I
nd
o
ne
s
ia
n J
o
urna
l o
f
E
lect
rica
l En
g
ineering
a
nd
Co
m
pu
t
er
Science
Vo
l.
3
7
,
No
.
2
,
Feb
r
u
ar
y
20
2
5
,
p
p
.
995
~
1
0
0
4
I
SS
N:
2
502
-
4
7
52
,
DOI
: 1
0
.
1
1
5
9
1
/ijee
cs
.v
3
7
.
i
2
.
pp
995
-
1
0
0
4
995
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
ee
cs
.
ia
esco
r
e.
co
m
Co
mpa
ra
tive
stu
dy
of potho
le d
ete
ction using
de
ep l
ea
rning
on
sm
a
rtphone
Achy
a
r
Ulul Am
ri
,
G
ede
P
utr
a
K
us
um
a
D
e
p
a
r
t
me
n
t
o
f
C
o
mp
u
t
e
r
S
c
i
e
n
c
e
,
B
I
N
U
S
G
r
a
d
u
a
t
e
P
r
o
g
r
a
m
-
M
a
s
t
e
r
o
f
C
o
mp
u
t
e
r
S
c
i
e
n
c
e
,
B
i
n
a
N
u
sa
n
t
a
r
a
U
n
i
v
e
r
si
t
y
,
Jak
a
r
t
a
,
I
n
d
o
n
e
si
a
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
May
15
,
2
0
2
4
R
ev
is
ed
Sep
16
,
2
0
2
4
Acc
ep
ted
Sep
29
,
2
0
2
4
P
o
th
o
les
p
re
se
n
t
a
sig
n
ifi
c
a
n
t
p
r
o
b
lem
in
m
a
n
y
c
o
u
n
t
ries
,
lea
d
i
n
g
t
o
v
e
h
icle
d
a
m
a
g
e
a
n
d
traffic
a
c
c
id
e
n
ts.
Th
e
se
ro
a
d
imp
e
rfe
c
ti
o
n
s
p
o
se
sa
fe
ty
risk
s
a
n
d
imp
o
se
e
c
o
n
o
m
ic
b
u
rd
e
n
s
.
De
sp
i
te
e
x
isti
n
g
d
e
tec
ti
o
n
m
e
th
o
d
s
u
si
n
g
se
n
so
rs
a
n
d
c
o
m
p
u
ter
v
isi
o
n
d
e
e
p
lea
rn
in
g
p
ro
c
e
ss
e
d
o
n
P
Cs,
a
g
a
p
r
e
m
a
in
s
in
d
e
p
lo
y
i
n
g
c
o
st
-
e
ffe
c
ti
v
e
,
wi
d
e
ly
a
c
c
e
ss
ib
le
so
lu
ti
o
n
s.
Th
is
stu
d
y
a
ims
t
o
b
rid
g
e
t
h
is
g
a
p
b
y
d
e
v
e
lo
p
in
g
d
e
e
p
lea
rn
i
n
g
m
o
d
e
ls
o
p
ti
m
ize
d
fo
r
sm
a
rtp
h
o
n
e
s,
re
d
u
c
in
g
c
o
sts
a
n
d
e
n
h
a
n
c
i
n
g
d
e
p
lo
y
m
e
n
t
f
e
a
sib
il
it
y
.
We
d
e
v
e
lo
p
e
d
m
u
lt
i
p
le
m
o
d
e
ls
fo
r
p
o
th
o
le
d
e
tec
ti
o
n
,
u
ti
li
z
i
n
g
tran
sfe
r
lea
rn
in
g
a
n
d
Ba
y
e
sia
n
h
y
p
e
rp
a
ra
m
e
ter
tu
n
i
n
g
t
o
o
p
ti
m
ize
d
e
tec
ti
o
n
a
c
c
u
ra
c
y
a
n
d
re
so
u
rc
e
e
fficie
n
c
y
.
Ou
r
e
v
a
lu
a
ti
o
n
s
f
o
c
u
se
d
o
n
c
o
m
p
u
tati
o
n
a
ll
y
li
g
h
t
m
o
d
e
ls
su
c
h
a
s
YO
LOv
8
s
m
a
ll
,
YO
LOv
8
-
n
a
n
o
,
YO
L
Ov
7
t
in
y
,
a
n
d
f
a
ste
r
R
-
CNN
M
o
b
il
e
Ne
tV3
.
In
term
s
o
f
d
e
tec
ti
o
n
a
c
c
u
ra
c
y
,
YO
LOv
8
s
m
a
ll
a
n
d
YO
LOv
8
n
a
n
o
sto
o
d
o
u
t
,
a
c
h
iev
in
g
a
v
e
ra
g
e
p
re
c
isi
o
n
s
(AP)
o
f
8
3
.
5
%
a
n
d
8
2
.
5
%
,
re
sp
e
c
ti
v
e
l
y
.
YO
LOv
8
n
a
n
o
p
ro
v
e
d
th
e
m
o
st
e
fficie
n
t,
o
ffe
rin
g
h
i
g
h
d
e
tec
ti
o
n
a
c
c
u
ra
c
y
,
a
fil
e
siz
e
t
h
re
e
ti
m
e
s
sm
a
ll
e
r
th
a
n
YO
LOv
8
s
m
a
ll
in
TF
Li
te
f
o
rm
a
t,
a
n
d
th
e
fa
ste
st
i
n
fe
re
n
c
e
ti
m
e
o
f
0
.
7
2
se
c
o
n
d
s
p
e
r
ima
g
e
.
Th
is
stu
d
y
h
i
g
h
li
g
h
ts
th
e
p
o
t
e
n
ti
a
l
o
f
sm
a
rtp
h
o
n
e
s
in
u
r
b
a
n
p
o
t
h
o
le
d
e
tec
ti
o
n
,
c
o
n
tr
ib
u
ti
n
g
t
o
imp
r
o
v
e
d
ro
a
d
m
a
in
ten
a
n
c
e
a
n
d
u
r
b
a
n
p
o
li
c
y
.
K
ey
w
o
r
d
s
:
B
ay
esian
s
ea
r
ch
Dee
p
lear
n
in
g
Hy
p
er
p
ar
a
m
eter
tu
n
in
g
Po
th
o
le
d
etec
tio
n
Sm
ar
tp
h
o
n
e
r
eso
u
r
ce
u
s
ag
e
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
:
Ach
y
ar
Ulu
l A
m
r
i
Dep
ar
tm
en
t o
f
C
o
m
p
u
ter
Scie
n
ce
,
B
I
NUS
Gr
ad
u
ate
Pro
g
r
a
m
-
Ma
s
ter
o
f
C
o
m
p
u
te
r
Scien
c
e
B
in
a
Nu
s
an
tar
a
Un
iv
er
s
ity
J
ak
ar
ta,
I
n
d
o
n
esia
E
m
ail: a
ch
y
ar
.
am
r
i@
b
in
u
s
.
ac
.
id
1.
I
NT
RO
D
UCT
I
O
N
R
o
ad
in
f
r
astru
ctu
r
e
co
n
d
itio
n
s
ar
e
a
k
e
y
s
af
ety
f
ac
t
o
r
th
at
n
ee
d
s
to
b
e
s
p
ec
ially
d
esig
n
e
d
to
r
e
d
u
ce
r
is
k
s
f
o
r
all
r
o
ad
u
s
er
s
.
Ho
we
v
er
,
m
an
y
r
o
ad
s
in
lo
w
-
to
m
id
d
le
-
in
co
m
e
co
u
n
tr
ies
h
av
e
n
o
t
y
et
m
et
th
ese
[
1
]
.
T
h
e
d
eter
io
r
atio
n
o
f
u
r
b
an
in
f
r
astru
ctu
r
e,
i
n
clu
d
i
n
g
r
o
a
d
s
,
ca
n
h
av
e
s
ig
n
if
ican
t
s
o
cial
im
p
licatio
n
s
,
af
f
ec
tin
g
th
e
q
u
ality
o
f
life
,
p
u
b
lic
s
af
ety
,
an
d
ac
ce
s
s
to
s
er
v
ices
[
2
]
.
R
o
ad
p
o
th
o
les
ar
e
o
n
e
o
f
m
o
s
t
is
s
u
e
th
at
p
r
esen
t
a
s
ig
n
if
ican
t
p
r
o
b
lem
in
m
an
y
co
u
n
tr
ies,
lead
in
g
to
v
eh
icle
d
am
ag
e
an
d
tr
af
f
ic
ac
cid
e
n
ts
[
3
]
.
T
h
e
p
r
esen
ce
o
f
p
o
th
o
les
ca
n
s
o
m
etim
es
b
e
d
if
f
icu
lt
to
id
en
tify
m
an
u
ally
,
esp
ec
ially
wh
ile
d
r
iv
in
g
[
4
]
.
T
h
e
ex
is
tan
ce
o
f
f
ast
p
o
th
o
le
d
etec
tio
n
m
et
h
o
d
s
is
an
ticip
ated
to
aid
d
r
i
v
er
s
in
av
o
id
in
g
p
o
t
h
o
le
r
o
a
d
an
d
ca
n
p
r
o
v
id
e
v
alu
ab
l
e
d
ata
to
th
e
g
o
v
e
r
n
m
en
t
in
p
r
io
r
itizin
g
r
o
ad
r
ep
air
s
,
u
ltima
tely
h
elp
i
n
g
t
o
r
e
d
u
c
e
th
e
lik
elih
o
o
d
o
f
ac
cid
en
ts
[
5
]
.
T
h
er
e
ar
e
m
an
y
m
eth
o
d
s
h
a
v
e
b
ee
n
d
e
v
elo
p
ed
to
d
etec
t
p
o
th
o
les,
in
clu
d
i
n
g
m
a
n
u
al
in
s
p
ec
tio
n
,
s
en
s
o
r
u
tili
za
tio
n
an
d
c
o
m
p
u
ter
v
is
io
n
[
6
]
,
[
7
]
.
E
ar
ly
a
d
o
p
t
io
n
o
f
s
en
s
o
r
s
s
u
ch
as
ac
ce
ler
o
m
eter
s
o
r
GPS
in
v
eh
icles
to
d
etec
t
p
o
th
o
les
b
ased
o
n
v
ib
r
ati
o
n
s
o
r
ir
r
e
g
u
l
ar
ities
.
L
iDAR
i
s
a
m
o
r
e
ad
v
an
ce
d
an
d
ac
cu
r
ate
s
en
s
o
r
tech
n
o
lo
g
y
th
at
u
s
es
laser
lig
h
t
to
cr
ea
te
h
ig
h
-
r
eso
lu
t
io
n
3
D
m
ap
s
o
f
th
e
r
o
ad
s
u
r
f
a
ce
alb
eit
at
a
h
ig
h
e
r
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
5
0
2
-
4
7
52
In
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
,
Vo
l.
3
7
,
No
.
2
,
Feb
r
u
a
r
y
20
2
5
:
9
9
5
-
1
0
0
4
996
co
s
t
[
8
]
.
C
o
m
p
u
ter
v
is
io
n
d
ee
p
lear
n
in
g
-
b
ased
c
o
m
p
u
ter
v
is
io
n
b
ec
am
e
m
o
s
t
p
o
p
u
lar
m
et
h
o
d
c
o
n
s
id
er
in
g
th
e
ac
cu
r
ac
y
o
f
d
etec
tio
n
o
f
f
e
r
ed
an
d
co
s
t e
f
f
icien
c
y
[
9
]
.
T
h
er
e
ar
e
s
ev
er
al
p
o
th
o
le
d
ete
ctio
n
r
esear
ch
s
tu
d
ies
th
at
u
til
ize
d
ee
p
lear
n
in
g
u
s
in
g
alg
o
r
ith
m
s
f
r
o
m
th
e
R
-
C
NN
f
am
ily
s
u
ch
as
R
-
FC
N,
s
p
ar
s
e
R
-
C
NN,
f
aster
R
-
C
NN,
an
d
t
h
e
YOL
O
f
am
ily
with
v
ar
io
u
s
co
n
f
ig
u
r
atio
n
v
a
r
iatio
n
s
.
Mo
s
t
o
f
th
e
p
o
t
h
o
le
d
etec
tio
n
m
o
d
el
s
d
ev
elo
p
ed
ar
e
s
till
o
n
PC
s
,
with
v
ar
iatio
n
s
in
th
e
d
atasets
u
s
ed
,
in
clu
d
i
n
g
ad
d
in
g
p
er
s
o
n
al
d
ata
[
1
0
]
,
[
1
1
]
,
n
o
is
e
tr
ea
tm
en
t,
b
r
i
g
h
tn
ess
v
ar
iaety
[
1
2
]
,
an
d
co
m
b
in
in
g
r
eso
u
r
ce
s
f
r
o
m
m
u
ltip
le
im
ag
e
d
atasets
[
1
3
]
.
I
n
p
r
ev
io
u
s
r
esear
ch
h
as
b
ee
n
d
ev
elo
p
e
d
p
o
th
o
le
d
etec
tio
n
m
o
d
el
u
s
in
g
f
in
e
tu
n
in
g
d
ee
p
lear
n
in
g
f
aster
R
-
C
NN
R
e
s
N
et1
0
1
,
wh
ich
ac
ie
v
ed
ac
cu
r
ac
y
AP
o
f
8
2
%
an
d
in
f
er
e
n
ce
s
p
ee
d
5
-
6
FP
S
(
f
r
am
e
p
er
s
ec
o
n
d
)
wh
e
n
im
p
lem
en
ted
o
n
a
co
m
p
u
te
r
Nv
id
ia
Dr
iv
ePX2
[
1
4
]
.
B
esid
es,
a
p
o
t
h
o
le
d
ete
ctio
n
m
o
d
el,
S
PFPPN
-
YOL
O
v
4
h
as
b
ee
n
d
ev
el
o
p
ed
an
d
a
ch
iev
ed
a
p
r
e
cisi
o
n
s
co
r
e
o
f
8
9
% with
FP
S v
alu
e
o
f
3
8
,
im
p
lem
en
te
d
o
n
a
PC
with
C
PU I
n
tel
i7
-
1
0
7
0
0
k
C
PU@
3
.
8
0
GHZ
[
1
5
]
.
Po
th
o
le
d
etec
tio
n
o
n
m
o
b
ile
d
ev
ices
is
r
elativ
ely
r
ar
e
,
with
o
n
e
ex
am
p
le
u
s
in
g
YOL
Ov
3
o
n
an
I
o
T
d
ev
ice,
ac
h
iev
in
g
4
0
%
AP
an
d
0
.
8
3
s
ec
o
n
d
s
p
ee
d
[
1
6
]
.
W
h
ile
r
ec
en
t
r
esear
ch
h
as
ex
p
lo
r
ed
m
o
b
ile
im
p
lem
en
tatio
n
,
m
o
s
t
s
tu
d
ies
f
o
cu
s
o
n
PC
-
b
ased
m
o
d
el
co
m
p
ar
is
o
n
s
with
f
u
tu
r
e
m
o
b
ile
d
ep
lo
y
m
en
t
p
lan
s
[
1
5
]
,
[
1
7
]
,
[
1
8
]
.
C
o
m
p
a
r
ativ
e
s
tu
d
ies
h
av
e
e
x
am
in
e
d
th
e
p
er
f
o
r
m
an
ce
o
f
YOL
O
an
d
f
aster
R
-
C
NN
f
o
r
p
o
th
o
le
d
etec
tio
n
[
1
9
]
,
[
2
0
]
.
C
o
m
p
ar
ativ
e
s
tu
d
ies
s
h
o
w
YOL
Ov
7
o
u
tp
er
f
o
r
m
s
YOL
Ov
6
an
d
v
5
in
d
etec
tio
n
task
[
2
1
]
.
YOL
Ov
7
ca
n
also
b
e
co
m
b
in
ed
with
s
m
ar
tp
h
o
n
e
s
en
s
o
r
s
s
u
ch
as
ac
ce
ler
o
m
eter
s
to
im
p
r
o
v
e
d
etec
tio
n
p
er
f
o
r
m
a
n
ce
[
2
2
]
.
New
ar
ch
itectu
r
e
YOL
Ov
8
c
o
u
ld
ac
h
ie
v
es
9
2
%
d
etec
tio
n
ac
cu
r
ac
y
o
n
PC
.
YOL
Ov
8
ca
n
b
e
f
u
r
th
e
r
en
h
an
ce
d
with
C
NNs
f
o
r
im
p
r
o
v
ed
d
etec
tio
n
p
er
f
o
r
m
an
c
e
[
2
3
]
.
Ad
d
itio
n
ally
,
YOL
Ov
8
ca
n
b
e
co
m
b
in
ed
with
s
m
ar
tp
h
o
n
e
s
en
s
o
r
s
lik
e
GPS
to
tr
ac
k
p
o
th
o
le
lo
ca
tio
n
s
[
2
4
]
.
R
ec
en
tly
,
r
esear
ch
h
as
f
o
c
u
s
ed
o
n
im
p
r
o
v
in
g
t
h
e
lig
h
tweig
h
t
YOL
Ov
8
n
f
o
r
m
o
b
ile
d
ep
lo
y
m
en
t,
ac
h
i
ev
in
g
a
n
AP
v
alu
e
o
f
5
3
.
1
% o
n
t
h
e
R
DD2
0
2
2
d
at
ase
t
[
2
5
]
.
On
th
e
o
th
er
h
a
n
d
,
m
o
r
e
r
es
ea
r
ch
ed
h
av
e
b
ee
n
d
ev
elo
p
e
d
f
o
r
g
e
n
er
al
o
b
ject
d
etec
tio
n
o
n
m
o
b
ile
d
ev
ice
in
cl
u
d
in
g
s
m
ar
tp
h
o
n
e.
Ob
ject
d
etec
tio
n
m
o
d
els
h
av
e
b
ee
n
a
p
p
lied
to
t
h
e
d
iv
er
s
e
C
OC
O
d
ataset
u
s
in
g
An
d
r
o
id
s
m
ar
tp
h
o
n
es
an
d
I
p
h
o
n
e
u
s
in
g
tf
lite
f
o
r
m
at
o
f
f
p
1
6
an
d
in
t8
o
p
tim
izatio
n
[
2
6
]
,
[
2
7
]
.
I
n
ad
d
itio
n
,
th
er
e
ar
e
s
ev
er
al
o
b
ject
d
etec
tio
n
s
o
n
s
m
aller
o
b
jects,
in
cl
u
d
in
g
f
o
o
d
[
2
8
]
,
as
well
as
f
ly
im
ag
in
g
d
atasets
[
2
9
]
a
n
d
p
ests
also
u
s
in
g
s
m
a
r
tp
h
o
n
es
[
3
0
]
.
Fu
r
th
e
r
m
o
r
e
,
t
h
er
e
ar
e
ap
p
licatio
n
s
o
n
m
o
b
i
le
d
ev
ices
s
u
ch
as
R
asp
b
er
r
y
T
PU
f
o
r
p
ea
ch
cu
ltiv
ar
d
etec
tio
n
[
3
1
]
,
to
m
ato
v
ar
iety
d
etec
tio
n
[
3
2
]
an
d
J
e
ts
o
n
T
X2
f
o
r
cr
o
p
s
ee
d
lin
g
d
etec
tio
n
[
3
3
]
.
So
m
e
r
ec
en
t
r
esear
ch
r
elate
d
to
o
b
ject
d
etec
tio
n
m
o
d
el
o
n
s
m
ar
tp
h
o
n
e
o
r
m
o
b
il
e
d
ev
ice
ar
e
T
u
n
a
-
YOL
O
Ob
jec
t
d
etec
tio
n
m
o
d
el
wh
ich
h
as
b
ee
n
d
ev
elo
p
ed
to
d
etec
t
tu
n
a
v
ar
ieties
o
n
m
o
b
ile
d
ev
ices,
th
is
alg
o
r
ith
m
b
ased
o
n
YOL
Ov
3
u
s
in
g
Mo
b
ileNet
an
d
tr
an
s
f
er
lear
n
in
g
m
eth
o
d
.
B
ased
o
n
th
e
test
r
esu
lt
it
was
f
o
u
n
d
th
at
th
e
h
i
g
h
es
m
AP
v
alu
e
o
f
8
5
.
7
4
%
with
p
er
f
o
r
m
an
ce
o
f
1
5
.
3
2
F
PS
[
3
4
]
.
On
e
o
f
th
e
latest
r
esear
ch
is
cr
ea
tin
g
a
co
r
n
s
ee
d
lin
g
d
etec
tio
n
m
o
d
el
with
m
o
b
ile
d
ev
ice
u
s
in
g
v
ar
iatio
n
s
o
f
th
e
YOL
Ov
7
an
d
YOL
Ov
8
m
o
d
e
ls
,
wh
ich
s
u
cc
ess
f
u
lly
ac
h
iev
e
d
ac
cu
r
ac
y
o
f
ar
o
u
n
d
9
3
%
an
d
FP
S
v
alu
es
o
f
9
0
an
d
1
2
5
r
esp
ec
tiv
ely
o
n
m
o
b
il
e
d
ev
ice
[
3
5
]
.
T
h
e
liter
atu
r
e
r
e
v
iew
h
ig
h
lig
h
ts
a
n
ee
d
f
o
r
p
o
th
o
le
d
etec
t
io
n
s
y
s
tem
s
th
at
o
p
er
ate
e
f
f
i
cien
tly
o
n
d
ev
ices
with
lim
ited
co
m
p
u
tatio
n
al
r
eso
u
r
ce
s
,
s
u
ch
as
s
m
ar
tp
h
o
n
es.
Alth
o
u
g
h
p
r
e
v
io
u
s
r
esear
ch
s
h
o
ws
th
at
th
e
YOL
Ov
8
alg
o
r
ith
m
c
o
u
l
d
o
u
tp
e
r
f
o
r
m
s
o
th
er
s
tate
o
f
th
e
ar
t
(
So
T
A)
m
o
d
els
lik
e
YOL
Ov
7
an
d
f
aster
R
-
C
NN,
th
eir
ap
p
licatio
n
in
m
o
b
ile
p
o
th
o
le
d
etec
tio
n
r
e
m
ain
s
u
n
d
e
r
ex
p
l
o
r
ed
.
T
h
is
r
esear
ch
aim
s
to
c
o
n
d
u
ct
in
cr
em
en
tal
in
n
o
v
atio
n
b
y
r
et
r
ain
in
g
a
n
d
e
v
alu
atin
g
YOL
Ov
8
,
YOL
Ov
7
,
an
d
f
aster
R
-
C
NN
m
o
d
els
u
s
in
g
tr
an
s
f
er
lear
n
in
g
an
d
B
ay
esia
n
h
y
p
er
p
ar
am
eter
tu
n
i
n
g
[
3
6
]
,
[
3
7
]
.
T
h
e
f
o
cu
s
will
b
e
o
n
s
m
all,
tin
y
,
o
r
n
an
o
v
ar
ian
ts
with
less
th
an
2
0
m
illi
o
n
p
ar
am
eter
s
to
en
s
u
r
e
ef
f
i
cien
t
d
ep
lo
y
m
e
n
t
o
n
s
m
ar
tp
h
o
n
es.
T
h
e
g
o
al
is
to
ass
es
s
an
d
co
m
p
ar
e
th
ese
m
o
d
els
in
ter
m
s
o
f
b
o
th
d
ete
ctio
n
p
er
f
o
r
m
a
n
ce
a
n
d
c
o
m
p
u
tatio
n
al
r
eso
u
r
ce
ef
f
icien
cy
,
th
er
eb
y
ad
v
an
cin
g
in
tellig
en
t
tr
a
n
s
p
o
r
tatio
n
s
y
s
tem
s
.
T
h
e
r
esear
ch
is
e
x
p
ec
ted
t
o
en
a
b
le
r
ea
l
-
tim
e
p
o
th
o
le
d
etec
tio
n
o
n
m
o
b
ile
d
ev
ices
an
d
p
r
o
v
id
e
v
alu
ab
le
d
ata
f
o
r
i
n
f
r
astru
ctu
r
e
m
ain
ten
a
n
ce
an
d
p
lan
n
in
g
.
2.
M
E
T
H
O
D
2
.
1
.
P
r
o
po
s
ed
s
y
s
t
em
Au
to
m
atic
p
o
th
o
le
d
etec
tio
n
is
in
ten
d
ed
to
h
elp
p
r
ev
en
t
an
d
m
in
im
ize
tr
af
f
ic
ac
ci
d
en
t
r
is
k
s
.
T
h
is
s
tu
d
y
in
tr
o
d
u
ce
s
ad
v
an
c
ed
m
o
d
els
th
at
ca
n
b
e
im
p
le
m
en
ted
o
n
d
ev
ices
with
lim
ited
co
m
p
u
tatio
n
al
ca
p
ac
ity
,
lik
e
t
h
e
wid
ely
-
u
s
ed
s
m
ar
tp
h
o
n
es.
T
h
e
r
esear
ch
in
v
o
l
v
e
d
tr
ain
i
n
g
th
ese
m
o
d
els
th
r
o
u
g
h
tr
an
s
f
er
lear
n
in
g
an
d
h
y
p
er
p
ar
a
m
eter
t
u
n
in
g
p
r
o
ce
d
u
r
e
o
n
So
T
A
alg
o
r
ith
m
s
,
in
clu
d
in
g
f
aster
R
-
C
NN
Mo
b
ileNetV3
,
YOL
Ov
7
t
in
y
,
an
d
YOL
Ov
8
s
m
all,
u
s
in
g
Go
o
g
le
C
o
lab
T
4
an
d
p
y
th
o
n
p
r
o
g
r
am
m
in
g
la
n
g
u
ag
e
.
Ad
d
itio
n
ally
,
th
e
im
p
lem
en
tatio
n
p
r
o
ce
d
u
r
e
was
test
ed
o
n
an
An
d
r
o
id
s
m
ar
tp
h
o
n
e
,
s
p
ec
if
ically
th
e
R
ed
m
i
No
te
1
2
.
T
h
e
ev
alu
atio
n
ass
ess
ed
b
o
th
d
etec
tio
n
ac
cu
r
ac
y
an
d
th
e
c
o
m
p
u
tatio
n
al
r
eso
u
r
ce
u
s
ag
e,
ai
m
in
g
to
i
d
en
tify
t
h
e
m
o
d
el
th
at
o
f
f
er
ed
th
e
b
est
b
alan
ce
o
f
d
etec
tio
n
ac
c
u
r
ac
y
an
d
ef
f
icien
c
y
f
o
r
r
ea
l
-
wo
r
ld
p
o
th
o
le
d
etec
tio
n
o
n
m
o
b
ile
d
ev
ices
[
2
6
]
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
I
SS
N:
2
5
0
2
-
4
7
52
C
o
mp
a
r
a
tive
s
tu
d
y
o
f p
o
th
o
le
d
etec
tio
n
u
s
in
g
d
ee
p
lea
r
n
in
g
o
n
s
ma
r
tp
h
o
n
e
(
A
ch
ya
r
Ulu
l
A
mri
)
997
T
h
e
s
tep
s
in
v
o
lv
ed
in
th
e
p
r
o
p
o
s
ed
m
o
d
el,
as illu
s
tr
ated
in
t
h
e
Fig
u
r
e
1
:
-
Data
in
p
u
t: c
o
llect
r
aw
im
ag
e
d
ata
f
o
r
a
n
aly
s
is
.
-
Data
p
r
ep
r
o
ce
s
s
in
g
: p
r
ep
ar
e
i
m
ag
es th
r
o
u
g
h
cr
o
p
p
i
n
g
,
r
esizin
g
in
to
8
0
0
×
800
,
an
d
n
o
r
m
ali
za
tio
n
.
-
L
ab
el
re
-
f
o
r
m
attin
g
: c
o
n
v
er
t
l
ab
els to
YOL
O
an
d
Pas
ca
l V
OC
f
o
r
m
ats s
u
itab
le
f
o
r
d
if
f
er
en
t m
o
d
els
.
-
Data
s
p
litt
in
g
: d
iv
id
e
th
e
p
r
e
p
r
o
ce
s
s
ed
d
ata
in
to
tr
ain
in
g
,
v
ali
d
atio
n
,
an
d
test
in
g
s
u
b
s
ets.
-
T
r
ain
in
g
a
n
d
f
in
e
tu
n
i
n
g
:
tr
a
in
an
d
f
i
n
e
-
tu
n
e
m
o
d
els
lik
e
YOL
Ov
7
t
in
y
,
YOL
Ov
8
s
m
all,
an
d
f
aster
R
-
C
NN
Mo
b
ileNetV3
u
s
in
g
a
B
ay
esian
s
ea
r
ch
m
eth
o
d
,
with
1
0
iter
atio
n
s
co
n
d
u
cted
f
o
r
ea
ch
m
o
d
el.
-
Mo
d
el
co
n
v
er
s
io
n
: c
o
n
v
er
t
ea
ch
b
est tu
n
e
m
o
d
els to
T
FLite
f
o
r
m
ats (
FP
1
6
,
I
NT
8
)
wh
er
e
a
p
p
licab
le.
-
Dete
ctio
n
ac
cu
r
ac
y
test
in
g
: te
s
t
ac
cu
r
ac
y
o
n
m
o
b
ile
-
o
p
tim
ize
d
m
o
d
els.
-
Mo
d
el
ap
p
d
ep
lo
y
m
en
t: d
e
p
lo
y
m
o
d
els with
in
a
a
n
d
r
o
id
ap
p
wh
ich
cr
ea
ted
u
s
in
g
an
d
r
o
id
s
tu
d
io
.
-
Sam
p
lin
g
f
o
r
test
s
im
u
latio
n
:
cr
ea
te
1
0
test
s
u
s
ets with
5
0
im
ag
es e
ac
h
f
o
r
r
ea
l
-
wo
r
ld
s
im
u
latio
n
s
.
-
R
eso
u
r
ce
u
s
ag
e
test
in
g
:
m
ea
s
u
r
e
ea
ch
m
o
d
el
’
s
in
f
er
en
ce
ti
m
e
an
d
r
eso
u
r
ce
u
s
ag
e
o
n
s
m
ar
tp
h
o
n
es
u
s
in
g
An
d
r
o
id
s
tu
d
i
o
p
r
o
f
iler
.
Fig
u
r
e
1
.
Flo
w
d
ia
g
r
am
o
f
th
e
p
o
th
o
le
d
etec
tio
n
s
y
s
tem
f
r
o
m
d
ev
elo
p
m
en
t to
s
m
ar
tp
h
o
n
e
d
ep
lo
y
m
en
t
2
.
2
.
Da
t
a
s
et
W
HO
n
o
tes
th
at
lo
w
-
to
m
id
d
le
-
in
co
m
e
co
u
n
tr
ies
o
f
ten
o
f
ten
f
ailin
g
to
m
ee
t
s
af
ety
s
tan
d
ar
d
s
[
1
]
.
T
o
r
ef
lect
th
ese
c
o
n
d
itio
n
s
,
t
h
is
s
tu
d
y
u
s
es
So
u
th
Af
r
ica
a
s
a
r
ep
r
esen
tativ
e
co
u
n
tr
y
,
d
r
awin
g
o
n
a
d
ataset
f
r
o
m
Nian
a
b
er
et
al.
f
o
c
u
s
ed
o
n
p
o
t
h
o
le
d
etec
ti
o
n
[
6
]
.
T
h
e
d
ata
was
co
llected
u
s
in
g
a
Go
Pro
Her
o
3
+
ca
m
er
a
m
o
u
n
ted
as
a
d
ash
ca
m
,
ca
p
t
u
r
in
g
im
ag
es
at
a
r
eso
lu
tio
n
o
f
3
6
8
0
×
2
7
6
0
p
ix
els
in
J
PG
f
o
r
m
at.
T
h
e
d
ataset
co
n
s
is
ts
o
f
1
3
.
4
K
im
ag
es,
b
u
t o
n
ly
3
.
8
K
im
ag
es
co
n
tain
in
g
l
ab
eled
p
o
th
o
les
wer
e
u
s
ed
,
s
u
p
p
lem
en
ted
b
y
4
0
0
n
eg
ativ
e
class
im
ag
es.
T
h
is
r
esu
lts
in
a
to
tal
o
f
4
.
1
K
im
ag
es
an
aly
ze
d
f
o
r
th
is
s
tu
d
y
,
with
ea
ch
im
ag
e
lab
eled
b
y
ex
p
er
ts
to
id
en
tify
p
o
th
o
les,
r
an
g
in
g
f
r
o
m
0
to
m
o
r
e
th
an
1
p
er
im
a
g
e,
as illu
s
tr
ated
in
F
ig
u
r
e
2
.
Neg
ati
ve
(
0
)
Po
s
iti
ve
(
1
)
Po
s
iti
ve
(
>1
)
Fig
u
r
e
2
.
Sam
p
le
d
ataset
o
f
p
o
th
o
le
r
o
ad
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
5
0
2
-
4
7
52
In
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
,
Vo
l.
3
7
,
No
.
2
,
Feb
r
u
a
r
y
20
2
5
:
9
9
5
-
1
0
0
4
998
2
.
3
.
Da
t
a
p
re
pa
ra
t
i
o
n
T
h
e
o
r
ig
in
al
d
ataset
im
ag
es
ar
e
tak
en
f
r
o
m
r
elativ
el
y
f
ix
ed
p
o
s
itio
n
d
ash
ca
m
it
is
n
o
t
o
n
ly
ca
p
tu
r
ed
th
e
r
ao
d
b
u
t
also
ca
r
d
ash
b
o
ar
d
an
d
s
k
y
.
T
h
e
p
r
ev
i
o
u
s
r
esear
ch
s
u
g
g
est
to
d
o
wn
s
ca
le
an
d
cr
o
p
th
e
im
ag
e
a
n
d
f
o
cu
s
o
n
th
e
r
o
ad
s
ec
tio
n
[
3
8
]
.
T
h
en
it
was
tr
an
s
f
o
r
m
e
d
u
s
in
g
a
Py
th
o
n
p
r
o
g
r
am
.
T
h
i
s
p
r
o
ce
s
s
in
clu
d
ed
re
-
lab
elin
g
d
ata
in
to
Pas
ca
l
VOC
an
d
YOL
O
f
o
r
m
ats,
cr
o
p
p
in
g
im
ag
es
to
f
o
cu
s
o
n
t
h
e
r
o
a
d
,
r
esizin
g
t
o
800
×
8
0
0
[
1
4
]
,
n
o
r
m
alizin
g
i
m
ag
es,
an
d
s
p
litt
in
g
th
e
d
ata
in
to
1
,
6
7
2
tr
ain
,
1
,
2
5
4
v
alid
a
tio
n
,
an
d
1
,
2
5
4
test
im
ag
es.
2
.
4
.
F
a
s
t
er
R
-
CNN
Fas
ter
R
-
C
NN
i
s
a
two
-
s
tep
o
b
ject
d
etec
tio
n
m
eth
o
d
t
h
at
f
ir
s
t p
r
o
p
o
s
es o
b
ject
r
eg
io
n
s
f
r
o
m
an
im
ag
e
b
ef
o
r
e
p
er
f
o
r
m
in
g
class
if
icatio
n
[
3
9
]
.
I
n
th
e
co
n
tex
t
o
f
p
o
th
o
le
d
etec
tio
n
,
th
e
m
o
d
el
is
tr
ai
n
ed
u
s
in
g
a
d
ataset
o
f
im
ag
es
co
n
tain
in
g
p
o
th
o
l
es,
wh
er
e
ea
ch
im
ag
e
is
an
n
o
tated
to
in
d
icate
th
e
lo
ca
tio
n
o
f
th
e
p
o
th
o
les.
Mo
b
ileNetV3
,
a
v
ar
iatio
n
o
f
f
aster
R
-
C
NN,
h
as
a
r
elativ
ely
s
m
all
n
u
m
b
er
o
f
p
a
r
am
eter
s
,
ar
o
u
n
d
2
0
m
illi
o
n
,
an
d
o
f
f
er
s
g
o
o
d
d
etec
tio
n
ac
cu
r
ac
y
,
m
ak
i
n
g
it su
itab
le
f
o
r
u
s
e
o
n
m
o
b
ile
d
ev
ices
[
4
0
]
.
2
.
5
.
YO
L
O
v
7
YOL
Ov
7
is
a
s
in
g
le
-
s
tep
o
b
ject
d
etec
tio
n
m
eth
o
d
t
h
at
in
c
lu
d
es
a
b
ac
k
b
o
n
e,
h
ea
d
,
an
d
n
ec
k
,
an
d
u
tili
ze
s
ex
ten
d
ed
ef
f
icien
t
lay
er
ag
g
r
eg
atio
n
n
etwo
r
k
(E
-
E
L
AN)
b
lo
ck
s
an
d
s
ca
lab
le
b
a
g
o
f
f
r
ee
b
ies
(
B
o
F)
tech
n
iq
u
es
[
4
1
]
.
W
h
en
d
ev
el
o
p
in
g
a
p
o
th
o
le
d
etec
tio
n
m
o
d
el,
YOL
Ov
7
is
tr
ain
ed
o
n
a
n
an
n
o
tated
im
ag
e
d
ataset
th
at
m
ar
k
s
th
e
lo
ca
tio
n
s
o
f
p
o
th
o
les.
T
h
e
m
o
d
el
ef
f
icien
tly
lear
n
s
to
d
etec
t
an
d
l
o
ca
lize
p
o
th
o
les
in
o
n
e
p
ass
th
r
o
u
g
h
th
e
im
ag
e.
Y
OL
Ov
7
t
in
y
,
with
6
.
2
m
illi
o
n
p
ar
am
eter
s
,
co
u
ld
o
f
f
er
g
o
o
d
d
etec
tio
n
ac
cu
r
ac
y
,
an
d
s
u
itab
le
f
o
r
im
p
lem
en
tatio
n
o
n
s
m
ar
tp
h
o
n
es.
2
.
6
.
YO
L
O
v
8
YOL
Ov
8
is
a
s
in
g
le
-
s
tep
o
b
je
ct
d
etec
tio
n
m
et
h
o
d
co
m
p
r
is
in
g
a
b
ac
k
b
o
n
e,
h
ea
d
,
an
d
n
ec
k
,
u
tili
zin
g
C
o
n
v
Mo
d
u
le
b
lo
ck
s
,
C
SP
L
ay
er
2
c
o
n
v
(
c2
f
)
,
an
d
SP
PF
.
T
h
is
m
o
d
el
is
also
an
c
h
o
r
-
f
r
ee
,
en
h
an
cin
g
its
d
etec
tio
n
ca
p
ab
ilit
ies
[
4
2
]
.
Fo
r
d
ev
elo
p
in
g
a
p
o
th
o
le
d
etec
tio
n
m
o
d
el,
YOL
Ov
8
is
tr
ain
ed
o
n
a
d
ataset
with
an
n
o
tated
p
o
th
o
le
lo
ca
tio
n
s
,
e
n
ab
lin
g
th
e
m
o
d
el
to
ef
f
icien
t
ly
d
etec
t
an
d
lo
ca
lize
p
o
th
o
le
s
in
a
s
in
g
le
p
ass
.
YOL
Ov
8
s
m
all,
with
1
1
.
2
m
illi
o
n
p
ar
a
m
eter
s
,
an
d
YO
L
Ov
8
n
a
n
o
,
with
3
.
2
m
illi
o
n
p
ar
a
m
eter
s
,
o
f
f
er
r
elativ
ely
g
o
o
d
d
etec
tio
n
ac
cu
r
ac
y
an
d
well
-
s
u
ited
f
o
r
m
o
b
il
e
d
ev
ices
im
p
lem
en
tatio
n
.
2
.
7
.
T
ra
ns
f
er
lea
rning
a
nd
hy
perpa
ra
m
et
er
t
un
ing
I
n
th
is
s
tu
d
y
,
tr
an
s
f
er
lear
n
in
g
was
em
p
lo
y
e
d
b
y
f
in
e
-
tu
n
in
g
p
r
e
-
tr
ain
ed
m
o
d
els
o
n
th
e
tar
g
et
d
ataset.
T
h
is
ap
p
r
o
ac
h
lev
e
r
ag
es
th
e
k
n
o
wled
g
e
f
r
o
m
m
o
d
els
tr
ain
ed
o
n
lar
g
e
d
atasets
,
allo
win
g
f
o
r
q
u
ick
er
co
n
v
er
g
en
ce
an
d
im
p
r
o
v
e
d
p
e
r
f
o
r
m
a
n
ce
o
n
th
e
s
p
ec
if
ic
task
[
4
3
]
.
T
o
f
u
r
th
e
r
en
h
a
n
ce
th
e
m
o
d
els,
B
ay
esian
o
p
tim
izatio
n
was
u
s
ed
f
o
r
h
y
p
e
r
p
ar
am
ete
r
tu
n
i
n
g
,
ch
o
s
en
f
o
r
its
ef
f
icien
cy
in
id
en
tify
in
g
o
p
tim
al
co
n
f
ig
u
r
atio
n
s
.
T
h
e
W
an
d
b
li
b
r
ar
y
f
ac
ilit
ated
m
o
n
ito
r
i
n
g
a
n
d
an
aly
zi
n
g
th
e
im
p
ac
t
o
f
ea
ch
h
y
p
er
p
ar
a
m
eter
[
4
4
]
.
In
th
is
s
tu
d
y
,
h
y
p
er
p
a
r
a
m
eter
tu
n
in
g
was
p
er
f
o
r
m
ed
f
o
r
1
0
iter
atio
n
s
o
n
ea
c
h
m
o
d
e
l
with
s
ea
r
ch
s
p
ac
e
in
clu
d
e
o
p
tim
izer
s
test
ed
b
ein
g
Ad
am
W
an
d
SGD.
T
h
e
tr
ai
n
in
g
r
an
f
o
r
1
5
0
ep
o
c
h
s
,
with
a
p
atien
ce
o
f
2
0
f
o
r
ea
r
ly
s
to
p
p
in
g
.
B
atch
s
izes
o
f
8
,
1
6
,
a
n
d
3
2
,
al
o
n
g
s
id
e
a
lea
r
n
in
g
r
ate
s
am
p
led
u
n
if
o
r
m
ly
b
etwe
en
0
.
0
0
1
a
n
d
0
.
0
1
.
M
o
m
en
tu
m
with
in
a
u
n
if
o
r
m
r
a
n
g
e
o
f
0
.
8
to
0
.
9
8
.
2
.
8
.
T
ens
o
rF
lo
w
l
it
e
T
o
d
e
p
lo
y
o
b
ject
d
etec
tio
n
m
o
d
els
lik
e
th
o
s
e
b
u
ilt
with
Py
T
o
r
ch
o
r
YOL
O
o
n
an
A
n
d
r
o
id
ap
p
,
th
e
m
o
d
els
n
ee
d
to
b
e
c
o
n
v
e
r
ted
t
o
T
en
s
o
r
Flo
w
l
ite
f
o
r
m
at.
T
h
e
co
n
v
er
s
io
n
p
r
o
ce
s
s
ty
p
ically
in
v
o
lv
es
ex
p
o
r
tin
g
th
e
Py
T
o
r
ch
o
r
YOL
O
m
o
d
e
l
to
o
p
e
n
n
e
u
r
al
n
etwo
r
k
ex
c
h
a
nge
(
ONNX
)
f
o
r
m
at
f
ir
s
t,
wh
ich
s
er
v
es
as
an
in
ter
m
ed
iate
r
ep
r
esen
tatio
n
.
T
h
en
th
e
m
o
d
el
b
ein
g
co
n
v
e
r
ted
to
T
e
n
s
o
r
Flo
w
f
o
r
m
at,
f
o
llo
wed
b
y
a
f
in
al
co
n
v
er
s
io
n
to
T
e
n
s
o
r
Flo
w
l
ite
u
s
in
g
T
en
s
o
r
Flo
w
l
ite
co
n
v
er
ter
.
Du
r
in
g
th
is
p
r
o
ce
s
s
,
q
u
a
n
tizatio
n
tech
n
iq
u
es
li
k
e
FP
1
6
o
r
I
NT
8
ar
e
ap
p
li
ed
to
r
e
d
u
ce
th
e
m
o
d
el
s
ize
an
d
co
m
p
u
tatio
n
r
eq
u
i
r
em
en
t
s
,
o
p
tim
izin
g
it
f
o
r
ef
f
icien
t o
n
-
d
ev
ice
in
f
er
en
ce
[
2
6
]
,
[
2
7
]
.
2
.
9
.
Andro
id
s
t
ud
io
An
d
r
o
id
s
tu
d
io
,
th
e
o
f
f
icia
l
in
teg
r
ated
d
ev
elo
p
m
en
t
e
n
v
ir
o
n
m
en
t
(
I
DE
)
f
o
r
An
d
r
o
id
ap
p
d
ev
elo
p
m
e
n
t,
p
r
o
v
id
es
r
o
b
u
s
t
to
o
ls
f
o
r
co
d
i
n
g
,
d
eb
u
g
g
in
g
,
test
in
g
,
an
d
o
p
tim
izin
g
a
p
p
lic
atio
n
s
.
T
h
e
p
o
th
o
le
d
etec
tio
n
ap
p
was
d
e
v
elo
p
e
d
u
s
in
g
t
h
e
Gir
af
f
e
v
e
r
s
io
n
o
f
An
d
r
o
id
s
tu
d
io
.
A
d
d
itio
n
al
ly
,
An
d
r
o
id
s
tu
d
io
in
clu
d
es
a
f
ea
tu
r
e
ca
lled
An
d
r
o
id
s
tu
d
io
p
r
o
f
iler
,
wh
ic
h
allo
ws
d
ev
elo
p
er
s
to
m
ea
s
u
r
e
a
p
p
p
er
f
o
r
m
a
n
ce
in
ter
m
s
o
f
C
PU,
R
AM
,
an
d
b
atter
y
u
s
ag
e,
e
n
s
u
r
in
g
e
f
f
icien
t
,
a
n
d
o
p
tim
ize
d
ap
p
b
eh
a
v
io
r
[
4
5
]
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
I
SS
N:
2
5
0
2
-
4
7
52
C
o
mp
a
r
a
tive
s
tu
d
y
o
f p
o
th
o
le
d
etec
tio
n
u
s
in
g
d
ee
p
lea
r
n
in
g
o
n
s
ma
r
tp
h
o
n
e
(
A
ch
ya
r
Ulu
l
A
mri
)
999
2
.
10
.
E
v
a
lua
t
io
n
Pro
ce
s
s
to
id
en
tify
th
e
m
o
s
t
ef
f
ec
tiv
e
d
ee
p
lear
n
in
g
m
o
d
e
l
f
o
r
p
o
th
o
le
d
etec
tio
n
o
n
s
m
a
r
tp
h
o
n
e
r
eq
u
ir
es
ev
alu
atin
g
b
o
th
d
ete
ctio
n
p
er
f
o
r
m
an
ce
a
n
d
r
eso
u
r
ce
ef
f
icien
cy
.
E
ac
h
m
o
d
el
is
an
aly
ze
d
b
ased
o
n
th
ese
cr
iter
ia,
an
d
r
esu
lts
ar
e
c
o
m
p
ar
ed
.
T
h
is
p
r
o
ce
s
s
id
en
tifie
s
m
o
s
t su
itab
le
m
o
d
el
f
o
r
d
ep
lo
y
m
en
t
[
2
6
]
.
Dete
ctio
n
ac
cu
r
ac
y
is
m
ea
s
u
r
ed
u
s
in
g
th
e
av
e
r
ag
e
p
r
ec
is
io
n
(
AP)
m
etr
ic,
wh
ich
is
b
a
s
ed
o
n
th
e
co
n
ce
p
ts
o
f
I
o
U,
p
r
ec
is
io
n
,
an
d
r
ec
all
[
2
4
]
.
I
n
ter
s
ec
tio
n
o
v
e
r
u
n
i
o
n
(
I
o
U)
q
u
a
n
tifie
s
th
e
o
v
er
lap
b
etwe
en
th
e
p
r
ed
icted
b
o
u
n
d
in
g
b
o
x
a
n
d
th
e
g
r
o
u
n
d
tr
u
th
,
co
n
s
id
er
in
g
o
n
ly
p
r
ed
ictio
n
s
th
at
ex
ce
e
d
a
s
et
co
n
f
id
en
ce
th
r
esh
o
ld
,
ty
p
ically
0
.
5
,
to
id
en
tify
tr
u
e
p
o
s
itiv
es
(
T
P),
as
s
ee
n
in
1
.
Fals
e
p
o
s
itiv
es
(
FP
)
o
cc
u
r
wh
e
n
th
e
mod
el
in
co
r
r
ec
tly
id
en
tifie
s
a
p
o
th
o
le
wh
er
e
n
o
n
e
ex
is
ts
,
wh
ile
f
alse
n
eg
ativ
es
(
FN)
ar
is
e
wh
en
th
e
m
o
d
el
m
is
s
es a
n
ac
tu
al
p
o
th
o
le.
W
ith
T
P,
FP
,
an
d
FN d
ef
in
ed
,
p
r
ec
i
s
io
n
an
d
r
ec
all
ar
e
ca
lcu
lated
-
p
r
ec
is
io
n
in
d
icate
s
th
e
p
er
ce
n
tag
e
o
f
c
o
r
r
ec
t
p
r
e
d
ictio
n
s
,
as
s
ee
n
in
2
,
an
d
r
ec
all
m
ea
s
u
r
es
th
e
p
er
ce
n
tag
e
o
f
a
ctu
al
p
o
s
itiv
es
th
a
t
ar
e
ac
cu
r
ately
d
etec
ted
,
as
s
ee
n
in
3
.
Fin
ally
,
AP
is
ca
lcu
lated
as
th
e
AP
ac
r
o
s
s
all
r
ec
all
lev
els,
r
ep
r
esen
ted
b
y
th
e
ar
ea
u
n
d
e
r
th
e
p
r
ec
is
io
n
-
r
ec
all
cu
r
v
e,
as
s
ee
n
in
4
.
T
h
is
d
etec
ti
o
n
ac
cu
r
ac
y
ev
alu
at
io
n
is
p
er
f
o
r
m
ed
o
n
th
e
test
d
ata
to
o
b
jectiv
ely
ass
ess
th
e
m
o
d
el
’
s
p
er
f
o
r
m
an
ce
.
(
1
)
(
)
=
+
(
2
)
(
)
=
+
(
3
)
(
)
=
∫
(
)
1
=
0
(
4
)
Sm
ar
tp
h
o
n
e
r
eso
u
r
ce
u
s
ag
e
f
o
r
ea
ch
m
o
d
el
is
ev
alu
ated
b
y
m
ea
s
u
r
in
g
in
f
e
r
en
ce
s
p
ee
d
,
b
atter
y
u
s
ag
e,
an
d
C
PU/R
AM
co
n
s
u
m
p
tio
n
u
s
in
g
An
d
r
o
id
p
r
o
f
iler
[
4
5
]
.
I
n
f
er
e
n
ce
s
p
ee
d
,
m
e
asu
r
ed
in
s
ec
o
n
d
s
,
r
ef
lects
th
e
tim
e
r
e
q
u
ir
ed
f
o
r
t
h
e
m
o
d
el
to
g
en
er
ate
p
r
ed
ictio
n
s
f
r
o
m
ea
ch
im
a
g
e
as
s
ee
n
i
n
5
.
B
atter
y
u
s
ag
e
is
esti
m
ated
b
y
An
d
r
o
id
en
e
r
g
y
p
r
o
f
iler
,
ca
teg
o
r
izin
g
e
n
er
g
y
co
n
s
u
m
p
tio
n
in
to
lig
h
t,
m
ed
iu
m
,
an
d
h
ea
v
y
in
ten
s
ity
lev
els.
T
h
e
en
er
g
y
u
s
ag
e
p
er
ce
n
ta
g
e
with
in
th
e
lig
h
t
o
r
less
r
eso
u
r
ce
-
in
ten
s
iv
e
task
s
ca
teg
o
r
y
s
ee
n
in
6
.
C
PU
u
s
ag
e
is
r
ec
o
r
d
ed
b
y
C
PU
p
r
o
f
iler
as
a
p
er
ce
n
tag
e
o
f
to
tal
av
ailab
le
C
PU
tim
e
as
s
ee
n
in
7
,
wh
ile
R
AM
u
s
ag
e
is
tr
ac
k
ed
b
y
m
o
n
ito
r
in
g
th
e
p
r
iv
ate
m
e
m
o
r
y
p
ag
es
allo
ca
ted
to
th
e
ap
p
as
s
ee
n
in
8
.
T
h
e
ev
alu
atio
n
is
co
n
d
u
cted
o
n
1
0
s
u
b
s
ets
o
f
5
0
im
ag
es
ea
ch
,
f
o
cu
s
in
g
o
n
av
er
a
g
e
in
f
e
r
en
ce
tim
e,
av
er
ag
e
b
atter
y
u
s
ag
e,
a
n
d
m
ax
im
u
m
C
PU
an
d
R
AM
u
s
ag
e.
T
h
ese
m
etr
ics
d
eter
m
in
e
o
v
e
r
all
p
e
r
f
o
r
m
an
ce
,
h
elp
i
n
g
to
id
en
tify
th
e
m
o
d
el
with
th
e
b
e
s
t in
f
er
en
ce
s
p
ee
d
a
n
d
lo
west
r
eso
u
r
ce
co
n
s
u
m
p
tio
n
[
2
6
]
.
(
)
=
−
(
5
)
(
%
)
=
ℎ
ℎ
ℎ
×
100
(
6
)
(
%
)
=
ℎ
ℎ
×
100
(
7
)
(
)
=
ℎ
(
8
)
3.
RE
SU
L
T
S
Po
th
o
le
d
etec
tio
n
m
o
d
el
h
as
b
ee
n
d
ev
el
o
p
ed
o
n
PC
u
s
in
g
s
ev
er
al
p
r
etr
ain
ed
m
o
d
el
o
f
alg
o
r
ith
m
wh
ich
ar
e
YOL
Ov
8
-
s
m
all
,
Y
OL
Ov
8
-
n
an
o
,
YOL
Ov
7
-
t
in
y
a
n
d
f
aster
R
-
C
NN
Mo
b
ileNet
V3
.
E
ac
h
m
o
d
el
h
as
ex
p
er
ien
ce
1
0
iter
atio
n
o
f
h
y
p
er
p
ar
am
eter
tu
n
in
g
s
u
c
h
as
b
atch
s
ize,
lear
n
in
g
r
at
e,
an
d
m
o
m
en
tu
m
.
B
est h
y
p
er
p
ar
am
eter
s
f
o
r
ea
ch
m
o
d
el
b
ased
o
n
tu
n
in
g
r
esu
lt sh
o
wn
in
T
ab
l
e
1
.
T
h
e
T
ab
le
1
in
d
icate
s
th
at
o
n
th
e
v
alid
atio
n
d
ataset,
YOL
Ov
8
Sm
all
ac
h
iev
ed
th
e
h
i
g
h
est
AP
at
I
o
U
0
.
5
(
AP@
0
.
5
)
s
co
r
e
o
f
8
7
.
8
%.
YOL
Ov
8
n
an
o
f
o
llo
we
d
with
a
s
tr
o
n
g
8
4
.
9
%,
b
o
th
b
en
ef
itin
g
f
r
o
m
t
h
e
Ad
am
W
o
p
tim
izer
.
I
n
co
n
tr
a
s
t,
YOL
Ov
7
t
in
y
an
d
f
aster
R
-
C
NN
Mo
b
ileNetV3
s
h
o
we
d
lo
wer
v
alid
atio
n
p
er
f
o
r
m
an
ce
,
with
AP@
0
.
5
s
co
r
es
o
f
7
8
.
8
%
an
d
6
6
.
1
%
r
e
s
p
ec
tiv
ely
,
with
th
e
latter
u
s
i
n
g
th
e
less
ef
f
ec
tiv
e
SGD
o
p
tim
izer
.
T
h
is
h
ig
h
lig
h
t
s
th
e
ef
f
ec
tiv
en
ess
o
f
YOL
Ov
8
m
o
d
els
an
d
th
e
Ad
am
W
o
p
ti
m
izer
in
ac
h
iev
in
g
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
5
0
2
-
4
7
52
In
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
,
Vo
l.
3
7
,
No
.
2
,
Feb
r
u
a
r
y
20
2
5
:
9
9
5
-
1
0
0
4
1000
b
etter
v
alid
atio
n
r
esu
lts
.
Nex
t,
ea
ch
o
f
th
e
m
o
d
els
wer
e
tr
y
to
b
e
c
o
n
v
e
r
ted
to
T
FLite
f
o
r
m
at
(
f
lo
at
1
6
an
d
in
t8
)
an
d
ev
alu
ated
o
n
test
d
ataset.
T
h
e
m
o
d
els
d
etec
tio
n
p
er
f
o
r
m
a
n
ce
o
n
PC
in
Fig
u
r
e
3
its
o
r
ig
in
al
f
o
r
m
at
an
d
af
ter
co
n
v
er
s
io
n
to
T
FLite
is
s
h
o
wn
in
Fig
u
r
e
3
(
a
)
,
wh
i
le
th
e
m
o
d
els
f
ile
s
ize
co
m
p
ar
is
o
n
is
illu
s
tr
ated
in
Fig
u
r
e
3
(
b
)
.
T
ab
l
e
1.
B
est
h
y
p
er
p
ar
am
eter
tu
n
in
g
r
esu
lt f
o
r
ea
c
h
m
o
d
el
M
o
d
e
l
I
mg
s
i
z
e
B
a
t
c
h
Ep
o
c
h
s
l
r
0
M
o
me
n
t
u
m
O
p
t
i
mi
z
e
r
A
P
@
0
.
5
v
a
l
i
d
a
t
i
o
n
Y
O
LO
v
8
s
ma
l
l
8
0
0
×
8
0
0
8
1
5
0
0
.
0
0
1
4
0
.
9
0
8
A
d
a
mW
8
7
.
8
%
Y
O
LO
v
8
n
a
n
o
8
0
0
×
8
0
0
32
1
5
0
0
.
0
0
1
0
0
.
8
8
3
A
d
a
mW
8
4
.
9
%
Y
O
LO
v
7
t
i
n
y
8
0
0
×
8
0
0
16
1
5
0
0
.
0
0
7
6
0
.
9
5
7
A
d
a
mW
7
8
.
8
%
F
a
st
e
r
R
-
C
N
N
M
v
3
8
0
0
×
8
0
0
16
1
5
0
0
.
0
0
8
2
0
.
8
9
4
S
G
D
6
6
.
1
%
(
a)
(
b
)
Fig
u
r
e
3
.
C
o
m
p
a
r
is
o
n
o
f
m
o
d
els d
etec
tio
n
p
er
f
o
r
m
an
ce
o
n
(
a)
test
d
ataset
an
d
(
b
)
f
ile
s
ize
As
s
h
o
wn
in
Fig
u
r
e
3
,
th
e
YOL
Ov
8
s
m
all
an
d
n
an
o
m
o
d
els
ac
h
iev
ed
th
e
h
ig
h
est
d
etec
tio
n
ac
cu
r
ac
y
o
n
th
e
test
d
ataset
an
d
m
ain
t
ain
ed
p
er
f
o
r
m
a
n
ce
af
ter
co
n
v
er
s
io
n
wh
ich
h
ig
h
er
th
an
p
r
e
v
io
u
s
r
esear
ch
[
2
5
]
.
I
n
co
n
tr
ast,
YOL
Ov
7
t
in
y
m
o
d
el
h
av
e
a
s
ig
n
if
ican
t
a
cc
u
r
ac
y
d
r
o
p
a
f
ter
FP
1
6
c
o
n
v
e
r
s
io
n
,
s
o
I
NT
8
co
n
v
er
s
io
n
was
n
o
t
p
u
r
s
u
ed
.
T
h
e
f
aster
R
-
C
NN
Mo
b
ileNetV3
m
o
d
el
w
as
ex
clu
d
ed
f
r
o
m
T
FLite
co
n
v
er
s
io
n
d
u
e
to
its
o
r
lo
w
ac
cu
r
ac
y
a
n
d
lar
g
e
f
ile
s
ize.
No
tab
ly
,
YOL
Ov
8
T
FLite
-
I
NT
8
r
ed
u
ce
d
its
o
r
ig
in
al
f
ile
s
ize
b
y
5
0
%.
Fig
u
r
e
4
s
h
o
wca
s
es
a
s
am
p
le
co
m
p
ar
is
o
n
o
f
d
etec
tio
n
r
esu
lts
ac
r
o
s
s
d
if
f
er
en
t
m
o
d
els
u
n
d
er
v
ar
y
in
g
lig
h
tin
g
co
n
d
itio
n
s
.
I
n
b
r
ig
h
t l
ig
h
t,
b
o
th
YOL
Ov
8
s
m
all
an
d
n
an
o
m
o
d
els ac
cu
r
ately
d
etec
t a
ll p
o
th
o
les,
wh
ile
YOL
Ov
7
m
is
s
es
o
n
e.
Ho
wev
er
,
in
lo
w
-
lig
h
t
co
n
d
itio
n
s
with
s
h
ad
o
ws,
YOL
Ov
8
s
m
all
r
em
ain
s
p
r
ec
is
e,
id
en
tify
in
g
all
p
o
t
h
o
les,
w
h
er
ea
s
YOL
Ov
8
n
an
o
d
etec
ts
o
n
ly
f
o
u
r
,
an
d
YOL
Ov
7
c
ap
tu
r
es
ju
s
t
th
r
ee
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
I
SS
N:
2
5
0
2
-
4
7
52
C
o
mp
a
r
a
tive
s
tu
d
y
o
f p
o
th
o
le
d
etec
tio
n
u
s
in
g
d
ee
p
lea
r
n
in
g
o
n
s
ma
r
tp
h
o
n
e
(
A
ch
ya
r
Ulu
l
A
mri
)
1001
T
h
is
co
m
p
ar
is
o
n
h
i
g
h
lig
h
ts
th
e
m
o
d
els
’
s
tr
en
g
th
s
an
d
lim
ita
tio
n
s
,
p
ar
ticu
lar
ly
in
d
etec
tin
g
s
m
aller
o
r
d
is
tan
t
p
o
th
o
les u
n
d
er
p
o
o
r
lig
h
tin
g
c
o
n
d
itio
n
s
[
4
]
.
Fig
u
r
e
4
.
Sam
p
le
c
o
m
p
a
r
is
o
n
o
f
p
o
t
h
o
le
d
etec
tio
n
u
n
d
er
v
a
r
y
in
g
lig
h
tin
g
co
n
d
itio
n
s
Acc
o
r
d
in
g
to
Fig
u
r
e
5
s
h
o
ws
th
e
s
am
p
le
o
f
Fig
u
r
e
5
(
a)
s
h
o
wca
s
es
th
e
p
o
th
o
le
d
etec
tio
n
m
o
d
els
im
p
lem
en
tatio
n
o
n
a
s
m
ar
tp
h
o
n
e.
T
h
e
m
o
d
els
wer
e
test
ed
ac
r
o
s
s
ten
d
atasets
to
as
s
ess
i
n
f
er
en
ce
s
p
ee
d
a
n
d
d
etec
tio
n
ac
cu
r
ac
y
.
W
h
ile
Fi
g
u
r
e
5
(
b
)
p
r
o
v
id
es
in
s
ig
h
ts
in
to
r
eso
u
r
ce
u
s
ag
e
m
o
n
ito
r
e
d
v
ia
th
e
An
d
r
o
i
d
Pro
f
iler
,
h
ig
h
lig
h
tin
g
k
e
y
m
et
r
ics
s
u
ch
as
m
ax
im
u
m
C
PU
an
d
R
AM
u
s
ag
e
alo
n
g
with
en
er
g
y
c
o
n
s
u
m
p
tio
n
o
v
er
tim
e.
I
t
was
f
o
u
n
d
th
at
m
o
s
t
o
f
th
e
en
er
g
y
u
s
ag
e
f
ell
in
t
o
lig
h
t
en
e
r
g
y
c
o
n
s
u
m
p
tio
n
ca
teg
o
r
y
.
T
h
er
ef
o
r
e,
th
e
en
er
g
y
u
s
ag
e
is
s
h
o
wn
as
a
p
er
ce
n
ta
g
e
r
ate
with
in
th
e
li
g
h
t
en
e
r
g
y
ca
teg
o
r
y
f
o
r
m
o
r
e
p
r
ec
is
e
co
m
p
a
r
is
o
n
.
T
h
e
m
o
n
ito
r
in
g
r
esu
lts
o
f
in
f
e
r
en
ce
s
p
ee
d
a
n
d
r
eso
u
r
ce
u
s
ag
e
ar
e
s
h
o
wn
in
T
ab
le
2
.
(
a)
(
b
)
Fig
u
r
e
5
.
Sam
p
le
o
f
(
a
)
s
m
ar
tp
h
o
n
e
im
p
lem
en
tatio
n
r
esu
lt a
n
d
(
b
)
m
o
n
ito
r
in
g
u
s
in
g
an
d
r
o
i
d
p
r
o
f
iler
T
ab
l
e
2.
C
o
m
p
a
r
is
o
n
o
f
in
f
er
e
n
ce
tim
e
an
d
r
eso
u
r
ce
u
s
ag
e
f
o
r
each
m
o
d
el
M
e
t
r
i
c
e
s
Y
O
LO
v
8
Y
O
LO
v
8
Y
O
LO
v
7
M
o
d
e
l
TFLi
t
e
t
y
p
e
S
mal
l
I
N
T8
N
a
n
o
I
N
T8
Ti
n
y
F
P
1
6
A
v
e
r
a
g
e
i
n
f
e
r
e
n
c
e
t
i
me
1
.
3
3
se
c
o
n
d
s
0
.
7
2
se
c
o
n
d
s
1
.
0
4
se
c
o
n
d
s
A
v
e
r
a
g
e
b
a
t
e
r
a
i
u
s
a
g
e
4
8
%
4
8
%
5
4
%
M
a
x
i
m
u
m
o
f
C
P
U
u
sa
g
e
3
9
%
3
5
%
4
4
%
M
a
x
i
m
u
m
o
f
R
A
M
u
s
a
g
e
1
6
6
M
B
1
3
3
M
B
2
6
3
M
B
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
5
0
2
-
4
7
52
In
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
,
Vo
l.
3
7
,
No
.
2
,
Feb
r
u
a
r
y
20
2
5
:
9
9
5
-
1
0
0
4
1002
Acc
o
r
d
in
g
to
T
a
b
le
2
,
it
co
m
p
ar
es
in
f
er
en
ce
tim
e
an
d
r
eso
u
r
ce
u
s
ag
e
ac
r
o
s
s
m
o
d
els.
YOL
Ov
8
n
a
n
o
I
NT
8
is
th
e
f
astes
t
an
d
m
o
s
t
ef
f
icien
t,
with
a
0
.
7
2
s
ec
o
n
d
s
in
f
er
e
n
ce
tim
e
an
d
th
e
lo
we
s
t
C
PU
an
d
R
AM
u
s
ag
e.
YOL
Ov
8
s
m
all
I
NT
8
is
s
lig
h
tly
s
lo
wer
b
u
t
also
r
eso
u
r
ce
-
ef
f
icien
t.
I
n
co
n
tr
ast,
Y
OL
Ov
7
t
in
y
FP
1
6
,
th
o
u
g
h
q
u
ic
k
er
th
a
n
YOL
Ov
8
s
m
all,
h
as
th
e
h
ig
h
est
b
atter
y
an
d
R
AM
u
s
ag
e,
m
ak
i
n
g
it
les
s
ef
f
icien
t
o
v
er
all.
T
h
is
s
u
g
g
ests
YOL
Ov
8
n
an
o
-
I
NT
8
is
th
e
b
est ch
o
ice
f
o
r
b
alan
cin
g
s
p
ee
d
a
n
d
r
eso
u
r
ce
ef
f
i
cien
cy
.
4.
DIS
CU
SS
I
O
N
T
h
is
s
tu
d
y
i
n
v
esti
g
ated
th
e
e
f
f
ec
tiv
en
ess
o
f
d
ep
l
o
y
in
g
p
o
th
o
le
d
etec
tio
n
m
o
d
els
o
n
d
e
v
ices
w
ith
lim
ited
co
m
p
u
tatio
n
al
r
eso
u
r
c
es,
s
u
ch
as
s
m
ar
tp
h
o
n
es.
W
h
ile
ea
r
lier
s
tu
d
ies h
av
e
ex
p
lo
r
ed
th
e
im
p
ac
t o
f
d
ee
p
lear
n
in
g
m
o
d
els lik
e
f
aster
R
-
C
NN
an
d
YOL
Ov
7
f
o
r
p
o
th
o
l
e
d
etec
tio
n
,
th
ey
h
av
e
n
o
t e
x
p
l
icitly
ad
d
r
ess
ed
th
e
o
p
tim
izatio
n
an
d
d
ep
lo
y
m
en
t
o
f
th
ese
m
o
d
els
o
n
m
o
b
ile
p
latf
o
r
m
s
with
co
n
s
tr
ain
ed
r
eso
u
r
ce
s
,
s
u
ch
as
s
m
ar
tp
h
o
n
es
[
2
1
]
.
Alth
o
u
g
h
p
r
ev
io
u
s
r
esear
ch
s
h
o
ws
th
at
th
e
YOL
Ov
8
alg
o
r
ith
m
co
u
ld
o
u
tp
er
f
o
r
m
s
o
th
er
So
T
A
m
o
d
els
lik
e
YOL
Ov
7
an
d
f
aster
R
-
C
NN,
th
eir
ap
p
licatio
n
in
m
o
b
ile
p
o
th
o
le
d
etec
tio
n
r
em
ain
s
u
n
d
er
e
x
p
lo
r
e
d
[
2
3
]
.
T
h
is
s
tu
d
y
aim
s
to
ac
h
iev
e
in
cr
e
m
en
tal
in
n
o
v
atio
n
b
y
r
etr
ai
n
in
g
an
d
ev
al
u
atin
g
lig
h
tweig
h
t
m
o
d
els
(
less
th
an
2
0
m
illi
o
n
p
a
r
am
eter
s
)
o
f
YOL
Ov
8
,
YOL
Ov
7
,
an
d
f
a
s
ter
R
-
C
NN
u
s
in
g
tr
an
s
f
er
lear
n
in
g
an
d
B
ay
esian
h
y
p
er
p
ar
am
eter
t
u
n
in
g
[
3
7
]
.
T
h
is
s
tu
d
y
’
s
k
ey
f
i
n
d
in
g
s
d
e
m
o
n
s
tr
ate
th
at
th
e
YOL
Ov
8
-
b
ased
alg
o
r
ith
m
is
th
e
m
o
s
t
ef
f
ec
tiv
e
m
o
d
el
f
o
r
p
o
th
o
le
d
etec
tio
n
,
o
u
tp
er
f
o
r
m
in
g
o
t
h
er
s
tate
-
of
-
th
e
-
ar
t
m
o
d
els
s
u
ch
as
Y
OL
Ov
7
an
d
f
aster
R
-
C
NN.
W
ith
an
AP@
0
.
5
s
co
r
e
o
f
8
0
.
2
%
an
d
an
i
n
f
er
en
c
e
tim
e
o
f
j
u
s
t
0
.
7
2
s
ec
o
n
d
s
,
YOL
Ov
8
n
an
o
I
NT
8
s
tan
d
s
o
u
t
f
o
r
b
o
th
its
p
r
ec
is
io
n
an
d
s
p
ee
d
,
alo
n
g
with
m
i
n
im
al
r
eso
u
r
ce
c
o
n
s
u
m
p
tio
n
.
T
h
e
p
r
o
p
o
s
ed
m
eth
o
d
d
em
o
n
s
tr
ated
a
s
ig
n
if
ican
tl
y
h
ig
h
er
p
r
o
p
o
r
tio
n
o
f
ac
cu
r
ate
d
etec
tio
n
s
wh
ile
m
ain
tain
in
g
lo
w
co
m
p
u
tatio
n
al
d
em
an
d
s
,
in
d
i
ca
tin
g
its
p
o
ten
tial
f
o
r
r
ea
l
-
tim
e
ap
p
licatio
n
s
o
n
r
eso
u
r
ce
-
co
n
s
tr
ain
ed
d
ev
ices
lik
e
s
m
ar
tp
h
o
n
es.
Ou
r
s
tu
d
y
s
u
g
g
est
th
at
t
h
e
YOL
Ov
8
n
an
o
I
NT
8
o
u
tp
er
f
o
r
m
s
p
r
e
v
io
u
s
m
o
d
els
u
s
ed
in
p
o
th
o
le
d
etec
tio
n
.
Fo
r
in
s
tan
ce
,
Z
e
n
g
an
d
Z
h
o
n
g
[
2
5
]
wh
ic
h
r
e
p
o
r
t
ed
an
AP@
0
.
5
s
co
r
e
o
f
5
3
.
1
%
o
n
th
e
R
DD2
0
2
2
d
ataset.
Ad
d
itio
n
ally
,
in
ter
m
s
o
f
in
f
er
en
ce
tim
e,
YOL
Ov
8
n
an
o
I
NT
8
is
f
aster
th
an
th
e
YOL
Ov
3
m
o
d
el
d
ev
elo
p
e
d
b
y
Gajjar
et
a
l.
[
1
6
]
,
w
h
ich
h
a
d
a
n
AP@
0
.
5
o
f
3
4
.
7
%
with
a
0
.
8
4
s
ec
o
n
d
s
in
f
er
en
ce
tim
e.
T
h
e
p
r
o
p
o
s
ed
m
eth
o
d
b
e
n
ef
its
f
r
o
m
h
ig
h
er
ac
c
u
r
a
cy
an
d
f
aster
i
n
f
er
en
ce
with
o
u
t
ad
v
e
r
s
ely
im
p
ac
tin
g
co
m
p
u
tatio
n
al
ef
f
icien
cy
,
m
ak
in
g
it
m
o
r
e
s
u
it
ab
le
f
o
r
d
ep
lo
y
m
en
t
o
n
d
ev
ices
with
lim
ited
p
r
o
ce
s
s
in
g
p
o
wer
.
T
h
is
s
tu
d
y
ex
p
lo
r
ed
th
e
o
p
ti
m
izatio
n
o
f
p
o
th
o
le
d
etec
tio
n
m
o
d
els
f
o
r
d
ep
l
o
y
m
en
t
o
n
s
m
ar
tp
h
o
n
es
u
n
d
er
s
tan
d
ar
d
co
n
d
itio
n
s
.
Ho
wev
er
,
f
u
r
th
e
r
an
d
in
-
d
ep
th
s
tu
d
ies
m
ay
b
e
n
ee
d
ed
to
c
o
n
f
ir
m
its
ef
f
ec
tiv
en
ess
in
v
ar
ied
en
v
ir
o
n
m
en
tal
co
n
d
itio
n
s
,
esp
ec
ially
r
eg
ar
d
i
n
g
p
er
f
o
r
m
an
ce
o
n
wet
r
o
ad
s
,
e
x
tr
em
e
lig
h
tin
g
,
o
r
d
if
f
er
en
t
r
o
a
d
tex
t
u
r
es.
Su
c
h
f
a
cto
r
s
co
u
ld
p
o
ten
tially
im
p
ac
t
t
h
e
m
o
d
el
’
s
d
etec
tio
n
ac
cu
r
ac
y
an
d
g
en
er
aliza
b
ilit
y
ac
r
o
s
s
d
iv
e
r
s
e
r
ea
l
-
wo
r
ld
s
ce
n
ar
i
o
s
.
Ou
r
s
tu
d
y
d
em
o
n
s
tr
ates
th
at
m
o
d
els
o
p
tim
ized
f
o
r
m
o
b
i
le
d
ev
ices,
lik
e
YOL
Ov
8
n
a
n
o
I
NT
8
,
ar
e
m
o
r
e
ef
f
icie
n
t
an
d
p
r
ac
tical
f
o
r
r
ea
l
-
wo
r
l
d
ap
p
licatio
n
s
th
an
h
ea
v
ier
m
o
d
els.
Fu
tu
r
e
s
tu
d
ies
m
ay
ex
p
lo
r
e
ex
p
an
d
i
n
g
th
e
d
ataset
to
in
clu
d
e
d
iv
er
s
e
en
v
ir
o
n
m
en
tal
s
ce
n
ar
io
s
an
d
in
v
esti
g
ate
f
ea
s
ib
le
way
s
o
f
en
h
an
cin
g
m
o
d
el
r
o
b
u
s
tn
ess
.
Ad
d
itio
n
ally
,
ex
p
lo
r
in
g
ad
v
a
n
ce
d
o
p
tim
i
za
tio
n
tech
n
iq
u
es
an
d
ad
ap
tiv
e
alg
o
r
ith
m
s
co
u
ld
f
u
r
th
er
im
p
r
o
v
e
d
etec
tio
n
ac
c
u
r
ac
y
with
o
u
t c
o
m
p
r
o
m
is
in
g
r
eso
u
r
ce
ef
f
icien
c
y
.
R
ec
en
t
o
b
s
er
v
atio
n
s
s
u
g
g
est
t
h
at
d
ep
lo
y
in
g
o
p
tim
ized
p
o
t
h
o
le
d
etec
tio
n
m
o
d
els
o
n
s
m
ar
tp
h
o
n
es
is
b
o
th
f
ea
s
ib
le
an
d
e
f
f
ec
tiv
e.
O
u
r
f
in
d
i
n
g
s
p
r
o
v
id
e
co
n
clu
s
iv
e
ev
i
d
en
ce
th
at
th
e
YOL
Ov
8
n
an
o
I
NT
8
m
o
d
e
l
s
ig
n
if
ican
tly
im
p
r
o
v
es
d
etec
t
io
n
ac
c
u
r
ac
y
an
d
s
p
ee
d
with
o
u
t
r
e
q
u
ir
in
g
ele
v
ated
c
o
m
p
u
tatio
n
al
r
eso
u
r
ce
s
.
T
h
is
ad
v
an
ce
m
e
n
t
co
n
tr
ib
u
tes
to
r
o
a
d
s
af
ety
a
n
d
i
n
f
r
astru
ct
u
r
e
m
ain
ten
a
n
ce
b
y
e
n
ab
lin
g
r
ea
l
-
tim
e,
ac
ce
s
s
ib
le
p
o
th
o
le
d
etec
tio
n
th
r
o
u
g
h
wid
ely
av
ailab
le
m
o
b
ile
tech
n
o
lo
g
y
lik
e
s
m
ar
tp
h
o
n
e
.
5.
CO
NCLU
SI
O
N
T
h
is
s
tu
d
y
h
as
s
h
o
wn
th
at
YO
L
Ov
8
n
an
o
I
NT
8
is
th
e
m
o
s
t
ef
f
ec
tiv
e
m
o
d
el
f
o
r
p
o
th
o
le
d
e
tectio
n
o
n
s
m
ar
tp
h
o
n
es,
o
u
tp
er
f
o
r
m
in
g
o
th
er
m
o
d
els
lik
e
YOL
Ov
7
a
n
d
f
aster
R
-
C
NN
in
b
o
th
d
et
ec
tio
n
p
er
f
o
r
m
an
ce
an
d
r
eso
u
r
ce
ef
f
icien
c
y
.
I
t
h
as
s
u
cc
ess
f
u
lly
d
em
o
n
s
tr
ates
th
e
f
ea
s
ib
ilit
y
o
f
d
e
p
lo
y
in
g
p
o
t
h
o
le
d
etec
tio
n
u
s
in
g
d
ee
p
lear
n
in
g
o
n
r
eso
u
r
ce
-
co
n
s
tr
ain
ed
d
ev
ices lik
e
s
m
ar
tp
h
o
n
e,
m
ak
in
g
it m
o
r
e
ac
ce
s
s
ib
le
an
d
p
r
ac
tical.
Desp
ite
th
ese
ac
h
iev
em
en
ts
,
th
e
s
tu
d
y
h
as
lim
itatio
n
s
,
in
clu
d
in
g
p
er
f
o
r
m
a
n
ce
v
ar
ia
b
ilit
y
u
n
d
er
d
if
f
er
en
t
en
v
ir
o
n
m
e
n
tal
co
n
d
itio
n
s
an
d
a
p
r
im
ar
y
f
o
c
u
s
o
n
p
o
th
o
le
d
etec
tio
n
.
T
h
er
e
is
r
o
o
m
f
o
r
f
u
r
th
er
im
p
r
o
v
em
e
n
t in
s
m
ar
tp
h
o
n
e
p
r
o
ce
s
s
in
g
,
p
ar
ticu
lar
ly
in
m
o
d
el
o
p
tim
izatio
n
f
o
r
f
aster
in
f
e
r
en
ce
s
p
ee
d
s
.
T
h
e
im
p
licatio
n
s
o
f
th
is
r
esear
ch
ar
e
s
ig
n
if
ican
t
f
o
r
en
h
an
ci
n
g
r
o
ad
s
af
ety
an
d
u
r
b
a
n
tr
a
n
s
p
o
r
tatio
n
s
y
s
tem
s
.
B
y
in
teg
r
atin
g
s
m
ar
t
p
h
o
n
e
-
b
ased
d
etec
tio
n
s
y
s
tem
s
in
to
ex
is
tin
g
in
f
r
astru
ctu
r
e,
cities
ca
n
d
ev
elo
p
m
o
r
e
in
tellig
en
t
tr
an
s
p
o
r
tatio
n
p
o
licies,
h
elp
in
g
to
p
r
e
v
en
t
a
cc
id
en
ts
an
d
p
r
o
tect
b
o
t
h
d
r
iv
er
s
an
d
p
ed
estrian
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
I
SS
N:
2
5
0
2
-
4
7
52
C
o
mp
a
r
a
tive
s
tu
d
y
o
f p
o
th
o
le
d
etec
tio
n
u
s
in
g
d
ee
p
lea
r
n
in
g
o
n
s
ma
r
tp
h
o
n
e
(
A
ch
ya
r
Ulu
l
A
mri
)
1003
Fu
tu
r
e
wo
r
k
s
h
o
u
ld
f
o
cu
s
o
n
im
p
r
o
v
in
g
m
o
d
el
r
o
b
u
s
tn
ess
u
n
d
er
d
iv
e
r
s
e
co
n
d
itio
n
s
an
d
ex
p
lo
r
in
g
ad
d
itio
n
al
ap
p
licatio
n
s
s
u
ch
as
s
m
ar
t
an
d
d
ash
ca
m
,
an
d
h
az
a
r
d
d
etec
t
io
n
.
T
o
f
u
r
th
er
ad
v
an
ce
i
n
tellig
en
t
tr
an
s
p
o
r
tatio
n
s
y
s
tem
s
an
d
s
af
er
u
r
b
a
n
en
v
ir
o
n
m
en
ts
.
RE
F
E
R
E
NC
E
S
[
1
]
W
H
O
,
“
G
e
n
e
v
a
:
g
l
o
b
a
l
st
a
t
u
s
r
e
p
o
r
t
o
n
r
o
a
d
saf
e
t
y
2
0
2
3
,
”
2
0
2
3
.
h
t
t
p
s:
/
/
w
w
w
.
w
h
o
.
i
n
t
/
t
e
a
ms
/
so
c
i
a
l
-
d
e
t
e
r
mi
n
a
n
t
s
-
of
-
h
e
a
l
t
h
/
safe
t
y
-
a
n
d
-
mo
b
i
l
i
t
y
/
g
l
o
b
a
l
-
s
t
a
t
u
s
-
r
e
p
o
r
t
-
on
-
r
o
a
d
-
safe
t
y
-
2
0
2
3
(
a
c
c
e
sse
d
A
u
g
.
2
7
,
2
0
2
4
)
.
[
2
]
G
.
R
i
t
z
e
r
,
“
B
l
a
c
k
w
e
l
l
e
n
c
y
c
l
o
p
e
d
i
a
o
f
so
c
i
o
l
o
g
y
,
”
i
n
T
h
e
B
l
a
c
k
w
e
l
l
E
n
c
y
c
l
o
p
e
d
i
a
o
f
S
o
c
i
o
l
o
g
y
,
G
.
R
i
t
z
e
r
,
Ed
.
O
x
f
o
r
d
,
U
K
,
M
a
l
d
e
n
,
U
S
A
a
n
d
C
a
r
l
t
o
n
,
A
u
s
t
r
a
l
i
a
:
B
l
a
c
k
w
e
l
l
P
u
b
l
i
s
h
i
n
g
L
t
d
,
2
0
0
7
,
p
p
.
v
i
i
–
v
i
i
.
[
3
]
R
.
B
i
b
i
e
t
a
l
.
,
“
E
d
g
e
A
I
-
b
a
se
d
a
u
t
o
ma
t
e
d
d
e
t
e
c
t
i
o
n
a
n
d
c
l
a
ssi
f
i
c
a
t
i
o
n
o
f
r
o
a
d
a
n
o
ma
l
i
e
s
i
n
V
A
N
E
T
u
si
n
g
d
e
e
p
l
e
a
r
n
i
n
g
,
”
C
o
m
p
u
t
a
t
i
o
n
a
l
I
n
t
e
l
l
i
g
e
n
c
e
a
n
d
N
e
u
r
o
sc
i
e
n
c
e
,
v
o
l
.
2
0
2
1
,
n
o
.
1
,
J
a
n
.
2
0
2
1
,
d
o
i
:
1
0
.
1
1
5
5
/
2
0
2
1
/
6
2
6
2
1
9
4
.
[
4
]
B
.
B
u
č
k
o
,
E.
L
i
e
s
k
o
v
s
k
á
,
K
.
Z
á
b
o
v
sk
á
,
a
n
d
M
.
Z
á
b
o
v
sk
ý
,
“
C
o
m
p
u
t
e
r
v
i
si
o
n
b
a
s
e
d
p
o
t
h
o
l
e
d
e
t
e
c
t
i
o
n
u
n
d
e
r
c
h
a
l
l
e
n
g
i
n
g
c
o
n
d
i
t
i
o
n
s
,
”
S
e
n
s
o
rs
,
v
o
l
.
2
2
,
n
o
.
2
2
,
p
.
8
8
7
8
,
N
o
v
.
2
0
2
2
,
d
o
i
:
1
0
.
3
3
9
0
/
s
2
2
2
2
8
8
7
8
.
[
5
]
R
.
R
a
j
,
B
.
K
.
V
a
d
a
p
u
,
R
.
N
a
r
e
s
h
,
S
.
G
u
n
d
u
,
a
n
d
R
.
R
.
L
e
k
sh
m
i
,
“
A
n
a
r
t
i
f
i
c
i
a
l
i
n
t
e
l
l
i
g
e
n
c
e
b
a
se
d
a
ss
i
st
a
n
t
t
o
d
e
t
e
c
t
,
n
o
t
i
f
y
a
n
d
u
p
d
a
t
e
p
o
t
h
o
l
e
s
,
”
i
n
2
0
2
4
I
EEE
I
n
t
e
rn
a
t
i
o
n
a
l
C
o
n
f
e
r
e
n
c
e
o
n
I
n
t
e
r
d
i
sc
i
p
l
i
n
a
r
y
A
p
p
r
o
a
c
h
e
s
i
n
T
e
c
h
n
o
l
o
g
y
a
n
d
Ma
n
a
g
e
m
e
n
t
f
o
r
S
o
c
i
a
l
I
n
n
o
v
a
t
i
o
n
,
I
ATM
S
I
2
0
2
4
,
M
a
r
.
2
0
2
4
,
p
p
.
1
–
6
,
d
o
i
:
1
0
.
1
1
0
9
/
I
A
TM
S
I
6
0
4
2
6
.
2
0
2
4
.
1
0
5
0
3
1
2
3
.
[
6
]
S
.
N
i
e
n
a
b
e
r
,
M
.
J.
(
Th
i
n
u
s)
B
o
o
y
se
n
,
a
n
d
R
.
K
r
o
o
n
,
“
D
e
t
e
c
t
i
n
g
p
o
t
h
o
l
e
s
u
si
n
g
s
i
mp
l
e
i
ma
g
e
p
r
o
c
e
ssi
n
g
t
e
c
h
n
i
q
u
e
s
a
n
d
r
e
a
l
-
w
o
r
l
d
f
o
o
t
a
g
e
,
”
Pro
c
e
e
d
i
n
g
s
o
f
t
h
e
3
4
t
h
S
o
u
t
h
e
r
n
Af
r
i
c
a
n
T
r
a
n
sp
o
rt
C
o
n
f
e
re
n
c
e
(
S
ATC
2
0
1
5
)
,
p
p
.
1
5
3
–
1
6
4
,
2
0
1
5
,
d
o
i
:
1
0
.
1
3
1
4
0
/
R
G
.
2
.
1
.
3
1
2
1
.
8
4
0
8
.
[
7
]
M
.
R
a
t
h
e
e
,
B
.
B
a
č
i
ć
,
a
n
d
M
.
D
o
b
o
r
j
e
h
,
“
A
u
t
o
m
a
t
e
d
r
o
a
d
d
e
f
e
c
t
a
n
d
a
n
o
mal
y
d
e
t
e
c
t
i
o
n
f
o
r
t
r
a
f
f
i
c
s
a
f
e
t
y
:
a
s
y
st
e
ma
t
i
c
r
e
v
i
e
w
,
”
S
e
n
so
rs
,
v
o
l
.
2
3
,
n
o
.
1
2
,
p
.
5
6
5
6
,
J
u
n
.
2
0
2
3
,
d
o
i
:
1
0
.
3
3
9
0
/
s2
3
1
2
5
6
5
6
.
[
8
]
S
.
A
.
T
a
l
h
a
,
D
.
M
a
n
a
sr
e
h
,
a
n
d
M
.
D
.
N
a
z
z
a
l
,
“
Th
e
u
se
o
f
l
i
d
a
r
a
n
d
a
r
t
i
f
i
c
i
a
l
i
n
t
e
l
l
i
g
e
n
c
e
a
l
g
o
r
i
t
h
ms
f
o
r
d
e
t
e
c
t
i
o
n
a
n
d
s
i
z
e
e
st
i
mat
i
o
n
o
f
p
o
t
h
o
l
e
s
,
”
Bu
i
l
d
i
n
g
s
,
v
o
l
.
1
4
,
n
o
.
4
,
p
.
1
0
7
8
,
A
p
r
.
2
0
2
4
,
d
o
i
:
1
0
.
3
3
9
0
/
b
u
i
l
d
i
n
g
s
1
4
0
4
1
0
7
8
.
[
9
]
Y
.
M
.
K
i
m,
Y
.
G
.
K
i
m,
S
.
Y
.
S
o
n
,
S
.
Y
.
Li
m,
B
.
Y
.
C
h
o
i
,
a
n
d
D
.
H
.
C
h
o
i
,
“
R
e
v
i
e
w
o
f
r
e
c
e
n
t
a
u
t
o
ma
t
e
d
p
o
t
h
o
l
e
-
d
e
t
e
c
t
i
o
n
met
h
o
d
s,
”
A
p
p
l
i
e
d
S
c
i
e
n
c
e
s (S
w
i
t
zer
l
a
n
d
)
,
v
o
l
.
1
2
,
n
o
.
1
1
,
p
.
5
3
2
0
,
M
a
y
2
0
2
2
,
d
o
i
:
1
0
.
3
3
9
0
/
a
p
p
1
2
1
1
5
3
2
0
.
[
1
0
]
J.
J.
K
o
h
e
t
a
l
.
,
“
A
u
t
o
n
o
m
o
u
s
r
o
a
d
p
o
t
h
o
l
e
s
d
e
t
e
c
t
i
o
n
o
n
v
i
d
e
o
,
”
i
n
L
e
c
t
u
r
e
N
o
t
e
s
i
n
El
e
c
t
ri
c
a
l
E
n
g
i
n
e
e
r
i
n
g
,
v
o
l
.
4
8
1
,
2
0
1
9
,
p
p
.
1
3
7
–
1
4
3
.
[
1
1
]
P
.
A
.
C
h
i
t
a
l
e
,
K
.
Y
.
K
e
k
r
e
,
H
.
R
.
S
h
e
n
a
i
,
R
.
K
a
r
a
n
i
,
a
n
d
J
.
P
.
G
a
l
a
,
“
P
o
t
h
o
l
e
d
e
t
e
c
t
i
o
n
a
n
d
d
i
me
n
si
o
n
e
st
i
ma
t
i
o
n
s
y
s
t
e
m
u
si
n
g
d
e
e
p
l
e
a
r
n
i
n
g
(
Y
O
LO
)
a
n
d
i
m
a
g
e
p
r
o
c
e
ssi
n
g
,
”
i
n
I
n
t
e
rn
a
t
i
o
n
a
l
C
o
n
f
e
r
e
n
c
e
I
m
a
g
e
a
n
d
V
i
si
o
n
C
o
m
p
u
t
i
n
g
N
e
w
Z
e
a
l
a
n
d
,
N
o
v
.
2
0
2
0
,
v
o
l
.
2
0
2
0
-
N
o
v
e
m
b
e
r
,
p
p
.
1
–
6
,
d
o
i
:
1
0
.
1
1
0
9
/
I
V
C
N
Z5
1
5
7
9
.
2
0
2
0
.
9
2
9
0
5
4
7
.
[
1
2
]
R
.
K
a
n
n
a
n
,
C
.
J.
Ji
a
n
,
a
n
d
X
.
N
.
G
u
o
,
“
A
d
v
e
r
sa
ri
a
l
e
v
a
s
i
o
n
n
o
i
s
e
a
t
t
a
c
k
s
a
g
a
i
n
s
t
Te
n
so
r
F
l
o
w
o
b
j
e
c
t
d
e
t
e
c
t
i
o
n
A
P
I
,
”
i
n
2
0
2
0
1
5
t
h
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
r
e
n
c
e
f
o
r
I
n
t
e
rn
e
t
T
e
c
h
n
o
l
o
g
y
a
n
d
S
e
c
u
r
e
d
T
r
a
n
s
a
c
t
i
o
n
s,
I
C
I
T
S
T
2
0
2
0
,
D
e
c
.
2
0
2
0
,
p
p
.
1
–
4
,
d
o
i
:
1
0
.
2
3
9
1
9
/
I
C
I
TST5
1
0
3
0
.
2
0
2
0
.
9
3
5
1
3
3
1
.
[
1
3
]
K
.
R
.
A
h
me
d
,
“
S
mart
p
o
t
h
o
l
e
d
e
t
e
c
t
i
o
n
u
s
i
n
g
d
e
e
p
l
e
a
r
n
i
n
g
b
a
se
d
o
n
d
i
l
a
t
e
d
c
o
n
v
o
l
u
t
i
o
n
,
”
S
e
n
s
o
rs
,
v
o
l
.
2
1
,
n
o
.
2
4
,
p
.
8
4
0
6
,
D
e
c
.
2
0
2
1
,
d
o
i
:
1
0
.
3
3
9
0
/
s2
1
2
4
8
4
0
6
.
[
1
4
]
J.
J.
Y
e
b
e
s
,
D
.
M
o
n
t
e
r
o
,
a
n
d
I
.
A
r
r
i
o
l
a
,
“
Le
a
r
n
i
n
g
t
o
a
u
t
o
ma
t
i
c
a
l
l
y
c
a
t
c
h
p
o
t
h
o
l
e
s
i
n
w
o
r
l
d
w
i
d
e
r
o
a
d
sce
n
e
i
m
a
g
e
s
,
”
I
EEE
I
n
t
e
l
l
i
g
e
n
t
T
r
a
n
sp
o
rt
a
t
i
o
n
S
y
st
e
m
s M
a
g
a
zi
n
e
,
v
o
l
.
1
3
,
n
o
.
3
,
p
p
.
1
9
2
–
2
0
5
,
2
0
2
1
,
d
o
i
:
1
0
.
1
1
0
9
/
M
I
TS.2
0
1
9
.
2
9
2
6
3
7
0
.
[
1
5
]
D
.
H
.
H
e
o
,
J.
Y
.
C
h
o
i
,
S
.
B
.
K
i
m,
T.
O
.
Ta
k
,
a
n
d
S
.
P
.
Z
h
a
n
g
,
“
I
mag
e
-
b
a
s
e
d
p
o
t
h
o
l
e
d
e
t
e
c
t
i
o
n
u
si
n
g
m
u
l
t
i
-
sc
a
l
e
f
e
a
t
u
r
e
n
e
t
w
o
r
k
a
n
d
r
i
s
k
a
ssess
me
n
t
,
”
E
l
e
c
t
r
o
n
i
c
s (S
w
i
t
zer
l
a
n
d
)
,
v
o
l
.
1
2
,
n
o
.
4
,
p
.
8
2
6
,
F
e
b
.
2
0
2
3
,
d
o
i
:
1
0
.
3
3
9
0
/
e
l
e
c
t
r
o
n
i
c
s
1
2
0
4
0
8
2
6
.
[
1
6
]
K
.
G
a
j
j
a
r
,
T.
V
.
N
i
e
k
e
r
k
,
T.
W
i
l
m,
a
n
d
P
.
M
e
r
c
o
r
e
l
l
i
,
“
V
i
si
o
n
-
b
a
s
e
d
d
e
e
p
l
e
a
r
n
i
n
g
a
l
g
o
r
i
t
h
m
f
o
r
d
e
t
e
c
t
i
n
g
p
o
t
h
o
l
e
s,”
J
o
u
rn
a
l
o
f
Ph
y
s
i
c
s:
C
o
n
f
e
r
e
n
c
e
S
e
ri
e
s
,
v
o
l
.
2
1
6
2
,
n
o
.
1
,
p
.
0
1
2
0
1
9
,
Ja
n
.
2
0
2
2
,
d
o
i
:
1
0
.
1
0
8
8
/
1
7
4
2
-
6
5
9
6
/
2
1
6
2
/
1
/
0
1
2
0
1
9
.
[
1
7
]
B
.
M
.
P
r
a
k
a
s
h
a
n
d
K
.
C
.
S
r
i
h
a
r
i
p
r
i
y
a
,
“
E
n
h
a
n
c
e
d
p
o
t
h
o
l
e
d
e
t
e
c
t
i
o
n
s
y
s
t
e
m
u
s
i
n
g
Y
O
LO
X
a
l
g
o
r
i
t
h
m,
”
A
u
t
o
n
o
m
o
u
s
I
n
t
e
l
l
i
g
e
n
t
S
y
s
t
e
m
s
,
v
o
l
.
2
,
n
o
.
1
,
p
.
2
2
,
A
u
g
.
2
0
2
2
,
d
o
i
:
1
0
.
1
0
0
7
/
s4
3
6
8
4
-
0
2
2
-
0
0
0
3
7
-
z.
[
1
8
]
M
.
H
.
A
sa
d
,
S
.
K
h
a
l
i
q
,
M
.
H
.
Y
o
u
saf,
M
.
O
.
U
l
l
a
h
,
a
n
d
A
.
A
h
ma
d
,
“
P
o
t
h
o
l
e
d
e
t
e
c
t
i
o
n
u
si
n
g
d
e
e
p
l
e
a
r
n
i
n
g
:
a
r
e
a
l
-
tim
e
a
n
d
A
I
-
on
-
t
h
e
-
e
d
g
e
p
e
r
sp
e
c
t
i
v
e
,
”
Ad
v
a
n
c
e
s
i
n
C
i
v
i
l
E
n
g
i
n
e
e
r
i
n
g
,
v
o
l
.
2
0
2
2
,
n
o
.
1
,
Ja
n
.
2
0
2
2
,
d
o
i
:
1
0
.
1
1
5
5
/
2
0
2
2
/
9
2
2
1
2
1
1
.
[
1
9
]
K
.
G
a
y
a
t
h
r
i
a
n
d
S
.
T
h
a
n
g
a
v
e
l
u
,
“
N
o
v
e
l
d
e
e
p
l
e
a
r
n
i
n
g
m
o
d
e
l
f
o
r
v
e
h
i
c
l
e
a
n
d
p
o
t
h
o
l
e
d
e
t
e
c
t
i
o
n
,
”
I
n
d
o
n
e
s
i
a
n
J
o
u
r
n
a
l
o
f
El
e
c
t
ri
c
a
l
En
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
E
EC
S
)
,
v
o
l
.
2
3
,
n
o
.
3
,
p
p
.
1
5
7
6
–
1
5
8
2
,
S
e
p
.
2
0
2
1
,
d
o
i
:
1
0
.
1
1
5
9
1
/
i
j
e
e
c
s
.
v
2
3
.
i
3
.
p
p
1
5
7
6
-
1
5
8
2
.
[
2
0
]
C
.
K
.
C
h
i
u
,
J
.
C
.
L
i
u
,
Y
.
W
.
C
h
a
n
,
a
n
d
C
.
T
.
Y
a
n
g
,
“
P
a
v
e
me
n
t
d
i
st
r
e
ss
d
e
t
e
c
t
i
o
n
u
s
i
n
g
Y
O
LO
a
n
d
f
a
st
e
r
R
C
N
N
o
n
e
d
g
e
d
e
v
i
c
e
s,
”
i
n
L
e
c
t
u
r
e
N
o
t
e
s
i
n
E
l
e
c
t
r
i
c
a
l
En
g
i
n
e
e
ri
n
g
,
v
o
l
.
1
1
3
4
,
2
0
2
4
,
p
p
.
2
4
6
–
2
5
2
.
[
2
1
]
M
.
O
mar
a
n
d
P
.
K
u
mar
,
“
P
D
-
I
TS:
p
o
t
h
o
l
e
d
e
t
e
c
t
i
o
n
u
s
i
n
g
Y
O
LO
v
a
r
i
a
n
t
s
f
o
r
i
n
t
e
l
l
i
g
e
n
t
t
r
a
n
s
p
o
r
t
s
y
st
e
m
,
”
S
N
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
,
v
o
l
.
5
,
n
o
.
5
,
p
.
5
5
2
,
M
a
y
2
0
2
4
,
d
o
i
:
1
0
.
1
0
0
7
/
s
4
2
9
7
9
-
024
-
0
2
8
8
7
-
1.
[
2
2
]
C
.
C
.
H
si
e
h
,
H
.
W
.
J
i
a
,
W
.
H
.
H
u
a
n
g
,
a
n
d
M
.
H
.
H
si
h
,
“
D
e
e
p
l
e
a
r
n
i
n
g
-
b
a
s
e
d
r
o
a
d
p
a
v
e
m
e
n
t
i
n
sp
e
c
t
i
o
n
b
y
i
n
t
e
g
r
a
t
i
n
g
v
i
su
a
l
i
n
f
o
r
mat
i
o
n
a
n
d
I
M
U
,
”
I
n
f
o
rm
a
t
i
o
n
(
S
w
i
t
z
e
rl
a
n
d
)
,
v
o
l
.
1
5
,
n
o
.
4
,
p
.
2
3
9
,
A
p
r
.
2
0
2
4
,
d
o
i
:
1
0
.
3
3
9
0
/
i
n
f
o
1
5
0
4
0
2
3
9
.
[
2
3
]
M
.
D
i
v
y
a
,
G
.
D
i
v
y
a
s
h
r
e
e
,
a
n
d
B
.
U
.
M
a
h
e
sw
a
r
i
,
“
P
o
t
h
o
l
e
d
e
t
e
c
t
i
o
n
u
s
i
n
g
Y
O
LO
v
8
a
n
d
C
N
N
,
”
i
n
2
0
2
4
3
r
d
I
n
t
e
rn
a
t
i
o
n
a
l
C
o
n
f
e
r
e
n
c
e
f
o
r
I
n
n
o
v
a
t
i
o
n
i
n
T
e
c
h
n
o
l
o
g
y
,
I
N
O
C
O
N
2
0
2
4
,
M
a
r
.
2
0
2
4
,
p
p
.
1
–
6
,
d
o
i
:
1
0
.
1
1
0
9
/
I
N
O
C
O
N
6
0
7
5
4
.
2
0
2
4
.
1
0
5
1
2
0
0
4
.
[
2
4
]
Y
.
M
a
t
o
u
q
,
D
.
M
a
n
a
sr
e
h
,
a
n
d
M
.
D
.
N
a
z
z
a
l
,
“
A
I
-
d
r
i
v
e
n
a
p
p
r
o
a
c
h
f
o
r
a
u
t
o
mat
e
d
r
e
a
l
-
t
i
m
e
p
o
t
h
o
l
e
d
e
t
e
c
t
i
o
n
,
l
o
c
a
l
i
z
a
t
i
o
n
,
a
n
d
a
r
e
a
e
st
i
mat
i
o
n
,
”
T
r
a
n
sp
o
r
t
a
t
i
o
n
Re
se
a
rc
h
Re
c
o
r
d
,
M
a
y
2
0
2
4
,
d
o
i
:
1
0
.
1
1
7
7
/
0
3
6
1
1
9
8
1
2
4
1
2
4
6
9
9
3
.
[
2
5
]
J.
Ze
n
g
a
n
d
H
.
Z
h
o
n
g
,
“
Y
O
LO
v
8
-
P
D
:
a
n
i
m
p
r
o
v
e
d
r
o
a
d
d
a
m
a
g
e
d
e
t
e
c
t
i
o
n
a
l
g
o
r
i
t
h
m
b
a
se
d
o
n
Y
O
LO
v
8
n
mo
d
e
l
,
”
S
c
i
e
n
t
i
f
i
c
Re
p
o
r
t
s
,
v
o
l
.
1
4
,
n
o
.
1
,
p
.
1
2
0
5
2
,
M
a
y
2
0
2
4
,
d
o
i
:
1
0
.
1
0
3
8
/
s
4
1
5
9
8
-
0
2
4
-
6
2
9
3
3
-
z.
[
2
6
]
I
.
M
a
r
t
i
n
e
z
-
A
l
p
i
s
t
e
,
G
.
G
o
l
c
a
r
e
n
a
r
e
n
j
i
,
Q
.
W
a
n
g
,
a
n
d
J
.
M
.
A
l
c
a
r
a
z
-
C
a
l
e
r
o
,
“
S
m
a
r
t
p
h
o
n
e
-
b
a
se
d
r
e
a
l
-
t
i
m
e
o
b
j
e
c
t
r
e
c
o
g
n
i
t
i
o
n
a
r
c
h
i
t
e
c
t
u
r
e
f
o
r
p
o
r
t
a
b
l
e
a
n
d
c
o
n
s
t
r
a
i
n
e
d
sy
s
t
e
ms
,
”
J
o
u
rn
a
l
o
f
Re
a
l
-
T
i
m
e
I
m
a
g
e
Pro
c
e
ss
i
n
g
,
v
o
l
.
1
9
,
n
o
.
1
,
p
p
.
1
0
3
–
1
1
5
,
F
e
b
.
2
0
2
2
,
d
o
i
:
1
0
.
1
0
0
7
/
s1
1
5
5
4
-
0
2
1
-
0
1
1
6
4
-
1.
[
2
7
]
D
.
D
l
u
ž
n
e
v
sk
i
j
,
P
.
S
t
e
f
a
n
o
v
i
č
,
a
n
d
S
.
R
a
ma
n
a
u
s
k
a
i
t
ė
,
“
I
n
v
e
s
t
i
g
a
t
i
o
n
o
f
Y
O
LO
v
5
e
f
f
i
c
i
e
n
c
y
i
n
i
P
h
o
n
e
su
p
p
o
r
t
e
d
s
y
st
e
ms,
”
Ba
l
t
i
c
J
o
u
r
n
a
l
o
f
M
o
d
e
rn
C
o
m
p
u
t
i
n
g
,
v
o
l
.
9
,
n
o
.
3
,
p
p
.
3
3
3
–
3
4
4
,
2
0
2
1
,
d
o
i
:
1
0
.
2
2
3
6
4
/
b
j
mc.
2
0
2
1
.
9
.
3
.
0
7
.
[
2
8
]
A
.
F
a
k
h
r
o
u
,
J
.
K
u
n
h
o
t
h
,
a
n
d
S
.
A
l
M
a
a
d
e
e
d
,
“
S
m
a
r
t
p
h
o
n
e
-
b
a
se
d
f
o
o
d
r
e
c
o
g
n
i
t
i
o
n
sy
s
t
e
m
u
si
n
g
mu
l
t
i
p
l
e
d
e
e
p
C
N
N
mo
d
e
l
s,”
Mu
l
t
i
m
e
d
i
a
T
o
o
l
s
a
n
d
Ap
p
l
i
c
a
t
i
o
n
s
,
v
o
l
.
8
0
,
n
o
.
2
1
–
2
3
,
p
p
.
3
3
0
1
1
–
3
3
0
3
2
,
S
e
p
.
2
0
2
1
,
d
o
i
:
1
0
.
1
0
0
7
/
s1
1
0
4
2
-
021
-
1
1
3
2
9
-
6.
[
2
9
]
M
.
A
.
G
e
n
a
e
v
e
t
a
l
.
,
“
C
l
a
ss
i
f
i
c
a
t
i
o
n
o
f
f
r
u
i
t
f
l
i
e
s
b
y
g
e
n
d
e
r
i
n
i
m
a
g
e
s
u
s
i
n
g
sm
a
r
t
p
h
o
n
e
s
a
n
d
t
h
e
Y
O
LO
v
4
-
t
i
n
y
n
e
u
r
a
l
n
e
t
w
o
r
k
,
”
Ma
t
h
e
m
a
t
i
c
s
,
v
o
l
.
1
0
,
n
o
.
3
,
p
.
2
9
5
,
Ja
n
.
2
0
2
2
,
d
o
i
:
1
0
.
3
3
9
0
/
m
a
t
h
1
0
0
3
0
2
9
5
.
[
3
0
]
S
.
K
h
a
l
i
d
,
H
.
M
.
O
q
a
i
b
i
,
M
.
A
q
i
b
,
a
n
d
Y
.
H
a
f
e
e
z
,
“
S
ma
l
l
p
e
st
s
d
e
t
e
c
t
i
o
n
i
n
f
i
e
l
d
c
r
o
p
s
u
si
n
g
d
e
e
p
l
e
a
r
n
i
n
g
o
b
j
e
c
t
d
e
t
e
c
t
i
o
n
,
”
S
u
s
t
a
i
n
a
b
i
l
i
t
y
(
S
w
i
t
zer
l
a
n
d
)
,
v
o
l
.
1
5
,
n
o
.
8
,
p
.
6
8
1
5
,
A
p
r
.
2
0
2
3
,
d
o
i
:
1
0
.
3
3
9
0
/
s
u
1
5
0
8
6
8
1
5
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
5
0
2
-
4
7
52
In
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
,
Vo
l.
3
7
,
No
.
2
,
Feb
r
u
a
r
y
20
2
5
:
9
9
5
-
1
0
0
4
1004
[
3
1
]
E.
A
ss
u
n
ç
ã
o
e
t
a
l
.
,
“
R
e
a
l
-
t
i
me
i
ma
g
e
d
e
t
e
c
t
i
o
n
f
o
r
e
d
g
e
d
e
v
i
c
e
s
:
a
p
e
a
c
h
f
r
u
i
t
d
e
t
e
c
t
i
o
n
a
p
p
l
i
c
a
t
i
o
n
,
”
F
u
t
u
re
I
n
t
e
rn
e
t
,
v
o
l
.
1
4
,
n
o
.
1
1
,
p
.
3
2
3
,
N
o
v
.
2
0
2
2
,
d
o
i
:
1
0
.
3
3
9
0
/
f
i
1
4
1
1
0
3
2
3
.
[
3
2
]
R
.
L
i
,
Z
.
Ji
,
S
.
H
u
,
X
.
H
u
a
n
g
,
J
.
Y
a
n
g
,
a
n
d
W
.
L
i
,
“
To
mat
o
m
a
t
u
r
i
t
y
r
e
c
o
g
n
i
t
i
o
n
m
o
d
e
l
b
a
s
e
d
o
n
i
m
p
r
o
v
e
d
Y
O
LO
v
5
i
n
G
r
e
e
n
h
o
u
s
e
,
”
A
g
ro
n
o
m
y
,
v
o
l
.
1
3
,
n
o
.
2
,
p
.
6
0
3
,
F
e
b
.
2
0
2
3
,
d
o
i
:
1
0
.
3
3
9
0
/
a
g
r
o
n
o
m
y
1
3
0
2
0
6
0
3
.
[
3
3
]
S
.
K
o
n
g
,
J.
Li
,
Y
.
Z
h
a
i
,
Z
.
G
a
o
,
Y
.
Zh
o
u
,
a
n
d
Y
.
X
u
,
“
R
e
a
l
-
t
i
me
d
e
t
e
c
t
i
o
n
o
f
c
r
o
p
s
w
i
t
h
d
e
n
s
e
p
l
a
n
t
i
n
g
u
si
n
g
d
e
e
p
l
e
a
r
n
i
n
g
a
t
see
d
l
i
n
g
s
t
a
g
e
,
”
Ag
r
o
n
o
m
y
,
v
o
l
.
1
3
,
n
o
.
6
,
p
.
1
5
0
3
,
M
a
y
2
0
2
3
,
d
o
i
:
1
0
.
3
3
9
0
/
a
g
r
o
n
o
m
y
1
3
0
6
1
5
0
3
.
[
3
4
]
Y
.
L
i
u
e
t
a
l
.
,
“
A
n
I
mp
r
o
v
e
d
T
u
n
a
-
Y
O
LO
M
o
d
e
l
B
a
se
d
o
n
Y
O
LO
v
3
f
o
r
r
e
a
l
-
t
i
m
e
t
u
n
a
d
e
t
e
c
t
i
o
n
c
o
n
s
i
d
e
r
i
n
g
l
i
g
h
t
w
e
i
g
h
t
d
e
p
l
o
y
me
n
t
,
”
J
o
u
r
n
a
l
o
f
Ma
r
i
n
e
S
c
i
e
n
c
e
a
n
d
E
n
g
i
n
e
e
ri
n
g
,
v
o
l
.
1
1
,
n
o
.
3
,
p
.
5
4
2
,
M
a
r
.
2
0
2
3
,
d
o
i
:
1
0
.
3
3
9
0
/
j
mse
1
1
0
3
0
5
4
2
.
[
3
5
]
K
.
Z
h
a
o
,
L.
Zh
a
o
,
Y
.
Z
h
a
o
,
a
n
d
H
.
D
e
n
g
,
“
S
t
u
d
y
o
n
l
i
g
h
t
w
e
i
g
h
t
m
o
d
e
l
o
f
ma
i
z
e
s
e
e
d
l
i
n
g
o
b
j
e
c
t
d
e
t
e
c
t
i
o
n
b
a
se
d
o
n
Y
O
LO
v
7
,
”
Ap
p
l
i
e
d
S
c
i
e
n
c
e
s
(
S
w
i
t
zer
l
a
n
d
)
,
v
o
l
.
1
3
,
n
o
.
1
3
,
p
.
7
7
3
1
,
J
u
n
.
2
0
2
3
,
d
o
i
:
1
0
.
3
3
9
0
/
a
p
p
1
3
1
3
7
7
3
1
.
[
3
6
]
M
.
C
o
c
c
i
a
,
“
S
o
u
r
c
e
s
o
f
d
i
sr
u
p
t
i
v
e
t
e
c
h
n
o
l
o
g
i
e
s f
o
r
i
n
d
u
s
t
r
i
a
l
c
h
a
n
g
e
,
”
I
n
d
u
st
ri
a
,
v
o
l
.
3
8
,
n
o
.
1
,
p
p
.
9
7
–
1
2
0
,
2
0
1
7
.
[
3
7
]
M
.
C
o
c
c
i
a
,
“
S
o
u
r
c
e
s
o
f
t
e
c
h
n
o
l
o
g
i
c
a
l
i
n
n
o
v
a
t
i
o
n
:
r
ad
i
c
a
l
a
n
d
i
n
c
r
e
me
n
t
a
l
i
n
n
o
v
a
t
i
o
n
p
r
o
b
l
e
m
-
d
r
i
v
e
n
t
o
su
p
p
o
r
t
c
o
m
p
e
t
i
t
i
v
e
a
d
v
a
n
t
a
g
e
o
f
f
i
r
ms,
”
T
e
c
h
n
o
l
o
g
y
An
a
l
y
si
s
a
n
d
S
t
ra
t
e
g
i
c
M
a
n
a
g
e
m
e
n
t
,
v
o
l
.
2
9
,
n
o
.
9
,
p
p
.
1
0
4
8
–
1
0
6
1
,
O
c
t
.
2
0
1
7
,
d
o
i
:
1
0
.
1
0
8
0
/
0
9
5
3
7
3
2
5
.
2
0
1
6
.
1
2
6
8
6
8
2
.
[
3
8
]
S
.
N
i
e
n
a
b
e
r
,
M
.
J
.
(
T
h
i
n
u
s)
B
o
o
y
s
e
n
,
a
n
d
S
.
K
r
o
o
n
,
“
D
a
t
a
se
t
o
f
i
ma
g
e
s
u
s
e
d
f
o
r
p
o
t
h
o
l
e
d
e
t
e
c
t
i
o
n
,
”
i
n
Pr
o
c
e
e
d
i
n
g
s
o
f
t
h
e
3
4
t
h
S
o
u
t
h
e
rn
Af
r
i
c
a
n
T
r
a
n
s
p
o
rt
C
o
n
f
e
r
e
n
c
e
(
S
ATC
2
0
1
5
)
,
2
0
1
5
,
p
p
.
1
5
3
–
1
6
4
.
[
3
9
]
X
.
Ji
a
n
g
,
A
.
H
a
d
i
d
,
Y
.
P
a
n
g
,
E.
G
r
a
n
g
e
r
,
a
n
d
X
.
F
e
n
g
,
D
e
e
p
l
e
a
r
n
i
n
g
i
n
o
b
j
e
c
t
d
e
t
e
c
t
i
o
n
a
n
d
re
c
o
g
n
i
t
i
o
n
.
S
i
n
g
a
p
o
r
e
:
S
p
r
i
n
g
e
r
S
i
n
g
a
p
o
r
e
,
2
0
1
9
.
[
4
0
]
M
.
F
r
o
mm
,
M
.
S
c
h
u
b
e
r
t
,
G
.
C
a
s
t
i
l
l
a
,
J.
Li
n
k
e
,
a
n
d
G
.
M
c
D
e
r
m
i
d
,
“
A
u
t
o
mat
e
d
d
e
t
e
c
t
i
o
n
o
f
c
o
n
i
f
e
r
s
e
e
d
l
i
n
g
s
i
n
d
r
o
n
e
i
m
a
g
e
r
y
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
s,
”
Re
m
o
t
e
S
e
n
si
n
g
,
v
o
l
.
1
1
,
n
o
.
2
1
,
p
.
2
5
8
5
,
N
o
v
.
2
0
1
9
,
d
o
i
:
1
0
.
3
3
9
0
/
r
s
1
1
2
1
2
5
8
5
.
[
4
1
]
C.
-
Y
.
W
a
n
g
,
A
.
B
o
c
h
k
o
v
sk
i
y
,
a
n
d
H
.
-
Y
.
M
.
L
i
a
o
,
“
Y
O
LO
v
7
:
t
r
a
i
n
a
b
l
e
b
a
g
-
of
-
f
r
e
e
b
i
e
s
se
t
s
n
e
w
st
a
t
e
-
of
-
t
h
e
-
a
r
t
f
o
r
r
e
a
l
-
t
i
me
o
b
j
e
c
t
d
e
t
e
c
t
o
r
s,”
i
n
2
0
2
3
I
EE
E/
C
VF
C
o
n
f
e
r
e
n
c
e
o
n
C
o
m
p
u
t
e
r
Vi
si
o
n
a
n
d
Pa
t
t
e
rn
Re
c
o
g
n
i
t
i
o
n
(
C
VPR)
,
J
u
n
.
2
0
2
3
,
p
p
.
7
4
6
4
–
7
4
7
5
,
d
o
i
:
1
0
.
1
1
0
9
/
c
v
p
r
5
2
7
2
9
.
2
0
2
3
.
0
0
7
2
1
.
[
4
2
]
G
.
Jo
c
h
e
r
,
A
.
C
h
a
u
r
a
s
i
a
,
a
n
d
J.
Q
i
u
,
“
U
l
t
r
a
l
y
t
i
c
s YO
LO
,
”
g
i
t
h
u
b
,
2
0
1
3
.
h
t
t
p
s:
/
/
g
i
t
h
u
b
.
c
o
m/
u
l
t
r
a
l
y
t
i
c
s
/
u
l
t
r
a
l
y
t
i
c
s
.
[
4
3
]
W
.
Y
.
Le
e
,
S
.
M
.
P
a
r
k
,
a
n
d
K
.
B
.
S
i
m,
“
O
p
t
i
m
a
l
h
y
p
e
r
p
a
r
a
met
e
r
t
u
n
i
n
g
o
f
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
u
r
a
l
n
e
t
w
o
r
k
s
b
a
s
e
d
o
n
t
h
e
p
a
r
a
m
e
t
e
r
-
se
t
t
i
n
g
-
f
r
e
e
h
a
r
mo
n
y
se
a
r
c
h
a
l
g
o
r
i
t
h
m,”
O
p
t
i
k
,
v
o
l
.
1
7
2
,
p
p
.
3
5
9
–
3
6
7
,
N
o
v
.
2
0
1
8
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
i
j
l
e
o
.
2
0
1
8
.
0
7
.
0
4
4
.
[
4
4
]
R
.
S
u
n
k
a
r
a
a
n
d
T.
Lu
o
,
“
N
o
mo
r
e
s
t
r
i
d
e
d
c
o
n
v
o
l
u
t
i
o
n
s
o
r
po
o
l
i
n
g
:
a
n
e
w
C
N
N
b
u
i
l
d
i
n
g
b
l
o
c
k
f
o
r
l
o
w
-
r
e
s
o
l
u
t
i
o
n
i
ma
g
e
s
a
n
d
smal
l
o
b
j
e
c
t
s,”
i
n
L
e
c
t
u
re
N
o
t
e
s
i
n
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
(
i
n
c
l
u
d
i
n
g
s
u
b
ser
i
e
s
L
e
c
t
u
r
e
N
o
t
e
s
i
n
Art
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
a
n
d
L
e
c
t
u
r
e
N
o
t
e
s
i
n
B
i
o
i
n
f
o
rm
a
t
i
c
s)
,
v
o
l
.
1
3
7
1
5
LN
A
I
,
2
0
2
3
,
p
p
.
4
4
3
–
4
5
9
.
[
4
5
]
T.
H
a
g
o
s,
“
A
n
d
r
o
i
d
st
u
d
i
o
p
r
o
f
i
l
e
r
,
”
An
d
r
o
i
d
S
t
u
d
i
o
I
D
E
Q
u
i
c
k
Re
f
e
r
e
n
c
e
,
p
p
.
7
3
–
8
2
,
2
0
1
9
,
d
o
i
:
1
0
.
1
0
0
7
/
9
7
8
-
1
-
4
8
4
2
-
4
9
5
3
-
6
_
7
.
B
I
O
G
RAP
H
I
E
S O
F
AUTH
O
RS
Ac
h
y
a
r
Ulu
l
Am
r
i
i
s
th
e
a
d
v
a
n
c
e
d
a
n
a
ly
ti
c
s
lea
d
a
t
o
n
e
o
f
I
n
d
o
n
e
sia
’
s
larg
e
s
t
c
o
n
g
l
o
m
e
ra
tes
.
Ho
ld
i
n
g
a
b
a
c
h
e
lo
r
’
s
d
e
g
re
e
in
M
a
th
e
m
a
ti
c
s
fro
m
Un
iv
e
rsitas
G
a
d
jah
M
a
d
a
,
h
e
h
a
s
e
x
ten
siv
e
e
x
p
e
rien
c
e
in
re
tail
a
n
d
a
u
t
o
m
o
ti
v
e
a
n
a
ly
ti
c
s
a
c
ro
ss
m
u
lt
ip
le
c
o
u
n
tri
e
s.
Cu
rre
n
tl
y
,
h
e
is
p
u
rsu
in
g
a
M
a
ste
r
’
s
d
e
g
re
e
in
Co
m
p
u
ter
S
c
ien
c
e
a
t
Bin
a
Nu
sa
n
tara
Un
iv
e
rsity
,
I
n
d
o
n
e
sia
,
w
h
e
re
h
e
i
s
p
a
rt
o
f
th
e
B
in
u
s
G
ra
d
u
a
te
P
ro
g
ra
m
u
n
d
e
r
t
h
e
g
u
i
d
a
n
c
e
o
f
M
r.
G
e
d
e
.
His
re
se
a
rc
h
in
tere
sts
fo
c
u
s
o
n
c
o
m
p
u
ter
sc
ien
c
e
,
p
a
rti
c
u
larly
in
m
a
c
h
in
e
lea
rn
i
n
g
,
d
e
e
p
lea
rn
in
g
,
a
n
d
c
o
m
p
u
ter
v
isio
n
.
He
c
a
n
b
e
c
o
n
tac
ted
at
e
m
a
il
:
a
c
h
y
a
r.
a
m
ri@b
in
u
s.a
c
.
id
.
G
e
d
e
Pu
tr
a
K
u
sum
a
re
c
e
iv
e
d
P
h
.
D
.
d
e
g
re
e
in
El
e
c
tri
c
a
l
a
n
d
El
e
c
tr
o
n
ic
En
g
i
n
e
e
rin
g
fr
o
m
Na
n
y
a
n
g
Tec
h
n
o
lo
g
ica
l
Un
iv
e
rsity
(NTU),
S
in
g
a
p
o
re
,
in
2
0
1
3
.
He
is
c
u
rre
n
tl
y
wo
rk
in
g
a
s
a
lec
tu
re
r
a
n
d
h
e
a
d
o
f
De
p
a
rtme
n
t
o
f
Co
m
p
u
ter
S
c
ien
c
e
,
BINU
S
G
ra
d
u
a
te
P
ro
g
ra
m
,
Bi
n
a
Nu
sa
n
t
a
ra
Un
iv
e
rsity
,
I
n
d
o
n
e
sia
.
Be
fo
re
jo
i
n
in
g
Bi
n
a
Nu
sa
n
tara
Un
iv
e
rsity
,
h
e
wa
s
wo
rk
in
g
a
s
a
re
se
a
rc
h
sc
ien
ti
st
in
I2
R
–
A*
S
TAR,
S
in
g
a
p
o
re
.
His
re
se
a
rc
h
in
tere
sts
in
c
lu
d
e
c
o
m
p
u
ter
v
isi
o
n
,
p
a
tt
e
r
n
re
c
o
g
n
it
i
o
n
,
d
e
e
p
lea
rn
in
g
,
fa
c
e
re
c
o
g
n
it
io
n
,
a
n
d
a
p
p
e
a
ra
n
c
e
-
b
a
se
d
o
b
jec
t
re
c
o
g
n
it
i
o
n
.
He
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
in
e
g
a
ra
@b
in
u
s.e
d
u
.
Evaluation Warning : The document was created with Spire.PDF for Python.