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.
42
,
No
.
3
,
J
u
n
e
2
0
2
6
,
p
p
.
809
~
81
7
I
SS
N:
2
5
0
2
-
4
7
5
2
,
DOI
:
1
0
.
1
1
5
9
1
/ijeecs.v
42
.i
3
.
pp
809
-
81
7
809
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
ee
cs.ia
esco
r
e.
co
m
H
y
brid
pl
ug
in f
o
r
det
ec
ting illici
t
i
ma
g
es o
n t
he in
te
rnet
usi
ng
E
ff
icien
t
N
e
t
co
nv
o
lutiona
l neural n
etworks
Chris
t
ine
L
a
ure
M
a
na
ng
a
,
F
elix
P
a
un
e,
L
éa
nd
re
Nnem
e
Nnem
e
C
o
m
p
u
t
e
r
E
n
g
i
n
e
e
r
i
n
g
D
e
p
a
r
t
me
n
t
,
A
d
v
a
n
c
e
d
T
e
a
c
h
e
r
s T
r
a
i
n
i
n
g
C
o
l
l
e
g
e
f
o
r
Te
c
h
n
i
c
a
l
E
d
u
c
a
t
i
o
n
,
U
n
i
v
e
r
si
t
y
o
f
D
o
u
a
l
a
,
D
o
u
a
l
a
,
C
a
m
e
r
o
o
n
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
J
u
l 6
,
2
0
2
5
R
ev
is
ed
Ma
y
4
,
2
0
2
6
Acc
ep
ted
Ma
y
2
6
,
2
0
2
6
Th
e
p
ro
li
fe
ra
ti
o
n
o
f
i
ll
icit
v
i
su
a
l
c
o
n
ten
t
o
n
th
e
i
n
tern
e
t,
su
c
h
a
s
p
o
r
n
o
g
ra
p
h
y
a
n
d
v
io
le
n
t
ima
g
e
ry
,
p
re
se
n
ts
a
g
ro
wi
n
g
so
c
ieta
l
c
o
n
c
e
rn
.
Th
is
p
a
p
e
r
p
ro
p
o
se
s
t
h
e
d
e
sig
n
a
n
d
im
p
lem
e
n
tatio
n
o
f
a
li
g
h
twe
ig
h
t
b
ro
ws
e
r
-
in
teg
ra
te
d
p
lu
g
i
n
t
h
a
t
u
ti
li
se
s
a
h
y
b
ri
d
a
p
p
r
o
a
c
h
c
o
m
b
i
n
in
g
c
o
n
ten
t
-
b
a
se
d
fil
terin
g
wit
h
c
o
n
v
o
l
u
ti
o
n
a
l
n
e
u
ra
l
n
e
two
r
k
s
(CNN
s),
sp
e
c
ifi
c
a
ll
y
th
e
Eff
icie
n
tNe
tB7
a
rc
h
it
e
c
tu
re
,
t
o
d
e
tec
t
a
n
d
b
lo
c
k
il
li
c
it
ima
g
e
s
in
re
a
l
ti
m
e
.
De
v
e
lo
p
e
d
u
sin
g
P
y
th
o
n
a
n
d
T
e
n
so
rF
lo
w
,
th
e
p
lu
g
in
wa
s
trai
n
e
d
o
n
a
c
u
ra
ted
d
a
tas
e
t
c
o
m
p
risin
g
NSF
W,
De
e
p
Nu
d
e
,
a
n
d
sa
fe
-
fo
r
-
wo
r
k
(S
F
W)
ima
g
e
s.
Ex
p
e
rime
n
tal
re
su
lt
s o
n
a
d
a
tas
e
t
o
f
1
,
0
6
4
ra
n
d
o
m
l
y
se
lec
ted
ima
g
e
s
d
e
m
o
n
stra
ted
a
d
e
tec
ti
o
n
a
c
c
u
ra
c
y
o
f
9
9
%
,
wit
h
a
p
ro
c
e
ss
in
g
ti
m
e
o
f
9
2
se
c
o
n
d
s
a
n
d
a
7
%
c
o
m
b
i
n
e
d
fa
lse
p
o
sit
iv
e
a
n
d
fa
lse
n
e
g
a
ti
v
e
ra
te.
Th
e
p
lu
g
in
is
c
o
m
p
a
ti
b
le
with
Ch
r
o
m
e
b
ro
ws
e
rs
a
n
d
c
o
n
tri
b
u
tes
to
sa
fe
r
o
n
li
n
e
e
x
p
e
rien
c
e
s,
p
a
rti
c
u
larly
fo
r
c
h
il
d
re
n
,
e
d
u
c
a
t
o
rs,
a
n
d
u
se
rs
in
se
n
siti
v
e
e
n
v
iro
n
m
e
n
ts.
K
ey
w
o
r
d
s
:
C
o
n
v
o
lu
tio
n
al
n
eu
r
al
n
etwo
r
k
s
E
f
f
icien
tNetB
7
Hy
b
r
id
f
ilter
in
g
I
llicit im
ag
e
d
etec
tio
n
Plu
g
in
d
ev
elo
p
m
en
t
T
h
is i
s
a
n
o
p
e
n
a
c
c
e
ss
a
rticle
u
n
d
e
r th
e
CC B
Y
-
SA
li
c
e
n
se
.
C
o
r
r
e
s
p
o
nd
ing
A
uth
o
r
:
C
h
r
is
tin
e
L
au
r
e
Ma
n
an
g
a
Dep
ar
tm
en
t o
f
C
o
m
p
u
ter
E
n
g
i
n
ee
r
in
g
,
A
d
v
an
ce
d
T
ea
ch
er
s
T
r
ain
in
g
C
o
lleg
e
f
o
r
T
ec
h
n
ical
E
d
u
ca
tio
n
Un
iv
er
s
ity
o
f
Do
u
ala
1
8
7
2
,
Do
u
ala,
C
am
er
o
o
n
E
m
ail:
cm
an
an
g
a
3
5
@
g
m
ail.
co
m
1.
I
NT
RO
D
UCT
I
O
N
T
h
e
m
ass
iv
e
ex
p
a
n
s
io
n
o
f
d
i
g
ital
co
n
ten
t
h
as
m
a
d
e
h
i
g
h
-
s
p
ee
d
in
ter
n
et
ac
ce
s
s
a
s
tan
d
ar
d
,
b
u
t
it
h
as
co
n
cu
r
r
en
tly
s
im
p
lifie
d
th
e
d
is
s
em
in
atio
n
o
f
h
ar
m
f
u
l
m
ater
ials
[
1
]
.
Ma
n
u
al
co
n
ten
t
m
o
d
er
atio
n
is
n
o
lo
n
g
e
r
v
iab
le
d
u
e
to
th
e
s
h
ee
r
v
o
lu
m
e
o
f
d
a
ta
u
p
l
o
ad
ed
ev
er
y
s
ec
o
n
d
.
C
o
n
s
eq
u
en
tly
,
a
u
to
m
a
ted
s
y
s
tem
s
u
s
in
g
ar
tific
ial
in
tellig
en
ce
(
AI
)
h
av
e
b
ec
o
m
e
i
n
d
is
p
en
s
ab
le.
T
h
is
r
esear
ch
in
tr
o
d
u
ce
s
a
“
h
y
b
r
i
d
f
ilter
in
g
”
ap
p
r
o
ac
h
,
wh
ich
r
ef
e
r
s
to
th
e
i
n
teg
r
ate
d
u
s
e
o
f
co
n
te
n
t
-
b
ased
attr
i
b
u
tes
(
m
etad
ata,
s
ig
n
at
u
r
es)
a
n
d
s
tr
u
ctu
r
al
f
ea
t
u
r
e
ex
tr
ac
tio
n
v
ia
co
n
v
o
l
u
tio
n
al
n
eu
r
al
n
etwo
r
k
s
(
C
NN)
[
2
]
–
[
4
]
.
Pre
v
io
u
s
s
tu
d
ies
p
r
im
ar
ily
f
o
cu
s
ed
o
n
s
er
v
er
-
s
id
e
f
ilter
in
g
,
wh
ich
o
f
ten
in
tr
o
d
u
ce
s
laten
c
y
a
n
d
p
r
iv
ac
y
c
o
n
ce
r
n
s
[
5
]
.
Ou
r
o
b
jectiv
e
is
t
o
d
e
v
elo
p
a
clien
t
-
s
id
e
s
o
lu
tio
n
v
ia
a
b
r
o
wser
p
lu
g
in
th
at
e
n
s
u
r
es
r
ea
l
-
tim
e
p
r
o
tectio
n
with
o
u
t
r
e
ly
in
g
o
n
e
x
ter
n
al
s
er
v
er
r
esp
o
n
s
e
tim
es.
T
o
ac
h
iev
e
th
is
,
we
lev
er
ag
e
th
e
u
n
iq
u
e
ca
p
ab
ilit
ies
o
f
E
f
f
icien
tNetB
7
,
an
ar
c
h
itectu
r
e
t
h
at
u
tili
ze
s
a
co
m
p
o
u
n
d
s
ca
lin
g
m
eth
o
d
t
o
o
p
tim
ally
b
alan
ce
n
etwo
r
k
d
ep
th
,
wid
th
,
a
n
d
r
eso
lu
tio
n
.
T
h
e
ce
n
t
r
al
is
s
u
e
ad
d
r
ess
ed
in
th
is
s
tu
d
y
is
en
s
u
r
in
g
th
at
u
s
er
s
p
a
r
ticu
lar
ly
v
u
ln
er
ab
le
g
r
o
u
p
s
ca
n
n
o
t
ac
ce
s
s
illi
cit
o
r
u
n
au
th
o
r
ized
co
n
ten
t,
s
p
ec
if
i
ca
lly
im
ag
es
r
elate
d
to
v
io
le
n
ce
an
d
p
o
r
n
o
g
r
ap
h
y
.
Un
lik
e
ex
is
tin
g
s
tan
d
alo
n
e
to
o
ls
,
o
u
r
s
o
lu
tio
n
is
d
esig
n
ed
to
b
e
lig
h
tweig
h
t,
in
teg
r
a
b
le
with
v
ar
io
u
s
web
p
latf
o
r
m
s
,
an
d
ca
p
ab
le
o
f
r
ea
l
-
tim
e
class
if
icatio
n
.
T
h
is
p
ap
e
r
p
r
esen
ts
th
e
d
esig
n
a
n
d
im
p
lem
en
tatio
n
o
f
a
C
NN
-
b
ased
p
lu
g
in
f
o
r
Go
o
g
le
C
h
r
o
m
e.
T
h
e
k
ey
co
n
tr
ib
u
tio
n
s
o
f
th
is
wo
r
k
in
clu
d
e:
T
h
e
d
e
v
elo
p
m
en
t
o
f
a
r
o
b
u
s
t
d
etec
ti
o
n
alg
o
r
ith
m
b
ased
o
n
E
f
f
icien
tNetB
7
in
teg
r
ate
d
d
ir
ec
tly
in
to
th
e
b
r
o
wser
.
An
a
n
aly
s
is
o
f
web
co
n
ten
t
(
im
ag
e
s
an
d
v
id
eo
f
r
am
es)
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
SN
:
2
5
0
2
-
4
7
5
2
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
,
Vo
l.
42
,
No
.
3
,
J
u
n
e
20
2
6
:
8
0
9
-
81
7
810
to
class
if
y
it
as
e
ith
er
illi
cit
o
r
p
er
m
is
s
ib
le
in
r
ea
l
-
tim
e.
A
co
m
p
r
eh
en
s
iv
e
ev
alu
atio
n
o
f
th
e
p
lu
g
in
’
s
p
er
f
o
r
m
an
ce
in
t
er
m
s
o
f
ac
cu
r
ac
y
an
d
ex
ec
u
tio
n
e
f
f
icien
cy
.
T
h
e
p
r
o
p
o
s
al
o
f
an
a
d
ap
tiv
e
s
y
s
tem
tailo
r
ed
t
o
th
e
s
p
ec
if
ic
n
ee
d
s
o
f
d
iv
er
s
e
u
s
er
ca
teg
o
r
ies,
in
clu
d
in
g
p
ar
e
n
ts
,
ch
ild
r
en
,
teac
h
er
s
,
a
n
d
s
tu
d
en
ts
.
2.
T
H
E
P
RO
P
O
SE
D
M
E
T
H
O
D
T
h
e
p
r
o
p
o
s
ed
s
y
s
tem
u
tili
ze
s
a
m
u
ltil
ev
el
d
e
cisi
o
n
lo
g
i
c
d
esig
n
ed
to
m
i
n
im
ize
co
m
p
u
tatio
n
al
o
v
er
h
ea
d
an
d
e
n
s
u
r
e
lo
w
late
n
cy
.
T
h
e
p
r
o
ce
s
s
b
eg
in
s
b
y
in
ter
ce
p
tin
g
HT
T
P
r
eq
u
ests
f
o
r
im
ag
e
ass
ets
at
th
e
b
r
o
wser
lev
el.
A
f
allb
ac
k
m
ec
h
an
is
m
is
im
p
lem
en
ted
:
if
th
e
in
itial
co
n
ten
t
s
ca
n
,
b
a
s
ed
o
n
lig
h
tweig
h
t
m
etad
ata
o
r
s
ig
n
atu
r
es,
is
i
n
co
n
clu
s
iv
e,
th
e
im
ag
e
is
th
en
p
ass
ed
to
th
e
C
NN
f
o
r
s
tr
u
ctu
r
al
an
aly
s
is
.
A
class
if
icatio
n
th
r
esh
o
ld
is
s
et
at
0
.
8
5
;
an
y
im
ag
e
s
co
r
in
g
ab
o
v
e
th
is
v
alu
e
in
an
illi
cit
ca
teg
o
r
y
(
e.
g
.
,
n
u
d
ity
o
r
wea
p
o
n
s
)
is
au
to
m
atica
lly
b
lu
r
r
ed
o
r
r
e
p
lace
d
b
y
a
war
n
in
g
p
lace
h
o
l
d
er
to
p
r
o
tect
th
e
u
s
er
.
2
.
1
.
Co
m
pa
riso
n wit
h e
x
is
t
ing
m
et
ho
ds
Pre
v
io
u
s
m
eth
o
d
s
f
o
r
co
n
ten
t
m
o
d
er
atio
n
p
r
im
ar
ily
r
elied
o
n
s
k
in
-
to
n
e
d
etec
tio
n
,
m
etad
at
a
f
ilter
in
g
,
an
d
k
e
y
wo
r
d
an
aly
s
is
,
all
o
f
wh
ich
h
av
e
d
em
o
n
s
tr
ated
s
ig
n
if
ican
t
lim
itatio
n
s
in
p
r
ec
is
io
n
an
d
ad
a
p
tab
ilit
y
.
I
n
co
n
tr
ast,
m
ac
h
i
n
e
lear
n
i
n
g
(
ML
)
h
as
f
u
n
d
am
en
tally
ad
v
an
ce
d
th
e
f
ield
o
f
co
n
ten
t
f
ilter
in
g
[
3
]
.
Fo
r
in
s
tan
ce
,
Yah
o
o
’
s
NSFW
class
if
ier
an
d
v
ar
io
u
s
T
en
s
o
r
Flo
w
-
b
ased
im
p
lem
en
tatio
n
s
h
av
e
d
em
o
n
s
tr
ated
th
at
C
NN
s
s
ig
n
if
ican
tly
o
u
tp
e
r
f
o
r
m
tr
ad
itio
n
al
alg
o
r
ith
m
ic
m
o
d
els.
Ho
wev
er
,
m
an
y
ex
is
t
in
g
s
y
s
tem
s
r
em
ain
s
tan
d
alo
n
e
to
o
ls
o
r
s
er
v
er
-
s
id
e
ap
p
licatio
n
s
th
at
lack
s
ea
m
less
in
teg
r
atio
n
with
m
ain
s
tr
ea
m
u
s
er
en
v
ir
o
n
m
en
ts
.
Ou
r
ap
p
r
o
ac
h
ad
d
r
ess
es
th
is
g
ap
b
y
em
b
e
d
d
in
g
C
NN
-
p
o
wer
ed
d
etec
tio
n
d
ir
ec
tly
in
to
web
b
r
o
wser
s
as
a
clien
t
-
s
id
e
p
lu
g
in
.
T
h
is
ar
ch
itectu
r
e
o
f
f
er
s
a
p
r
ac
tical,
r
ea
l
-
tim
e,
an
d
ac
ce
s
s
ib
le
s
o
lu
tio
n
f
o
r
au
to
m
ated
co
n
ten
t
m
o
d
e
r
a
t
i
o
n
,
e
n
s
u
r
i
n
g
i
m
m
e
d
i
a
t
e
p
r
o
t
ec
t
i
o
n
w
i
t
h
o
u
t
t
h
e
l
a
te
n
c
y
i
n
h
e
r
e
n
t
i
n
e
x
t
e
r
n
a
l
s
e
r
v
e
r
-
s
i
d
e
p
r
o
c
e
s
s
i
n
g
[
4
]
.
2
.
2
.
P
lug
in a
rc
hite
ct
ure
T
h
e
p
l
u
g
in
is
d
ev
el
o
p
ed
in
Py
th
o
n
a
n
d
in
teg
r
ate
d
v
ia
a
R
E
STf
u
l
API
i
n
to
we
b
p
latf
o
r
m
s
.
I
t
f
ea
tu
r
es
a
f
r
o
n
t
-
en
d
in
ter
f
ac
e
f
o
r
im
ag
e
u
p
lo
ad
an
d
s
ca
n
n
in
g
,
wh
ile
th
e
b
ac
k
-
en
d
p
r
o
ce
s
s
es
i
m
ag
es
th
r
o
u
g
h
th
e
E
f
f
icien
tNetB
7
m
o
d
el
to
ass
ess
co
n
ten
t.
R
ea
l
-
tim
e
aler
ts
ar
e
g
en
er
ated
f
o
r
f
lag
g
ed
im
ag
es [
5
]
.
2
.
3
.
Da
t
a
s
et
a
nd
p
re
pro
ce
s
s
i
ng
W
e
ass
em
b
led
o
u
r
tr
ain
in
g
d
ata
f
r
o
m
a
c
o
m
b
in
atio
n
o
f
p
u
b
lic
d
atasets
,
in
clu
d
in
g
NSFW
,
Dee
p
Nu
d
e,
an
d
SF
W
/NS
FW
s
o
u
r
ce
s
.
I
m
ag
es
wer
e
r
esized
to
2
2
4
×2
2
4
p
ix
els
an
d
n
o
r
m
alis
ed
.
Au
g
m
en
tatio
n
te
ch
n
iq
u
es
s
u
ch
as
r
o
tatio
n
,
f
l
ip
p
in
g
,
an
d
b
r
ig
h
tn
ess
ad
ju
s
tm
en
t
wer
e
u
s
ed
to
in
cr
ea
s
e
d
ataset
d
iv
er
s
ity
an
d
im
p
r
o
v
e
m
o
d
el
g
en
er
aliza
tio
n
[
5
]
.
2
.
4
.
CNN
c
o
nfig
ura
t
io
n
E
f
f
icien
tNetB
7
was
s
elec
ted
f
o
r
its
s
tate
-
of
-
t
h
e
-
ar
t
b
alan
ce
o
f
ac
cu
r
ac
y
an
d
co
m
p
u
tatio
n
al
ef
f
icien
cy
.
T
h
e
ar
ch
itectu
r
e
i
n
clu
d
es
co
n
v
o
lu
tio
n
al
lay
e
r
s
,
R
eL
U
ac
tiv
atio
n
s
,
b
atc
h
n
o
r
m
alis
atio
n
,
d
r
o
p
o
u
t
r
eg
u
lar
is
atio
n
,
an
d
a
s
ig
m
o
i
d
o
u
tp
u
t la
y
e
r
f
o
r
b
in
a
r
y
class
if
icatio
n
[
6
]
.
2
.
5
.
T
ra
ini
ng
a
nd
v
a
lid
a
t
io
n
Mu
ltip
le
m
o
d
el
v
ar
ian
ts
wer
e
test
ed
,
with
Mo
d
el
4
y
iel
d
in
g
t
h
e
h
i
g
h
est
p
er
f
o
r
m
a
n
c
e.
T
r
ain
in
g
in
v
o
lv
ed
u
s
in
g
C
I
FAR
-
1
0
an
d
C
I
FAR
-
1
0
0
d
atasets
,
an
d
m
etr
ics
s
u
ch
as
p
r
ec
is
io
n
,
r
ec
all,
an
d
F1
-
s
co
r
e
wer
e
ev
alu
ated
u
s
in
g
Scik
it
-
lear
n
.
T
h
e
co
n
f
u
s
io
n
m
atr
ix
a
n
aly
s
is
in
d
icate
d
a
s
tr
o
n
g
ab
ilit
y
to
d
if
f
er
en
tiate
b
etwe
en
class
es,
in
clu
d
in
g
n
ak
ed
/clo
th
ed
in
d
iv
id
u
als an
d
v
a
r
io
u
s
we
ap
o
n
s
[
7
]
–
[
9
]
.
Step
1
: Co
n
s
tr
u
ctio
n
o
f
th
e
co
n
v
o
lu
tio
n
n
eu
r
al
n
etwo
r
k
#
Sep
ar
atio
n
o
f
th
e
tr
ain
i
n
g
s
et
f
r
o
m
t
h
e
test
s
et
#
W
e
u
s
e
Ker
as to
im
p
o
r
t th
e
im
ag
es to
tr
ain
o
u
r
n
e
u
r
al
n
etw
o
r
k
.
#
I
m
p
o
r
tin
g
m
o
d
u
les s
p
ec
if
ic
t
o
co
n
v
o
lu
t
io
n
n
eu
r
al
n
etwo
r
k
s
#
I
n
itialis
e
o
u
r
co
n
v
o
lu
tio
n
n
e
u
r
al
n
etwo
r
k
#
Ad
d
in
g
t
h
e
co
n
v
o
lu
tio
n
lay
er
#
C
o
n
v
er
t th
e
im
a
g
e
in
to
a
m
at
r
ix
with
n
u
m
b
er
s
f
o
r
ea
ch
p
ix
el
#
T
h
en
a
p
p
ly
a
f
ea
tu
r
e
d
etec
to
r
to
th
e
m
atr
ix
#
W
e
’
ll o
b
tain
a
f
ea
tu
r
e
m
ap
w
h
ich
will f
o
r
m
o
u
r
c
o
n
v
o
lu
ti
o
n
lay
er
#
R
E
L
U
o
r
R
ec
tifie
r
ac
tiv
atio
n
f
u
n
ctio
n
th
at
ad
d
s
n
o
n
-
lin
ea
r
i
ty
to
o
u
r
m
o
d
el.
#
o
u
r
m
o
d
el
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:
2502
-
4
7
5
2
Hyb
r
id
p
lu
g
in
fo
r
d
etec
tin
g
illi
cit
ima
g
es o
n
th
e
in
tern
et
u
s
i
n
g
E
fficien
tN
et
…
(
C
h
r
is
tin
e
L
a
u
r
e
Ma
n
a
n
g
a
)
811
Step
2
: Po
o
lin
g
b
eg
in
s
#
Ma
x
p
o
o
lin
g
#
T
h
is
in
v
o
l
v
es
tak
in
g
th
e
f
ea
tu
r
e
m
ap
s
a
n
d
f
illi
n
g
in
th
e
co
n
v
o
l
u
tio
n
m
at
r
ix
.
W
e
o
b
tain
a
m
u
ch
s
m
alle
r
f
ea
tu
r
e
m
ap
,
wh
ich
allo
ws u
s
t
o
o
b
tain
t
h
e
p
o
o
lin
g
lay
e
r
class
if
ier
.
ad
d
(
Ma
x
Po
o
lin
g
2
D(
p
o
o
l_
s
ize=
(
2
,
2
)
)
)
#
Ad
d
a
s
ec
o
n
d
co
n
v
o
lu
tio
n
la
y
er
class
if
ier
.
ad
d
(
C
o
n
v
o
lu
tio
n
2
D(
f
ilter
s
=3
2
,
k
er
n
el
_
s
ize=
3
,
s
tr
id
es=1
,
ac
tiv
atio
n
=
‘
r
elu
’
))
class
if
ier
.
ad
d
(
Ma
x
Po
o
lin
g
2
D(
p
o
o
l_
s
ize=
(
2
,
2
)
)
)
S
tep
3
: Flaten
n
in
g
#
C
r
ea
te
a
s
in
g
le
1
D
v
ec
to
r
an
d
th
en
c
o
n
n
ec
t it
to
t
h
e
f
ir
s
t h
i
d
d
en
lay
e
r
to
s
tar
t c
lass
if
y
in
g
.
class
if
ier
.
ad
d
(
Flatten
(
)
)
Step
4
: A
d
d
in
g
a
f
u
lly
c
o
n
n
ec
t
ed
n
eu
r
al
n
etwo
r
k
#
C
o
m
p
ile
th
e
n
eu
r
al
n
etwo
r
k
#
W
h
ich
Gr
ad
ien
t a
lg
o
r
ith
m
to
u
s
e
#
C
o
s
t
f
u
n
ctio
n
to
u
s
e
#
Me
tr
ics to
ev
alu
ate
#
T
r
ain
in
g
th
e
C
NN
o
n
o
u
r
im
ag
es
#
Use in
Ker
as d
o
cu
m
e
n
tatio
n
to
p
r
ev
e
n
t o
v
e
r
tr
ain
in
g
.
On
ce
th
e
tr
ain
in
g
p
h
ase
is
co
m
p
lete,
th
e
m
o
d
el
m
u
s
t
b
e
v
a
lid
ated
u
s
in
g
a
test
in
g
d
ata
b
a
s
e
d
is
tin
ct
f
r
o
m
th
e
tr
ain
in
g
s
et.
T
h
is
ev
alu
atio
n
is
cr
u
cial
f
o
r
ass
ess
in
g
th
e
n
eu
r
al
s
y
s
tem
’
s
p
er
f
o
r
m
a
n
ce
an
d
id
en
tif
y
in
g
s
p
ec
if
ic
d
ata
ty
p
es
th
at
m
ay
ca
u
s
e
m
is
cla
s
s
if
icatio
n
.
I
f
th
e
r
esu
lts
ar
e
u
n
s
ati
s
f
ac
to
r
y
,
ad
ju
s
tm
en
ts
ar
e
r
eq
u
ir
ed
,
s
u
ch
as
m
o
d
if
y
in
g
th
e
n
etwo
r
k
ar
ch
itectu
r
e
o
r
f
i
n
e
-
tu
n
in
g
h
y
p
e
r
p
ar
am
eter
s
,
i
n
clu
d
in
g
ac
tiv
atio
n
f
u
n
ctio
n
s
,
lear
n
i
n
g
r
ates,
o
r
th
e
co
m
p
o
s
itio
n
o
f
th
e
tr
a
in
in
g
b
ase.
T
h
e
co
n
f
u
s
io
n
m
atr
ix
s
er
v
es
as
a
co
m
p
r
eh
e
n
s
iv
e
s
u
m
m
ar
y
o
f
p
r
ed
ictio
n
r
esu
lts
f
o
r
th
e
class
if
icatio
n
p
r
o
b
lem
.
I
t
f
ac
ilit
ates
a
d
ir
ec
t
co
m
p
ar
is
o
n
b
etwe
en
th
e
ac
t
u
al
tar
g
et
v
alu
es
an
d
th
e
p
r
e
d
ictio
n
s
g
e
n
er
at
ed
b
y
th
e
m
o
d
el.
B
y
b
r
ea
k
in
g
d
o
wn
co
r
r
ec
t
an
d
in
co
r
r
ec
t p
r
ed
ictio
n
s
b
y
class
,
it a
llo
ws f
o
r
a
d
etailed
an
al
y
s
is
ag
ain
s
t d
ef
in
ed
g
r
o
u
n
d
tr
u
th
v
alu
es.
Als
o
r
ef
er
r
ed
to
as
a
c
o
n
tin
g
en
cy
tab
le,
th
e
c
o
n
f
u
s
io
n
m
a
t
r
ix
is
a
f
u
n
d
a
m
en
tal
to
o
l
f
o
r
ev
alu
atin
g
class
if
icatio
n
p
er
f
o
r
m
an
ce
.
I
t
v
is
u
ally
r
ep
r
esen
ts
h
o
w
“
c
o
n
f
u
s
ed
”
a
m
o
d
el
m
ay
b
e
wh
en
d
is
tin
g
u
is
h
in
g
b
etwe
en
d
if
f
er
en
t
ca
te
g
o
r
ies.
W
h
ile
it
ap
p
ea
r
s
as
a
s
im
p
le
2
×
2
m
atr
ix
in
b
in
ar
y
ca
s
es,
r
o
ws
an
d
co
l
u
m
n
s
ar
e
ad
d
ed
to
ac
co
m
m
o
d
ate
m
o
r
e
c
o
m
p
lex
,
m
u
lti
-
class
class
if
icat
io
n
p
r
o
b
lem
s
[
1
0
]
.
Fro
m
th
is
m
atr
ix
,
v
ar
io
u
s
p
er
f
o
r
m
an
ce
m
etr
ics
ar
e
d
er
iv
e
d
to
q
u
an
tify
th
e
m
o
d
el
’
s
d
etec
ti
o
n
q
u
ality
.
T
h
ese
m
etr
ics
-
in
clu
d
in
g
p
r
ec
is
io
n
,
r
ec
all
,
an
d
F1
-
s
co
r
e
-
ar
e
ca
lcu
lated
b
ased
o
n
tr
u
e
p
o
s
itiv
e
(
T
P),
tr
u
e
n
eg
ativ
e
(
T
N)
,
f
alse p
o
s
itiv
e
(
FP
)
,
an
d
f
alse n
eg
ativ
e
(
FN)
v
alu
es.
Fig
u
r
e
1
illu
s
tr
ates th
e
d
is
tr
ib
u
tio
n
o
f
th
ese
m
etr
ics
ac
r
o
s
s
ea
ch
clas
s
in
o
u
r
s
tu
d
y
.
T
h
e
d
ata
class
if
icatio
n
r
ep
o
r
t
f
o
r
m
o
d
el
1
p
r
o
ce
s
s
ed
b
y
Py
th
o
n
’
s
Scik
it
-
lear
n
lib
r
ar
y
is
s
h
o
wn
in
T
ab
le
1
.
Fig
u
r
e
1
.
C
o
n
f
u
s
io
n
m
atr
i
x
o
f
o
u
r
m
o
d
el
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
SN
:
2
5
0
2
-
4
7
5
2
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
,
Vo
l.
42
,
No
.
3
,
J
u
n
e
20
2
6
:
8
0
9
-
81
7
812
T
ab
le
1
.
L
ea
r
n
in
g
c
lass
if
icattio
n
r
atio
f
o
r
o
u
r
m
o
d
el
P
r
e
c
i
s
i
o
n
R
e
c
a
l
l
F1
-
s
c
o
r
e
S
u
p
p
o
r
t
R
i
f
f
l
e
s
0
.
9
1
0
.
9
8
0
.
9
5
1
1
6
Ta
n
k
s
0
.
9
7
0
.
9
2
0
.
9
4
1
2
3
G
u
n
s
0
.
9
9
0
.
9
9
0
.
9
9
1
5
7
K
n
i
f
e
i
m
a
g
e
s
0
.
9
7
0
.
9
8
0
.
9
8
2
2
0
N
a
k
e
d
p
e
o
p
l
e
0
.
9
9
0
.
9
7
0
.
9
8
3
2
1
C
l
o
t
h
e
d
p
e
o
p
l
e
0
.
9
9
1
.
0
0
1
.
0
0
1
2
7
A
c
c
u
r
a
c
y
0
.
9
8
1
0
6
4
M
a
c
r
o
a
v
g
.
0
.
9
7
0
.
9
7
0
.
9
7
1
0
6
4
W
e
i
g
h
t
e
d
a
v
g
0
.
9
8
0
.
9
8
0
.
9
8
1
0
6
4
3.
M
E
T
H
O
DS (
M
)
3
.
1
.
Da
t
a
s
et
s
o
urce
a
nd
cha
ra
ct
er
is
t
ics
T
h
e
d
ataset
co
n
s
is
ts
o
f
1
,
0
6
4
im
ag
es
cu
r
ated
f
r
o
m
p
u
b
lic
r
ep
o
s
ito
r
ies
,
in
clu
d
in
g
NSFW
an
d
SF
W
d
atasets
.
T
h
e
im
ag
es
co
v
er
s
ix
d
is
tin
ct
class
e
s
:
R
if
le,
T
an
k
,
Gu
n
s
,
Kn
if
e
im
ag
es
(
f
o
r
v
io
l
en
ce
d
etec
tio
n
)
,
a
n
d
Nak
ed
/C
lo
th
ed
p
eo
p
le
(
f
o
r
ad
u
lt
co
n
ten
t
d
etec
tio
n
)
.
T
h
e
lab
elin
g
p
r
o
ce
s
s
was
co
n
d
u
cted
m
a
n
u
ally
to
estab
lis
h
a
r
ig
o
r
o
u
s
g
r
o
u
n
d
tr
u
th
.
As
n
o
ted
b
y
r
ev
iewe
r
s
[
6
]
,
[
1
0
]
,
d
ev
el
o
p
in
g
g
r
o
u
n
d
tr
u
th
is
ex
p
en
s
iv
e,
an
d
wh
il
e
o
u
r
d
ataset
is
r
o
b
u
s
t,
i
t
ex
clu
d
es
n
o
n
-
h
u
m
an
o
b
jects
with
s
k
in
-
lik
e
tex
tu
r
es
(
e.
g
.
,
ce
r
tain
an
im
als),
wh
ich
is
a
n
o
ted
lim
itatio
n
.
3
.
2
.
P
r
o
ce
du
re
a
nd
im
plem
e
nta
t
io
n
As
s
h
o
wn
in
Fig
u
r
e
1
,
th
e
a
r
ch
itectu
r
e
f
o
llo
ws
a
s
tan
d
ar
d
C
NN
p
ip
elin
e
b
u
t
is
o
p
tim
ized
u
s
in
g
E
f
f
icien
tNetB
7
[
7
]
.
T
h
e
o
v
er
all
Fig
u
r
e
2
r
ep
r
esen
ts
th
e
wo
r
k
f
lo
w
o
f
t
h
e
p
lu
g
in
.
Sp
ec
i
f
ically
,
illu
s
tr
ates
th
e
im
ag
e
p
r
ep
r
o
ce
s
s
in
g
p
h
ase
w
h
er
e
im
ag
es
ar
e
n
o
r
m
alize
d
t
o
2
2
4
×
2
2
4
p
ix
els.
[
I
MA
GE
OF
FLO
W
C
HA
R
T
:
I
NPUT
-
>
PR
E
P
R
OC
E
SS
-
>
E
FF
I
C
I
E
NT
NE
T
-
>
C
L
AS
SI
FICATI
ON
-
>
AC
T
I
ON]
4.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
N
4
.
1
.
Ana
ly
s
is
o
f
e
x
perim
ent
a
l r
esu
lt
s
T
h
e
p
er
f
o
r
m
an
ce
o
f
th
e
v
ar
io
u
s
ex
p
er
im
en
tal
m
o
d
els
was
ev
alu
ated
b
ased
o
n
m
u
ltip
le
m
etr
ics:
er
r
o
r
r
ate,
ac
cu
r
ac
y
,
v
alid
atio
n
er
r
o
r
,
v
alid
atio
n
ac
cu
r
ac
y
,
an
d
ex
ec
u
tio
n
tim
e.
A
s
ig
n
if
ican
t
o
b
s
er
v
atio
n
ac
r
o
s
s
all
test
s
wa
s
th
e
h
ig
h
co
m
p
u
tatio
n
al
d
em
an
d
;
t
h
e
s
u
b
s
tan
tial
s
ize
o
f
th
e
d
ataset
n
ec
ess
itated
th
e
u
s
e
o
f
a
GPU
in
s
tead
o
f
a
C
PU
to
en
s
u
r
e
ef
f
icien
t
p
r
o
ce
s
s
in
g
an
d
m
an
ag
ea
b
le
tr
ain
in
g
cy
cles.
Am
o
n
g
th
e
test
ed
co
n
f
ig
u
r
atio
n
s
,
Mo
d
el
4
ac
h
iev
ed
th
e
h
ig
h
est
o
v
er
all
p
e
r
f
o
r
m
a
n
ce
.
T
h
ese
s
u
p
er
io
r
r
esu
lts
ar
e
d
ir
ec
tly
co
r
r
elate
d
with
th
e
o
p
tim
ized
n
u
m
b
er
o
f
ep
o
ch
s
a
n
d
t
h
e
d
e
p
th
o
f
th
e
co
n
v
o
lu
tio
n
al
lay
er
s
.
Ho
wev
er
,
th
is
g
ain
in
ac
cu
r
ac
y
ca
m
e
at
th
e
c
o
s
t
o
f
e
x
ten
d
e
d
ex
e
c
u
tio
n
tim
es
,
p
r
im
ar
ily
d
u
e
to
th
e
h
i
g
h
n
u
m
b
er
o
f
t
r
ain
in
g
ep
o
ch
s
r
eq
u
ir
ed
f
o
r
c
o
n
v
e
r
g
e
n
ce
.
Ou
r
f
in
d
in
g
s
co
n
f
ir
m
th
a
t
th
e
d
ep
th
o
f
a
co
n
v
o
lu
tio
n
al
n
eu
r
al
n
etwo
r
k
is
a
cr
itical
f
ac
to
r
f
o
r
s
u
cc
ess
;
f
o
r
in
s
tan
ce
,
o
m
itti
n
g
ju
s
t
o
n
e
i
n
ter
m
ed
iate
lay
er
r
esu
lted
in
an
ap
p
r
o
x
im
ate
5
%
d
ec
r
ea
s
e
in
p
e
r
f
o
r
m
an
ce
,
h
ig
h
lig
h
tin
g
t
h
e
n
ec
ess
ity
o
f
a
d
ee
p
ar
ch
itectu
r
e
f
o
r
co
m
p
lex
class
if
icatio
n
task
s
.
T
h
e
m
o
d
els
’
ef
f
ec
tiv
en
ess
im
p
r
o
v
e
d
p
r
o
g
r
ess
iv
ely
as
th
e
n
etwo
r
k
d
ep
th
was
in
cr
ea
s
ed
an
d
ep
o
ch
in
ter
v
als
wer
e
r
ef
in
ed
.
Fu
r
th
er
m
o
r
e,
th
e
s
ize
an
d
q
u
ality
o
f
th
e
tr
ai
n
in
g
d
ataset
p
r
o
v
ed
to
b
e
p
iv
o
tal,
r
ein
f
o
r
cin
g
th
e
p
r
in
cip
le
th
at
a
lar
g
e,
d
i
v
er
s
e
d
ataset
is
ess
en
tia
l f
o
r
attain
in
g
o
p
tim
al
p
e
r
f
o
r
m
an
ce
in
d
ee
p
lear
n
in
g
m
o
d
els.
T
h
e
im
p
lem
en
tatio
n
b
ased
o
n
E
f
f
icien
tNetB
7
d
em
o
n
s
tr
ated
ex
ce
p
tio
n
al
r
esu
lts
in
d
etec
tin
g
‘
v
io
len
ce
’
an
d
‘
p
o
r
n
o
g
r
ap
h
y
’
class
es.
T
h
e
m
o
d
el
ac
h
iev
ed
h
ig
h
lev
els
o
f
p
r
ec
is
io
n
,
s
en
s
itiv
ity
,
an
d
s
p
ec
if
icity
,
wh
ich
ca
n
b
e
at
tr
ib
u
ted
to
th
e
m
in
im
al
o
cc
u
r
r
en
ce
o
f
f
alse
p
o
s
itiv
es
an
d
f
alse
n
eg
ativ
es.
R
em
ar
k
ab
ly
,
a
p
r
ec
is
io
n
s
co
r
e
o
f
9
9
%
was
r
ea
c
h
ed
af
ter
o
n
l
y
9
ep
o
c
h
s
.
T
h
e
co
n
s
is
ten
tly
h
ig
h
p
r
ec
is
io
n
a
n
d
s
en
s
itiv
ity
v
alu
es,
b
o
th
av
er
ag
in
g
n
ea
r
9
9
%,
co
n
f
ir
m
th
at
th
e
m
o
d
el
ef
f
ec
tiv
ely
m
in
im
izes
m
is
class
if
icatio
n
s
.
Sp
ec
if
ically
:
i.
Vio
len
ce
d
etec
tio
n
was
s
u
cc
ess
f
u
l
in
id
en
tify
in
g
o
b
jects
an
d
p
atter
n
s
ty
p
ica
lly
ass
o
ciate
d
with
v
io
len
t
b
eh
av
i
o
r
,
an
d
ii.
Po
r
n
o
g
r
a
p
h
ic
co
n
te
n
t
d
etec
tio
n
r
elied
o
n
th
e
ac
cu
r
ate
r
ec
o
g
n
itio
n
o
f
n
u
d
ity
b
y
d
is
tin
g
u
is
h
in
g
b
etwe
en
cl
o
th
e
d
an
d
u
n
clo
th
e
d
in
d
i
v
id
u
als.
T
ab
le
2
s
u
m
m
ar
izes
th
e
p
er
f
o
r
m
an
ce
m
etr
ics
o
b
tain
ed
d
u
r
in
g
th
e
f
in
al
tr
ain
in
g
cy
cle
f
o
r
e
ac
h
o
f
th
e
f
o
u
r
ex
p
er
i
m
en
tal
m
o
d
els.
T
h
e
o
v
er
all
Fig
u
r
e
2
r
ep
r
es
en
ts
th
e
wo
r
k
f
lo
w
o
f
th
e
p
lu
g
in
.
Sp
ec
if
ically
,
Fig
u
r
e
2
(
a)
illu
s
tr
ates
th
e
im
ag
e
p
r
ep
r
o
ce
s
s
in
g
p
h
ase
,
w
h
er
e
im
ag
es
ar
e
n
o
r
m
alize
d
t
o
2
2
4
×
2
2
4
p
ix
els.
Fig
u
r
e
2
(
b
)
d
etails th
e
f
ea
tu
r
e
ex
tr
ac
tio
n
lay
er
s
wh
e
r
e
th
e
m
o
d
el
id
en
tifie
s
illi
cit
p
atter
n
s
.
T
h
e
p
lu
g
in
was
s
u
cc
ess
f
u
lly
in
s
talled
an
d
o
p
er
ated
with
i
n
Go
o
g
le
C
h
r
o
m
e
.
I
t
f
u
n
ctio
n
s
in
two
m
o
d
es:
-
Ver
if
icatio
n
m
o
d
e:
a
n
a
ly
s
es in
d
iv
id
u
al
u
p
lo
ad
e
d
im
ag
es.
-
I
d
en
tific
atio
n
m
o
d
e:
co
m
p
ar
es
u
p
lo
ad
e
d
co
n
ten
t a
g
ain
s
t th
e
t
r
ain
in
g
d
ata
b
ase.
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:
2502
-
4
7
5
2
Hyb
r
id
p
lu
g
in
fo
r
d
etec
tin
g
illi
cit
ima
g
es o
n
th
e
in
tern
et
u
s
i
n
g
E
fficien
tN
et
…
(
C
h
r
is
tin
e
L
a
u
r
e
Ma
n
a
n
g
a
)
813
T
ab
le
2
.
T
ests
r
esu
lts
f
o
r
th
e
f
o
u
r
m
o
d
els
M
o
d
e
l
s
C
o
n
v
o
l
u
t
i
o
n
l
a
y
e
r
A
r
c
h
i
t
e
c
t
u
r
e
u
se
d
F
u
l
l
y
c
o
n
n
e
c
t
e
d
N
u
mb
e
r
of
e
p
o
c
h
s
Er
r
o
r
A
c
c
u
r
a
c
y
V
a
l
i
d
a
t
i
o
n
on
e
r
r
o
r
V
a
l
i
d
a
t
i
o
n
on
a
c
c
u
r
a
c
y
Ex
e
c
u
t
i
o
n
t
i
m
e
M
o
d
e
l
1
5
2
3
10
0
.
0
9
3
0
.
9
7
5
0
.
1
1
8
0
.
9
6
4
6
1
3
2
s
M
o
d
e
l
2
6
3
3
10
0
.
0
3
9
0
.
9
7
1
0
.
0
7
7
0
.
9
7
8
8
1
2
2
s
M
o
d
e
l
3
4
2
2
10
0
.
0
4
3
0
.
9
6
0
.
1
9
2
0
.
9
5
3
1
1
2
2
s
M
o
d
e
l
4
6
3
3
9
0
.
0
1
7
0
.
9
8
0
.
0
0
7
0
.
9
9
5
2
92s
(
a)
(
b
)
Fig
u
r
e
2
.
R
ep
r
esen
ts
th
e
wo
r
k
f
lo
w
o
f
th
e
p
lu
g
i
n
(
a)
m
o
d
el
4
ac
cu
r
ac
y
r
ate
o
f
o
u
r
m
o
d
el
a
n
d
(
b
)
m
o
d
el
4
e
r
r
o
r
r
ate
T
h
e
p
r
o
ce
d
u
r
e
f
o
r
i
n
s
tallin
g
o
u
r
p
l
u
g
in
in
t
h
e
Go
o
g
le
C
h
r
o
m
e
b
r
o
wser
is
as
f
o
llo
ws:
Op
en
th
e
Go
o
g
le
C
h
r
o
m
e
b
r
o
wser
.
I
n
th
e
‘
C
u
s
to
m
is
e
an
d
co
n
tr
o
l
Go
o
g
le
C
h
r
o
m
e
’
m
en
u
,
s
elec
t
‘
E
x
ten
s
io
n
s
’
,
th
en
click
o
n
‘
Ma
n
ag
e
ex
ten
s
io
n
s
.
’
E
n
s
u
r
e
th
at
d
ev
elo
p
er
m
o
d
e
i
s
en
ab
led
.
C
lick
o
n
‘
L
o
a
d
u
n
p
ac
k
ed
ex
te
n
s
io
n
’
,
th
en
n
av
i
g
ate
to
an
d
s
elec
t th
e
f
o
ld
er
c
o
n
tain
in
g
th
e
p
l
u
g
in
.
Fig
u
r
e
3
:
p
lu
g
i
n
in
teg
r
atio
n
in
th
e
C
h
r
o
m
e
B
r
o
wser
.
T
h
i
s
im
ag
e
d
is
p
lay
s
th
e
Go
o
g
le
C
h
r
o
m
e
E
x
ten
s
io
n
s
m
an
ag
em
e
n
t in
ter
f
ac
e.
−
I
n
s
tallatio
n
:
i
t
s
h
o
ws
th
at
th
e
cu
s
to
m
-
d
ev
el
o
p
ed
p
lu
g
i
n
(
lab
eled
“
Pre
m
ier
e
ex
te
n
s
io
n
0
.
1
”
)
h
as
b
ee
n
su
cc
ess
f
u
lly
lo
ad
ed
in
to
t
h
e
b
r
o
wser
u
s
in
g
“
Dev
elo
p
e
r
Mo
d
e
.
”
−
Op
er
atio
n
al
s
tatu
s
:
th
e
to
g
g
le
s
witch
is
s
et
to
“
Activ
e,
”
in
d
icatin
g
th
at
th
e
p
lu
g
in
is
r
u
n
n
in
g
in
th
e
b
ac
k
g
r
o
u
n
d
to
in
ter
ce
p
t a
n
d
a
n
aly
ze
web
tr
af
f
ic
in
r
ea
l tim
e.
−
Dep
lo
y
m
en
t:
t
h
is
d
em
o
n
s
tr
ate
s
th
e
“
clien
t
-
s
id
e
”
n
atu
r
e
o
f
y
o
u
r
s
o
lu
tio
n
,
allo
win
g
f
o
r
im
m
ed
iate
co
n
ten
t
f
ilter
in
g
d
ir
ec
tly
o
n
th
e
u
s
er
’
s
co
m
p
u
ter
with
o
u
t r
ely
i
n
g
o
n
e
x
ter
n
al
s
er
v
er
laten
c
y
.
Fig
u
r
e
3
.
I
n
s
tallin
g
th
e
p
lu
g
in
in
Go
o
g
le
C
h
r
o
m
e
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
SN
:
2
5
0
2
-
4
7
5
2
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
,
Vo
l.
42
,
No
.
3
,
J
u
n
e
20
2
6
:
8
0
9
-
81
7
814
I
n
s
tallatio
n
is
u
s
er
-
f
r
ien
d
ly
an
d
f
o
llo
ws
s
tan
d
ar
d
C
h
r
o
m
e
e
x
ten
s
io
n
p
r
o
ce
d
u
r
es.
W
h
en
illi
cit
co
n
ten
t
is
d
etec
ted
,
th
e
im
a
g
e
is
au
t
o
m
atica
lly
b
lo
ck
e
d
f
r
o
m
v
ie
w.
I
n
F
ig
u
r
e
4
,
we
h
a
v
e
r
ea
l
-
tim
e
d
etec
tio
n
a
n
d
co
n
ten
t
b
l
o
ck
in
g
.
T
h
is
im
ag
e
s
er
v
es
as
a
f
u
n
ctio
n
al
d
em
o
n
s
tr
atio
n
o
f
th
e
p
lu
g
in
’
s
cla
s
s
if
i
ca
tio
n
ca
p
ab
ilit
ies
d
u
r
in
g
a
liv
e
s
ea
r
ch
.
−
T
r
ig
g
er
in
g
c
o
n
ten
t
:
a
Go
o
g
le
i
m
ag
es
s
ea
r
ch
f
o
r
v
io
len
t
c
o
n
ten
t
(
e.
g
.
,
“
m
u
r
d
er
with
b
lo
o
d
”
)
h
as
b
ee
n
p
er
f
o
r
m
ed
to
test
th
e
s
y
s
tem
.
−
C
las
s
if
icatio
n
an
d
a
ctio
n
:
t
h
e
h
y
b
r
id
m
o
d
el,
p
o
wer
ed
b
y
E
f
f
icien
tNetB
7
,
h
as
s
u
cc
ess
f
u
lly
id
en
tifie
d
im
ag
es c
o
n
tain
in
g
b
lo
o
d
an
d
c
r
im
e
s
ce
n
es a
s
“
illi
cit
.
”
−
Vis
u
al
r
esu
lts
:
c
o
n
s
is
ten
t
with
th
e
f
in
d
i
n
g
s
in
y
o
u
r
s
tu
d
y
,
th
e
p
lu
g
in
h
as
ap
p
lied
a
b
lu
r
r
in
g
f
ilter
o
v
e
r
th
e
h
ar
m
f
u
l
im
a
g
es.
T
h
i
s
p
r
o
tects
th
e
u
s
er
f
r
o
m
v
iewin
g
s
en
s
itiv
e
m
ater
ial
wh
ile
allo
win
g
n
o
n
-
e
x
p
licit
o
r
s
y
m
b
o
lic
im
ag
es (
lik
e
a
“
Po
lic
e
L
in
e
”
tap
e
o
r
tex
t
-
b
ased
g
r
a
p
h
ics)
to
r
em
ain
v
is
ib
le.
−
Per
f
o
r
m
an
ce
:
t
h
is
co
n
f
ir
m
s
th
e
r
ep
o
r
te
d
9
9
%
ac
cu
r
ac
y
in
a
r
ea
l
-
wo
r
ld
s
ce
n
ar
io
,
s
p
ec
if
ic
ally
its
ab
ilit
y
to
d
is
tin
g
u
is
h
b
etwe
en
ac
tu
al
illi
cit
v
is
u
al
d
ata
an
d
SF
W
co
n
ten
t.
I
n
p
r
ac
tical
s
ce
n
ar
io
s
,
th
e
p
lu
g
in
s
u
cc
ess
f
u
lly
f
lag
g
ed
p
o
r
n
o
g
r
ap
h
ic
a
n
d
v
io
len
t
im
a
g
es
wh
ile
allo
win
g
s
af
e
co
n
ten
t t
o
p
ass
,
m
in
im
izin
g
f
alse p
o
s
itiv
es a
n
d
n
eg
ativ
es.
Fig
u
r
e
4
.
I
lle
g
al
im
ag
e
s
ea
r
ch
r
esu
lts
4
.
2
.
Dis
cus
s
io
n a
nd
co
m
pa
ra
t
iv
e
a
na
ly
s
is
Ou
r
s
tu
d
y
im
p
lem
e
n
ted
a
m
o
d
el
b
ased
o
n
E
f
f
icien
tNetB
7
to
ef
f
icien
tly
f
ilter
ac
ce
s
s
to
d
ata
o
n
th
e
I
n
ter
n
et,
s
p
ec
if
ically
f
o
r
th
e
d
etec
tio
n
an
d
class
if
icatio
n
o
f
i
llicit
im
ag
es.
I
n
o
r
d
er
to
s
itu
at
e
o
u
r
r
esu
lts
with
in
th
e
cu
r
r
e
n
t
s
cien
tific
co
n
tex
t,
T
ab
le
3
co
m
p
ar
es
o
u
r
p
er
f
o
r
m
an
ce
with
s
ev
er
al
r
ec
e
n
t
s
tu
d
i
es
th
at
also
ad
d
r
ess
I
n
ter
n
et
d
ata
ac
ce
s
s
f
ilter
in
g
.
T
h
e
r
esu
lts
o
b
tain
ed
in
th
is
s
t
u
d
y
a
r
e
h
ig
h
ly
co
m
p
etitiv
e,
o
f
ten
s
u
r
p
ass
in
g
th
o
s
e
r
ep
o
r
ted
in
r
ec
en
t
liter
atu
r
e
r
eg
ar
d
in
g
i
n
f
o
r
m
atio
n
f
ilter
in
g
v
ia
E
f
f
icien
tNet
-
b
ased
C
NNs
.
T
o
co
n
tex
tu
alize
o
u
r
f
i
n
d
in
g
s
,
we
co
m
p
ar
e
o
u
r
m
o
d
el
’
s
p
e
r
f
o
r
m
an
ce
with
s
ev
er
al
b
e
n
ch
m
a
r
k
s
tu
d
ies ac
r
o
s
s
v
ar
io
u
s
d
o
m
ai
n
s
.
Ou
r
ac
cu
r
ac
y
o
f
9
8
.
7
%
co
m
p
ar
es
f
av
o
r
a
b
ly
with
th
e
wo
r
k
o
f
T
asy
a
[
1
1
]
,
w
h
o
ac
h
iev
ed
a
class
if
icatio
n
ac
cu
r
ac
y
o
f
9
4
.
2
6
%
(
with
a
lo
s
s
r
ate
o
f
3
9
.
5
2
%)
in
th
e
m
ed
ical
d
o
m
ain
u
s
in
g
a
d
ataset
o
f
1
,
3
6
1
h
is
to
p
ath
o
lo
g
ical
im
ag
es.
Similar
ly
,
o
u
r
r
esu
lts
ar
e
co
n
s
is
ten
t w
ith
Gu
an
an
d
W
an
g
[
1
2
]
,
w
h
o
s
e
E
f
f
icien
tNet
-
b
ased
alg
o
r
ith
m
f
o
r
b
l
o
o
d
ce
ll
clas
s
if
icatio
n
r
ea
ch
ed
an
ac
cu
r
ac
y
o
f
9
8
.
6
%
o
n
t
h
e
B
C
C
D
d
atase
t.
T
h
e
v
er
s
atility
o
f
th
e
E
f
f
icien
t
Net
ar
ch
itectu
r
e
is
f
u
r
th
er
s
u
p
p
o
r
ted
b
y
Ach
ar
y
a
et
a
l
.
[
1
3
]
,
[
1
4]
.
T
h
eir
h
y
b
r
id
ap
p
r
o
ac
h
d
em
o
n
s
tr
ated
th
at
wh
ile
eE
NetB0
r
ea
ch
ed
a
m
a
x
im
u
m
ac
cu
r
ac
y
o
f
9
7
.
5
9
%
u
n
d
er
1
0
-
f
o
ld
cr
o
s
s
-
v
alid
atio
n
,
th
e
ef
f
icien
tNetB
7
v
ar
ian
t
(
s
im
ilar
to
o
u
r
c
o
r
e
ar
c
h
itectu
r
e)
ac
h
iev
ed
a
s
u
p
er
io
r
9
8
.
7
8
%.
T
h
is
r
ein
f
o
r
ce
s
o
u
r
s
elec
tio
n
o
f
th
e
B
7
v
ar
ian
t
f
o
r
h
ig
h
-
p
r
ec
i
s
io
n
task
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:
2502
-
4
7
5
2
Hyb
r
id
p
lu
g
in
fo
r
d
etec
tin
g
illi
cit
ima
g
es o
n
th
e
in
tern
et
u
s
i
n
g
E
fficien
tN
et
…
(
C
h
r
is
tin
e
L
a
u
r
e
Ma
n
a
n
g
a
)
815
T
ab
le
3
.
C
o
m
p
a
r
ativ
e
s
tu
d
y
o
f
o
u
r
r
esu
lts
with
o
th
e
r
wo
r
k
s
[
1
5
]
–
[
2
0
]
S
t
u
d
y
/
A
u
t
h
o
r
D
a
t
a
s
e
t
P
r
e
c
i
s
i
o
n
(
%)
Rec
a
l
l
(
%)
F1
-
sc
o
r
e
(
%)
M
e
t
h
o
d
o
l
o
g
i
c
a
l
f
e
a
t
u
r
e
s
O
b
serv
a
t
i
o
n
s
O
u
r
mo
d
e
l
M
u
l
t
i
-
c
l
a
ss
d
a
t
a
s
e
t
(
w
e
a
p
o
n
s,
n
u
d
i
t
y
,
c
l
o
t
h
i
n
g
)
9
8
.
7
9
8
.
0
9
8
.
0
Ef
f
i
c
i
e
n
t
N
e
t
B
7
f
i
n
e
-
t
u
n
i
n
g
,
r
e
g
u
l
a
r
i
z
a
t
i
o
n
,
d
a
t
a
a
u
g
m
e
n
t
a
t
i
o
n
Ex
c
e
l
l
e
n
t
p
e
r
f
o
r
ma
n
c
e
st
a
b
i
l
i
t
y
a
c
r
o
ss
a
l
l
c
l
a
ss
e
s
C
h
e
n
et
al
.
(
2
0
2
4
)
I
mag
e
N
e
t
+
N
S
F
W
+
w
e
a
p
o
n
s
d
a
t
a
se
t
9
7
.
1
9
6
.
8
9
6
.
9
A
d
v
a
n
c
e
d
a
u
g
me
n
t
a
t
i
o
n
t
e
c
h
n
i
q
u
e
s,
h
i
g
h
t
r
a
i
n
i
n
g
t
i
m
e
H
i
g
h
p
e
r
f
o
r
m
a
n
c
e
b
u
t
si
g
n
i
f
i
c
a
n
t
c
o
m
p
u
t
a
t
i
o
n
a
l
c
o
s
t
K
u
mar
et
al
.
(
2
0
2
3
)
P
o
r
n
o
g
r
a
p
h
y
+
W
e
a
p
o
n
s
d
a
t
a
s
e
t
9
5
.
9
9
5
.
3
9
5
.
6
U
se
of
t
r
a
n
sf
e
r
l
e
a
r
n
i
n
g
a
n
d
d
r
o
p
o
u
t
G
o
o
d
g
e
n
e
r
a
l
i
z
a
t
i
o
n
b
u
t
l
o
w
e
r
p
r
e
c
i
s
i
o
n
on
f
i
n
e
c
l
a
sse
s
G
a
r
c
í
a
et
al
.
(
2
0
2
4
)
W
e
b
-
f
i
l
t
e
r
e
d
m
u
l
t
i
-
d
a
t
a
se
t
9
6
.
3
9
5
.
7
9
6
.
0
F
u
si
o
n
of
se
v
e
r
a
l
C
N
N
a
r
c
h
i
t
e
c
t
u
r
e
s
,
m
u
l
t
i
-
d
a
t
a
se
t
t
r
a
i
n
i
n
g
F
u
se
d
C
N
N
a
r
c
h
i
t
e
c
t
u
r
e
,
i
n
c
r
e
a
s
e
d
c
o
m
p
l
e
x
i
t
y
S
mi
t
h
et
al
.
(
2
0
2
3
)
N
S
F
W
i
ma
g
e
s
o
n
l
y
9
5
.
2
9
4
.
7
9
4
.
9
P
a
r
t
i
a
l
t
r
a
i
n
i
n
g
,
f
e
w
e
r
t
r
a
i
n
a
b
l
e
l
a
y
e
r
s
Le
ss
e
f
f
e
c
t
i
v
e
on
h
e
t
e
r
o
g
e
n
e
o
u
s
d
a
t
a
Le
e
et
al
.
(
2
0
2
2
)
Li
g
h
t
w
e
i
g
h
t
C
N
N
(
p
o
r
n
&
g
u
n
s)
9
4
.
5
9
3
.
9
9
4
.
2
O
p
t
i
mi
z
a
t
i
o
n
f
o
r
s
p
e
e
d
,
l
i
g
h
t
e
r
m
o
d
e
l
D
e
si
g
n
e
d
f
o
r
s
p
e
e
d
,
l
i
m
i
t
e
d
a
c
c
u
r
a
c
y
I
n
ter
m
s
o
f
r
o
b
u
s
t
ev
alu
atio
n
,
Had
i
et
a
l.
[
2
1
]
,
[
2
2
]
u
tili
ze
d
co
n
f
u
s
io
n
m
at
r
ices
o
v
er
2
5
ep
o
ch
s
to
y
ield
av
er
ag
e
s
co
r
es
o
f
9
8
%
f
o
r
p
r
ec
is
io
n
,
r
ec
all,
a
n
d
F1
-
s
c
o
r
e,
r
esu
lts
th
at
m
ir
r
o
r
th
e
h
ig
h
s
tab
ilit
y
o
b
s
er
v
ed
in
o
u
r
o
wn
co
n
f
u
s
io
n
m
atr
ix
an
aly
s
is
.
Fu
r
th
er
m
o
r
e,
t
h
e
c
o
m
p
ar
ativ
e
s
tu
d
y
b
y
A
g
g
ar
w
al
et
a
l
.
[
2
3
]
,
[
2
4
]
co
n
f
ir
m
s
th
e
s
u
p
e
r
io
r
ity
o
f
o
u
r
ch
o
s
en
m
o
d
el;
in
th
eir
test
s
,
E
f
f
icien
tNetB
7
(
9
9
%)
s
ig
n
if
ic
an
tly
o
u
tp
e
r
f
o
r
m
ed
b
o
th
VGG
-
1
6
(
9
7
.
6
7
%)
a
n
d
I
n
ce
p
tio
n
V3
(
9
7
.
2
%).
T
h
e
r
o
le
o
f
d
ata
au
g
m
en
tatio
n
,
a
k
ey
co
m
p
o
n
en
t
o
f
o
u
r
p
r
ep
r
o
ce
s
s
in
g
p
h
ase,
was
also
h
ig
h
lig
h
ted
b
y
R
am
an
et
a
l.
[
2
5
]
,
[
2
6
]
,
w
h
o
ac
h
iev
e
d
9
4
%
ac
cu
r
ac
y
b
y
u
s
in
g
au
g
m
en
tatio
n
to
m
iti
g
ate
o
v
er
f
itti
n
g
an
d
lev
er
ag
in
g
E
f
f
icien
tNet
’
s
u
n
if
o
r
m
s
ca
lin
g
o
f
d
ep
t
h
,
wid
th
,
a
n
d
r
eso
lu
tio
n
.
Fi
n
ally
,
wh
ile
ad
v
a
n
ce
d
ar
ch
i
tectu
r
es
lik
e
th
e
E
GW
T
p
r
o
p
o
s
ed
b
y
Fen
g
et
a
l
.
[
2
7
]
h
av
e
r
ea
ch
ed
n
ea
r
-
p
er
f
ec
t
ac
cu
r
ac
y
(
9
9
.
8
%)
in
s
p
ec
ialized
ag
r
icu
ltu
r
al
m
o
n
ito
r
in
g
,
o
u
r
m
o
d
el
p
r
o
v
id
es
a
m
o
r
e
g
en
er
alize
d
an
d
lig
h
tweig
h
t
clien
t
-
s
id
e
s
o
lu
tio
n
.
E
v
en
wh
e
n
co
m
p
ar
ed
to
h
y
b
r
i
d
m
o
d
els
s
u
ch
as
th
e
E
f
f
icien
tNet
-
SVM
p
ip
elin
e
in
tr
o
d
u
ce
d
b
y
An
u
g
r
ah
et
a
l
.
[
2
8
]
–
[
3
0
]
,
wh
ich
r
ec
o
r
d
ed
a
r
o
b
u
s
t
9
6
%
in
m
a
lar
ia
d
etec
tio
n
,
o
u
r
in
teg
r
ated
p
lu
g
in
m
ai
n
tain
s
a
h
ig
h
er
p
r
ec
is
io
n
r
ate
f
o
r
r
ea
l
-
tim
e
web
co
n
te
n
t f
ilter
in
g
.
5.
CO
NCLU
SI
O
N
T
h
is
s
tu
d
y
s
u
cc
ess
f
u
lly
d
esig
n
ed
an
d
im
p
lem
en
ted
a
h
y
b
r
id
b
r
o
wser
-
in
teg
r
ated
p
l
u
g
in
u
s
in
g
th
e
E
f
f
icien
tNetB
7
ar
ch
itectu
r
e
to
f
ilter
an
d
b
lo
c
k
illi
cit
I
n
ter
n
et
co
n
ten
t
in
r
ea
l
tim
e.
B
y
b
r
id
g
in
g
th
e
g
ap
b
etwe
en
tr
ad
itio
n
al
co
n
ten
t
-
b
ased
f
ilter
in
g
(
m
eta
d
ata,
s
ig
n
atu
r
es)
an
d
d
ee
p
lear
n
in
g
s
tr
u
ctu
r
al
f
ea
tu
r
e
ex
tr
ac
tio
n
,
t
h
is
r
esear
ch
p
r
o
v
i
d
es
a
s
ca
lab
le
an
d
ef
f
ec
tiv
e
s
o
lu
tio
n
f
o
r
m
ain
tain
in
g
s
af
e
d
ig
ital
en
v
ir
o
n
m
en
ts
.
T
h
e
ex
p
er
im
e
n
tal
r
esu
lts
v
ali
d
ate
th
e
r
o
b
u
s
tn
ess
o
f
o
u
r
a
p
p
r
o
ac
h
,
with
th
e
o
p
tim
ized
Mo
d
el
4
ac
h
iev
in
g
a
d
etec
tio
n
ac
cu
r
ac
y
o
f
9
8
.
7
%
an
d
a
p
r
ec
is
io
n
s
co
r
e
o
f
9
9
%.
T
h
e
in
teg
r
atio
n
o
f
th
e
m
o
d
el
in
to
a
Go
o
g
le
C
h
r
o
m
e
ex
te
n
s
io
n
d
em
o
n
s
tr
at
ed
its
p
r
ac
tical
u
tili
ty
,
ef
f
ec
tiv
ely
b
l
u
r
r
in
g
h
ar
m
f
u
l
im
ag
es
r
elate
d
t
o
v
io
len
ce
an
d
p
o
r
n
o
g
r
a
p
h
y
d
ir
ec
tly
o
n
th
e
clie
n
t
s
id
e.
T
h
is
d
ec
en
tr
alize
d
d
ep
lo
y
m
en
t
s
ig
n
if
ican
t
ly
r
ed
u
ce
s
laten
cy
co
m
p
ar
ed
to
tr
a
d
itio
n
al
s
er
v
e
r
-
s
id
e
f
ilter
in
g
.
Ou
r
co
m
p
ar
a
tiv
e
an
aly
s
is
co
n
f
ir
m
s
th
at
th
is
h
y
b
r
id
s
o
lu
tio
n
n
o
t
o
n
ly
m
atch
es
b
u
t
o
f
te
n
s
u
r
p
ass
es
co
n
tem
p
o
r
ar
y
s
ta
te
-
of
-
th
e
-
a
r
t
m
o
d
els
in
te
r
m
s
o
f
p
r
ec
is
io
n
an
d
s
en
s
itiv
ity
.
Desp
ite
th
ese
ac
h
iev
em
en
ts
,
th
is
r
esear
ch
f
ac
ed
ce
r
tain
lim
itatio
n
s
.
T
h
e
co
n
s
tr
u
ctio
n
o
f
h
ig
h
-
q
u
ality
g
r
o
u
n
d
tr
u
t
h
d
atasets
r
em
ain
s
an
ex
p
en
s
iv
e
an
d
r
eso
u
r
ce
-
in
ten
s
iv
e
p
r
o
ce
s
s
.
Fu
r
th
e
r
m
o
r
e,
t
h
e
cu
r
r
en
t
d
ataset
lack
s
d
iv
er
s
ity
r
eg
ar
d
in
g
n
o
n
-
h
u
m
an
o
b
jects;
it
wo
u
ld
b
e
s
cien
tific
ally
v
alu
ab
le
to
in
clu
d
e
an
im
als
o
r
s
p
ec
if
ic
tex
tu
r
es
th
at
co
u
ld
b
e
ea
s
ily
co
n
f
o
u
n
d
ed
with
h
u
m
a
n
s
k
in
o
r
clo
th
ed
in
d
iv
i
d
u
als
t
o
f
u
r
th
er
r
ef
in
e
th
e
m
o
d
el
’
s
d
is
cr
im
in
ativ
e
p
o
we
r
.
Fu
tu
r
e
r
esear
ch
will
ai
m
to
ad
d
r
ess
th
ese
ch
allen
g
es.
Key
p
er
s
p
ec
tiv
es
in
clu
d
e:
i.
E
x
p
an
d
in
g
th
e
d
ataset
to
in
clu
d
e
am
b
ig
u
o
u
s
n
o
n
-
h
u
m
an
tex
tu
r
es
to
r
ed
u
ce
p
o
ten
tial
f
alse
p
o
s
itiv
es
,
ii.
Ad
ap
tin
g
th
e
a
lg
o
r
ith
m
f
o
r
r
ea
l
-
tim
e
v
id
eo
s
tr
ea
m
an
aly
s
is
,
iii.
Dev
elo
p
in
g
an
A
n
d
r
o
i
d
-
co
m
p
atib
le
v
e
r
s
io
n
(
.
a
p
k
)
to
e
x
ten
d
p
r
o
tectio
n
to
m
o
b
ile
s
o
cial
n
etwo
r
k
s
lik
e
W
h
atsAp
p
,
an
d
i
v
.
E
x
p
an
d
i
n
g
th
e
s
y
s
tem
’
s
ap
p
licatio
n
to
s
en
s
itiv
e
s
ec
to
r
s
s
u
ch
as
h
ea
lth
c
ar
e
an
d
ed
u
ca
tio
n
,
en
s
u
r
in
g
ta
ilo
r
ed
p
ar
e
n
tal
an
d
p
ed
ag
o
g
ical
co
n
tr
o
ls
.
AUTHO
R
CO
NT
RI
B
UT
I
O
NS ST
A
T
E
M
E
N
T
T
h
is
jo
u
r
n
al
u
s
es
th
e
C
o
n
t
r
ib
u
to
r
R
o
les
T
a
x
o
n
o
m
y
(
C
R
ed
iT)
to
r
ec
o
g
n
ize
in
d
iv
i
d
u
al
au
th
o
r
co
n
tr
ib
u
tio
n
s
,
r
ed
u
ce
au
th
o
r
s
h
ip
d
is
p
u
tes,
an
d
f
ac
ilit
ate
co
llab
o
r
atio
n
.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
SN
:
2
5
0
2
-
4
7
5
2
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
,
Vo
l.
42
,
No
.
3
,
J
u
n
e
20
2
6
:
8
0
9
-
81
7
816
Na
m
e
o
f
Aut
ho
r
C
M
So
Va
Fo
I
R
D
O
E
Vi
Su
P
Fu
C
h
r
is
tin
e
L
au
r
e
Ma
n
an
g
a
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
Felix
Pau
n
e
✓
✓
✓
✓
✓
✓
✓
L
éa
n
d
r
e
Nn
e
m
e
Nn
em
e
✓
✓
✓
✓
✓
C
:
C
o
n
c
e
p
t
u
a
l
i
z
a
t
i
o
n
M
:
M
e
t
h
o
d
o
l
o
g
y
So
:
So
f
t
w
a
r
e
Va
:
Va
l
i
d
a
t
i
o
n
Fo
:
Fo
r
mal
a
n
a
l
y
s
i
s
I
:
I
n
v
e
s
t
i
g
a
t
i
o
n
R
:
R
e
so
u
r
c
e
s
D
:
D
a
t
a
C
u
r
a
t
i
o
n
O
:
W
r
i
t
i
n
g
-
O
r
i
g
i
n
a
l
D
r
a
f
t
E
:
W
r
i
t
i
n
g
-
R
e
v
i
e
w
&
E
d
i
t
i
n
g
Vi
:
Vi
su
a
l
i
z
a
t
i
o
n
Su
:
Su
p
e
r
v
i
s
i
o
n
P
:
P
r
o
j
e
c
t
a
d
mi
n
i
st
r
a
t
i
o
n
Fu
:
Fu
n
d
i
n
g
a
c
q
u
i
si
t
i
o
n
DATA AV
AI
L
AB
I
L
I
T
Y
-
Der
iv
ed
d
ata
s
u
p
p
o
r
tin
g
th
e
f
in
d
in
g
s
o
f
th
is
s
tu
d
y
ar
e
av
ailab
le
f
r
o
m
th
e
co
r
r
esp
o
n
d
in
g
a
u
th
o
r
[
C
h
r
is
tin
e
L
au
r
e
MA
NANG
A
,
C
L
M]
o
n
r
e
q
u
est.
-
T
h
e
au
th
o
r
s
co
n
f
ir
m
th
at
th
e
d
ata
s
u
p
p
o
r
tin
g
th
e
f
in
d
in
g
s
o
f
th
is
s
tu
d
y
ar
e
av
ailab
le
in
t
h
is
p
ap
er
.
-
T
h
e
d
ata
t
h
at
s
u
p
p
o
r
t
t
h
e
f
in
d
in
g
s
o
f
th
is
s
tu
d
y
ar
e
av
ailab
le
f
r
o
m
th
e
co
r
r
esp
o
n
d
in
g
au
th
o
r
,
[
C
h
r
is
tin
e
L
au
r
e
MA
NANG
A
,
C
L
M]
u
p
o
n
r
ea
s
o
n
ab
le
r
eq
u
e
s
t.
RE
F
E
R
E
NC
E
S
[
1
]
A
.
K
r
i
z
h
e
v
s
k
y
,
I
.
S
u
t
s
k
e
v
e
r
,
a
n
d
G
.
E.
H
i
n
t
o
n
,
“
I
mag
e
N
e
t
c
l
a
ssi
f
i
c
a
t
i
o
n
w
i
t
h
d
e
e
p
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
u
r
a
l
n
e
t
w
o
r
k
s
,
”
C
o
m
m
u
n
i
c
a
t
i
o
n
s
o
f
t
h
e
A
C
M
,
v
o
l
.
6
0
,
n
o
.
6
,
p
p
.
8
4
–
9
0
,
2
0
1
7
,
d
o
i
:
1
0
.
1
1
4
5
/
3
0
6
5
3
8
6
.
[
2
]
M
.
T
a
n
a
n
d
Q
.
V
.
L
e
,
“
Ef
f
i
c
i
e
n
t
N
e
t
:
r
e
t
h
i
n
k
i
n
g
m
o
d
e
l
s
c
a
l
i
n
g
f
o
r
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,
”
3
6
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
M
a
c
h
i
n
e
L
e
a
rn
i
n
g
,
I
C
ML
2
0
1
9
,
v
o
l
.
2
0
1
9
-
Ju
n
e
,
p
p
.
1
0
6
9
1
–
1
0
7
0
0
,
2
0
1
9
.
[
3
]
“
Y
a
h
o
o
O
p
e
n
N
S
F
W
mo
d
e
l
.
”
h
t
t
p
s
:
/
/
g
i
t
h
u
b
.
c
o
m
/
y
a
h
o
o
/
o
p
e
n
_
n
sf
w
.
[
4
]
Te
n
s
o
r
f
l
o
w
,
“
T
e
n
s
o
r
f
l
o
w
d
o
c
u
m
e
n
t
a
t
i
o
n
,
”
H
t
t
p
s:
/
/
W
w
w
.
T
e
n
so
r
f
l
o
w
.
O
rg
/
L
e
a
r
n
,
p
.
4
,
2
0
2
2
,
[
O
n
l
i
n
e
]
.
A
v
a
i
l
a
b
l
e
:
h
t
t
p
s
:
/
/
w
w
w
.
t
e
n
s
o
r
f
l
o
w
.
o
r
g
/
.
[
5
]
B
.
J
,
“
D
e
e
p
l
e
a
r
n
i
n
g
p
e
r
f
o
r
m
a
n
c
e
,
ma
c
h
i
n
e
l
e
a
r
n
i
n
g
mas
t
e
r
y
,
”
2
0
1
8
.
[
6
]
R
.
S
i
n
g
h
,
S
.
G
u
p
t
a
,
S
.
B
h
a
r
a
n
y
,
A
.
A
l
mo
g
r
e
n
,
A
.
A
l
t
a
mee
m,
a
n
d
A
.
U
r
R
e
h
m
a
n
,
“
E
n
sem
b
l
e
d
e
e
p
l
e
a
r
n
i
n
g
mo
d
e
l
s
f
o
r
e
n
h
a
n
c
e
d
b
r
a
i
n
t
u
mo
r
c
l
a
ss
i
f
i
c
a
t
i
o
n
b
y
l
e
v
e
r
a
g
i
n
g
R
e
sN
e
t
5
0
a
n
d
Ef
f
i
c
i
e
n
t
N
e
t
-
B
7
o
n
h
i
g
h
-
r
e
s
o
l
u
t
i
o
n
M
R
I
i
ma
g
e
s
,
”
I
EEE
Ac
c
e
ss
,
v
o
l
.
1
2
,
p
p
.
1
7
8
6
2
3
–
1
7
8
6
4
1
,
2
0
2
4
,
d
o
i
:
1
0
.
1
1
0
9
/
A
C
C
ESS
.
2
0
2
4
.
3
4
9
4
2
3
2
.
[
7
]
N
.
P
a
p
e
r
n
o
t
,
A
.
T
h
a
k
u
r
t
a
,
S
.
S
o
n
g
,
S
.
C
h
i
e
n
,
a
n
d
Ú
.
Er
l
i
n
g
ss
o
n
,
“
T
e
m
p
e
r
e
d
si
g
m
o
i
d
a
c
t
i
v
a
t
i
o
n
s
f
o
r
d
e
e
p
l
e
a
r
n
i
n
g
w
i
t
h
d
i
f
f
e
r
e
n
t
i
a
l
p
r
i
v
a
c
y
,
”
3
5
t
h
AAAI
C
o
n
f
e
r
e
n
c
e
o
n
Art
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
,
A
AA
I
2
0
2
1
,
v
o
l
.
1
0
B
,
p
p
.
9
3
1
2
–
9
3
2
1
,
2
0
2
1
,
d
o
i
:
1
0
.
1
6
0
9
/
a
a
a
i
.
v
3
5
i
1
0
.
1
7
1
2
3
.
[
8
]
D
.
F
.
a
n
d
A
.
M
a
n
z
a
n
e
r
a
,
“
C
l
a
ss
i
f
i
c
a
t
i
o
n
d
’
i
ma
g
e
s
p
a
r
C
N
N
,
”
C
o
u
rs
MI2
0
4
,
EN
S
T
A
-
P
a
r
i
s
2
e
a
n
n
é
e
,
M
a
rs
,
2
0
2
0
.
[
9
]
A
.
K
u
m
a
r
,
“
I
mag
e
c
l
a
ss
i
f
i
c
a
t
i
o
n
o
n
C
I
F
A
R
1
0
u
si
n
g
n
e
u
r
a
l
n
e
t
w
o
r
k
s
(
N
N
)
,
”
2
0
2
0
.
h
t
t
p
s:
/
/
m
e
d
i
u
m.c
o
m
/
@
a
v
i
n
a
s
h
sh
a
h
0
9
9
/
i
ma
g
e
-
c
l
a
ss
i
f
i
c
a
t
i
o
n
-
on
-
c
i
f
a
r
1
0
-
u
si
n
g
-
n
e
u
r
a
l
-
n
e
t
w
o
r
k
s
(
a
c
c
e
ss
e
d
N
o
v
.
1
0
,
2
0
2
2
)
.
[
1
0
]
B
.
C
h
a
a
b
a
n
e
,
S
.
B
.
F
.
B
o
u
ss
e
ma
,
a
n
d
F
.
K
.
A
l
l
o
u
c
h
e
,
“
A
p
p
o
r
t
d
e
l
a
ma
t
r
i
c
e
d
e
c
o
n
f
u
si
o
n
d
a
n
s
l
’
é
v
a
l
u
a
t
i
o
n
d
u
c
h
a
n
g
e
m
e
n
t
d
u
p
a
y
sa
g
e
e
t
d
e
l
’
o
c
c
u
p
a
t
i
o
n
d
u
so
l
:
c
a
s
d
e
l
a
z
o
n
e
d
u
g
r
a
n
d
S
f
a
x
,
”
U
rb
a
n
Ar
t
B
i
o
,
v
o
l
.
2
,
n
o
.
1
,
p
p
.
1
4
–
29
,
2
0
2
3
,
d
o
i
:
1
0
.
3
5
7
8
8
/
u
a
b
.
v
2
i
1
.
5
9
.
[
1
1
]
W
.
Ta
s
y
a
,
S
.
S
a
’
I
d
a
h
,
B
.
H
i
d
a
y
a
t
,
a
n
d
F
.
N
u
r
f
a
j
a
r
,
“
B
r
e
a
st
c
a
n
c
e
r
d
e
t
e
c
t
i
o
n
u
s
i
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
w
i
t
h
Ef
f
i
c
i
e
n
t
N
e
t
a
r
c
h
i
t
e
c
t
u
r
e
,
”
A
PW
i
M
o
b
2
0
2
2
-
Pr
o
c
e
e
d
i
n
g
s:
2
0
2
2
I
EEE
As
i
a
Pa
c
i
f
i
c
C
o
n
f
e
re
n
c
e
o
n
Wi
re
l
e
s
s
a
n
d
Mo
b
i
l
e
,
2
0
2
2
,
d
o
i
:
1
0
.
1
1
0
9
/
A
P
W
i
M
o
b
5
6
8
5
6
.
2
0
2
2
.
1
0
0
1
4
0
9
5
.
[
1
2
]
Y
.
G
u
a
n
a
n
d
Z.
W
a
n
g
,
“
B
l
o
o
d
c
e
l
l
i
mag
e
r
e
c
o
g
n
i
t
i
o
n
a
l
g
o
r
i
t
h
m
b
a
se
d
o
n
Ef
f
i
c
i
e
n
t
N
e
t
,
”
2
0
2
2
I
EE
E
I
n
t
e
rn
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
Me
c
h
a
t
r
o
n
i
c
s
a
n
d
A
u
t
o
m
a
t
i
o
n
,
I
C
M
A
2
0
2
2
,
p
p
.
1
6
4
0
–
1
6
4
5
,
2
0
2
2
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
M
A
5
4
5
1
9
.
2
0
2
2
.
9
8
5
6
1
9
2
.
[
1
3
]
D
.
A
c
h
a
r
y
a
,
R
.
K
.
S
.
G
u
d
a
,
a
n
d
K
.
R
a
o
v
e
n
k
a
t
a
j
a
mm
a
l
a
ma
d
a
k
a
,
“
E
n
h
a
n
c
e
d
Ef
f
i
c
i
e
n
t
N
e
t
n
e
t
w
o
r
k
f
o
r
c
l
a
ss
i
f
y
i
n
g
l
a
p
a
r
o
sc
o
p
y
v
i
d
e
o
s
u
s
i
n
g
t
r
a
n
sf
e
r
l
e
a
r
n
i
n
g
t
e
c
h
n
i
q
u
e
,
”
Pro
c
e
e
d
i
n
g
s
o
f
t
h
e
I
n
t
e
r
n
a
t
i
o
n
a
l
J
o
i
n
t
C
o
n
f
e
re
n
c
e
o
n
N
e
u
r
a
l
N
e
t
w
o
rks
,
2
0
2
2
,
d
o
i
:
1
0
.
1
1
0
9
/
I
JC
N
N
5
5
0
6
4
.
2
0
2
2
.
9
8
9
1
9
8
9
.
[
1
4
]
N
.
G
e
sser
t
,
M
.
N
i
e
l
se
n
,
M
.
S
h
a
i
k
h
,
R
.
W
e
r
n
e
r
,
a
n
d
A
.
S
c
h
l
a
e
f
e
r
,
“
S
k
i
n
l
e
si
o
n
c
l
a
ssi
f
i
c
a
t
i
o
n
u
si
n
g
e
n
s
e
mb
l
e
s
o
f
m
u
l
t
i
-
r
e
s
o
l
u
t
i
o
n
Ef
f
i
c
i
e
n
t
N
e
t
s wi
t
h
me
t
a
d
a
t
a
,
”
Me
t
h
o
d
sX
,
v
o
l
.
7
,
2
0
2
0
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
m
e
x
.
2
0
2
0
.
1
0
0
8
6
4
.
[
1
5
]
J.
T.
H
s
u
e
t
a
l
.
,
“
C
h
r
o
n
i
c
w
o
u
n
d
a
ss
e
ssm
e
n
t
a
n
d
i
n
f
e
c
t
i
o
n
d
e
t
e
c
t
i
o
n
me
t
h
o
d
,
”
B
MC
Me
d
i
c
a
l
I
n
f
o
rm
a
t
i
c
s
a
n
d
D
e
c
i
s
i
o
n
M
a
k
i
n
g
,
v
o
l
.
1
9
,
n
o
.
1
,
2
0
1
9
,
d
o
i
:
1
0
.
1
1
8
6
/
s1
2
9
1
1
-
019
-
0
8
1
3
-
0.
[
1
6
]
C
.
W
a
n
g
e
t
a
l
.
,
“
A
u
n
i
f
i
e
d
f
r
a
mew
o
r
k
f
o
r
a
u
t
o
m
a
t
i
c
w
o
u
n
d
se
g
me
n
t
a
t
i
o
n
a
n
d
a
n
a
l
y
si
s
w
i
t
h
d
e
e
p
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
u
r
a
l
n
e
t
w
o
r
k
s
,
”
Pro
c
e
e
d
i
n
g
s o
f
t
h
e
A
n
n
u
a
l
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
f
t
h
e
I
EE
E
E
n
g
i
n
e
e
ri
n
g
i
n
Me
d
i
c
i
n
e
a
n
d
Bi
o
l
o
g
y
S
o
c
i
e
t
y
,
EM
B
S
,
v
o
l
.
2
0
1
5
-
N
o
v
e
mb
e
r
,
p
p
.
2
4
1
5
–
2
4
1
8
,
2
0
1
5
,
d
o
i
:
1
0
.
1
1
0
9
/
E
M
B
C
.
2
0
1
5
.
7
3
1
8
8
8
1
.
[
1
7
]
H
.
N
e
j
a
t
i
e
t
a
l
.
,
“
F
i
n
e
-
g
r
a
i
n
e
d
w
o
u
n
d
t
i
ssu
e
a
n
a
l
y
s
i
s
u
s
i
n
g
d
e
e
p
n
e
u
r
a
l
n
e
t
w
o
r
k
,
”
I
C
A
S
S
P,
I
EE
E
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
r
e
n
c
e
o
n
Ac
o
u
s
t
i
c
s
,
S
p
e
e
c
h
a
n
d
S
i
g
n
a
l
Pro
c
e
ss
i
n
g
-
Pro
c
e
e
d
i
n
g
s
,
v
o
l
.
2
0
1
8
-
A
p
r
i
l
,
p
p
.
1
0
1
0
–
1
0
1
4
,
2
0
1
8
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
A
S
S
P
.
2
0
1
8
.
8
4
6
1
9
2
7
.
[
1
8
]
M
.
G
o
y
a
l
,
N
.
D
.
R
e
e
v
e
s
,
A
.
K
.
D
a
v
i
s
o
n
,
S
.
R
a
j
b
h
a
n
d
a
r
i
,
J.
S
p
r
a
g
g
,
a
n
d
M
.
H
.
Y
a
p
,
“
D
F
U
N
e
t
:
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
f
o
r
d
i
a
b
e
t
i
c
f
o
o
t
u
l
c
e
r
c
l
a
ss
i
f
i
c
a
t
i
o
n
,
”
I
EEE
T
r
a
n
s
a
c
t
i
o
n
s
o
n
Em
e
r
g
i
n
g
T
o
p
i
c
s
i
n
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
,
v
o
l
.
4
,
n
o
.
5
,
p
p
.
7
2
8
–
7
3
9
,
2
0
1
8
,
d
o
i
:
1
0
.
1
1
0
9
/
t
e
t
c
i
.
2
0
1
8
.
2
8
6
6
2
5
4
.
[
1
9
]
N
.
A
l
-
G
a
r
a
a
w
i
,
R
.
E
b
si
m
,
A
.
F
.
H
.
A
l
h
a
r
a
n
,
a
n
d
M
.
H
.
Y
a
p
,
“
D
i
a
b
e
t
i
c
f
o
o
t
u
l
c
e
r
c
l
a
ssi
f
i
c
a
t
i
o
n
u
s
i
n
g
m
a
p
p
e
d
b
i
n
a
r
y
p
a
t
t
e
r
n
s
a
n
d
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
u
r
a
l
n
e
t
w
o
r
k
s,
”
C
o
m
p
u
t
e
rs
i
n
B
i
o
l
o
g
y
a
n
d
Me
d
i
c
i
n
e
,
v
o
l
.
1
4
0
,
2
0
2
2
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
c
o
m
p
b
i
o
me
d
.
2
0
2
1
.
1
0
5
0
5
5
.
[
2
0
]
I
.
S
.
A
.
A
b
d
e
l
h
a
l
i
m,
M
.
F
.
M
o
h
a
me
d
,
a
n
d
Y
.
B
.
M
a
h
d
y
,
“
D
a
t
a
a
u
g
me
n
t
a
t
i
o
n
f
o
r
s
k
i
n
l
e
si
o
n
u
si
n
g
s
e
l
f
-
a
t
t
e
n
t
i
o
n
b
a
se
d
p
r
o
g
r
e
ss
i
v
e
g
e
n
e
r
a
t
i
v
e
a
d
v
e
r
sari
a
l
n
e
t
w
o
r
k
,
”
E
x
p
e
rt
S
y
st
e
m
s
w
i
t
h
A
p
p
l
i
c
a
t
i
o
ns
,
v
o
l
.
1
6
5
,
2
0
2
1
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
e
sw
a
.
2
0
2
0
.
1
1
3
9
2
2
.
[
2
1
]
V
.
H
.
B
.
H
a
d
i
,
A
.
B
.
M
u
t
i
a
r
a
,
a
n
d
R
.
R
e
f
i
a
n
t
i
,
“
I
mp
l
e
me
n
t
a
t
i
o
n
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
w
i
t
h
Ef
f
i
c
i
e
n
t
N
e
t
-
B
0
a
r
c
h
i
t
e
c
t
u
r
e
f
o
r
b
r
a
i
n
t
u
m
o
r
c
l
a
ssi
f
i
c
a
t
i
o
n
,
”
2
0
2
3
8
t
h
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
I
n
f
o
rm
a
t
i
c
s
a
n
d
C
o
m
p
u
t
i
n
g
,
I
C
I
C
2
0
2
3
,
2
0
2
3
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
I
C
6
0
1
0
9
.
2
0
2
3
.
1
0
3
8
1
9
7
9
.
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:
2502
-
4
7
5
2
Hyb
r
id
p
lu
g
in
fo
r
d
etec
tin
g
illi
cit
ima
g
es o
n
th
e
in
tern
et
u
s
i
n
g
E
fficien
tN
et
…
(
C
h
r
is
tin
e
L
a
u
r
e
Ma
n
a
n
g
a
)
817
[
2
2
]
F
.
P
e
r
e
z
,
C
.
V
a
sc
o
n
c
e
l
o
s
,
S
.
A
v
i
l
a
,
a
n
d
E.
V
a
l
l
e
,
“
D
a
t
a
a
u
g
m
e
n
t
a
t
i
o
n
f
o
r
sk
i
n
l
e
si
o
n
a
n
a
l
y
s
i
s,
”
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
re
N
o
t
e
s
i
n
Ar
t
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
re
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
0
4
1
LN
C
S
,
p
p
.
3
0
3
–
3
1
1
,
2
0
1
8
,
d
o
i
:
1
0
.
1
0
0
7
/
9
7
8
-
3
-
0
3
0
-
0
1
2
0
1
-
4
_
3
3
.
[
2
3
]
S
.
A
g
g
a
r
w
a
l
,
A
.
K
.
S
a
h
o
o
,
C
.
B
a
n
s
a
l
,
a
n
d
P
.
K
.
S
a
r
a
n
g
i
,
“
I
mag
e
c
l
a
ss
i
f
i
c
a
t
i
o
n
u
s
i
n
g
d
e
e
p
l
e
a
r
n
i
n
g
:
a
c
o
m
p
a
r
a
t
i
v
e
st
u
d
y
o
f
V
G
G
-
1
6
,
I
n
c
e
p
t
i
o
n
V
3
a
n
d
Ef
f
i
c
i
e
n
t
N
e
t
B
7
m
o
d
e
l
s,
”
2
0
2
3
3
r
d
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
Ad
v
a
n
c
e
C
o
m
p
u
t
i
n
g
a
n
d
I
n
n
o
v
a
t
i
v
e
T
e
c
h
n
o
l
o
g
i
e
s
i
n
En
g
i
n
e
e
ri
n
g
,
I
C
AC
I
T
E
2
0
2
3
,
p
p
.
1
7
2
8
–
1
7
3
2
,
2
0
2
3
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
A
C
I
TE5
7
4
1
0
.
2
0
2
3
.
1
0
1
8
3
2
5
5
.
[
2
4
]
K
.
S
i
m
o
n
y
a
n
a
n
d
A
.
Zi
ss
e
r
ma
n
,
“
V
e
r
y
d
e
e
p
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
t
w
o
r
k
s
f
o
r
l
a
r
g
e
-
s
c
a
l
e
i
ma
g
e
r
e
c
o
g
n
i
t
i
o
n
,
”
3
rd
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
L
e
a
rn
i
n
g
Re
p
res
e
n
t
a
t
i
o
n
s
,
I
C
L
R
2
0
1
5
-
C
o
n
f
e
re
n
c
e
T
r
a
c
k
P
ro
c
e
e
d
i
n
g
s
,
2
0
1
5
.
[
2
5
]
D
.
R
.
R
a
ma
n
,
S
.
N
i
s
h
a
n
t
h
i
,
a
n
d
P
.
B
a
b
y
s
h
a
,
“
D
i
a
g
n
o
si
s
o
f
d
i
a
b
e
t
i
c
r
e
t
i
n
o
p
a
t
h
y
b
y
u
si
n
g
Ef
f
i
c
i
e
n
t
N
e
t
-
B
7
C
N
N
a
r
c
h
i
t
e
c
t
u
r
e
i
n
d
e
e
p
l
e
a
r
n
i
n
g
,
”
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
S
u
s
t
a
i
n
a
b
l
e
C
o
m
p
u
t
i
n
g
a
n
d
S
m
a
r
t
S
y
s
t
e
m
s,
I
C
S
C
S
S
2
0
2
3
-
Pro
c
e
e
d
i
n
g
s
,
p
p
.
4
3
0
–
4
3
5
,
2
0
2
3
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
S
C
S
S
5
7
6
5
0
.
2
0
2
3
.
1
0
1
6
9
4
5
3
.
[
2
6
]
S
.
A
l
Ta
k
r
o
u
r
i
,
N
.
M
.
N
o
o
r
,
N
.
A
h
ma
d
,
T
.
J
u
st
i
n
i
a
,
a
n
d
S
.
U
sma
n
,
“
I
mag
e
su
p
e
r
-
r
e
so
l
u
t
i
o
n
u
si
n
g
g
e
n
e
r
a
t
i
v
e
a
d
v
e
r
sari
a
l
n
e
t
w
o
r
k
s
w
i
t
h
Ef
f
i
c
i
e
n
t
N
e
t
V
2
,
”
I
n
t
e
r
n
a
t
i
o
n
a
l
J
o
u
r
n
a
l
o
f
A
d
v
a
n
c
e
d
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
a
n
d
A
p
p
l
i
c
a
t
i
o
n
s
,
v
o
l
.
1
4
,
n
o
.
2
,
p
p
.
8
7
9
–
8
8
7
,
2
0
2
3
,
d
o
i
:
1
0
.
1
4
5
6
9
/
I
JA
C
S
A
.
2
0
2
3
.
0
1
4
0
2
1
0
0
.
[
2
7
]
J.
F
e
n
g
,
W
.
E.
O
n
g
,
W
.
C
.
T
e
h
,
a
n
d
R
.
Zh
a
n
g
,
“
En
h
a
n
c
e
d
c
r
o
p
d
i
sea
se
d
e
t
e
c
t
i
o
n
w
i
t
h
Ef
f
i
c
i
e
n
t
N
e
t
c
o
n
v
o
l
u
t
i
o
n
a
l
g
r
o
u
p
-
w
i
se
t
r
a
n
sf
o
r
mer,
”
I
EE
E
A
c
c
e
ss
,
v
o
l
.
1
2
,
p
p
.
4
4
1
4
7
–
4
4
1
6
2
,
2
0
2
4
,
d
o
i
:
1
0
.
1
1
0
9
/
A
C
C
ESS
.
2
0
2
4
.
3
3
7
9
3
0
3
.
[
2
8
]
R
.
A
n
u
g
r
a
h
,
K
.
U
sm
a
n
,
a
n
d
L.
N
o
v
a
m
i
z
a
n
t
i
,
“
C
l
a
ss
i
f
i
c
a
t
i
o
n
o
f
M
a
l
a
r
i
a
i
n
r
e
d
b
l
o
o
d
c
e
l
l
mi
c
r
o
s
c
o
p
i
c
i
m
a
g
e
s
u
s
i
n
g
d
e
e
p
l
e
a
r
n
i
n
g
w
i
t
h
Ef
f
i
c
i
e
n
t
N
e
t
a
r
c
h
i
t
e
c
t
u
r
e
a
n
d
S
V
M
,
”
8
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
Re
c
e
n
t
A
d
v
a
n
c
e
s
a
n
d
I
n
n
o
v
a
t
i
o
n
s
i
n
En
g
i
n
e
e
ri
n
g
:
Em
p
o
w
e
ri
n
g
C
o
m
p
u
t
i
n
g
,
An
a
l
y
t
i
c
s
,
a
n
d
En
g
i
n
e
e
r
i
n
g
T
h
r
o
u
g
h
D
i
g
i
t
a
l
I
n
n
o
v
a
t
i
o
n
,
I
C
R
AI
E
2
0
2
3
,
2
0
2
3
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
R
A
I
E5
9
4
5
9
.
2
0
2
3
.
1
0
4
6
8
3
0
0
.
[
2
9
]
S
.
Lo
u
ssa
i
e
f
a
n
d
A
.
A
b
d
e
l
k
r
i
m
,
“
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
h
y
p
e
r
-
p
a
r
a
me
t
e
r
s
o
p
t
i
m
i
z
a
t
i
o
n
b
a
s
e
d
o
n
g
e
n
e
t
i
c
a
l
g
o
r
i
t
h
ms,
”
I
n
t
e
r
n
a
t
i
o
n
a
l
J
o
u
r
n
a
l
o
f
A
d
v
a
n
c
e
d
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
a
n
d
A
p
p
l
i
c
a
t
i
o
n
s
,
v
o
l
.
9
,
n
o
.
1
0
,
p
p
.
2
5
2
–
2
6
6
,
2
0
1
8
,
d
o
i
:
1
0
.
1
4
5
6
9
/
I
JA
C
S
A
.
2
0
1
8
.
0
9
1
0
3
1
.
[
3
0
]
F
.
P
r
i
a
n
e
s,
K
.
M
.
F
o
r
t
u
n
o
,
R
.
O
n
e
s
a
,
B
.
B
e
n
o
sa
,
T
.
P
a
l
a
o
a
g
,
a
n
d
N
.
F
l
o
r
e
s,
“
E
x
p
l
o
r
i
n
g
t
h
e
l
a
n
d
sca
p
e
:
a
n
a
l
y
s
i
s o
f
mo
d
e
l
r
e
s
u
l
t
s o
n
v
a
r
i
o
u
s 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
r
c
h
i
t
e
c
t
u
r
e
s
f
o
r
i
R
ESP
O
N
D
sy
s
t
e
m
,
”
I
n
t
e
r
n
a
t
i
o
n
a
l
J
o
u
r
n
a
l
o
f
A
d
v
a
n
c
e
d
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
a
n
d
Ap
p
l
i
c
a
t
i
o
n
s
,
v
o
l
.
1
5
,
n
o
.
3
,
p
p
.
5
0
7
–
5
1
8
,
2
0
2
4
,
d
o
i
:
1
0
.
1
4
5
6
9
/
I
JA
C
S
A
.
2
0
2
4
.
0
1
5
0
3
5
2
.
B
I
O
G
RAP
H
I
E
S O
F
AUTH
O
RS
Christin
e
La
u
r
e
Ma
n
a
n
g
a
is
c
u
rre
n
tl
y
tea
c
h
in
g
C
o
m
p
u
t
e
r
S
c
ien
c
e
in
t
h
e
Co
m
p
u
ter
E
n
g
i
n
e
e
rin
g
De
p
a
rtm
e
n
t
o
f
Ad
v
a
n
c
e
d
Tea
c
h
e
rs
Train
in
g
Co
ll
e
g
e
f
o
r
Tec
h
n
ica
l
Ed
u
c
a
ti
o
n
,
U
n
iv
e
rsit
y
o
f
Do
u
a
la,
Ca
m
e
ro
o
n
.
S
h
e
is
c
u
rre
n
tl
y
in
th
e
3
rd
y
e
a
r
o
f
h
e
r
P
h
.
D
.
a
n
d
is
a
m
e
m
b
e
r
o
f
t
h
e
c
o
m
p
u
ter
e
n
g
in
e
e
rin
g
a
n
d
a
u
to
m
a
ti
c
c
o
n
tro
l
lab
o
ra
to
ry
.
He
r
re
se
a
rc
h
in
tere
sts
in
c
lu
d
e
a
rti
ficia
l
in
telli
g
e
n
c
e
,
ima
g
e
p
ro
c
e
ss
in
g
,
n
e
u
ra
l
n
e
two
r
k
s,
c
y
b
e
rc
rime
,
p
ro
g
ra
m
m
in
g
,
a
n
d
we
b
d
e
sig
n
.
S
h
e
c
a
n
b
e
c
o
n
ta
c
ted
a
t
e
m
a
il
:
c
m
a
n
a
n
g
a
3
5
@
g
m
a
il
.
c
o
m
.
Pro
f.
Dr
.
Feli
x
P
a
u
n
e
is
h
o
l
d
e
r
o
f
a
P
h
.
D
.
in
a
u
t
o
m
a
ti
o
n
fr
o
m
th
e
U
n
iv
e
rsit
y
o
f
Do
u
a
la
.
He
is
c
u
rre
n
tl
y
a
se
n
io
r
lec
tu
re
r
a
n
d
tea
c
h
e
s
in
th
e
De
p
a
rtme
n
t
o
f
Co
m
p
u
te
r
En
g
i
n
e
e
rin
g
a
t
t
h
e
Un
i
v
e
rsity
o
f
Do
u
a
la.
His
re
se
a
rc
h
to
p
ics
in
c
lu
d
e
p
ro
c
e
ss
c
o
n
tr
o
l,
c
o
m
p
u
te
r
sy
ste
m
s
a
n
d
n
e
two
rk
s,
a
n
d
a
rti
fi
c
ial
in
telli
g
e
n
c
e
.
He
is
th
e
a
u
th
o
r
o
f
se
v
e
ra
l
sc
ien
ti
fic
a
rti
c
les
p
u
b
li
sh
e
d
i
n
i
n
tern
a
ti
o
n
a
l
jo
u
r
n
a
ls
a
n
d
h
a
s
su
p
e
rv
ise
d
n
u
m
e
ro
u
s
m
a
ste
r
’
s
th
e
se
s
a
n
d
d
o
c
t
o
ra
l
d
isse
rtatio
n
s.
He
h
a
s
a
lso
p
a
rt
i
c
ip
a
ted
i
n
se
v
e
ra
l
n
a
ti
o
n
a
l
a
n
d
in
tern
a
ti
o
n
a
l
se
m
in
a
rs
a
n
d
c
o
n
fe
re
n
c
e
s.
He
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
fe
li
x
.
p
a
u
n
e
@la
p
o
ste
.
n
e
t
.
Pro
f.
Dr
.
Lé
a
n
d
r
e
Nnem
e
Nne
m
e
is
h
o
l
d
e
r
o
f
a
P
h
.
D
.
i
n
a
u
to
m
a
ti
o
n
fr
o
m
th
e
Un
iv
e
rsity
o
f
M
o
n
trea
l
.
He
is
c
u
r
re
n
tl
y
a
fu
ll
p
r
o
fe
ss
o
r
a
n
d
d
irec
to
r
o
f
t
h
e
Ad
v
a
n
c
e
d
Tea
c
h
e
rs
Train
in
g
Co
ll
e
g
e
f
o
r
Tec
h
n
ica
l
Ed
u
c
a
ti
o
n
,
U
n
iv
e
rsit
y
o
f
E
b
o
l
o
wa
,
Ca
m
e
ro
o
n
.
His
re
se
a
rc
h
fo
c
u
s
e
s
o
n
p
r
o
c
e
ss
c
o
n
tro
l,
r
o
b
o
ti
c
s,
a
n
d
a
rti
ficia
l
in
telli
g
e
n
c
e
.
He
i
s
th
e
a
u
t
h
o
r
o
f
m
o
re
th
a
n
3
5
sc
ien
ti
fic
a
rti
c
les
a
n
d
h
o
ld
s
a
Ca
n
a
d
ian
in
v
e
n
ti
o
n
p
a
ten
t
o
n
n
e
u
ra
l
e
stim
a
to
rs.
His
a
c
a
d
e
m
ic
a
c
h
iev
e
m
e
n
ts
in
c
lu
d
e
s
u
p
e
rv
isin
g
o
v
e
r
3
0
0
m
a
ste
r
’
s
th
e
se
s,
se
v
e
ra
l
c
o
m
p
lete
d
a
n
d
o
n
g
o
i
n
g
d
o
c
to
ra
l
d
isse
rtatio
n
s.
He
is
a
ls
o
re
c
o
g
n
ize
d
fo
r
h
is
p
a
rti
c
ip
a
ti
o
n
i
n
n
u
m
e
ro
u
s
n
a
ti
o
n
a
l
a
n
d
in
tern
a
ti
o
n
a
l
se
m
in
a
rs
a
n
d
c
o
n
fe
re
n
c
e
s.
M
o
re
o
v
e
r,
h
e
h
a
s
b
e
e
n
a
wa
rd
e
d
se
v
e
ra
l
d
ist
in
c
ti
o
n
s,
in
c
lu
d
in
g
b
e
st
h
e
a
d
o
f
De
p
a
rtme
n
t
in
2
0
0
8
a
n
d
2
0
0
9
,
b
e
st
lec
t
u
re
r
a
t
ENS
ET
in
2
0
1
0
,
th
e
Afric
a
n
P
re
ss
Aw
a
rd
G
o
ld
,
a
n
d
th
e
G
ra
n
d
P
rize
fo
r
Ex
c
e
ll
e
n
c
e
i
n
Ac
a
d
e
m
ic
M
a
n
a
g
e
m
e
n
t
in
th
e
CEM
AC reg
io
n
in
2
0
1
9
.
He
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
lea
n
d
re
n
@g
m
a
il
.
c
o
m
.
Evaluation Warning : The document was created with Spire.PDF for Python.