T
E
L
K
O
M
N
I
K
A
T
elec
o
m
m
un
ica
t
io
n,
Co
m
pu
t
ing
,
E
lect
ro
nics
a
nd
Co
ntr
o
l
Vo
l.
18
,
No
.
6
,
Dec
em
b
er
2
0
2
0
,
p
p
.
3
0
19
~3
0
2
5
I
SS
N:
1
6
9
3
-
6
9
3
0
,
ac
cr
ed
ited
First Gr
ad
e
b
y
Kem
en
r
is
tek
d
i
k
ti,
Dec
r
ee
No
: 2
1
/E/KPT
/2
0
1
8
DOI
: 1
0
.
1
2
9
2
8
/TE
L
KOM
NI
K
A.
v
1
8
i6
.
1
6
1
3
3
3019
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//jo
u
r
n
a
l.u
a
d
.
a
c.
id
/in
d
ex
.
p
h
p
/TELK
OM
N
I
K
A
Vehicle lo
g
o
reco
g
nitio
n using
hist
o
g
ra
ms
of oriente
d gra
dient
descripto
r
a
nd
sp
a
rsity
sco
re
K
it
t
ik
hu
n M
ee
t
ho
ng
j
a
n
1
,
T
h
o
ng
cha
i Surinwa
ra
ng
k
o
o
n
2
,
Vinh
T
r
uo
ng
H
o
a
ng
3
1
F
a
c
u
lt
y
o
f
S
c
ien
c
e
a
n
d
Tec
h
n
o
l
o
g
y
,
S
u
a
n
S
u
n
a
n
d
h
a
Ra
jab
h
a
t
Un
i
v
e
rsity
,
Th
a
il
a
n
d
2
F
a
c
u
lt
y
o
f
M
a
n
a
g
e
m
e
n
t
S
c
ien
c
e
,
S
u
a
n
S
u
n
a
n
d
h
a
Ra
jab
h
a
t
Un
i
v
e
rsity
,
T
h
a
il
a
n
d
3
F
a
c
u
lt
y
o
f
C
o
m
p
u
ter
S
c
ien
c
e
,
Ho
Ch
i
M
i
n
h
Ci
ty
O
p
e
n
Un
iv
e
rsit
y
,
Vie
tn
a
m
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Ma
r
2
4
,
2
0
2
0
R
ev
is
ed
Ma
y
2
4
,
2
0
2
0
Acc
ep
ted
J
u
n
2
5
,
2
0
2
0
M
o
st
o
f
v
e
h
icle
h
a
v
e
th
e
sim
il
a
r
stru
c
tu
re
s
a
n
d
d
e
si
g
n
s.
It
is
e
x
trem
e
ly
c
o
m
p
li
c
a
ted
a
n
d
d
iffi
c
u
l
t
to
id
e
n
t
i
fy
a
n
d
c
las
sify
v
e
h
icle
b
ra
n
d
s
b
a
s
e
d
o
n
t
h
e
ir
stru
c
tu
re
a
n
d
sh
a
p
e
.
As
we
re
q
u
i
re
a
q
u
ick
a
n
d
re
li
a
b
le
re
sp
o
n
se
,
so
v
e
h
icle
lo
g
o
s
a
re
a
n
a
lt
e
rn
a
ti
v
e
m
e
th
o
d
o
f
d
e
term
in
in
g
t
h
e
t
y
p
e
o
f
a
v
e
h
i
c
le.
In
th
is
p
a
p
e
r,
we
p
ro
p
o
se
a
m
e
th
o
d
f
o
r
v
e
h
icle
lo
g
o
re
c
o
g
n
it
i
o
n
b
a
se
d
o
n
fe
a
tu
re
se
lec
ti
o
n
m
e
th
o
d
in
a
h
y
b
ri
d
wa
y
.
Ve
h
icle
lo
g
o
ima
g
e
s
a
re
first
c
h
a
ra
c
teriz
e
d
b
y
His
to
g
ra
m
s
o
f
Orie
n
ted
G
ra
d
ien
t
d
e
sc
rip
to
rs
a
n
d
t
h
e
fin
a
l
fe
a
tu
re
s
v
e
c
to
r
a
re
th
e
n
a
p
p
li
e
d
fe
a
tu
re
se
lec
ti
o
n
m
e
th
o
d
t
o
re
d
u
c
e
th
e
irrele
v
a
n
t
in
fo
rm
a
ti
o
n
.
M
o
re
o
v
e
r,
we
re
lea
se
a
n
e
w
b
e
n
c
h
m
a
rk
d
a
tas
e
t
f
o
r
v
e
h
icle
lo
g
o
r
e
c
o
g
n
it
i
o
n
a
n
d
re
tri
e
v
a
l
tas
k
n
a
m
e
ly
,
VLR
-
4
0
.
T
h
e
e
x
p
e
rime
n
tal
re
su
lt
s a
re
e
v
a
lu
a
ted
o
n
th
is d
a
tab
a
se
wh
ic
h
sh
o
w t
h
e
e
ffi
c
ien
c
y
o
f
t
h
e
p
r
o
p
o
se
d
a
p
p
ro
a
c
h
.
K
ey
w
o
r
d
s
:
Featu
r
e
s
elec
tio
n
HOG
d
escr
ip
to
r
I
m
ag
e
class
if
icatio
n
Sp
ar
s
ity
s
co
r
e
Veh
icle
lo
g
o
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
:
Vin
h
T
r
u
o
n
g
Ho
a
n
g
,
Facu
lty
o
f
C
o
m
p
u
ter
Scien
ce
,
Ho
C
h
i M
in
h
C
ity
Op
en
Un
iv
er
s
ity
,
9
7
Vo
Van
T
an
Stre
et,
W
ar
d
6
,
Dis
tr
ict
3
,
HC
M
C
ity
,
Vietn
a
m
.
E
m
ail:
v
in
h
.
th
@
o
u
.
e
d
u
.
v
n
1.
I
NT
RO
D
UCT
I
O
N
Ho
w
to
id
en
tify
a
b
r
an
d
an
d
d
is
tin
g
u
is
h
with
th
e
o
th
er
s
v
is
u
ally
?
E
ac
h
b
r
an
d
h
as
its
o
w
n
lo
g
o
th
at
r
ep
r
esen
t
a
tr
ad
em
a
r
k
,
co
n
tai
n
a
ce
r
tain
m
ea
n
in
g
s
y
m
b
o
lizin
g
th
at
b
r
a
n
d
a
n
d
its
m
an
u
f
ac
tu
r
er
.
T
o
cr
ea
te
an
im
p
r
ess
iv
e
lo
g
o
,
s
p
ec
if
ic
ch
ar
ac
ter
is
tics
n
ee
d
to
p
ay
atten
ti
o
n
to
m
a
n
y
d
etails
in
clu
d
in
g
:
l
ay
o
u
t,
co
lo
r
s
,
lin
es,
an
g
les,
an
d
all
in
f
o
r
m
atio
n
m
u
s
t
b
e
ar
r
an
g
ed
in
co
h
er
en
t
a
n
d
h
ar
m
o
n
y
way
.
T
r
ad
itio
n
al
v
eh
icle
r
ec
o
g
n
itio
n
s
y
s
tem
s
id
en
tify
v
eh
icle
b
ased
o
n
m
an
u
al
h
u
m
an
o
b
s
er
v
ati
o
n
s
v
ia
licen
s
e
p
late
o
r
m
o
d
e
l
o
f
v
eh
icles.
T
h
u
s
,
au
to
m
atic
v
eh
icle
id
en
tific
atio
n
is
a
k
ey
p
r
o
b
lem
in
in
tellig
en
t
tr
an
s
p
o
r
tatio
n
s
y
s
tem
.
E
ac
h
v
eh
icle
h
as
a
u
n
iq
u
e
licen
s
e
p
late,
b
u
t
it
is
d
if
f
icu
l
t
to
tr
ac
k
an
d
id
e
n
tify
s
in
ce
t
h
ese
im
ag
es
ar
e
in
v
er
y
l
o
w
q
u
ality
in
in
s
m
ar
t
s
u
r
v
eilla
n
ce
s
y
s
tem
s
.
Var
io
u
s
wo
r
k
s
h
av
e
b
ee
n
p
r
o
p
o
s
ed
f
o
r
v
e
h
icle
lo
g
o
r
ec
o
g
n
iti
o
n
in
th
e
p
ast.
W
e
b
r
ief
ly
r
ev
i
ew
s
ev
er
al
wo
r
k
s
in
th
is
f
ield
.
Fo
r
ex
am
p
le,
L
lo
r
ca
et
a
l.
[
1
]
ap
p
l
y
h
is
to
g
r
am
s
o
f
o
r
ien
ted
g
r
ad
ien
ts
an
d
s
u
p
p
o
r
t
v
ec
to
r
m
ac
h
in
es
(
SVM
)
class
if
ier
f
o
r
v
e
h
icle
m
an
u
f
ac
tu
r
er
r
ec
o
g
n
itio
n
.
A
r
eg
i
o
n
o
f
in
ter
est
is
ap
p
lied
b
ef
o
r
e
ex
tr
ac
tin
g
h
is
to
g
r
am
s
o
f
o
r
ien
t
ed
g
r
a
d
ien
t
(
HOG
)
f
ea
t
u
r
es
f
r
o
m
lo
g
o
im
ag
es.
Hu
a
n
g
et
a
l.
[
2
]
p
r
o
p
o
s
e
a
s
y
s
tem
f
o
r
ca
r
lo
g
o
s
eg
m
e
n
tatio
n
a
n
d
r
ec
o
g
n
itio
n
b
ased
o
n
an
e
f
f
icien
t
p
r
e
-
t
r
ain
ed
co
n
v
o
lu
tio
n
al
n
eu
r
al
n
etwo
r
k
(
C
NN
)
m
o
d
el.
Hu
an
et
a
l.
[
3
]
p
r
esen
t
a
n
ew
al
g
o
r
ith
m
b
ase
d
o
n
H
o
u
g
h
tr
an
s
f
o
r
m
an
d
De
ep
L
ea
r
n
in
g
is
f
o
r
v
eh
icle
lo
g
o
r
etr
iev
al
task
wh
ich
co
m
b
i
n
e
s
h
ap
e
d
etec
tio
n
an
d
d
ee
p
b
elief
n
etwo
r
k
s
.
Pan
et
a
l.
[
4
]
u
s
e
s
ca
le
in
v
ar
ian
t
f
ea
t
u
r
e
tr
an
s
f
o
r
m
(
S
I
FT
)
d
escr
ip
to
r
a
n
d
C
NN
as
th
e
f
ea
tu
r
e
e
x
tr
ac
tio
n
m
et
h
o
d
f
r
o
m
v
eh
icle
lo
g
o
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
1
6
9
3
-
6
9
3
0
T
E
L
KOM
NI
KA
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l Co
n
tr
o
l
,
Vo
l.
18
,
No
.
6
,
Dec
em
b
e
r
2
0
2
0
:
30
19
-
302
5
3020
im
ag
es.
An
en
h
an
ce
d
lo
g
o
-
r
ec
o
g
n
itio
n
s
y
s
tem
is
p
r
esen
ted
b
y
Ps
y
llo
s
et
a
l.
[
5
]
o
n
th
e
Me
d
ialab
licen
s
e
p
late
r
ec
o
g
n
itio
n
(
L
PR
)
d
ataset.
Hu
an
g
et
a
l.
[
6
]
ap
p
l
y
Fas
ter
-
R
C
NN
m
o
d
el
with
two
d
if
f
er
e
n
t
C
NNs
(
VGG
-
1
6
an
d
R
esNet
-
5
0
)
f
o
r
v
eh
icle
lo
g
o
r
ec
o
g
n
itio
n
.
So
th
ee
s
war
an
an
d
R
am
an
an
[
7
]
p
r
esen
t
a
s
tu
d
y
f
o
cu
s
es
o
n
l
o
ca
l
f
ea
tu
r
es
th
at
d
escr
ib
e
s
tr
u
ctu
r
a
l
ch
ar
ac
ter
is
tics
o
f
th
e
lo
g
o
o
f
a
ca
r
u
s
in
g
a
c
o
ar
s
e
-
to
-
f
i
n
e
s
tr
ateg
y
.
N
ie
et
a
l.
[
8
]
p
r
esen
t
a
n
ew
VL
R
m
eth
o
d
b
ased
o
n
f
o
r
eg
r
o
u
n
d
-
b
ac
k
g
r
o
u
n
d
p
ix
el
-
p
air
f
ea
tu
r
e
.
Dif
f
er
en
t
h
an
d
-
cr
af
te
d
d
escr
ip
to
r
s
(
HOG,
L
B
P
an
d
SI
FT)
ar
e
ap
p
lied
to
ex
tr
ac
t
f
ea
tu
r
es.
C
y
g
an
ek
an
d
M.
W
o
´
zn
ia
k
[
9
]
u
s
e
en
s
em
b
le
class
if
ier
s
b
ased
o
n
h
ig
h
e
r
-
o
r
d
er
s
i
n
g
u
lar
v
alu
e
d
ec
o
m
p
o
s
itio
n
to
class
if
y
v
e
h
icle
lo
g
o
.
Mo
r
e
r
ec
e
n
tly
,
Z
h
ao
an
d
W
an
g
[
1
0
]
i
n
tr
o
d
u
ce
a
m
o
d
if
ied
v
er
s
io
n
o
f
HU
in
v
ar
ian
t
m
o
m
en
t to
r
ep
r
esen
t c
a
r
lo
g
o
im
ag
es.
I
n
d
ee
d
,
m
a
n
y
m
ac
h
in
e
lear
n
in
g
p
r
o
b
lem
s
in
co
m
p
u
ter
v
is
io
n
an
d
s
ev
er
al
r
elate
d
d
o
m
ain
s
n
ee
d
to
d
ea
l
with
v
er
y
h
i
g
h
d
i
m
en
s
io
n
al
d
a
ta.
Ma
n
y
o
f
th
ese
f
ea
tu
r
es
m
a
y
n
o
t
b
e
r
elev
a
n
t
f
o
r
th
e
f
in
al
p
r
ed
ictio
n
task
a
n
d
d
eg
r
ad
e
th
e
class
if
icatio
n
p
er
f
o
r
m
an
ce
.
Mu
ltip
le
s
tu
d
ies
h
av
e
s
h
o
wn
th
at
th
e
class
if
icatio
n
p
er
f
o
r
m
an
ce
ca
n
b
e
im
p
r
o
v
e
d
b
y
elim
in
atin
g
th
es
e
f
ea
tu
r
es.
T
h
ese
is
s
u
es
ca
n
b
e
s
o
lv
ed
b
y
th
e
m
eth
o
d
o
f
th
e
d
im
en
s
io
n
ality
r
ed
u
ctio
n
.
Fo
r
t
h
is
p
u
r
p
o
s
e,
th
e
d
im
en
s
io
n
ality
r
ed
u
ctio
n
ca
n
b
e
ac
h
ie
v
ed
eith
e
r
b
y
f
ea
tu
r
e
e
x
tr
ac
tio
n
o
r
f
ea
tu
r
e
s
elec
tio
n
to
a
lo
w
d
im
en
s
io
n
al
s
p
ac
e.
Featu
r
e
ex
tr
ac
tio
n
r
ef
e
r
s
to
th
e
m
eth
o
d
s
th
at
cr
ea
te
a
s
et
o
f
n
ew
f
ea
tu
r
es
b
ased
o
n
th
e
lin
ea
r
o
r
n
o
n
-
lin
ea
r
co
m
b
in
atio
n
s
o
f
th
e
o
r
ig
in
al
f
ea
tu
r
es.
Fu
r
t
h
er
a
n
aly
s
is
is
p
r
o
b
lem
atic
s
in
ce
we
ca
n
n
o
t
g
et
th
e
p
h
y
s
ical
m
e
an
in
g
s
o
f
th
ese
f
ea
tu
r
es
in
th
e
tr
an
s
f
o
r
m
ed
s
p
ac
e.
E
x
a
m
p
les
o
f
f
e
atu
r
e
ex
tr
ac
tio
n
m
eth
o
d
s
in
clu
d
e
p
r
in
ci
p
al
co
m
p
o
n
e
n
t a
n
aly
s
is
(
PC
A)
[
1
1
]
,
lo
ca
lity
p
r
eser
v
in
g
p
r
o
jectio
n
s
(
L
PP
)
[
1
2
]
.
I
n
c
o
n
t
r
a
s
t
,
t
h
e
f
e
a
t
u
r
e
s
e
l
e
c
t
i
o
n
m
e
t
h
o
d
s
a
i
m
a
t
f
i
n
d
i
n
g
a
d
e
q
u
a
t
e
s
u
b
s
e
t
s
o
f
f
e
a
t
u
r
e
s
b
y
k
e
e
p
i
n
g
s
o
m
e
o
r
i
g
i
n
a
l
f
e
a
t
u
r
e
s
a
n
d
t
h
e
r
e
f
o
r
e
m
a
i
n
t
a
i
n
s
t
h
e
p
h
y
s
i
c
a
l
m
e
a
n
i
n
g
s
o
f
t
h
e
f
e
a
t
u
r
e
s
.
T
h
e
u
s
e
o
f
b
o
t
h
m
e
t
h
o
d
s
h
a
s
t
h
e
a
d
v
a
n
t
a
g
e
o
f
i
m
p
r
o
v
i
n
g
p
e
r
f
o
r
m
a
n
c
e
o
f
c
l
a
s
s
i
f
i
c
a
t
i
o
n
a
n
d
i
n
c
r
e
a
s
i
n
g
c
o
m
p
u
t
a
t
i
o
n
a
l
e
f
f
i
c
i
e
n
c
y
.
R
e
c
e
n
t
l
y
,
f
e
a
t
u
r
e
s
e
l
e
c
t
i
o
n
h
a
s
g
a
i
n
e
d
i
n
c
r
e
a
s
i
n
g
i
n
t
e
r
e
s
t
i
n
t
h
e
f
i
e
l
d
o
f
m
a
c
h
i
n
e
l
e
a
r
n
i
n
g
[
1
3
-
1
6
]
,
d
a
t
a
a
n
a
l
y
s
i
s
[
1
7
-
1
9
]
,
a
n
d
s
u
c
c
e
s
s
f
u
l
l
y
a
p
p
l
i
e
d
i
n
c
o
m
p
u
t
e
r
v
i
s
i
o
n
s
u
c
h
a
s
i
n
f
o
r
m
a
t
i
o
n
r
e
t
r
i
e
v
a
l
[
2
0
-
2
2
]
o
r
v
i
s
u
a
l
o
b
j
e
c
t
t
r
a
c
k
i
n
g
[
2
3
-
2
5
]
.
I
n
t
h
i
s
w
o
r
k
,
w
e
f
o
c
u
s
o
n
t
h
e
a
p
p
l
i
c
a
t
i
o
n
o
f
f
e
a
t
u
r
e
s
e
l
e
c
t
i
o
n
m
e
t
h
o
d
s
t
o
v
e
h
i
c
l
e
l
o
g
o
i
m
a
g
e
s
c
l
a
s
s
i
f
i
c
a
t
i
o
n
b
y
s
p
a
r
s
i
t
y
s
c
o
r
e
.
T
h
is
p
ap
er
is
o
r
g
an
ized
a
n
d
s
tr
u
ctu
r
e
d
as f
o
llo
ws.
Sectio
n
2
in
tr
o
d
u
ce
s
th
e
f
ea
tu
r
e
ex
tr
ac
tin
g
m
eth
o
d
s
b
ased
o
n
th
r
ee
lo
ca
l
i
m
ag
e
d
escr
ip
to
r
s
.
Sectio
n
2
a
n
d
3
p
r
esen
t
p
r
o
p
o
s
ed
ap
p
r
o
ac
h
an
d
e
x
p
er
im
e
n
tal
r
esu
lts
.
Fin
al
ly
,
th
e
co
n
clu
s
io
n
is
d
is
cu
s
s
ed
in
s
ec
tio
n
4
.
2.
T
H
E
F
E
AT
UR
E
E
XT
R
ACT
I
O
N
AND
S
E
L
E
C
T
I
O
N
2
.
1
.
H
is
t
o
g
ra
m
s
o
f
o
rient
ed
g
ra
dient
des
cr
ipto
r
H
i
s
t
o
g
r
a
m
s
o
f
o
r
i
e
n
t
e
d
g
r
a
d
i
e
n
t
(
H
O
G
)
d
e
s
c
r
i
p
t
o
r
is
a
p
p
l
i
e
d
f
o
r
d
i
f
f
e
r
e
n
t
p
r
o
b
l
e
m
s
i
n
m
a
c
h
i
n
e
v
i
s
i
o
n
[
2
6
-
3
2
]
.
H
O
G
f
e
a
t
u
r
e
i
s
e
x
t
r
a
ct
e
d
b
y
c
o
u
n
t
i
n
g
t
h
e
o
c
c
u
r
r
e
n
c
e
s
o
f
g
r
a
d
i
e
n
t
o
r
i
e
n
t
a
t
i
o
n
b
as
e
o
n
t
h
e
g
r
a
d
i
e
n
t
a
n
g
l
e
a
n
d
t
h
e
g
r
a
d
i
e
n
t
m
a
g
n
i
t
u
d
e
o
f
l
o
c
a
l
p
a
t
c
h
e
s
o
f
a
n
i
m
a
g
e
.
T
h
e
g
r
a
d
i
e
n
t
a
n
g
l
e
a
n
d
m
a
g
n
i
t
u
d
e
a
t
e
a
c
h
p
i
x
e
l
a
r
e
c
o
m
p
u
t
e
d
i
n
a
n
8
×
8
p
i
x
e
ls
p
a
t
c
h
.
N
e
x
t
,
6
4
g
r
a
d
i
e
n
t
f
e
a
t
u
r
e
v
e
c
t
o
r
s
a
r
e
d
i
v
i
d
e
d
i
n
t
o
9
a
n
g
u
l
a
r
b
i
n
s
0
-
180
°
(
20
°
e
a
c
h
)
.
T
h
e
g
r
a
d
i
e
n
t
m
a
g
n
i
t
u
d
e
a
n
d
a
n
g
l
e
a
t
ea
c
h
p
o
s
it
i
o
n
(
,
ℎ
)
f
r
o
m
a
n
i
m
a
g
e
a
r
e
c
o
m
p
u
t
e
d
a
s
f
o
l
l
o
w
s
:
∆
=
|
(
−
1
,
ℎ
)
−
(
+
1
,
ℎ
)
|
(
1
)
∆
ℎ
=
|
(
,
ℎ
−
1
)
−
(
,
ℎ
+
1
)
|
(
2
)
(
,
ℎ
)
=
√
∆
2
+
∆
2
(
3
)
(
,
ℎ
)
=
(
∆
∆
)
(
4
)
2
.
2
.
F
ea
t
ure
s
elec
t
io
n
B
a
s
e
d
o
n
t
h
e
a
v
ailab
ilit
y
o
f
s
u
p
er
v
is
ed
in
f
o
r
m
atio
n
(
i.e
.
lab
els),
f
ea
tu
r
e
s
elec
tio
n
tech
n
iq
u
es
ca
n
b
e
g
r
o
u
p
ed
in
to
two
lar
g
e
ca
teg
o
r
ies:
s
u
p
er
v
is
ed
an
d
u
n
s
u
p
er
v
i
s
ed
co
n
tex
t
[
3
3
]
.
Ad
d
itio
n
ally
,
d
if
f
er
en
t
s
tr
ateg
ies
o
f
f
ea
tu
r
e
s
elec
tio
n
ar
e
p
r
o
p
o
s
ed
b
ased
o
n
ev
alu
atio
n
p
r
o
ce
s
s
s
u
ch
as
f
ilter
,
wr
ap
p
er
,
an
d
h
y
b
r
id
m
eth
o
d
s
[
3
4
]
.
Hy
b
r
id
a
p
p
r
o
ac
h
es
in
co
r
p
o
r
ate
b
o
th
f
ilter
an
d
wr
a
p
p
er
i
n
to
a
s
in
g
le
s
tr
u
ctu
r
e,
to
g
iv
e
an
e
f
f
ec
tiv
e
s
o
lu
tio
n
f
o
r
d
im
en
s
io
n
ality
r
ed
u
ctio
n
[
3
5
]
.
L
iu
et
a
l.
ex
ten
d
th
e
u
n
s
u
p
er
v
is
ed
s
p
ar
s
ity
s
co
r
e
to
s
u
p
er
v
is
ed
co
n
tex
t
b
y
u
tili
zin
g
th
e
class
lab
el
in
f
o
r
m
atio
n
[
3
6
,
3
7
]
.
L
et
d
e
n
o
tes
th
e
ℎ
f
ea
tu
r
e
o
f
ℎ
in
s
tan
ce
in
class
,
̂
is
th
e
elem
en
t
o
f
s
p
ar
s
e
s
im
ilar
ity
m
atr
ix
wh
ich
is
co
n
s
tr
u
cted
with
in
th
e
class
,
is
a
N
-
d
im
en
s
io
n
al
v
ec
to
r
with
=1
,
if
b
elo
n
g
s
to
th
e
class
an
d
0
o
th
er
wis
e.
T
h
e
two
p
r
o
p
o
s
ed
s
u
p
er
v
is
ed
s
p
ar
s
ity
s
co
r
e
o
f
th
e
ℎ
f
ea
tu
r
e,
d
en
o
ted
Sp
a
r
s
e
S
c
or
e
,
wh
ich
s
h
o
u
l
d
b
e
m
i
n
im
ized
,
ar
e
d
ef
in
e
d
as f
o
llo
ws:
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KOM
NI
KA
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l Co
n
tr
o
l
V
eh
icle
lo
g
o
r
ec
o
g
n
itio
n
u
s
in
g
h
is
to
g
r
a
ms o
f o
r
ien
ted
g
r
a
d
ien
t d
escr
ip
to
r
a
n
d
… (
K
itti
kh
u
n
Meeth
o
n
g
ja
n
)
3021
Sp
a
r
s
e
S
c
or
e
1
=
∑
∑
(
−
∑
̂
=
1
)
=
1
=
1
(
5
)
Sp
a
r
s
e
S
c
or
e
2
=
∑
∑
(
−
∑
̂
−
=
1
)
=
1
=
1
∑
∑
(
−
µ
)
2
=
1
=
1
(
6
)
Af
ter
ca
lcu
latin
g
th
e
s
co
r
e
f
o
r
ea
ch
f
ea
tu
r
e,
th
e
y
ar
e
s
o
r
ted
in
th
e
ascen
d
in
g
o
r
d
er
o
f
Sp
a
r
s
e
Score
to
s
elec
t
th
e
r
elev
an
t
o
n
es.
I
n
th
e
class
if
icatio
n
ex
p
e
r
im
en
ts
,
L
iu
et
a
l.
ha
v
e
d
e
m
o
n
s
tr
ated
th
at
th
i
s
s
co
r
e
o
u
t
p
er
f
o
r
m
s
o
th
er
m
eth
o
d
s
in
m
o
s
t c
ases
,
esp
ec
ially
f
o
r
m
u
lti
-
class
p
r
o
b
lem
s
[
3
6
]
.
3.
E
XP
E
R
I
M
E
N
T
A
L
RE
SUL
T
S
3
.
1
.
E
x
perim
ent
a
l set
up
D
e
s
p
it
e
t
h
e
v
e
h
i
c
l
e
l
o
g
o
r
e
c
o
g
n
i
t
i
o
n
p
r
o
b
l
e
m
h
a
s
b
e
e
n
s
t
u
d
i
e
d
f
o
r
m
a
n
y
y
e
a
r
s
,
a
f
e
w
p
u
b
l
i
c
l
y
a
v
a
i
l
a
b
l
e
i
s
a
v
a
il
a
b
l
e
f
o
r
t
h
e
c
o
m
p
u
t
e
r
v
i
s
i
o
n
c
o
m
m
u
n
i
t
y
.
T
h
e
r
e
a
r
e
a
f
e
w
d
a
t
as
e
ts
is
a
p
p
l
i
e
d
f
o
r
l
o
g
o
d
e
t
e
c
t
i
o
n
s
u
c
h
as
v
e
h
i
c
l
e
l
o
g
o
d
a
t
a
s
et
w
i
t
h
3
0
c
la
s
s
e
s
,
n
a
m
e
l
y
V
L
D
-
3
0
[
3
8
]
.
T
a
b
l
e
1
a
n
a
l
y
s
e
s
t
h
e
e
x
i
s
t
i
n
g
v
e
h
i
c
l
e
l
o
g
o
d
a
t
a
s
et
s
.
T
h
e
f
i
r
s
t
c
o
l
u
m
n
i
n
d
i
c
a
t
es
t
h
e
d
a
t
a
s
et
n
a
m
e
a
n
d
i
t
s
r
e
f
e
r
e
n
c
e
.
T
h
e
s
e
c
o
n
d
c
o
l
u
m
n
s
h
o
w
t
h
e
a
v
a
i
l
a
b
i
li
t
y
o
f
t
h
e
p
r
e
-
d
e
f
i
n
e
d
t
r
a
i
n
i
n
g
a
n
d
t
e
s
t
i
n
g
s
e
t
.
T
h
i
s
i
n
f
o
r
m
a
t
i
o
n
i
s
i
m
p
o
r
ta
n
t
f
o
r
c
o
m
p
a
r
e
t
h
e
c
l
a
s
s
i
f
i
c
a
ti
o
n
r
e
s
u
l
ts
b
e
c
a
u
s
e
o
t
h
er
r
e
s
e
a
r
c
h
e
r
s
c
a
n
u
s
e
t
h
is
d
e
c
o
m
p
o
s
i
t
i
o
n
t
o
t
r
a
i
n
a
n
d
t
es
t
t
h
e
i
r
m
o
d
e
l
s
i
n
s
t
e
a
d
o
f
u
s
i
n
g
c
r
o
s
s
-
v
a
l
i
d
a
t
i
o
n
m
et
h
o
d
s
.
T
h
e
n
e
x
t
c
o
l
u
m
n
s
r
e
p
r
e
s
e
n
t
t
h
e
n
u
m
b
e
r
o
f
l
o
g
o
s
,
t
o
t
a
l
i
m
a
g
e
s
,
a
n
d
i
t
s
r
e
s
o
l
u
ti
o
n
.
T
h
e
l
a
s
t
c
o
l
u
m
n
s
h
o
w
s
t
h
e
a
v
a
i
l
a
b
i
li
t
y
o
f
t
h
e
c
o
r
r
es
p
o
n
d
i
n
g
d
a
t
a
s
e
t
.
T
h
e
d
r
a
w
b
a
c
k
o
f
t
h
e
s
e
d
a
t
a
s
e
t
s
i
s
t
h
e
l
a
c
k
o
f
p
r
e
-
d
e
f
i
n
e
d
d
e
c
o
m
p
o
s
i
t
i
o
n
f
o
r
t
r
a
i
n
i
n
g
a
n
d
t
e
s
t
i
n
g
s
e
t
a
n
d
m
a
g
e
s
a
r
e
r
e
s
i
z
e
d
t
o
t
h
e
s
a
m
e
r
e
s
o
l
u
t
i
o
n
.
T
o
t
h
i
s
e
n
d
,
w
e
c
o
l
l
e
c
t
e
d
a
n
d
o
r
g
a
n
i
z
e
d
a
l
a
r
g
e
-
s
c
a
l
e
a
n
d
c
o
m
p
r
e
h
e
n
s
i
v
e
i
m
a
g
e
d
a
t
a
b
a
s
e
c
a
l
l
e
d
V
L
R
-
4
0
.
T
h
i
s
d
a
t
a
s
e
t
c
o
n
t
a
i
n
s
i
m
a
g
e
s
t
h
a
t
w
e
r
e
t
a
k
e
n
b
y
d
i
f
f
e
r
e
n
t
u
s
e
r
s
,
i
n
u
n
c
o
n
s
t
r
a
i
n
e
d
c
o
n
d
i
t
i
o
n
.
T
h
e
d
a
t
a
w
a
s
g
a
t
h
e
r
e
d
b
y
c
r
a
w
l
i
n
g
f
r
o
m
w
e
b
p
a
g
e
s
a
n
d
c
r
o
p
p
e
d
s
e
m
i
-
a
u
t
o
m
a
t
i
c
a
l
l
y
.
W
e
k
e
e
p
t
h
e
o
r
i
g
i
n
a
l
r
e
s
o
l
u
t
i
o
n
f
r
o
m
c
r
o
p
p
e
d
l
o
g
o
.
I
t
c
a
n
b
e
d
o
w
n
l
o
a
d
e
d
p
u
b
l
i
c
l
y
a
t
:
h
t
t
p
s
:
/
/
d
a
t
a
.
m
e
n
d
e
l
e
y
.
c
o
m
/
d
a
t
a
s
e
t
s
/
d
r
2
3
3
n
s
3
g
6
/
3
w
h
i
c
h
c
o
n
t
a
i
n
s
t
o
t
a
l
4
,
0
0
0
c
o
l
o
r
i
m
a
g
e
s
o
f
v
e
h
i
c
l
e
l
o
g
o
(
s
e
e
s
e
v
e
r
a
l
e
x
a
m
p
l
e
i
m
a
g
e
s
f
r
o
m
t
h
i
s
d
a
t
a
s
e
t
i
n
1
)
.
T
h
e
s
i
z
e
s
o
f
e
a
c
h
i
m
a
g
e
a
r
e
t
o
t
a
l
l
y
d
i
f
f
e
r
e
n
t
.
T
h
e
r
e
i
s
a
t
o
t
a
l
o
f
4
0
c
l
a
s
s
e
s
o
f
l
o
g
o
i
n
t
h
i
s
o
r
i
g
i
n
a
l
d
a
t
a
s
e
t
.
T
h
e
d
a
t
a
i
s
s
p
l
i
t
i
n
h
a
l
f
t
o
b
e
u
s
e
d
a
s
t
r
a
i
n
i
n
g
a
n
d
t
e
s
t
i
n
g
s
e
t
s
f
o
r
c
l
a
s
s
i
f
i
c
a
t
i
o
n
t
a
s
k
.
V
e
h
i
c
l
e
l
o
g
o
i
m
a
g
e
s
a
r
e
c
r
o
p
p
e
d
f
r
o
m
p
h
o
t
o
o
f
c
a
r
i
m
a
g
e
s
s
e
m
i
-
a
u
t
o
m
a
t
i
c
a
l
l
y
.
T
h
e
y
a
l
s
o
c
o
n
t
a
i
n
b
a
c
k
g
r
o
u
n
d
a
n
d
f
o
r
e
g
r
o
u
n
d
a
n
d
m
a
k
e
t
h
i
s
d
a
t
a
s
e
t
i
s
m
o
r
e
c
h
a
l
l
e
n
g
i
n
g
t
h
a
n
t
h
e
o
t
h
e
r
s
.
F
i
g
u
r
e
1
i
l
l
u
s
t
r
a
t
e
s
t
h
e
t
w
o
-
v
e
h
i
c
l
e
l
o
g
o
s
a
r
e
c
r
o
p
p
e
d
f
r
o
m
a
n
o
r
i
g
i
n
a
l
i
m
a
g
e
.
T
ab
le
1
.
Su
m
m
a
r
y
o
f
th
e
a
v
ailab
le
lo
g
o
d
atab
ases
in
liter
atu
r
e
D
a
t
a
s
e
t
N
a
me
Pre
-
d
e
f
i
n
e
d
Tr
a
i
n
/
T
e
st
N
u
mb
e
r
o
f
l
o
g
o
s
To
t
a
l
i
m
a
g
e
s
R
e
s
o
l
u
t
i
o
n
P
u
b
l
i
c
l
y
Ji
a
n
g
S
u
[
4
]
No
16
4
0
0
1
4
0
×
1
0
0
No
B
e
i
j
i
n
g
U
n
i
v
e
r
si
t
y
[
6
]
Y
e
s
8
4
,
0
0
0
1
0
0
0
×
6
0
0
No
VAP
-
R
o
a
d
F
i
g
[
3
]
No
30
2
0
0
,
0
0
0
v
a
r
i
e
d
No
I
TS
-
V
LR
[
8
]
Y
e
s
35
1
4
,
0
0
0
5
0
×
5
0
No
C
V
LD
[
3
9
]
Y
e
s
13
1
4
,
9
5
0
1
0
×
1
0
1
5
0
×
1
5
0
No
P
r
e
t
r
a
i
n
i
n
g
S
t
r
a
t
e
g
y
[
2
]
No
10
1
1
,
5
0
0
7
0
×
7
0
Y
e
s
X
i
a
me
n
U
n
i
v
e
r
s
i
t
y
v
e
h
i
c
l
e
l
o
g
o
[
4
0
]
Y
e
s
10
1
1
,
5
0
0
7
0
×
7
0
Y
e
s
H
F
U
T
-
V
L
[
4
1
]
N
o
80
3
2
,
0
0
0
6
4
×
6
4
6
4
×
9
6
Y
e
s
X
i
’
a
n
Ji
a
o
t
o
n
g
L
i
v
e
r
p
o
o
l
U
n
i
v
e
r
si
t
y
V
L
[
4
2
]
No
15
1
9
,
7
8
0
6
4
×
6
4
No
V
e
h
i
c
l
e
l
o
g
o
r
e
c
o
g
n
i
t
i
o
n
(
V
LR
-
40
)
[
4
3
]
Y
e
s
40
4
,
0
0
0
v
a
r
i
e
d
Y
e
s
Fig
u
r
e
1
.
E
x
am
p
le
o
f
v
e
h
icle
lo
g
o
im
a
g
es e
x
tr
ac
ted
f
r
o
m
r
ea
l
-
life
s
ce
n
ar
io
.
T
h
e
o
r
ig
i
n
al
im
ag
e
is
o
n
th
e
le
f
t a
n
d
two
lo
g
o
s
ar
e
c
r
o
p
p
ed
f
r
o
m
th
is
im
ag
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
1
6
9
3
-
6
9
3
0
T
E
L
KOM
NI
KA
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l Co
n
tr
o
l
,
Vo
l.
18
,
No
.
6
,
Dec
em
b
e
r
2
0
2
0
:
30
19
-
302
5
3022
F
i
g
u
r
e
2
i
l
l
u
s
t
r
a
t
es
t
h
e
c
r
o
p
p
e
d
i
m
a
g
e
s
f
r
o
m
o
r
i
g
i
n
a
l
i
m
a
g
e
c
r
a
w
l
e
d
f
r
o
m
t
h
e
w
e
b
.
W
e
c
a
n
o
b
s
e
r
v
e
t
h
a
t
th
e
t
w
o
c
r
o
p
p
e
d
i
m
a
g
e
s
o
f
t
h
e
F
e
r
r
a
r
i
l
o
g
o
a
r
e
d
i
f
f
e
r
e
n
t
v
i
s
u
a
l
l
y
b
y
t
h
e
r
e
s
o
l
u
t
i
o
n
,
r
o
t
a
t
i
o
n
,
a
n
d
c
o
l
o
r
.
U
n
d
e
r
t
h
e
c
u
r
r
e
n
t
t
e
c
h
n
o
l
o
g
y
c
o
n
d
i
t
io
n
,
t
h
e
s
e
p
r
o
b
l
e
m
s
m
a
k
e
t
h
is
d
a
t
a
s
et
is
m
o
r
e
c
h
a
l
le
n
g
i
n
g
t
h
a
n
o
t
h
e
r
s
.
T
a
b
l
e
2
p
r
e
s
e
n
t
s
e
x
a
m
p
l
e
i
m
a
g
es
o
f
4
0
d
i
f
f
e
r
e
n
t
l
o
g
o
f
r
o
m
V
L
R
-
4
0
d
at
a
s
et
.
Fig
u
r
e
2
.
Sev
er
al
ca
teg
o
r
ies o
f
VL
R
-
4
0
d
ataset
3
.
2
.
Resul
t
s
T
h
e
r
e
a
r
e
m
a
n
y
c
o
l
o
r
s
p
a
ce
s
a
r
e
p
r
o
p
o
s
e
d
i
n
t
h
e
li
t
e
r
a
t
u
r
e
f
o
r
d
i
f
f
e
r
e
n
t
a
p
p
l
i
c
at
i
o
n
s
.
T
h
e
H
OG
d
e
s
c
r
i
p
t
o
r
i
s
a
p
p
l
i
e
d
t
o
e
x
t
r
a
c
t
f
e
a
t
u
r
es
f
r
o
m
v
e
h
i
c
l
e
l
o
g
o
i
m
a
g
e
o
f
e
a
c
h
c
o
l
o
r
c
o
m
p
o
n
e
n
t
.
T
h
e
f
i
n
a
l
f
e
a
t
u
r
e
i
s
o
b
t
a
i
n
e
d
b
y
f
u
s
i
n
g
a
l
l
f
ea
t
u
r
e
s
f
r
o
m
t
h
r
ee
c
o
m
p
o
n
e
n
t
s
.
T
h
r
e
e
c
o
l
o
r
s
p
ac
e
s
(
R
GB
,
HS
V
,
YC
b
C
r
)
a
r
e
co
n
s
i
d
e
r
e
d
t
o
e
n
c
o
d
e
i
m
a
g
e
s
i
n
c
e
t
h
e
s
e
s
p
a
c
es
a
r
e
wi
d
e
l
y
u
s
e
d
f
o
r
p
a
t
t
e
r
n
r
e
c
o
g
n
i
tio
n
a
p
p
l
i
c
a
t
i
o
n
.
T
h
e
t
r
a
i
n
i
n
g
s
e
t i
s
u
s
e
d
t
o
c
o
m
p
u
t
e
s
p
a
r
s
e
s
c
o
r
e
b
y
(
5
)
a
n
d
(
6
)
.
T
h
e
v
a
l
u
e
o
f
t
h
e
s
e
s
c
o
r
e
s
a
r
e
t
h
e
n
a
p
p
l
i
e
d
t
o
r
a
n
k
f
e
at
u
r
e
s
o
f
t
r
a
in
i
n
g
a
n
d
t
es
t
i
n
g
s
e
t.
H
e
r
e
,
w
e
u
s
e
t
h
e
c
u
t
-
o
f
f
r
at
i
o
is
1
%
n
u
m
b
e
r
o
f
f
e
a
t
u
r
es
t
o
d
et
e
r
m
i
n
e
t
h
e
o
p
t
i
m
u
m
d
i
m
e
n
s
i
o
n
.
T
a
b
l
e
2
p
r
es
e
n
ts
t
h
e
c
l
as
s
i
f
i
c
a
ti
o
n
r
e
s
u
lt
s
o
n
t
h
e
V
L
R
-
4
0
d
a
t
as
e
t
.
T
h
e
f
i
r
s
t
c
o
l
u
m
n
i
n
d
i
c
a
t
es
t
h
e
c
o
l
o
r
s
p
a
c
e
u
s
e
d
t
o
e
n
c
o
d
e
v
e
h
i
c
l
e
l
o
g
o
i
m
a
g
e
s
.
T
h
e
s
e
c
o
n
d
c
o
l
u
m
n
s
h
o
w
s
t
h
e
a
c
c
u
r
a
c
y
a
c
h
i
e
v
ed
o
f
e
a
c
h
s
p
a
c
e
a
n
d
i
ts
d
i
m
e
n
s
io
n
w
h
e
n
n
o
s
e
l
e
c
ti
o
n
m
e
t
h
o
d
i
s
a
p
p
l
i
e
d
.
T
h
e
n
u
m
b
e
r
o
f
f
ea
tu
r
es
is
1
1
,
5
3
2
×3
=
3
4
,
5
9
6
f
ea
tu
r
es.
W
e
s
ee
th
at
th
e
ac
cu
r
ac
y
v
ar
ies
o
n
d
if
f
er
en
t
co
lo
r
s
p
ac
e.
T
h
e
s
ec
o
n
d
an
d
th
ir
d
co
lu
m
n
p
r
esen
t
th
e
class
if
icat
io
n
r
esu
lts
b
y
u
s
in
g
s
p
ar
s
e
s
co
r
e
1
an
d
s
p
ar
s
e
s
co
r
e
2
,
r
esp
ec
tiv
ely
.
T
h
e
s
p
ar
s
e
s
co
r
e
1
clea
r
l
y
o
u
tp
er
f
o
r
m
s
o
th
e
r
m
eth
o
d
s
b
y
g
iv
in
g
th
e
b
est
ac
cu
r
ac
y
(
7
5
.
2
5
%)
b
y
u
s
in
g
7
7
%
(
2
6
,
6
3
8
f
ea
t
u
r
es)
n
u
m
b
er
o
f
f
ea
tu
r
es.
T
h
e
s
p
ar
s
e
s
co
r
e
2
g
iv
e
th
e
ac
c
u
r
ac
y
clo
s
e
to
th
e
r
esu
lts
wh
en
n
o
s
elec
tio
n
m
eth
o
d
is
ap
p
lied
.
Ho
wev
er
,
it
lar
g
ely
r
e
d
u
ce
s
n
u
m
b
er
o
f
f
ea
tu
r
es
co
m
p
ar
in
g
with
s
p
ar
s
e
s
co
r
e
1
.
Fo
r
ex
am
p
le,
s
p
ar
s
e
s
co
r
e
2
o
n
ly
u
s
es
2
0
%
n
u
m
b
e
r
o
f
f
ea
t
u
r
es
o
n
HSV
s
p
ac
e
wh
ile
g
iv
in
g
b
etter
p
e
r
f
o
r
m
an
ce
.
B
y
o
b
s
er
v
i
n
g
t
h
is
tab
l
e,
we
s
ee
t
h
at
f
ea
t
u
r
e
s
elec
tio
n
m
eth
o
d
g
iv
es
th
e
ac
cu
r
ac
y
as
p
er
f
o
r
m
in
g
w
h
en
n
o
s
elec
tio
n
m
eth
o
d
,
b
u
t it
allo
ws to
r
ed
u
ce
th
e
d
im
en
s
io
n
s
p
ac
e.
T
ab
le
2
.
C
lass
if
icatio
n
r
esu
lts
o
n
th
e
VL
R
-
4
0
d
ataset
with
two
f
ea
tu
r
e
s
elec
tio
n
m
eth
o
d
s
C
o
l
o
r
S
p
a
c
e
W
i
t
h
o
u
t
se
l
e
c
t
i
o
n
S
p
a
r
se
S
c
o
r
e
1
S
p
a
r
se
S
c
o
r
e
2
A
c
c
u
r
a
c
y
D
i
me
n
si
o
n
A
c
c
u
r
a
c
y
D
i
me
n
si
o
n
A
c
c
u
r
a
c
y
D
i
me
n
si
o
n
R
G
B
7
2
.
9
5
3
4
,
5
9
6
7
5
.
2
5
7
7
%
7
2
.
9
5
7
0
%
H
S
V
6
6
.
3
0
3
4
,
5
9
6
6
7
.
5
0
8
3
%
6
6
.
9
0
2
0
%
Y
C
b
C
r
7
1
.
2
0
3
4
,
5
9
6
7
4
.
4
0
7
2
%
7
1
.
3
0
8
6
%
Ad
d
itio
n
ally
,
Fig
u
r
e
3
co
m
p
ar
es
th
e
p
er
f
o
r
m
an
ce
o
f
two
s
p
a
r
s
e
s
co
r
e
1
an
d
2
o
n
t
h
r
ee
d
if
f
er
en
t
co
lo
r
s
p
ac
es.
T
h
e
co
m
b
i
n
atio
n
o
f
HSV
co
lo
r
s
p
ac
e
an
d
s
p
ar
s
e
co
r
e
1
g
iv
e
th
e
wo
r
s
t
p
er
f
o
r
m
an
ce
co
m
p
ar
ed
with
o
t
h
er
m
eth
o
d
s
.
T
h
e
R
GB
s
p
ac
e
an
d
s
p
ar
s
e
s
co
r
e
2
g
iv
e
a
g
o
o
d
p
e
r
f
o
r
m
a
n
ce
at
ea
r
ly
s
tag
e
s
in
ce
it
o
n
ly
n
ee
d
f
ew
e
r
th
an
1
0
%
n
u
m
b
er
o
f
f
ea
tu
r
es
to
r
ea
ch
an
ac
cu
r
ac
y
m
o
r
e
th
a
n
7
0
%.
I
n
c
o
n
tr
ast,
th
e
YC
b
C
r
an
d
HSV
s
p
ac
es
co
m
b
in
ed
with
s
p
ar
s
e
s
co
r
e
1
ac
h
iev
e
a
v
er
y
lo
w
ac
cu
r
ac
y
a
t
th
e
b
e
g
in
n
in
g
wh
en
n
u
m
b
er
o
f
s
elec
ted
f
ea
tu
r
es
is
f
ewe
r
th
an
5
5
%.
So
,
ex
p
er
i
m
en
tal
r
esu
lts
s
h
o
w
th
at
it
s
h
o
u
ld
b
e
in
ter
esti
n
g
to
f
in
d
a
s
u
i
tab
le
co
lo
r
s
p
ac
e
to
en
co
d
e
v
eh
icle
lo
g
o
im
ag
es a
n
d
an
ap
p
r
o
p
r
iate
f
ea
tu
r
e
s
elec
tio
n
m
eth
o
d
to
r
e
m
o
v
e
i
r
r
elev
a
n
t f
ea
tu
r
es.
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KOM
NI
KA
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l Co
n
tr
o
l
V
eh
icle
lo
g
o
r
ec
o
g
n
itio
n
u
s
in
g
h
is
to
g
r
a
ms o
f o
r
ien
ted
g
r
a
d
ien
t d
escr
ip
to
r
a
n
d
… (
K
itti
kh
u
n
Meeth
o
n
g
ja
n
)
3023
Fig
u
r
e
3
.
C
lass
if
icatio
n
p
er
f
o
r
m
an
ce
o
f
Sp
a
r
s
e
S
c
or
e
1
an
d
Spa
r
s
e
Scor
e
2
o
n
VL
R
-
4
0
d
at
aset b
y
d
if
f
er
e
n
t
co
lo
r
s
p
ac
es
4.
CO
NCLU
SI
O
N
T
h
i
s
p
ap
er
p
r
esen
ts
a
v
e
h
icle
l
o
g
o
r
ec
o
g
n
itio
n
b
ased
o
n
HO
G
d
escr
ip
to
r
an
d
f
ea
tu
r
e
s
elec
tio
n
v
ia
two
s
p
ar
s
e
s
co
r
e.
W
e
also
r
elea
s
e
a
n
ew
b
en
c
h
m
ar
k
v
eh
icle
lo
g
o
i
m
ag
e
(
VL
R
-
4
0
)
d
ataset
f
o
r
r
esear
ch
co
m
m
u
n
ity
.
T
h
e
ex
p
er
im
en
tal
r
esu
lts
s
h
o
w
th
at
th
e
s
p
ar
s
e
s
co
r
e
1
g
iv
es
th
e
b
est
ac
cu
r
ac
y
o
n
t
h
e
R
GB
co
lo
r
s
p
ac
e
an
d
lar
g
ely
r
ed
u
ce
n
u
m
b
er
o
f
f
ea
t
u
r
es.
T
h
is
s
tu
d
y
is
n
o
w
ex
ten
d
ed
to
co
m
p
ar
e
th
e
p
er
f
o
r
m
an
c
e
an
d
f
in
d
a
s
u
itab
le
co
lo
r
s
p
ac
e
f
o
r
en
c
o
d
in
g
v
eh
ic
le
lo
g
o
im
ag
es.
ACK
NO
WL
E
DG
E
M
E
NT
S
T
h
is
wo
r
k
was su
p
p
o
r
ted
b
y
Su
an
Su
n
an
d
h
a
R
ajab
h
at
Un
i
v
e
r
s
ity
,
T
h
ailan
d
.
RE
F
E
R
E
NC
E
S
[1
]
D.
F
.
Ll
o
rc
a
,
R.
Arro
y
o
,
a
n
d
M
.
A
.
S
o
tel
o
,
“
Ve
h
icle
lo
g
o
re
c
o
g
n
i
ti
o
n
in
traffic
ima
g
e
s
u
sin
g
HO
G
fe
a
tu
re
s
a
n
d
S
V
M
,”
In
1
6
th
In
ter
n
a
ti
o
n
a
l
IEE
E
Co
n
fe
re
n
c
e
o
n
In
telli
g
e
n
t
T
r
a
n
sp
o
rta
t
io
n
S
y
ste
ms
(IT
S
C
2
0
1
3
),
pp.
2
2
2
9
-
2
2
3
4
,
2
0
1
3
.
[2
]
Yu
e
Hu
a
n
g
,
R
u
iwe
n
Wu
,
Ye
S
u
n
,
Wei
Wan
g
,
a
n
d
Xin
g
h
a
o
Din
g
,
“
Ve
h
icle
L
o
g
o
Re
c
o
g
n
i
ti
o
n
S
y
ste
m
Ba
se
d
o
n
Co
n
v
o
l
u
ti
o
n
a
l
Ne
u
ra
l
Ne
two
r
k
s
with
a
P
re
trai
n
in
g
S
t
ra
teg
y
,”
I
EE
E
T
ra
n
sa
c
ti
o
n
s
o
n
In
telli
g
e
n
t
T
ra
n
s
p
o
rt
a
ti
o
n
S
y
ste
ms
,
v
o
l.
16
,
n
o
.
4
,
p
p
.
1
9
5
1
-
1
9
6
0
,
2
0
1
5
.
[3
]
Li
Hu
a
n
,
Wa
n
g
Li
,
a
n
d
Qin
Yu
ji
a
n
,
“
Ve
h
icle
L
o
g
o
Re
tr
iev
a
l
B
a
se
d
o
n
Ho
u
g
h
Tran
sf
o
rm
a
n
d
De
e
p
Lea
rn
in
g
,”
In
ter
n
a
t
io
n
a
l
C
o
n
fer
e
n
c
e
o
n
C
o
mp
u
ter
Vi
si
o
n
W
o
rk
sh
o
p
s (ICC
VW
),
p
p
.
9
6
7
-
9
7
3
,
2
0
1
7
.
[4
]
Ch
u
n
P
a
n
,
M
in
g
x
ia
S
u
n
,
Z
h
ig
u
o
Ya
n
,
Jie
S
h
a
o
,
Xia
o
m
i
n
g
Xu
,
a
n
d
Di
W
u
,
“
Ve
h
icle
l
o
g
o
re
c
o
g
n
i
ti
o
n
b
a
se
d
o
n
d
e
e
p
lea
rn
in
g
a
rc
h
it
e
c
tu
re
i
n
v
id
e
o
su
r
v
e
il
lan
c
e
fo
r
in
telli
g
e
n
t
traffic sy
s
tem
,
”
IET
In
ter
n
a
ti
o
n
a
l
Co
n
fer
e
n
c
e
o
n
S
ma
rt a
n
d
S
u
sta
in
a
b
le Ci
ty 2
0
1
3
(ICS
S
C
2
0
1
3
),
p
p.
132
-
1
3
5
,
2
0
1
3
.
[5
]
Ap
o
sto
l
o
s
P
.
P
sy
ll
o
s,
Ch
rist
o
s
-
Ni
k
o
lao
s
E
.
An
a
g
n
o
st
o
p
o
u
l
o
s,
a
n
d
El
e
fth
e
rio
s
Ka
y
a
fa
s
,
“
Ve
h
icle
Lo
g
o
Re
c
o
g
n
it
i
o
n
Us
in
g
a
S
IF
T
-
Ba
se
d
E
n
h
a
n
c
e
d
M
a
tch
in
g
S
c
h
e
m
e
,”
IEE
E
T
ra
n
s
a
c
ti
o
n
s
o
n
In
tel
li
g
e
n
t
T
ra
n
sp
o
rt
a
ti
o
n
S
y
ste
ms
,
v
o
l.
11
,
n
o
.
2
,
p
p
.
3
2
2
-
3
2
8
,
2
0
1
0
.
[6
]
Z
.
Hu
a
n
g
,
M
.
F
u
,
K
.
Ni,
H
.
S
u
n
,
a
n
d
S
.
S
u
n
,
“
Re
c
o
g
n
i
ti
o
n
o
f
Ve
h
icle
Lo
g
o
Ba
se
d
o
n
F
a
ste
r
-
RCNN
,”
S
o
n
g
li
n
S
u
n
,
e
d
it
o
r,
S
ig
n
a
l
a
n
d
In
f
o
rm
a
ti
o
n
Pr
o
c
e
ss
in
g
,
Ne
two
rk
i
n
g
a
n
d
C
o
mp
u
ter
s
,
v
o
l
.
4
9
4
,
p
p
.
75
-
8
3
,
2
0
1
9
.
[7
]
S
.
S
o
th
e
e
sw
a
ra
n
,
A.
Ra
m
a
n
a
n
,
“
A
Co
a
rse
-
to
-
F
in
e
S
trate
g
y
f
o
r
V
e
h
icle
Lo
g
o
Re
c
o
g
n
it
i
o
n
fr
o
m
F
ro
n
tal
-
Vie
w
Ca
r
Im
a
g
e
s
,”
Pa
tt
e
rn
Rec
o
g
n
it
i
o
n
a
n
d
Ima
g
e
An
a
lys
is
,
v
o
l.
28
,
n
o
.
1
,
p
p
.
1
4
2
-
1
5
4
,
2
0
1
8
.
[8
]
Z
.
Nie
,
Y
.
Yu
,
a
n
d
Q
.
Jin
,
“
A
Ve
h
icle
Lo
g
o
Re
c
o
g
n
it
io
n
Ap
p
r
o
a
c
h
Ba
se
d
o
n
F
o
re
g
r
o
u
n
d
-
Ba
c
k
g
r
o
u
n
d
P
i
x
e
l
-
P
a
ir
F
e
a
tu
re
,”
T
ra
n
s
a
c
ti
o
n
s o
n
E
d
u
t
a
i
n
me
n
t
XII
I
,
p
p
.
2
0
4
-
2
1
4
,
2
0
1
7
.
[9
]
B
.
Cy
g
a
n
e
k
a
n
d
M
.
Wo
´
z
n
iak
,
“
V
e
h
icle
Lo
g
o
Re
c
o
g
n
it
i
o
n
wit
h
a
n
En
se
m
b
le
o
f
Clas
sifiers
,”
Bo
o
tstr
a
p
p
in
g
a
n
d
Ru
le
-
Ba
se
d
M
o
d
e
l
f
o
r
Re
c
o
g
n
izi
n
g
Vi
e
tn
a
m
e
se
Na
m
e
d
En
ti
t
y
,
”
p
p
.
1
1
7
-
126
,
2
0
1
4
.
[1
0
]
J
.
Zh
a
o
a
n
d
X
.
Wan
g
,
“
Ve
h
icle
-
l
o
g
o
re
c
o
g
n
it
io
n
b
a
se
d
o
n
m
o
d
ifi
e
d
HU
in
v
a
ria
n
t
m
o
m
e
n
ts
a
n
d
S
V
M
,”
M
u
l
ti
me
d
i
a
T
o
o
ls
a
n
d
A
p
p
li
c
a
ti
o
n
s
,
v
o
l.
78
,
n
o
.
1
,
p
p
.
75
-
9
7
,
2
0
1
9
.
[1
1
]
K.
F
u
k
u
n
a
g
a
,
“
In
tr
o
d
u
c
ti
o
n
t
o
sta
ti
stica
l
p
a
tt
e
rn
re
c
o
g
n
it
io
n
.
Co
m
p
u
ter
sc
ien
c
e
a
n
d
sc
ien
ti
fic
c
o
m
p
u
ti
n
g
,”
Aca
d
e
mic
Pre
ss
,
Bo
sto
n
,
2
n
d
e
d
e
d
it
io
n
,
1
9
9
0
.
[
1
2
]
X
.
H
e
a
n
d
P
.
N
i
y
o
g
i
,
“
L
o
c
a
l
i
t
y
p
r
e
s
e
r
v
i
n
g
p
r
o
j
e
c
t
i
o
n
s
,”
A
d
v
a
n
c
e
s
i
n
n
e
u
r
a
l
i
n
f
o
r
m
a
t
i
o
n
p
r
o
c
e
s
s
i
n
g
s
y
s
t
e
m
s
,
p
p.
153
-
1
6
0
,
2
0
0
4
.
[1
3
]
S
.
K.
S
h
e
v
a
d
e
a
n
d
S
.
S
.
Ke
e
rth
i
,
“
A
sim
p
le
a
n
d
e
fficie
n
t
a
lg
o
rit
h
m
f
o
r
g
e
n
e
se
lec
ti
o
n
u
sin
g
sp
a
rse
lo
g
i
stic
re
g
re
ss
io
n
,”
Bi
o
in
fo
rm
a
t
ics
,
v
o
l.
19
,
n
o
.
17
,
p
p
.
2
2
4
6
-
2
2
5
3
,
2
0
0
3
.
[1
4
]
T. Li,
C.
Zh
a
n
g
, a
n
d
M
. Ogi
h
a
ra
,
“
A co
m
p
a
ra
ti
v
e
stu
d
y
o
f
fe
a
tu
re
s
e
lec
ti
o
n
a
n
d
m
u
lt
icla
ss
c
las
sifica
ti
o
n
m
e
th
o
d
s f
o
r
ti
ss
u
e
c
las
sifica
ti
o
n
b
a
se
d
o
n
g
e
n
e
e
x
p
re
ss
io
n
,”
Bi
o
i
n
f
o
rm
a
ti
c
s
,
v
o
l
.
20
,
n
o
.
15
,
p
p
.
2
4
2
9
-
2
4
3
7
,
2
0
0
4
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
1
6
9
3
-
6
9
3
0
T
E
L
KOM
NI
KA
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l Co
n
tr
o
l
,
Vo
l.
18
,
No
.
6
,
Dec
em
b
e
r
2
0
2
0
:
30
19
-
302
5
3024
[1
5
]
M
.
Ya
n
g
,
F
.
Wan
g
,
a
n
d
P
.
Ya
n
g
,
“
A
n
o
v
e
l
fe
a
tu
re
se
lec
ti
o
n
a
lg
o
rit
h
m
b
a
se
d
o
n
h
y
p
o
t
h
e
sis
m
a
rg
in
,”
J
o
u
rn
a
l
o
f
Co
mp
u
ter
s
,
v
o
l
.
3
,
n
o
.
12
,
p
p
.
27
-
3
4
,
2
0
0
8
.
[1
6
]
W.
M
e
g
c
h
e
len
b
ri
k
,
“
Re
li
e
f
-
Ba
se
d
fe
a
tu
re
se
lec
ti
o
n
in
b
io
in
f
o
rm
a
ti
c
s:
d
e
tec
ti
n
g
f
u
n
c
ti
o
n
a
l
sp
e
c
ifi
c
it
y
re
sid
u
e
s
fro
m
m
u
lt
ip
le se
q
u
e
n
c
e
a
li
g
n
m
e
n
ts
,”
R
a
d
b
o
u
d
U
n
iv
e
rsit
y
,
Nijme
g
e
n
,
2
0
1
0
.
[1
7
]
M
.
Da
sh
a
n
d
H.
L
iu
,
“
F
e
a
tu
re
se
lec
ti
o
n
f
o
r
c
las
sifica
ti
o
n
,”
I
n
telli
g
e
n
t
d
a
ta
a
n
a
lys
is
,
v
o
l
.
1
,
n
o
.
3
,
p
p
.
1
3
1
-
1
5
6
,
1
9
9
7
.
[1
8
]
Y.
S
u
n
,
S
.
T
o
d
o
ro
v
ic,
a
n
d
S
.
G
o
o
d
iso
n
,
“
Lo
c
a
l
-
lea
rn
i
n
g
-
b
a
se
d
fe
a
tu
re
se
lec
ti
o
n
f
o
r
h
ig
h
-
d
ime
n
si
o
n
a
l
d
a
ta
a
n
a
ly
sis
,”
IEE
E
T
ra
n
sa
c
ti
o
n
s
o
n
Pa
t
ter
n
An
a
lys
is a
n
d
M
a
c
h
in
e
In
telli
g
e
n
c
e
,
v
o
l.
32
,
n
o
.
9
,
p
p
.
1
6
1
0
-
1
6
2
6
,
2
0
1
0
.
[1
9
]
D.
De
rn
o
n
c
o
u
rt,
B.
Ha
n
c
z
a
r,
a
n
d
J.
D.
Zu
c
k
e
r
,
“
An
a
ly
sis
o
f
fe
a
tu
r
e
se
lec
ti
o
n
sta
b
il
it
y
o
n
h
i
g
h
d
ime
n
sio
n
a
n
d
sm
a
ll
sa
m
p
le d
a
ta
,”
Co
mp
u
ta
ti
o
n
a
l
S
ta
t
isti
c
s &
Da
ta
A
n
a
lys
is
,
v
o
l.
71
,
p
p
.
6
8
1
-
6
9
3
,
2
0
1
4
.
[2
0
]
G
.
Ro
ffo
,
C.
S
e
g
a
li
n
,
A.
Vi
n
c
iare
ll
i,
V.
M
u
rin
o
,
a
n
d
M
.
Cristan
i
,
“
Re
a
d
in
g
b
e
twe
e
n
th
e
t
u
rn
s:
sta
ti
stic
a
l
m
o
d
e
li
n
g
fo
r
id
e
n
ti
t
y
re
c
o
g
n
it
io
n
a
n
d
v
e
rifi
c
a
ti
o
n
i
n
c
h
a
ts
,”
Pro
c
e
e
d
in
g
s
o
f
th
e
1
0
t
h
IEE
E
In
ter
n
a
ti
o
n
a
l
C
o
n
fer
e
n
c
e
o
n
Ad
v
a
n
c
e
d
Vi
d
e
o
a
n
d
S
ig
n
a
l
Ba
se
d
S
u
rv
e
il
la
n
c
e
,
p
p.
99
-
1
0
4
,
2
0
1
3
.
[2
1
]
G
.
Ro
ffo
,
C.
G
io
rg
e
tt
a
,
R
.
F
e
rra
r
io
,
W
.
Riv
iera
,
a
n
d
M
.
Crista
n
i
,
“
S
tatisti
c
a
l
a
n
a
ly
sis
o
f
p
e
rso
n
a
li
t
y
a
n
d
id
e
n
ti
t
y
in
c
h
a
ts
u
sin
g
a
k
e
y
l
o
g
g
in
g
p
latf
o
r
m
,”
Pro
c
e
e
d
in
g
s
o
f
th
e
1
6
t
h
In
te
rn
a
ti
o
n
a
l
Co
n
fer
e
n
c
e
o
n
M
u
l
ti
m
o
d
a
l
In
ter
a
c
ti
o
n
,
p
p.
2
2
4
-
231
,
2
0
1
4
.
[2
2
]
G
.
Ro
ffo
,
M
.
Cristan
i
,
L
.
Ba
z
z
a
n
i
,
H.
Q.
M
in
h
,
a
n
d
V.
M
u
r
in
o
,
“
Tru
stin
g
s
k
y
p
e
:
Lea
rn
i
n
g
t
h
e
Way
P
e
o
p
le
C
h
a
t
f
o
r
F
a
st
Us
e
r
Re
c
o
g
n
it
i
o
n
a
n
d
Ve
rifi
c
a
ti
o
n
,
”
Pr
o
c
e
e
d
in
g
s
o
f
th
e
IEE
E
In
ter
n
a
ti
o
n
a
l
Co
n
fer
e
n
c
e
o
n
C
o
mp
u
ter
V
isio
n
W
o
rk
sh
o
p
s
,
p
p.
7
4
8
-
7
5
4
,
2
0
1
3
.
[2
3
]
J.
Yu
a
n
a
n
d
F
.
B.
Ba
sta
n
i
,
“
Ro
b
u
s
t
o
b
jec
t
trac
k
in
g
v
ia
o
n
li
n
e
i
n
fo
rm
a
ti
v
e
fe
a
tu
re
se
lec
ti
o
n
,”
Pr
o
c
e
e
d
in
g
s
o
f
t
h
e
IEE
E
In
ter
n
a
t
io
n
a
l
C
o
n
fer
e
n
c
e
o
n
Im
a
g
e
Pro
c
e
ss
in
g
,
p
p.
4
7
1
-
4
7
5
,
2
0
1
4
.
[2
4
]
K.
Zh
a
n
g
,
L.
Zh
a
n
g
,
a
n
d
M
.
H.
Ya
n
g
,
“
Re
a
l
-
Ti
m
e
Ob
jec
t
Trac
k
i
n
g
Via
On
li
n
e
Disc
rimin
a
ti
v
e
F
e
a
tu
re
S
e
lec
ti
o
n
,”
IEE
E
T
ra
n
sa
c
ti
o
n
s
o
n
Ima
g
e
Pro
c
e
ss
in
g
,
v
o
l.
22
,
n
o
.
12
,
p
p
.
4
6
6
4
-
4
6
7
7
,
2
0
1
3
.
[2
5
]
O
.
De
n
iz,
G
.
B
u
e
n
o
,
J
.
S
a
li
d
o
,
a
n
d
F
.
De
la T
o
rre
,
“
F
a
c
e
re
c
o
g
n
it
i
o
n
u
sin
g
h
isto
g
ra
m
s o
f
o
rie
n
ted
g
r
a
d
ien
ts
,”
P
a
tt
e
rn
Rec
o
g
n
it
io
n
L
e
tt
e
rs
,
v
o
l.
32
,
n
o
.
1
2
,
p
p
.
1
5
9
8
-
1
6
0
3
,
2
0
1
1
.
[2
6
]
D
.
P
.
V
.
Ho
a
i,
T
.
S
u
rin
wa
ra
n
g
k
o
o
n
,
V
.
T
.
Ho
a
n
g
,
H
.
T
.
Du
o
n
g
,
a
n
d
K
.
M
e
e
th
o
n
g
jan
,
“
A
c
o
m
p
a
ra
ti
v
e
stu
d
y
o
f
rice
v
a
riety
c
las
sifica
ti
o
n
b
a
se
d
o
n
d
e
e
p
lea
rn
in
g
a
n
d
h
a
n
d
-
c
ra
fted
fe
a
tu
re
s
,”
ECT
I
T
ra
n
sa
c
ti
o
n
s
o
n
Co
mp
u
ter
a
n
d
In
fo
rm
a
t
io
n
T
e
c
h
n
o
l
o
g
y
(EC
T
I
-
C
IT
)
,
v
o
l.
14
,
n
o
.
1
,
p
p
.
1
-
1
0
,
2
0
2
0
.
[2
7
]
H.
T.
M
.
Nh
a
t
a
n
d
V.
T.
Ho
a
n
g
,
“
F
e
a
tu
re
fu
sio
n
b
y
u
sin
g
LBP
,
H
OG
,
G
IS
T
d
e
sc
rip
to
rs
a
n
d
Ca
n
o
n
ica
l
Co
rre
latio
n
An
a
ly
sis
fo
r
fa
c
e
re
c
o
g
n
it
i
o
n
,
”
2
6
th
In
ter
n
a
ti
o
n
a
l
C
o
n
fer
e
n
c
e
o
n
T
e
lec
o
mm
u
n
ica
ti
o
n
s (ICT
),
p
p.
3
7
1
-
3
7
5
,
2
0
1
9
.
[2
8
]
T.
N.
Va
n
a
n
d
V.
T.
H
o
a
n
g
,
“
Ki
n
sh
ip
Ve
rifi
c
a
ti
o
n
b
a
se
d
o
n
L
o
c
a
l
Bin
a
ry
P
a
tt
e
rn
fe
a
tu
re
s
c
o
d
in
g
i
n
d
iffere
n
t
c
o
l
o
r
sp
a
c
e
,
”
2
6
th
In
ter
n
a
ti
o
n
a
l
Co
n
fer
e
n
c
e
o
n
T
e
lec
o
mm
u
n
ica
t
io
n
s (IC
T
),
p
p.
3
7
6
-
3
8
0
,
2
0
1
9
.
[2
9
]
K.
M
e
e
th
o
n
g
jan
,
M
.
Dz
u
l
k
ifl
i
,
P
.
K.
Re
e
,
a
n
d
M
.
Y
.
Na
m
,
“
F
u
si
o
n
a
ffin
e
m
o
m
e
n
t
in
v
a
ri
a
n
ts
a
n
d
wa
v
e
let
p
a
c
k
e
t
fe
a
tu
re
s
se
lec
ti
o
n
fo
r
fa
c
e
v
e
rifi
c
a
ti
o
n
,”
J
o
u
rn
a
l
o
f
T
h
e
o
re
ti
c
a
l
a
n
d
Ap
p
li
e
d
I
n
fo
rm
a
ti
o
n
T
e
c
h
n
o
l
o
g
y
,
v
o
l.
64
,
n
o
.
3
,
p
p
.
6
0
6
-
6
1
5
,
2
0
1
4
.
[3
0
]
T.
S
u
r
in
wa
ra
n
g
k
o
o
n
,
S
.
Nitsu
wa
t,
a
n
d
J.
El
v
in
,
“
A
traffic
sig
n
d
e
tec
ti
o
n
a
n
d
re
c
o
g
n
it
io
n
sy
ste
m
,”
In
ter
n
a
ti
o
n
a
l
J
o
u
rn
a
l
o
f
Circ
u
it
s,
S
y
ste
ms
a
n
d
S
ig
n
a
l
Pro
c
e
ss
in
g
,
v
o
l.
7
,
n
o
.
1
,
p
p
.
58
-
6
5
,
2
0
1
3
.
[3
1
]
N.
Da
lal
a
n
d
B.
Tri
g
g
s
,
“
Histo
g
ra
m
s
o
f
o
rien
ted
g
ra
d
ien
ts
fo
r
h
u
m
a
n
d
e
tec
ti
o
n
,
“
2
0
0
5
IEE
E
Co
m
p
u
ter
S
o
c
iety
Co
n
fe
re
n
c
e
o
n
Co
m
p
u
ter Visi
o
n
a
n
d
P
a
tt
e
rn
Re
c
o
g
n
it
i
o
n
(CVPR’
0
5
),
”
vol
.
1
,
p
p.
8
8
6
-
8
9
3
,
2
0
0
5
.
[3
2
]
K.
Be
n
a
b
d
e
sle
m
a
n
d
M
.
Hin
d
a
wi
,
“
Co
n
stra
in
e
d
Lap
lac
ian
sc
o
re
fo
r
se
m
i
-
su
p
e
rv
ise
d
fe
a
tu
re
se
lec
ti
o
n
,”
M
a
c
h
in
e
L
e
a
rn
in
g
a
n
d
Kn
o
wled
g
e
Disc
o
v
e
ry
in
Da
t
a
b
a
se
s
, p
p.
2
0
4
-
2
1
8
,
2
0
1
1
.
[3
3
]
I.
G
u
y
o
n
a
n
d
A.
El
isse
e
ff
,
“
An
in
tro
d
u
c
ti
o
n
t
o
v
a
riab
le
a
n
d
fe
a
tu
re
se
lec
ti
o
n
,”
T
h
e
J
o
u
rn
a
l
o
f
M
a
c
h
in
e
L
e
a
rn
i
n
g
Res
e
a
rc
h
,
v
o
l.
3
,
n
o
.
1
1
5
7
-
1
1
8
2
,
2
0
0
3
.
[3
4
]
J
.
Ca
i,
J
.
L
u
o
,
S
.
Wan
g
,
a
n
d
S
.
Y
a
n
g
,
“
F
e
a
tu
re
se
lec
ti
o
n
i
n
m
a
c
h
i
n
e
lea
rn
in
g
:
A
n
e
w
p
e
rsp
e
c
ti
v
e
,”
Ne
u
ro
c
o
mp
u
ti
n
g
,
v
o
l.
3
0
0
,
p
p
.
7
0
-
9
,
2
0
1
8
.
[3
5
]
M
.
L
iu
a
n
d
D.
Zh
a
n
g
,
“
S
p
a
rsit
y
s
c
o
re
:
a
n
o
v
e
l
g
ra
p
h
-
p
re
se
rv
in
g
fe
a
tu
re
se
lec
ti
o
n
m
e
th
o
d
,”
I
n
ter
n
a
ti
o
n
a
l
J
o
u
rn
a
l
o
f
Pa
tt
e
rn
Rec
o
g
n
it
io
n
a
n
d
Arti
fi
c
i
a
l
In
telli
g
e
n
c
e
,
v
o
l.
28
,
n
o
.
0
4
,
2
0
1
4
.
[3
6
]
C
.
Zu
,
L
.
Zh
u
,
a
n
d
D
.
Zh
a
n
g
,
“
It
e
ra
ti
v
e
sp
a
rsity
sc
o
re
fo
r
fe
a
tu
re
se
lec
ti
o
n
a
n
d
it
s
e
x
te
n
sio
n
fo
r
m
u
lt
imo
d
a
l
d
a
ta
,”
Ne
u
ro
c
o
mp
u
ti
n
g
,
v
o
l
.
2
5
9
,
p
p
.
1
4
6
,
5
3
,
2
0
1
7
.
[3
7
]
S
.
Ya
n
g
,
J
.
Zh
a
n
g
,
C
.
B
o
,
M
.
Wa
n
g
,
a
n
d
L
.
Ch
e
n
,
“
F
a
st
v
e
h
icle
lo
g
o
d
e
tec
ti
o
n
i
n
c
o
m
p
le
x
sc
e
n
e
s
,”
Op
ti
c
s
&
L
a
se
r
T
e
c
h
n
o
l
o
g
y
,
v
o
l.
1
1
0
,
p
p
.
1
9
6
-
2
0
1
,
2
0
1
8
.
[3
8
]
R
.
Li
u
,
Q
.
Ha
n
,
W
.
M
i
n
,
L
.
Zh
o
u
,
a
n
d
J
.
Xu
,
“
Ve
h
icle
Lo
g
o
Re
c
o
g
n
it
io
n
Ba
se
d
o
n
En
h
a
n
c
e
d
M
a
t
c
h
in
g
fo
r
S
m
a
ll
Ob
jec
ts,
Co
n
stra
i
n
e
d
Re
g
i
o
n
a
n
d
S
S
F
P
D Ne
two
r
k
,”
S
e
n
so
rs
,
v
o
l
.
19
,
n
o
.
20
,
p
p
.
1
-
1
8
,
2
0
1
9
.
[3
9
]
F
.
Tafa
z
z
o
li
,
H
.
F
ri
g
u
i,
a
n
d
K
.
Nish
iy
a
m
a
,
“
A
Larg
e
a
n
d
Di
v
e
rse
Da
tas
e
t
fo
r
Im
p
ro
v
e
d
Ve
h
icle
M
a
k
e
a
n
d
M
o
d
e
l
Re
c
o
g
n
it
i
o
n
,”
2
0
1
7
IE
EE
Co
n
f
e
re
n
c
e
o
n
Co
mp
u
ter
Vi
si
o
n
a
n
d
P
a
tt
e
rn
Rec
o
g
n
it
i
o
n
W
o
rk
s
h
o
p
s
(CV
PR
W
),
pp.
8
7
4
-
8
8
1
,
2
0
1
7
.
[4
0
]
Y
.
Yu
,
J
.
Wan
g
,
J
.
L
u
,
Y
.
Xie
,
a
n
d
Z
.
Nie
,
“
Ve
h
icle
l
o
g
o
re
c
o
g
n
it
io
n
b
a
se
d
o
n
o
v
e
rlap
p
i
n
g
e
n
h
a
n
c
e
d
p
a
tt
e
rn
s
o
f
o
rien
ted
e
d
g
e
m
a
g
n
it
u
d
e
s
,”
Co
mp
u
ter
s &
El
e
c
trica
l
En
g
in
e
e
rin
g
,
v
o
l.
71
,
p
p
.
2
7
3
-
2
8
3
,
2
0
1
8
.
[4
1
]
Y
.
Xia
,
J
.
F
e
n
g
,
a
n
d
B
.
Zh
a
n
g
,
“
Ve
h
icle
Lo
g
o
Re
c
o
g
n
i
ti
o
n
a
n
d
a
tt
ri
b
u
tes
p
re
d
icti
o
n
b
y
m
u
l
ti
-
tas
k
lea
rn
in
g
with
CNN
,”
2
0
1
6
1
2
th
In
ter
n
a
ti
o
n
a
l
C
o
n
fer
e
n
c
e
o
n
Na
tu
ra
l
C
o
mp
u
t
a
ti
o
n
,
F
u
zz
y
S
y
ste
ms
a
n
d
K
n
o
wle
d
g
e
Di
sc
o
v
e
ry
(ICNC
-
FS
KD),
p
p.
6
6
8
-
6
7
2
,
2
0
1
6
.
[4
2
]
Xia
Y,
F
e
n
g
J,
Z
h
a
n
g
B.
Ve
h
icle
lo
g
o
re
c
o
g
n
it
io
n
a
n
d
a
tt
ri
b
u
te
s
p
re
d
icti
o
n
b
y
m
u
lt
i
-
tas
k
lea
rn
i
n
g
wit
h
CNN
.
P
ro
c
e
e
d
in
g
s
o
f
1
2
t
h
IE
EE
i
n
tern
a
ti
o
n
a
l
c
o
n
fe
re
n
c
e
o
n
n
a
t
u
ra
l
c
o
m
p
u
tati
o
n
,
fu
z
z
y
s
y
ste
m
s
a
n
d
k
n
o
wle
d
g
e
d
isc
o
v
e
ry
(ICNC
-
F
S
KD
).
2
0
1
6
.
p
.
6
6
8
–
72.
[4
3
]
V
.
T
.
Ho
a
n
g
,
“
HG
M
-
4
:
A
n
e
w m
u
lt
i
-
c
a
m
e
ra
s d
a
tas
e
t
fo
r
h
a
n
d
g
e
st
u
re
re
c
o
g
n
it
io
n
,
”
v
o
l.
3
0
,
2
0
2
0
.
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KOM
NI
KA
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l Co
n
tr
o
l
V
eh
icle
lo
g
o
r
ec
o
g
n
itio
n
u
s
in
g
h
is
to
g
r
a
ms o
f o
r
ien
ted
g
r
a
d
ien
t d
escr
ip
to
r
a
n
d
… (
K
itti
kh
u
n
Meeth
o
n
g
ja
n
)
3025
B
I
O
G
RAP
H
I
E
S O
F
AUTH
O
RS
K
itti
k
h
u
n
Mee
th
o
n
g
j
a
n
is
a
f
u
ll
lec
tu
re
r
in
th
e
C
o
m
p
u
ter
S
c
ie
n
c
e
P
ro
g
ra
m
,
He
a
d
o
f
A
p
p
l
y
S
c
ien
c
e
De
p
a
rtme
n
t,
F
a
c
u
lt
y
o
f
S
c
ien
c
e
a
n
d
Tec
h
n
o
lo
g
y
,
S
u
a
n
S
u
n
a
n
d
h
a
Ra
jab
h
a
t
Un
iv
e
rsity
(S
S
RU),
Th
a
il
a
n
d
.
H
e
re
c
e
iv
e
d
h
is
B.
S
c
.
De
g
re
e
in
Co
m
p
u
ter
S
c
ien
c
e
fro
m
S
S
RU
i
n
1
9
9
0
,
a
m
a
ste
r’s
d
e
g
re
e
in
c
o
m
p
u
ter
sc
ien
c
e
fr
o
m
Ki
n
g
M
o
n
g
k
u
t'
s
Un
i
v
e
rsity
o
f
Tec
h
n
o
l
o
g
y
Th
o
n
b
u
ri
(KMUT
T)
in
2
0
0
0
a
n
d
a
P
h
D
d
e
g
re
e
i
n
Co
m
p
u
ter
G
ra
p
h
ic
fro
m
th
e
Un
iv
e
rsity
o
f
Tec
h
n
o
lo
g
y
M
a
lay
sia
in
2
0
1
3
.
His
re
se
a
rc
h
in
tere
st
is
C
o
m
p
u
ter
G
ra
p
h
ic,
Im
a
g
e
P
ro
c
e
ss
in
g
,
Art
ifi
c
ial
In
tel
li
g
e
n
t,
Bio
m
e
tri
c
s,
P
a
tt
e
r
n
Re
c
o
g
n
it
io
n
,
S
o
ft
Co
m
p
u
ti
n
g
tec
h
n
iq
u
e
s
a
n
d
a
p
p
li
c
a
ti
o
n
.
C
u
rr
e
n
tl
y
,
h
e
is
a
n
a
d
v
isi
n
g
o
f
M
a
s
ter
a
n
d
P
h
D
stu
d
e
n
t
i
n
F
o
re
n
sic
S
c
ien
c
e
o
f
S
S
RU.
Th
o
n
g
c
h
a
i
S
u
r
in
wa
r
a
n
g
k
o
o
n
w
a
s
b
o
rn
in
S
u
ra
tt
h
a
n
i,
Th
a
il
a
n
d
i
n
1
9
7
2
.
He
o
b
tai
n
e
d
B.
S
c
.
d
e
g
re
e
in
m
a
th
e
m
a
ti
c
s
fro
m
Ch
ian
g
M
a
i
Un
i
v
e
rsity
,
Th
a
il
a
n
d
i
n
1
9
9
5
.
He
re
c
e
iv
e
d
M
.
S
c
.
d
e
g
re
e
in
m
a
n
a
g
e
m
e
n
t
o
f
in
f
o
rm
a
ti
o
n
tec
h
n
o
l
o
g
y
fro
m
Wala
il
a
k
Un
iv
e
rsity
,
Th
a
il
a
n
d
i
n
2
0
0
5
a
n
d
P
h
.
D
.
in
i
n
fo
rm
a
ti
o
n
te
c
h
n
o
l
o
g
y
fro
m
Kin
g
M
o
n
g
k
u
t'
s
Un
iv
e
rsity
o
f
Tec
h
n
o
lo
g
y
No
rth
Ba
n
g
k
o
k
,
Th
a
il
a
n
d
i
n
2
0
1
4
.
He
is
n
o
w
a
Lec
tu
re
r
in
De
p
a
rtme
n
t
o
f
Bu
sin
e
ss
Co
m
p
u
ter,
S
u
a
n
S
u
n
a
n
d
h
a
Ra
jab
h
a
t
Un
i
v
e
rsity
,
Ba
n
g
k
o
k
,
T
h
a
il
a
n
d
.
His
re
se
a
rc
h
in
tere
sts
in
c
lu
d
e
d
ig
i
tal
ima
g
e
p
ro
c
e
ss
in
g
,
m
a
c
h
in
e
lea
rn
i
n
g
,
a
rti
ficia
l
in
telli
g
e
n
c
e
,
b
u
si
n
e
ss
in
telli
g
e
n
c
e
,
a
n
d
m
o
b
i
le ap
p
l
ica
ti
o
n
d
e
v
e
lo
p
m
e
n
t
f
o
r
b
u
sin
e
ss
.
Vin
h
Tr
u
o
n
g
H
o
a
n
g
re
c
e
iv
e
d
h
is
m
a
ste
r’s
d
e
g
re
e
fro
m
th
e
Un
iv
e
rsity
o
f
M
o
n
tp
e
ll
ier
i
n
2
0
0
9
a
n
d
h
is
P
h
D
d
e
g
re
e
in
c
o
m
p
u
ter
sc
ien
c
e
fro
m
th
e
Un
i
v
e
rsity
o
f
th
e
Li
tt
o
ra
l
Op
a
l
C
o
a
st,
F
ra
n
c
e
.
He
is
c
u
rre
n
tl
y
a
n
a
ss
ista
n
t
p
ro
fe
ss
o
r
a
n
d
He
a
d
o
f
Im
a
g
e
P
ro
c
e
ss
in
g
a
n
d
C
o
m
p
u
te
r
G
ra
p
h
ics
De
p
a
rtme
n
t
a
t
th
e
Ho
Ch
i
M
i
n
h
Cit
y
Op
e
n
Un
i
v
e
rsity
,
Vie
tn
a
m
.
His
re
se
a
rc
h
in
tere
sts
in
c
lu
d
e
ima
g
e
a
n
a
ly
sis a
n
d
fe
a
tu
re
se
lec
ti
o
n
.
Evaluation Warning : The document was created with Spire.PDF for Python.