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
.
3
,
Ma
r
ch
20
2
5
,
p
p
.
1
6
9
2
~
1
701
I
SS
N:
2
502
-
4
7
52
,
DOI
: 1
0
.
1
1
5
9
1
/ijee
cs
.v
3
7
.
i
3
.
pp
1
6
9
2
-
1
7
0
1
1692
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
ee
cs
.
ia
esco
r
e.
co
m
A new
deep
learn
ing
mo
del bas
ed o
n conv
o
lutiona
l n
eura
l
network a
nd
resi
dua
l blocks
f
o
r
dr
iv
er drowsi
ness
d
etec
tion
Abdelfe
t
t
a
h So
ulta
na
,
F
a
o
uz
ia
B
ena
bb
o
u,
Na
wa
l Sa
el,
So
uk
a
ina
B
o
hs
is
s
in
La
b
o
r
a
t
o
r
y
o
f
M
o
d
e
l
i
n
g
a
n
d
I
n
f
o
r
ma
t
i
o
n
T
e
c
h
n
o
l
o
g
y
,
F
a
c
u
l
t
y
o
f
S
c
i
e
n
c
e
s
B
e
n
M
’
S
I
K
,
U
n
i
v
e
r
si
t
y
H
a
ssa
n
I
I
o
f
C
a
sa
b
l
a
n
c
a
,
C
a
sa
b
l
a
n
c
a
,
M
o
r
o
c
c
o
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
J
u
n
29
,
2
0
2
4
R
ev
is
ed
Sep
19
,
2
0
2
4
Acc
ep
ted
Oct
7
,
2
0
2
4
Re
c
o
g
n
izin
g
th
e
p
i
v
o
tal
imp
o
rta
n
c
e
o
f
m
o
n
it
o
r
in
g
d
riv
e
r
i
n
a
tt
e
n
t
io
n
i
n
t
h
e
q
u
e
st t
o
m
in
imiz
e
a
c
c
id
e
n
ts
a
n
d
e
n
h
a
n
c
e
sa
fe
ty
a
n
d
se
c
u
rit
y
,
it
is
e
ss
e
n
ti
a
l
to
h
ig
h
li
g
h
t
th
e
i
n
h
e
re
n
t
d
a
n
g
e
r
p
o
se
d
b
y
d
ro
ws
i
n
e
ss
-
a
sp
e
c
ifi
c
fo
rm
o
f
in
a
tt
e
n
ti
o
n
t
h
a
t
c
a
n
sig
n
ifi
c
a
n
tl
y
c
o
n
tri
b
u
te
t
o
a
c
c
id
e
n
ts.
To
a
d
d
re
ss
th
is
issu
e
,
se
v
e
ra
l
p
ro
p
o
si
ti
o
n
s
in
v
o
l
v
in
g
a
rti
ficia
l
i
n
telli
g
e
n
c
e
h
a
v
e
b
e
e
n
p
u
t
fo
rth
t
o
e
ffe
c
ti
v
e
ly
m
o
n
i
to
r
a
n
d
id
e
n
ti
fy
i
n
sta
n
c
e
s
o
f
d
ri
v
e
r
d
r
o
ws
in
e
ss
.
Ho
we
v
e
r,
c
h
a
ll
e
n
g
e
s
p
e
rsist
in
t
h
e
fo
rm
o
f
re
a
l
-
ti
m
e
p
ro
c
e
ss
in
g
c
o
n
stra
in
ts,
th
e
in
tr
ica
te
n
a
tu
re
o
f
m
o
d
e
l
p
a
ra
m
e
ters
,
a
n
d
th
e
re
sp
o
n
se
ti
m
e
o
f
t
h
e
m
o
d
e
l
.
Th
e
p
ro
p
o
se
d
m
e
t
h
o
d
o
l
o
g
y
fo
c
u
s
e
s
o
n
u
si
n
g
c
o
n
v
o
lu
ti
o
n
a
l
n
e
u
ra
l
n
e
two
rk
s
(CNN
s)
a
n
d
Re
sid
u
a
l
b
lo
c
k
s
a
s
ro
b
u
st
a
n
d
e
ffe
c
ti
v
e
d
e
e
p
-
lea
rn
in
g
m
o
d
e
l
s
fo
r
t
h
e
re
a
l
-
ti
m
e
d
e
tec
ti
o
n
o
f
d
r
iv
e
r
d
r
o
ws
in
e
ss
.
T
h
e
in
t
e
g
ra
ti
o
n
o
f
CNN
s
a
n
d
Re
sid
u
a
l
b
lo
c
k
s
e
n
h
a
n
c
e
s
th
e
m
o
d
e
l'
s
p
re
c
isio
n
,
str
ik
i
n
g
a
we
ll
-
b
a
lan
c
e
d
sy
n
e
rg
y
b
e
twe
e
n
c
o
m
p
u
tati
o
n
a
l
e
fficie
n
c
y
a
n
d
p
e
rfo
rm
a
n
c
e
.
No
t
a
b
ly
,
t
h
is
a
p
p
ro
a
c
h
d
e
m
o
n
stra
ted
a
n
imp
r
e
ss
iv
e
a
c
c
u
ra
c
y
o
f
9
6
.
0
9
%
a
lo
n
g
wit
h
a
re
c
a
ll
,
f1
-
sc
o
re
,
a
n
d
p
re
c
isio
n
a
l
l
a
t
9
6
%
wh
e
n
e
v
a
l
u
a
ted
o
n
th
e
p
u
b
l
icly
a
v
a
il
a
b
le
e
y
e
_
d
a
tas
e
t.
F
u
r
th
e
rm
o
re
,
th
e
r
u
n
t
ime
o
f
th
e
d
e
v
e
lo
p
e
d
m
o
d
e
l
is
a
m
e
re
7
0
m
s.
To
fu
rth
e
r
v
a
li
d
a
te
th
e
e
fficie
n
c
y
o
f
o
u
r
p
ro
p
o
se
d
m
o
d
e
l,
we
c
o
n
d
u
c
ted
a
c
o
m
p
a
ra
ti
v
e
a
n
a
ly
si
s
with
v
a
ri
o
u
s
p
re
-
train
e
d
re
si
d
u
a
l
n
e
u
ra
l
n
e
two
rk
s,
i
n
c
lu
d
in
g
Re
sN
e
t1
5
2
,
Re
sN
e
t5
0
,
Re
sN
e
t5
0
v
2
,
Re
sN
e
t1
0
1
,
a
n
d
Re
sN
e
t1
0
1
v
2
.
K
ey
w
o
r
d
s
:
C
o
m
p
u
ter
v
is
io
n
C
o
n
v
o
lu
tio
n
n
eu
r
al
n
etwo
r
k
Deep
-
lear
n
in
g
Dr
iv
er
d
r
o
wsi
n
ess
R
esid
u
al
n
eu
r
al
n
etwo
r
k
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
:
Ab
d
elf
ettah
So
u
ltan
a
L
ab
o
r
ato
r
y
o
f
Mo
d
elin
g
an
d
I
n
f
o
r
m
atio
n
T
ec
h
n
o
lo
g
y
,
Facu
l
ty
o
f
Scien
ce
s
B
en
M’
SIK
Un
iv
er
s
ity
Hass
an
I
I
o
f
C
asab
lan
ca
C
asab
lan
ca
,
Mo
r
o
cc
o
E
m
ail:
ab
d
elf
ettah
.
s
o
u
ltan
a
-
et
u
@
etu
.
u
n
iv
h
2
c
.
m
a
1.
I
NT
RO
D
UCT
I
O
N
E
v
er
y
y
ea
r
,
a
r
is
in
g
r
ate
o
f
a
cc
id
en
ts
is
o
b
s
er
v
e
d
g
lo
b
ally
.
Sev
er
al
f
ac
to
r
s
ca
n
co
n
tr
ib
u
te
to
t
h
is
o
b
s
er
v
atio
n
,
in
clu
d
in
g
n
o
n
-
c
o
m
p
lian
ce
with
tr
af
f
ic
r
eg
u
lati
o
n
s
,
tr
af
f
ic
c
o
n
g
esti
o
n
,
en
v
ir
o
n
m
en
tal
co
n
d
itio
n
s
,
an
d
n
o
ta
b
ly
,
th
e
d
r
iv
er
'
s
s
tat
e.
I
n
d
ee
d
,
th
e
d
r
iv
er
'
s
co
n
tex
t
ca
n
b
e
a
s
ig
n
if
ican
t
s
o
u
r
ce
o
f
r
o
a
d
ac
cid
en
ts
.
Facto
r
s
s
u
ch
as
in
to
x
icatio
n
,
n
eg
ativ
e
em
o
tio
n
s
,
an
d
s
tates
o
f
in
atten
tio
n
,
lik
e
d
is
tr
ac
tio
n
an
d
d
r
o
wsi
n
ess
,
lead
to
d
an
g
er
o
u
s
an
d
d
is
astro
u
s
s
itu
atio
n
s
.
I
n
tellig
en
t
tr
an
s
p
o
r
tatio
n
s
y
s
tem
s
(
I
T
S)
p
lay
a
p
iv
o
tal
r
o
le
in
m
o
d
er
n
izin
g
an
d
en
h
a
n
cin
g
r
o
ad
s
af
ety
.
T
h
ese
in
teg
r
ated
tech
n
o
lo
g
ies
lev
er
ag
e
d
ata
an
d
c
o
m
m
u
n
icati
o
n
s
y
s
tem
s
to
o
p
tim
ize
tr
af
f
ic
f
lo
w,
im
p
r
o
v
e
d
r
iv
e
r
b
eh
av
io
r
,
a
n
d
u
ltima
tely
r
ed
u
ce
ac
cid
en
ts
.
On
e
n
o
tab
le
ap
p
licati
o
n
in
v
o
lv
es
th
e
im
p
lem
en
tatio
n
o
f
ad
v
an
ce
d
d
r
iv
er
m
o
n
ito
r
in
g
s
y
s
tem
s
(
DM
S),
wh
ich
u
tili
ze
cu
ttin
g
-
ed
g
e
s
en
s
o
r
s
an
d
alg
o
r
ith
m
s
to
tr
ac
k
an
d
a
n
aly
z
e
a
d
r
iv
er
'
s
b
eh
a
v
io
r
an
d
aler
t
n
ess
.
Ad
d
itio
n
ally
,
I
T
S
i
n
co
r
p
o
r
ates
s
o
p
h
is
ticated
aler
t
s
y
s
tem
s
,
s
u
ch
as
lan
e
d
e
p
ar
tu
r
e
war
n
i
n
g
(
L
DW
)
[
1
]
a
n
d
f
o
r
war
d
co
llis
io
n
war
n
in
g
(
FC
W
)
[
2
]
,
wh
ich
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
A
n
ew d
ee
p
lea
r
n
in
g
mo
d
el
b
a
s
ed
o
n
co
n
vo
lu
tio
n
a
l n
eu
r
a
l
n
etw
o
r
k
…
(
A
b
d
elfetta
h
S
o
u
lta
n
a
)
1693
p
r
o
v
id
e
r
ea
l
-
tim
e
n
o
tific
atio
n
s
to
d
r
iv
e
r
s
ab
o
u
t
p
o
ten
tial
h
az
ar
d
s
.
Dr
o
wsi
n
ess
is
o
f
ten
u
n
d
er
s
to
o
d
to
m
ea
n
s
leep
in
ess
in
in
ap
p
r
o
p
r
iate
co
n
tex
ts
.
E
v
en
if
d
r
o
wsi
n
ess
las
ts
o
n
ly
a
f
ew
m
in
u
tes,
t
h
e
c
o
n
s
eq
u
en
ce
s
ca
n
b
e
d
ev
astatin
g
.
T
ir
ed
d
r
iv
e
r
s
ar
e
as
d
an
g
er
o
u
s
as
d
r
u
n
k
d
r
iv
e
r
s
b
ec
au
s
e
th
ey
h
av
e
s
lo
wer
r
ea
ctio
n
tim
es
an
d
s
u
f
f
er
f
r
o
m
r
ed
u
ce
d
atten
tio
n
,
awa
r
en
ess
,
an
d
a
b
ilit
y
to
co
n
tr
o
l
t
h
eir
v
e
h
icles
[
3
]
.
A
cc
o
r
d
in
g
to
th
e
US
Natio
n
al
Hig
h
way
T
r
a
f
f
ic
S
af
ety
Ad
m
in
is
tr
atio
n
[
4
]
,
5
0
,
0
0
0
in
j
u
r
ies
an
d
n
ea
r
ly
8
0
0
d
ea
th
s
h
av
e
b
ee
n
r
ep
o
r
ted
,
with
9
1
,
0
0
0
tr
af
f
ic
ac
cid
en
ts
r
elate
d
to
d
r
o
wsi
n
e
s
s
.
A
s
tu
d
y
b
y
t
h
e
AAA
Fo
u
n
d
atio
n
f
o
r
T
r
af
f
ic
Saf
ety
esti
m
ated
th
at
3
2
8
,
0
0
0
d
r
o
wsy
d
r
iv
i
n
g
c
r
ash
es
o
c
cu
r
an
n
u
ally
[
5
]
.
T
h
e
a
f
o
r
e
m
en
tio
n
ed
alar
m
in
g
s
tatis
t
ics
h
av
e
s
h
o
wn
th
e
n
e
ce
s
s
ity
o
f
a
s
y
s
tem
f
o
r
d
r
iv
er
d
r
o
wsi
n
ess
m
o
n
ito
r
in
g
an
d
aler
tin
g
,
th
e
r
eb
y
p
r
ev
en
tin
g
u
n
f
o
r
t
u
n
ate
tr
af
f
ic
ac
cid
en
ts
f
r
o
m
h
ap
p
en
in
g
.
R
ec
en
tly
,
m
an
y
m
o
d
els
h
av
e
b
ee
n
d
ev
elo
p
ed
f
o
r
au
to
m
atic
d
r
o
wsi
n
ess
d
etec
tio
n
s
y
s
tem
s
.
Of
ten
,
t
h
e
in
p
u
ts
o
f
th
e
d
etec
tio
n
s
y
s
tem
s
ar
e
im
ag
es
ca
p
tu
r
e
d
f
r
o
m
a
ca
m
er
a
p
lace
d
in
f
r
o
n
t
o
f
t
h
e
d
r
iv
e
r
f
ac
e.
T
h
e
ey
es
p
la
y
a
ce
n
tr
al
r
o
le
as
th
e
p
r
im
ar
y
f
ac
to
r
in
d
etec
tin
g
s
ig
n
s
o
f
f
atig
u
e,
in
clu
d
in
g
d
r
o
o
p
in
g
ey
elid
s
o
r
e
x
ten
d
e
d
p
er
io
d
s
o
f
cl
o
s
u
r
e.
Dee
p
c
o
n
v
o
lu
ti
o
n
al
n
e
u
r
al
n
etwo
r
k
s
(
C
NNs)
h
av
e
r
ec
en
tl
y
p
er
f
o
r
m
ed
e
x
ce
p
tio
n
ally
well
o
n
d
if
f
e
r
en
t ta
s
k
s
in
co
m
p
u
te
r
v
is
io
n
,
in
clu
d
in
g
o
b
ject
d
etec
tio
n
,
im
a
g
e
class
if
icatio
n
,
im
ag
e
s
eg
m
en
tati
o
n
,
o
b
ject
tr
ac
k
in
g
,
etc.
I
n
p
ar
ticu
lar
,
s
ev
er
al
p
r
o
p
o
s
ed
d
r
o
wsi
n
ess
d
etec
tio
n
s
y
s
tem
s
em
p
lo
y
ed
d
if
f
er
e
n
t
co
n
v
o
l
u
tio
n
al
n
etwo
r
k
ar
c
h
itectu
r
es
an
d
ac
h
iev
ed
ef
f
ec
tiv
e
an
d
r
eliab
le
r
esu
lts
.
T
h
e
p
o
wer
o
f
C
NN
m
o
d
els
ex
ce
l
is
th
e
e
x
tr
ac
tio
n
o
f
in
tr
icate
f
ea
tu
r
es
f
r
o
m
im
ag
es
an
d
v
id
e
o
s
,
co
n
tr
i
b
u
ti
n
g
to
en
h
an
ce
d
ac
cu
r
ac
y
in
id
en
tify
in
g
s
ig
n
s
o
f
d
r
o
wsi
n
ess
.
Am
o
n
g
th
e
m
y
r
ia
d
f
ea
tu
r
es
an
aly
ze
d
,
t
h
e
ey
es
e
m
er
g
e
as
a
f
o
ca
l
p
o
in
t
in
d
r
o
wsi
n
ess
d
ete
ctio
n
.
T
h
e
id
en
tif
icatio
n
an
d
m
itig
atio
n
o
f
d
r
o
wsi
n
ess
ar
e
ad
d
r
ess
ed
th
r
o
u
g
h
d
iv
er
s
e
ap
p
r
o
ac
h
es,
en
co
m
p
ass
in
g
v
ar
i
o
u
s
asp
ec
ts
o
f
b
o
th
p
h
y
s
io
lo
g
ical
an
d
b
eh
av
i
o
r
al
in
d
icato
r
s
.
Ph
y
s
io
lo
g
ically
,
th
e
ac
tiv
ity
o
f
th
e
b
r
ain
is
s
cr
u
tin
ized
u
s
i
n
g
E
E
G,
p
r
o
v
i
d
in
g
in
s
ig
h
ts
in
to
th
e
in
d
iv
id
u
al'
s
c
o
g
n
itiv
e
s
tate.
Simu
ltan
eo
u
s
ly
,
b
eh
av
io
r
al
cu
es
f
r
o
m
th
e
v
e
h
icle
its
elf
,
s
u
ch
as
lan
e
d
ep
ar
tu
r
e
p
atter
n
s
,
co
n
tr
ib
u
te
v
alu
ab
le
in
f
o
r
m
atio
n
r
e
g
ar
d
in
g
t
h
e
d
r
i
v
er
'
s
atten
tiv
e
n
ess
.
Ad
d
itio
n
ally
,
f
ac
ial
an
aly
s
is
s
er
v
es
as
a
k
ey
d
im
e
n
s
io
n
i
n
ass
ess
in
g
d
r
o
w
s
in
ess
,
ca
p
tu
r
in
g
n
u
an
ce
d
ex
p
r
ess
io
n
s
an
d
s
u
b
tle
s
ig
n
s
th
at
m
ay
in
d
icate
a
d
ec
lin
e
in
aler
tn
ess
.
.
.
Neu
r
al
n
etwo
r
k
s
h
a
v
e
b
ec
o
m
e
m
u
ch
m
o
r
e
p
r
o
f
o
u
n
d
,
with
s
tate
-
of
-
th
e
-
ar
t
n
etwo
r
k
s
[
6
]
,
ev
o
lv
in
g
f
r
o
m
h
av
in
g
ju
s
t
a
f
ew
la
y
er
s
(
e.
g
.
,
Ale
x
Net)
to
o
v
er
a
h
u
n
d
r
e
d
lay
e
r
s
.
Dete
ctin
g
d
r
o
wsi
n
ess
with
C
NN
in
tr
o
d
u
ce
s
ch
allen
g
es
th
at
d
i
r
ec
tly
im
p
ac
t
th
e
m
o
d
el
an
d
its
r
esp
o
n
s
e
tim
e.
R
ea
l
-
tim
e
co
n
s
tr
ain
ts
ar
e
a
s
ig
n
if
ican
t
f
ac
to
r
,
d
em
a
n
d
in
g
alg
o
r
ith
m
s
th
at
ca
n
s
wif
tly
p
r
o
ce
s
s
v
id
eo
s
tr
ea
m
s
in
a
tim
el
y
m
an
n
er
.
T
h
e
ar
ch
itectu
r
e'
s
co
m
p
lex
ity
p
lay
s
a
cr
u
cial
r
o
le,
as
in
tr
icate
m
o
d
els
m
ay
h
in
d
er
q
u
ic
k
r
esp
o
n
s
e
tim
es.
Ad
d
itio
n
ally
,
t
h
e
n
u
m
b
e
r
o
f
tr
ain
ed
p
ar
am
eter
s
in
C
NNs
d
ir
ec
tly
in
f
lu
en
ce
s
th
e
m
o
d
el
'
s
r
esp
o
n
s
iv
en
ess
,
af
f
ec
tin
g
its
ab
ilit
y
t
o
p
r
o
m
p
tly
d
ete
ct
an
d
r
esp
o
n
d
to
s
ig
n
s
o
f
d
r
o
wsi
n
ess
.
B
alan
cin
g
th
e
in
tr
icac
y
o
f
t
h
e
m
o
d
el
with
th
e
n
ee
d
f
o
r
r
ap
i
d
r
esp
o
n
s
es
is
ess
en
tial
f
o
r
d
ev
elo
p
in
g
ef
f
ec
tiv
e
d
r
o
wsi
n
ess
d
etec
tio
n
s
y
s
tem
s
an
d
en
s
u
r
in
g
th
eir
p
r
ac
tical
u
tili
ty
in
r
ea
l
-
wo
r
ld
s
ce
n
ar
io
s
.
R
esNet
h
as
ac
co
m
p
lis
h
ed
ex
ce
p
tio
n
al
ex
ec
u
tio
n
in
p
ictu
r
e
ac
k
n
o
wled
g
m
e
n
t
ass
ig
n
m
en
ts
,
o
u
tp
er
f
o
r
m
in
g
th
e
c
ap
ab
ilit
ies
o
f
p
ast
C
NN
d
esig
n
s
[
7
]
,
[
8
]
.
Du
r
in
g
ea
ch
iter
atio
n
o
f
tr
ain
in
g
a
n
e
u
r
al
n
etwo
r
k
,
all
weig
h
ts
r
ec
eiv
e
an
u
p
d
ate
p
r
o
p
o
r
tio
n
al
to
t
h
e
p
ar
tial
d
er
iv
ativ
e
o
f
th
e
er
r
o
r
f
u
n
ctio
n
co
n
ce
r
n
in
g
t
h
e
c
u
r
r
en
t
weig
h
t.
I
f
t
h
e
g
r
ad
ien
t
is
m
in
im
al,
th
e
weig
h
ts
will
n
o
t
b
e
ch
an
g
ed
e
f
f
ec
tiv
ely
a
n
d
m
a
y
s
to
p
th
e
n
eu
r
al
n
etwo
r
k
f
r
o
m
f
u
r
th
e
r
tr
ain
in
g
.
T
h
e
p
h
en
o
m
en
o
n
is
ca
lled
v
an
is
h
in
g
g
r
ad
ien
ts
.
Mo
r
e
s
p
e
cif
ically
,
th
e
d
ata
d
is
ap
p
ea
r
s
th
r
o
u
g
h
th
e
lay
e
r
s
o
f
th
e
d
ee
p
n
eu
r
al
n
etwo
r
k
d
u
e
to
a
v
er
y
s
lo
w
g
r
ad
ien
t
d
escen
t.
T
h
e
i
n
teg
r
atio
n
o
f
R
esid
u
al
b
lo
ck
s
is
m
o
tiv
ated
b
y
th
e
c
h
allen
g
e
o
f
tr
ain
in
g
d
ee
p
er
n
etwo
r
k
s
wh
ile
a
v
o
i
d
in
g
th
e
v
an
is
h
in
g
g
r
ad
ien
t
p
r
o
b
lem
[
9
]
.
R
esid
u
al
b
l
o
c
k
s
,
with
th
ei
r
s
k
ip
co
n
n
ec
tio
n
s
,
f
ac
ilit
ate
th
e
f
l
o
w
o
f
g
r
ad
ien
ts
th
r
o
u
g
h
t
h
e
n
et
wo
r
k
,
e
n
ab
lin
g
th
e
t
r
ain
in
g
o
f
d
ee
p
er
a
r
ch
itectu
r
es
with
o
u
t
s
ac
r
if
icin
g
p
e
r
f
o
r
m
an
ce
.
R
esid
u
al
n
etwo
r
k
s
,
lik
e
R
esNet,
h
av
e
co
n
s
is
ten
tly
ac
h
ie
v
ed
s
tate
-
of
-
th
e
-
ar
t
r
esu
lts
in
s
ig
n
if
ican
t
co
m
p
u
te
r
v
is
io
n
b
en
ch
m
ar
k
s
an
d
co
m
p
etitio
n
s
[
1
0
]
.
T
h
eir
p
e
r
f
o
r
m
an
ce
o
n
task
s
lik
e
I
m
ag
eNe
t
class
if
icatio
n
d
em
o
n
s
tr
ates
th
eir
ef
f
icac
y
in
lear
n
in
g
in
tr
icate
f
ea
tu
r
es
f
r
o
m
d
iv
e
r
s
e
im
ag
e
d
atasets
.
T
h
is
s
tr
o
n
g
tr
ac
k
r
ec
o
r
d
r
ei
n
f
o
r
ce
s
th
eir
r
eliab
ilit
y
an
d
m
ak
e
s
th
em
a
co
m
p
ellin
g
ch
o
ice
f
o
r
v
ar
io
u
s
co
m
p
u
te
r
v
is
io
n
task
s
.
T
h
e
aim
o
b
jectiv
e
o
f
th
is
s
tu
d
y
is
to
e
f
f
ec
tiv
ely
tack
le
th
e
cr
u
cial
ch
allen
g
e
o
f
d
r
iv
er
d
r
o
wsi
n
ess
d
etec
tio
n
,
a
m
atter
o
f
u
tm
o
s
t
im
p
o
r
tan
ce
f
o
r
r
o
ad
s
af
ety
t
h
at
n
ec
ess
itates
r
ea
l
-
tim
e
in
f
er
en
ce
ca
p
a
b
ilit
ies.
T
h
r
o
u
g
h
th
e
in
teg
r
atio
n
o
f
C
NNs
wi
th
R
e
s
id
u
al
b
lo
ck
s
,
we
n
o
t
o
n
ly
en
h
an
ce
th
e
m
o
d
el
'
s
ac
cu
r
ac
y
b
u
t
also
s
tr
ik
e
a
d
elica
te
b
alan
ce
b
et
wee
n
co
m
p
u
tatio
n
al
ef
f
icien
cy
an
d
p
er
f
o
r
m
an
ce
.
T
h
is
b
alan
ce
is
esp
ec
ially
cr
u
cial
wh
en
co
n
s
id
er
in
g
th
e
d
ep
lo
y
m
en
t
o
f
th
e
m
o
d
el
in
r
eso
u
r
ce
-
co
n
s
tr
ain
ed
e
n
v
ir
o
n
m
en
ts
,
s
u
ch
as
em
b
ed
d
e
d
s
y
s
tem
s
with
in
v
eh
icles.
Ma
n
y
s
o
lu
tio
n
s
h
av
e
b
ee
n
p
r
esen
ted
to
tack
le
th
e
ch
allen
g
e
o
f
d
etec
tin
g
d
r
iv
er
d
r
o
wsi
n
ess
.
I
n
th
is
co
n
tex
t,
we
h
ig
h
lig
h
t
a
s
elec
tio
n
o
f
r
ec
e
n
t
wo
r
k
s
th
at
co
n
tr
ib
u
te
to
th
is
f
ield
.
C
h
ir
r
a
et
a
l.
[
1
1
]
p
r
esen
ted
a
s
y
s
tem
lev
er
ag
in
g
d
ee
p
lea
r
n
i
n
g
tec
h
n
iq
u
es
f
o
r
d
etec
tin
g
d
r
iv
er
d
r
o
wsi
n
ess
,
f
o
cu
s
in
g
o
n
ass
es
s
in
g
ey
e
s
tate.
T
h
e
in
itial
s
tep
s
o
f
th
e
ap
p
r
o
a
ch
in
v
o
l
v
ed
em
p
lo
y
in
g
th
e
Vi
o
la
-
J
o
n
es
alg
o
r
ith
m
f
o
r
task
s
en
co
m
p
ass
in
g
f
ac
e
d
etec
tio
n
an
d
lo
ca
liza
tio
n
o
f
t
h
e
ey
e
r
eg
io
n
s
.
I
f
th
e
class
if
ier
d
etec
ts
ten
co
n
s
ec
u
tiv
e
in
s
tan
ce
s
o
f
clo
s
ed
ey
es,
it
in
d
icate
s
th
at
th
e
in
d
iv
id
u
al
is
ex
h
ib
itin
g
s
ig
n
s
o
f
d
r
o
wsi
n
ess
.
A
C
NN
m
o
d
el
was
tr
ai
n
ed
u
s
in
g
a
d
ataset
co
m
p
r
is
ed
o
f
2
8
5
0
i
m
ag
es,
e
n
co
m
p
ass
in
g
two
d
is
tin
ct
cla
s
s
es:
o
p
en
ey
es
a
n
d
cl
o
s
ed
e
y
es,
ac
h
iev
in
g
a
n
ac
cu
r
ac
y
o
f
9
6
.
4
2
%.
C
NN
m
o
d
el
was
also
u
s
ed
b
y
R
af
id
et
a
l.
[
1
2
]
to
d
ev
elo
p
a
d
r
o
wsi
n
ess
d
etec
tio
n
s
y
s
tem
.
T
h
is
ap
p
r
o
ac
h
em
p
l
o
y
ed
th
e
Vio
la
-
J
o
n
es
alg
o
r
ith
m
f
o
r
p
r
ec
is
e
lo
ca
lizatio
n
o
f
th
e
ey
es.
T
h
e
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
.
3
,
Ma
r
ch
20
2
5
:
1
692
-
1
7
0
1
1694
d
r
o
wsi
n
ess
s
tate
if
co
n
f
ir
m
ed
wh
en
th
e
class
if
ier
id
en
tifie
s
3
0
co
n
s
ec
u
tiv
e
in
s
tan
ce
s
o
f
clo
s
ed
ey
es.
T
h
e
C
NN
m
o
d
el
ac
h
iev
ed
an
ac
cu
r
ac
y
r
ate
o
f
9
4
.
5
%
u
s
in
g
th
e
MRL
E
y
e
Data
s
et.
Sh
ak
ee
l
et
a
l
.
[
1
3
]
i
n
tr
o
d
u
ce
d
a
Mo
b
ileNet
m
o
d
el
to
d
etec
t
d
r
o
wsi
n
ess
ca
s
e
s
.
T
h
e
m
eth
o
d
o
lo
g
y
in
co
r
p
o
r
ated
T
en
s
o
r
Flo
w
d
etec
tio
n
m
o
d
el
zo
o
f
o
r
b
o
t
h
f
ac
e
d
etec
tio
n
an
d
e
y
es
lo
ca
lizatio
n
.
T
h
e
m
o
d
el
was
tr
ain
e
d
o
n
a
d
ataset
co
m
p
r
is
in
g
6
0
0
0
im
ag
es a
n
d
ac
h
iev
ed
an
ac
cu
r
ac
y
o
f
7
6
.
9
%.
T
h
e
d
etec
tio
n
o
f
s
ig
n
s
o
f
d
r
o
wsi
n
ess
i
s
co
n
f
ir
m
ed
if
th
e
class
if
ier
i
d
en
tifie
s
ten
co
n
s
ec
u
tiv
e
in
s
tan
ce
s
o
f
clo
s
ed
ey
es.
R
ajam
o
h
an
a
et
a
l.
[
1
4
]
in
tr
o
d
u
ce
d
an
h
y
b
r
id
ap
p
r
o
ac
h
th
at
co
m
b
in
es
a
C
NN
m
o
d
el
with
B
id
ir
ec
tio
n
al
lo
n
g
s
h
o
r
t
-
ter
m
m
em
o
r
y
(
B
iLST
M)
alg
o
r
it
h
m
to
d
etec
t
d
r
iv
er
d
r
o
wsi
n
ess
b
ased
o
n
e
y
e
s
tate.
T
h
e
ey
es
r
e
g
io
n
ex
tr
ac
tio
n
p
r
o
ce
s
s
was
f
ac
ilit
ated
b
y
e
m
p
lo
y
in
g
th
e
DL
I
B
lib
r
ar
y
.
T
h
e
d
ataset,
co
m
p
r
is
in
g
2
2
0
8
im
a
g
es,
s
er
v
ed
as
th
e
f
o
u
n
d
atio
n
f
o
r
th
eir
r
ese
ar
ch
,
lead
in
g
to
a
n
ac
cu
r
ac
y
r
ate
o
f
9
4
%.
T
h
e
m
eth
o
d
o
lo
g
y
in
itiates
a
tr
ig
g
er
ac
tio
n
wh
en
ey
es
r
em
ain
clo
s
ed
f
o
r
a
n
ex
ten
d
ed
p
er
io
d
.
J
o
s
e
et
a
l.
[
1
5
]
u
s
ed
E
f
f
icien
tNetV2
lig
h
tweig
h
t
co
n
v
o
lu
tio
n
al
n
etwo
r
k
m
o
d
el
f
o
r
d
etec
tin
g
d
r
o
wsi
n
ess
b
y
em
p
lo
y
in
g
a
co
n
cu
r
r
en
tly
with
f
ac
ial
lan
d
m
ar
k
d
etec
tio
n
.
T
h
eir
ap
p
r
o
ac
h
in
v
o
lv
ed
th
e
Me
d
iaPip
e
Face
Me
s
h
alg
o
r
it
h
m
to
ex
tr
ac
t
ey
e
r
e
g
io
n
s
.
T
h
e
m
o
d
el
was
tr
ain
ed
u
s
in
g
a
d
ataset
co
m
p
r
is
in
g
1
0
,
0
7
6
im
ag
es,
ac
h
iev
in
g
an
ac
cu
r
ac
y
o
f
9
6
.
9
2
%.
A
p
iv
o
t
al
asp
ec
t
o
f
th
eir
s
tr
ateg
y
in
v
o
lv
es
th
e
d
ec
is
io
n
-
m
ak
in
g
p
r
o
ce
s
s
.
T
h
e
n
u
m
b
e
r
o
f
f
r
am
es
co
n
s
id
er
ed
f
o
r
co
n
f
i
r
m
in
g
d
r
o
wsi
n
ess
d
etec
tio
n
is
6
0
f
r
am
e
s
.
I
n
th
eir
r
esear
ch
,
B
ek
h
o
u
ch
e
et
a
l.
[
1
6
]
ex
p
lo
r
ed
v
id
e
o
-
b
ased
d
r
o
w
s
in
ess
d
etec
tio
n
b
y
lev
e
r
ag
in
g
th
e
NT
HU
d
r
o
wsy
d
r
iv
er
d
etec
tio
n
d
ataset
(
NT
HU
-
DDD)
.
T
h
eir
in
v
esti
g
ati
o
n
f
o
cu
s
ed
o
n
id
e
n
tify
in
g
t
h
e
m
o
s
t
ef
f
ec
tiv
e
co
n
f
ig
u
r
atio
n
s
co
n
ce
r
n
in
g
f
r
a
m
e
co
u
n
t
an
d
in
ter
-
f
r
am
e
s
k
i
p
s
.
T
h
e
e
x
p
er
im
en
ts
led
to
a
n
ac
cu
r
ac
y
o
f
8
6
%
wh
en
co
m
b
in
i
n
g
a
R
esNet
-
5
0
with
s
u
p
p
o
r
t
v
ec
t
o
r
m
ac
h
in
e
(
SVM)
class
if
ier
s
.
Fo
r
p
r
ec
is
e
ey
e
r
eg
io
n
lo
ca
lizatio
n
,
T
in
y
-
YOL
O
m
o
d
el
was
em
p
lo
y
ed
.
T
h
e
f
in
d
in
g
s
h
i
g
h
lig
h
ted
th
at
th
e
b
est
co
n
f
ig
u
r
atio
n
,
co
n
s
id
er
in
g
co
m
p
u
tatio
n
tim
e,
en
tailed
u
tili
zin
g
n
in
e
f
r
a
m
es
an
d
s
k
i
p
p
in
g
two
i
n
ter
-
f
r
am
es.
C
iv
ik
a
n
d
Yu
zg
ec
[
1
7
]
p
r
esen
ted
a
C
NN
-
b
ased
m
o
d
el
f
o
r
d
r
o
wsi
n
e
s
s
d
etec
tio
n
,
s
p
ec
if
ically
tar
g
etin
g
th
e
ey
es
an
d
m
o
u
th
r
e
g
io
n
s
.
T
h
e
DL
I
B
lib
r
ar
y
was
u
s
ed
to
ex
tr
ac
t
th
e
f
ac
ial
r
eg
io
n
s
,
an
d
th
e
Yaw
D
D
d
atasetto
tr
ain
th
e
m
o
d
el
ex
h
ib
itin
g
an
ac
c
u
r
ac
y
o
f
9
4
.
5
%.
Ho
wev
er
,
th
e
s
p
e
cif
ic
th
r
esh
o
ld
f
o
r
d
ef
i
n
in
g
d
r
o
wsi
n
ess
was
n
o
t
p
r
o
v
id
e
d
in
th
is
wo
r
k
.
Flo
r
ez
et
a
l.
[
1
8
]
f
o
c
u
s
ed
s
p
ec
if
ically
o
n
ey
e
clo
s
u
r
e
d
u
r
atio
n
as
a
cr
u
cia
l
in
d
icato
r
to
d
etec
t
d
r
o
wsi
n
ess
.
Fo
r
p
r
ec
is
e
ey
e
r
eg
io
n
d
elin
ea
tio
n
,
th
e
M
ed
iaPip
e
Face
Me
s
h
al
g
o
r
ith
m
was
u
s
ed
,
an
d
f
o
r
class
if
icatio
n
th
e
I
n
ce
p
tio
n
V3
,
VGG1
6
,
an
d
R
esNet5
0
V2
m
o
d
els,
wer
e
em
p
lo
y
e
d
.
T
h
e
b
est
ac
cu
r
ac
y
o
f
9
8
.
1
5
%
was
o
b
tain
ed
u
s
in
g
th
e
R
esNe
t
5
0
V2
m
o
d
el
o
n
th
e
NI
T
YM
E
D
d
ataset,
with
a
r
e
s
p
o
n
s
e
tim
e
o
f
1
0
6
.
5
m
illi
s
ec
o
n
d
s
.
T
h
e
tr
ac
k
in
g
d
u
r
atio
n
o
f
e
y
e
clo
s
u
r
e
is
3
0
0
m
illi
s
ec
o
n
d
s
,
an
d
wh
e
n
ex
ee
d
ed
i
t
is
a
co
n
f
ir
m
atio
n
o
f
d
r
o
wsi
n
ess
s
tate.
Ph
an
et
a
l.
[
1
9
]
co
m
p
a
r
ed
m
u
ltip
le
d
ee
p
-
lear
n
i
n
g
m
o
d
els
f
o
r
d
r
o
wsi
n
ess
d
etec
tio
n
,
in
clu
d
in
g
L
STM
,
VGG
-
1
6
,
Den
s
eNe
t,
an
d
I
n
ce
p
tio
n
-
V3
.
T
h
e
SS
D
-
R
esNet
-
1
0
m
o
d
el
was
em
p
lo
y
ed
f
o
r
ef
f
icien
t
ey
e
r
eg
io
n
lo
ca
lizatio
n
.
Usi
n
g
a
lar
g
e
d
ataset
o
f
1
3
,
7
2
9
im
a
g
es,
th
e
f
u
s
io
n
ap
p
r
o
ac
h
ac
h
iev
ed
9
8
%
ac
cu
r
ac
y
u
s
in
g
Den
s
eNe
t.
Ho
wev
er
,
th
e
s
tu
d
y
d
id
n
o
t
p
r
o
v
i
d
e
th
e
r
esp
o
n
s
e
tim
e
n
eith
er
th
e
th
r
esh
o
ld
u
s
ed
to
d
eter
m
in
e
th
e
d
r
o
wsy
s
tate.
Acc
o
r
d
in
g
to
t
h
e
an
aly
s
is
o
f
th
e
s
tate
o
f
th
e
ar
t,
v
ar
io
u
s
alg
o
r
ith
m
s
ar
e
d
e
p
lo
y
ed
f
o
r
d
r
o
wsi
n
ess
d
etec
tio
n
,
in
clu
d
in
g
m
o
d
els
l
ik
e
C
NN,
R
e
s
Net5
0
V2
,
Mo
b
ile
Net,
E
f
f
icien
tNetV2
,
an
d
o
th
er
s
.
I
n
ter
m
s
o
f
p
er
f
o
r
m
an
ce
,
th
e
R
esNet5
0
V
2
alg
o
r
ith
m
,
u
s
ed
b
y
Flo
r
ez
e
t
a
l.
[
1
8
]
,
h
as
d
em
o
n
s
tr
ated
o
u
ts
tan
d
in
g
ac
c
u
r
ac
y
at
9
8
.
1
5
%,
s
h
o
wca
s
in
g
its
ef
f
ec
tiv
en
ess
.
Ho
wev
er
,
it's
cr
u
cial
to
n
o
te
th
at
th
e
m
in
im
u
m
ac
cu
r
ac
y
lev
el
m
ay
v
ar
y
ac
r
o
s
s
s
tu
d
ies,
u
n
d
e
r
s
co
r
in
g
th
e
im
p
o
r
tan
ce
o
f
ch
o
o
s
in
g
an
alg
o
r
ith
m
b
ased
o
n
s
p
ec
if
ic
n
ee
d
s
.
W
h
ile
th
e
p
er
f
o
r
m
an
ce
s
ar
e
p
r
o
m
is
in
g
,
co
n
s
id
er
in
g
t
h
e
r
esp
o
n
s
e
ti
m
e
o
f
th
e
tr
ain
e
d
m
o
d
el
is
im
p
er
ativ
e.
So
m
e
alg
o
r
ith
m
s
,
d
esp
ite
th
eir
h
ig
h
ac
cu
r
ac
y
,
m
ay
h
a
v
e
lo
n
g
e
r
r
esp
o
n
s
e
tim
es,
a
cr
itical
f
ac
to
r
f
o
r
r
ea
l
-
tim
e
ap
p
licatio
n
s
s
u
ch
as
d
r
o
wsi
n
ess
d
etec
tio
n
wh
ile
d
r
iv
in
g
.
T
h
er
ef
o
r
e,
th
e
s
elec
tio
n
o
f
a
n
alg
o
r
ith
m
s
h
o
u
ld
b
e
g
u
id
ed
n
o
t
o
n
ly
b
y
its
in
tr
in
s
ic
p
er
f
o
r
m
an
ce
b
u
t
also
b
y
t
h
e
r
elev
an
ce
o
f
its
r
esp
o
n
s
e
tim
e
in
th
e
in
ten
d
ed
u
s
ag
e
co
n
tex
t.
R
esNet
ad
d
r
ess
es
th
e
ch
allen
g
es
o
f
d
ee
p
n
et
wo
r
k
tr
ain
i
n
g
b
y
in
c
o
r
p
o
r
atin
g
s
k
ip
co
n
n
ec
tio
n
s
o
r
r
esid
u
al
b
l
o
ck
s
.
T
h
is
in
n
o
v
ativ
e
d
esig
n
,
s
ee
n
in
v
ar
ia
n
ts
lik
e
R
esNet5
0
an
d
R
esN
et1
0
1
,
en
a
b
les
th
e
tr
ain
in
g
o
f
m
u
ch
d
ee
p
er
n
etw
o
r
k
s
,
ef
f
ec
tiv
ely
m
itig
atin
g
is
s
u
es
s
u
ch
as
v
a
n
is
h
in
g
g
r
ad
ien
ts
.
T
h
e
ar
ch
itectu
r
e
n
o
t
o
n
ly
e
n
h
an
ce
s
ac
c
u
r
ac
y
b
u
t
also
im
p
r
o
v
es
tr
ain
in
g
ef
f
icien
cy
,
m
ak
i
n
g
it
p
ar
ti
cu
lar
ly
s
u
itab
le
f
o
r
ap
p
licatio
n
s
wh
er
e
p
r
o
ce
s
s
in
g
tim
e
is
a
cr
itical
f
ac
to
r
.
R
esNet
'
s
ab
ili
ty
to
b
alan
ce
d
ep
th
an
d
ef
f
icien
c
y
r
en
d
er
s
it
a
co
m
p
ellin
g
c
h
o
i
ce
,
esp
ec
ially
in
s
ce
n
ar
io
s
w
h
er
e
p
r
o
ce
s
s
in
g
tim
e
is
a
cr
u
cial
co
n
s
id
er
atio
n
.
T
h
er
ef
o
r
e,
wh
en
co
n
s
id
er
i
n
g
alg
o
r
ith
m
s
f
o
r
d
r
o
wsi
n
ess
d
etec
tio
n
,
th
e
u
n
iq
u
e
f
ea
tu
r
es
an
d
a
d
v
an
tag
es
o
f
R
esNet
s
h
o
u
ld
b
e
tak
e
n
in
to
a
cc
o
u
n
t a
lo
n
g
s
id
e
th
eir
r
esp
ec
ti
v
e
p
er
f
o
r
m
an
ce
s
.
T
h
e
r
em
ain
d
e
r
o
f
th
is
p
ap
e
r
is
s
tr
u
ctu
r
ed
as
f
o
llo
ws.
Sectio
n
2
p
r
esen
ts
th
e
p
r
o
p
o
s
ed
m
e
th
o
d
o
lo
g
y
b
ased
o
n
a
n
ew
C
NN
ar
ch
itectu
r
e.
Sectio
n
3
p
r
esen
ts
th
e
r
e
s
u
lts
an
d
d
is
cu
s
s
io
n
.
Fin
ally
,
s
ec
tio
n
4
p
r
o
v
id
es
a
co
n
clu
s
io
n
an
d
f
u
tu
r
e
wo
r
k
s
.
2.
RE
S
E
ARCH
M
E
T
H
O
D
T
h
e
o
b
jectiv
e
o
f
th
is
r
esear
c
h
is
to
p
r
o
p
o
s
e
a
C
NN
b
ase
d
m
o
d
el
f
o
r
d
r
o
wsi
n
ess
d
etec
tio
n
th
at
p
r
o
d
u
ce
b
etter
r
esp
o
n
s
e
tim
e
a
n
d
th
at
h
a
v
e
less
h
y
p
er
p
a
r
am
e
ter
s
.
T
h
e
a
p
p
r
o
ac
h
s
tar
ts
b
y
ca
p
tu
r
in
g
s
u
cc
ess
iv
e
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
A
n
ew d
ee
p
lea
r
n
in
g
mo
d
el
b
a
s
ed
o
n
co
n
vo
lu
tio
n
a
l n
eu
r
a
l
n
etw
o
r
k
…
(
A
b
d
elfetta
h
S
o
u
lta
n
a
)
1695
v
id
eo
f
r
a
m
es
f
r
o
m
a
ca
m
er
a
an
d
p
r
o
ce
s
s
in
g
ea
ch
f
o
llo
wi
n
g
a
s
et
o
f
s
tep
s
.
T
h
e
in
itial
s
tep
s
in
v
o
lv
e
th
e
ap
p
licatio
n
o
f
Haa
r
ca
s
ca
d
es
t
o
d
etec
t
f
ac
ial
a
n
d
o
cu
lar
f
ea
tu
r
es
ac
cu
r
ately
.
Sp
ec
if
ically
,
p
r
e
-
tr
ain
ed
ca
s
ca
d
e
class
if
ier
s
ar
e
em
p
lo
y
ed
f
o
r
r
ec
o
g
n
izin
g
f
ac
ial
co
n
to
u
r
s
as
well
as
th
e
d
is
tin
ctiv
e
ch
ar
ac
ter
is
tics
o
f
th
e
ey
es.
A
R
e
s
Net1
8
n
eu
r
al
n
etwo
r
k
m
o
d
el
is
in
tr
o
d
u
ce
d
to
d
eter
m
in
e
th
e
ey
es'
s
tate.
T
h
is
m
o
d
el
an
aly
ze
s
th
e
ex
tr
ac
ted
ey
e
r
eg
io
n
s
an
d
p
r
e
d
icts
wh
e
th
er
th
ey
ar
e
o
p
e
n
o
r
clo
s
ed
.
As
th
e
v
id
e
o
f
r
am
es
ar
e
p
r
o
c
ess
ed
,
th
e
co
d
e
m
ai
n
tain
s
a
s
co
r
e
th
at
is
in
cr
em
en
te
d
ea
ch
tim
e
a
n
in
s
tan
ce
s
o
f
clo
s
ed
-
ey
e
is
d
etec
ted
.
W
h
en
t
h
e
s
co
r
e
ex
ce
ed
s
a
p
r
ed
ef
in
e
d
th
r
esh
o
l
d
o
f
1
0
co
n
s
ec
u
tiv
e
im
ag
es with
th
e
ey
es c
lo
s
ed
,
t
h
e
s
y
s
tem
p
r
o
v
id
es a
s
ig
n
al
to
th
e
d
r
iv
e
r
ab
o
u
t
h
is
s
tate
o
f
d
r
o
wsi
n
ess
in
r
ea
l
tim
e.
T
h
e
Fig
u
r
e
1
p
r
esen
ts
two
s
tep
s
:
th
e
o
f
f
lin
e
m
o
d
el
co
n
s
tr
u
ctio
n
w
h
ich
is
th
e
co
n
s
tr
u
ctio
n
p
h
ase
o
f
th
e
p
r
etr
ain
e
d
m
o
d
el
an
d
th
e
o
n
lin
e
d
r
o
wsi
n
ess
d
etec
tio
n
th
at
ca
p
tu
r
e
in
r
ea
l tim
e
d
r
iv
e
r
v
id
eo
an
d
ass
ess
th
e
d
r
o
wsi
n
ess
ca
s
es.
Fig
u
r
e
1
.
B
lo
ck
d
iag
r
am
o
f
th
e
p
r
o
p
o
s
ed
d
r
o
wsi
n
ess
d
etec
ti
o
n
s
y
s
tem
2
.
1
.
Da
t
a
s
et
T
h
e
d
ee
p
lear
n
in
g
m
o
d
el
d
ev
elo
p
ed
h
er
e
is
tr
ai
n
ed
o
n
im
a
g
es
f
r
o
m
th
e
Kag
g
le
e
y
es
d
at
aset
[
2
0
]
.
T
h
e
ey
es
d
ataset
co
m
p
r
is
es
a
to
tal
o
f
4
,
8
5
0
g
r
ay
s
ca
le
im
ag
es.
Am
o
n
g
th
ese,
3
,
7
9
0
im
ag
es
ar
e
allo
ca
ted
f
o
r
tr
ain
in
g
,
d
is
tr
ib
u
ted
ac
r
o
s
s
two
class
es:
"c
lo
s
ed
"
an
d
"o
p
en
"
ey
es.
Ad
d
itio
n
ally
,
1
,
0
6
0
im
ag
es
ar
e
d
esig
n
ated
f
o
r
test
in
g
,
co
v
er
in
g
b
o
th
"
o
p
en
"
an
d
"c
lo
s
ed
"
ey
e
class
es.
T
h
ese
im
ag
es
ex
h
ib
it
d
iv
er
s
ity
in
lig
h
tin
g
co
n
d
itio
n
s
a
n
d
s
u
b
ject
v
ar
iab
i
lity
an
d
ev
e
n
in
cl
u
d
e
in
s
tan
ce
s
o
f
i
n
d
iv
id
u
als
wea
r
in
g
g
lass
es.
E
ac
h
i
m
ag
e
h
as
d
im
en
s
io
n
s
o
f
24
×
24
p
ix
els.
Giv
en
th
ese
ch
ar
ac
ter
is
tics
,
th
is
d
ataset
p
r
o
v
es
to
b
e
well
-
s
u
ited
f
o
r
ef
f
ec
tiv
el
y
tr
ain
in
g
an
d
co
n
s
tr
u
ctin
g
a
r
e
s
ilien
t
m
o
d
el
ca
p
ab
le
o
f
ac
cu
r
ately
tr
ac
k
in
g
ey
e
clo
s
u
r
e.
T
h
e
F
ig
u
r
e
2
p
r
esen
ts
s
o
m
e
ex
am
p
les o
f
im
a
g
es f
r
o
m
th
e
d
ataset,
s
h
o
wca
s
in
g
b
o
th
ca
s
es
o
f
o
p
e
n
an
d
clo
s
ed
ey
e
s
.
Fig
u
r
e
2
.
R
ep
r
esen
tatio
n
o
f
a
n
n
o
tated
im
ag
es o
f
th
e
ey
es_
d
a
taset
2
.
2
.
O
f
f
lin
e
m
o
del c
o
ns
t
ruct
io
n
T
h
e
o
f
f
lin
e
tr
ain
in
g
d
iag
r
am
s
h
o
ws
h
o
w
th
e
m
o
d
el
is
b
u
ilt.
I
n
d
ee
d
,
th
e
p
r
o
p
o
s
ed
m
o
d
el
in
co
r
p
o
r
ates
r
esid
u
al
b
lo
c
k
s
to
f
ac
ilit
ate
d
ee
p
er
r
ep
r
esen
tatio
n
s
lear
n
i
n
g
a
n
d
i
m
p
r
o
v
e
t
h
e
m
o
d
el
p
er
f
o
r
m
an
ce
.
T
h
e
ar
ch
itectu
r
e
co
n
s
is
ts
o
f
th
e
f
o
l
lo
win
g
s
tag
es:
−
I
n
p
u
t
p
r
ep
a
r
atio
n
an
d
in
itial
c
o
n
v
o
lu
ti
o
n
al
lay
er
:
t
h
e
in
p
u
t
i
m
ag
es
h
av
e
d
im
en
s
io
n
s
o
f
(
1
4
5
,
1
4
5
,
3
)
an
d
a
ze
r
o
-
p
a
d
d
in
g
is
u
s
ed
to
m
ain
t
ain
s
ize.
Su
b
s
eq
u
e
n
tly
,
a
c
o
n
v
o
lu
tio
n
al
lay
e
r
with
a
7
×
7
k
er
n
el
an
d
a
s
tr
id
e
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
.
3
,
Ma
r
ch
20
2
5
:
1
692
-
1
7
0
1
1696
o
f
2
is
in
tr
o
d
u
ce
d
to
r
ed
u
ce
th
e
d
im
e
n
s
io
n
s
.
B
atch
n
o
r
m
aliza
tio
n
an
d
R
eL
U
ac
tiv
at
io
n
ar
e
ap
p
lied
,
f
o
llo
wed
b
y
m
ax
-
p
o
o
li
n
g
to
d
o
wn
s
ize
th
e
f
ea
tu
r
e
m
a
p
s
f
u
r
t
h
er
.
−
R
esid
u
al
b
lo
ck
(
s
tag
e
2
)
:
t
h
e
in
p
u
t
p
ass
es
th
r
o
u
g
h
a
r
esid
u
al
b
lo
ck
with
two
c
o
n
v
o
lu
tio
n
al
lay
er
s
a
n
d
a
s
h
o
r
tcu
t
co
n
n
ec
tio
n
.
T
h
e
b
lo
ck
en
h
an
ce
s
th
e
m
o
d
el'
s
ab
ilit
y
to
lear
n
i
n
tr
icate
f
ea
tu
r
es.
B
atch
n
o
r
m
aliza
tio
n
an
d
R
eL
U
ac
ti
v
atio
n
f
o
llo
w
ea
ch
co
n
v
o
l
u
tio
n
,
an
d
th
e
r
esu
lt
is
ad
d
ed
ele
m
en
t
-
wis
e
to
th
e
o
r
ig
in
al
in
p
u
t b
ef
o
r
e
a
f
in
al
R
eL
U
ac
tiv
atio
n
.
−
R
esid
u
al
bl
o
ck
(
s
tag
e
3
)
:
s
im
ilar
to
s
tag
e
2
,
th
e
in
p
u
t
g
o
es
th
r
o
u
g
h
an
o
th
er
r
esid
u
al
b
lo
ck
with
in
cr
ea
s
ed
f
ilter
s
,
cr
ea
tin
g
a
d
ee
p
e
r
u
n
d
er
s
tan
d
in
g
o
f
f
ea
tu
r
es.
T
h
e
s
am
e
p
atter
n
o
f
b
atch
n
o
r
m
aliza
tio
n
,
R
eL
U
ac
tiv
atio
n
,
an
d
elem
en
t
-
wis
e
ad
d
itio
n
is
m
ain
tain
ed
.
−
Av
er
ag
e
p
o
o
li
n
g
an
d
f
in
al
cl
ass
if
icatio
n
:
t
h
e
av
er
ag
e
p
o
o
lin
g
lay
er
r
ed
u
ce
s
f
ea
tu
r
e
m
ap
d
im
en
s
io
n
s
,
p
r
ep
ar
in
g
th
e
m
f
o
r
t
h
e
f
in
al
cl
ass
if
icatio
n
.
T
h
e
f
latten
e
d
f
ea
t
u
r
es
ar
e
f
ed
in
to
a
f
u
ll
y
co
n
n
e
cted
lay
er
with
a
s
o
f
tm
ax
ac
tiv
atio
n
f
u
n
ctio
n
t
o
o
u
tp
u
t c
lass
p
r
o
b
ab
ilit
ies.
2
.
3
.
O
nli
ne
dro
wsi
nes
s
det
ec
t
io
n
T
h
e
o
n
lin
e
d
r
o
wsi
n
ess
d
etec
tio
n
p
r
o
ce
s
s
in
itiates
with
in
p
u
t
f
r
o
m
a
v
id
eo
s
tr
ea
m
,
f
o
l
lo
wed
b
y
f
r
am
e
-
by
-
f
r
a
m
e
e
x
tr
ac
tio
n
.
T
h
e
Haa
r
ca
s
ca
d
e
alg
o
r
ith
m
is
ap
p
lied
f
o
r
f
ac
e
d
etec
tio
n
,
an
d
if
a
f
ac
e
is
d
etec
ted
,
th
e
p
r
o
ce
s
s
p
r
o
ce
e
d
s
to
ey
e
d
etec
tio
n
;
o
th
er
wis
e,
it
m
o
v
es
t
o
th
e
n
ex
t
f
r
a
m
e.
Af
te
r
ex
t
r
ac
tin
g
th
e
ey
e
r
eg
io
n
u
s
in
g
th
e
Haa
r
ca
s
ca
d
e,
it u
n
d
er
g
o
es c
l
ass
if
icatio
n
u
s
in
g
o
u
r
p
r
e
-
tr
ain
ed
m
o
d
el,
d
etailed
in
th
e
f
ir
s
t sectio
n
,
to
d
eter
m
in
e
th
e
s
tate
o
f
t
h
e
ey
e
s
as
o
p
en
o
r
cl
o
s
ed
.
A
co
u
n
te
r
is
in
cr
em
en
ted
wh
e
n
clo
s
ed
ey
es
ar
e
d
etec
ted
,
an
d
wh
e
n
it
r
ea
ch
es
a
p
r
ed
et
er
m
in
ed
th
r
esh
o
ld
o
v
er
1
0
co
n
s
ec
u
tiv
e
f
r
a
m
es,
th
e
s
y
s
tem
co
n
clu
d
es
t
h
at
th
e
d
r
iv
er
is
in
a
s
tate
o
f
d
r
o
wsi
n
ess
.
At
th
is
p
o
in
t,
an
aler
t
is
s
en
t
to
th
e
d
r
iv
er
,
n
o
tify
in
g
t
h
em
o
f
th
eir
d
r
o
wsy
s
tate
an
d
p
r
o
m
p
tin
g
ap
p
r
o
p
r
ia
te
ac
tio
n
s
to
en
s
u
r
e
s
af
ety
wh
i
le
d
r
iv
in
g
.
T
h
is
co
n
tin
u
o
u
s
an
d
r
ea
l
-
tim
e
p
r
o
ce
s
s
e
n
s
u
r
es c
o
n
s
tan
t m
o
n
it
o
r
in
g
o
f
th
e
d
r
i
v
er
'
s
co
n
d
itio
n
d
u
r
in
g
th
eir
jo
u
r
n
ey
.
2
.
4
. P
r
o
po
s
ed
m
o
del det
a
ils
R
esNet
r
ev
o
lu
tio
n
ized
th
e
C
NN
ar
ch
itectu
r
al
r
ac
e
b
y
in
tr
o
d
u
cin
g
th
e
co
n
ce
p
t
o
f
r
esid
u
al
lear
n
in
g
in
C
NNs
an
d
d
ev
is
ed
an
e
f
f
icien
t
m
eth
o
d
o
lo
g
y
f
o
r
th
e
t
r
ain
i
n
g
o
f
d
ee
p
n
etwo
r
k
s
[
2
1
]
.
T
h
ey
ar
e
c
h
ar
ac
ter
ized
b
y
th
eir
d
ee
p
ar
c
h
itectu
r
es,
wh
ich
ca
n
p
r
o
d
u
ce
lo
w
er
r
o
r
r
at
es.
T
h
e
b
asic
b
u
ild
in
g
b
lo
ck
s
o
f
a
r
esid
u
al
n
eu
r
al
n
etwo
r
k
ar
e
th
e
s
o
-
ca
lled
r
esid
u
al
b
lo
ck
s
.
T
h
e
b
asic
id
ea
h
er
e
is
th
at
s
o
-
ca
lled
“sk
ip
co
n
n
ec
tio
n
s
”
ar
e
b
u
ilt
in
to
th
e
n
etwo
r
k
.
T
h
ese
en
s
u
r
e
th
at
th
e
ac
tiv
atio
n
o
f
a
la
y
er
is
ad
d
ed
to
g
eth
er
with
th
e
o
u
t
p
u
t
o
f
a
later
lay
e
r
.
L
im
o
n
o
v
a
et
a
l.
[
2
2
]
h
as
co
n
clu
s
iv
ely
d
em
o
n
s
tr
ated
th
at
th
e
im
p
lem
en
tatio
n
o
f
th
e
R
es
Net
m
o
d
el
d
em
an
d
s
n
o
tab
ly
f
ewe
r
lo
g
ic
g
ates,
r
ed
u
cin
g
th
e
r
eq
u
ir
e
m
en
t
b
y
a
f
a
cto
r
o
f
2
.
1
-
2
.
9
tim
es
in
co
m
p
ar
is
o
n
to
alter
n
ativ
e
m
o
d
els.
R
esid
u
al
n
etwo
r
k
(
R
esNet
)
is
a
n
etwo
r
k
m
o
d
el
with
b
ette
r
p
er
f
o
r
m
an
ce
an
d
s
m
aller
m
o
d
el
s
ize
an
d
p
ar
am
eter
s
.
I
t
o
cc
u
p
ies
less
m
em
o
r
y
s
p
ac
e
wh
ile
m
ain
tain
in
g
ac
c
u
r
ac
y
an
d
h
as
s
tr
o
n
g
m
o
d
u
lar
ity
an
d
p
o
r
tab
ilit
y
.
T
h
is
ar
ch
itectu
r
e
a
llo
ws
th
e
n
etwo
r
k
to
s
k
ip
s
p
e
cif
ic
lay
er
s
s
ig
n
if
ican
tly
if
th
e
y
d
o
n
o
t
co
n
tr
i
b
u
te
to
a
b
etter
r
esu
lt.
A
r
esid
u
al
n
eu
r
al
n
etwo
r
k
is
co
m
p
o
s
ed
o
f
s
ev
er
al
o
f
th
ese
s
o
-
ca
lled
r
esid
u
al
b
lo
ck
s
[
2
3
]
,
[
2
4
]
.
I
n
R
esNets
,
two
p
r
im
ar
y
b
lo
ck
t
y
p
es
co
m
e
in
to
p
lay
b
ased
o
n
wh
eth
e
r
th
e
in
p
u
t
a
n
d
o
u
tp
u
t
d
im
en
s
io
n
s
alig
n
o
r
d
if
f
er
.
B
o
th
b
lo
c
k
t
y
p
es,
n
am
ely
th
e
"id
e
n
tity
b
l
o
ck
"
an
d
th
e
"c
o
n
v
o
lu
tio
n
al
b
lo
ck
,
"
ar
e
in
te
g
r
al
co
m
p
o
n
en
ts
o
f
th
e
im
p
lem
en
t
atio
n
.
T
h
e
id
e
n
tity
b
lo
c
k
is
t
h
e
s
tan
d
ar
d
ch
o
ice
f
o
r
R
esNets
,
em
p
lo
y
ed
wh
en
th
e
in
p
u
t
an
d
o
u
tp
u
t
ac
tiv
atio
n
s
s
h
ar
e
th
e
s
am
e
d
im
en
s
io
n
s
.
B
elo
w,
we
illu
s
tr
ate
a
n
e
x
a
m
p
le
o
f
a
n
id
e
n
tity
b
lo
ck
an
d
co
n
v
o
lu
tio
n
al
b
lo
c
k
in
o
u
r
m
o
d
el
f
ea
tu
r
in
g
th
e
"sh
o
r
tcu
t
p
ath
"
an
d
th
e
"m
ai
n
p
ath
.
"
T
h
ese
two
b
lo
ck
s
to
g
et
h
er
f
o
r
m
th
e
f
o
u
n
d
atio
n
al
a
r
ch
itectu
r
e
o
f
th
e
p
r
o
p
o
s
ed
m
o
d
el
[
2
5
]
.
T
h
e
c
o
n
v
o
lu
ti
o
n
al
b
l
o
ck
ca
p
tu
r
es
in
tr
icate
f
ea
tu
r
es,
w
h
ile
th
e
I
d
en
tity
B
lo
ck
en
s
u
r
es
ef
f
icien
t
g
r
ad
ie
n
t
f
lo
w,
allo
wi
n
g
f
o
r
th
e
ad
e
q
u
ate
tr
ain
in
g
an
d
o
p
tim
izatio
n
o
f
d
ee
p
n
etwo
r
k
s
.
T
h
e
R
E
S
b
lo
ck
in
co
r
p
o
r
ates
b
o
th
th
e
co
n
v
o
l
u
tio
n
b
lo
c
k
an
d
th
e
id
en
tity
B
lo
ck
,
as d
ep
icted
i
n
Fig
u
r
e
3
,
to
f
o
r
m
an
in
te
g
r
al
p
ar
t o
f
th
e
f
in
al
m
o
d
el
ar
ch
itect
u
r
e.
Fig
u
r
e
3
.
Fin
al
m
o
d
el
with
R
esNet
b
lo
ck
,
id
en
tity
b
l
o
ck
,
a
n
d
co
n
v
o
l
u
tio
n
b
lo
ck
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
A
n
ew d
ee
p
lea
r
n
in
g
mo
d
el
b
a
s
ed
o
n
co
n
vo
lu
tio
n
a
l n
eu
r
a
l
n
etw
o
r
k
…
(
A
b
d
elfetta
h
S
o
u
lta
n
a
)
1697
3.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
N
I
n
th
is
s
ec
tio
n
o
f
r
esu
lts
an
d
d
is
cu
s
s
io
n
,
we
h
av
e
p
r
esen
t
ed
th
e
v
ar
i
o
u
s
o
u
tco
m
es
o
f
t
h
e
m
o
d
els
tr
ain
ed
o
n
t
h
e
ey
es_
d
ataset,
e
m
p
h
asizin
g
f
o
u
r
m
etr
ics:
ac
cu
r
ac
y
,
p
r
ec
is
io
n
,
r
ec
all,
an
d
F1
-
s
co
r
e.
Ad
d
itio
n
ally
,
we
h
av
e
d
is
p
lay
ed
th
e
c
o
n
f
u
s
i
o
n
m
atr
ix
f
o
r
th
e
m
o
s
t
p
r
o
f
ici
en
t
m
o
d
els
alo
n
g
with
th
eir
r
e
s
p
ec
tiv
e
m
o
d
el
lo
s
s
an
d
m
o
d
el
ac
cu
r
ac
y
.
Ad
d
itio
n
ally
,
a
d
is
cu
s
s
io
n
o
f
th
ese
r
esu
lts
is
p
r
o
v
id
ed
.
3
.
1
.
P
er
f
o
r
m
a
nce
m
e
t
rics
Sev
er
al
p
er
f
o
r
m
a
n
ce
s
wer
e
u
s
ed
to
ev
alu
ate
th
e
p
r
o
p
o
s
ed
m
o
d
el.
T
h
e
co
n
f
u
s
io
n
m
atr
ix
m
ea
s
u
r
es th
e
q
u
ality
o
f
a
class
if
icatio
n
s
y
s
tem
b
ased
o
n
s
co
r
es
o
f
tr
u
e
p
o
s
itiv
e
(
T
P),
tr
u
e
n
eg
ativ
e
(
T
N)
,
f
alse
p
o
s
it
iv
e
(
FP
)
,
an
d
f
alse
p
o
s
itiv
e
(
FP
)
.
Ad
d
itio
n
ally
,
ac
cu
r
ac
y
,
p
r
ec
is
io
n
,
r
ec
all
an
d
F1
-
s
co
r
e
wer
e
ca
p
tu
r
ed
to
h
a
v
e
a
g
lo
b
al
v
is
o
n
o
f
m
o
d
el
p
e
r
f
o
r
m
an
ce
:
A
c
c
ur
a
c
y
(
%
)
=
+
+
+
+
×
100
(
1
)
R
e
c
a
l
l
(
%
)
=
+
×
100
(
2
)
Pr
e
c
ision
(
%
)
=
+
×
100
(
3
)
F1
-
s
c
or
e
(
%
)
=
2
×
Pr
e
c
isi
o
n
×
R
e
c
a
l
l
Pr
e
c
isi
o
n
+
R
e
c
a
l
l
×
100
(
4
)
3
.
2
.
E
x
perim
ent
re
s
ults
I
n
a
d
d
itio
n
o
f
th
e
p
r
o
p
o
s
ed
m
o
d
el
ev
al
u
atio
n
u
s
in
g
k
ey
m
et
r
ics
s
u
ch
as
ac
c
u
r
ac
y
,
p
r
ec
is
io
n
,
r
ec
all,
an
d
F1
-
s
co
r
e,
a
co
m
p
ar
is
o
n
w
ith
th
e
p
er
f
o
r
m
an
ce
o
f
o
th
er
p
r
e
-
tr
ain
ed
R
esNet
m
o
d
els
is
c
o
n
d
u
cte
d
in
cl
u
d
in
g
R
esNet1
5
2
,
R
es
Net5
0
,
R
esNe
t5
0
v
2
,
R
esNet1
0
1
,
an
d
R
e
s
N
et1
0
1
v
2
.
T
h
e
T
ab
le
1
p
r
o
v
i
d
es
th
e
p
er
f
o
r
m
an
ce
co
m
p
ar
is
o
n
o
f
th
e
p
r
o
p
o
s
ed
m
o
d
el
an
d
th
e
R
esNet
p
r
e
-
tr
ain
ed
m
o
d
els.
Ad
d
itio
n
ally
,
th
e
tab
le
in
clu
d
es
in
f
o
r
m
atio
n
o
n
r
esp
o
n
s
e
tim
e
in
m
illi
s
ec
o
n
d
s
an
d
to
tal
p
ar
am
eter
s
,
p
r
o
v
id
in
g
a
co
m
p
r
eh
en
s
iv
e
o
v
er
v
iew
o
f
co
m
p
u
tatio
n
al
e
f
f
icien
cy
a
n
d
co
m
p
lex
ity
o
f
ea
ch
m
o
d
el.
T
ab
le
1
.
R
ep
o
r
t c
lass
if
icatio
n
A
l
g
o
r
i
t
h
ms
A
c
c
u
r
a
c
y
P
r
e
c
i
s
i
o
n
r
e
c
a
l
l
F1
-
sc
o
r
e
R
e
s
p
o
n
se
t
i
me
(
ms)
To
t
a
l
p
a
r
a
ms
R
e
sN
e
t
1
5
2
5
1
.
9
5
%
7
4
%
5
1
%
3
5
%
1
5
0
5
8
6
3
3
3
4
5
R
e
sN
e
t
5
0
9
5
.
3
1
%
9
6
%
9
6
%
9
6
%
1
1
0
2
3
8
5
0
1
1
3
R
e
sN
e
t
5
0
v
2
9
6
.
4
8
%
9
6
%
9
6
%
9
6
%
1
0
0
2
3
8
2
7
2
0
1
R
e
sN
e
t
1
0
1
9
4
.
1
4
%
9
5
%
9
5
%
9
5
%
1
2
0
4
2
9
2
0
5
7
7
R
e
sN
e
t
1
0
1
v
2
9
4
.
9
2
%
9
5
%
9
5
%
9
5
%
1
2
0
4
2
8
8
8
9
6
1
P
r
o
p
o
se
d
m
o
d
e
l
9
6
.
0
9
%
9
6
%
9
6
%
9
6
%
70
1
1
9
4
1
1
4
T
h
e
p
r
o
p
o
s
ed
m
o
d
el,
R
esNet5
0
,
R
esNet5
0
v
2
,
R
esNet1
0
1
,
an
d
R
esNet1
0
1
v
2
,
e
x
h
ib
it
s
tr
o
n
g
a
n
d
b
alan
ce
d
p
e
r
f
o
r
m
an
ce
ac
r
o
s
s
ac
cu
r
ac
y
,
p
r
ec
is
io
n
,
r
ec
all,
an
d
F1
-
s
co
r
e,
m
ak
in
g
th
em
r
eliab
le
ch
o
ices
f
o
r
d
r
o
wsi
n
ess
d
etec
tio
n
task
.
O
u
r
m
o
d
el
ac
h
iev
e
d
a
h
ig
h
a
cc
u
r
ac
y
o
f
9
6
.
0
9
%,
in
d
icatin
g
its
ca
p
ab
ilit
y
to
class
if
y
in
s
tan
ce
s
co
r
r
ec
tly
.
Ad
d
itio
n
ally
,
th
e
p
r
ec
is
io
n
,
r
ec
a
ll,
an
d
F1
-
s
co
r
e
a
r
e
all
eq
u
all
y
im
p
r
ess
iv
e
with
a
s
co
r
e
o
f
9
6
%,
r
e
f
lectin
g
a
well
-
r
o
u
n
d
e
d
ab
ilit
y
to
co
r
r
ec
tly
class
if
y
p
o
s
itiv
e
in
s
tan
ce
s
in
th
e
d
ataset
.
R
esNet1
5
2
p
er
f
o
r
m
ed
r
elativ
ely
h
ig
h
p
r
ec
is
io
n
o
f
7
4
%,
b
u
t
it
s
u
f
f
er
s
f
r
o
m
a
lo
wer
r
ec
all
an
d
F1
-
s
co
r
e
o
f
5
1
%
an
d
3
5
%,
r
esp
ec
tiv
ely
,
r
esu
ltin
g
in
a
n
o
v
er
all
ac
c
u
r
ac
y
o
f
5
1
.
9
5
%.
T
h
is
s
u
g
g
est
s
th
at
alth
o
u
g
h
t
h
e
alg
o
r
ith
m
is
p
r
ec
is
e
in
class
if
y
in
g
p
o
s
itiv
e
i
n
s
tan
ce
s
,
it
n
ee
d
s
h
elp
to
ca
p
tu
r
e
a
s
u
b
s
tan
tia
l
p
o
r
tio
n
o
f
r
elev
an
t
in
s
tan
ce
s
f
r
o
m
th
e
d
ataset.
I
n
co
n
tr
ast,
R
esNet5
0
,
R
esNet5
0
v
2
,
an
d
R
esNet1
0
1
,
d
em
o
n
s
tr
ate
s
im
ilar
ly
h
ig
h
lev
els
o
f
ac
cu
r
ac
y
r
a
n
g
in
g
f
r
o
m
9
4
.
1
4
%
to
9
6
%
an
d
R
esNet5
0
v
2
o
u
tp
e
r
f
o
r
m
ed
with
an
ac
cu
r
ac
y
o
f
9
6
.
4
8
%.
All
p
r
ec
is
io
n
,
r
ec
all,
a
n
d
F1
-
s
co
r
es
ar
e
all
co
n
s
is
ten
t
at
9
6
%,
in
d
icatin
g
b
alan
ce
d
p
e
r
f
o
r
m
a
n
ce
b
etwe
en
ac
cu
r
ately
class
if
y
in
g
p
o
s
itiv
e
in
s
tan
ce
s
an
d
co
m
p
r
eh
e
n
s
iv
ely
ca
p
tu
r
in
g
r
elev
a
n
t
in
s
tan
ce
s
f
r
o
m
th
e
d
ataset.
C
o
n
s
id
er
in
g
ex
ec
u
tio
n
t
im
e,
t
h
e
p
r
o
p
o
s
ed
m
o
d
el
o
b
tain
ed
a
r
em
ar
k
ab
le
r
esp
o
n
s
e
tim
e
o
f
ju
s
t 7
0
m
illi
s
ec
o
n
d
s
,
a
s
ig
n
if
ican
t
leap
in
s
p
ee
d
co
m
p
ar
ed
t
o
all
o
th
er
m
o
d
els
i
n
th
e
co
m
p
ar
is
o
n
.
T
h
is
tr
an
s
lates
in
to
a
lev
el
o
f
p
r
o
ce
s
s
in
g
s
wif
tn
ess
th
at
is
cr
u
cial
i
n
r
ea
l
-
tim
e
o
r
tim
e
-
s
en
s
i
tiv
e
ap
p
licatio
n
s
,
s
ettin
g
it
a
p
ar
t
as
a
n
ex
ce
p
tio
n
al
c
h
o
ice.
I
n
c
o
n
tr
a
s
t,
wh
ile
o
th
er
m
o
d
els
lik
e
R
esNet5
0
an
d
R
esNet5
0
v
2
ex
h
ib
it
co
m
m
e
n
d
ab
le
p
er
f
o
r
m
an
ce
,
th
eir
r
esp
o
n
s
e
tim
es
o
f
1
1
0
a
n
d
1
0
0
m
i
llis
ec
o
n
d
s
r
esp
ec
tiv
ely
,
w
h
ile
s
till
r
esp
ec
tab
le,
lag
b
eh
i
n
d
th
e
r
ap
id
ity
ac
h
i
ev
ed
b
y
o
u
r
ap
p
r
o
ac
h
.
I
n
a
d
d
itio
n
,
o
u
r
m
o
d
el
h
as
th
e
s
m
allest
n
u
m
b
er
o
f
p
ar
am
eter
s
h
ig
h
lig
h
tin
g
its
ef
f
icien
cy
a
n
d
r
eso
u
r
ce
o
p
tim
izatio
n
.
T
h
is
u
n
d
er
s
co
r
es
th
e
h
ig
h
eq
u
ilib
r
iu
m
s
tr
u
ck
b
y
th
e
p
r
o
p
o
s
ed
m
o
d
el,
s
ea
m
less
ly
m
ar
r
y
in
g
h
ig
h
p
er
f
o
r
m
a
n
ce
with
o
p
tim
al
r
esp
o
n
s
e
tim
e,
r
en
d
e
r
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
.
3
,
Ma
r
ch
20
2
5
:
1
692
-
1
7
0
1
1698
it a
h
ig
h
ly
p
r
o
m
is
in
g
ca
n
d
id
at
e
f
o
r
ap
p
licatio
n
s
wh
er
e
s
p
ee
d
is
cr
itical.
T
h
e
Fig
u
r
e
4
illu
s
tr
ates th
e
m
o
d
el
lo
s
s
an
d
ac
cu
r
ac
y
f
o
r
th
e
two
m
o
d
els th
at
ex
h
ib
it th
e
m
o
s
t sig
n
if
ican
t m
etr
ic
v
alu
es.
Fig
u
r
e
4
.
M
o
d
el
lo
s
s
an
d
m
o
d
el
ac
cu
r
ac
y
T
h
e
g
r
ap
h
s
h
o
ws
th
at
th
e
p
r
ec
is
io
n
an
d
lo
s
s
o
f
o
u
r
m
o
d
el
h
av
e
a
lear
n
in
g
p
r
o
g
r
ess
io
n
o
v
er
3
0
ep
o
ch
s
.
T
h
e
m
o
d
el'
s
tr
ain
in
g
ac
cu
r
ac
y
s
tar
ts
at
ar
o
u
n
d
7
9
.
8
%
an
d
s
tead
ily
im
p
r
o
v
es,
r
ea
c
h
in
g
clo
s
e
to
1
0
0
%
ac
cu
r
ac
y
b
y
th
e
1
3
th
ep
o
c
h
.
T
h
e
v
alid
atio
n
ac
c
u
r
ac
y
,
wh
i
ch
b
e
g
in
s
at
a
p
p
r
o
x
im
ately
4
9
.
2
%,
u
n
d
er
g
o
es
a
s
u
b
s
tan
tial
im
p
r
o
v
em
e
n
t
an
d
s
tab
ilizes
at
a
h
ig
h
lev
el
o
f
9
6
%
ac
cu
r
ac
y
af
te
r
a
f
ew
ep
o
ch
s
.
T
h
is
tr
e
n
d
in
d
icate
s
th
at
th
e
m
o
d
el
ef
f
ec
tiv
ely
lear
n
s
f
r
o
m
t
h
e
tr
ain
in
g
d
ata
an
d
g
e
n
er
a
lizes
well
to
th
e
v
alid
atio
n
d
ata.
Ho
wev
er
,
th
e
r
elativ
ely
s
m
all
g
ap
b
etwe
en
th
e
t
r
ain
in
g
a
n
d
v
alid
atio
n
ac
cu
r
ac
y
s
u
g
g
ests
th
e
m
o
d
el
c
o
u
ld
b
e
well
-
r
eg
u
lar
ized
a
n
d
c
o
n
tin
u
e
to
p
er
f
o
r
m
well
o
n
n
ew,
u
n
s
ee
n
d
ata.
T
h
e
p
r
o
v
id
ed
m
o
d
el
ac
cu
r
ac
y
a
n
d
m
o
d
el
lo
s
s
f
o
r
R
esNetNet5
0
v
2
illu
s
tr
ate
th
e
lear
n
in
g
t
r
ajec
to
r
y
o
f
a
n
eu
r
al
n
etwo
r
k
ac
r
o
s
s
2
5
ep
o
ch
s
.
I
n
itially
,
th
e
m
o
d
el
ac
h
iev
es
a
tr
ain
in
g
ac
c
u
r
ac
y
o
f
8
6
.
5
%,
wh
ich
s
u
b
s
eq
u
en
tly
ad
v
an
ce
s
co
n
s
is
ten
tly
,
cu
lm
in
atin
g
in
a
n
ac
cu
r
ac
y
o
f
a
r
o
u
n
d
9
6
.
4
8
%
b
y
th
e
2
5
th
e
p
o
ch
.
Me
an
wh
i
le,
th
e
v
alid
atio
n
ac
cu
r
ac
y
s
tar
ts
at
5
8
.
6
%
an
d
d
is
p
lay
s
s
o
m
e
o
s
cillatio
n
s
in
t
h
e
ea
r
ly
e
p
o
ch
s
.
H
o
wev
er
,
th
e
f
lu
ctu
atio
n
s
s
tab
ilize,
lead
in
g
to
a
c
o
n
v
e
r
g
en
ce
o
f
ar
o
u
n
d
9
6
.
4
8
%
as
th
e
tr
a
in
in
g
p
r
o
g
r
ess
es.
T
h
is
co
n
v
e
r
g
en
ce
s
ig
n
if
ies
th
e
m
o
d
el'
s
r
o
b
u
s
t
ca
p
a
city
to
g
en
er
alize
ef
f
ec
tiv
ely
to
n
ew
an
d
u
n
s
ee
n
d
ata.
T
h
e
co
n
f
u
s
i
o
n
m
atr
ices
f
o
r
th
e
two
m
o
d
el
s
ar
e
also
d
is
p
lay
ed
in
th
e
F
ig
u
r
e
5
.
Fig
u
r
e
5
.
C
o
n
f
u
s
io
n
m
atr
i
x
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
A
n
ew d
ee
p
lea
r
n
in
g
mo
d
el
b
a
s
ed
o
n
co
n
vo
lu
tio
n
a
l n
eu
r
a
l
n
etw
o
r
k
…
(
A
b
d
elfetta
h
S
o
u
lta
n
a
)
1699
T
h
is
m
atr
ix
s
h
o
ws
th
at
t
h
e
p
r
o
p
o
s
ed
m
o
d
el
p
er
f
o
r
m
s
q
u
ite
well
o
v
er
all,
with
h
ig
h
v
al
u
es
o
n
th
e
d
iag
o
n
al
i
n
d
icatin
g
ac
cu
r
ate
p
r
ed
ictio
n
s
.
T
h
e
m
o
d
el
s
ee
m
s
s
lig
h
tly
m
o
r
e
lik
ely
to
m
is
class
if
y
"c
lo
s
ed
"
in
s
tan
ce
s
as
"o
p
en
"
th
a
n
th
e
r
ev
er
s
e.
T
h
e
"
R
es
N
et5
0
v
2
"
m
o
d
el
also
p
er
f
o
r
m
s
well,
p
ar
ticu
lar
ly
in
th
e
"c
lo
s
ed
"
class
,
wh
er
e
it
ac
h
iev
es
p
er
f
ec
t
ac
cu
r
a
cy
.
Ho
we
v
e
r
,
th
er
e
ar
e
s
o
m
e
in
s
tan
ce
s
wh
er
e
it
m
is
class
if
ies
"o
p
en
"
in
s
tan
ce
s
as
"c
lo
s
ed
,
"
lead
in
g
to
a
n
o
n
-
ze
r
o
v
alu
e
in
th
e
to
p
r
ig
h
t
co
r
n
er
.
I
n
co
n
cl
u
s
io
n
,
b
o
th
m
o
d
els
p
er
f
o
r
m
r
ea
s
o
n
ab
ly
well
in
th
e
b
in
ar
y
class
if
icatio
n
task
b
u
t
h
av
e
s
lig
h
tly
d
if
f
er
en
t
m
is
class
if
icatio
n
p
atter
n
s
.
T
h
e
p
r
o
p
o
s
ed
m
o
d
el
h
as
a
b
a
lan
ce
d
m
is
class
if
icatio
n
p
atter
n
,
wh
ile
R
esNet5
0
v
2
ex
ce
ls
in
th
e
clo
s
ed
class
b
u
t
h
as
s
o
m
e
m
is
class
if
icatio
n
s
in
th
e
o
p
e
n
ca
teg
o
r
y
.
I
n
c
o
n
clu
s
io
n
,
b
o
th
m
o
d
els,
th
e
p
r
o
p
o
s
ed
m
o
d
el
an
d
R
esNet5
0
v
2
,
d
em
o
n
s
tr
ate
p
r
o
m
is
in
g
p
er
f
o
r
m
a
n
ce
f
o
r
d
r
o
w
s
in
ess
d
etec
tio
n
.
T
h
e
p
r
o
p
o
s
ed
m
o
d
el
s
tan
d
s
o
u
t
with
its
f
aster
r
esp
o
n
s
e
tim
e
an
d
r
ed
u
ce
d
n
u
m
b
e
r
o
f
p
ar
a
m
eter
s
,
m
ak
in
g
it
a
p
r
ef
er
r
ed
ch
o
ice
f
o
r
r
ea
l
-
tim
e
d
r
o
wsi
n
ess
d
etec
tio
n
in
d
r
iv
er
s
.
T
h
e
p
r
o
p
o
s
ed
m
o
d
el
d
em
o
n
s
tr
ates
ex
ce
p
tio
n
al
p
e
r
f
o
r
m
a
n
ce
o
n
th
e
p
r
im
ar
y
d
ataset,
ac
h
iev
in
g
an
ac
c
u
r
ac
y
o
f
9
6
.
0
9
%.
T
h
is
h
ig
h
lev
el
o
f
ac
cu
r
ac
y
is
s
u
p
p
o
r
ted
b
y
v
ar
io
u
s
ev
alu
atio
n
m
etr
ics,
in
clu
d
in
g
p
r
ec
is
io
n
,
r
ec
all,
F1
s
co
r
e,
an
d
a
d
etaile
d
co
n
f
u
s
io
n
m
atr
ix
.
T
h
e
s
tab
i
lity
o
f
th
e
tr
ain
i
n
g
p
r
o
ce
s
s
,
as e
v
id
en
ce
d
b
y
c
o
n
s
is
ten
t lo
s
s
an
d
ac
cu
r
ac
y
p
r
o
f
ile
s
,
f
u
r
th
er
v
alid
ates th
e
m
o
d
el'
s
ef
f
ec
tiv
en
ess
.
I
n
ad
d
itio
n
to
th
ese
m
etr
ics,
a
co
m
p
ar
ativ
e
a
n
aly
s
is
was
c
o
n
d
u
cte
d
with
s
ev
er
al
p
r
e
-
t
r
a
in
ed
C
NN
m
o
d
els
b
ased
o
n
th
e
R
esNet
ar
ch
itectu
r
e.
T
h
e
r
esu
lt
s
o
f
th
is
co
m
p
ar
is
o
n
r
ev
ea
l
th
at
o
u
r
p
r
o
p
o
s
ed
1
8
-
lay
er
m
o
d
el
s
ig
n
if
ican
tly
o
u
t
p
er
f
o
r
m
s
th
ese
alter
n
ativ
es,
ac
h
iev
in
g
a
co
m
p
ellin
g
le
v
el
o
f
ac
cu
r
a
cy
.
T
h
is
estab
lis
h
es
o
u
r
m
o
d
el'
s
s
u
p
er
io
r
ef
f
icac
y
in
th
e
d
o
m
ai
n
o
f
d
r
iv
er
d
r
o
wsi
n
ess
d
etec
tio
n
.
T
o
ass
ess
th
e
m
o
d
el'
s
r
o
b
u
s
tn
ess
an
d
g
en
er
aliza
tio
n
ca
p
a
b
ilit
ies,
we
co
n
d
u
cted
a
s
ec
o
n
d
ar
y
ev
alu
atio
n
u
s
in
g
an
alter
n
ati
v
e
Kag
g
le
d
ataset
co
m
p
r
is
in
g
1
,
8
7
0
im
ag
es
d
ep
i
ctin
g
b
o
th
o
p
e
n
an
d
clo
s
ed
e
y
es.
R
em
ar
k
ab
ly
,
th
e
m
o
d
el
ac
h
iev
ed
an
ac
cu
r
ac
y
o
f
9
8
.
2
8
%
o
n
th
is
in
d
ep
e
n
d
e
n
t
d
ataset.
T
h
is
im
p
r
ess
iv
e
p
e
r
f
o
r
m
a
n
ce
o
n
a
d
if
f
e
r
en
t
d
ata
s
et
u
n
d
er
s
co
r
es
th
e
m
o
d
el'
s
r
o
b
u
s
tn
ess
an
d
b
r
o
a
d
ap
p
licab
ilit
y
,
m
ak
in
g
it a
r
elia
b
le
s
o
lu
tio
n
f
o
r
v
a
r
io
u
s
r
ea
l
-
w
o
r
ld
s
ce
n
ar
io
s
.
4.
CO
NCLU
SI
O
N
I
n
co
n
clu
s
io
n
,
a
d
d
r
ess
in
g
th
e
cr
itical
is
s
u
e
o
f
d
r
o
wsy
d
r
iv
in
g
in
th
e
r
ea
lm
o
f
d
r
iv
er
s
af
ety
d
em
an
d
s
in
n
o
v
ativ
e
s
o
lu
tio
n
s
,
p
ar
ticu
la
r
ly
th
o
s
e
g
r
o
u
n
d
e
d
in
r
ea
l
-
tim
e
m
o
n
ito
r
in
g
s
y
s
tem
s
.
T
h
is
r
e
s
ea
r
ch
in
tr
o
d
u
ce
s
a
g
r
o
u
n
d
b
r
ea
k
in
g
ap
p
r
o
ac
h
u
tili
zin
g
co
n
v
o
lu
tio
n
al
n
eu
r
al
n
et
wo
r
k
s
(
C
NNs)
an
d
R
esid
u
al
b
lo
ck
s
to
p
r
o
m
p
tly
d
etec
t
d
r
iv
er
d
r
o
wsi
n
ess
.
T
h
e
f
u
s
io
n
o
f
C
NNs
with
R
esid
u
al
u
n
its
,
in
co
r
p
o
r
atin
g
cr
u
cial
an
aly
tical
co
m
p
o
n
en
ts
,
estab
lis
h
es
a
r
o
b
u
s
t
m
eth
o
d
o
lo
g
y
f
o
r
ad
d
r
es
s
in
g
th
is
cr
itical
ch
allen
g
e.
Ou
r
d
ev
is
ed
s
t
r
ateg
y
ex
h
ib
its
r
em
ar
k
a
b
le
ef
f
icac
y
,
b
o
asti
n
g
an
i
m
p
r
ess
iv
e
9
6
.
0
9
%
ac
cu
r
ac
y
wh
e
n
test
ed
o
n
th
e
wid
ely
ac
ce
s
s
ib
le
ey
e_
d
ataset.
No
tab
ly
,
o
u
r
p
r
o
p
o
s
ed
ap
p
r
o
ac
h
d
e
m
o
n
s
tr
a
tes
an
ex
ce
p
tio
n
al
r
esp
o
n
s
e
tim
e
o
f
7
0
m
s
,
u
n
d
er
s
co
r
i
n
g
its
ef
f
icien
cy
in
r
ea
l
-
tim
e
ap
p
lic
atio
n
s
.
F
u
r
th
er
m
o
r
e,
o
u
r
in
v
esti
g
atio
n
ex
ten
d
s
to
th
e
ex
p
lo
r
atio
n
o
f
d
iv
e
r
s
e
p
r
e
-
tr
ai
n
ed
r
esid
u
al
n
eu
r
al
n
etwo
r
k
s
,
in
clu
d
in
g
R
esNet1
5
2
an
d
R
esNet5
0
,
em
p
h
asizin
g
th
e
ad
ap
tab
ilit
y
o
f
o
u
r
m
eth
o
d
o
lo
g
y
.
L
o
o
k
in
g
f
o
r
war
d
,
o
u
r
e
n
v
is
io
n
ed
tr
ajec
to
r
y
in
v
o
lv
es
th
e
s
e
am
less
in
teg
r
atio
n
o
f
o
u
r
d
r
o
wsi
n
ess
d
etec
tio
n
s
y
s
tem
i
n
to
a
co
m
p
r
eh
en
s
iv
e
g
lo
b
al
d
r
iv
er
in
atten
tio
n
d
etec
tio
n
s
y
s
tem
.
T
h
is
ex
p
an
d
e
d
s
y
s
tem
tr
an
s
ce
n
d
s
th
e
s
in
g
u
lar
f
o
c
u
s
o
n
d
r
o
wsi
n
ess
,
en
co
m
p
ass
in
g
f
ac
to
r
s
s
u
c
h
as
f
atig
u
e,
n
eg
ativ
e
em
o
tio
n
s
,
a
n
d
d
r
iv
er
d
is
tr
ac
tio
n
.
B
y
b
r
o
ad
en
in
g
its
s
co
p
e,
we
asp
ir
e
to
estab
lis
h
a
m
o
r
e
r
esil
ien
t
an
d
p
r
o
ac
tiv
e
a
p
p
r
o
ac
h
to
b
o
ls
ter
in
g
d
r
iv
er
s
af
ety
an
d
atten
tiv
e
n
ess
o
n
th
e
r
o
ad
s
.
T
h
is
h
o
lis
tic
s
tr
ateg
y
,
co
u
p
led
with
a
n
im
p
r
ess
iv
e
r
esp
o
n
s
e
tim
e,
r
ef
lects
o
u
r
c
o
m
m
itm
en
t
t
o
ad
v
an
cin
g
n
o
t
o
n
ly
t
h
e
f
ield
o
f
d
r
o
wsy
d
r
iv
i
n
g
d
etec
tio
n
b
u
t
also
co
n
tr
i
b
u
tin
g
t
o
a
b
r
o
ad
er
f
r
am
ew
o
r
k
f
o
r
c
o
m
p
r
eh
en
s
iv
e
d
r
iv
er
s
af
ety
s
o
lu
tio
n
s
.
Fu
tu
r
e
r
esear
ch
ca
n
b
u
ild
u
p
o
n
th
ese
f
in
d
in
g
s
b
y
ex
p
lo
r
i
n
g
r
ea
l
-
tim
e
im
p
lem
e
n
tatio
n
,
en
h
an
cin
g
d
ataset
d
iv
er
s
ity
,
in
teg
r
atin
g
m
u
ltimo
d
al
d
ata,
in
v
esti
g
atin
g
tr
an
s
f
er
lear
n
in
g
,
an
d
ev
alu
atin
g
lo
n
g
-
ter
m
p
er
f
o
r
m
an
ce
.
T
h
is
h
o
lis
tic
s
tr
ateg
y
,
co
u
p
led
with
a
n
im
p
r
e
s
s
iv
e
r
esp
o
n
s
e
tim
e,
r
ef
lects
o
u
r
co
m
m
itm
en
t
to
ad
v
an
cin
g
n
o
t
o
n
l
y
th
e
f
ield
o
f
d
r
o
wsy
d
r
iv
in
g
d
etec
tio
n
b
u
t
also
co
n
tr
ib
u
tin
g
to
a
b
r
o
a
d
er
f
r
am
ewo
r
k
f
o
r
co
m
p
r
eh
e
n
s
iv
e
d
r
iv
e
r
s
af
ety
s
o
lu
tio
n
s
.
RE
F
E
R
E
NC
E
S
[
1
]
S
.
S
t
e
r
n
l
u
n
d
,
J.
S
t
r
a
n
d
r
o
t
h
,
M
.
R
i
z
z
i
,
A
.
Li
e
,
a
n
d
C
.
Ti
n
g
v
a
l
l
,
“
Th
e
e
f
f
e
c
t
i
v
e
n
e
ss o
f
l
a
n
e
d
e
p
a
r
t
u
r
e
w
a
r
n
i
n
g
s
y
st
e
ms
-
a
r
e
d
u
c
t
i
o
n
i
n
r
e
a
l
-
w
o
r
l
d
p
a
ss
e
n
g
e
r
c
a
r
i
n
j
u
r
y
c
r
a
sh
e
s,
”
T
r
a
f
f
i
c
I
n
j
u
r
y
Pre
v
e
n
t
i
o
n
,
v
o
l
.
1
8
,
n
o
.
2
,
p
p
.
2
2
5
–
2
2
9
,
F
e
b
.
2
0
1
7
,
d
o
i
:
1
0
.
1
0
8
0
/
1
5
3
8
9
5
8
8
.
2
0
1
6
.
1
2
3
0
6
7
2
.
[
2
]
L.
Y
u
e
,
M
.
A
b
d
e
l
-
A
t
y
,
Y
.
W
u
,
J
.
U
g
a
n
,
a
n
d
C
.
Y
u
a
n
,
“
Ef
f
e
c
t
s
o
f
f
o
r
w
a
r
d
c
o
l
l
i
si
o
n
w
a
r
n
i
n
g
t
e
c
h
n
o
l
o
g
y
i
n
d
i
f
f
e
r
e
n
t
p
r
e
-
c
r
a
sh
sce
n
a
r
i
o
s,
”
T
ra
n
s
p
o
r
t
a
t
i
o
n
R
e
se
a
rc
h
P
a
rt
F:
T
r
a
f
f
i
c
Ps
y
c
h
o
l
o
g
y
a
n
d
Be
h
a
v
i
o
u
r
,
v
o
l
.
7
6
,
p
p
.
3
3
6
–
3
5
2
,
Ja
n
.
2
0
2
1
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
t
r
f
.
2
0
2
0
.
1
2
.
0
0
4
.
[
3
]
L.
R
.
W
i
l
l
i
a
ms,
D
.
R
.
D
a
v
i
e
s,
K
.
T
h
i
e
l
e
,
J
.
R
.
D
a
v
i
d
s
o
n
,
a
n
d
A
.
W
.
M
a
c
Le
a
n
,
“
Y
o
u
n
g
d
r
i
v
e
r
s’
p
e
r
c
e
p
t
i
o
n
s
o
f
c
u
l
p
a
b
i
l
i
t
y
o
f
sl
e
ep
-
d
e
p
r
i
v
e
d
v
e
r
s
u
s
d
r
i
n
k
i
n
g
d
r
i
v
e
r
s
,
”
J
o
u
rn
a
l
o
f
S
a
f
e
t
y
Re
s
e
a
rc
h
,
v
o
l
.
4
3
,
n
o
.
2
,
p
p
.
1
1
5
–
1
2
2
,
2
0
1
2
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
j
sr
.
2
0
1
2
.
0
2
.
0
0
2
.
[
4
]
“
D
r
o
w
sy
d
r
i
v
i
n
g
,
”
N
H
TSA
.
A
c
c
e
sse
d
:
A
u
g
.
1
3
,
2
0
2
3
.
[
O
n
l
i
n
e
]
.
A
v
a
i
l
a
b
l
e
:
h
t
t
p
s
:
/
/
w
w
w
.
n
h
t
sa.
g
o
v
/
r
i
s
k
y
-
d
r
i
v
i
n
g
/
d
r
o
w
sy
-
d
r
i
v
i
n
g
[
5
]
B
.
C
.
T
e
f
f
t
,
“
P
r
e
v
a
l
e
n
c
e
o
f
m
o
t
o
r
v
e
h
i
c
l
e
c
r
a
s
h
e
s
i
n
v
o
l
v
i
n
g
d
r
o
w
s
y
d
r
i
v
e
r
s,
U
n
i
t
e
d
S
t
a
t
e
s,
1
9
9
9
-
2
0
0
8
,
”
Ac
c
i
d
e
n
t
A
n
a
l
y
s
i
s
a
n
d
Pre
v
e
n
t
i
o
n
,
v
o
l
.
4
5
,
p
p
.
1
8
0
–
1
8
6
,
M
a
r
.
2
0
1
2
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
a
a
p
.
2
0
1
1
.
0
5
.
0
2
8
.
[
6
]
Y
.
Le
c
u
n
,
Y
.
B
e
n
g
i
o
,
a
n
d
G
.
H
i
n
t
o
n
,
“
D
e
e
p
l
e
a
r
n
i
n
g
,
”
N
a
t
u
re
,
v
o
l
.
5
2
1
,
n
o
.
7
5
5
3
,
p
p
.
4
3
6
–
4
4
4
,
2
0
1
5
,
d
o
i
:
1
0
.
1
0
3
8
/
n
a
t
u
r
e
1
4
5
3
9
.
[
7
]
K
.
H
e
,
X
.
Z
h
a
n
g
,
S
.
R
e
n
,
a
n
d
J
.
S
u
n
,
“
D
e
e
p
r
e
si
d
u
a
l
l
e
a
r
n
i
n
g
f
o
r
i
ma
g
e
r
e
c
o
g
n
i
t
i
o
n
,
”
i
n
2
0
1
6
I
EEE
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
V
P
R)
,
I
EEE,
J
u
n
.
2
0
1
6
,
p
p
.
7
7
0
–
7
7
8
.
d
o
i
:
1
0
.
1
1
0
9
/
C
V
P
R
.
2
0
1
6
.
9
0
.
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
.
3
,
Ma
r
ch
20
2
5
:
1
692
-
1
7
0
1
1700
[
8
]
M
.
S
h
a
f
i
q
a
n
d
Z.
G
u
,
“
D
e
e
p
r
e
si
d
u
a
l
l
e
a
r
n
i
n
g
f
o
r
i
ma
g
e
r
e
c
o
g
n
i
t
i
o
n
:
a
su
r
v
e
y
,
”
A
p
p
l
i
e
d
S
c
i
e
n
c
e
s
(
S
w
i
t
z
e
rl
a
n
d
)
,
v
o
l
.
1
2
,
n
o
.
1
8
,
p
.
8
9
7
2
,
S
e
p
.
2
0
2
2
,
d
o
i
:
1
0
.
3
3
9
0
/
a
p
p
1
2
1
8
8
9
7
2
.
[
9
]
F
.
H
u
a
n
g
,
J.
T.
A
s
h
,
J.
L
a
n
g
f
o
r
d
,
a
n
d
R
.
E.
S
c
h
a
p
i
r
e
,
“
L
e
a
r
n
i
n
g
d
e
e
p
R
e
sN
e
t
b
l
o
c
k
s
s
e
q
u
e
n
t
i
a
l
l
y
u
si
n
g
b
o
o
s
t
i
n
g
t
h
e
o
r
y
,
”
i
n
3
5
t
h
I
n
t
e
rn
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
Ma
c
h
i
n
e
L
e
a
r
n
i
n
g
,
I
C
ML
2
0
1
8
,
2
0
1
8
,
p
p
.
3
2
7
2
–
3
2
9
0
.
[
1
0
]
Z.
W
u
,
C
.
S
h
e
n
,
a
n
d
A
.
v
a
n
d
e
n
H
e
n
g
e
l
,
“
W
i
d
e
r
o
r
d
e
e
p
e
r
:
r
e
v
i
si
t
i
n
g
t
h
e
R
e
sN
e
t
m
o
d
e
l
f
o
r
v
i
s
u
a
l
r
e
c
o
g
n
i
t
i
o
n
,
”
Pa
t
t
e
r
n
Re
c
o
g
n
i
t
i
o
n
,
v
o
l
.
9
0
,
p
p
.
1
1
9
–
1
3
3
,
J
u
n
.
2
0
1
9
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
p
a
t
c
o
g
.
2
0
1
9
.
0
1
.
0
0
6
.
[
1
1
]
V
.
R
.
R
e
d
d
y
C
h
i
r
r
a
,
S
.
R
.
U
y
y
a
l
a
,
a
n
d
V
.
K
.
K
i
sh
o
r
e
K
o
l
l
i
,
“
D
e
e
p
C
N
N
:
a
ma
c
h
i
n
e
l
e
a
r
n
i
n
g
a
p
p
r
o
a
c
h
f
o
r
d
r
i
v
e
r
d
r
o
w
s
i
n
e
ss
d
e
t
e
c
t
i
o
n
b
a
s
e
d
o
n
e
y
e
s
t
a
t
e
,
”
R
e
v
u
e
d
’
I
n
t
e
l
l
i
g
e
n
c
e
Art
i
f
i
c
i
e
l
l
e
,
v
o
l
.
3
3
,
n
o
.
6
,
p
p
.
4
6
1
–
4
6
6
,
D
e
c
.
2
0
1
9
,
d
o
i
:
1
0
.
1
8
2
8
0
/
r
i
a
.
3
3
0
6
0
9
.
[
1
2
]
A
.
U
.
I
.
R
a
f
i
d
,
A
.
I
.
C
h
o
w
d
h
u
r
y
,
A
.
R
.
N
i
l
o
y
,
a
n
d
N
.
S
h
a
r
m
i
n
,
“
A
d
e
e
p
l
e
a
r
n
i
n
g
b
a
se
d
a
p
p
r
o
a
c
h
f
o
r
r
e
a
l
-
t
i
m
e
d
r
i
v
e
r
d
r
o
w
si
n
e
ss
d
e
t
e
c
t
i
o
n
,
”
i
n
2
0
2
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
o
n
E
l
e
c
t
r
i
c
a
l
E
n
g
i
n
e
e
r
i
n
g
a
n
d
I
n
f
o
rm
a
t
i
o
n
a
n
d
C
o
m
m
u
n
i
c
a
t
i
o
n
T
e
c
h
n
o
l
o
g
y
,
I
C
EEICT
2
0
2
1
,
I
EEE,
N
o
v
.
2
0
2
1
,
p
p
.
1
–
5
.
d
o
i
:
1
0
.
1
1
0
9
/
I
C
EEI
C
T
5
3
9
0
5
.
2
0
2
1
.
9
6
6
7
9
4
4
.
[
1
3
]
M
.
F
.
S
h
a
k
e
e
l
,
N
.
A
.
B
a
j
w
a
,
A
.
M
.
A
n
w
a
a
r
,
A
.
S
o
h
a
i
l
,
A
.
K
h
a
n
,
a
n
d
H
a
r
o
o
n
-
ur
-
R
a
s
h
i
d
,
“
D
e
t
e
c
t
i
n
g
d
r
i
v
e
r
d
r
o
w
si
n
e
ss i
n
r
e
a
l
t
i
m
e
t
h
r
o
u
g
h
d
e
e
p
l
e
a
r
n
i
n
g
b
a
se
d
o
b
j
e
c
t
d
e
t
e
c
t
i
o
n
,
”
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
s
e
ri
e
s
L
e
c
t
u
re
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
1
5
0
6
LN
C
S
,
2
0
1
9
,
p
p
.
2
8
3
–
2
9
6
.
d
o
i
:
1
0
.
1
0
0
7
/
9
7
8
-
3
-
030
-
2
0
5
2
1
-
8
_
2
4
.
[
1
4
]
S
.
P
.
R
a
j
a
mo
h
a
n
a
,
E.
G
.
R
a
d
h
i
k
a
,
S
.
P
r
i
y
a
,
a
n
d
S
.
S
a
n
g
e
e
t
h
a
,
“
D
r
i
v
e
r
d
r
o
w
si
n
e
ss
d
e
t
e
c
t
i
o
n
s
y
st
e
m
u
s
i
n
g
h
y
b
r
i
d
a
p
p
r
o
a
c
h
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
a
n
d
b
i
d
i
r
e
c
t
i
o
n
a
l
l
o
n
g
s
h
o
r
t
t
e
r
m
m
e
m
o
r
y
(
C
N
N
_
B
I
LSTM
)
,
”
M
a
t
e
r
i
a
l
s
T
o
d
a
y
:
Pr
o
c
e
e
d
i
n
g
s
,
v
o
l
.
4
5
,
p
p
.
2
8
9
7
–
2
9
0
1
,
2
0
2
1
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
ma
t
p
r
.
2
0
2
0
.
1
1
.
8
9
8
.
[
1
5
]
J.
Jo
se,
A
.
J
,
K
.
R
a
i
m
o
n
d
,
S
.
V
i
n
c
e
n
t
,
a
n
d
J
.
E
u
n
i
c
e
R
,
“
S
l
e
e
p
y
W
h
e
e
l
s
:
a
n
e
n
s
e
mb
l
e
mo
d
e
l
f
o
r
d
r
o
w
si
n
e
ss
d
e
t
e
c
t
i
o
n
l
e
a
d
i
n
g
t
o
a
c
c
i
d
e
n
t
p
r
e
v
e
n
t
i
o
n
,
”
J
o
u
r
n
a
l
o
f
A
u
t
o
n
o
m
o
u
s In
t
e
l
l
i
g
e
n
c
e
,
v
o
l
.
7
,
n
o
.
3
,
Ja
n
.
2
0
2
4
,
d
o
i
:
1
0
.
3
2
6
2
9
/
j
a
i
.
v
7
i
3
.
1
1
1
7
.
[
1
6
]
S
.
E.
B
e
k
h
o
u
c
h
e
,
Y
.
R
u
i
c
h
e
k
,
a
n
d
F
.
D
o
r
n
a
i
k
a
,
“
D
r
i
v
e
r
d
r
o
w
s
i
n
e
ss
d
e
t
e
c
t
i
o
n
i
n
v
i
d
e
o
s
e
q
u
e
n
c
e
s
u
s
i
n
g
h
y
b
r
i
d
sel
e
c
t
i
o
n
o
f
d
e
e
p
f
e
a
t
u
r
e
s,
”
K
n
o
w
l
e
d
g
e
-
B
a
se
d
S
y
s
t
e
m
s
,
v
o
l
.
2
5
2
,
p
.
1
0
9
4
3
6
,
S
e
p
.
2
0
2
2
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
k
n
o
s
y
s.
2
0
2
2
.
1
0
9
4
3
6
.
[
1
7
]
E.
C
i
v
i
k
a
n
d
U
.
Y
u
z
g
e
c
,
“
R
e
a
l
-
t
i
m
e
d
r
i
v
e
r
f
a
t
i
g
u
e
d
e
t
e
c
t
i
o
n
s
y
s
t
e
m
w
i
t
h
d
e
e
p
l
e
a
r
n
i
n
g
o
n
a
l
o
w
-
c
o
st
e
m
b
e
d
d
e
d
s
y
st
e
m,
”
Mi
c
r
o
p
ro
c
e
ss
o
rs
a
n
d
M
i
c
r
o
sys
t
e
m
s
,
v
o
l
.
9
9
,
p
.
1
0
4
8
5
1
,
Ju
n
.
2
0
2
3
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
m
i
c
p
r
o
.
2
0
2
3
.
1
0
4
8
5
1
.
[
1
8
]
R
.
F
l
o
r
e
z
,
F
.
P
a
l
o
m
i
n
o
-
Q
u
i
s
p
e
,
R
.
J
.
C
o
a
q
u
i
r
a
-
C
a
st
i
l
l
o
,
J.
C
.
H
e
r
r
e
r
a
-
Le
v
a
n
o
,
T.
P
a
i
x
ã
o
,
a
n
d
A
.
B
.
A
l
v
a
r
e
z
,
“
A
C
N
N
-
b
a
s
e
d
a
p
p
r
o
a
c
h
f
o
r
d
r
i
v
e
r
d
r
o
w
si
n
e
s
s
d
e
t
e
c
t
i
o
n
b
y
r
e
a
l
-
t
i
m
e
e
y
e
st
a
t
e
i
d
e
n
t
i
f
i
c
a
t
i
o
n
,
”
A
p
p
l
i
e
d
S
c
i
e
n
c
e
s
(
S
w
i
t
ze
rl
a
n
d
)
,
v
o
l
.
1
3
,
n
o
.
1
3
,
p
.
7
8
4
9
,
Ju
l
.
2
0
2
3
,
d
o
i
:
1
0
.
3
3
9
0
/
a
p
p
1
3
1
3
7
8
4
9
.
[
1
9
]
A
.
C
.
P
h
a
n
,
T
.
N
.
Tr
i
e
u
,
a
n
d
T.
C
.
P
h
a
n
,
“
D
r
i
v
e
r
d
r
o
w
si
n
e
ss
d
e
t
e
c
t
i
o
n
a
n
d
s
mart
a
l
e
r
t
i
n
g
u
s
i
n
g
d
e
e
p
l
e
a
r
n
i
n
g
a
n
d
I
o
T,
”
I
n
t
e
r
n
e
t
o
f
T
h
i
n
g
s
(
N
e
t
h
e
rl
a
n
d
s)
,
v
o
l
.
2
2
,
p
.
1
0
0
7
0
5
,
J
u
l
.
2
0
2
3
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
i
o
t
.
2
0
2
3
.
1
0
0
7
0
5
.
[
2
0
]
S
.
R
a
j
u
,
“
y
a
w
n
_
e
y
e
_
d
a
t
a
s
e
t
_
n
e
w
.
”
A
c
c
e
ss
e
d
:
A
u
g
.
0
6
,
2
0
2
3
.
[
O
n
l
i
n
e
]
.
A
v
a
i
l
a
b
l
e
:
h
t
t
p
s
:
/
/
w
w
w
.
k
a
g
g
l
e
.
c
o
m
/
d
a
t
a
se
t
s/
s
e
r
e
n
a
r
a
j
u
/
y
a
w
n
-
e
y
e
-
d
a
t
a
set
-
n
e
w
[
2
1
]
A
.
K
h
a
n
,
A
.
S
o
h
a
i
l
,
U
.
Za
h
o
o
r
a
,
a
n
d
A
.
S
.
Q
u
r
e
s
h
i
,
“
A
s
u
r
v
e
y
o
f
t
h
e
r
e
c
e
n
t
a
r
c
h
i
t
e
c
t
u
r
e
s o
f
d
e
e
p
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
u
r
a
l
n
e
t
w
o
r
k
s,
”
Art
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
Re
v
i
e
w
,
v
o
l
.
5
3
,
n
o
.
8
,
p
p
.
5
4
5
5
–
5
5
1
6
,
D
e
c
.
2
0
2
0
,
d
o
i
:
1
0
.
1
0
0
7
/
s1
0
4
6
2
-
020
-
0
9
8
2
5
-
6.
[
2
2
]
E.
Li
m
o
n
o
v
a
,
D
.
A
l
f
o
n
s
o
,
D
.
N
i
k
o
l
a
e
v
,
a
n
d
V
.
V
.
A
r
l
a
z
a
r
o
v
,
“
R
e
sN
e
t
-
l
i
k
e
a
r
c
h
i
t
e
c
t
u
r
e
w
i
t
h
l
o
w
h
a
r
d
w
a
r
e
r
e
q
u
i
r
e
me
n
t
s
,
”
i
n
Pr
o
c
e
e
d
i
n
g
s
-
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
P
a
t
t
e
r
n
R
e
c
o
g
n
i
t
i
o
n
,
I
EEE,
J
a
n
.
2
0
2
0
,
p
p
.
6
2
0
4
–
6
2
1
1
.
d
o
i
:
1
0
.
1
1
0
9
/
I
C
P
R
4
8
8
0
6
.
2
0
2
1
.
9
4
1
3
1
8
6
.
[
2
3
]
Y
.
Z
h
o
n
g
,
S
.
Q
i
u
,
X
.
L
u
o
,
Z
.
M
e
n
g
,
a
n
d
J
.
L
i
u
,
“
F
a
c
i
a
l
e
x
p
r
e
ss
i
o
n
r
e
c
o
g
n
i
t
i
o
n
b
a
s
e
d
o
n
o
p
t
i
m
i
z
e
d
R
e
sN
e
t
,
”
i
n
2
0
2
0
2
n
d
Wo
r
l
d
S
y
m
p
o
s
i
u
m
o
n
Art
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
,
WS
AI
2
0
2
0
,
I
EEE,
J
u
n
.
2
0
2
0
,
p
p
.
8
4
–
9
1
.
d
o
i
:
1
0
.
1
1
0
9
/
W
S
A
I
4
9
6
3
6
.
2
0
2
0
.
9
1
4
3
2
8
7
.
[
2
4
]
J.
Z
h
a
n
g
,
Y
.
Zh
u
,
W
.
L
i
,
W
.
F
u
,
a
n
d
L.
C
a
o
,
“
D
R
N
e
t
:
a
d
e
e
p
n
e
u
r
a
l
n
e
t
w
o
r
k
w
i
t
h
m
u
l
t
i
-
l
a
y
e
r
r
e
s
i
d
u
a
l
b
l
o
c
k
s
i
m
p
r
o
v
e
s
i
ma
g
e
d
e
n
o
i
si
n
g
,
”
I
EEE
Ac
c
e
ss
,
v
o
l
.
9
,
p
p
.
7
9
9
3
6
–
7
9
9
4
6
,
2
0
2
1
,
d
o
i
:
1
0
.
1
1
0
9
/
A
C
C
ESS
.
2
0
2
1
.
3
0
8
4
9
5
1
.
[
2
5
]
F
.
H
e
,
T.
Li
u
,
a
n
d
D
.
Ta
o
,
“
W
h
y
R
e
sN
e
t
w
o
r
k
s?
r
e
si
d
u
a
l
s
g
e
n
e
r
a
l
i
z
e
,
”
I
EEE
T
r
a
n
sa
c
t
i
o
n
s
o
n
N
e
u
r
a
l
N
e
t
w
o
r
k
s
a
n
d
L
e
a
r
n
i
n
g
S
y
s
t
e
m
s
,
v
o
l
.
3
1
,
n
o
.
1
2
,
p
p
.
5
3
4
9
–
5
3
6
2
,
D
e
c
.
2
0
2
0
,
d
o
i
:
1
0
.
1
1
0
9
/
TN
N
LS
.
2
0
2
0
.
2
9
6
6
3
1
9
.
B
I
O
G
RAP
H
I
E
S O
F
AUTH
O
RS
Abd
e
lfetta
h
S
o
u
lta
n
a
re
c
e
n
t
ly
o
b
tain
e
d
h
is
d
o
c
t
o
ra
te
in
c
o
m
p
u
ter
sc
ien
c
e
.
H
e
c
o
m
p
lete
d
h
is
m
a
ste
r'
s
d
e
g
re
e
in
so
f
twa
re
q
u
a
li
t
y
a
t
Ha
ss
a
n
II
Un
iv
e
rsit
y
,
Ca
sa
b
la
n
c
a
,
M
o
ro
c
c
o
,
i
n
2
0
1
5
.
His
th
e
sis,
ti
tl
e
d
"
To
wa
rd
s
a
c
o
n
tex
tu
a
l
sy
ste
m
fo
r
sm
a
rt
c
a
r
m
a
n
a
g
e
m
e
n
t,
"
m
a
rk
e
d
a
sig
n
ifi
c
a
n
t
m
il
e
sto
n
e
in
h
is
a
c
a
d
e
m
ic
jo
u
r
n
e
y
.
Cu
rre
n
tl
y
,
h
e
is
d
e
e
p
l
y
e
n
g
a
g
e
d
i
n
h
i
s
d
o
c
to
ra
l
stu
d
ies
a
t
th
e
Lab
o
ra
t
o
ry
o
f
In
f
o
rm
a
ti
o
n
P
r
o
c
e
ss
in
g
a
n
d
M
o
d
e
li
n
g
(LT
I
M
)
wit
h
i
n
t
h
e
Be
n
M
’sik
F
a
c
u
lt
y
o
f
S
c
ien
c
e
.
His
re
se
a
rc
h
fo
c
u
se
s
o
n
m
a
c
h
in
e
lea
rn
in
g
,
d
e
e
p
lea
rn
in
g
,
a
n
d
th
e
a
p
p
li
c
a
ti
o
n
o
f
In
tern
e
t
o
f
Th
in
g
s
tec
h
n
o
lo
g
ies
fo
r
d
ri
v
e
r
c
o
n
t
e
x
t
m
o
n
it
o
rin
g
.
He
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
so
u
lt
a
n
a
.
a
b
d
e
l
fe
tt
a
h
@g
m
a
il
.
c
o
m
.
Fa
o
u
z
i
a
Be
n
a
b
b
o
u
tea
c
h
e
r
-
re
se
a
rc
h
e
r
sin
c
e
1
9
9
4
,
Au
t
h
o
rize
d
P
r
o
fe
ss
o
r
si
n
c
e
2
0
0
8
a
n
d
P
ro
fe
ss
o
r
o
f
Hi
g
h
e
r
E
d
u
c
a
ti
o
n
in
t
h
e
De
p
a
rtme
n
t
o
f
M
a
th
e
m
a
ti
c
s
a
n
d
Co
m
p
u
ter
S
c
ien
c
e
a
t
th
e
Be
n
M
'
S
ick
F
a
c
u
lt
y
o
f
S
c
ien
c
e
s
in
Ca
sa
b
lan
c
a
sin
c
e
2
0
1
5
.
S
h
e
is
a
m
e
m
b
e
r
o
f
th
e
In
f
o
rm
a
ti
o
n
Tec
h
n
o
l
o
g
y
a
n
d
M
o
d
e
li
n
g
Lab
o
ra
t
o
ry
a
n
d
lea
d
e
r
o
f
t
h
e
ICCNSE
(Cl
o
u
d
Co
m
p
u
ti
n
g
,
Ne
two
r
k
a
n
d
S
y
ste
m
s
En
g
i
n
e
e
rin
g
)
tea
m
.
His
re
se
a
rc
h
a
re
a
s
in
c
lu
d
e
c
lo
u
d
Co
m
p
u
ti
n
g
,
d
a
ta
m
in
i
n
g
,
m
a
c
h
i
n
e
lea
rn
in
g
,
a
n
d
n
a
tu
ra
l
la
n
g
u
a
g
e
p
ro
c
e
ss
in
g
.
Sh
e
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
fa
o
u
z
i
a
.
b
e
n
a
b
b
o
u
@
u
n
iv
h
2
c
.
m
a
.
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
A
n
ew d
ee
p
lea
r
n
in
g
mo
d
el
b
a
s
ed
o
n
co
n
vo
lu
tio
n
a
l n
eu
r
a
l
n
etw
o
r
k
…
(
A
b
d
elfetta
h
S
o
u
lta
n
a
)
1701
Na
wa
l
S
a
e
l
tea
c
h
e
r
-
re
se
a
rc
h
e
r
sin
c
e
2
0
1
2
,
A
u
th
o
rize
d
P
r
o
fe
ss
o
r
sin
c
e
2
0
1
4
a
n
d
P
ro
fe
ss
o
r
o
f
Hi
g
h
e
r
Ed
u
c
a
ti
o
n
i
n
th
e
De
p
a
rtme
n
t
o
f
M
a
th
e
m
a
ti
c
s
a
n
d
Co
m
p
u
ter
S
c
ien
c
e
a
t
th
e
Be
n
M
'
S
ick
F
a
c
u
lt
y
o
f
S
c
ien
c
e
s
i
n
Ca
sa
b
lan
c
a
si
n
c
e
2
0
2
0
a
n
d
h
e
r
e
n
g
in
e
e
r
d
e
g
re
e
i
n
s
o
ftwa
re
e
n
g
in
e
e
rin
g
fro
m
ENS
IAS,
M
o
r
o
c
c
o
,
in
2
0
0
2
.
He
re
re
se
a
rc
h
in
tere
sts
in
c
lu
d
e
d
a
ta
m
in
in
g
,
e
d
u
c
a
ti
o
n
a
l
d
a
ta
m
in
i
n
g
,
m
a
c
h
in
e
lea
rn
in
g
,
d
e
e
p
lea
rn
in
g
a
n
d
i
n
t
e
rn
e
t
o
f
t
h
in
g
s
.
Sh
e
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
sa
e
ln
a
wa
l@h
o
tma
il
.
c
o
m
.
S
o
u
k
a
in
a
B
o
u
h
siss
in
re
c
e
i
v
e
d
a
B.
S
c
.
d
e
g
re
e
i
n
m
a
th
e
m
a
ti
c
a
l
sc
ien
c
e
s
a
n
d
c
o
m
p
u
ter
sc
ien
c
e
fro
m
th
e
F
a
c
u
lt
y
o
f
S
c
ien
c
e
s
Be
n
M
’S
ick
,
Ha
ss
a
n
II
Un
iv
e
rsity
o
f
Ca
sa
b
lan
c
a
,
M
o
ro
c
c
o
,
i
n
2
0
1
8
,
a
n
d
a
n
M
.
S
c
.
d
e
g
re
e
in
d
a
ta
sc
ien
c
e
a
n
d
b
ig
d
a
ta
fro
m
Ha
ss
a
n
II
Un
i
v
e
rsity
o
f
Ca
sa
b
lan
c
a
,
in
2
0
2
0
,
wh
e
re
sh
e
is
c
u
rre
n
tl
y
p
u
rsu
in
g
t
h
e
P
h
.
D.d
e
g
re
e
in
c
o
m
p
u
ter
sc
ien
c
e
.
He
r
re
se
a
rc
h
in
tere
sts
i
n
c
lu
d
e
d
riv
e
r
b
e
h
a
v
i
o
r
c
las
sifica
ti
o
n
,
i
n
telli
g
e
n
t
tran
sp
o
rt
sy
ste
m
s,
m
a
c
h
in
e
lea
rn
i
n
g
,
d
e
e
p
lea
rn
i
n
g
,
ima
g
e
c
las
sifica
ti
o
n
,
a
n
d
tem
p
o
ra
l
se
ries
.
Sh
e
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
b
o
u
h
siss
in
.
s
o
u
k
a
in
a
@g
m
a
il
.
c
o
m
.
Evaluation Warning : The document was created with Spire.PDF for Python.