu
lts
.
So
m
e
ex
am
p
les a
r
e
p
r
esen
ted
in
T
ab
le
6
.
T
ab
le
5
.
C
o
m
p
a
r
is
o
n
o
f
p
r
e
-
t
r
ain
ed
m
o
d
els
[
5
6
]
M
o
d
e
l
/
P
a
p
e
r
S
i
z
e
(
MB
)
P
a
r
a
me
t
e
r
s
To
p
-
1
a
c
c
u
r
a
c
y
(
%)
To
p
-
5
a
c
c
u
r
a
c
y
(
%
)
D
e
p
t
h
D
e
f
a
u
l
t
i
ma
g
e
s
i
z
e
X
c
e
p
t
i
o
n
/
C
h
o
l
l
e
t
e
t
a
l
.
[
5
7
]
88
2
2
,
9
1
0
,
4
8
0
79
9
4
.
5
1
2
6
2
2
9
×
2
2
9
VGG
-
1
6
/
S
i
mo
n
y
a
n
e
t
a
l
.
[
5
8
]
5
2
8
1
3
8
,
3
5
7
,
5
4
4
7
1
.
3
9
0
.
1
23
2
2
4
×
2
2
4
VGG
-
19/
S
i
mo
n
y
a
n
a
n
d
Z
i
sserm
a
n
[
5
8
]
5
4
9
1
4
3
,
6
6
7
,
2
4
0
7
1
.
3
90
26
2
2
4
×
2
2
4
R
e
sN
e
t
-
5
0
/
H
e
e
t
a
l
.
[
5
9
]
98
2
5
,
6
3
6
,
7
1
2
7
4
.
9
9
2
.
1
-
2
2
4
×
2
2
4
I
n
c
e
p
t
i
o
n
-
V
3
/
S
z
e
g
e
d
y
e
t
a
l
.
[
6
0
]
92
2
3
,
8
5
1
,
7
8
4
7
7
.
9
9
3
.
7
1
5
9
2
2
9
×
2
2
9
M
o
b
i
l
e
N
e
t
-
V
2
/
S
a
n
d
l
e
r
e
t
a
l
.
[
6
1
]
14
3
,
5
3
8
,
9
8
4
7
1
.
3
9
0
.
1
88
2
2
4
×
2
2
4
T
ab
le
6
.
Pre
–
tr
ai
n
ed
m
o
d
els u
s
ed
in
s
tu
d
ies
A
u
t
h
o
r
s
M
e
a
su
r
e
Pre
-
t
r
a
i
n
e
d
m
o
d
e
l
s
B
e
h
a
v
i
o
r
D
i
st
r
a
c
t
i
o
n
t
y
p
e
A
c
c
u
r
a
c
y
(
%)
G
u
e
t
a
l
.
[
1
7
]
Ey
e
/
m
o
u
t
h
A
l
e
x
N
e
t
,
R
e
sN
e
t
,
M
S
P
-
N
ET
(
d
e
v
e
l
o
p
e
d
n
e
w
)
F
a
t
i
g
u
e
V
i
su
a
l
9
7
.
1
2
/
9
8
.
6
,
9
7
.
8
/
9
8
.
5
,
9
8
.
1
/
9
8
.
9
F
l
o
r
e
z
e
t
a
l
.
[
4
5
]
Ey
e
R
e
s
N
e
t
-
50
-
V2
,
I
n
c
e
p
t
i
o
n
-
V3
,
VGG
-
16
F
a
t
i
g
u
e
V
i
su
a
l
9
9
.
7
,
9
9
.
3
,
9
9
.
4
C
h
e
n
e
t
a
l
.
[
4
]
D
i
st
r
a
c
t
i
o
n
b
e
h
a
v
i
o
r
VGG
-
16
T
e
x
t
i
n
g
,
p
h
o
n
e
,
r
a
d
i
o
,
r
e
a
c
h
i
n
g
b
a
c
k
w
a
r
d
,
a
d
j
u
s
t
i
n
g
h
a
i
r
o
r
m
a
k
e
u
p
,
a
n
d
i
n
t
e
r
a
c
t
i
n
g
w
i
t
h
p
a
s
s
e
n
g
e
r
s
M
a
n
u
a
l
I
n
c
r
e
a
se
t
h
e
a
c
c
u
r
a
c
y
r
a
t
e
b
y
n
e
a
r
l
y
3
0
%
A
y
t
e
k
i
n
e
t
a
l
.
[
3
0
]
Ey
e
a
n
d
mo
u
t
h
VGG
-
16
F
a
t
i
g
u
e
V
i
su
a
l
91
4.
DATAS
E
T
S
On
e
o
f
th
e
m
ain
ch
allen
g
es
in
th
is
p
r
o
ce
s
s
i
s
id
en
tify
in
g
a
s
u
f
f
icien
tly
lar
g
e
p
u
b
lic
d
ataset
th
at
ad
eq
u
ately
r
ep
r
esen
ts
th
e
ex
p
ec
ted
o
u
tco
m
es
f
o
r
s
u
ch
s
y
s
tem
s
.
T
h
er
ef
o
r
e,
th
is
s
ec
tio
n
p
r
o
v
id
es
a
lis
t
o
f
s
o
m
e
p
u
b
lic
d
atasets
in
th
is
f
ield
with
d
etailed
in
f
o
r
m
atio
n
ab
o
u
t e
ac
h
wh
ich
ca
n
b
e
u
s
ed
as a
b
e
n
ch
m
ar
k
.
i)
Natio
n
al
T
s
in
g
Hu
a
Un
iv
er
s
ity
(
NT
HU
)
d
ataset
:
th
is
d
ata
s
e
t
co
m
p
r
is
es
2
2
s
u
b
jects
f
r
o
m
d
iv
er
s
e
eth
n
ic
b
ac
k
g
r
o
u
n
d
s
,
r
ec
o
r
d
ed
u
n
d
er
b
o
th
d
ay
tim
e
an
d
n
ig
h
ttime
co
n
d
itio
n
s
.
Simu
lated
d
r
iv
in
g
s
ce
n
ar
io
s
in
clu
d
ed
b
eh
a
v
io
r
s
s
u
ch
as
Ya
wn
in
g
,
r
ed
u
ce
d
b
lin
k
r
ate,
h
ea
d
n
o
d
d
in
g
,
c
o
n
v
e
r
s
atio
n
s
,
an
d
d
r
o
wsy
ey
es
.
Vid
eo
s
wer
e
ca
p
tu
r
ed
u
s
in
g
a
n
in
f
r
ar
e
d
ca
m
er
a
at
a
r
eso
lu
ti
o
n
o
f
6
4
0
×
4
8
0
p
ix
els an
d
3
0
FPS
[
6
2
]
.
ii)
So
u
th
est
Un
iv
er
s
ity
Dr
iv
in
g
-
p
o
s
tu
r
e
(
SEU
)
d
r
i
v
in
g
-
p
o
s
tu
r
e
d
ataset
:
d
ev
elo
p
ed
b
y
Z
h
ao
et
a
l.
[
6
3
]
.
t
h
is
d
ataset
co
n
s
is
t
s
o
f
v
id
eo
s
ca
p
tu
r
ed
u
s
in
g
a
s
id
e
-
m
o
u
n
ted
L
o
g
itech
C
9
0
5
C
C
D
ca
m
er
a
u
n
d
er
d
ay
lig
h
t
co
n
d
itio
n
s
,
with
a
r
eso
lu
tio
n
o
f
6
4
0
×4
8
0
p
i
x
els.
A
to
tal
o
f
2
0
d
r
iv
er
s
i
n
th
e
d
ataset,
co
m
p
r
is
in
g
ten
m
ales a
n
d
ten
f
em
ales.
iii)
Z
J
U
ey
eb
lin
k
d
ataset
:
c
o
n
s
is
ts
o
f
8
0
v
id
eo
clip
s
f
r
o
m
2
0
in
d
iv
id
u
als,
with
ea
c
h
in
d
iv
id
u
a
l
co
n
tr
ib
u
tin
g
f
o
u
r
clip
s
:
f
r
o
n
tal
v
iew
with
o
u
t
g
lass
es,
f
r
o
n
tal
v
iew
with
b
lack
-
f
r
am
e
g
lass
es,
an
d
u
p
war
d
v
iew
with
o
u
t
g
lass
es.
E
y
e
im
ag
es
a
r
e
class
if
ied
as
o
p
en
o
r
clo
s
ed
an
d
s
ep
ar
ated
in
to
tr
ain
i
n
g
a
n
d
test
in
g
s
ets.
T
h
e
d
ataset
co
n
tain
s
7
,
0
0
0
o
p
en
-
ey
e
im
ag
es
(
5
,
7
7
0
f
o
r
tr
ain
in
g
,
1
,
2
3
0
f
o
r
test
in
g
)
an
d
5
,
5
7
0
clo
s
ed
-
ey
e
im
ag
es (
1
,
5
7
4
f
o
r
tr
ain
in
g
an
d
4
1
0
f
o
r
test
in
g
)
,
with
ea
ch
im
ag
e
s
ized
2
4
×2
4
p
ix
els [
6
4
]
.
iv
)
C
lo
s
ed
ey
es
in
th
e
wild
(
C
E
W
)
d
ataset
:
th
is
d
ataset
co
n
tain
s
o
n
lin
e
im
a
g
es
o
f
ap
p
r
o
x
im
atel
y
2
,
4
2
3
p
ar
ticip
a
n
ts
f
r
o
m
m
u
ltip
le
r
ac
ial
g
r
o
u
p
s
,
in
cl
u
d
in
g
Asi
an
s
an
d
lig
h
t
-
s
k
in
n
e
d
n
o
n
-
A
s
ian
s
.
Am
o
n
g
th
ese,
1
,
1
9
2
im
a
g
es
f
ea
tu
r
e
clo
s
ed
ey
es
an
d
1
,
2
3
1
s
h
o
w
o
p
en
ey
es.
T
h
e
im
ag
es
wer
e
s
elec
ted
f
r
o
m
th
e
lab
eled
f
ac
es in
th
e
wild
(
L
FW
)
d
atab
ase
[
6
5
]
.
v)
DR
OZ
Y
d
ataset
(
UL
g
m
u
ltimo
d
ality
d
r
o
wsi
n
ess
d
atab
ase
)
:
in
clu
d
es
1
4
p
ar
ticip
an
ts
(
3
m
ales
an
d
1
1
f
em
ales)
,
ea
ch
co
n
tr
ib
u
tin
g
v
id
eo
s
r
o
u
g
h
ly
1
0
m
in
u
tes
lo
n
g
,
ac
co
m
p
an
ied
b
y
p
s
y
ch
o
m
o
to
r
v
ig
ilan
ce
test
(
PVT)
s
co
r
es
m
ea
s
u
r
in
g
d
r
o
wsi
n
ess
.
T
im
e
-
s
y
n
ch
r
o
n
ize
d
K
ar
o
lin
s
k
a
s
leep
in
ess
s
ca
le
(
KSS)
r
atin
g
s
ar
e
p
r
o
v
id
ed
f
o
r
ea
c
h
p
ar
ticip
a
n
t [
6
5
]
.
v
i)
E
y
e
an
d
m
o
u
t
h
d
etec
tio
n
(
E
MD
)
d
atas
et:
co
m
p
r
is
es
3
6
,
7
6
4
ey
e
s
am
p
les
an
d
1
5
,
1
8
5
m
o
u
th
s
am
p
le
s
f
r
o
m
2
1
v
o
lu
n
teer
s
.
T
h
e
d
ata
s
et
co
v
er
s
r
ea
l
-
wo
r
l
d
d
r
iv
in
g
co
n
d
itio
n
s
,
in
clu
d
i
n
g
p
ar
ticip
an
ts
with
o
r
with
o
u
t g
lass
es,
f
r
o
n
tal
an
d
later
al
v
iews,
as we
ll a
s
d
ay
an
d
n
ig
h
t e
n
v
ir
o
n
m
e
n
ts
[
1
7
]
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
1
4
,
No
.
6
,
Dec
em
b
er
2
0
2
5
:
4
4
1
5
-
4
4
2
6
4422
v
i
i
)
Y
a
w
n
d
e
t
e
c
t
i
o
n
d
a
t
a
s
e
t
(
Y
a
w
n
D
D
)
d
a
t
a
s
e
t:
c
o
n
t
a
i
n
s
r
e
c
o
r
d
i
n
g
s
o
f
3
2
2
d
r
i
v
e
r
s
i
n
r
e
a
l
-
c
a
r
c
o
n
d
i
t
i
o
n
s
,
c
a
p
t
u
r
e
d
w
i
t
h
i
n
-
c
a
r
c
a
m
e
r
a
s
.
T
h
e
d
a
t
a
s
e
t
i
s
d
i
v
i
d
e
d
i
n
t
o
f
o
u
r
c
l
a
s
s
e
s
—
y
a
w
n
,
n
o
-
y
a
w
n
,
o
p
e
n
-
e
y
e
,
a
n
d
c
l
o
s
e
d
-
e
y
e
—
t
o
t
a
l
i
n
g
2
,
9
0
0
s
a
m
p
l
e
s
:
7
2
6
c
l
o
s
e
d
-
e
y
e
,
7
2
6
o
p
e
n
-
e
y
e
,
7
2
5
n
o
-
y
a
w
n
,
a
n
d
7
2
3
y
a
w
n
i
m
a
g
e
s
[
3
0
]
.
5.
M
E
T
H
O
D
I
n
th
is
s
ec
tio
n
,
p
r
o
p
o
s
ed
a
p
p
r
o
ac
h
es a
n
d
m
ec
h
an
is
m
s
ad
o
p
t
ed
in
4
p
ap
er
s
to
d
etec
t d
r
iv
e
r
d
is
tr
ac
tio
n
will b
e
lis
ted
,
in
co
m
p
letio
n
o
f
th
e
tech
n
iq
u
es a
n
d
m
eth
o
d
s
p
r
esen
ted
in
th
e
p
r
ev
io
u
s
s
ec
tio
n
s
.
5
.
1
.
Chen
et
a
l.
[
2
4
]
/2
0
2
0
T
h
ey
in
tr
o
d
u
ce
d
a
d
r
iv
er
d
r
o
wsi
n
ess
d
etec
tio
n
m
o
d
el
th
at
in
teg
r
ates
f
ac
to
r
ized
b
ilin
ea
r
f
ea
tu
r
e
f
u
s
io
n
with
an
L
STM
-
b
ased
r
ec
u
r
r
en
t
co
n
v
o
lu
tio
n
al
n
etwo
r
k
to
ac
cu
r
ately
id
en
tify
s
ig
n
s
o
f
d
r
iv
er
s
leep
in
ess
,
as illu
s
tr
ated
in
Fig
u
r
e
6
.
T
h
e
p
r
im
ar
y
c
o
n
tr
ib
u
tio
n
s
o
f
t
h
eir
r
esear
ch
in
clu
d
e:
i)
Desig
n
o
f
a
n
o
v
el
m
u
ltil
ev
el
d
r
iv
er
d
r
o
wsi
n
ess
esti
m
atio
n
s
y
s
tem
co
m
p
o
s
ed
o
f
th
e
f
o
llo
win
g
m
ain
co
m
p
o
n
en
ts
:
ex
tr
ac
tio
n
o
f
d
e
ep
f
ea
tu
r
e
r
ep
r
esen
tatio
n
s
as
s
o
ciate
d
with
th
e
d
r
iv
e
r
’
s
ey
es
an
d
m
o
u
t
h
f
r
o
m
th
e
d
ataset
;
f
u
s
io
n
o
f
f
ea
tu
r
es
r
elate
d
to
f
atig
u
e
in
d
icato
r
s
;
an
d
tem
p
o
r
al
m
o
d
eli
n
g
o
f
f
atig
u
e
f
ea
tu
r
es th
r
o
u
g
h
a
lo
n
g
s
h
o
r
t
-
t
er
m
r
ec
u
r
r
en
t c
o
n
v
o
l
u
tio
n
al
n
etwo
r
k
(
L
STM
)
.
ii)
R
eg
ar
d
in
g
f
atig
u
e
f
ea
tu
r
e
f
u
s
io
n
,
th
ey
p
r
o
p
o
s
ed
a
n
ew
f
ac
to
r
ized
b
ilin
ea
r
f
ea
tu
r
e
f
u
s
io
n
m
o
d
el
s
u
itab
le
f
o
r
m
u
lti
-
m
o
d
el
f
ea
tu
r
e
in
p
u
t
an
d
p
er
f
o
r
m
e
d
b
ilin
ea
r
f
u
s
io
n
o
f
th
e
ex
tr
ac
ted
d
ee
p
f
ea
t
u
r
e
r
ep
r
esen
tatio
n
s
o
f
ey
es a
n
d
m
o
u
t
h
to
s
o
lv
e
t
h
e
lim
itatio
n
s
o
f
th
e
f
ea
tu
r
e
lin
e
ar
f
u
s
io
n
p
r
o
ce
s
s
.
Fig
u
r
e
6
.
Ov
e
r
v
iew
o
f
th
e
p
r
o
p
o
s
ed
f
r
am
ew
o
r
k
a
r
ch
itectu
r
e
[
2
4
]
5
.
2
.
M
a
g
á
n
et
a
l.
[
4
2
]
/2
0
2
2
T
h
e
a
i
m
o
f
t
h
is
w
o
r
k
is
t
o
d
ev
e
l
o
p
a
s
y
s
t
e
m
c
a
p
a
b
l
e
o
f
e
s
tim
a
t
i
n
g
d
r
i
v
e
r
f
a
t
i
g
u
e
u
s
i
n
g
s
eq
u
e
n
c
e
s
o
f
i
m
a
g
e
s
i
n
w
h
i
c
h
t
h
e
s
u
b
j
e
c
t
’
s
f
a
c
e
i
s
c
l
e
a
r
l
y
v
is
i
b
le
.
T
h
e
s
y
s
t
em
i
n
c
l
u
d
e
s
t
h
e
f
o
ll
o
w
i
n
g
c
o
m
p
o
n
e
n
t
s
:
i)
T
h
ey
p
er
f
o
r
m
f
atig
u
e
d
etec
tio
n
task
s
at
a
g
iv
en
m
o
m
en
t
b
a
s
ed
o
n
th
e
an
aly
s
is
o
f
a
s
eq
u
e
n
ce
o
f
im
ag
es
f
o
r
th
e
last
6
0
s
.
ii)
I
n
th
is
s
tu
d
y
,
two
alter
n
ativ
e
s
o
lu
tio
n
s
ar
e
p
r
esen
ted
,
f
o
cu
s
in
g
o
n
m
i
n
im
izin
g
f
alse p
o
s
itiv
es.
iii)
T
h
e
f
ir
s
t a
p
p
r
o
ac
h
em
p
lo
y
s
a
co
m
b
in
atio
n
o
f
a
R
NN
an
d
C
NN
.
iv
)
T
h
e
s
ec
o
n
d
a
p
p
r
o
ac
h
u
tili
ze
s
d
ee
p
lear
n
i
n
g
tech
n
iq
u
es
to
ex
tr
ac
t
n
u
m
er
ical
f
ea
tu
r
es
f
r
o
m
th
e
im
ag
es,
wh
ich
ar
e
th
en
p
r
o
ce
s
s
ed
b
y
a
f
u
zz
y
lo
g
ic
-
b
ased
s
y
s
tem
.
v)
A
Gau
s
s
ian
b
lu
r
was
ap
p
lied
t
o
th
e
o
r
ig
in
al
im
ag
e
to
m
i
n
im
ize
n
o
is
e
an
d
s
o
f
ten
ed
g
es,
en
s
u
r
in
g
th
at
th
e
m
ain
co
n
ten
t a
n
d
s
tr
u
ctu
r
e
o
f
th
e
im
ag
e
r
em
ai
n
ed
lar
g
ely
u
n
af
f
ec
ted
.
v
i)
T
h
e
DL
I
B
lib
r
ar
y
was u
s
ed
to
d
etec
t th
e
f
ac
ial
r
eg
io
n
with
in
ea
ch
im
ag
e
.
v
ii)
Face
co
o
r
d
in
ates w
er
e
d
ete
r
m
i
n
ed
u
s
in
g
HOG
f
ea
tu
r
es in
co
n
ju
n
ctio
n
with
a
lin
ea
r
SVM.
5
.
3
.
Chen
et
a
l.
[
4
]
/2
0
2
0
T
h
ey
d
e
v
elo
p
ed
a
two
-
s
tr
ea
m
C
NN
ar
ch
itectu
r
e
f
o
r
d
is
tr
ac
ti
o
n
d
etec
tio
n
,
as d
ep
icted
i
n
Fig
u
r
e
7
.
T
h
e
ar
ch
itectu
r
e
o
p
er
ates a
s
f
o
llo
ws:
i)
T
h
e
m
o
d
el
co
m
p
r
is
es
th
r
e
e
m
ai
n
s
u
b
-
n
etw
o
r
k
s
:
a
s
p
ati
al
s
tr
ea
m
c
o
n
v
o
lu
ti
o
n
al
n
etw
o
r
k
,
a
te
m
p
o
r
a
l
s
tr
e
am
c
o
n
v
o
lu
ti
o
n
al
n
etw
o
r
k
,
a
n
d
a
f
u
s
i
o
n
n
etw
o
r
k
.
ii)
T
h
e
s
p
at
ial
a
n
d
te
m
p
o
r
a
l
s
tr
e
am
s
a
r
e
r
es
p
o
n
s
i
b
le
f
o
r
ex
t
r
ac
ti
n
g
s
p
a
tia
l
a
n
d
te
m
p
o
r
al
f
e
at
u
r
es,
r
es
p
e
cti
v
e
ly
,
w
h
ic
h
a
r
e
s
u
b
s
eq
u
e
n
tl
y
c
o
m
b
in
ed
w
it
h
i
n
t
h
e
f
u
s
io
n
n
e
tw
o
r
k
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
A
r
ev
iew
o
f d
r
iver d
is
t
r
a
ctio
n
d
etec
tio
n
w
h
ile
d
r
ivin
g
b
a
s
ed
o
n
co
n
vo
lu
tio
n
a
l n
eu
r
a
l
(
Gh
a
d
y
A
lh
a
ma
d
)
4423
iii)
L
e
v
e
r
a
g
i
n
g
tr
a
n
s
f
er
lea
r
n
i
n
g
,
t
h
e
s
p
at
ial
s
tr
ea
m
n
etw
o
r
k
is
co
n
s
t
r
u
ct
ed
b
as
ed
o
n
t
h
e
w
ell
-
k
n
o
w
n
VGG
-
1
6
ar
ch
ite
ct
u
r
e,
a
ll
o
wi
n
g
t
h
e
u
s
e
o
f
p
r
e
-
tr
ai
n
e
d
we
ig
h
ts
f
r
o
m
t
h
e
I
m
ag
eN
et
d
atas
et.
iv
)
I
n
iti
all
y
,
a
v
e
r
ag
e
p
o
o
li
n
g
a
cr
o
s
s
th
e
t
em
p
o
r
al
d
im
en
s
io
n
is
ap
p
li
e
d
t
o
1
0
co
n
s
ec
u
ti
v
e
R
G
B
f
r
a
m
es
,
a
n
d
th
e
o
u
t
p
u
t is
t
h
e
n
f
o
r
wa
r
d
e
d
t
o
t
h
e
s
u
b
s
e
q
u
e
n
t
l
ay
e
r
s
o
f
t
h
e
s
p
at
ial
s
t
r
e
am
n
et
wo
r
k
.
v)
Fin
all
y
,
t
h
e
f
u
s
i
o
n
n
et
wo
r
k
c
o
m
p
r
is
in
g
tw
o
c
o
n
v
o
l
u
t
io
n
a
l
la
y
e
r
s
a
n
d
t
wo
f
u
ll
y
co
n
n
e
ct
ed
lay
e
r
s
-
m
e
r
g
es
th
e
s
p
at
ial
a
n
d
te
m
p
o
r
a
l
r
e
p
r
es
en
t
ati
o
n
s
to
c
lass
i
f
y
te
n
d
is
ti
n
c
t t
y
p
es
o
f
d
is
t
r
ac
te
d
d
r
i
v
i
n
g
b
e
h
a
v
i
o
r
s
.
5
.
4
.
G
ha
z
a
l
et
a
l
.
[
4
6
]
/2
0
1
8
A
lo
w
-
co
s
t
an
d
r
ea
l
-
tim
e
em
b
ed
d
ed
s
y
s
tem
f
o
r
f
ati
g
u
e
d
e
tectio
n
is
p
r
o
p
o
s
ed
in
t
h
is
s
tu
d
y
u
s
in
g
C
NN
.
T
h
e
p
ap
er
in
clu
d
es m
an
y
id
ea
s
,
as f
o
llo
ws:
i)
Pre
-
t
r
ai
n
ed
m
o
d
el
wei
g
h
ts
we
r
e
lo
ad
e
d
t
o
l
ev
er
a
g
e
tr
an
s
f
e
r
lea
r
n
i
n
g
a
n
d
r
ed
u
ce
r
el
ia
n
c
e
o
n
la
r
g
e
-
s
c
ale
d
at
ase
ts
.
ii)
Face
d
et
ec
t
io
n
was
a
p
p
li
e
d
to
d
e
f
i
n
e
t
h
e
R
O
I
.
iii)
T
h
e
s
t
u
d
y
f
o
c
u
s
e
d
o
n
s
o
l
u
ti
o
n
s
s
u
ita
b
l
e
f
o
r
r
es
o
u
r
c
e
-
c
o
n
s
t
r
ai
n
ed
d
e
v
i
ce
s
,
ai
m
i
n
g
f
o
r
a
s
ca
l
ab
le
,
c
o
s
t
-
ef
f
e
cti
v
e
,
a
n
d
r
o
b
u
s
t
i
m
p
le
m
e
n
ta
ti
o
n
.
Fo
r
t
h
is
p
u
r
p
o
s
e,
a
R
a
s
p
b
e
r
r
y
Pi
3
M
o
d
el
B
was
e
m
p
l
o
y
e
d
d
u
e
t
o
its
af
f
o
r
d
a
b
i
lit
y
.
Fig
u
r
e
7
.
Dr
iv
e
r
d
is
tr
ac
tio
n
s
y
s
tem
f
lo
wch
ar
t
[
4
]
6.
CO
NCLU
SI
O
N
T
h
e
d
r
i
v
er
d
is
tr
ac
tio
n
d
etec
ti
o
n
s
y
s
tem
s
p
lay
an
ass
is
tan
t
r
o
le
in
p
r
o
tectin
g
th
e
d
r
iv
er
f
r
o
m
tr
a
f
f
ic
ac
cid
en
ts
.
Dee
p
lear
n
in
g
m
o
d
els
h
av
e
th
e
p
o
ten
tial
to
b
ec
o
m
e
a
m
ain
s
tr
ea
m
s
o
lu
tio
n
d
u
e
to
th
eir
h
i
g
h
ac
cu
r
ac
y
an
d
s
tab
le
r
o
b
u
s
tn
ess
,
p
r
o
v
id
e
d
th
at
s
u
f
f
icien
tly
la
r
g
e
d
atasets
ar
e
av
ailab
le
f
o
r
t
r
ain
in
g
.
T
h
is
p
ap
er
r
ev
iews
a
s
elec
tio
n
o
f
ca
s
e
s
tu
d
ies
co
n
d
u
cted
b
etwe
en
2
0
1
4
an
d
2
0
2
4
th
at
f
o
cu
s
o
n
d
ee
p
l
ea
r
n
in
g
an
d
C
NN
-
b
ased
ap
p
r
o
ac
h
es.
Ma
n
y
s
u
cc
ess
f
u
l
s
tu
d
ies
h
av
e
d
em
o
n
s
tr
ated
th
e
e
f
f
ec
tiv
en
ess
o
f
C
NNs
in
d
etec
tin
g
d
r
iv
er
d
is
tr
ac
tio
n
,
s
o
m
etim
es
co
m
b
in
ed
with
o
th
er
tech
n
i
q
u
es,
h
av
e
h
ig
h
ac
c
u
r
ac
y
,
a
n
d
h
a
v
e
ad
v
an
tag
es
an
d
d
is
ad
v
an
tag
es.
T
h
e
m
eth
o
d
o
lo
g
ies
p
r
o
p
o
s
ed
s
o
f
ar
in
all
th
e
p
ap
er
s
ar
e
ef
f
ec
tiv
e
an
d
ca
n
in
teg
r
ate
to
g
eth
er
to
lau
n
ch
n
ew
cr
ea
tiv
e
f
in
d
in
g
s
an
d
id
ea
s
.
T
h
er
e
ar
e
th
r
ee
ty
p
es
th
at
ca
n
b
e
u
s
ed
to
m
ea
s
u
r
e
lev
el
o
f
d
is
tr
ac
tio
n
(
co
g
n
itiv
e,
v
is
u
al,
an
d
m
an
u
al
)
.
I
t
ca
n
b
e
n
o
ted
th
at
th
e
p
ap
er
s
r
ev
iewe
d
f
o
cu
s
ed
m
o
r
e
o
n
v
is
u
al
an
d
m
an
u
al
d
is
tr
ac
tio
n
th
an
o
n
co
g
n
itiv
e
d
is
tr
ac
tio
n
,
b
ec
au
s
e
r
esear
ch
er
s
d
id
n
o
t
p
ay
s
u
f
f
icien
t
atten
tio
n
to
co
g
n
itiv
e
d
is
tr
ac
tio
n
.
Per
h
ap
s
th
e
lar
g
e
s
t
p
r
o
b
lem
h
o
wev
e
r
is
th
at
th
e
d
ef
in
itio
n
s
ty
p
ically
r
e
f
er
to
d
is
tr
ac
tio
n
as
a
r
ed
u
ctio
n
o
f
atten
tio
n
,
b
u
t
n
e
v
er
d
e
f
in
e
atten
tio
n
.
Actu
ally
,
c
o
g
n
itiv
e
d
is
tr
ac
tio
n
h
as
a
n
eg
ativ
e
im
p
ac
t
o
n
d
r
iv
in
g
an
d
ca
n
lead
to
s
er
io
u
s
ac
cid
en
ts
.
As
a
r
esu
lt,
w
e
r
e
co
m
m
en
d
d
ig
g
in
g
more
in
to
c
o
g
n
itiv
e
d
is
tr
ac
tio
n
b
y
tak
in
g
ad
v
an
tag
e
o
f
C
NN
an
d
th
e
tech
n
iq
u
es
r
ev
iewe
d
.
I
t
is
also
r
ec
o
m
m
en
d
ed
to
b
u
ild
lar
g
e
p
u
b
lic
d
ataset
co
v
er
in
g
t
h
is
asp
ec
t to
s
u
p
p
o
r
t b
r
o
ad
e
r
s
tu
d
ies.
F
UNDING
I
NF
O
R
M
A
T
I
O
N
Au
th
o
r
s
s
tate
n
o
f
u
n
d
in
g
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
tr
ib
u
to
r
R
o
les
T
ax
o
n
o
m
y
(
C
R
ed
iT)
to
r
ec
o
g
n
ize
in
d
iv
id
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
Aut
ho
r
C
M
So
Va
Fo
I
R
D
O
E
Vi
Su
P
Fu
Gh
ad
y
A
lh
am
ad
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
M
o
h
a
m
a
d
-
B
a
s
s
a
m
K
u
r
d
y
✓
✓
✓
✓
✓
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
1
4
,
No
.
6
,
Dec
em
b
er
2
0
2
5
:
4
4
1
5
-
4
4
2
6
4424
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
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
e
y
h
av
e
n
o
co
n
f
licts
o
f
in
ter
est r
e
lated
to
th
is
wo
r
k
.
DATA AV
AI
L
AB
I
L
I
T
Y
T
h
is
p
ap
er
p
r
esen
ts
s
ev
en
d
ata
s
ets th
at
h
av
e
b
ee
n
u
s
ed
a
n
d
a
r
e
av
ailab
le
at:
‒
NT
HU
d
at
a
s
et
at
h
tt
p
s
:
//ww
w.
k
a
g
g
le
.
c
o
m
/
d
at
ase
ts
/
b
a
n
u
d
ee
p
/n
t
h
u
d
d
d
2
.
‒
SEU
d
atas
et
is
n
o
t
a
v
a
ila
b
l
e
p
u
b
lic
.
‒
Z
J
U
d
a
ta
s
et
at
h
tt
p
s
://
g
it
h
u
b
.
c
o
m
/el
n
i
n
o
9
y
k
l
/ZJU
-
D
atas
et
.
‒
C
E
W
d
ataset
at
h
ttp
s
://p
ar
n
ec
.
n
u
aa
.
e
d
u
.
c
n
/_
u
p
lo
ad
/tp
l/0
2
/d
b
/
7
3
1
/tem
p
late7
3
1
/p
ag
es/x
tan
/C
lo
s
ed
E
y
eDa
tab
ases
.
h
tm
l
.
‒
DR
OZ
Y
d
atas
et
at
h
tt
p
:
//ww
w
.
d
r
o
z
y
.
u
lg
.
a
c.
b
e
/
.
‒
E
MD
d
at
ase
t
at
h
tt
p
s
:
//ww
w.
k
ag
g
l
e.
co
m
/
d
a
tase
ts
/
u
m
a
p
r
i
y
asr
/ey
e
-
m
o
u
t
h
-
d
e
te
cti
o
n
-
d
a
tas
et
.
‒
Yaw
n
DD
d
at
a
s
et
at
h
tt
p
s
:
//
q
u
a
lin
et
.
g
it
h
u
b
.
i
o
/
d
at
a
b
ases
/v
id
e
o
/y
a
wd
d
_
a
_
y
aw
n
i
n
g
_
d
et
ec
t
io
n
_
d
at
aset
/
.
RE
F
E
R
E
NC
E
S
[
1
]
Q
.
H
u
a
,
L.
Ji
n
,
Y
.
Ji
a
n
g
,
M
.
G
a
o
,
a
n
d
B
.
G
u
o
,
“
C
o
g
n
i
t
i
v
e
d
i
st
r
a
c
t
i
o
n
s
t
a
t
e
r
e
c
o
g
n
i
t
i
o
n
o
f
d
r
i
v
e
r
s a
t
a
n
o
n
s
i
g
n
a
l
i
z
e
d
i
n
t
e
r
s
e
c
t
i
o
n
i
n
a
mi
x
e
d
t
r
a
f
f
i
c
e
n
v
i
r
o
n
m
e
n
t
,
”
A
d
v
a
n
c
e
s i
n
C
i
v
i
l
E
n
g
i
n
e
e
ri
n
g
,
v
o
l
.
2
0
2
1
,
p
p
.
1
6
,
2
0
2
1
,
d
o
i
:
1
0
.
1
1
5
5
/
2
0
2
1
/
6
6
7
6
8
0
7
[
2
]
N
H
TSA
,
“
D
i
st
r
a
c
t
e
d
d
r
i
v
i
n
g
,
”
N
a
t
i
o
n
a
l
H
i
g
h
w
a
y
T
r
a
f
f
i
c
S
a
f
e
t
y
Ad
m
i
n
i
st
r
a
t
i
o
n
,
U
.
S
.
,
2
0
2
2
.
A
c
c
e
ss
e
d
:
D
e
c
.
3
1
,
2
0
2
3
.
[
O
n
l
i
n
e
]
.
A
v
a
i
l
a
b
l
e
:
h
t
t
p
s:
/
/
w
w
w
.
n
h
t
s
a
.
g
o
v
/
r
i
sk
y
-
d
r
i
v
i
n
g
/
d
i
st
r
a
c
t
e
d
-
d
r
i
v
i
n
g
[
3
]
T
r
a
f
f
i
c
S
a
f
e
t
y
M
a
r
k
e
t
i
n
g
,
“
2
0
2
3
d
i
s
t
r
a
c
t
i
o
n
f
a
c
t
s
h
e
e
t
,
”
T
r
a
f
f
i
c
S
a
f
e
t
y
M
a
r
k
e
t
i
n
g
,
2
0
2
3
.
A
c
c
e
s
s
e
d
:
D
e
c
.
3
1
,
2
0
2
3
.
[
O
n
l
i
n
e
]
.
A
v
a
i
l
a
b
l
e
:
h
t
t
p
s
:
/
/
w
w
w
.
t
r
a
f
f
i
c
s
a
f
e
t
y
m
a
r
k
e
t
i
n
g
.
g
o
v
/
s
i
t
e
s
/
t
s
m
.
g
o
v
/
f
i
l
e
s
/
2
0
2
3
-
0
2
/
1
5
7
7
6
a
-
D
i
s
t
r
a
c
t
i
o
n
%
2
0
F
a
c
t
%
2
0
S
h
e
e
t
_
%
2
0
0
2
0
3
2
3
v
2
-
t
a
g
_
0
.
d
o
c
x
[
4
]
J.
-
C
.
C
h
e
n
,
C
.
-
Y
.
Le
e
,
P
.
-
Y
.
H
u
a
n
g
,
a
n
d
C
.
-
R
.
L
i
n
,
“
D
r
i
v
e
r
b
e
h
a
v
i
o
r
a
n
a
l
y
s
i
s v
i
a
t
w
o
-
s
t
r
e
a
m d
e
e
p
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
,
”
Ap
p
l
i
e
d
S
c
i
e
n
c
e
s
,
v
o
l
.
1
0
,
n
o
.
6
,
2
0
2
0
,
d
o
i
:
1
0
.
3
3
9
0
/
a
p
p
1
0
0
6
1
9
0
8
.
[
5
]
A
.
D
a
sg
u
p
t
a
,
A
.
G
e
o
r
g
e
,
S
.
L.
H
a
p
p
y
,
a
n
d
A
.
R
o
u
t
r
a
y
,
“
A
v
i
s
i
o
n
-
b
a
se
d
s
y
s
t
e
m
f
o
r
m
o
n
i
t
o
r
i
n
g
t
h
e
l
o
ss
o
f
a
t
t
e
n
t
i
o
n
i
n
a
u
t
o
m
o
t
i
v
e
d
r
i
v
e
r
s
,
”
I
EEE
T
r
a
n
s
a
c
t
i
o
n
s
o
n
I
n
t
e
l
l
i
g
e
n
t
T
ra
n
s
p
o
rt
a
t
i
o
n
S
y
s
t
e
m
s,
v
o
l
.
1
4
,
n
o
.
4
,
p
p
.
1
8
2
5
-
1
8
3
8
,
2
0
1
3
,
d
o
i
:
1
0
.
1
1
0
9
/
TI
TS.2
0
1
3
.
2
2
7
1
0
5
2
.
[
6
]
J.
-
J
.
Y
a
n
,
H
.
-
H
.
K
u
o
,
Y
.
-
F
.
L
i
n
,
a
n
d
T
.
-
L
.
L
i
a
o
,
“
Real
-
t
i
m
e
d
r
i
v
e
r
d
r
o
w
s
i
n
e
s
s
d
e
t
e
c
t
i
o
n
s
y
s
t
e
m
b
a
s
e
d
o
n
P
E
R
C
L
O
S
a
n
d
g
r
a
y
s
c
a
l
e
i
m
a
g
e
p
r
o
c
e
s
s
i
n
g
,
”
i
n
2
0
1
6
I
n
t
e
r
n
a
t
i
o
n
a
l
S
y
m
p
o
s
i
u
m
o
n
C
o
m
p
u
t
e
r
,
C
o
n
s
u
m
e
r
a
n
d
C
o
n
t
r
o
l
,
2
0
1
6
,
p
p
.
2
4
3
-
2
4
6
,
d
o
i
:
1
0
.
1
1
0
9
/
I
S
3
C
.
2
0
1
6
.
7
2
.
[
7
]
W
.
Z
h
a
n
g
,
B
.
C
h
e
n
g
,
a
n
d
Y
.
L
i
n
,
“
D
r
i
v
e
r
d
r
o
w
si
n
e
ss r
e
c
o
g
n
i
t
i
o
n
b
a
s
e
d
o
n
c
o
m
p
u
t
e
r
v
i
si
o
n
t
e
c
h
n
o
l
o
g
y
,
”
S
c
i
e
n
c
e
a
n
d
T
e
c
h
n
o
l
o
g
y
,
v
o
l
.
1
7
,
n
o
.
3
,
p
p
.
3
5
4
-
3
6
2
,
2
0
1
2
,
d
o
i
:
1
0
.
1
1
0
9
/
TS
T.
2
0
1
2
.
6
2
1
6
7
6
8
.
[
8
]
M
.
S
h
a
h
v
e
r
d
y
,
M
.
F
a
t
h
y
,
R
.
B
e
r
a
n
g
i
,
a
n
d
M
.
S
a
b
o
k
r
o
u
,
“
D
r
i
v
e
r
b
e
h
a
v
i
o
r
d
e
t
e
c
t
i
o
n
a
n
d
c
l
a
ssi
f
i
c
a
t
i
o
n
u
si
n
g
d
e
e
p
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
,
”
Ex
p
e
r
t
S
y
s
t
e
m
s
w
i
t
h
A
p
p
l
i
c
a
t
i
o
n
s,
v
o
l
.
1
4
9
,
2
0
2
0
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
e
sw
a
.
2
0
2
0
.
1
1
3
2
4
0
.
[
9
]
L.
Li
,
B
.
Z
h
o
n
g
,
C
.
H
.
Jr
,
Y
.
Li
a
n
g
,
W
.
J.
H
o
r
r
e
y
,
a
n
d
X
.
X
u
,
“
D
e
t
e
c
t
i
o
n
o
f
d
r
i
v
e
r
ma
n
u
a
l
d
i
s
t
r
a
c
t
i
o
n
v
i
a
i
ma
g
e
-
b
a
s
e
d
h
a
n
d
a
n
d
e
a
r
r
e
c
o
g
n
i
t
i
o
n
,
”
Ac
c
i
d
e
n
t
An
a
l
y
s
i
s &
Pre
v
e
n
t
i
o
n
,
v
o
l
.
1
3
7
,
2
0
2
0
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
a
a
p
.
2
0
2
0
.
1
0
5
4
3
2
.
[
1
0
]
C
.
H
u
a
n
g
,
X
.
W
a
n
g
,
J.
C
a
o
,
S
.
W
a
n
g
,
a
n
d
Y
.
Z
h
a
n
g
,
“
H
C
F
:
a
h
y
b
r
i
d
C
N
N
f
r
a
mew
o
r
k
f
o
r
b
e
h
a
v
i
o
r
d
e
t
e
c
t
i
o
n
o
f
d
i
st
r
a
c
t
e
d
d
r
i
v
e
r
s,
”
I
EE
E
a
c
c
e
ss,
v
o
l
.
8
,
p
p
.
1
0
9
3
3
5
-
1
0
9
3
4
9
,
2
0
2
0
,
d
o
i
:
1
0
.
1
1
0
9
/
A
C
C
ESS
.
2
0
2
0
.
3
0
0
1
1
5
9
.
[
1
1
]
E.
M
u
h
r
e
r
a
n
d
M
.
V
o
l
l
r
a
t
h
,
“
Th
e
e
f
f
e
c
t
o
f
v
i
su
a
l
a
n
d
c
o
g
n
i
t
i
v
e
d
i
st
r
a
c
t
i
o
n
o
n
d
r
i
v
e
r
’
s
a
n
t
i
c
i
p
a
t
i
o
n
i
n
a
si
m
u
l
a
t
e
d
c
a
r
f
o
l
l
o
w
i
n
g
sce
n
a
r
i
o
,
”
T
r
a
n
sp
o
r
t
a
t
i
o
n
r
e
se
a
rc
h
p
a
r
t
F:
t
ra
f
f
i
c
p
syc
h
o
l
o
g
y
a
n
d
b
e
h
a
v
i
o
u
r,
v
o
l
.
1
4
,
n
o
.
6
,
p
p
.
5
5
5
-
5
6
6
,
2
0
1
1
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
t
r
f
.
2
0
1
1
.
0
6
.
0
0
3
.
[
1
2
]
Y
.
Y
a
n
g
,
Z.
Y
e
,
S
.
M
.
Ea
sa,
Y
.
F
e
n
g
,
a
n
d
X
.
Zh
e
n
g
,
“
Ef
f
e
c
t
o
f
d
r
i
v
i
n
g
d
i
s
t
r
a
c
t
i
o
n
s
o
n
d
r
i
v
e
r
m
e
n
t
a
l
w
o
r
k
l
o
a
d
i
n
w
o
r
k
z
o
n
e
’
s
w
a
r
n
i
n
g
a
r
e
a
,
”
T
ra
n
s
p
o
r
t
a
t
i
o
n
r
e
s
e
a
r
c
h
p
a
rt
F:
t
r
a
f
f
i
c
p
sy
c
h
o
l
o
g
y
a
n
d
b
e
h
a
v
i
o
u
r,
v
o
l
.
9
5
,
p
p
.
1
1
2
-
1
2
8
,
2
0
2
3
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
t
r
f
.
2
0
2
3
.
0
3
.
0
1
8
.
[
1
3
]
M
.
C
h
a
n
a
n
d
A
.
S
i
n
g
h
a
l
,
“
Th
e
e
m
o
t
i
o
n
a
l
si
d
e
o
f
c
o
g
n
i
t
i
v
e
d
i
s
t
r
a
c
t
i
o
n
:
I
mp
l
i
c
a
t
i
o
n
s
f
o
r
r
o
a
d
saf
e
t
y
,
”
Ac
c
i
d
e
n
t
A
n
a
l
y
si
s
&
Pre
v
e
n
t
i
o
n
,
v
o
l
.
5
0
,
p
p
.
1
4
7
-
1
5
4
,
2
0
1
3
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
a
a
p
.
2
0
1
2
.
0
4
.
0
0
4
.
[
1
4
]
H
.
V
.
C
h
a
n
d
a
n
d
J
.
K
a
r
t
h
i
k
e
y
a
n
,
“
C
N
N
b
a
se
d
d
r
i
v
e
r
d
r
o
w
si
n
e
ss
d
e
t
e
c
t
i
o
n
s
y
st
e
m u
si
n
g
e
m
o
t
i
o
n
a
n
a
l
y
s
i
s
,
”
I
n
t
e
l
l
i
g
e
n
t
A
u
t
o
m
a
t
i
o
n
& S
o
f
t
C
o
m
p
u
t
i
n
g
,
v
o
l
.
3
1
,
n
o
.
2
,
2
0
2
2
,
d
o
i
:
1
0
.
3
2
6
0
4
/
i
a
s
c
.
2
0
2
2
.
0
2
0
0
0
8
.
[
1
5
]
A
.
K
.
B
i
sw
a
l
,
D
.
S
i
n
g
h
,
B
.
K
.
P
a
t
t
a
n
a
y
a
k
,
D
.
S
a
ma
n
t
a
,
a
n
d
M
.
-
H
.
Y
a
n
g
,
“
I
o
T‐
b
a
se
d
s
mart
a
l
e
r
t
s
y
s
t
e
m
f
o
r
d
r
o
w
sy
d
r
i
v
e
r
d
e
t
e
c
t
i
o
n
,
”
W
i
re
l
e
ss
c
o
m
m
u
n
i
c
a
t
i
o
n
s
a
n
d
m
o
b
i
l
e
c
o
m
p
u
t
i
n
g
,
v
o
l
.
2
0
2
1
,
d
o
i
:
1
0
.
1
1
5
5
/
2
0
2
1
/
6
6
2
7
2
1
7
.
[
1
6
]
S
.
M
e
h
t
a
,
S
.
D
a
d
h
i
c
h
,
S
.
G
u
m
b
e
r
,
a
n
d
A
.
J.
B
h
a
t
t
,
“
R
e
a
l
-
t
i
me
d
r
i
v
e
r
d
r
o
w
si
n
e
ss
d
e
t
e
c
t
i
o
n
s
y
st
e
m
u
si
n
g
e
y
e
a
s
p
e
c
t
r
a
t
i
o
a
n
d
e
y
e
c
l
o
su
r
e
r
a
t
i
o
,
”
i
n
Pr
o
c
e
e
d
i
n
g
s
o
f
i
n
t
e
rn
a
t
i
o
n
a
l
c
o
n
f
e
re
n
c
e
o
n
s
u
st
a
i
n
a
b
l
e
c
o
m
p
u
t
i
n
g
i
n
s
c
i
e
n
c
e
,
t
e
c
h
n
o
l
o
g
y
a
n
d
m
a
n
a
g
e
m
e
n
t
,
Am
i
t
y
U
n
i
v
e
rsi
t
y
Ra
j
a
st
h
a
n
,
2
0
1
9
,
p
p
.
1
3
3
3
-
1
3
3
9
,
d
o
i
:
1
0
.
2
1
3
9
/
ssr
n
.
3
3
5
6
4
0
1
.
[
1
7
]
W
.
H
.
G
u
,
Y
.
Z
h
u
,
X
.
D
.
C
h
e
n
,
L
.
F
.
H
e
,
a
n
d
B
.
B
.
Zh
e
n
g
,
“
H
i
e
r
a
r
c
h
i
c
a
l
C
N
N
‐
b
a
se
d
r
e
a
l
‐
t
i
me
f
a
t
i
g
u
e
d
e
t
e
c
t
i
o
n
sy
st
e
m b
y
v
i
s
u
a
l
‐
b
a
s
e
d
t
e
c
h
n
o
l
o
g
i
e
s
u
s
i
n
g
M
S
P
mo
d
e
l
,
”
I
ET
I
m
a
g
e
Pr
o
c
e
ssi
n
g
,
v
o
l
.
1
2
,
n
o
.
1
2
,
p
p
.
2
3
1
9
-
2
3
2
9
,
2
0
1
8
,
d
o
i
:
1
0
.
1
0
4
9
/
i
e
t
-
i
p
r
.
2
0
1
8
.
5
2
4
5
.
[
1
8
]
S
.
Ze
p
f
,
T.
S
t
r
a
c
k
e
,
A
.
S
c
h
m
i
t
t
,
F
.
v
.
d
.
C
a
m
p
,
a
n
d
J
.
B
e
y
e
r
e
r
,
“
To
w
a
r
d
s
r
e
a
l
-
t
i
me
d
e
t
e
c
t
i
o
n
a
n
d
m
i
t
i
g
a
t
i
o
n
o
f
d
r
i
v
e
r
f
r
u
s
t
r
a
t
i
o
n
u
si
n
g
S
V
M
,
”
i
n
2
0
1
9
1
8
t
h
I
EE
E
I
n
t
e
rn
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
M
a
c
h
i
n
e
L
e
a
r
n
i
n
g
a
n
d
A
p
p
l
i
c
a
t
i
o
n
s
,
B
o
c
a
R
a
t
o
n
,
F
l
o
r
i
d
a
,
2
0
1
9
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
M
LA
.
2
0
1
9
.
0
0
0
3
9
.
[
1
9
]
J.
F
.
-
M
o
n
r
o
y
,
M
.
N
.
-
M
i
y
a
t
a
k
e
,
G
.
S
.
-
P
e
r
e
z
,
a
n
d
H
.
P
.
-
M
e
a
n
a
,
“
V
i
su
a
l
-
b
a
s
e
d
r
e
a
l
t
i
m
e
d
r
i
v
e
r
d
r
o
w
s
i
n
e
ss
d
e
t
e
c
t
i
o
n
s
y
s
t
e
m
u
si
n
g
C
N
N
,
”
i
n
2
0
2
1
1
8
t
h
I
n
t
e
rn
a
t
i
o
n
a
l
C
o
n
f
e
r
e
n
c
e
o
n
El
e
c
t
ri
c
a
l
E
n
g
i
n
e
e
ri
n
g
,
C
o
m
p
u
t
i
n
g
S
c
i
e
n
c
e
a
n
d
A
u
t
o
m
a
t
i
c
C
o
n
t
r
o
l
,
M
e
x
i
c
o
C
i
t
y
,
M
e
x
i
c
o
,
2
0
2
1
,
d
o
i
:
1
0
.
1
1
0
9
/
C
C
E5
3
5
2
7
.
2
0
2
1
.
9
6
3
3
0
8
2
.
Evaluation Warning : The document was created with Spire.PDF for Python.