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
6
,
No
.
1
,
Octo
b
er
20
24
,
p
p
.
5
0
9
~
5
1
6
I
SS
N:
2
502
-
4
7
52
,
DOI
: 1
0
.
1
1
5
9
1
/ijee
cs
.v
3
6
.
i
1
.
pp
509
-
5
1
6
509
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
ee
cs
.
ia
esco
r
e.
co
m
Ey
e
disea
se dete
ct
io
n using
t
ra
nsfe
r
learning
bas
ed o
n ret
ina
l
fundus ima
g
e dat
a
H
elm
i Im
a
du
dd
in
,
Aliv
ia
Ra
hm
a
Sa
k
ina
D
e
p
a
r
t
me
n
t
o
f
I
n
f
o
r
m
a
t
i
c
s,
U
n
i
v
e
r
si
t
a
s M
u
h
a
mm
a
d
i
y
a
h
S
u
r
a
k
a
r
t
a
,
S
u
r
a
k
a
r
t
a
,
I
n
d
o
n
e
si
a
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Dec
2
,
2
0
2
3
R
ev
is
ed
J
un
6
,
2
0
2
4
Acc
ep
ted
J
u
n
25
,
2
0
2
4
Th
e
e
sc
a
latin
g
g
l
o
b
a
l
p
re
v
a
len
c
e
o
f
b
li
n
d
n
e
ss
re
m
a
in
s
a
p
re
ss
in
g
c
o
n
c
e
rn
,
with
e
y
e
d
ise
a
se
s
re
p
re
se
n
ti
n
g
t
h
e
p
rima
ry
c
u
l
p
rit
s
b
e
h
in
d
t
h
is
iss
u
e
.
Visio
n
is
in
teg
ra
l
to
v
a
ri
o
u
s
a
sp
e
c
ts
o
f
h
u
m
a
n
li
fe
,
u
n
d
e
rsc
o
rin
g
t
h
e
si
g
n
i
fica
n
c
e
o
f
e
ffe
c
ti
v
e
e
y
e
d
ise
a
se
d
e
t
e
c
ti
o
n
.
P
re
se
n
tl
y
,
d
ise
a
se
d
e
tec
ti
o
n
re
li
e
s
l
a
rg
e
ly
o
n
m
a
n
u
a
l
m
e
th
o
d
s,
wh
ich
a
re
su
sc
e
p
ti
b
le
to
m
isd
iag
n
o
sis.
H
o
we
v
e
r,
th
e
a
d
v
e
n
t
o
f
tec
h
n
o
l
o
g
y
h
a
s
p
a
v
e
d
th
e
wa
y
fo
r
d
ise
a
se
d
e
tec
ti
o
n
th
ro
u
g
h
th
e
a
p
p
li
c
a
ti
o
n
o
f
d
e
e
p
lea
rn
in
g
m
e
th
o
d
o
lo
g
ies
.
De
e
p
lea
rn
in
g
e
x
h
ib
it
s
su
b
sta
n
ti
a
l
p
o
ten
ti
a
l
in
d
ise
a
se
d
e
tec
ti
o
n
,
p
a
rti
c
u
larl
y
w
h
e
n
a
p
p
li
e
d
to
ima
g
e
d
a
ta,
a
s
a
tt
e
ste
d
b
y
it
s
a
c
c
u
ra
c
y
in
a
lg
o
r
it
h
m
ic
a
ss
e
ss
m
e
n
ts.
Th
is
re
se
a
r
c
h
in
tro
d
u
c
e
s
a
n
o
v
e
l
a
p
p
r
o
a
c
h
t
o
d
ise
a
se
d
e
tec
ti
o
n
,
s
p
e
c
ifi
c
a
ll
y
tra
n
sfe
r
lea
rn
in
g
-
b
a
se
d
d
e
e
p
le
a
rn
in
g
.
T
h
e
stu
d
y
se
e
k
s
to
e
v
a
lu
a
te
a
n
d
c
o
m
p
a
re
th
e
p
e
rfo
rm
a
n
c
e
o
f
v
a
rio
u
s
m
o
d
e
ls,
in
c
l
u
d
i
n
g
Eff
icie
n
tNe
tB3
,
De
n
s
e
Ne
t
-
1
2
1
,
VGG
-
1
6
,
a
n
d
Re
sN
e
t
-
1
5
2
,
i
n
id
e
n
ti
f
y
in
g
th
re
e
p
re
v
a
len
t
e
y
e
d
ise
a
se
s:
c
a
tara
c
t,
d
iab
e
ti
c
re
ti
n
o
p
a
t
h
y
,
a
n
d
g
lau
c
o
m
a
,
u
ti
li
z
i
n
g
re
ti
n
a
l
fu
n
d
u
s
ima
g
e
d
a
ta.
E
x
ten
siv
e
e
x
p
e
rime
n
tati
o
n
re
v
e
a
ls
t
h
a
t
th
e
De
n
se
Ne
t
-
1
2
1
m
o
d
e
l
a
c
h
iev
e
s
th
e
h
ig
h
e
st
a
c
c
u
ra
c
y
l
e
v
e
ls,
b
o
a
stin
g
p
re
c
isio
n
,
re
c
a
ll
,
F
1
-
sc
o
re
,
a
n
d
a
c
c
u
ra
c
y
v
a
l
u
e
s
o
f
9
6
.
5
%
,
9
6
%
,
9
6
.
2
5
%
,
a
n
d
9
6
.
2
0
%
,
re
sp
e
c
ti
v
e
ly
.
Th
e
se
re
su
lt
s
d
e
m
o
n
stra
te
th
e
su
p
e
rio
r
p
e
rfo
rm
a
n
c
e
o
f
t
h
e
e
m
p
lo
y
e
d
tran
sfe
r
lea
rn
in
g
m
o
d
e
l,
sig
n
ify
i
n
g
it
s e
ffi
c
a
c
y
in
d
e
tec
ti
n
g
e
y
e
d
ise
a
se
s.
K
ey
w
o
r
d
s
:
Dee
p
lear
n
in
g
Den
s
eNe
t
-
121
E
y
e
d
is
ea
s
e
R
esNet
-
152
R
etin
al
im
ag
e
T
r
an
s
f
er
lear
n
i
n
g
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
:
Helm
i I
m
ad
u
d
d
in
Dep
ar
tm
en
t o
f
I
n
f
o
r
m
atics,
U
n
iv
er
s
itas
Mu
h
am
m
ad
iy
a
h
Su
r
ak
ar
ta
Su
r
ak
ar
ta,
I
n
d
o
n
esia
E
m
ail: h
elm
i.im
ad
u
d
d
in
@
u
m
s
.
ac
.
id
1.
I
NT
RO
D
UCT
I
O
N
Vis
io
n
p
lay
s
a
p
iv
o
tal
r
o
le
th
r
o
u
g
h
o
u
t
o
n
e'
s
life
s
p
an
.
Vis
u
al
im
p
air
m
en
t
o
r
ey
e
d
is
ea
s
es
ca
n
m
an
if
est
ac
r
o
s
s
v
a
r
io
u
s
a
g
e
g
r
o
u
p
s
,
s
ig
n
i
f
ican
tly
im
p
ac
tin
g
t
h
e
q
u
ality
o
f
life
.
No
ta
b
ly
,
ey
e
d
is
ea
s
es
s
tan
d
as
a
p
r
im
ar
y
ca
u
s
e
o
f
b
lin
d
n
es
s
,
f
u
r
th
er
ex
ac
er
b
atin
g
th
e
c
h
allen
g
es
f
ac
ed
b
y
i
n
d
iv
id
u
a
ls
[
1
]
.
T
h
e
g
lo
b
al
p
r
ev
alen
ce
o
f
v
is
u
al
im
p
air
m
en
t
h
as
s
ee
n
a
co
n
ce
r
n
in
g
r
is
e,
with
d
ata
f
r
o
m
th
e
W
o
r
ld
Hea
lth
Or
g
an
izatio
n
(
W
HO)
in
2
0
1
9
r
ev
ea
lin
g
th
at
o
v
er
2
.
2
b
illi
o
n
p
eo
p
le
wo
r
l
d
wid
e
s
u
f
f
er
f
r
o
m
v
is
u
al
im
p
ai
r
m
en
t
o
r
b
li
n
d
n
ess
.
Alar
m
in
g
ly
,
I
n
d
o
n
esia
r
an
k
s
th
ir
d
g
lo
b
ally
in
ter
m
s
o
f
b
lin
d
n
ess
p
r
ev
alen
ce
,
with
a
r
ate
o
f
1
.
4
7
%,
an
d
with
in
So
u
th
ea
s
t A
s
ia,
it p
o
s
s
es
s
es th
e
h
ig
h
est r
ate
at
3
% [
2
]
.
An
aly
zin
g
d
ata
s
p
an
n
in
g
f
r
o
m
1
9
9
0
to
2
0
1
5
,
ca
tar
ac
ts
em
er
g
e
as
th
e
lead
in
g
ca
u
s
e
o
f
b
lin
d
n
ess
,
ac
co
u
n
tin
g
f
o
r
3
6
.
5
8
%
o
f
ca
s
es.
Fo
llo
win
g
clo
s
ely
is
u
n
d
er
co
r
r
ec
ted
r
ef
r
ac
tiv
e
er
r
o
r
at
3
6
.
4
3
%,
wh
ile
o
th
er
co
n
d
itio
n
s
s
u
ch
as
g
la
u
co
m
a
(
5
.
8
1
%),
ag
e
-
r
elate
d
m
ac
u
lar
d
eg
en
er
atio
n
(
2
.
4
4
%),
c
o
r
n
ea
l
d
is
ea
s
e
(
2
.
4
3
%)
,
d
iab
etic
r
etin
o
p
ath
y
(
0
.
1
6
%),
an
d
tr
ac
h
o
m
a
(
0
.
0
4
%)
co
n
s
titu
te
th
e
r
em
ain
in
g
ca
u
s
es
o
f
b
lin
d
n
ess
[
3
]
.
T
h
ese
s
tatis
t
ics
u
n
d
er
s
co
r
e
ca
tar
ac
ts
as
th
e
p
r
e
d
o
m
in
a
n
t
co
n
tr
ib
u
t
o
r
to
b
lin
d
n
ess
.
C
atar
ac
ts
in
v
o
lv
e
a
d
eg
en
er
ati
v
e
p
r
o
ce
s
s
ch
ar
ac
ter
ized
b
y
o
p
ac
ity
in
th
e
e
y
e'
s
len
s
f
i
b
er
s
.
Ad
d
itio
n
a
lly
,
co
n
d
itio
n
s
lik
e
g
lau
co
m
a
,
ch
ar
ac
ter
ized
b
y
o
p
tic
d
is
c
cu
p
p
in
g
a
n
d
r
esu
ltin
g
f
r
o
m
elev
ated
in
tr
ao
cu
lar
p
r
ess
u
r
e,
p
lay
a
s
ig
n
if
ican
t
r
o
le
i
n
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
6
,
No
.
1
,
Octo
b
er
20
24
:
509
-
5
1
6
510
v
is
io
n
lo
s
s
.
Diab
etic
r
etin
o
p
ath
y
,
m
ar
k
e
d
b
y
d
am
ag
e
to
th
e
ey
e'
s
b
lo
o
d
v
ess
els,
i
s
an
o
th
er
p
r
o
m
in
en
t
ca
u
s
e
o
f
bl
in
d
n
ess
[
4
]
.
T
h
ese
h
ig
h
in
c
id
en
ce
r
ates
h
i
g
h
lig
h
t
a
co
n
c
er
n
in
g
is
s
u
e:
m
a
n
y
in
d
iv
id
u
a
ls
af
f
ec
ted
b
y
ey
e
d
is
ea
s
es
r
em
ain
u
n
awa
r
e
o
f
th
eir
co
n
d
itio
n
,
lea
d
in
g
to
d
elay
ed
d
iag
n
o
s
is
an
d
tr
ea
tm
en
t,
o
f
ten
wo
r
s
en
in
g
th
e
p
r
o
g
n
o
s
is
.
C
o
n
v
er
s
ely
,
th
e
p
r
o
ce
s
s
o
f
d
iag
n
o
s
in
g
ey
e
d
is
o
r
d
e
r
s
ca
n
b
e
tim
e
-
co
n
s
u
m
in
g
,
p
ar
ticu
la
r
ly
wh
en
r
ely
in
g
o
n
m
an
u
al
o
b
s
er
v
ati
o
n
s
,
wh
ich
m
ay
lead
to
m
is
d
iag
n
o
s
is
.
Ho
wev
er
,
with
t
h
e
ad
v
an
ce
m
e
n
t
o
f
tech
n
o
lo
g
y
,
d
is
ea
s
e
id
en
tific
atio
n
ca
n
b
e
f
ac
ilit
ated
th
r
o
u
g
h
t
h
e
u
s
e
o
f
tech
n
o
lo
g
ical
m
ea
n
s
,
with
d
ee
p
lear
n
in
g
em
e
r
g
in
g
as
a
p
r
o
m
i
n
en
t
ap
p
r
o
ac
h
.
Dee
p
lear
n
in
g
,
a
s
u
b
s
et
o
f
m
ac
h
in
e
lear
n
in
g
,
is
ch
ar
ac
ter
ized
b
y
its
f
o
cu
s
o
n
a
r
tific
ial
in
tellig
e
n
ce
alg
o
r
ith
m
s
in
s
p
ir
e
d
b
y
th
e
n
eu
r
al
s
tr
u
ctu
r
e
an
d
f
u
n
ctio
n
in
g
o
f
th
e
b
r
ain
,
k
n
o
wn
as
ar
tific
ial
n
eu
r
al
n
et
wo
r
k
s
[
5
]
.
W
ith
in
th
e
r
ea
lm
o
f
d
ee
p
lea
r
n
in
g
,
o
n
e
m
eth
o
d
th
at
s
tan
d
s
o
u
t
is
tr
an
s
f
er
lear
n
in
g
.
T
h
is
ap
p
r
o
ac
h
h
ar
n
ess
es
th
e
p
o
ten
tial
to
ac
cu
r
ately
id
en
tify
d
is
ea
s
es,
th
u
s
o
f
f
er
in
g
s
ig
n
if
ican
t
ass
is
tan
ce
in
th
e
f
ield
o
f
m
ed
icin
e
.
T
r
a
n
s
f
er
lear
n
in
g
s
er
v
es
as
a
tech
n
iq
u
e
to
ex
p
ed
ite
th
e
tr
ain
in
g
p
r
o
ce
s
s
with
in
co
n
v
o
l
u
tio
n
al
n
eu
r
al
n
etwo
r
k
s
(
C
NNs),
a
k
ey
co
m
p
o
n
e
n
t
o
f
d
ee
p
lea
r
n
in
g
,
d
esig
n
ed
to
ad
d
r
ess
lim
itatio
n
s
in
h
er
en
t in
p
r
ev
io
u
s
m
eth
o
d
s
[
6
]
.
I
n
a
s
tu
d
y
c
o
n
d
u
cted
b
y
Sar
k
i
et
a
l.
[
7
]
,
th
e
ap
p
licatio
n
o
f
t
r
an
s
f
er
lear
n
in
g
was
em
p
lo
y
ed
to
d
iag
n
o
s
e
d
iab
etic
ey
e
d
is
ea
s
es
b
y
an
aly
zin
g
r
etin
al
f
u
n
d
u
s
im
a
g
es
u
s
in
g
C
NN
ar
ch
itectu
r
es,
s
p
ec
if
ically
VGG
-
1
6
.
T
h
e
f
in
d
in
g
s
i
n
d
ica
ted
th
at
th
e
r
e
was
a
n
ac
cu
r
ac
y
r
ate
o
f
8
3
.
4
3
%.
I
n
ad
d
itio
n
,
Pin
co
n
d
u
cte
d
a
s
tu
d
y
wh
er
ein
th
e
R
esNet5
0
m
o
d
el
was
em
p
lo
y
ed
f
o
r
th
e
p
u
r
p
o
s
e
o
f
d
etec
tin
g
ey
e
p
r
o
b
lem
s
[
8
]
.
T
h
e
r
esu
lts
in
d
ica
ted
an
ac
cu
r
ac
y
r
ate
o
f
8
5
.
7
9
%
wh
e
n
ap
p
lied
to
a
d
ataset
co
n
s
is
tin
g
o
f
1
,
3
0
4
f
u
n
d
u
s
im
a
g
es.
I
n
a
s
ep
ar
ate
s
tu
d
y
,
Su
g
en
o
d
id
r
esear
ch
o
n
th
e
ap
p
licatio
n
o
f
E
f
f
icien
tNetB
3
f
o
r
th
e
id
en
tific
atio
n
o
f
ey
e
illn
ess
es
[
9
]
.
T
h
e
f
in
d
in
g
s
o
f
t
h
is
in
q
u
ir
y
r
e
v
ea
led
an
ac
cu
r
a
cy
r
ate
o
f
8
4
.
4
2
%.
Ad
d
itio
n
all
y
,
T
aşar
em
p
lo
y
ed
th
e
T
r
an
s
f
er
L
ea
r
n
in
g
m
eth
o
d
o
lo
g
y
to
d
is
ce
r
n
v
ar
io
u
s
m
ed
i
ca
l
co
n
d
itio
n
s
,
s
u
ch
as
s
k
in
ca
n
ce
r
,
th
r
o
u
g
h
th
e
ex
am
in
atio
n
o
f
d
e
r
m
o
s
co
p
y
im
ag
es
[
1
0
]
.
I
n
t
h
is
p
ar
ticu
lar
ca
s
e,
th
e
Den
s
eNe
t
-
1
2
1
ar
ch
itectu
r
e
wa
s
em
p
lo
y
ed
,
r
esu
ltin
g
in
a
r
em
ar
k
ab
le
ac
cu
r
ac
y
r
ate
o
f
9
4
.
2
9
%
.
B
ased
o
n
th
e
af
o
r
e
m
en
tio
n
e
d
f
in
d
in
g
s
,
th
is
s
tu
d
y
aim
s
to
id
en
tify
s
p
ec
if
ic
ey
e
d
is
o
r
d
er
s
,
n
am
ely
ca
tar
ac
t,
g
lau
co
m
a,
a
n
d
d
ia
b
e
tic
r
etin
o
p
ath
y
.
T
h
is
r
esear
ch
p
r
o
p
o
s
e
s
to
co
n
d
u
ct
a
c
o
m
p
ar
ativ
e
an
aly
s
is
o
f
v
ar
io
u
s
alg
o
r
ith
m
s
u
s
in
g
tr
an
s
f
er
lear
n
in
g
m
eth
o
d
o
lo
g
y
,
s
p
e
cif
ically
f
o
cu
s
in
g
o
n
E
f
f
icien
t
NetB3
,
Den
s
eNe
t
-
1
2
1
,
VGG
-
1
6
,
a
n
d
R
esNet
-
1
5
2
,
u
s
in
g
r
etin
al
f
u
n
d
u
s
im
a
g
e
d
ata.
T
h
e
u
ltima
te
g
o
al
i
s
to
d
eter
m
in
e
th
e
alg
o
r
ith
m
t
h
at
ex
h
ib
its
th
e
h
ig
h
est
lev
e
l
o
f
ac
cu
r
ac
y
in
d
etec
tin
g
v
ar
i
o
u
s
ey
e
d
is
ea
s
es.
T
h
is
wo
r
k
p
r
esen
ts
a
n
o
v
el
tr
an
s
f
er
lear
n
in
g
ap
p
r
o
ac
h
u
s
in
g
d
if
f
er
e
n
t
h
y
p
er
p
ar
a
m
eter
s
.
W
e
h
av
e
co
n
d
u
cted
1
2
ex
p
er
im
e
n
ts
b
y
m
o
d
if
y
in
g
d
i
f
f
er
en
t
lear
n
in
g
r
ate
an
d
e
p
o
ch
v
alu
es
f
o
r
ea
c
h
alg
o
r
ith
m
,
th
en
th
e
r
esu
lts
o
f
th
e
class
if
icatio
n
will
b
e
co
m
p
ar
ed
to
f
in
d
o
u
t
wh
ich
tr
an
s
f
er
lear
n
in
g
m
eth
o
d
h
as
th
e
b
est
p
er
f
o
r
m
an
ce
f
o
r
ey
e
d
is
ea
s
e
im
ag
e
d
ata.
So
th
at
th
e
b
est r
esu
lts
ar
e
o
b
tain
ed
f
o
r
t
h
e
class
if
icatio
n
o
f
e
y
e
d
is
ea
s
e
im
ag
es
.
2.
M
E
T
H
O
D
I
n
t
h
i
s
s
t
u
d
y
,
t
h
e
t
r
a
n
s
f
e
r
l
e
ar
n
i
n
g
t
e
c
h
n
i
q
u
e
w
as
u
t
il
i
z
e
d
to
c
l
a
s
s
i
f
y
i
m
a
g
e
d
a
t
a
p
e
r
t
a
i
n
in
g
t
o
e
y
e
d
i
s
e
as
e
s
.
T
h
i
s
a
p
p
r
o
a
c
h
i
n
v
o
l
v
e
s
u
t
i
li
z
i
n
g
a
p
r
e
-
t
r
a
i
n
e
d
m
o
d
e
l
a
n
d
a
d
j
u
s
ti
n
g
i
ts
p
a
r
a
m
et
e
r
s
t
o
c
a
t
e
r
t
o
t
h
e
s
p
e
c
i
f
i
c
c
h
a
r
a
c
t
e
r
is
ti
c
s
o
f
t
h
e
n
e
w
c
a
s
e
,
w
h
i
c
h
,
i
n
t
h
is
c
o
n
t
ex
t
,
r
e
l
a
t
es
t
o
e
y
e
d
is
e
a
s
e
c
l
as
s
if
i
c
a
t
i
o
n
.
T
h
is
s
t
u
d
y
a
p
p
l
i
e
s
t
h
e
t
r
a
n
s
f
e
r
l
e
a
r
n
i
n
g
p
r
o
c
e
s
s
t
o
c
o
m
p
a
r
e
t
h
e
p
e
r
f
o
r
m
a
n
c
e
o
f
v
a
r
i
o
u
s
m
o
d
e
l
s
w
i
t
h
p
a
r
a
m
e
t
e
r
a
d
j
u
s
t
m
e
n
t
s
a
c
c
o
r
d
i
n
g
t
o
n
e
w
c
as
e
s
i
n
e
y
e
d
i
s
e
as
e
s
.
T
h
i
s
r
es
e
a
r
c
h
w
i
ll
i
d
e
n
t
i
f
y
t
h
r
e
e
e
y
e
d
is
e
as
e
s
n
a
m
el
y
c
a
t
a
r
a
c
t
,
d
i
a
b
e
ti
c
r
e
t
i
n
o
p
a
t
h
y
,
a
n
d
g
l
a
u
c
o
m
a
.
L
e
e
t
a
l
.
[
1
1
]
m
e
n
t
i
o
n
e
d
t
h
a
t
u
s
i
n
g
t
r
a
n
s
f
e
r
l
e
a
r
n
i
n
g
w
il
l
r
e
s
u
l
t
i
n
s
t
r
o
n
g
e
r
c
l
a
s
s
i
f
ic
a
t
i
o
n
.
T
h
is
is
o
n
e
o
f
t
h
e
r
e
f
e
r
e
n
c
e
s
f
o
r
r
e
s
e
a
r
c
h
e
r
s
t
o
u
s
e
t
r
a
n
s
f
e
r
l
e
a
r
n
i
n
g
i
n
e
y
e
d
is
e
a
s
e
cl
a
s
s
i
f
ic
a
t
i
o
n
.
T
h
e
em
p
lo
y
e
d
m
eth
o
d
o
l
o
g
y
ca
n
b
e
d
escr
ib
ed
as
f
o
llo
w
s
:
I
t
b
eg
in
s
with
th
e
g
ath
er
in
g
o
f
th
e
n
ec
ess
ar
y
d
atasets
.
Af
ter
war
d
s
,
d
ata
p
r
e
p
r
o
ce
s
s
in
g
is
ca
r
r
ied
o
u
t
to
p
r
ep
ar
e
th
e
d
ata
f
o
r
i
n
teg
r
atio
n
with
th
e
s
elec
ted
m
o
d
el.
T
h
er
e
ar
e
t
h
r
ee
s
u
b
s
ets
with
in
th
e
d
ataset:
tr
ain
in
g
d
ata,
t
esti
n
g
d
ata,
a
n
d
v
alid
atio
n
d
ata.
I
n
th
e
tr
ain
in
g
p
h
ase,
th
e
d
esig
n
ated
tr
ain
in
g
d
ataset
is
u
s
ed
t
o
tr
ain
th
e
m
o
d
el.
Du
r
i
n
g
ea
c
h
tr
ain
in
g
iter
atio
n
,
th
e
v
alid
atio
n
d
ataset
is
u
s
ed
to
ev
alu
ate
th
e
p
er
f
o
r
m
an
ce
o
f
th
e
m
o
d
el.
Af
ter
th
e
tr
ain
in
g
p
r
o
ce
s
s
is
co
m
p
lete,
th
e
m
o
d
el
is
test
ed
u
s
in
g
th
e
test
in
g
d
ataset
[
1
2
]
.
Fig
u
r
e
1
is
a
v
is
u
al
r
ep
r
esen
tatio
n
o
f
th
e
o
v
er
all
wo
r
k
f
lo
w
f
o
r
th
e
m
eth
o
d
u
s
ed
in
th
is
s
tu
d
y
.
2
.
1
.
Da
t
a
c
o
llect
io
n
T
h
e
d
ataset
u
tili
ze
d
in
th
is
s
tu
d
y
co
m
p
r
is
es
p
u
b
licly
av
ailab
le
r
etin
al
f
u
n
d
u
s
im
ag
es
o
f
e
y
e
d
is
ea
s
es
,
s
o
u
r
ce
d
f
r
o
m
Kag
g
le.
T
h
is
d
ataset
am
alg
am
ates
d
ata
f
r
o
m
d
iv
er
s
e
o
r
ig
in
s
,
in
clu
d
in
g
th
e
I
n
d
ian
Diab
etic
R
etin
o
p
ath
y
I
m
a
g
e
Data
s
et
(
I
DR
iD)
,
o
cu
lu
r
r
ec
o
g
n
itio
n
,
an
d
h
i
g
h
-
r
eso
lu
tio
n
f
u
n
d
u
s
(
HR
F)
d
atasets
.
T
h
e
d
ataset
em
p
lo
y
ed
in
t
h
is
r
esear
ch
en
co
m
p
ass
es a
to
tal
o
f
4
,
2
1
7
im
ag
es,
wh
ich
h
av
e
b
ee
n
c
ateg
o
r
ized
in
to
f
o
u
r
d
is
tin
ct
class
e
s
.
T
h
ese
clas
s
e
s
co
m
p
r
is
e
1
,
0
3
8
im
a
g
es
o
f
ca
tar
ac
t
ca
s
es,
1
,
0
9
8
im
ag
es
d
ep
ictin
g
d
iab
etic
r
etin
o
p
ath
y
,
1
,
0
0
7
im
ag
es
s
h
o
wca
s
in
g
g
lau
co
m
a,
an
d
1
,
0
7
4
im
ag
e
s
r
ep
r
esen
tin
g
n
o
r
m
al
ey
e
co
n
d
itio
n
s
.
A
co
m
p
r
eh
e
n
s
iv
e
b
r
ea
k
d
o
wn
o
f
th
e
d
ataset
d
is
tr
ib
u
tio
n
ac
r
o
s
s
th
ese
class
es i
s
p
r
esen
ted
in
T
ab
le
1
.
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
E
ye
d
is
ea
s
e
d
etec
tio
n
u
s
in
g
tr
a
n
s
fer lea
r
n
in
g
b
a
s
ed
o
n
r
etin
a
l fu
n
d
u
s
ima
g
e
d
a
t
a
(
Helmi
I
ma
d
u
d
d
i
n
)
511
Fig
u
r
e
1
.
R
esear
ch
m
eth
o
d
wo
r
k
f
lo
w
T
ab
le
1
.
Div
is
io
n
o
f
ey
e
d
is
ea
s
e
d
ataset
C
l
a
s
s
D
a
t
a
C
a
t
a
r
a
c
t
s
1
,
0
3
8
D
i
a
b
e
t
i
c
r
e
t
i
n
o
p
a
t
h
y
1
,
0
9
8
G
l
a
u
c
o
ma
1
,
0
0
7
N
o
r
mal
1
,
0
7
4
2
.
2
.
I
ma
g
e
prepro
ce
s
s
ing
T
h
e
im
ag
e
p
r
e
p
r
o
ce
s
s
in
g
s
tag
e
en
co
m
p
ass
es
a
s
er
ies
o
f
tec
h
n
iq
u
es
ap
p
lied
to
th
e
d
ataset
to
en
h
a
n
ce
its
u
s
ab
ilit
y
f
o
r
s
u
b
s
eq
u
e
n
t
p
r
o
ce
s
s
es.
A
cr
itical
asp
ec
t
o
f
th
is
s
tag
e
in
v
o
lv
es
n
o
is
e
r
em
o
v
al
to
en
s
u
r
e
d
at
a
q
u
ality
.
Ad
d
itio
n
ally
,
s
in
ce
th
e
lo
ad
ed
d
ata
ex
h
ib
its
v
ar
iatio
n
s
in
r
eso
lu
tio
n
s
izes,
s
tan
d
ar
d
izatio
n
is
ac
h
iev
ed
b
y
r
esizin
g
all
im
ag
es
to
a
co
n
s
is
ten
t
d
im
en
s
io
n
o
f
2
2
4
×
224
×
3
.
I
n
th
e
r
ea
lm
o
f
d
ee
p
lea
r
n
in
g
,
a
s
u
b
s
tan
tial
v
o
lu
m
e
o
f
d
ata
is
ty
p
ically
ess
en
tial
f
o
r
o
p
tim
al
m
o
d
el
p
er
f
o
r
m
an
ce
.
Ho
wev
e
r
,
im
ag
e
class
if
icatio
n
m
eth
o
d
s
o
f
ten
en
co
u
n
ter
lim
itatio
n
s
d
u
e
to
in
s
u
f
f
icien
t
d
ata
av
ailab
ilit
y
f
o
r
m
o
d
el
tr
ain
in
g
.
C
o
n
s
eq
u
en
tly
,
im
ag
e
au
g
m
en
tatio
n
e
m
er
g
es
as
an
e
f
f
ec
t
iv
e
tech
n
iq
u
e,
p
a
r
ticu
lar
l
y
wh
en
d
ea
lin
g
with
d
atasets
lack
in
g
am
p
le
d
ata
p
o
in
ts
.
T
h
is
m
eth
o
d
p
r
o
v
es
v
alu
ab
le
b
y
au
g
m
en
tin
g
th
e
tr
ain
in
g
d
ataset
with
o
u
t
th
e
n
e
ce
s
s
ity
o
f
ac
q
u
ir
in
g
ad
d
itio
n
al
d
ata,
th
er
e
b
y
cir
cu
m
v
en
tin
g
th
e
n
ee
d
f
o
r
ex
tr
a
s
to
r
ag
e
ca
p
ac
ity
.
I
n
th
e
co
n
tex
t
o
f
th
is
r
esear
ch
,
th
e
Ker
as
lib
r
ar
y
is
h
ar
n
ess
ed
to
lev
er
ag
e
th
e
im
ag
e
d
ata
g
e
n
er
ato
r
f
u
n
ctio
n
,
wh
ich
p
lay
s
a
p
iv
o
tal
r
o
le
in
p
r
ev
en
tin
g
o
v
er
f
itti
n
g
.
T
h
is
f
u
n
ctio
n
en
co
m
p
ass
es
v
ar
io
u
s
g
r
ap
h
ical
p
ar
am
ete
r
s
d
e
s
ig
n
ed
to
g
e
n
er
ate
s
y
n
th
etic
im
ag
e
s
[
1
3
]
.
W
ith
in
th
is
s
tu
d
y
,
s
p
ec
if
ic
p
ar
am
e
ter
s
ar
e
em
p
lo
y
ed
,
in
cl
u
d
in
g
th
e
p
r
ep
r
o
ce
s
s
in
g
f
u
n
ctio
n
=scala
r
an
d
h
o
r
izo
n
tal
f
lip
=tr
u
e.
2
.
3
.
Da
t
a
s
pli
t
t
ing
Du
r
in
g
t
h
e
d
ata
p
r
o
ce
s
s
in
g
s
tag
e,
a
d
ata
s
p
litt
in
g
p
r
o
ce
s
s
,
c
o
m
m
o
n
l
y
r
ef
e
r
r
ed
to
as
d
ata
d
iv
is
io
n
,
is
c
o
n
d
u
cte
d
in
o
r
d
er
to
ac
q
u
ir
e
a
d
ataset
th
at
is
p
r
o
p
o
r
tio
n
al
ly
r
ep
r
esen
tativ
e.
T
h
e
u
tili
za
t
io
n
o
f
th
is
s
tag
e
is
also
em
p
lo
y
ed
to
m
itig
ate
b
ia
s
d
u
r
in
g
th
e
ass
ess
m
en
t
o
f
m
o
d
el
p
er
f
o
r
m
an
ce
.
T
h
e
d
ata
d
i
v
is
io
n
r
atio
u
tili
ze
d
in
th
is
s
tu
d
y
in
v
o
lv
es
allo
ca
ti
n
g
9
0
%
o
f
t
h
e
d
ata
f
o
r
tr
ain
in
g
p
u
r
p
o
s
es,
wh
ile
5
%
is
allo
c
ated
f
o
r
test
in
g
an
d
an
o
th
er
5
%
f
o
r
v
alid
atio
n
.
T
h
e
tr
ain
in
g
d
ata
co
n
s
is
ted
o
f
3
,
7
9
5
in
s
tan
ce
s
,
wh
ile
th
e
test
an
d
v
alid
atio
n
d
atasets
ea
ch
co
n
tain
ed
2
1
1
in
s
tan
ce
s
.
T
h
e
d
ata
will
s
u
b
s
eq
u
en
tly
b
e
p
ar
titi
o
n
ed
r
an
d
o
m
ly
u
tili
zin
g
th
e
r
an
d
o
m
s
tate
f
u
n
ctio
n
s
et
to
1
2
3
.
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
6
,
No
.
1
,
Octo
b
er
20
24
:
509
-
5
1
6
512
2
.
4
.
T
ra
ns
f
er
lea
rning
T
r
an
s
f
er
lear
n
in
g
is
a
wid
el
y
r
ec
o
g
n
ized
m
eth
o
d
th
at
u
tili
ze
s
th
e
s
tr
u
ctu
r
al
f
r
a
m
ewo
r
k
o
f
a
p
r
e
-
ex
is
tin
g
m
o
d
el
to
im
p
r
o
v
e
th
e
p
r
ec
is
io
n
o
f
m
o
d
els,
esp
ec
ially
w
h
en
f
ac
ed
with
d
atasets
th
at
ar
e
n
o
t
lar
g
e
en
o
u
g
h
f
o
r
ex
ten
s
iv
e
t
r
ain
in
g
.
T
r
a
n
s
f
er
lear
n
in
g
lev
er
a
g
es
f
ea
tu
r
e
e
x
tr
ac
tio
n
tech
n
iq
u
es
d
er
iv
e
d
f
r
o
m
p
r
e
-
ex
is
tin
g
m
o
d
els.
T
h
e
tr
an
s
f
er
lear
n
in
g
m
o
d
els
h
av
e
b
ee
n
tr
ain
ed
o
n
ex
ten
s
iv
e
d
atasets
s
u
ch
as
I
m
ag
eNe
t,
wh
ich
co
n
s
is
ts
o
f
1
.
2
m
illi
o
n
im
ag
es
b
elo
n
g
in
g
to
1
,
0
0
0
d
i
f
f
er
en
t
class
es.
T
h
is
tr
ain
in
g
p
r
o
ce
s
s
en
ab
les
t
h
e
m
o
d
els to
ac
q
u
i
r
e
s
ig
n
if
ican
t f
ea
tu
r
es th
at
ca
n
b
e
a
p
p
lied
to
n
ew
tar
g
et
d
ata
[
1
4
]
,
[
1
5
]
.
On
e
o
f
th
e
m
ain
b
en
e
f
its
o
f
u
tili
zin
g
tr
an
s
f
er
lear
n
i
n
g
is
its
ca
p
ac
ity
to
o
p
tim
ize
th
e
d
u
r
atio
n
o
f
tr
ain
in
g
an
d
r
e
d
u
ce
g
e
n
er
aliz
atio
n
er
r
o
r
s
,
as
ev
id
en
ce
d
b
y
p
r
ev
io
u
s
s
tu
d
ies
[
1
6
]
.
W
ith
in
th
e
s
co
p
e
o
f
th
is
r
esear
ch
,
th
e
u
tili
za
tio
n
o
f
tr
an
s
f
er
lear
n
i
n
g
m
eth
o
d
o
lo
g
y
is
o
b
s
er
v
e
d
,
em
p
lo
y
in
g
f
o
u
r
p
r
e
-
tr
ain
ed
C
NN
ar
ch
itectu
r
e
m
o
d
els:
E
f
f
icien
t
NetB3
,
Den
s
eNe
t
-
1
2
1
,
VGG
-
1
6
,
a
n
d
R
esNet
-
1
5
2
.
T
h
ese
m
o
d
els
h
av
e
b
ee
n
s
p
ec
if
ically
cu
s
to
m
ized
f
o
r
t
h
e
p
u
r
p
o
s
e
o
f
class
if
y
in
g
ey
e
d
is
ea
s
es.
T
h
e
a
p
p
licatio
n
o
f
tr
a
n
s
f
er
lear
n
in
g
m
o
d
els
in
a
s
tr
ateg
ic
m
an
n
er
g
r
ea
tly
en
h
a
n
ce
s
th
e
ef
f
icien
c
y
an
d
ef
f
ec
tiv
en
ess
o
f
th
e
class
if
icatio
n
task
b
ein
g
p
er
f
o
r
m
ed
.
2
.
4
.
1
.
E
f
f
icient
Net
B
3
E
f
f
icien
tNetB
3
is
a
C
NN
m
o
d
el
th
at
in
clu
d
es
th
r
ee
cr
itical
e
lem
en
ts
in
its
ar
ch
itectu
r
e:
wid
th
,
d
ep
t
h
,
an
d
r
eso
lu
tio
n
.
T
h
is
co
m
b
in
atio
n
is
s
tr
ateg
ically
d
e
s
ig
n
ed
to
ac
h
iev
e
h
ig
h
er
lev
els
o
f
ac
cu
r
ac
y
wh
ile
s
im
u
ltan
eo
u
s
ly
m
in
im
izin
g
b
o
th
th
e
o
p
tim
al
p
ar
am
eter
s
ize
an
d
th
e
n
u
m
b
er
o
f
f
lo
ati
n
g
-
p
o
in
t
o
p
e
r
atio
n
s
(
FLOPs)
[
1
7
]
.
E
f
f
icien
tNetB
3
'
s
ar
ch
itectu
r
e
in
clu
d
es
tw
o
co
n
v
o
l
u
tio
n
lay
er
s
,
s
ev
en
m
o
b
ile
b
o
ttlen
ec
k
co
n
v
o
l
u
tio
n
lay
e
r
s
,
o
n
e
p
o
o
li
n
g
lay
er
,
an
d
o
n
e
f
u
lly
c
o
n
n
ec
ted
lay
er
.
W
ith
in
ea
ch
co
n
v
o
lu
tio
n
la
y
er
,
th
e
p
r
o
ce
s
s
b
eg
i
n
s
with
a
co
n
v
o
lu
tio
n
al
o
p
e
r
atio
n
with
a
3
×
3
k
e
r
n
el
s
ize
an
d
a
f
ilter
v
alu
e
o
f
2
4
.
Fo
llo
win
g
t
h
at,
th
e
r
ec
tifie
d
lin
ea
r
u
n
it
(
R
eL
U)
ac
tiv
atio
n
f
u
n
ctio
n
is
u
s
ed
,
wh
ich
is
f
o
llo
we
d
b
y
th
e
m
a
x
p
o
o
lin
g
o
p
er
atio
n
,
wh
ich
is
also
f
o
llo
wed
b
y
th
e
R
eL
U
ac
tiv
atio
n
f
u
n
ctio
n
.
T
h
i
s
o
p
er
atio
n
s
eq
u
en
ce
c
o
n
tin
u
e
s
u
n
til
th
e
n
etwo
r
k
r
ea
ch
es
th
e
7
th
co
n
v
o
lu
tio
n
lay
er
.
Fo
llo
win
g
a
s
er
ies
o
f
co
n
v
o
lu
tio
n
s
,
R
eL
U
ac
tiv
atio
n
f
u
n
ctio
n
s
,
an
d
m
a
x
p
o
o
lin
g
,
th
e
n
ex
t
s
tep
in
v
o
lv
e
s
cr
ea
tin
g
a
f
u
lly
co
n
n
ec
ted
la
y
er
an
d
a
p
p
ly
in
g
th
e
So
f
tMa
x
f
u
n
ctio
n
.
T
h
is
f
in
al
s
tep
is
cr
itical
in
class
if
y
in
g
th
e
ex
tr
ac
ted
f
ea
tu
r
es,
y
ield
in
g
an
o
u
tp
u
t
in
d
icatin
g
th
e
im
ag
e
class
u
n
d
e
r
co
n
s
id
er
atio
n
[
1
8
].
2
.
4
.
2
.
DenseNet
-
121
Den
s
eNe
t
r
ep
r
esen
ts
a
C
NN
ar
ch
itectu
r
e
d
is
tin
g
u
is
h
ed
b
y
its
u
n
iq
u
e
a
p
p
r
o
ac
h
o
f
co
n
n
e
ctin
g
ea
ch
lay
er
to
all
s
u
b
s
eq
u
en
t
lay
er
s
i
n
a
f
ee
d
-
f
o
r
war
d
m
a
n
n
er
.
T
h
i
s
co
n
n
ec
tiv
ity
en
s
u
r
es
th
at
ea
c
h
s
u
b
s
eq
u
en
t
lay
er
r
ec
eiv
es
in
p
u
t
f
ea
tu
r
e
-
m
a
p
s
f
r
o
m
all
p
r
ec
ed
in
g
lay
er
s
.
T
h
e
u
tili
za
tio
n
o
f
Den
s
eNe
t
o
f
f
er
s
s
ev
er
al
ad
v
an
tag
es,
in
clu
d
in
g
th
e
m
itig
atio
n
o
f
g
r
ad
ien
t
-
r
elate
d
ch
allen
g
es,
f
o
r
tific
atio
n
o
f
f
ea
tu
r
e
p
r
o
p
ag
ati
o
n
th
r
o
u
g
h
o
u
t
th
e
n
etwo
r
k
,
an
d
r
ed
u
ctio
n
in
th
e
o
v
er
all
n
u
m
b
er
o
f
p
a
r
a
m
eter
s
r
eq
u
ir
ed
f
o
r
ef
f
ec
tiv
e
o
p
er
atio
n
[
1
9
]
.
Fu
r
th
er
m
o
r
e
,
a
p
r
o
m
in
e
n
t
ch
a
r
ac
ter
is
tic
o
f
Den
s
eNe
t
is
th
e
s
eq
u
en
tial
co
n
n
ec
tio
n
d
e
n
s
ity
o
f
lay
er
s
with
in
t
h
e
n
etwo
r
k
.
I
n
th
e
c
o
n
tex
t
o
f
th
is
s
tu
d
y
,
Den
s
eNe
t
-
1
2
1
is
th
e
c
h
o
s
en
v
a
r
ian
t,
c
h
ar
ac
ter
ized
b
y
f
o
u
r
d
en
s
e
b
lo
c
k
s
,
th
r
ee
tr
an
s
itio
n
lay
er
s
,
an
d
a
t
o
tal
o
f
1
2
1
la
y
er
s
.
T
h
is
lay
er
co
m
p
o
s
itio
n
en
co
m
p
ass
es
1
1
7
co
n
v
o
lu
tio
n
lay
er
s
,
th
r
ee
tr
an
s
itio
n
lay
er
s
,
an
d
o
n
e
class
if
icatio
n
lay
er
.
E
ac
h
co
n
v
o
l
u
tio
n
lay
er
is
ass
o
ciate
d
with
a
co
m
p
o
s
ite
o
p
er
atio
n
s
eq
u
en
ce
,
c
o
n
s
is
tin
g
o
f
b
atch
n
o
r
m
aliza
tio
n
(
B
N)
,
R
eL
U
,
an
d
co
n
v
o
lu
tio
n
o
p
er
atio
n
s
.
T
h
e
class
if
icatio
n
s
u
b
n
etwo
r
k
o
f
Den
s
eNe
t
-
1
2
1
en
co
m
p
ass
es
g
lo
b
al
av
er
ag
e
p
o
o
lin
g
at
7
7
%,
f
o
llo
wed
b
y
a
f
u
lly
-
co
n
n
ec
ted
la
y
er
with
1
0
0
0
d
i
m
en
s
io
n
s
,
co
n
clu
d
in
g
with
th
e
So
f
tMa
x
f
u
n
ctio
n
f
o
r
class
if
y
in
g
th
e
ex
tr
ac
ted
f
ea
tu
r
es a
n
d
p
r
o
d
u
cin
g
th
e
d
es
ir
ed
im
ag
e
class
o
u
tp
u
t
[
20
].
2
.
4
.
3
.
VG
G
-
16
VGG
-
1
6
s
tan
d
s
as
a
p
r
o
m
i
n
en
t
C
NN
ar
ch
itectu
r
e
m
o
d
el
th
at
h
as
d
em
o
n
s
tr
ated
r
e
m
ar
k
ab
le
p
er
f
o
r
m
an
ce
o
n
th
e
I
m
ag
eNe
t d
ataset.
No
tab
ly
,
it
s
ec
u
r
ed
th
e
to
p
p
o
s
it
io
n
in
th
e
I
m
a
g
eNe
t
2
0
1
4
v
is
u
al
v
is
io
n
ch
allen
g
e
.
T
h
e
a
r
ch
itectu
r
e
o
f
VGG
-
1
6
is
ch
ar
ac
ter
ized
b
y
a
s
p
ec
if
ic
co
n
v
o
lu
tio
n
al
f
ilter
s
p
ec
if
icatio
n
,
em
p
lo
y
in
g
3
×
3
f
ilter
s
.
VGG
-
1
6
is
d
esig
n
ed
to
p
r
o
ce
s
s
R
GB
im
ag
es
with
d
im
en
s
io
n
s
o
f
2
2
4
×
2
2
4
p
ix
els.
As
an
in
itia
l
s
tep
,
it
n
o
r
m
alize
s
th
e
p
ix
el
v
alu
es,
wh
ich
ty
p
ic
ally
r
an
g
e
f
r
o
m
0
to
2
5
5
,
to
th
e
n
o
r
m
alize
d
s
ca
le
o
f
0
to
1
.
Su
b
s
eq
u
en
tly
,
th
e
im
ag
e
u
n
d
er
g
o
es
a
s
er
ies
o
f
co
n
v
o
l
u
tio
n
al
lay
er
s
an
d
f
u
ll
y
co
n
n
ec
ted
lay
er
s
.
VGG
-
1
6
b
o
asts
a
to
tal
o
f
1
3
c
o
n
v
o
lu
ti
o
n
la
y
er
s
an
d
3
f
u
lly
co
n
n
ec
ted
lay
er
s
with
in
its
s
tr
u
ctu
r
e.
T
o
m
an
ag
e
im
ag
e
s
ize
r
ed
u
ctio
n
a
n
d
au
g
m
en
t
f
ilter
d
ep
th
,
VGG
-
1
6
in
co
r
p
o
r
ates
2
D
Ma
x
Po
o
l
lay
er
s
s
tr
ateg
ically
.
T
h
e
n
u
m
b
er
o
f
f
ilter
s
p
r
o
g
r
ess
iv
ely
escalate
s
with
th
e
d
ep
th
o
f
th
e
m
o
d
el,
co
m
m
en
cin
g
at
6
4
f
ilte
r
s
an
d
p
r
o
g
r
ess
iv
ely
in
cr
ea
s
in
g
t
o
1
2
8
,
2
5
6
,
an
d
5
1
2
f
ilter
s
as
f
ea
tu
r
es
ar
e
ex
tr
ac
ted
f
r
o
m
th
e
in
p
u
t
im
ag
e.
T
h
e
u
ltima
te
o
u
tp
u
t
o
f
th
e
VGG
-
1
6
m
o
d
el
co
n
s
is
ts
o
f
a
f
ea
t
u
r
e
r
ep
r
esen
tatio
n
d
er
iv
e
d
f
r
o
m
th
e
in
p
u
t
im
ag
e,
r
en
d
er
in
g
it we
ll
-
s
u
ited
f
o
r
a
v
ar
iety
o
f
class
if
icatio
n
o
r
d
ete
ctio
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:
2
5
0
2
-
4
7
52
E
ye
d
is
ea
s
e
d
etec
tio
n
u
s
in
g
tr
a
n
s
fer lea
r
n
in
g
b
a
s
ed
o
n
r
etin
a
l fu
n
d
u
s
ima
g
e
d
a
t
a
(
Helmi
I
ma
d
u
d
d
i
n
)
513
2
.
4
.
4
.
Rest
Net
-
1
5
2
R
esNet,
s
h
o
r
t
f
o
r
r
esid
u
al
n
etwo
r
k
s
,
r
ep
r
esen
ts
a
C
NN
ar
ch
itectu
r
e
m
eticu
lo
u
s
ly
en
g
in
ee
r
ed
to
ad
d
r
ess
th
e
n
o
to
r
io
u
s
p
r
o
b
le
m
o
f
g
r
ad
ien
t
lo
s
s
en
co
u
n
te
r
ed
in
d
ee
p
n
eu
r
al
n
etwo
r
k
s
.
T
h
is
in
n
o
v
atio
n
is
a
ch
iev
ed
th
r
o
u
g
h
th
e
im
p
lem
en
tatio
n
o
f
s
k
ip
co
n
n
ec
tio
n
s
b
etwe
en
lay
er
s
,
a
tech
n
iq
u
e
r
ef
er
r
ed
t
o
as
r
esid
u
al
lear
n
in
g
.
Su
ch
ar
c
h
itectu
r
al
en
h
an
ce
m
e
n
t
cu
lm
in
ates
in
a
n
etwo
r
k
th
at
is
co
n
s
id
er
ab
ly
m
o
r
e
tr
ac
tab
le
d
u
r
in
g
th
e
tr
ain
in
g
p
r
o
ce
s
s
.
T
h
is
,
in
t
u
r
n
,
f
ac
i
litates
th
e
d
esig
n
o
f
d
ee
p
er
n
etwo
r
k
s
,
an
d
it
ex
e
r
ts
a
f
av
o
r
ab
le
im
p
ac
t
o
n
o
v
er
all
m
o
d
el
ac
c
u
r
ac
y
.
Am
o
n
g
t
h
e
v
a
r
io
u
s
R
esNet
ar
ch
itectu
r
es
av
ailab
le,
R
esN
et
-
1
5
2
s
tan
d
s
o
u
t,
b
o
asti
n
g
th
e
lo
west
r
ec
o
r
d
ed
t
o
p
-
1
er
r
o
r
an
d
to
p
-
5
er
r
o
r
r
at
es,
ap
p
r
o
x
im
ately
m
ea
s
u
r
i
n
g
2
1
.
4
3
%
an
d
5
.
7
1
%,
r
esp
ec
tiv
ely
[
21
]
.
T
h
e
R
esNet
-
1
5
2
a
r
ch
itectu
r
e
its
elf
u
n
f
o
ld
s
with
s
p
ec
if
ic
s
p
ec
if
icatio
n
s
:
it
co
m
m
en
ce
s
with
a
7
×
7
co
n
v
o
lu
tio
n
o
p
er
atio
n
,
f
ea
tu
r
in
g
a
s
tr
id
e
o
f
2
.
Su
b
s
eq
u
en
tly
,
a
3
×
3
m
a
x
-
p
o
o
lin
g
o
p
er
atio
n
with
a
s
tr
id
e
o
f
2
is
ex
ec
u
t
ed
.
Fu
r
th
er
m
o
r
e,
th
e
m
o
d
el
in
co
r
p
o
r
ates
b
a
tch
n
o
r
m
aliza
tio
n
(
1
×
1
)
,
f
o
llo
wed
b
y
a
R
eL
U
ac
tiv
atio
n
(
1
×
1
)
an
d
a
3
×
3
co
n
v
o
l
u
tio
n
o
p
er
atio
n
with
in
ea
ch
o
f
th
e
c
o
n
v
2
_
x
,
co
n
v
3
_
x
,
co
n
v
4
_
x
,
a
n
d
co
n
v
5
_
x
s
tag
es.
T
h
e
f
i
n
al
s
tep
s
in
v
o
lv
e
av
er
a
g
e
p
o
o
lin
g
an
d
th
e
So
f
tm
ax
f
u
n
ctio
n
t
o
y
ield
class
if
icatio
n
r
esu
lts
.
2
.
5
.
T
ra
ini
ng
m
o
del
T
h
e
p
r
im
ar
y
f
r
am
ewo
r
k
s
f
o
r
d
ev
elo
p
in
g
th
e
m
o
d
el
in
th
is
s
tu
d
y
ar
e
Py
th
o
n
a
n
d
T
e
n
s
o
r
Flo
w.
T
h
e
r
u
n
tim
e
e
n
v
ir
o
n
m
en
t
u
s
ed
is
GPU,
wh
ich
p
r
o
v
id
es
s
ig
n
if
ic
an
t
b
en
e
f
its
in
te
r
m
s
o
f
im
p
r
o
v
ed
p
er
f
o
r
m
an
ce
an
d
r
ed
u
ce
d
tr
ain
in
g
tim
e,
esp
e
cially
wh
en
d
e
alin
g
with
c
o
m
p
lex
n
eu
r
al
n
etwo
r
k
m
o
d
els
[
2
2
]
.
I
n
th
e
ex
p
er
im
en
tal
p
h
ase,
th
e
m
o
d
e
l
is
p
r
etr
ain
ed
b
y
u
s
in
g
th
e
I
m
ag
eNe
t
p
u
b
lic
d
ataset
to
in
itialize
weig
h
t
v
alu
es.
C
o
m
p
r
eh
en
s
iv
e
h
y
p
er
p
ar
a
m
et
er
tu
n
in
g
is
p
er
f
o
r
m
e
d
to
im
p
r
o
v
e
th
e
m
o
d
el'
s
ef
f
icien
cy
.
T
h
is
en
tails
f
in
e
-
tu
n
in
g
a
v
ar
iety
o
f
p
a
r
am
eter
s
s
u
ch
as
lear
n
in
g
r
ate,
m
o
m
en
tu
m
,
ep
o
c
h
co
u
n
t,
b
atc
h
s
ize,
an
d
o
th
er
s
[
2
3
]
.
A
c
u
s
t
o
m
K
e
r
as
c
a
l
l
b
a
c
k
s
u
b
cla
s
s
is
c
r
e
a
t
e
d
t
o
e
n
s
u
r
e
e
f
f
i
ci
en
t
p
e
r
f
o
r
m
a
n
c
e
m
o
n
i
t
o
r
i
n
g
t
h
r
o
u
g
h
o
u
t
t
h
e
t
r
a
i
n
i
n
g
p
r
o
c
e
s
s
.
T
h
is
ca
l
l
b
a
c
k
s
y
s
t
e
m
i
s
c
r
it
i
c
al
i
n
m
o
n
i
t
o
r
i
n
g
t
h
e
m
o
d
e
l
'
s
t
r
ai
n
i
n
g
p
r
o
g
r
e
s
s
a
n
d
f
ac
i
l
i
ta
t
i
n
g
p
a
r
a
m
et
e
r
a
d
j
u
s
t
m
e
n
ts
t
o
i
m
p
r
o
v
e
t
h
e
t
r
a
in
i
n
g
m
o
n
i
t
o
r
i
n
g
p
r
o
c
e
s
s
.
T
a
b
le
2
s
h
o
w
s
t
h
e
h
y
p
e
r
p
a
r
a
m
e
te
r
t
u
n
i
n
g
u
s
e
d
.
T
ab
le
2
.
T
u
n
in
g
h
y
p
er
p
ar
a
m
eter
C
l
a
s
s
D
a
t
a
Lo
ss
f
u
n
c
t
i
o
n
C
a
t
e
g
o
r
i
c
a
l
c
r
o
ss
-
e
n
t
r
o
p
y
O
p
t
i
mi
z
e
r
A
d
a
ma
x
A
c
t
i
v
a
t
i
o
n
S
o
f
t
ma
x
Le
a
r
n
i
n
g
r
a
t
e
[
(
0
.
0
1
)
,
(
0
.
0
0
1
)
,
(
0
.
0
0
0
1
)
]
Ep
o
c
h
[
(
1
0
)
,
(
2
0
)
,
(
3
0
)
]
B
a
t
c
h
s
i
z
e
40
M
o
me
n
t
u
m
0
.
99
2
.
6
.
E
v
a
lua
t
i
o
n
E
v
alu
atin
g
a
m
o
d
el'
s
p
er
f
o
r
m
an
ce
is
an
im
p
o
r
ta
n
t
s
tep
in
th
e
p
r
o
ce
s
s
,
esp
ec
ially
wh
en
d
eter
m
in
in
g
th
e
b
est
m
o
d
el.
T
h
e
co
n
f
u
s
io
n
m
atr
ix
is
a
v
alu
ab
le
tech
n
iq
u
e
f
o
r
ev
alu
atin
g
cl
ass
if
icatio
n
m
o
d
el
p
er
f
o
r
m
an
ce
.
T
h
is
m
atr
ix
r
e
p
r
esen
ts
th
e
m
o
d
el'
s
p
r
ed
ictio
n
s
in
r
elatio
n
to
th
e
ac
tu
al
d
ata
co
n
d
itio
n
s
in
d
etail.
T
wo
f
u
n
d
am
e
n
tal
m
etr
ics,
ac
cu
r
ac
y
an
d
th
e
F1
-
s
co
r
e
,
ca
n
b
e
d
er
iv
ed
f
r
o
m
th
e
co
n
f
u
s
io
n
m
atr
ix
.
T
h
e
p
r
o
p
o
r
tio
n
o
f
co
r
r
ec
tly
p
r
ed
ict
ed
s
am
p
les
in
r
elatio
n
to
th
e
to
tal
s
am
p
le
co
u
n
t
is
r
ef
er
r
ed
t
o
as
ac
cu
r
ac
y
.
T
h
e
F1
-
s
co
r
e,
o
n
th
e
o
th
er
h
a
n
d
,
p
r
o
v
id
es
a
co
m
b
i
n
ed
m
ea
s
u
r
e
t
h
at
b
alan
ce
s
p
r
ec
is
io
n
an
d
r
ec
all.
Pre
cisi
o
n
is
th
e
r
atio
o
f
co
r
r
ec
tly
p
r
ed
icted
p
o
s
itiv
e
ca
s
es
to
to
tal
p
o
s
itiv
e
p
r
ed
ictio
n
s
,
wh
er
ea
s
r
ec
all
i
s
th
e
r
atio
o
f
tr
u
e
p
o
s
itiv
e
p
r
ed
ictio
n
s
to
t
o
tal
tr
u
e
p
o
s
itiv
e
d
ata
in
s
tan
ce
s
.
T
h
e
o
u
tco
m
es
o
f
th
ese
co
m
p
u
tatio
n
s
s
er
v
e
as
a
m
ea
n
s
to
ass
es
s
th
e
alg
o
r
ith
m
's
ef
f
icac
y
.
Acc
u
r
ac
y
v
alu
es
f
all
with
in
a
r
an
g
e
o
f
0
to
1
o
r
0
to
1
0
0
%
wh
en
ex
p
r
ess
ed
as
a
p
er
ce
n
tag
e.
A
h
ig
h
er
ac
cu
r
ac
y
v
alu
e
is
in
d
icativ
e
o
f
a
m
o
r
e
e
f
f
ec
tiv
e
u
tili
za
tio
n
o
f
th
e
al
g
o
r
ith
m
[
2
4
]
.
3.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
N
T
h
e
p
r
im
ar
y
o
b
jectiv
e
o
f
em
p
lo
y
in
g
tr
an
s
f
er
lear
n
in
g
m
o
d
e
ls
,
in
co
n
ju
n
ctio
n
with
r
ig
o
r
o
u
s
ly
test
ed
h
y
p
er
p
ar
am
eter
v
alu
es,
is
to
f
ac
ilit
ate
a
co
m
p
ar
ativ
e
ev
alu
atio
n
o
f
v
ar
io
u
s
m
o
d
el
ar
ch
itectu
r
es.
T
h
is
co
m
p
ar
ativ
e
a
n
aly
s
is
in
v
esti
g
ates
th
e
p
er
f
o
r
m
an
ce
o
f
E
f
f
icien
tNetB
3
,
Den
s
eNe
t
-
1
2
1
,
VG
G
-
1
6
,
an
d
R
esNet
-
1
5
2
,
with
a
s
p
ec
if
ic
em
p
h
as
is
o
n
ass
ess
in
g
ac
cu
r
ac
y
a
n
d
o
v
er
all
ef
f
icac
y
.
I
n
o
r
d
er
to
ac
co
m
p
lis
h
th
is
o
b
jectiv
e,
m
o
d
if
icatio
n
s
ar
e
m
ad
e
to
th
e
lear
n
in
g
r
ate
f
u
n
ctio
n
,
ta
k
in
g
in
to
co
n
s
id
er
at
io
n
th
e
s
u
b
s
tan
tial
in
f
lu
en
ce
o
f
th
e
lear
n
in
g
r
ate
o
n
ac
cu
r
ac
y
.
An
in
cr
ea
s
ed
lea
r
n
in
g
r
ate
ac
ce
ler
ates
th
e
tr
ai
n
in
g
p
r
o
ce
d
u
r
e
b
u
t
m
ay
p
o
te
n
tially
u
n
d
e
r
m
in
e
th
e
ac
cu
r
ac
y
o
f
th
e
n
etwo
r
k
,
wh
er
ea
s
a
r
ed
u
ce
d
lear
n
i
n
g
r
ate
r
esu
lts
in
a
s
lo
wer
y
et
p
o
ten
tially
m
o
r
e
p
r
ec
is
e
o
r
s
tab
le
tr
ain
in
g
p
r
o
ce
s
s
.
I
n
ad
d
itio
n
,
a
r
an
g
e
o
f
d
iv
er
s
e
ep
o
c
h
v
alu
es
is
u
tili
ze
d
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
6
,
No
.
1
,
Octo
b
er
20
24
:
509
-
5
1
6
514
in
th
is
s
tu
d
y
,
ac
k
n
o
wled
g
in
g
th
e
s
ig
n
if
ican
t
im
p
ac
t
th
at
e
p
o
ch
s
ettin
g
s
ca
n
h
av
e
o
n
th
e
p
er
f
o
r
m
an
ce
o
f
t
h
e
m
o
d
el.
E
p
o
ch
s
a
r
e
o
f
u
tm
o
s
t
im
p
o
r
tan
ce
in
th
e
co
n
tex
t
o
f
m
ac
h
in
e
lear
n
in
g
,
s
er
v
in
g
as
iter
ativ
e
s
tag
es
th
at
in
d
icate
th
e
lev
el
o
f
tr
ai
n
in
g
th
e
alg
o
r
ith
m
h
as
u
n
d
e
r
g
o
n
e
o
n
th
e
co
m
p
lete
d
ataset.
I
n
s
u
f
f
icien
t
n
u
m
b
er
s
o
f
tr
ain
in
g
iter
atio
n
s
ca
n
p
o
ten
ti
ally
lead
to
th
e
m
o
d
el'
s
in
ab
ilit
y
to
ca
p
tu
r
e
in
t
r
icate
d
ata
p
atter
n
s
,
r
esu
ltin
g
in
th
e
p
h
en
o
m
en
o
n
k
n
o
wn
as
u
n
d
er
f
itti
n
g
.
On
th
e
c
o
n
tr
ar
y
,
an
ex
ce
s
s
iv
e
n
u
m
b
er
o
f
ep
o
ch
s
ca
n
lead
to
o
v
er
f
itti
n
g
o
f
th
e
m
o
d
el,
w
h
er
ein
it m
em
o
r
izes th
e
tr
ain
i
n
g
s
et.
T
h
er
ef
o
r
e,
th
e
e
v
alu
atio
n
p
r
o
ce
s
s
ass
u
m
es
s
ig
n
if
ican
t
im
p
o
r
tan
ce
in
th
is
r
esear
ch
.
T
h
e
ev
alu
a
tio
n
m
eth
o
d
ch
o
s
en
f
o
r
th
is
s
tu
d
y
in
v
o
lv
es
th
e
im
p
lem
en
tatio
n
o
f
a
co
n
f
u
s
io
n
m
atr
ix
,
wh
ich
in
co
r
p
o
r
ates
v
ar
io
u
s
m
etr
ics
s
u
ch
as
p
r
ec
is
io
n
,
r
ec
all,
F1
-
s
co
r
e,
an
d
ac
cu
r
ac
y
.
E
v
er
y
in
d
iv
id
u
al
v
al
u
e
with
in
th
e
m
atr
ix
p
lay
s
a
s
ig
n
if
ican
t
r
o
le
in
ass
ess
in
g
t
h
e
m
o
d
el'
s
ef
f
ec
tiv
en
ess
an
d
f
ac
ilit
ates
co
m
p
r
e
h
en
s
io
n
o
f
th
e
s
p
ec
if
ic
er
r
o
r
s
m
ad
e
b
y
th
e
m
o
d
el.
T
h
e
c
o
m
p
r
eh
en
s
iv
e
p
r
esen
tatio
n
o
f
th
e
r
esear
ch
f
in
d
i
n
g
s
,
wh
ic
h
in
clu
d
es
th
e
m
atr
ix
v
alu
es f
o
r
ea
c
h
m
o
d
el,
ca
n
b
e
f
o
u
n
d
in
T
a
b
le
3
.
T
ab
le
3
.
Mo
d
el
co
m
p
ar
is
o
n
r
e
s
u
l
ts
M
o
d
e
l
Le
a
r
n
i
n
g
r
a
t
e
Ep
o
c
h
P
r
e
c
i
s
i
o
n
(
%)
R
e
c
a
l
l
(
%)
F1
-
sc
o
r
e
(
%)
A
c
c
u
r
a
c
y
(
%)
Ef
f
i
c
i
e
n
t
N
e
t
B
3
0
.
0
1
10
89
8
8
.
7
5
8
8
.
5
8
8
.
6
3
0
.
0
0
1
20
9
3
.
7
5
9
3
.
7
5
94
9
3
.
8
4
0
.
0
0
0
1
30
9
2
.
5
9
2
.
5
9
2
.
5
9
2
.
4
2
D
e
n
seN
e
t
-
1
2
1
0
.
0
1
10
7
7
.
7
5
6
6
.
5
6
6
.
7
5
6
8
.
2
4
0
.
0
0
1
20
9
5
.
5
9
5
.
7
5
9
5
.
7
5
9
5
.
7
3
0
.
0
0
0
1
30
9
6
.
5
96
9
6
.
2
5
9
6
.
2
0
VGG
-
16
0
.
0
1
10
7
4
.
2
5
4
9
.
7
5
4
7
.
2
5
5
1
.
1
8
0
.
0
0
1
20
9
1
.
2
5
9
1
.
2
5
91
91
0
.
0
0
0
1
30
9
6
.
2
5
9
5
.
7
5
96
9
5
.
7
3
R
e
sN
e
t
-
152
0
.
0
1
10
2
1
.
5
2
4
.
7
1
1
.
2
8
2
9
.
8
5
0
.
0
0
1
20
95
9
4
.
7
5
9
4
.
7
5
9
4
.
7
8
0
.
0
0
0
1
30
95
95
95
9
3
.
8
4
T
h
e
r
esu
lts
o
f
t
h
e
c
o
m
p
ar
ativ
e
an
aly
s
is
d
em
o
n
s
tr
ate
th
at
th
e
E
f
f
icien
tNetB
3
m
o
d
el
ac
h
iev
e
s
s
u
p
er
io
r
p
er
f
o
r
m
an
ce
wh
en
tr
ain
ed
with
a
lear
n
in
g
r
ate
o
f
0
.
0
0
1
an
d
an
ep
o
c
h
c
o
u
n
t
o
f
2
0
.
Den
s
eNe
t
-
1
2
1
ex
h
ib
its
s
u
p
er
io
r
p
e
r
f
o
r
m
an
ce
c
o
m
p
ar
ed
to
o
th
e
r
m
o
d
els,
esp
ec
i
ally
wh
en
u
tili
zin
g
a
lear
n
in
g
r
at
e
o
f
0
.
0
0
0
1
an
d
an
ep
o
ch
v
alu
e
o
f
3
0
.
Fu
r
th
e
r
m
o
r
e,
it
is
wo
r
th
n
o
tin
g
th
at
b
o
th
VGG
-
1
6
an
d
R
esNet
-
1
5
2
ex
h
ib
it
r
em
ar
k
ab
le
p
er
f
o
r
m
an
ce
w
h
en
a
lear
n
in
g
r
ate
o
f
0
.
0
0
0
1
an
d
an
ep
o
c
h
co
u
n
t
o
f
3
0
a
r
e
em
p
lo
y
ed
.
T
h
e
af
o
r
em
en
tio
n
ed
r
esu
lts
h
ig
h
lig
h
t
th
e
s
ig
n
if
ican
t
in
f
lu
en
ce
o
f
d
if
f
er
en
t
lear
n
in
g
r
ates
an
d
e
p
o
ch
v
alu
es
o
n
th
e
r
esu
ltan
t
ac
cu
r
ac
y
m
ea
s
u
r
e.
I
t
is
i
m
p
er
ativ
e
to
ac
k
n
o
wled
g
e
t
h
at
th
e
r
e
ex
is
ts
a
p
o
s
itiv
e
co
r
r
elatio
n
b
etwe
en
ac
cu
r
ac
y
an
d
e
p
o
ch
co
u
n
t,
s
u
g
g
esti
n
g
a
u
n
id
ir
ec
tio
n
al
ass
o
ciatio
n
.
Pu
t
s
im
p
ly
,
wh
en
th
e
n
u
m
b
er
o
f
e
p
o
ch
s
is
in
cr
ea
s
ed
,
th
e
ac
c
u
r
ac
y
v
alu
es
f
o
r
b
o
th
th
e
tr
ain
in
g
an
d
v
ali
d
atio
n
d
atasets
also
in
cr
ea
s
e.
On
th
e
o
th
e
r
h
a
n
d
,
a
n
e
g
a
t
i
v
e
c
o
r
r
e
l
a
t
i
o
n
c
a
n
b
e
o
b
s
e
r
v
e
d
b
e
t
w
e
e
n
t
h
e
n
u
m
b
e
r
o
f
e
p
o
c
h
s
a
n
d
t
h
e
l
o
s
s
m
e
t
r
i
c
.
T
h
i
s
s
u
g
g
e
s
t
s
t
h
at
a
s
t
h
e
n
u
m
b
e
r
o
f
e
p
o
c
h
s
i
n
c
r
e
a
s
es,
t
h
e
r
e
i
s
a
c
o
r
r
e
s
p
o
n
d
i
n
g
d
e
c
r
ea
s
e
i
n
t
h
e
l
o
s
s
v
a
l
u
e
o
b
s
e
r
v
e
d
in
t
h
e
t
r
a
i
n
i
n
g
d
a
t
a
.
Up
o
n
ev
alu
atin
g
th
e
ac
cu
r
ac
y
v
alu
es
o
b
tain
ed
f
r
o
m
th
e
tr
ain
ed
m
o
d
els,
it
is
ev
id
en
t
th
at
th
e
Den
s
eNe
t
-
1
2
1
m
o
d
el
ac
h
ie
v
ed
th
e
h
ig
h
est
ac
cu
r
ac
y
,
r
ea
ch
in
g
a
n
im
p
r
ess
iv
e
9
6
.
2
0
%.
T
h
is
ex
ce
p
tio
n
al
p
er
f
o
r
m
an
ce
was
ac
h
iev
ed
with
a
lear
n
i
n
g
r
ate
s
et
at
0
.
0
0
0
1
an
d
3
0
e
p
o
ch
s
.
Fo
llo
win
g
clo
s
ely
,
VGG
-
1
6
attain
ed
an
ac
cu
r
ac
y
o
f
9
5
.
7
3
%
u
n
d
er
th
e
s
am
e
co
n
f
ig
u
r
atio
n
o
f
a
lear
n
in
g
r
ate
o
f
0
.
0
0
0
1
an
d
3
0
ep
o
ch
s
.
R
esNet
-
1
5
2
s
ec
u
r
ed
an
ac
c
u
r
ac
y
o
f
9
4
.
7
6
%
u
s
in
g
a
l
ea
r
n
in
g
r
ate
o
f
0
.
0
0
1
a
n
d
2
0
ep
o
c
h
s
,
wh
ile
E
f
f
icien
tNetB
3
y
ield
ed
an
ac
c
u
r
ac
y
o
f
9
3
.
8
4
%
with
th
e
s
am
e
lear
n
in
g
r
ate
a
n
d
ep
o
ch
s
ettin
g
s
.
I
t
is
ess
en
tial
to
ac
k
n
o
wled
g
e
th
at
th
e
ar
c
h
ite
ctu
r
al
d
esig
n
o
f
t
h
ese
m
o
d
els
ca
n
ex
er
t
a
p
r
o
f
o
u
n
d
in
f
lu
en
ce
o
n
th
eir
ac
cu
r
ac
y
.
Var
iab
les
s
u
ch
as
lay
er
s
ize,
n
u
m
b
er
o
f
lay
er
s
,
lay
er
ty
p
e,
an
d
th
e
in
ter
p
l
ay
b
etwe
en
lay
er
s
s
ig
n
if
ican
tly
im
p
ac
t th
e
m
o
d
el
s
'
ab
ili
ty
to
d
is
ce
r
n
an
d
in
ter
p
r
et
p
atter
n
s
with
in
t
h
e
d
ata.
Pre
v
io
u
s
s
tu
d
ies
in
th
is
d
o
m
ai
n
h
a
v
e
also
y
ield
ed
n
o
tewo
r
th
y
r
esu
lts
.
Fo
r
in
s
tan
ce
,
I
s
lam
et
a
l.
[
2
5
]
r
ep
o
r
ted
an
ac
c
u
r
ac
y
o
f
9
6
.
2
5
%
in
d
etec
tin
g
g
lau
c
o
m
a
d
is
ea
s
e
u
s
in
g
E
f
f
icien
tNetB
3
.
Similar
ly
,
Par
ad
is
a
et
a
l.
[
2
6
]
u
tili
ze
d
Den
s
eNe
t
-
1
2
1
+I
n
ce
p
t
io
n
-
R
esNetV2
ar
ch
itectu
r
e
to
d
etec
t
d
iab
etic
ey
e
d
is
ea
s
e,
g
ar
n
er
in
g
an
ac
cu
r
ac
y
o
f
9
1
%
.
No
tab
ly
,
th
ese
s
tu
d
ies
o
f
ten
f
o
cu
s
ed
o
n
s
in
g
le
-
e
y
e
d
is
ea
s
e
d
etec
tio
n
o
r
ca
r
r
ied
o
u
t
th
e
class
if
icatio
n
s
ep
ar
ately
.
I
n
co
n
tr
ast,
th
is
r
esear
ch
en
co
m
p
ass
ed
th
e
s
im
u
ltan
eo
u
s
class
if
icatio
n
o
f
th
r
ee
ey
e
d
is
ea
s
e
ca
teg
o
r
ies
ca
tar
ac
t,
g
lau
co
m
a,
a
n
d
d
iab
etic
r
etin
o
p
ath
y
y
ield
i
n
g
a
n
a
v
er
ag
e
ac
c
u
r
ac
y
o
f
9
6
.
2
0
%
with
t
h
e
Den
s
eNe
t
-
1
2
1
ar
ch
itectu
r
e,
s
u
r
p
ass
in
g
o
t
h
er
m
o
d
els,
in
clu
d
in
g
VGG
-
1
6
,
E
f
f
icien
tNetB
3
,
an
d
R
esNet
-
152.
T
h
ese
ac
cu
r
ac
y
r
esu
lts
ac
r
o
s
s
v
ar
io
u
s
m
o
d
els
af
f
ir
m
th
eir
ef
f
ec
tiv
en
ess
in
class
if
y
in
g
r
etin
al
f
u
n
d
u
s
im
ag
e
d
ata
f
o
r
th
e
d
etec
tio
n
o
f
th
e
af
o
r
em
en
tio
n
ed
ey
e
d
is
ea
s
es.
Sev
er
al
f
ac
to
r
s
ca
n
i
n
f
l
u
en
ce
th
e
ac
cu
r
ac
y
o
u
tco
m
es,
in
clu
d
in
g
d
ataset
s
i
ze
,
d
ata
p
r
o
ce
s
s
in
g
s
tag
es
lik
e
im
ag
e
au
g
m
en
tatio
n
an
d
d
at
a
s
p
litt
in
g
,
an
d
th
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
E
ye
d
is
ea
s
e
d
etec
tio
n
u
s
in
g
tr
a
n
s
fer lea
r
n
in
g
b
a
s
ed
o
n
r
etin
a
l fu
n
d
u
s
ima
g
e
d
a
t
a
(
Helmi
I
ma
d
u
d
d
i
n
)
515
co
n
f
ig
u
r
atio
n
o
f
h
y
p
er
p
ar
am
e
ter
s
d
u
r
in
g
t
h
e
m
o
d
el
tr
ain
in
g
p
r
o
ce
s
s
.
T
h
ese
f
ac
t
o
r
s
will
af
f
ec
t
th
e
ac
cu
r
ac
y
p
r
o
d
u
ce
d
b
y
ea
ch
m
o
d
el.
T
h
e
b
etter
t
h
e
d
ata
p
r
e
p
ar
atio
n
d
o
n
e
b
ef
o
r
e
m
o
d
el
t
r
ain
in
g
.
T
h
en
th
e
b
etter
th
e
ac
cu
r
ac
y
p
r
o
d
u
ce
d
b
y
ea
c
h
m
o
d
el
.
Nag
p
al
et
a
l.
[
2
7
]
s
tates
th
at
d
ata
p
r
ep
r
o
ce
s
s
in
g
o
r
d
ata
p
r
ep
ar
atio
n
b
ef
o
r
e
d
ata
is
f
u
r
th
er
p
r
o
ce
s
s
ed
will
r
ed
u
ce
n
o
is
e
in
th
e
im
a
g
e
to
b
e
p
r
o
ce
s
s
ed
,
th
is
will
r
esu
lt
in
b
etter
class
if
icatio
n
.
4.
CO
NCLU
SI
O
N
T
h
e
p
r
o
ce
s
s
o
f
d
etec
tin
g
ey
e
d
is
ea
s
es
in
o
u
r
r
esear
c
h
f
ac
ed
s
ev
er
al
ch
allen
g
es,
with
li
m
ited
d
ata
av
ailab
ilit
y
b
ein
g
a
s
ig
n
if
ican
t
im
p
ed
im
en
t
th
at
co
u
l
d
p
o
ten
t
ially
af
f
ec
t
th
e
ac
cu
r
ac
y
o
f
o
u
r
m
o
d
els.
T
r
an
s
f
e
r
lear
n
in
g
,
a
f
u
n
d
am
e
n
tal
co
n
c
ep
t
em
p
l
o
y
ed
i
n
th
is
s
tu
d
y
,
in
v
o
lv
es
lev
er
ag
i
n
g
p
r
e
-
tr
ain
e
d
m
o
d
els
th
at
h
av
e
b
ee
n
tr
ain
ed
o
n
ex
ten
s
iv
e
d
at
asets
to
f
ac
ilit
ate
th
e
u
n
d
er
s
tan
d
in
g
o
f
n
ew
d
ata.
T
h
is
ap
p
r
o
ac
h
was
ad
o
p
ted
to
en
h
an
ce
th
e
ac
cu
r
ac
y
o
f
o
u
r
m
o
d
els.
T
h
e
p
r
o
p
o
s
ed
m
eth
o
d
p
er
f
o
r
m
s
well
with
s
ev
er
al
tr
an
s
f
er
lear
n
in
g
m
o
d
els,
in
clu
d
i
n
g
E
f
f
icien
tNetB
3
,
Den
s
eNe
t
-
1
2
1
,
VGG
-
1
6
,
an
d
R
esNet
-
1
5
2
,
c
o
u
p
led
with
th
e
ex
p
l
o
r
atio
n
o
f
v
ar
io
u
s
h
y
p
er
p
ar
a
m
eter
s
,
we
ac
h
iev
ed
th
e
h
i
g
h
est
ac
cu
r
ac
y
with
th
e
Den
s
eNe
t
-
1
2
1
m
o
d
el.
I
t
d
em
o
n
s
tr
ate
d
p
r
ec
is
io
n
,
r
ec
all,
an
d
F
1
-
s
co
r
e
v
alu
es
o
f
9
6
.
5
%,
9
6
%,
a
n
d
9
6
.
2
5
%,
r
esp
ec
tiv
ely
,
alo
n
g
with
an
o
v
e
r
all
ac
cu
r
ac
y
o
f
9
6
.
2
0
%.
Fu
r
th
er
m
o
r
e,
o
u
r
in
v
esti
g
atio
n
h
ig
h
lig
h
ted
th
e
s
ig
n
if
ican
t
im
p
ac
t
o
f
d
if
f
er
e
n
t
lear
n
in
g
r
ates
an
d
ep
o
c
h
s
o
n
ac
cu
r
ac
y
.
No
tab
ly
,
th
e
u
s
e
o
f
a
lear
n
i
n
g
r
ate
s
et
at
0
.
0
0
0
1
an
d
an
ep
o
ch
co
u
n
t
o
f
3
0
co
n
s
is
ten
tly
y
ield
ed
h
ig
h
er
a
cc
u
r
ac
y
v
al
u
es
co
m
p
ar
e
d
to
o
th
er
co
n
f
ig
u
r
atio
n
s
.
C
o
n
s
eq
u
en
tly
,
o
u
r
r
esear
ch
d
em
o
n
s
tr
ates
th
e
e
f
f
ec
tiv
en
ess
o
f
t
h
e
s
elec
ted
tr
a
n
s
f
er
lear
n
in
g
m
o
d
el
in
class
if
y
in
g
r
etin
al
f
u
n
d
u
s
im
ag
e
d
at
a
f
o
r
th
e
d
etec
tio
n
o
f
ca
tar
ac
t,
g
lau
co
m
a,
a
n
d
d
iab
etic
r
etin
o
p
ath
y
d
is
ea
s
es.
T
h
is
s
tu
d
y
s
er
v
es
as
a
v
alu
ab
le
b
en
ch
m
ar
k
f
o
r
f
u
tu
r
e
r
esear
c
h
en
d
ea
v
o
u
r
s
,
o
f
f
e
r
in
g
in
s
ig
h
t
s
in
to
p
o
ten
tial
av
en
u
es
f
o
r
im
p
r
o
v
i
n
g
ac
cu
r
ac
y
.
Fu
tu
r
e
in
v
esti
g
atio
n
s
c
o
u
ld
e
x
p
lo
r
e
a
b
r
o
ad
er
s
p
ec
tr
u
m
o
f
tr
an
s
f
er
lear
n
in
g
m
o
d
els
a
n
d
lev
e
r
ag
e
lar
g
er
d
atasets
to
f
u
r
th
er
en
h
an
ce
th
e
class
if
icatio
n
p
er
f
o
r
m
a
n
ce
in
th
is
cr
itical
d
o
m
ain
o
f
ey
e
d
is
ea
s
e
d
etec
tio
n
.
ACK
NO
WL
E
DG
E
M
E
NT
S
T
h
e
au
th
o
r
s
ex
p
r
ess
g
r
atitu
d
e
to
th
e
Un
iv
er
s
itas
Mu
h
am
m
ad
iy
ah
Su
r
ak
a
r
ta
f
o
r
p
r
o
v
id
in
g
r
esear
ch
s
u
p
p
o
r
t,
e
n
ab
lin
g
th
e
co
m
p
letio
n
o
f
th
is
r
esear
ch
.
RE
F
E
R
E
NC
E
S
[
1
]
S
.
R
.
F
l
a
x
ma
n
e
t
a
l
.
,
“
G
l
o
b
a
l
c
a
u
s
e
s
o
f
b
l
i
n
d
n
e
ss
a
n
d
d
i
st
a
n
c
e
v
i
si
o
n
i
m
p
a
i
r
me
n
t
1
9
9
0
–
2
0
2
0
:
a
s
y
s
t
e
ma
t
i
c
r
e
v
i
e
w
a
n
d
m
e
t
a
-
a
n
a
l
y
si
s
,
”
L
a
n
c
e
t
G
l
o
b
.
H
e
a
l
.
,
v
o
l
.
5
,
n
o
.
1
2
,
p
p
.
e
1
2
2
1
–
e
1
2
3
4
,
D
e
c
.
2
0
1
7
,
d
o
i
:
1
0
.
1
0
1
6
/
S
2
2
1
4
-
1
0
9
X
(
1
7
)
3
0
3
9
3
-
5.
[
2
]
E.
M
a
r
d
a
l
e
n
a
a
n
d
F
.
H
a
y
a
t
i
,
“
P
r
e
v
a
l
e
n
c
e
o
f
b
l
i
n
d
n
e
ss
d
u
e
t
o
c
a
t
a
r
a
c
t
s
i
n
R
S
U
D
d
r
.
Za
i
n
o
e
l
A
b
i
d
i
n
B
a
n
d
a
A
c
e
h
,
”
J
u
r
n
a
l
Ri
s
e
t
S
a
i
n
s
d
a
n
T
e
k
n
o
l
o
g
i
J
RS
T
,
v
o
l
.
1
1
,
n
o
.
3
,
p
p
.
5
9
7
–
6
0
2
,
2
0
2
1
,
d
o
i
:
1
0
.
4
7
6
4
7
/
j
sr
.
v
1
1
i
3
.
7
9
4
[
3
]
V
.
N
a
n
g
i
a
e
t
a
l
.
,
“
P
r
e
v
a
l
e
n
c
e
a
n
d
c
a
u
ses
o
f
b
l
i
n
d
n
e
ss a
n
d
v
i
s
i
o
n
i
m
p
a
i
r
me
n
t
:
m
a
g
n
i
t
u
d
e
,
t
e
mp
o
r
a
l
t
r
e
n
d
s
a
n
d
p
r
o
j
e
c
t
i
o
n
s i
n
so
u
t
h
a
n
d
C
e
n
t
r
a
l
A
s
i
a
,
”
Bri
t
i
s
h
J
o
u
r
n
a
l
o
f
O
p
h
t
h
a
l
m
o
l
o
g
y
,
v
o
l
.
1
0
3
,
n
o
.
7
,
p
p
.
8
7
1
–
8
7
7
,
N
o
v
.
2
0
1
8
,
d
o
i
:
1
0
.
1
1
3
6
/
b
j
o
p
h
t
h
a
l
m
o
l
-
2
0
1
8
-
3
1
2
2
9
2
[
4
]
R
.
S
.
S
a
l
v
i
,
S
.
R
.
La
b
h
set
w
a
r
,
P
.
A
.
K
o
l
t
e
,
V
.
S
.
V
e
n
k
a
t
e
sh
,
a
n
d
A
.
M
.
B
a
r
e
t
t
o
,
“
P
r
e
d
i
c
t
i
v
e
a
n
a
l
y
s
i
s
o
f
d
i
a
b
e
t
i
c
r
e
t
i
n
o
p
a
t
h
y
w
i
t
h
t
r
a
n
sf
e
r
l
e
a
r
n
i
n
g
,
”
i
n
2
0
2
1
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
N
a
s
c
e
n
t
T
e
c
h
n
o
l
o
g
i
e
s
i
n
E
n
g
i
n
e
e
r
i
n
g
,
I
C
N
ET
2
0
2
1
-
P
r
o
c
e
e
d
i
n
g
s
,
2
0
2
1
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
N
TE
5
1
1
8
5
.
2
0
2
1
.
9
4
8
7
7
8
9
.
[
5
]
W
.
S
u
p
r
i
y
a
n
t
i
a
n
d
D
.
A
.
A
n
g
g
o
r
o
,
“
C
l
a
ssi
f
i
c
a
t
i
o
n
o
f
p
a
n
d
a
v
a
s
f
i
g
u
r
e
i
n
s
h
a
d
o
w
p
u
p
p
e
t
i
m
a
g
e
s
u
s
i
n
g
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
u
r
a
l
ne
t
w
o
r
k
s,”
K
h
a
z
a
n
a
h
I
n
f
o
rm
a
t
i
k
a
:
J
u
rn
a
l
I
l
m
u
K
o
m
p
u
t
e
r
d
a
n
I
n
f
o
rm
a
t
i
k
a
,
v
o
l
.
7
,
n
o
.
1
,
p
p
.
1
8
–
2
4
,
2
0
2
1
,
d
o
i
:
1
0
.
2
3
9
1
7
/
k
h
i
f
.
v
7
i
1
.
1
2
4
8
4
.
[
6
]
I
.
D
.
A
p
o
s
t
o
l
o
p
o
u
l
o
s
a
n
d
T
.
A
.
M
p
e
si
a
n
a
,
“
C
O
V
I
D
-
1
9
:
a
u
t
o
ma
t
i
c
d
e
t
e
c
t
i
o
n
f
r
o
m
X
-
r
a
y
i
m
a
g
e
s
u
t
i
l
i
z
i
n
g
t
r
a
n
s
f
e
r
l
e
a
r
n
i
n
g
w
i
t
h
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,”
Ph
y
si
c
a
l
a
n
d
En
g
i
n
e
e
ri
n
g
S
c
i
e
n
c
e
s
i
n
M
e
d
i
c
i
n
e
,
v
o
l
.
4
3
,
n
o
.
2
,
p
p
.
6
3
5
–
6
4
0
,
Ju
n
.
2
0
2
0
,
d
o
i
:
1
0
.
1
0
0
7
/
s
1
3
2
4
6
-
0
2
0
-
0
0
8
6
5
-
4.
[
7
]
R
.
S
a
r
k
i
e
t
a
l
.
,
“
I
mag
e
p
r
e
p
r
o
c
e
ss
i
n
g
i
n
c
l
a
ssi
f
i
c
a
t
i
o
n
a
n
d
i
d
e
n
t
i
f
i
c
a
t
i
o
n
o
f
d
i
a
b
e
t
i
c
e
y
e
d
i
se
a
ses
,
”
D
a
t
a
S
c
i
e
n
c
e
a
n
d
E
n
g
i
n
e
e
ri
n
g
,
v
o
l
.
6
,
n
o
.
4
,
p
p
.
4
5
5
–
4
7
1
,
A
u
g
.
2
0
2
1
.
d
o
i
:
1
0
.
1
0
0
7
/
s4
1
0
1
9
-
0
2
1
-
0
0
1
6
7
-
z
.
[
8
]
K
.
P
i
n
,
J.
H
o
C
h
a
n
g
,
a
n
d
Y
.
N
a
m,
“
C
o
mp
a
r
a
t
i
v
e
st
u
d
y
o
f
t
r
a
n
sf
e
r
l
e
a
r
n
i
n
g
mo
d
e
l
s
f
o
r
r
e
t
i
n
a
l
d
i
s
e
a
se
d
i
a
g
n
o
si
s
f
r
o
m
f
u
n
d
u
s
i
ma
g
e
s,
”
C
o
m
p
u
t
e
rs,
Ma
t
e
r
i
a
l
s,
C
o
n
t
i
n
u
a
,
v
o
l
.
7
0
,
n
o
.
3
,
p
p
.
5
8
2
1
–
5
8
3
4
,
2
0
2
2
,
d
o
i
:
1
0
.
3
2
6
0
4
/
c
mc.
2
0
2
2
.
0
2
1
9
4
3
.
[
9
]
A
.
S
u
g
e
n
o
,
Y
.
I
sh
i
k
a
w
a
,
T
.
O
h
sh
i
m
a
,
a
n
d
R
.
M
u
r
a
m
a
t
s
u
,
“
S
i
mp
l
e
me
t
h
o
d
s
f
o
r
t
h
e
l
e
s
i
o
n
d
e
t
e
c
t
i
o
n
a
n
d
se
v
e
r
i
t
y
g
r
a
d
i
n
g
o
f
d
i
a
b
e
t
i
c
r
e
t
i
n
o
p
a
t
h
y
b
y
i
ma
g
e
p
r
o
c
e
s
si
n
g
a
n
d
t
r
a
n
sf
e
r
l
e
a
r
n
i
n
g
,
”
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
3
7
,
p
.
1
0
4
7
9
5
,
O
c
t
.
2
0
2
1
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
c
o
mp
b
i
o
m
e
d
.
2
0
2
1
.
1
0
4
7
9
5
.
[
1
0
]
B
.
T
a
şar,
“
S
k
i
n
C
a
n
c
e
r
N
e
t
:
A
u
t
o
mat
e
d
c
l
a
ss
i
f
i
c
a
t
i
o
n
o
f
sk
i
n
l
e
s
i
o
n
u
si
n
g
d
e
e
p
t
r
a
n
sf
e
r
l
e
a
r
n
i
n
g
me
t
h
o
d
,
”
T
r
a
i
t
e
m
e
n
t
D
u
S
i
g
n
a
l
,
v
o
l
.
4
0
,
n
o
.
1
,
p
p
.
2
8
5
–
2
9
5
,
F
e
b
.
2
0
2
3
,
d
o
i
:
1
0
.
1
8
2
8
0
/
t
s
.
4
0
0
1
2
8
.
[
1
1
]
D
.
Le
e
t
a
l
.
,
“
Tr
a
n
sf
e
r
l
e
a
r
n
i
n
g
f
o
r
a
u
t
o
ma
t
e
d
O
C
TA
d
e
t
e
c
t
i
o
n
o
f
d
i
a
b
e
t
i
c
r
e
t
i
n
o
p
a
t
h
y
,
”
T
r
a
n
s
l
a
t
i
o
n
a
l
Vi
s
i
o
n
S
c
i
e
n
c
e
and
T
e
c
h
n
o
l
o
g
y
,
v
o
l
.
9
,
n
o
.
2
,
p
.
3
5
,
J
u
l
.
2
0
2
0
,
d
o
i
:
1
0
.
1
1
6
7
/
t
v
s
t
.
9
.
2
.
3
5
.
[
1
2
]
H
.
I
mad
u
d
d
i
n
a
n
d
B
.
A
.
H
e
r
ma
n
sy
a
h
,
“
Tr
a
n
sf
e
r
l
e
a
r
n
i
n
g
f
o
r
d
e
t
e
c
t
i
n
g
C
O
V
I
D
-
1
9
o
n
x
-
r
a
y
u
s
i
n
g
d
e
e
p
r
e
si
d
u
a
l
n
e
t
w
o
r
k
,
”
B
u
l
l
e
t
i
n
o
f
El
e
c
t
ri
c
a
l
E
n
g
i
n
e
e
ri
n
g
a
n
d
I
n
f
o
rm
a
t
i
c
s
,
v
o
l
.
1
1
,
n
o
.
6
,
2
0
2
2
,
d
o
i
:
1
0
.
1
1
5
9
1
/
e
e
i
.
v
1
1
i
6
.
4
3
3
4
.
[
1
3
]
S
.
G
u
,
M
.
P
e
d
n
e
k
a
r
,
a
n
d
R
.
S
l
a
t
e
r
,
“
I
mp
r
o
v
e
i
ma
g
e
c
l
a
ss
i
f
i
c
a
t
i
o
n
u
s
i
n
g
d
a
t
a
a
u
g
m
e
n
t
a
t
i
o
n
a
n
d
n
e
u
r
a
l
n
e
t
w
or
k
s,”
S
MU
D
a
t
a
S
c
i
e
n
c
e
Re
v
i
e
w
,
v
o
l
.
2
,
n
o
.
2
,
A
u
g
.
2
0
1
9
,
A
c
c
e
sse
d
:
A
p
r
.
1
7
,
2
0
2
3
,
[
O
n
l
i
n
e
]
.
A
v
a
i
l
a
b
l
e
:
h
t
t
p
s
:
/
/
sc
h
o
l
a
r
.
smu
.
e
d
u
/
d
a
t
a
sc
i
e
n
c
e
r
e
v
i
e
w
/
v
o
l
2
/
i
ss
2
/
1
.
[
1
4
]
M
.
H
u
h
,
P
.
A
g
r
a
w
a
l
,
a
n
d
A
.
A
.
Ef
r
o
s,
“
W
h
a
t
ma
k
e
s
I
mag
e
N
e
t
g
o
o
d
f
o
r
t
r
a
n
sf
e
r
l
e
a
r
n
i
n
g
?
,
”
Be
rk
e
l
e
y
Ar
t
i
f
.
I
n
t
e
l
l
.
Re
s.
L
a
b
.
U
C
Be
rk
e
l
e
y
,
A
u
g
.
2
0
1
6
,
A
c
c
e
sse
d
:
A
p
r
.
1
7
,
2
0
2
3
.
[
O
n
l
i
n
e
]
.
A
v
a
i
l
a
b
l
e
:
h
t
t
p
s
:
/
/
a
r
x
i
v
.
o
r
g
/
a
b
s
/
1
6
0
8
.
0
8
6
1
4
v
2
.
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
6
,
No
.
1
,
Octo
b
er
20
24
:
509
-
5
1
6
516
[
1
5
]
R
.
G
o
d
a
s
u
,
D
.
Z
e
n
g
,
a
n
d
K
.
S
u
t
r
a
v
e
,
“
Tr
a
n
sf
e
r
l
e
a
r
n
i
n
g
i
n
m
e
d
i
c
a
l
i
m
a
g
e
c
l
a
ssi
f
i
c
a
t
i
o
n
:
c
h
a
l
l
e
n
g
e
s
a
n
d
o
p
p
o
r
t
u
n
i
t
i
e
s
o
p
p
o
r
t
u
n
i
t
i
e
s,
”
MW
AI
S
2
0
2
0
P
ro
c
e
e
d
i
n
g
s
.
,
p
p
.
5
–
2
8
,
2
0
2
0
,
A
c
c
e
sse
d
:
A
p
r
.
1
8
,
2
0
2
3
.
[
1
6
]
A
.
Jar
a
d
a
t
e
t
a
l
.
,
“
A
u
t
o
m
a
t
e
d
m
o
n
k
e
y
p
o
x
s
k
i
n
l
e
si
o
n
d
e
t
e
c
t
i
o
n
u
s
i
n
g
d
e
e
p
l
e
a
r
n
i
n
g
a
n
d
t
r
a
n
sf
e
r
l
e
a
r
n
i
n
g
t
e
c
h
n
i
q
u
e
s
,
”
I
n
t
e
r
n
a
t
i
o
n
a
l
J
o
u
r
n
a
l
o
f
En
v
i
r
o
n
m
e
n
t
a
l
R
e
se
a
rc
h
a
n
d
P
u
b
l
i
c
H
e
a
l
t
h
,
v
o
l
.
2
0
,
n
o
.
5
,
p
.
4
4
2
2
,
M
a
r
.
2
0
2
3
,
d
o
i
:
1
0
.
3
3
9
0
/
i
j
e
r
p
h
2
0
0
5
4
4
2
2
.
[
1
7
]
A
.
B
h
a
t
t
,
A
.
G
a
n
a
t
r
a
,
a
n
d
K
.
K
o
t
e
c
h
a
,
“
C
O
V
I
D
-
1
9
p
u
l
mo
n
a
r
y
c
o
n
s
o
l
i
d
a
t
i
o
n
s
d
e
t
e
c
t
i
o
n
i
n
c
h
e
st
X
-
r
a
y
u
si
n
g
p
r
o
g
r
e
ssi
v
e
r
e
s
i
z
i
n
g
a
n
d
t
r
a
n
sf
e
r
l
e
a
r
n
i
n
g
t
e
c
h
n
i
q
u
e
s,”
H
e
l
i
y
o
n
,
v
o
l
.
7
,
n
o
.
6
,
J
u
n
.
2
0
2
1
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
h
e
l
i
y
o
n
.
2
0
2
1
.
e
0
7
2
1
1
.
[
1
8
]
F
.
Za
e
l
a
n
i
a
n
d
Y
.
M
i
f
t
a
h
u
d
d
i
n
,
“
C
o
mp
a
r
i
s
o
n
o
f
Ef
f
i
c
i
e
n
t
N
e
t
B
3
a
n
d
M
o
b
i
l
e
N
e
t
V
2
me
t
h
o
d
s
f
o
r
i
d
e
n
t
i
f
i
c
a
t
i
o
n
o
f
f
r
u
i
t
t
y
p
e
s
u
si
n
g
l
e
a
f
f
e
a
t
u
r
e
s
,
”
J
I
T
T
ER
,
v
o
l
.
9
,
n
o
.
1
,
p
p
.
1
–
1
1
,
2
0
2
2
,
d
o
i
:
1
0
.
3
3
1
9
7
/
j
i
t
t
e
r
.
v
o
l
9
.
i
ss1
.
2
0
2
2
.
9
1
1
.
[
1
9
]
G
.
H
u
a
n
g
,
Z
.
L
i
u
,
a
n
d
L.
v
a
n
d
e
r
M
a
a
t
e
n
,
“
D
e
n
se
l
y
c
o
n
n
e
c
t
e
d
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
t
w
o
r
ks
,
”
I
n
Pr
o
c
e
e
d
i
n
g
s
o
f
t
h
e
I
EEE
C
o
n
f
e
re
n
c
e
o
n
C
o
m
p
u
t
e
r
Vi
s
i
o
n
a
n
d
P
a
t
t
e
r
n
R
e
c
o
g
n
i
t
i
o
n
,
p
p
.
4
7
0
0
-
4
7
0
8
.
2
0
1
7
,
d
o
i
:
1
0
.
4
8
5
5
0
/
a
r
X
i
v
.
1
6
0
8
.
0
6
9
9
3
.
[
2
0
]
X
.
L
i
,
X
.
S
h
e
n
,
Y
.
Zh
o
u
,
X
.
W
a
n
g
,
a
n
d
T.
Q
.
Li
,
“
C
l
a
ss
i
f
i
c
a
t
i
o
n
o
f
b
r
e
a
st
c
a
n
c
e
r
h
i
st
o
p
a
t
h
o
l
o
g
i
c
a
l
i
m
a
g
e
s
u
si
n
g
i
n
t
e
r
l
e
a
v
e
d
D
e
n
seN
e
t
w
i
t
h
S
EN
e
t
(
I
D
S
N
e
t
)
,
”
PL
o
S
O
n
e
,
v
o
l
.
1
5
,
n
o
.
5
,
M
a
y
2
0
2
0
,
d
o
i
:
1
0
.
1
3
7
1
/
j
o
u
r
n
a
l
.
p
o
n
e
.
0
2
3
2
1
2
7
.
[
2
1
]
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
s
i
d
u
a
l
l
e
a
r
n
i
n
g
f
o
r
i
m
a
g
e
r
e
c
o
g
n
i
t
i
o
n
,
”
Pro
c
e
e
d
i
n
g
s
o
f
t
h
e
I
EE
E
C
o
n
f
e
re
n
c
e
o
n
C
o
m
p
u
t
e
r
Vi
s
i
o
n
a
n
d
P
a
t
t
e
r
n
Re
c
o
g
n
i
t
i
o
n
,
p
p
.
7
7
0
–
7
7
8
,
D
e
c
.
2
0
1
6
,
d
o
i
:
1
0
.
1
1
0
9
/
C
V
P
R
.
2
0
1
6
.
9
0
.
[
2
2
]
R
.
D
.
N
u
r
f
i
t
a
,
G
.
A
r
i
y
a
n
t
o
,
“
Te
n
s
o
r
f
l
o
w
-
b
a
s
e
d
d
e
e
p
l
e
a
r
n
i
n
g
i
m
p
l
e
me
n
t
a
t
i
o
n
f
o
r
f
i
n
g
e
r
p
r
i
n
t
r
e
c
o
g
n
i
t
i
o
n
,
”
J
o
u
r
n
a
l
Em
i
t
,
v
o
l
.
1
8
,
n
o
.
0
1
,
p
p
.
2
2
–
27
,
2
0
1
8
,
d
o
i
:
1
0
.
2
3
9
1
7
/
e
mi
t
o
r
.
v
1
8
i
0
1
.
6
2
3
6
.
[
2
3
]
C
.
M
.
S
h
a
r
m
a
,
K
.
T
o
mar,
R
.
K
.
M
i
s
h
r
a
,
a
n
d
V
.
M
.
C
h
a
r
i
a
r
,
“
I
n
d
i
a
n
s
i
g
n
l
a
n
g
u
a
g
e
r
e
c
o
g
n
i
t
i
o
n
u
si
n
g
f
i
n
e
-
t
u
n
e
d
d
e
e
p
t
r
a
n
sf
e
r
l
e
a
r
n
i
n
g
m
o
d
e
l
,
”
Pr
o
c
e
e
d
i
n
g
s
o
f
t
h
e
1
st
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
n
o
v
a
t
i
o
n
i
n
C
o
m
p
u
t
e
r
a
n
d
I
n
f
o
rm
a
t
i
o
n
S
c
i
e
n
c
e
,
2
0
2
1
.
d
o
i
:
1
0
.
5
2
2
0
/
0
0
1
0
7
9
0
3
0
0
0
0
3
1
6
7
[
2
4
]
E.
F
.
O
h
a
t
a
,
J.
V
.
S
.
d
a
s
C
h
a
g
a
s,
G
.
M
.
B
e
z
e
r
r
a
,
M
.
M
.
H
a
ssa
n
,
V
.
H
.
C
.
d
e
A
l
b
u
q
u
e
r
q
u
e
,
a
n
d
P
.
P
.
R
.
F
i
l
h
o
,
“
A
n
o
v
e
l
t
r
a
n
sf
e
r
l
e
a
r
n
i
n
g
a
p
p
r
o
a
c
h
f
o
r
t
h
e
c
l
a
ss
i
f
i
c
a
t
i
o
n
o
f
h
i
s
t
o
l
o
g
i
c
a
l
i
ma
g
e
s
o
f
c
o
l
o
r
e
c
t
a
l
c
a
n
c
e
r
,
”
J
o
u
r
n
a
l
S
u
p
e
r
c
o
m
p
u
t
,
v
o
l
.
7
7
,
n
o
.
9
,
p
p
.
9
4
9
4
–
9
5
1
9
,
2
0
2
1
,
d
o
i
:
1
0
.
1
0
0
7
/
s
1
1
2
2
7
-
0
2
0
-
0
3
5
7
5
-
6.
[
2
5
]
M
.
T
.
I
sl
a
m,
S
.
T.
M
a
sh
f
u
,
A
.
F
a
i
sa
l
,
S
.
C
.
S
i
a
m
,
I
.
T.
N
a
h
e
e
n
a
n
d
R
.
K
h
a
n
,
“
D
e
e
p
l
e
a
r
n
i
n
g
-
b
a
s
e
d
g
l
a
u
c
o
ma
d
e
t
e
c
t
i
o
n
w
i
t
h
c
r
o
p
p
e
d
o
p
t
i
c
c
u
p
a
n
d
d
i
sc
a
n
d
b
l
o
o
d
v
e
sse
l
se
g
me
n
t
a
t
i
o
n
,
”
i
n
I
EEE
Ac
c
e
ss
,
v
o
l
.
1
0
,
p
p
.
2
8
2
8
-
2
8
4
1
,
2
0
2
2
,
d
o
i
:
1
0
.
1
1
0
9
/
A
C
C
ESS
.
2
0
2
1
.
3
1
3
9
1
6
0
.
[
2
6
]
R
.
H
.
P
a
r
a
d
i
sa,
A
.
B
u
st
a
mam
,
W
.
M
a
n
g
u
n
w
a
r
d
o
y
o
,
A
.
A
.
V
i
c
t
o
r
,
A
.
R
.
Y
u
d
a
n
t
h
a
,
a
n
d
P
.
A
n
k
i
,
“
D
e
e
p
f
e
a
t
u
r
e
v
e
c
t
o
r
s
c
o
n
c
a
t
e
n
a
t
i
o
n
f
o
r
e
y
e
d
i
se
a
se
d
e
t
e
c
t
i
o
n
u
si
n
g
f
u
n
d
u
s
i
m
a
g
e
,
”
E
l
e
c
t
r
o
n
i
c
s
,
v
o
l
.
1
1
,
n
o
.
1
,
2
0
2
2
,
d
o
i
:
1
0
.
3
3
9
0
/
e
l
e
c
t
r
o
n
i
c
s
1
1
0
1
0
0
2
.
[
2
7
]
D
.
N
a
g
p
a
l
,
N
.
A
l
s
u
b
a
i
e
,
B
.
O
.
S
o
u
f
i
e
n
e
,
M
.
S
.
A
l
q
a
h
t
a
n
i
,
M
.
A
b
b
a
s,
a
n
d
H
.
A
l
m
o
h
i
y
,
“
A
u
t
o
ma
t
i
c
d
e
t
e
c
t
i
o
n
o
f
d
i
a
b
e
t
i
c
h
y
p
e
r
t
e
n
si
v
e
r
e
t
i
n
o
p
a
t
h
y
i
n
F
u
n
d
u
s
i
mag
e
s
u
si
n
g
t
r
a
n
sf
e
r
Le
a
r
n
i
n
g
,
”
A
p
p
l
i
e
d
S
c
i
e
n
c
e
s
,
v
o
l
.
1
3
,
n
o
.
8
,
p
.
4
6
9
5
,
A
p
r
.
2
0
2
3
,
d
o
i
:
1
0
.
3
3
9
0
/
a
p
p
1
3
0
8
4
6
9
5
.
B
I
O
G
RAP
H
I
E
S O
F
AUTH
O
RS
H
e
lm
i
Im
a
d
u
d
d
i
n
is
c
u
rre
n
tl
y
a
lec
tu
re
r
a
t
th
e
F
a
c
u
lt
y
o
f
C
o
m
m
u
n
ica
ti
o
n
a
n
d
In
fo
rm
a
ti
c
s,
Un
i
v
e
rsitas
M
u
h
a
m
m
a
d
iy
a
h
S
u
ra
k
a
rta.
His
p
r
ima
ry
a
re
a
s
o
f
in
tere
st
in
re
se
a
rc
h
in
c
lu
d
e
n
a
tu
ra
l
lan
g
u
a
g
e
p
r
o
c
e
ss
in
g
,
ima
g
e
p
ro
c
e
ss
in
g
,
a
n
d
c
l
a
ss
ifi
c
a
ti
o
n
.
In
2
0
1
5
,
h
e
su
c
c
e
ss
fu
ll
y
c
o
m
p
l
e
ted
h
is
Ba
c
h
e
lo
r
o
f
S
c
ien
c
e
in
I
n
fo
rm
a
ti
c
s
En
g
in
e
e
rin
g
fr
o
m
Un
i
v
e
rsitas
M
u
h
a
m
m
a
d
iy
a
h
S
u
ra
k
a
rta.
S
u
b
s
e
q
u
e
n
tl
y
,
i
n
2
0
2
0
,
h
e
o
b
tain
e
d
h
is
M
a
ste
r
o
f
S
c
ien
c
e
in
In
fo
rm
a
ti
o
n
Tec
h
n
o
lo
g
y
fr
o
m
G
a
d
jah
M
a
d
a
Un
i
v
e
rsity
in
Y
o
g
y
a
k
a
rta.
He
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
h
e
l
m
i.
ima
d
u
d
d
i
n
@u
m
s.
a
c
.
id
.
Aliv
ia
Ra
h
m
a
S
a
k
in
a
is
a
n
in
fo
rm
a
ti
c
s
e
n
g
in
e
e
rin
g
st
u
d
e
n
t
a
t
M
u
h
a
m
m
a
d
iy
a
h
Un
iv
e
rsity
o
f
S
u
ra
k
a
rta.
S
h
e
h
a
s
re
se
a
rc
h
in
tere
sts
in
d
a
ta
e
n
g
i
n
e
e
rin
g
,
c
o
m
p
u
ter
v
isio
n
a
n
d
m
a
c
h
in
e
lea
rn
in
g
.
Re
c
e
n
tl
y
S
h
e
h
a
s
b
e
e
n
a
c
ti
v
e
ly
w
o
rk
i
n
g
a
s
a
d
a
ta
e
n
g
in
e
e
r
i
n
t
h
e
tele
c
o
m
m
u
n
ica
ti
o
n
s
n
e
two
rk
se
c
to
r.
S
h
e
c
a
n
b
e
c
o
n
tac
ted
at
e
m
a
il
:
a
li
v
iara
h
m
a
5
6
@g
m
a
il
.
c
o
m
.
Evaluation Warning : The document was created with Spire.PDF for Python.