I
nte
rna
t
io
na
l J
o
urna
l o
f
E
lect
rica
l a
nd
Co
m
p
ute
r
E
ng
in
ee
ring
(
I
J
E
CE
)
Vo
l.
12
,
No
.
1
,
Feb
r
u
ar
y
20
22
,
p
p
.
731
~
737
I
SS
N:
2088
-
8708
,
DOI
: 1
0
.
1
1
5
9
1
/
i
j
ec
e
.
v
12
i
1
.
pp
7
3
1
-
737
731
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
ec
e.
ia
esco
r
e.
co
m
A hierarchi
ca
l R
CNN
for v
ehicle
and v
ehicle
license
pla
te
detec
tion a
nd re
c
o
g
nition
Chun
l
ing
Tu
1
,
S
heng
zhi D
u
2
1
D
e
p
a
r
t
me
n
t
o
f
C
o
m
p
u
t
e
r
S
y
st
e
ms En
g
i
n
e
e
r
i
n
g
,
T
sh
w
a
n
e
U
n
i
v
e
r
si
t
y
o
f
T
e
c
h
n
o
l
o
g
y
,
P
r
e
t
o
r
i
a
,
S
o
u
t
h
A
f
r
i
c
a
2
D
e
p
a
r
t
me
n
t
o
f
El
e
c
t
r
i
c
a
l
En
g
i
n
e
e
r
i
n
g
,
T
sh
w
a
n
e
U
n
i
v
e
r
si
t
y
o
f
T
e
c
h
n
o
l
o
g
y
,
P
r
e
t
o
r
i
a
,
S
o
u
t
h
A
f
r
i
c
a
Art
icle
I
nfo
AB
ST
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Au
g
3
0
,
2
0
2
0
R
ev
i
s
ed
J
u
n
1
7
,
2
0
2
1
A
cc
ep
ted
J
u
n
3
0
,
2021
V
e
h
icle
a
n
d
v
e
h
icle
li
c
e
n
se
d
e
tec
ti
o
n
o
b
tain
e
d
i
n
c
re
d
i
b
le
a
c
h
iev
e
m
e
n
ts
d
u
ri
n
g
re
c
e
n
t
y
e
a
rs
th
a
t
a
re
a
lso
p
o
p
u
larly
u
se
d
in
re
a
l
traff
i
c
sc
e
n
a
rio
s,
su
c
h
a
s
in
telli
g
e
n
t
tra
ff
ic
m
o
n
it
o
rin
g
sy
ste
m
s,
a
u
to
p
a
rk
in
g
s
y
ste
m
s,
a
n
d
v
e
h
icle
se
rv
ice
s.
Co
m
p
u
ter
v
isio
n
a
tt
ra
c
ted
m
u
c
h
a
tt
e
n
ti
o
n
i
n
v
e
h
icle
a
n
d
v
e
h
icle
li
c
e
n
se
d
e
tec
ti
o
n
,
b
e
n
e
f
it
f
ro
m
im
a
g
e
p
ro
c
e
ss
in
g
a
n
d
m
a
c
h
in
e
lea
rn
in
g
tec
h
n
o
l
o
g
ies
.
Ho
w
e
v
e
r,
th
e
e
x
is
ti
n
g
m
e
th
o
d
s
stil
l
h
a
v
e
so
m
e
is
su
e
s
w
it
h
v
e
h
icle
a
n
d
v
e
h
icle
li
c
e
n
se
p
l
a
te
re
c
o
g
n
it
io
n
,
e
sp
e
c
ially
in
a
c
o
m
p
lex
e
n
v
iro
n
m
e
n
t.
In
t
h
is
p
a
p
e
r,
w
e
p
ro
p
o
se
a
m
u
lt
iv
e
h
icle
d
e
tec
ti
o
n
a
n
d
li
c
e
n
se
p
late
re
c
o
g
n
it
io
n
sy
ste
m
b
a
s
e
d
o
n
a
h
iera
rc
h
ica
l
re
g
io
n
c
o
n
v
o
l
u
ti
o
n
al
n
e
u
ra
l
n
e
tw
o
rk
(RCNN
).
F
irstl
y
,
a
h
ig
h
e
r
lev
e
l
o
f
RCNN
is
e
m
p
lo
y
e
d
to
e
x
trac
t
v
e
h
icle
s
f
ro
m
th
e
o
rig
in
a
l
ima
g
e
s
o
r
v
id
e
o
f
ra
m
e
s.
S
e
c
o
n
d
ly
,
th
e
re
g
io
n
s
o
f
th
e
d
e
tec
ted
v
e
h
icle
s
a
re
i
n
p
u
t
t
o
a
lo
w
e
r
lev
e
l
(s
m
a
ll
e
r)
RCNN
to
d
e
tec
t
th
e
li
c
e
n
se
p
late
.
T
h
ird
ly
,
th
e
d
e
tec
ted
li
c
e
n
se
p
late
is
sp
li
t
in
t
o
sin
g
le
n
u
m
b
e
rs.
F
in
a
ll
y
,
th
e
i
n
d
iv
id
u
a
l
n
u
m
b
e
rs
a
re
re
c
o
g
n
ize
d
b
y
a
n
e
v
e
n
sm
a
ll
e
r
RCNN
.
T
h
e
e
x
p
e
ri
m
e
n
ts
o
n
th
e
re
a
l
traf
fic
d
a
tab
a
se
v
a
li
d
a
ted
th
e
p
r
o
p
o
se
d
m
e
th
o
d
.
Co
m
p
a
re
d
w
it
h
th
e
c
o
m
m
o
n
ly
u
se
d
a
ll
-
in
-
o
n
e
d
e
e
p
lea
rn
i
n
g
stru
c
tu
re
,
th
e
p
ro
p
o
se
d
h
iera
rc
h
ica
l
m
e
th
o
d
d
e
a
ls
w
it
h
th
e
li
c
e
n
se
p
late
re
c
o
g
n
it
i
o
n
tas
k
in
m
u
lt
ip
le
lev
e
ls
f
o
r
su
b
-
tas
k
s,
w
h
ich
e
n
a
b
les
th
e
m
o
d
if
ica
ti
o
n
o
f
n
e
tw
o
rk
siz
e
a
n
d
stru
c
tu
re
a
c
c
o
rd
in
g
to
t
h
e
c
o
m
p
lex
it
y
o
f
su
b
-
tas
k
s.
T
h
e
r
e
f
o
re
,
th
e
c
o
m
p
u
tatio
n
l
o
a
d
is r
e
d
u
c
e
d
.
K
ey
w
o
r
d
s
:
Dee
p
lear
n
in
g
R
eg
io
n
co
n
v
o
lu
tio
n
n
e
u
r
al
n
et
w
o
r
k
Veh
icle
d
etec
tio
n
Veh
icle
lice
n
s
e
r
ec
o
g
n
itio
n
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
:
C
h
u
n
li
n
g
T
u
Dep
ar
t
m
en
t o
f
C
o
m
p
u
ter
S
y
s
t
e
m
s
E
n
g
i
n
ee
r
i
n
g
,
T
s
h
w
a
n
e
U
n
iv
er
s
it
y
o
f
T
ec
h
n
o
lo
g
y
2
A
u
b
r
e
y
Ma
tlak
ala
St,
So
s
h
a
n
g
u
v
e,
P
r
eto
r
ia,
So
u
th
Af
r
ica
E
m
ail: d
u
c
@
t
u
t.a
c.
za
1.
I
NT
RO
D
UCT
I
O
N
Veh
icle
a
n
d
v
eh
icle
lice
n
s
e
p
late
tr
ac
k
i
n
g
a
n
d
r
ec
o
g
n
i
tio
n
m
an
a
g
e
m
e
n
t
s
y
s
te
m
s
p
la
y
a
n
i
m
p
o
r
tan
t
r
o
le
in
a
n
i
n
telli
g
e
n
t
tr
a
n
s
p
o
r
tatio
n
s
y
s
te
m
,
it
is
co
m
m
o
n
l
y
u
s
ed
in
au
to
m
atic
d
r
iv
in
g
,
v
e
h
icle
th
e
f
t
p
r
ev
en
tio
n
,
ac
ce
s
s
co
n
tr
o
l
s
ec
u
r
it
y
,
h
i
g
h
-
e
f
f
icien
c
y
r
o
ad
w
a
y
s
er
v
ices,
an
d
s
o
o
n
.
Mo
s
t
o
f
th
e
s
e
s
y
s
te
m
s
ar
e
b
ased
o
n
tr
af
f
ic
i
m
a
g
e
o
r
v
id
eo
an
al
y
s
is
,
w
h
er
e
co
m
p
u
ter
v
is
io
n
tech
n
iq
u
es
w
er
e
co
m
m
o
n
l
y
e
m
p
lo
y
ed
.
A
h
u
g
e
r
ec
en
t
d
ev
elo
p
m
e
n
t
w
a
s
f
o
u
n
d
w
it
h
t
h
e
e
m
p
lo
y
m
e
n
t
o
f
m
ac
h
in
e
lear
n
i
n
g
a
n
d
d
ee
p
lear
n
in
g
tec
h
n
iq
u
es,
w
h
ic
h
d
e
m
o
n
s
tr
ate
h
ig
h
er
cla
s
s
i
f
icatio
n
ac
cu
r
ac
y
an
d
r
o
b
u
s
tn
e
s
s
.
B
o
th
d
ee
p
lear
n
in
g
an
d
co
m
p
u
ter
v
i
s
io
n
m
et
h
o
d
s
[
1
]
-
[
1
5
]
w
er
e
d
ev
elo
p
ed
in
v
ar
io
u
s
v
e
h
icle
-
r
elev
a
n
t
ap
p
licatio
n
s
,
s
u
ch
a
s
v
e
h
icle
d
etec
tio
n
,
v
eh
icle
class
i
f
icatio
n
,
v
e
h
icle
p
late
r
ec
o
g
n
itio
n
,
a
n
d
r
o
ad
co
n
d
iti
o
n
m
o
n
ito
r
i
n
g
[
1
6
]
.
C
o
m
p
ar
in
g
w
i
th
co
m
p
u
ter
v
is
io
n
m
et
h
o
d
s
,
th
e
n
e
w
l
y
d
e
v
elo
p
ed
d
ee
p
lear
n
in
g
tech
n
iq
u
es
h
a
v
e
s
o
m
e
ad
v
an
ta
g
e
s
o
n
th
e
g
en
er
al
izatio
n
ca
p
ac
it
y
an
d
r
o
b
u
s
tn
es
s
to
u
n
ce
r
tain
ties
,
n
o
i
s
e,
an
d
o
cc
lu
s
i
o
n
in
i
m
a
g
es,
in
t
h
e
co
s
t
o
f
h
ig
h
er
co
m
p
u
ta
tio
n
lo
ad
an
d
d
em
an
d
o
n
s
a
m
p
le
s
et
s
ize.
Sev
er
al
m
et
h
o
d
s
w
er
e
p
r
o
p
o
s
ed
to
d
etec
t
v
eh
icle
an
d
v
eh
icle
licen
s
e
p
lates
b
ased
o
n
t
h
ese
n
e
w
l
y
d
ev
elo
p
ed
tech
n
iq
u
e
s
,
s
u
c
h
a
s
t
h
e
d
ir
t
y
n
u
m
b
er
p
late
d
etec
tio
n
s
y
s
te
m
b
ased
o
n
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
n
t J
E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
12
,
No
.
1
,
Feb
r
u
ar
y
20
22
:
7
3
1
-
737
732
y
o
u
o
n
l
y
lo
o
k
o
n
ce
(
Y
OL
O
)
[
1
]
,
th
e
licen
s
e
n
u
m
b
er
p
late
r
ec
o
g
n
itio
n
s
y
s
te
m
b
ased
o
n
c
o
n
v
o
l
u
tio
n
al
n
e
u
r
al
n
et
w
o
r
k
(
C
N
N)
,
an
d
r
ec
u
r
r
en
t
n
eu
r
al
n
e
t
w
o
r
k
(
R
NN)
[
2
]
,
th
e
tech
n
iq
u
e
to
i
m
p
r
o
v
e
th
e
ca
r
p
late
r
ec
o
g
n
itio
n
r
ate
b
ased
o
n
s
u
p
p
o
r
t v
ec
to
r
m
ac
h
in
e
s
(
SVM
)
[
3
]
,
th
e
v
e
h
i
cle
d
etec
tio
n
s
y
s
t
e
m
b
ased
o
n
r
eg
io
n
co
n
v
o
lu
t
io
n
al
n
eu
r
al
n
et
w
o
r
k
(
R
C
NN)
[
4
]
,
th
e
v
e
h
icle
d
etec
tio
n
an
d
co
u
n
ti
n
g
s
y
s
te
m
u
s
in
g
t
h
e
co
n
v
o
lu
tio
n
a
l
r
eg
r
ess
io
n
n
eu
r
al
n
et
w
o
r
k
(
C
R
NN)
[
5
]
.
Ho
w
e
v
er
,
th
er
e
ar
e
s
o
m
e
c
h
all
en
g
e
s
to
d
ee
p
lear
n
in
g
-
b
ased
t
ec
h
n
iq
u
es i
n
v
eh
icle
an
d
v
eh
ic
le
licen
s
e
p
late
d
etec
tio
n
an
d
r
ec
o
g
n
it
io
n
.
F
ir
s
t
l
y
,
f
a
k
e
o
b
j
ec
ts
f
r
o
m
a
co
m
p
lex
e
n
v
ir
o
n
m
en
t
ar
o
u
n
d
th
e
v
e
h
icle
r
ed
u
c
e
th
e
ac
cu
r
ac
y
o
f
d
etec
tio
n
b
y
co
n
f
u
s
i
n
g
t
h
e
d
ee
p
n
eu
r
al
n
et
w
o
r
k
s
,
b
u
ild
i
n
g
s
,
tr
ee
s
,
p
ed
estrian
s
,
an
d
o
th
er
o
b
j
ec
ts
s
u
r
r
o
u
n
d
ed
th
e
v
e
h
icl
es.
S
ec
o
n
d
l
y
,
th
e
w
ea
t
h
er
an
d
illu
m
in
a
tio
n
b
lu
r
th
e
v
eh
icle
s
an
d
licen
s
e
p
late
,
s
tr
o
n
g
s
u
n
lig
h
t,
f
o
g
g
y
,
r
ai
n
y
,
an
d
s
h
ad
o
w
.
T
h
ir
d
l
y
,
th
e
h
u
m
an
f
ac
to
r
,
s
p
ee
d
o
f
th
e
v
e
h
icle
,
d
r
iv
er
b
eh
a
v
io
r
s
.
A
ll o
f
th
e
ab
o
v
e
m
a
k
e
th
e
d
i
f
f
icu
lt to
ac
h
ie
v
e
ac
c
u
r
ate
an
d
ef
f
icien
t d
etec
tio
n
.
I
n
th
is
p
ap
er
,
w
e
p
r
o
p
o
s
ed
a
h
ier
ar
ch
ical
R
C
NN
s
y
s
te
m
f
o
r
v
eh
icle
d
etec
tio
n
an
d
lic
en
s
e
p
late
r
ec
o
g
n
itio
n
u
n
d
er
co
m
p
le
x
tr
af
f
ic
b
ac
k
g
r
o
u
n
d
s
.
T
h
e
s
y
s
te
m
co
n
tain
s
f
o
u
r
s
tep
s
:
i
)
a
h
ig
h
-
le
v
el
R
C
NN
t
o
d
etec
t
v
eh
icles
f
r
o
m
tr
a
f
f
ic
i
m
ag
es/
v
id
eo
,
ii
)
a
lo
w
er
lev
el
R
C
NN
i
s
u
s
ed
to
lo
ca
te
th
e
lice
n
s
e
p
late
f
r
o
m
th
e
d
etec
ted
v
eh
icle
r
e
g
io
n
s
,
iii
)
a
s
m
a
ller
R
C
NN
i
s
tr
ai
n
ed
to
d
etec
t
in
d
iv
id
u
al
le
tter
f
r
o
m
th
e
d
etec
ted
lice
n
s
e
p
late,
iv
)
f
i
n
all
y
a
n
R
C
NN
cl
ass
i
f
ied
is
e
m
p
lo
y
ed
to
r
ec
o
g
n
ize
th
e
i
n
d
i
v
id
u
al
lette
r
.
T
o
r
ed
u
ce
co
m
p
u
tatio
n
,
th
e
tr
ain
i
n
g
d
ata
w
er
e
cr
o
p
p
ed
in
to
s
m
a
ller
b
lo
ck
s
.
T
h
e
m
a
in
i
n
n
o
v
atio
n
o
f
t
h
e
p
r
o
p
o
s
ed
m
et
h
o
d
is
m
an
if
e
s
ted
in
t
h
e
id
ea
o
f
h
ie
r
ar
ch
ical
s
tr
u
ct
u
r
e
w
it
h
m
u
lti
p
le
-
lev
e
l
n
et
w
o
r
k
s
f
o
r
s
u
b
-
ta
s
k
s
.
Dif
f
er
en
t
iati
n
g
f
r
o
m
t
h
e
co
m
m
o
n
all
-
in
-
o
n
e
d
ee
p
lear
n
in
g
s
tr
u
ctu
r
e,
t
h
e
p
r
o
p
o
s
ed
m
e
th
o
d
co
n
s
id
er
s
th
e
lice
n
s
e
p
late
r
ec
o
g
n
itio
n
as
m
u
lt
ip
le
s
u
b
-
t
ask
s
,
w
h
ich
en
ab
le
s
t
h
e
o
p
ti
m
izatio
n
o
f
n
et
w
o
r
k
d
ep
th
a
n
d
s
tr
u
ct
u
r
e
f
o
r
each
s
u
b
-
tas
k
.
E
ac
h
R
C
N
N
i
s
d
esi
g
n
ed
w
it
h
ca
r
e
f
u
l c
o
n
s
id
er
atio
n
o
f
t
h
e
co
m
p
lex
it
y
o
f
th
e
p
r
o
b
le
m
to
b
e
s
o
lv
ed
in
th
e
co
r
r
esp
o
n
d
in
g
le
v
el,
s
o
a
n
R
C
NN
o
f
ap
p
r
o
p
r
iate
d
ep
th
i
s
ch
o
s
en
f
o
r
ea
c
h
s
tep
.
T
h
e
ex
p
er
i
m
e
n
t
r
e
s
u
l
ts
s
h
o
w
t
h
e
p
r
o
p
o
s
ed
s
y
s
te
m
h
as
h
ig
h
er
s
p
ee
d
an
d
ac
cu
r
ac
y
.
T
h
e
r
est
o
f
th
e
p
ap
er
is
o
r
g
an
ized
as
f
o
llo
w
s
.
Sectio
n
2
i
n
tr
o
d
u
ce
s
t
h
e
r
elativ
e
w
o
r
k
.
Sectio
n
3
ex
p
lain
s
th
e
p
r
o
p
o
s
ed
m
eth
o
d
.
Sectio
n
4
d
em
o
n
s
tr
ate
s
th
e
e
x
p
er
i
m
e
n
ts
an
d
an
al
y
s
is
.
Secti
o
n
5
co
n
clu
d
es
th
e
p
ap
er
an
d
in
d
icate
s
f
u
tu
r
e
w
o
r
k
.
2.
RE
S
E
ARCH
M
E
T
H
O
D
2
.
1
.
Vehicle
a
nd
licens
e
pla
t
e
det
ec
t
io
n
Veh
icle
d
e
tectio
n
h
as
b
ee
n
p
o
p
u
lar
l
y
s
t
u
d
ied
in
liter
atu
r
e
b
ased
o
n
co
m
p
u
ter
v
is
io
n
,
w
h
ich
i
s
th
e
f
u
n
d
a
m
en
ta
l
s
tag
e
f
o
r
an
au
to
m
atic
d
r
iv
er
s
y
s
te
m
.
T
h
er
e
ar
e
m
a
n
y
al
g
o
r
ith
m
s
d
ev
e
lo
p
ed
.
A
cc
o
r
d
in
g
to
th
e
liter
atu
r
e,
v
eh
icle
d
etec
tio
n
is
ca
teg
o
r
ized
in
to
m
o
v
in
g
v
e
h
i
cle
d
etec
tio
n
an
d
s
tatic
v
eh
ic
l
e
d
etec
tio
n
.
I
n
th
is
p
ap
er
,
w
e
f
o
c
u
s
o
n
o
n
l
y
m
o
v
i
n
g
v
e
h
icle
d
etec
t
io
n
.
T
h
e
f
ea
t
u
r
es,
b
ac
k
g
r
o
u
n
d
s
u
b
tr
ac
tio
n
s
,
f
r
a
m
e
d
if
f
er
en
ce
,
o
p
tical
f
lo
w
,
m
ac
h
i
n
e
lear
n
i
n
g
,
an
d
co
m
b
in
ed
m
et
h
o
d
s
ar
e
u
s
ed
to
d
etec
t
m
o
v
in
g
d
etec
tio
n
.
A
m
eth
o
d
w
a
s
p
r
o
p
o
s
ed
[
6
]
u
s
i
n
g
o
p
tical
f
l
o
w
to
d
etec
t
a
m
o
v
i
n
g
v
e
h
ic
le.
T
h
e
co
lo
r
an
d
te
x
t
u
r
e
f
ea
tu
r
es
w
er
e
u
s
ed
to
r
ed
u
ce
th
e
ef
f
ec
t
s
f
r
o
m
th
e
co
m
p
lex
b
ac
k
g
r
o
u
n
d
,
an
d
th
e
f
u
zz
y
b
ac
k
g
r
o
u
n
d
s
u
b
tr
ac
tio
n
was d
ev
elo
p
ed
[
7
]
f
o
r
m
o
v
i
n
g
v
eh
icle
d
etec
tio
n
.
[
8
]
im
p
r
o
v
ed
th
e
f
r
a
m
e
d
if
f
er
en
ce
m
et
h
o
d
f
o
r
m
o
v
in
g
v
e
h
i
cle
d
etec
tio
n
u
s
i
n
g
i
m
a
g
e
co
n
tr
ast
an
d
m
o
r
p
h
o
lo
g
ical
f
ilter
.
T
h
e
o
p
tical
f
lo
w
w
a
s
co
m
b
in
ed
w
it
h
k
-
n
ea
r
e
s
t
n
eig
h
b
o
r
(
KNN)
to
class
i
f
y
th
e
d
i
f
f
er
en
t
k
i
n
d
s
o
f
m
o
v
i
n
g
v
e
h
icle
s
[
1
2
]
.
A
n
a
n
ti
-
tr
ac
k
i
n
g
s
y
s
te
m
[
1
7
]
w
as
p
r
o
p
o
s
ed
to
d
etec
t
th
e
v
eh
ic
le
b
ased
o
n
Haa
r
f
ea
t
u
r
es
an
d
m
o
d
if
ied
t
h
e
ad
ap
tiv
e
b
o
o
s
tin
g
alg
o
r
it
h
m
.
T
h
e
v
eh
icle
lice
n
s
e
p
late,
as
an
I
D,
ca
n
u
n
iq
u
el
y
id
e
n
ti
f
y
v
eh
ic
les.
So
t
h
e
au
to
m
atic
lic
en
s
e
p
lat
e
d
etec
tio
n
an
d
r
ec
o
g
n
it
io
n
i
s
o
n
e
o
f
th
e
i
m
p
o
r
ta
n
t ta
s
k
s
i
n
a
n
in
tel
lig
e
n
t
d
r
iv
i
n
g
s
y
s
te
m
.
Va
r
io
u
s
m
et
h
o
d
s
h
a
v
e
b
ee
n
d
ev
elo
p
ed
,
a
m
eth
o
d
[
1
0
]
w
a
s
d
ev
elo
p
ed
to
d
etec
t
v
e
h
icle
p
lates
u
s
i
n
g
s
a
lien
t
f
ea
t
u
r
es,
w
it
h
a
s
u
cc
es
s
r
ate
o
f
9
3
.
1
%
r
ep
o
r
ted
.
T
h
e
r
e
ar
v
eh
icle
lig
h
t
s
w
er
e
u
s
ed
to
d
etec
t
th
e
r
an
g
e
o
f
l
icen
s
e
p
lat
e
an
d
th
e
n
id
en
t
if
y
th
e
l
icen
s
e
p
lated
u
s
i
n
g
t
h
e
h
is
to
g
r
a
m
al
g
o
r
ith
m
[
1
1
]
,
w
h
ich
w
as
o
n
l
y
v
alid
ated
b
y
v
eh
ic
les
f
r
o
m
f
iv
e
co
u
n
tr
ies.
T
h
e
av
er
ag
e
s
u
cc
es
s
r
ate
is
9
0
.
4
%.
A
n
au
to
m
atic
licen
s
e
p
late
r
ec
o
g
n
it
io
n
[
1
2
]
w
a
s
p
r
o
p
o
s
ed
u
s
in
g
u
p
d
ated
YOL
O,
w
i
th
a
lice
n
s
e
p
late
r
ec
o
g
n
itio
n
ac
cu
r
ac
y
o
f
7
8
%.
Ho
w
ev
er
,
t
h
e
ac
cu
r
ate
r
ate
s
till
i
s
an
is
s
u
e
f
o
r
r
ea
l
ap
p
licatio
n
s
.
A
n
d
m
o
s
t
o
f
t
h
ese
m
et
h
o
d
s
h
a
v
e
h
ig
h
co
m
p
u
tatio
n
al
r
eq
u
ir
e
m
e
n
t
s
d
u
e
to
th
e
s
y
s
te
m
's
n
ee
d
to
s
ea
r
ch
th
e
v
e
h
icle
s
an
d
v
eh
icle
licen
s
e
p
late
f
r
o
m
lar
g
e
an
d
h
i
gh
-
r
eso
l
u
tio
n
i
m
ag
es
o
r
v
id
eo
s
th
a
t
w
er
e
ca
p
tu
r
ed
f
r
o
m
t
h
e
r
o
ad
.
2
.
2
.
Reg
i
o
n c
o
nv
o
lutio
n neura
l net
w
o
rk
(
RCNN
)
Dee
p
lear
n
in
g
is
o
n
e
o
f
th
e
al
g
o
r
ith
m
s
o
f
t
h
e
m
ac
h
i
n
e
lear
n
in
g
m
et
h
o
d
th
at
w
as
d
ev
elo
p
e
d
b
ased
o
n
th
e
n
eu
r
al
n
et
w
o
r
k
[
1
8
]
.
Var
io
u
s
d
ee
p
lear
n
i
n
g
n
et
w
o
r
k
s
s
t
r
u
ctu
r
e
h
av
e
b
ee
n
d
ep
leted
,
s
u
ch
as
Alex
Ne
t
[
1
9
]
,
Ov
er
f
ea
t
[
2
0
]
,
Go
o
g
L
eNe
t
[
2
1
]
,
R
esNet
[
2
2
]
.
R
C
NN
is
t
h
e
m
o
s
t
p
o
p
u
lar
m
o
d
el
o
f
d
ee
p
lear
n
in
g
b
ec
au
s
e
o
f
th
e
s
i
g
n
i
f
ican
t
s
u
cc
es
s
in
i
m
a
g
e
p
r
o
ce
s
s
in
g
.
T
h
e
R
C
NN
i
s
m
ad
e
u
p
o
f
n
e
u
r
o
n
s
t
h
at
h
a
v
e
tu
n
ab
le
w
eig
h
t
s
an
d
b
ias.
T
h
e
in
p
u
t
d
ata
ca
n
b
e
i
m
a
g
es,
t
h
e
la
y
er
s
o
f
R
C
NN
h
av
e
3
d
i
m
e
n
s
io
n
s
(
w
id
t
h
,
h
eig
h
t,
a
n
d
d
ep
th
)
.
R
C
N
Ns
ar
e
w
id
el
y
u
s
ed
i
n
v
e
h
icle
a
n
d
v
e
h
icle
n
u
m
b
er
p
lat
e
d
etec
tio
n
.
Fo
r
in
s
ta
n
ce
,
B
r
o
d
y
[
1
3
]
d
etec
ted
th
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8708
A
h
iera
r
ch
ica
l R
C
N
N
fo
r
ve
h
i
cle
a
n
d
ve
h
icle
licen
s
e
p
la
te
d
etec
tio
n
a
n
d
…
(
C
h
u
n
lin
g
Tu
)
733
lan
es a
n
d
v
e
h
icle
s
u
s
i
n
g
C
NN
.
T
h
e
f
r
o
n
t c
o
llis
io
n
w
as p
r
ed
icate
d
u
s
i
n
g
C
NN
w
it
h
ad
ap
tiv
e
r
eg
io
n
-
of
-
i
n
ter
es
t
[
1
4
]
.
Sev
er
al
r
ec
en
t
s
t
u
d
ies
[
1
5
]
,
[
2
3
]
d
ev
elo
p
ed
o
n
e
s
y
s
te
m
f
o
r
lice
n
s
e
p
late
r
ec
o
g
n
it
io
n
b
ased
o
n
R
C
N
N.
I
n
th
is
p
ap
er
,
a
h
ier
ar
ch
ical
R
C
NN
w
i
ll
b
e
e
m
p
lo
y
ed
to
o
b
tai
n
h
i
g
h
ac
cu
r
ac
y
a
n
d
f
ast
s
p
ee
d
f
o
r
v
eh
icle
lice
n
s
e
p
late
d
etec
tio
n
.
3.
P
RO
P
O
SE
D
M
E
T
H
O
D
Dee
p
lear
n
in
g
tec
h
n
iq
u
es
ar
e
co
m
m
o
n
l
y
a
h
u
n
g
er
f
o
r
co
m
p
u
tatio
n
r
eso
u
r
ce
s
.
C
o
m
p
ar
ed
to
C
NN,
th
e
R
C
NN
g
et
s
th
e
ad
v
a
n
ta
g
e
o
f
h
ig
h
er
co
m
p
u
tatio
n
ef
f
ic
ien
c
y
.
T
h
e
tr
ad
itio
n
al
all
-
in
-
o
n
e
m
et
h
o
d
d
id
n
o
t
leav
e
m
u
ch
s
p
ac
e
f
o
r
m
o
d
i
f
icatio
n
s
ac
co
r
d
in
g
to
t
h
e
r
ea
l
-
w
o
r
ld
s
ce
n
ar
io
,
s
u
ch
as
th
e
tas
k
co
m
p
le
x
it
y
,
f
ea
t
u
r
es
to
b
e
co
n
s
id
er
ed
,
i
m
ag
e
q
u
al
ities
,
a
n
d
t
h
e
r
elatio
n
s
h
ip
a
m
o
n
g
s
u
b
-
tas
k
s
.
T
h
is
p
ap
er
is
in
ten
d
i
n
g
to
p
r
o
p
o
s
e
a
h
ier
ar
ch
ical
s
tr
u
c
tu
r
e
to
s
ep
ar
ate
th
e
all
-
in
-
o
n
e
R
C
NN
i
n
to
m
u
ltip
le
lev
e
ls
,
w
it
h
ea
ch
la
y
er
f
o
cu
s
es
o
n
a
s
i
n
g
le
s
u
b
-
tas
k
.
T
h
er
ef
o
r
e,
th
e
n
et
w
o
r
k
in
ea
c
h
la
y
er
will
b
e
d
ev
elo
p
ed
ac
co
r
d
in
g
to
th
e
r
ea
l
d
e
m
a
n
d
o
f
th
e
s
u
b
-
ta
s
k
o
n
l
y
.
T
h
is
w
ill
le
ad
to
th
e
lo
w
er
co
m
p
le
x
it
y
o
f
th
e
to
tal
task
,
w
h
ic
h
r
esu
lt
s
in
a
s
m
aller
n
et
w
o
r
k
s
ize,
lo
w
er
co
m
p
u
tatio
n
lo
ad
,
an
d
h
i
g
h
er
to
tal
ac
c
u
r
ac
y
.
Usi
n
g
t
h
e
v
eh
icle
p
late
r
ec
o
g
n
itio
n
ap
p
licatio
n
s
,
t
h
i
s
p
ap
er
e
m
p
lo
y
s
R
C
NN
f
o
r
3
lev
el
s
u
b
-
tas
k
s
,
v
eh
ic
le
d
etec
tio
n
,
p
late
d
et
ec
t
io
n
,
an
d
c
h
ar
ac
ter
le
v
el
d
etec
t
io
n
a
n
d
r
ec
o
g
n
itio
n
.
T
h
ese
3
t
ask
s
h
av
e
d
if
f
er
en
t
co
m
p
le
x
it
y
o
n
th
e
b
ac
k
g
r
o
u
n
d
an
d
o
b
j
ec
t
f
ea
tu
r
es.
T
h
e
m
ai
n
id
ea
is
to
co
n
s
id
er
R
C
NNs
w
it
h
d
if
f
er
e
n
t
co
m
p
le
x
it
y
le
v
els
f
o
r
t
h
ese
t
ask
s
.
Fi
g
u
r
e
1
d
ep
icts
t
h
e
p
r
o
p
o
s
ed
h
ier
ar
c
h
ical
s
tr
u
ctu
r
e
o
f
th
e
s
o
lu
tio
n
f
o
r
v
eh
ic
le
-
p
late
-
c
h
ar
ac
ter
d
etec
tio
n
an
d
id
en
ti
f
icatio
n
,
w
h
er
e
m
u
l
tip
le
R
C
NNs
w
i
th
f
ea
s
i
b
le
co
m
p
le
x
it
y
ar
e
co
n
s
id
er
ed
f
o
r
d
if
f
er
en
t
tas
k
s
.
T
h
e
h
ier
ar
ch
ical
s
tr
u
ct
u
r
e
i
s
n
o
t
o
n
l
y
f
o
r
R
C
NN,
v
ar
io
u
s
p
r
e
-
tr
ain
ed
d
ee
p
n
eu
r
al
n
et
w
o
r
k
s
ca
n
b
e
co
n
s
i
d
er
ed
to
r
ep
lace
th
e
R
C
NN.
I
n
th
i
s
p
ap
er
,
th
e
R
C
NN
is
co
n
s
id
er
ed
j
u
s
t
as
an
ex
a
m
p
le
to
s
h
o
w
th
e
id
ea
.
T
h
e
v
eh
ic
le
le
v
el
R
C
NN
i
s
t
h
e
m
o
s
t
co
m
p
le
x
o
n
e
b
ec
au
s
e
t
h
e
v
eh
icles
ar
e
co
m
m
o
n
l
y
f
o
u
n
d
i
n
b
u
s
y
tr
af
f
ic
co
n
tex
t
s
w
h
e
n
s
m
ar
t
t
r
af
f
ic
m
o
n
i
to
r
i
n
g
s
y
s
te
m
s
ar
e
co
n
ce
r
n
ed
.
D
etec
t
in
g
v
eh
icl
es
w
i
th
ac
ce
p
tab
le
ac
cu
r
ac
y
is
cr
itical
f
o
r
to
tal
s
y
s
te
m
p
er
f
o
r
m
a
n
ce
.
I
n
t
h
is
le
v
el
,
a
1
5
la
y
er
R
C
NN
is
e
m
p
lo
y
ed
,
a
s
s
h
o
w
n
in
Fig
u
r
e
2
,
w
h
er
e
th
e
R
C
NN
h
a
s
3
f
o
ld
s
o
f
co
n
v
o
lu
tio
n
al
-
p
o
o
lin
g
la
y
er
s
,
w
h
ic
h
ar
e
p
r
e
-
tr
ain
ed
.
Af
ter
th
e
v
eh
icle
i
s
d
etec
ted
b
y
t
h
e
v
e
h
icle
lev
e
l
R
C
N
N,
w
e
ca
n
g
et
th
e
r
eg
io
n
o
f
th
e
v
eh
ic
le.
T
h
e
p
late
lev
el
R
C
NN
o
n
l
y
f
o
c
u
s
e
s
o
n
th
i
s
r
eg
io
n
,
i
n
s
tead
o
f
th
e
w
h
o
le
i
m
ag
e.
T
h
is
g
r
ea
tl
y
r
ed
u
ce
s
th
e
co
m
p
u
tatio
n
lo
ad
.
Me
an
w
h
il
e,
th
i
s
s
tr
ate
g
y
ca
n
a
v
o
id
m
i
s
-
h
it
in
p
late
d
etec
tio
n
,
b
ec
a
u
s
e
o
t
h
er
p
o
ten
tia
l
ca
n
d
id
ates in
t
h
e
b
ac
k
g
r
o
u
n
d
ar
e
ex
clu
d
ed
.
B
ec
au
s
e
t
h
e
p
r
o
b
le
m
to
d
ete
ct
a
lice
n
s
e
p
late
f
r
o
m
a
v
e
h
icle
is
r
ele
v
a
n
tl
y
s
i
m
p
ler
t
h
a
n
d
etec
ti
n
g
v
eh
ic
les
f
r
o
m
th
e
b
ac
k
g
r
o
u
n
d
,
th
is
p
late
le
v
el
R
C
N
N
ca
n
b
e
s
m
a
ller
(
h
a
v
i
n
g
f
e
w
er
la
y
er
s
)
.
Fig
u
r
e
3
s
h
o
ws
th
e
1
2
-
la
y
er
s
tr
u
c
tu
r
e
o
f
th
i
s
R
C
N
N,
w
h
er
e
o
n
l
y
2
f
o
ld
s
o
f
co
n
v
o
lu
tio
n
-
R
e
L
U
-
P
o
o
lin
g
l
a
y
er
s
ar
e
co
n
tain
ed
.
B
o
th
th
e
in
p
u
t a
n
d
t
h
e
f
i
n
al
la
y
er
s
ar
e
th
e
s
a
m
e
as t
h
e
v
eh
ic
l
e
lev
el
R
C
NN.
Fig
u
r
e
1
.
T
h
e
h
ier
ar
ch
ical
s
tr
u
ctu
r
e
o
f
th
e
p
r
o
p
o
s
ed
R
C
NN
v
eh
icle
id
en
t
if
ica
tio
n
s
y
s
te
m
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
n
t J
E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
12
,
No
.
1
,
Feb
r
u
ar
y
20
22
:
7
3
1
-
737
734
Fig
u
r
e
2
.
T
h
e
s
tr
u
ctu
r
e
o
f
t
h
e
v
eh
ic
le
lev
el
R
C
NN
Fig
u
r
e
3
.
T
h
e
s
tr
u
ctu
r
e
o
f
R
C
NN
in
t
h
e
ch
ar
ac
ter
an
d
t
h
e
p
late
lev
els
T
h
e
f
ea
tu
r
es
u
s
ed
to
d
etec
t
an
d
r
ec
o
g
n
ize
ch
ar
ac
ter
s
f
r
o
m
t
h
e
p
late
i
m
ag
e
ar
e
ev
e
n
les
s
.
So
th
eo
r
etica
ll
y
th
e
ch
ar
ac
ter
le
v
el
R
C
NN
ca
n
b
e
s
m
aller
t
h
an
th
e
p
late
le
v
el.
Ho
w
e
v
er
,
th
e
p
late
lev
el
R
C
NN
is
alr
ea
d
y
v
er
y
s
h
al
lo
w
(
o
n
l
y
2
co
n
v
o
lu
tio
n
la
y
er
s
)
,
it
is
n
o
t
f
ea
s
ib
le
to
r
ed
u
ce
th
e
la
y
er
s
f
u
r
th
er
to
av
o
id
th
e
lo
w
f
itti
n
g
ca
p
ac
it
y
o
f
th
e
R
C
NN.
T
h
e
ch
ar
ac
ter
lev
el
R
C
N
N
tak
es
th
e
s
a
m
e
s
tr
u
c
tu
r
e
as
th
e
p
late
le
v
el
o
n
e.
I
t
s
h
o
u
ld
b
e
n
o
ted
th
at
i
f
t
h
e
v
eh
icle
le
v
el
d
ee
p
n
et
w
o
r
k
h
as
m
o
r
e
co
n
v
o
l
u
tio
n
la
y
er
s
,
o
n
e
c
an
e
x
p
ec
t
th
e
ch
ar
ac
ter
lev
el
R
C
NN
ca
n
b
e
s
m
al
ler
th
a
n
th
e
p
late
le
v
el.
4.
E
XP
E
R
I
M
E
NT
S
AN
D
DIS
CUSS
I
O
N
T
o
v
alid
ate
th
e
p
r
o
p
o
s
ed
m
eth
o
d
,
ex
p
er
im
e
n
t
s
ar
e
d
esig
n
ed
b
ased
o
n
th
e
p
u
b
lic
ac
ce
s
s
ib
le
C
an
ad
ia
n
I
n
s
tit
u
te
f
o
r
A
d
v
an
ce
d
R
e
s
ea
r
ch
,
1
0
class
es
(
C
I
F
AR
-
10
)
da
t
ab
ase
[
2
4
]
.
4
.
1
.
E
x
peri
m
ent
env
iro
n
m
e
nt
A
ll
t
h
e
ex
p
er
i
m
en
ts
in
t
h
is
p
ap
er
ar
e
co
m
p
leted
o
n
a
L
en
o
v
o
T
h
in
k
p
ad
X1
lap
to
p
,
w
i
t
h
1
6
GB
R
A
M,
an
I
n
tel(
R
)
C
o
r
e(
T
M)
i
7
-
8
5
5
0
C
P
U
at
1
Gh
z.
T
h
e
o
p
er
atin
g
s
y
s
te
m
is
w
i
n
d
o
w
s
1
0
x
6
4
.
T
h
e
s
o
f
t
w
ar
e
is
Ma
tlab
R
2
0
1
9
a
w
it
h
t
h
e
f
o
llo
w
i
n
g
to
o
lb
o
x
es:
co
m
p
u
te
r
v
is
io
n
,
i
m
a
g
e
p
r
o
ce
s
s
i
n
g
,
d
ee
p
lear
n
in
g
,
an
d
s
tatis
t
ics a
n
d
m
ac
h
in
e
lear
n
i
n
g
.
T
h
e
R
C
NN
s
ar
e
m
o
d
if
ied
a
n
d
tr
ain
ed
b
ased
o
n
th
e
C
I
F
A
R
-
10
n
et
w
o
r
k
[
2
5
]
.
4
.
2
.
Reduce
t
he
la
y
er
s
o
f
R
CNN
Ass
o
ciati
n
g
th
e
co
m
p
lex
it
y
o
f
R
C
NN
to
th
e
tas
k
is
th
e
m
ain
ad
v
a
n
ta
g
e
o
f
t
h
e
p
r
o
p
o
s
ed
s
tr
ateg
y
.
T
h
is
ex
p
er
i
m
e
n
t
ai
m
s
to
d
e
m
o
n
s
tr
ate
t
h
e
f
ea
s
ib
ili
t
y
o
f
t
h
e
d
ed
u
ctio
n
o
f
th
e
C
I
F
AR
-
1
0
Netw
o
r
k
.
T
h
e
o
r
ig
i
n
al
C
I
F
A
R
-
1
0
Net
w
o
r
k
(
1
5
-
la
y
er
R
C
NN)
h
as
1
i
m
a
g
e
in
p
u
t
la
y
er
,
3
f
o
ld
s
o
f
C
o
n
v
o
lu
tio
n
-
R
eL
U
-
P
o
o
lin
g
la
y
er
s
as
th
e
m
id
d
le
la
y
er
,
a
n
d
th
e
f
i
n
al
la
y
er
s
,
a
s
s
h
o
w
n
in
Fi
g
u
r
e
2
.
T
h
e
f
ir
s
t
e
x
p
er
i
m
e
n
t
i
s
to
r
e
m
o
v
e
1
f
o
ld
o
f
th
e
C
o
n
v
o
lu
tio
n
R
e
L
U
-
P
o
o
lin
g
lay
er
s
f
r
o
m
t
h
e
m
id
d
le
la
y
er
s
.
T
h
e
m
o
d
if
ied
n
et
w
o
r
k
h
as
t
h
e
s
t
r
u
ctu
r
e
as
s
h
o
w
n
in
Fi
g
u
r
e
3
.
Af
ter
tr
ain
i
n
g
u
s
i
n
g
t
h
e
C
I
F
AR
-
1
0
d
ataset,
th
e
w
eig
h
t
o
f
t
h
e
f
ir
s
t
co
n
v
o
lu
tio
n
la
y
er
f
o
r
th
e
o
r
ig
i
n
al
n
et
w
o
r
k
(
3
co
n
v
o
l
u
tio
n
la
y
er
s
)
an
d
th
e
m
o
d
if
ied
n
et
w
o
r
k
1
2
-
la
y
er
R
C
NN
(
2
co
n
v
o
l
u
tio
n
la
y
er
s
)
ar
e
s
h
o
w
n
in
F
ig
u
r
e
4
a
n
d
Fi
g
u
r
e
5
r
e
s
p
ec
tiv
el
y
.
Fro
m
t
h
e
w
ei
g
h
ts
,
o
n
e
f
i
n
d
s
t
h
at
b
o
th
o
f
t
h
e
n
et
wo
r
k
s
h
a
v
e
ca
p
t
u
r
ed
th
e
b
asic
f
ea
t
u
r
es
o
f
t
h
e
i
m
ag
es
i
n
th
e
d
ataset.
T
h
er
ef
o
r
e,
th
e
m
o
d
if
icatio
n
o
f
th
e
s
tr
u
ct
u
r
e
d
o
es
n
o
t
s
ig
n
i
f
ica
n
tl
y
d
a
m
a
g
e
th
e
f
ea
t
u
r
e
ex
tr
ac
tio
n
ca
p
ac
it
y
.
T
h
e
n
ex
t
e
x
p
er
i
m
e
n
t
d
ep
icts
t
h
e
f
u
r
th
er
r
ed
u
c
tio
n
o
f
th
e
m
i
d
d
le
la
y
er
s
to
a
s
i
n
g
le
co
n
v
o
l
u
tio
n
la
y
er
(9
-
la
y
er
R
C
N
N)
.
T
h
e
w
eig
h
t
s
o
f
t
h
e
s
in
g
le
co
n
v
o
lu
tio
n
n
et
w
o
r
k
s
h
o
w
n
in
Fi
g
u
r
e
6
s
h
o
w
s
t
h
e
s
tr
o
n
g
r
an
d
o
m
d
is
tr
ib
u
tio
n
,
w
h
ic
h
m
ea
n
s
t
h
e
n
et
w
o
r
k
f
ailed
to
ca
p
tu
r
e
t
h
e
i
m
a
g
e
f
ea
t
u
r
es.
T
h
is
i
s
b
ec
a
u
s
e
t
h
e
r
e
m
o
v
al
o
f
t
w
o
co
n
v
o
lu
t
io
n
la
y
er
s
h
a
s
d
am
ag
ed
t
h
e
lear
n
i
n
g
ca
p
ac
it
y
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8708
A
h
iera
r
ch
ica
l R
C
N
N
fo
r
ve
h
i
cle
a
n
d
ve
h
icle
licen
s
e
p
la
te
d
etec
tio
n
a
n
d
…
(
C
h
u
n
lin
g
Tu
)
735
Fig
u
r
e
4
.
T
h
e
w
ei
g
h
t o
f
t
h
e
f
ir
s
t c
o
n
v
o
l
u
tio
n
la
y
er
i
n
th
e
1
5
-
la
y
er
R
C
N
N
Fig
u
r
e
5
.
T
h
e
w
ei
g
h
t o
f
t
h
e
f
ir
s
t c
o
n
v
o
l
u
tio
n
la
y
er
i
n
th
e
1
2
-
la
y
er
R
C
N
N
Fig
u
r
e
6
.
T
h
e
w
ei
g
h
t o
f
t
h
e
f
ir
s
t c
o
n
v
o
l
u
tio
n
la
y
er
i
n
th
e
9
-
la
y
er
R
C
NN
T
h
e
ac
cu
r
ac
y
d
u
r
in
g
t
h
e
tr
ain
i
n
g
iter
atio
n
s
i
n
Fig
u
r
e
7
co
n
f
i
r
m
ed
th
is
p
o
in
t,
w
h
er
e
t
h
e
1
2
-
la
y
er
a
n
d
15
-
la
y
er
R
C
NNs
h
av
e
s
i
m
ilar
ac
cu
r
ac
y
.
T
h
is
m
ea
n
s
t
h
e
r
ed
u
ctio
n
o
f
t
h
e
la
y
er
s
d
o
es
n
o
t
s
ig
n
i
f
ica
n
tl
y
a
f
f
ec
t
th
e
lear
n
i
n
g
ca
p
ac
it
y
.
Ho
w
ev
er
,
th
e
ac
c
u
r
ac
y
o
f
th
e
9
-
la
y
e
r
R
C
NN
w
a
s
s
i
g
n
i
f
ica
n
tl
y
lo
w
er
ed
.
C
o
n
s
id
er
i
n
g
th
at
th
e
d
ataset
o
f
C
I
F
AR
-
1
0
h
as
1
0
clas
s
es,
t
h
e
ac
c
u
r
ac
y
o
f
1
0
%
o
b
tain
ed
b
y
t
h
e
9
-
la
y
er
R
C
N
N
i
s
j
u
s
t
a
r
an
d
o
m
g
u
es
s.
Fig
u
r
e
7
.
T
h
e
m
i
n
i
-
b
atch
ac
c
u
r
ac
y
d
u
r
in
g
th
e
tr
ai
n
i
n
g
o
f
R
C
NN
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
n
t J
E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
12
,
No
.
1
,
Feb
r
u
ar
y
20
22
:
7
3
1
-
737
736
I
t
s
h
o
u
ld
b
e
n
o
ted
t
h
at,
d
u
e
t
o
th
e
ad
v
a
n
ta
g
e
o
f
t
h
e
tr
a
n
s
f
er
lear
n
in
g
s
tr
ate
g
y
,
th
e
R
C
N
Ns
d
o
n
o
t
n
ee
d
to
r
ea
ch
1
0
0
%
ac
cu
r
ac
y
w
h
e
n
tr
ain
i
n
g
t
h
e
m
id
d
le
la
y
er
s
,
as
lo
n
g
as
t
h
e
m
id
d
le
la
y
e
r
s
ca
n
ca
p
tu
r
e
th
e
i
m
a
g
e
f
ea
tu
r
e
s
.
T
ab
le.
1
s
h
o
w
s
th
e
ac
cu
r
ac
y
o
f
th
e
3
R
C
NN
s
af
ter
th
e
tr
ain
i
n
g
o
f
m
id
d
le
l
a
y
er
s
u
s
i
n
g
t
h
e
1
0
class
es.
W
h
en
t
h
e
co
m
p
u
tatio
n
lo
ad
is
co
n
ce
r
n
ed
,
t
h
e
m
o
d
i
f
ied
R
C
N
Ns
(
1
2
-
la
y
er
an
d
9
-
la
y
er
n
et
w
o
r
k
s
)
h
av
e
h
ig
h
er
ef
f
icie
n
c
y
.
C
o
n
s
i
d
er
in
g
b
o
th
ac
cu
r
ac
y
a
n
d
ti
m
e
e
f
f
icie
n
c
y
,
o
n
e
f
i
n
d
s
th
e
1
2
-
la
y
er
R
C
NN
ca
n
b
e
co
n
s
id
er
ed
as th
e
lo
w
er
-
le
v
el
class
i
f
ier
s
i
n
t
h
e
h
ier
ar
ch
ical
s
tr
u
ctu
r
e
i
n
Fi
g
u
r
e
1
.
T
ab
le
1.
C
o
m
p
ar
is
o
n
o
f
ac
cu
r
ac
y
an
d
co
m
p
u
tatio
n
o
f
th
e
R
C
NNs
R
C
N
N
15
-
l
a
y
e
r
12
-
l
a
y
e
r
9
-
l
a
y
e
r
T
r
a
i
n
i
n
g
T
i
me
(
mi
n
)
49
44
33
T
e
st
S
e
t
A
c
c
u
r
a
c
y
(
%)
5
6
.
7
5
2
.
5
10
4
.
3
.
Vehicle
licens
e
pla
t
e
det
ec
t
io
n a
nd
re
co
g
nitio
n
T
h
is
ex
p
er
i
m
en
t
f
o
c
u
s
e
s
o
n
th
e
s
p
ec
if
ic
ap
p
licatio
n
,
th
e
v
e
h
icle
l
icen
s
e
p
late
d
et
e
ctio
n
a
n
d
r
ec
o
g
n
itio
n
.
T
ab
le
2
s
h
o
w
s
t
h
e
ac
cu
r
ac
y
o
f
ea
ch
R
C
NN
i
n
th
e
p
r
o
p
o
s
ed
s
tr
u
ct
u
r
e.
Fro
m
T
ab
le
2
,
o
n
e
f
in
d
s
th
at
th
e
R
C
NN
s
in
v
e
h
icle
a
n
d
p
late
lev
els
h
av
e
s
i
m
ilar
ac
cu
r
ac
y
,
alth
o
u
g
h
th
e
cla
s
s
i
f
ic
atio
n
ca
p
ac
it
y
w
a
s
lo
w
er
ed
w
h
en
a
co
n
v
o
l
u
tio
n
la
y
er
w
a
s
r
e
m
o
v
ed
.
T
h
is
is
b
e
ca
u
s
e
th
e
p
r
o
b
le
m
co
m
p
lex
it
y
f
o
r
p
late
d
etec
tio
n
is
lo
w
er
th
a
n
t
h
e
v
e
h
icle
d
etec
tio
n
,
th
er
e
f
o
r
e,
a
s
m
al
ler
R
C
N
N
also
ca
n
g
et
s
i
m
i
lar
ac
cu
r
ac
y
.
T
h
is
also
m
ea
n
s
th
er
e
ar
e
s
o
m
e
s
p
ac
es
to
m
o
d
if
y
th
e
all
-
in
-
o
n
e
R
C
NN
with
o
u
t
s
i
g
n
i
f
ican
t
d
a
m
a
g
e
to
th
e
ac
c
u
r
ac
y
.
T
h
e
ch
ar
ac
ter
lev
el
ac
c
u
r
ac
y
i
s
lo
w
er
t
h
an
th
e
o
t
h
er
t
w
o
lev
el
s
alth
o
u
g
h
th
e
c
h
ar
ac
ter
s
h
a
v
e
f
e
w
er
f
ea
tu
r
es
th
a
n
th
e
p
late
an
d
v
e
h
icle.
T
h
is
i
s
b
ec
au
s
e
th
e
cla
s
s
i
f
icat
io
n
p
r
o
b
le
m
in
t
h
e
c
h
ar
ac
ter
lev
el
b
ec
o
m
e
s
a
m
u
lt
ip
le
-
class
(
3
6
class
e
s
)
p
r
o
b
le
m
.
T
h
e
f
i
n
al
la
y
er
s
o
f
t
h
e
c
h
ar
ac
ter
lev
el
s
h
o
u
ld
b
e
i
m
p
r
o
v
ed
,
a
n
d
m
o
r
e
tr
ain
in
g
s
e
t
s
s
h
o
u
ld
b
e
co
n
s
id
er
ed
.
T
h
e
ex
p
er
im
e
n
t
s
d
e
m
o
n
s
tr
ated
th
e
r
ed
u
ctio
n
o
f
R
C
NN
s
an
d
th
e
p
er
f
o
r
m
a
n
ce
o
f
t
h
e
to
tal
s
y
s
te
m
,
w
h
ic
h
v
al
id
ated
th
e
p
r
o
p
o
s
ed
s
tr
ateg
y
.
T
ab
le
2
.
A
cc
u
r
ac
y
o
f
t
h
e
p
r
o
p
o
s
ed
m
e
th
o
d
f
o
r
d
etec
tio
n
an
d
r
ec
o
g
n
itio
n
R
C
N
N
s
V
e
h
i
c
l
e
L
e
v
e
l
P
l
a
t
e
L
e
v
e
l
C
h
a
r
a
c
t
e
r
L
e
v
e
l
A
c
c
u
r
a
c
y
(
%)
9
8
.
5
97
85
5.
CO
NCLU
SI
O
N
T
h
is
p
ap
er
p
r
o
p
o
s
ed
a
h
ier
a
r
ch
ical
s
tr
ateg
y
f
o
r
v
e
h
icle
a
n
d
v
e
h
icle
lice
n
s
e
p
late
d
etec
tio
n
a
n
d
id
en
ti
f
icatio
n
b
ased
o
n
R
C
N
N.
T
h
e
co
m
p
lex
i
t
y
o
f
th
e
R
C
NN
w
as
a
s
s
o
ciate
d
w
it
h
th
e
task
s
.
I
n
th
i
s
w
a
y
,
m
u
ltip
le
co
m
p
lex
le
v
el
R
C
N
Ns
ca
n
b
e
e
m
p
lo
y
ed
in
th
e
s
a
m
e
s
y
s
te
m
.
A
s
a
n
ex
a
m
p
le
,
a
s
am
p
le
v
e
h
icle
d
etec
tio
n
s
y
s
te
m
w
as
d
e
v
elo
p
ed
f
o
r
s
m
ar
t
tr
af
f
ic
m
o
n
ito
r
in
g
,
th
er
ea
f
ter
th
e
licen
s
e
p
late
r
ec
o
g
n
itio
n
R
C
NN
w
a
s
co
n
s
id
er
ed
f
o
r
v
e
h
icle
id
e
n
ti
f
icatio
n
.
T
h
e
v
eh
icle
le
v
el
R
C
N
N
w
it
h
th
e
h
ig
h
est
co
m
p
lex
it
y
w
a
s
e
m
p
lo
y
ed
to
d
etec
t
th
e
v
eh
icle
f
r
o
m
t
h
e
b
ac
k
g
r
o
u
n
d
.
I
n
th
i
s
R
C
NN,
t
h
e
task
ca
n
b
e
co
n
s
id
er
ed
a
s
a
tw
o
-
clas
s
(
v
e
h
icle
an
d
b
ac
k
g
r
o
u
n
d
)
clas
s
i
f
icatio
n
p
r
o
b
lem
.
T
h
e
d
etec
ted
v
e
h
icle
r
eg
io
n
(
a
p
o
r
tio
n
o
f
t
h
e
o
r
ig
i
n
al
i
m
a
g
e)
w
as
th
e
n
i
n
p
u
tted
to
th
e
lice
n
s
e
p
late
lev
el
R
C
NN,
w
h
er
e
t
h
e
lice
n
s
e
p
late
w
a
s
d
etec
ted
.
T
h
is
i
s
also
a
t
w
o
-
clas
s
cla
s
s
i
f
icat
io
n
p
r
o
b
le
m
(
lice
n
s
e
p
late
an
d
v
e
h
icle
b
o
d
y
a
s
b
ac
k
g
r
o
u
n
d
)
.
I
n
th
i
s
w
a
y
,
t
h
e
co
m
p
u
tatio
n
lo
ad
i
s
g
r
ea
tl
y
r
ed
u
ce
d
.
Fu
r
th
er
m
o
r
e,
th
e
d
is
tu
r
b
an
ce
s
f
r
o
m
o
u
ts
id
e
o
f
th
e
v
eh
icle
w
er
e
e
x
cl
u
d
ed
,
w
h
ich
i
m
p
r
o
v
ed
th
e
s
u
cc
ess
r
at
e
o
f
t
h
e
p
late
le
v
el
R
C
N
N.
Fin
al
l
y
,
t
h
e
d
etec
ted
licen
s
e
p
late
b
ec
a
m
e
t
h
e
in
p
u
t
o
f
t
h
e
ch
ar
ac
ter
lev
e
l
R
C
N
N,
w
h
er
e
th
e
in
d
iv
id
u
al
c
h
ar
ac
ter
s
w
er
e
d
et
ec
ted
an
d
cla
s
s
i
f
ied
.
T
h
is
R
C
NN
s
o
l
v
ed
a
m
u
ltip
le
c
lass
cla
s
s
i
f
icatio
n
p
r
o
b
le
m
,
w
h
er
e
th
e
n
u
m
b
er
s
(
‘
0
’
to
‘
9
’
)
an
d
letter
s
(
‘
a
’
to
‘
z’
)
ar
e
th
e
class
i
f
ic
atio
n
tar
g
ets.
T
h
e
f
u
t
u
r
e
w
o
r
k
i
n
cl
u
d
es
:
(
1
)
s
ep
ar
ate
th
e
to
tal
tas
k
in
t
o
m
o
r
e
le
v
els
an
d
d
es
ig
n
th
e
n
et
w
o
r
k
i
n
ea
c
h
le
v
el
ac
co
r
d
in
g
to
t
h
e
s
p
ec
if
i
c
f
ea
t
u
r
es
i
n
t
h
e
co
r
r
esp
o
n
d
in
g
s
u
b
-
tas
k
;
(
2
)
i
m
p
r
o
v
e
t
h
e
f
i
n
al
la
y
er
s
f
o
r
th
e
c
h
ar
ac
ter
le
v
el
R
C
N
N
to
i
m
p
r
o
v
e
th
e
ac
cu
r
ac
y
o
f
p
late
r
ec
o
g
n
i
ti
o
n
.
RE
F
E
R
E
NC
E
S
[
1
]
R
.
L
a
r
o
c
a
e
t
a
l
.
,
"
A
r
o
b
u
st
r
e
a
l
-
t
i
me
a
u
t
o
mat
i
c
l
i
c
e
n
se
p
l
a
t
e
r
e
c
o
g
n
i
t
i
o
n
b
a
se
d
o
n
t
h
e
Y
O
L
O
d
e
t
e
c
t
o
r
,
"
2
0
1
8
I
n
t
e
rn
a
t
i
o
n
a
l
J
o
i
n
t
C
o
n
f
e
re
n
c
e
o
n
N
e
u
r
a
l
N
e
t
w
o
r
k
s (I
J
C
N
N
)
,
2
0
1
8
,
p
p
.
1
-
1
0
,
d
o
i
:
1
0
.
1
1
0
9
/
I
JC
N
N
.
2
0
1
8
.
8
4
8
9
6
2
9
.
[
2
]
T
.
K
.
C
h
e
a
n
g
,
Y
.
S
.
C
h
o
n
g
,
a
n
d
Y
.
H
T
a
y
,
“
S
e
g
me
n
t
a
t
i
o
n
-
f
r
e
e
v
e
h
i
c
l
e
l
i
c
e
n
se
p
l
a
t
e
r
e
c
o
g
n
i
t
i
o
n
u
si
n
g
C
o
n
v
N
e
t
-
R
N
N
,
”
2
0
1
7
,
a
rXi
v
:
1
7
0
1
.
0
6
4
3
9
.
[
3
]
M
.
S
.
F
a
r
a
q
,
M
.
M
.
M
.
El
D
i
n
,
a
n
d
H
.
A
.
El
sh
e
n
b
a
r
y
,
“
P
a
r
k
i
n
g
e
n
t
r
a
n
c
e
c
o
n
t
r
o
l
u
si
n
g
l
i
c
e
n
se
p
l
a
t
e
d
e
t
e
c
t
i
o
n
a
n
d
r
e
c
o
g
n
i
t
i
o
n
,”
I
n
d
o
n
e
si
a
n
J
o
u
rn
a
l
o
f
El
e
c
t
r
i
c
a
l
E
n
g
i
n
e
e
ri
n
g
a
n
d
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
(
I
J
EEC
S
)
,
v
o
l
.
1
5
,
n
o
.
1
,
p
p
.
4
7
6
-
4
8
3
,
Ju
l
.
2
0
1
9
,
d
o
i
:
1
0
.
1
1
5
9
1
/
i
j
e
e
c
s.v
1
5
.
i
1
.
p
p
4
7
6
-
4
8
3
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8708
A
h
iera
r
ch
ica
l R
C
N
N
fo
r
ve
h
i
cle
a
n
d
ve
h
icle
licen
s
e
p
la
te
d
etec
tio
n
a
n
d
…
(
C
h
u
n
lin
g
Tu
)
737
[
4
]
T
.
T
a
n
g
,
S
.
Z
h
o
u
,
Z
.
D
e
n
g
,
H
.
Z
o
u
,
a
n
d
L
.
L
e
i
,
“
V
e
h
i
c
l
e
d
e
t
e
c
t
i
o
n
i
n
a
e
r
i
a
l
i
mag
e
s b
a
se
d
o
n
r
e
g
i
o
n
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
a
n
d
h
a
r
d
n
e
g
a
t
i
v
e
e
x
a
mp
l
e
mi
n
i
n
g
,
”
S
e
n
s
o
rs
,
v
o
l
.
1
7
,
n
o
.
2
,
p
p
.
1
-
1
7
,
2
0
1
7
,
A
r
t
.
n
o
.
3
3
6
,
d
o
i
:
1
0
.
3
3
9
0
/
s
1
7
0
2
0
3
3
6
.
[
5
]
H
.
T
a
y
a
r
a
,
K
.
G
i
l
S
o
o
,
a
n
d
K
.
T
.
C
h
o
n
g
,
"
V
e
h
i
c
l
e
d
e
t
e
c
t
i
o
n
a
n
d
c
o
u
n
t
i
n
g
i
n
h
i
g
h
-
r
e
so
l
u
t
i
o
n
a
e
r
i
a
l
i
m
a
g
e
s
u
si
n
g
c
o
n
v
o
l
u
t
i
o
n
a
l
r
e
g
r
e
ssi
o
n
n
e
u
r
a
l
n
e
t
w
o
r
k
,
"
i
n
I
EEE
A
c
c
e
ss
,
v
o
l
.
6
,
p
p
.
2
2
2
0
-
2
2
3
0
,
2
0
1
8
,
d
o
i
:
1
0
.
1
1
0
9
/
A
C
C
ESS
.
2
0
1
7
.
2
7
8
2
2
6
0
.
[
6
]
Y
.
C
h
e
n
a
n
d
Q
.
W
u
,
"
M
o
v
i
n
g
v
e
h
i
c
l
e
d
e
t
e
c
t
i
o
n
b
a
se
d
o
n
o
p
t
i
c
a
l
f
l
o
w
e
st
i
mat
i
o
n
o
f
e
d
g
e
,
"
2
0
1
5
1
1
t
h
I
n
t
e
rn
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
N
a
t
u
r
a
l
C
o
m
p
u
t
a
t
i
o
n
(
I
C
N
C
)
,
2
0
1
5
,
p
p
.
7
5
4
-
7
5
8
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
N
C
.
2
0
1
5
.
7
3
7
8
0
8
5
.
[
7
]
X
.
L
u
,
T
.
I
z
u
mi
,
T
.
T
a
k
a
h
a
s
h
i
,
a
n
d
L
.
W
a
n
g
,
"
M
o
v
i
n
g
v
e
h
i
c
l
e
d
e
t
e
c
t
i
o
n
b
a
se
d
o
n
f
u
z
z
y
b
a
c
k
g
r
o
u
n
d
s
u
b
t
r
a
c
t
i
o
n
,
"
2
0
1
4
I
EE
E
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
F
u
zz
y
S
y
st
e
m
s (FU
Z
Z
-
I
EEE)
,
2
0
1
4
,
p
p
.
5
2
9
-
5
3
2
,
d
o
i
:
1
0
.
1
1
0
9
/
F
U
Z
Z
-
I
EEE.
2
0
1
4
.
6
8
9
1
5
7
8
.
[
8
]
S
.
S
a
r
i
k
a
n
a
n
d
A
.
M
.
O
z
b
a
y
o
g
l
u
,
“
A
n
o
mal
y
d
e
t
e
c
t
i
o
n
i
n
v
e
h
i
c
l
e
t
r
a
f
f
i
c
w
i
t
h
i
mag
e
p
r
o
c
e
ssi
n
g
a
n
d
mac
h
i
n
e
l
e
a
r
n
i
n
g
,”
P
ro
c
e
d
i
a
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
,
v
o
l
.
1
4
0
,
pp
.
64
-
6
9
,
2
0
1
8
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
p
r
o
c
s.2
0
1
8
.
1
0
.
2
9
3
.
[
9
]
B
.
H
u
v
a
l
e
t
a
l
.
,
“
A
n
e
mp
i
r
i
c
a
l
e
v
a
l
u
a
t
i
o
n
o
f
d
e
e
p
l
e
a
r
n
i
n
g
o
n
h
i
g
h
w
a
y
d
r
i
v
i
n
g
,”
2
0
1
5
,
a
r
Xi
v
:
1
5
0
4
.
0
1
7
1
6
.
[
1
0
]
M
.
R
.
A
si
f
,
Q
.
C
h
u
n
,
S
.
H
u
ssa
i
n
,
M
.
S
.
F
a
r
e
e
d
,
a
n
d
S
.
K
h
a
n
,
“
M
u
l
t
i
n
a
t
i
o
n
a
l
v
e
h
i
c
l
e
l
i
c
e
n
se
p
l
a
t
e
d
e
t
e
c
t
i
o
n
i
n
c
o
mp
l
e
x
b
a
c
k
g
r
o
u
n
d
s
,”
J
o
u
rn
a
l
o
f
Vi
s
u
a
l
C
o
m
m
u
n
i
c
a
t
i
o
n
a
n
d
I
m
a
g
e
R
e
p
res
e
n
t
a
t
i
o
n
,
v
o
l
.
4
6
,
p
p
.
1
7
6
-
1
8
6
,
2
0
1
7
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
j
v
c
i
r
.
2
0
1
7
.
0
3
.
0
2
0
.
[
1
1
]
A
.
K
u
mar,
A
.
S
h
a
r
ma,
a
n
d
R
.
K
.
S
i
n
g
l
a
,
“
A
n
e
f
f
i
c
i
e
n
t
l
i
c
e
n
se
p
l
a
t
e
t
e
x
t
e
x
t
r
a
c
t
i
o
n
t
e
c
h
n
i
q
u
e
,”
Am
b
i
e
n
t
C
o
m
m
u
n
i
c
a
t
i
o
n
s
a
n
d
C
o
m
p
u
t
e
r
S
y
st
e
m
s
,
v
o
l
.
9
0
4
,
p
p
.
4
8
7
-
4
9
5
,
2
0
1
9
,
d
o
i
:
1
0
.
1
0
0
7
/
9
7
8
-
9
8
1
-
13
-
5
9
3
4
-
7
_
4
3
.
[
1
2
]
H
e
n
d
r
y
a
n
d
R
.
C
.
C
h
e
n
,
“
A
u
t
o
mat
i
c
l
i
c
e
n
se
p
l
a
t
e
r
e
c
o
g
n
i
t
i
o
n
v
i
a
sl
i
d
i
n
g
-
w
i
n
d
o
w
d
a
r
k
n
e
t
-
Y
O
L
O
d
e
e
p
l
e
a
r
n
i
n
g
,
”
I
m
a
g
e
a
n
d
Vi
si
o
n
C
o
m
p
u
t
i
n
g
,
v
o
l
.
8
7
p
p
.
4
7
-
5
6
,
2
0
1
9
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
i
m
a
v
i
s.
2
0
1
9
.
0
4
.
0
0
7
.
[
1
3
]
J.
P
y
o
,
J.
B
a
n
g
a
n
d
Y
.
Je
o
n
g
,
"
F
r
o
n
t
c
o
l
l
i
si
o
n
w
a
r
n
i
n
g
b
a
se
d
o
n
v
e
h
i
c
l
e
d
e
t
e
c
t
i
o
n
u
s
i
n
g
C
N
N
,
"
2
0
1
6
I
n
t
e
r
n
a
t
i
o
n
a
l
S
o
C
D
e
s
i
g
n
C
o
n
f
e
re
n
c
e
(
I
S
O
C
C
)
,
2
0
1
6
,
p
p
.
1
6
3
-
1
6
4
,
d
o
i
:
1
0
.
1
1
0
9
/
I
S
O
C
C
.
2
0
1
6
.
7
7
9
9
8
4
2
.
[
1
4
]
C
.
G
o
u
,
K
.
W
a
n
g
,
Y
.
Y
a
o
a
n
d
Z
.
L
i
,
"
V
e
h
i
c
l
e
l
i
c
e
n
se
p
l
a
t
e
r
e
c
o
g
n
i
t
i
o
n
b
a
se
d
o
n
e
x
t
r
e
mal
r
e
g
i
o
n
s
a
n
d
r
e
st
r
i
c
t
e
d
B
o
l
t
z
man
n
mac
h
i
n
e
s
,
"
i
n
I
EEE
T
r
a
n
sa
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
7
,
n
o
.
4
,
p
p
.
1
0
9
6
-
1
1
0
7
,
A
p
r
.
2
0
1
6
,
d
o
i
:
1
0
.
1
1
0
9
/
TI
T
S
.
2
0
1
5
.
2
4
9
6
5
4
5
.
[
1
5
]
M
.
A
.
R
a
f
i
q
u
e
,
W
.
P
e
d
r
y
c
z
,
a
n
d
M
.
Je
o
n
,
“
V
e
h
i
c
l
e
l
i
c
e
n
se
p
l
a
t
e
d
e
t
e
c
t
i
o
n
u
s
i
n
g
r
e
g
i
o
n
-
b
a
se
d
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
u
r
a
l
n
e
t
w
o
r
k
s,”
S
o
f
t
C
o
m
p
u
t
i
n
g
,
v
o
l
.
2
2
,
p
p
.
6
4
2
9
-
6
4
4
0
,
2
0
1
8
,
d
o
i
:
1
0
.
1
0
0
7
/
s0
0
5
0
0
-
0
1
7
-
2
6
9
6
-
2
.
[
1
6
]
T
.
S
i
r
i
b
o
r
v
o
r
n
r
a
t
a
n
a
k
u
l
,
“
A
n
a
u
t
o
ma
t
i
c
r
o
a
d
d
i
s
t
r
e
ss
v
i
su
a
l
i
n
sp
e
c
t
i
o
n
sy
st
e
m
u
si
n
g
a
n
o
n
b
o
a
r
d
i
n
-
c
a
r
c
a
me
r
a
,”
Ad
v
a
n
c
e
s
i
n
Mu
l
t
i
m
e
d
i
a
,
v
o
l
.
2
0
1
8
,
2
0
1
8
,
A
r
t
.
n
o
.
2
5
6
1
9
5
3
,
d
o
i
:
1
0
.
1
1
5
5
/
2
0
1
8
/
2
5
6
1
9
5
3
.
[
1
7
]
O
.
O
h
e
k
a
a
n
d
C
.
T
u
,
"
F
a
st
a
n
d
I
mp
r
o
v
e
d
R
e
a
l
-
T
i
me
V
e
h
i
c
l
e
A
n
t
i
-
T
r
a
c
k
i
n
g
S
y
st
e
m,"
Ap
p
l
i
e
d
S
c
i
e
n
c
e
s
,
v
o
l
.
1
0
,
n
o
.
1
7
,
2
0
2
0
,
d
o
i
:
1
0
.
3
3
9
0
/
a
p
p
1
0
1
7
5
9
2
8
.
[
1
8
]
Y
.
L
e
C
u
n
,
Y
.
B
e
n
g
i
o
,
a
n
d
G
.
H
i
n
t
o
n
,
“
D
e
e
p
L
e
a
r
n
i
n
g
,
”
N
a
t
u
re
,
v
o
l
.
5
2
1
,
p
p
.
4
3
6
-
4
4
4
,
2
0
1
5
,
d
o
i
:
1
0
.
1
0
3
8
/
n
a
t
u
r
e
1
4
5
3
9
.
[
1
9
]
F
.
N
.
L
a
n
d
o
l
a
,
S
.
H
a
n
,
M
.
W
.
M
o
sk
e
w
i
c
z
,
K
.
A
sh
r
a
f
,
W
.
J.
D
a
l
l
y
,
a
n
d
K
.
K
e
u
t
z
e
r
,
"
S
q
u
e
e
z
e
N
e
t
:
A
l
e
x
N
e
t
-
l
e
v
e
l
a
c
c
u
r
a
c
y
w
i
t
h
5
0
x
f
e
w
e
r
p
a
r
a
me
t
e
r
s a
n
d
<
0
.
5
M
B
mo
d
e
l
si
z
e
,
”
2
0
1
6
,
a
r
Xi
v
:
1
6
0
2
.
0
7
3
6
0
.
[
2
0
]
P
.
S
e
r
man
e
t
,
D
.
E
i
g
e
n
,
X
.
Z
h
a
n
g
,
M
.
M
a
t
h
i
e
u
,
R
.
F
e
r
g
u
s,
a
n
d
Y
.
L
e
C
u
n
,
"
O
v
e
r
F
e
a
t
:
i
n
t
e
g
r
a
t
e
d
r
e
c
o
g
n
i
t
i
o
n
,
l
o
c
a
l
i
z
a
t
i
o
n
a
n
d
d
e
t
e
c
t
i
o
n
u
si
n
g
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
t
w
o
r
k
s
,
"
2
0
1
3
,
a
rX
i
v
:
1
3
1
2
.
6
2
2
9
.
[
2
1
]
C
.
S
z
e
g
e
d
y
e
t
a
l
.
,
"
G
o
i
n
g
d
e
e
p
e
r
w
i
t
h
c
o
n
v
o
l
u
t
i
o
n
s
,
"
2
0
1
5
I
EEE
C
o
n
f
e
ren
c
e
o
n
C
o
m
p
u
t
e
r
Vi
s
i
o
n
a
n
d
P
a
t
t
e
rn
Re
c
o
g
n
i
t
i
o
n
(
C
VPR)
,
2
0
1
5
,
p
p
.
1
-
9
,
d
o
i
:
1
0
.
1
1
0
9
/
C
V
P
R
.
2
0
1
5
.
7
2
9
8
5
9
4
.
[
2
2
]
K
.
H
e
,
X
.
Z
h
a
n
g
,
S
.
R
e
n
a
n
d
J.
S
u
n
,
"
D
e
e
p
r
e
si
d
u
a
l
l
e
a
r
n
i
n
g
f
o
r
i
ma
g
e
r
e
c
o
g
n
i
t
i
o
n
,
"
2
0
1
6
I
EEE
C
o
n
f
e
re
n
c
e
o
n
C
o
m
p
u
t
e
r
Vi
s
i
o
n
a
n
d
Pa
t
t
e
rn
Re
c
o
g
n
i
t
i
o
n
(
C
V
PR)
,
2
0
1
6
,
p
p
.
7
7
0
-
7
7
8
,
d
o
i
:
1
0
.
1
1
0
9
/
C
V
P
R
.
2
0
1
6
.
9
0
.
[
2
3
]
P
.
M
a
r
z
u
k
i
,
F
.
R
a
d
z
i
,
Y
.
C
.
W
o
n
g
,
N
.
A
.
H
a
mi
d
,
N
.
A
l
i
,
a
n
d
M
.
M
.
I
b
r
a
h
i
m,
"
A
d
e
si
g
n
o
f
l
i
c
e
n
se
p
l
a
t
e
r
e
c
o
g
n
i
t
i
o
n
sy
st
e
m
u
si
n
g
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
u
r
a
l
n
e
t
w
o
r
k
,
”
I
n
t
e
rn
a
t
i
o
n
a
l
J
o
u
rn
a
l
o
f
E
l
e
c
t
r
i
c
a
l
a
n
d
C
o
m
p
u
t
e
r
E
n
g
i
n
e
e
ri
n
g
(
I
J
E
C
E)
,
v
o
l
.
9
.
n
o
.
3
,
p
p
.
2
1
9
6
-
2
2
0
4
,
2
0
1
9
,
d
o
i
:
1
0
.
1
1
5
9
1
/
i
j
e
c
e
.
v
9
i
3
.
p
p
2
1
9
6
-
2
2
0
4
.
[
2
4
]
A
.
K
r
i
z
h
e
v
sk
y
,
V
.
N
a
i
r
,
a
n
d
G
.
H
i
n
t
o
n
.
“
T
h
e
C
I
F
A
R
-
1
0
d
a
t
a
se
t
.
”
T
o
r
o
n
t
o
.
e
d
u
.
h
t
t
p
s:
/
/
w
w
w
.
c
s.t
o
r
o
n
t
o
.
e
d
u
/
~
k
r
i
z
/
c
i
f
a
r
.
h
t
ml
(
a
c
c
e
sse
d
D
e
c
.
1
7
,
2
0
2
0
)
.
[
2
5
]
A
.
K
r
i
z
h
e
v
sk
y
,
I
.
S
u
t
sk
e
v
e
r
,
a
n
d
G
.
E.
H
i
n
t
o
n
,
"
I
mag
e
N
e
t
c
l
a
ssi
f
i
c
a
t
i
o
n
w
i
t
h
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
,
"
N
I
P
S
'
1
2
:
Pro
c
e
e
d
i
n
g
s
o
f
t
h
e
2
5
t
h
I
n
t
e
rn
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
N
e
u
r
a
l
I
n
f
o
rm
a
t
i
o
n
Pr
o
c
e
ssi
n
g
S
y
st
e
m
s
,
2
0
1
2
,
v
o
l
.
1
,
p
p
.
1
0
9
7
-
1
1
0
5
.
B
I
O
G
RAP
H
I
E
S O
F
AUTH
O
RS
Chu
n
li
n
g
Tu
re
c
e
iv
e
d
a
b
a
c
h
e
lo
r
d
e
g
re
e
in
c
o
m
p
u
ter
sc
ien
c
e
f
ro
m
T
ian
ji
n
Un
iv
e
rsit
y
o
f
T
e
c
h
n
o
lo
g
y
a
n
d
Ed
u
c
a
ti
o
n
,
Ch
i
n
a
in
2
0
0
2
;
M
T
e
c
h
a
n
d
M
S
c
d
e
g
re
e
s
in
El
e
c
tri
c
a
l
En
g
in
e
e
rin
g
f
ro
m
T
sh
w
a
n
e
Un
iv
e
rsit
y
o
f
Tec
h
n
o
lo
g
y
(S
o
u
t
h
A
f
rica
)
a
n
d
ES
IEE
P
a
ris
Un
iv
e
rsity
(F
ra
n
c
e
)
in
2
0
1
0
;
DT
e
c
h
a
n
d
P
h
.
D.
d
e
g
re
e
s
o
f
El
e
c
tri
c
a
l
En
g
in
e
e
rin
g
f
ro
m
th
e
T
sh
w
a
n
e
Un
iv
e
rsit
y
o
f
Tec
h
n
o
lo
g
y
a
n
d
Un
iv
e
rsity
P
a
ris
E
a
st,
F
ra
n
c
e
in
2
0
1
5
.
He
r
re
se
a
rc
h
in
tere
sts
in
c
lu
d
e
im
a
g
e
p
ro
c
e
ss
in
g
,
A
I,
in
d
u
strial
c
o
n
tro
l
,
m
a
c
h
in
e
lea
rn
in
g
,
d
e
e
p
lea
rn
in
g
,
a
n
d
p
a
tt
e
rn
re
c
o
g
n
it
i
o
n
.
S
h
e
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
d
u
c
@tu
t.
a
c
.
z
a
.
S
h
e
n
g
z
h
i
D
u
re
c
e
i
v
e
d
a
n
M
.
S
.
d
e
g
re
e
in
c
o
n
tro
l
th
e
o
ry
a
n
d
c
o
n
tro
l
e
n
g
in
e
e
rin
g
f
ro
m
T
ian
ji
n
P
o
ly
Tec
h
n
o
lo
g
y
U
n
iv
e
rsity
,
T
ian
ji
n
,
Ch
in
a
,
i
n
2
0
0
1
a
n
d
t
h
e
P
h
.
D.
d
e
g
re
e
in
c
o
n
tro
l
th
e
o
ry
a
n
d
c
o
n
tro
l
e
n
g
in
e
e
rin
g
f
ro
m
Na
n
k
a
i
Un
iv
e
rsit
y
,
T
i
a
n
ji
n
,
Ch
in
a
,
in
2
0
0
5
.
He
is
c
u
rre
n
tl
y
a
p
ro
f
e
ss
o
r
in
th
e
F
re
n
c
h
S
o
u
t
h
A
f
ric
a
In
stit
u
te
o
f
T
e
c
h
n
o
lo
g
y
(F
’S
AT
I),
T
sh
w
a
n
e
Un
iv
e
r
sit
y
o
f
T
e
c
h
n
o
l
o
g
y
,
S
o
u
th
Af
rica
.
His
re
s
e
a
rc
h
i
n
tere
sts
in
c
lu
d
e
c
o
m
p
u
ter
v
isio
n
,
A
I,
p
a
tt
e
rn
re
c
o
g
n
it
io
n
,
a
n
d
h
u
m
a
n
in
lo
o
p
sy
st
e
m
s.
H
e
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
d
u
s@
tu
t.
a
c
.
z
a
.
Evaluation Warning : The document was created with Spire.PDF for Python.