I
nd
o
ne
s
ia
n J
o
urna
l o
f
E
lect
rica
l En
g
ineering
a
nd
Co
m
p
u
t
er
Science
Vo
l.
3
8
,
No
.
1
,
A
p
r
il 2
0
2
5
,
p
p
.
2
9
2
~
3
0
2
I
SS
N:
2
5
0
2
-
4
7
5
2
,
DOI
: 1
0
.
1
1
5
9
1
/i
j
ee
cs.v
3
8
.
i1
.
p
p
2
9
2
-
3
0
2
292
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
ee
cs.ia
esco
r
e.
co
m
TAL
O
S:
opti
m
i
za
tion o
f
the
CN
N
for the
det
ection
o
f
the
to
m
a
to lea
f
disea
s
es
Sh
rut
hi K
i
kk
er
i Su
bra
m
a
ny
a
,
Na
v
ee
n B
e
t
t
a
ha
lli,
Na
v
ee
n K
a
lena
ha
lli B
ho
g
a
nn
a
D
e
p
a
r
t
me
n
t
o
f
E
l
e
c
t
r
o
n
i
c
s
a
n
d
C
o
mm
u
n
i
c
a
t
i
o
n
E
n
g
i
n
e
e
r
i
n
g
,
B
G
S
I
n
st
i
t
u
t
e
o
f
T
e
c
h
n
o
l
o
g
y
,
A
d
i
c
h
u
n
c
h
a
n
a
g
i
r
i
U
n
i
v
e
r
si
t
y
,
K
a
r
n
a
t
a
k
a
,
I
n
d
i
a
Art
icle
I
nfo
AB
ST
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
J
u
n
5
,
2
0
2
4
R
ev
i
s
ed
Oct
7
,
2
0
2
4
A
cc
ep
ted
Oct
28
,
2
0
2
4
Early
d
e
tec
ti
o
n
o
f
p
la
n
t
d
ise
a
se
s
u
sin
g
c
o
n
v
o
lu
t
io
n
a
l
n
e
u
ra
l
n
e
tw
o
rk
(CNN
)is
c
ru
c
ial
f
o
r
m
a
x
i
m
izin
g
c
ro
p
y
ield
a
n
d
m
in
im
izin
g
e
c
o
n
o
m
ic
lo
ss
e
s.
M
a
n
u
a
l
i
n
sp
e
c
ti
o
n
,
t
h
e
f
re
q
u
e
n
t
tec
h
n
iq
u
e
,
is
in
e
f
f
icie
n
t
a
n
d
e
rro
r
p
r
o
n
e
.
W
h
il
e
CNN
’s
o
ff
e
r
p
o
ten
ti
a
l
f
o
r
a
c
c
u
ra
te
a
n
d
q
u
ick
d
ise
a
se
re
c
o
g
n
it
io
n
,
th
e
ir
p
e
rf
o
rm
a
n
c
e
is
h
ig
h
l
y
d
e
p
e
n
d
e
n
t
o
n
e
ff
e
c
ti
v
e
h
y
p
e
rp
a
ra
m
e
t
e
r
tu
n
i
n
g
.
T
h
is
p
ro
c
e
ss
is
ti
m
e
c
o
n
su
m
in
g
,
re
so
u
rc
e
in
ten
siv
e
,
a
n
d
n
e
e
d
s
s
ig
n
if
ica
n
t
e
x
p
e
rti
se
d
u
e
to
th
e
v
a
st
h
y
p
e
rp
a
ra
m
e
ter
sp
a
c
e
,
sin
c
e
it
c
a
n
b
e
h
a
rd
t
o
f
ig
u
re
o
u
t
w
h
ich
is
i
d
e
a
l
f
o
r
o
p
ti
m
a
l
p
e
r
f
o
rm
a
n
c
e
.
A
n
e
ffe
c
ti
v
e
o
p
ti
m
iza
ti
o
n
to
o
l
,
tu
n
a
b
le
a
u
to
m
a
ted
h
y
p
e
r
p
a
ra
m
e
ter
lea
rn
in
g
o
p
ti
m
iza
ti
o
n
sy
ste
m
(TAL
OS),
is
p
ro
p
o
se
d
,
w
h
ich
a
u
to
m
a
tes
th
e
tu
n
in
g
o
f
h
y
p
e
rp
a
ra
m
e
ters
b
y
s
y
ste
m
a
ti
c
a
ll
y
e
x
p
lo
rin
g
th
e
h
y
p
e
rp
a
ra
m
e
ter
sp
a
c
e
a
n
d
e
v
a
lu
a
tes
d
if
fe
re
n
t
c
o
m
b
in
a
ti
o
n
s
o
f
p
a
ra
m
e
t
e
rs
to
f
i
n
d
t
h
e
o
p
t
im
a
l
c
o
n
f
i
g
u
ra
ti
o
n
th
a
t
m
a
x
i
m
i
z
e
th
e
m
o
d
e
l’s
p
e
rf
o
r
m
a
n
c
e
.
T
h
e
p
e
rf
o
r
m
a
n
c
e
o
f
th
is
a
p
p
ro
a
c
h
is
re
c
o
g
n
iza
b
le
th
ro
u
g
h
it
s
e
x
p
lo
ra
ti
o
n
o
f
f
iv
e
d
if
fe
re
n
t
h
y
p
e
rp
a
ra
m
e
ter
s
a
c
ro
ss
a
se
a
rc
h
sp
a
c
e
o
f
3
2
c
o
m
b
in
a
ti
o
n
s,
y
ield
in
g
o
p
ti
m
a
l
p
a
ra
m
e
ters
b
y
th
e
se
c
o
n
d
r
o
u
n
d
.
Us
in
g
3
0
3
0
to
m
a
to
lea
f
ima
g
e
s
f
ro
m
a
b
e
n
c
h
m
a
r
k
d
a
ta
se
t,
t
h
e
m
o
d
e
l
a
c
h
iev
e
s
a
r
e
m
a
r
k
a
b
le
9
4
.
7
%
v
a
li
d
a
ti
o
n
a
c
c
u
ra
c
y
w
it
h
3
3
6
4
7
train
a
b
le
p
a
ra
m
e
ters
.
T
h
u
s,
a
u
to
m
a
ted
h
y
p
e
rp
a
ra
m
e
ter
tu
n
in
g
a
p
p
r
o
a
c
h
n
o
t
o
n
ly
o
p
ti
m
ize
s
m
o
d
e
l
p
e
rf
o
r
m
a
n
c
e
b
u
t
a
lso
re
d
u
c
e
s
m
a
n
u
a
l
e
ff
o
rt
a
n
d
re
so
u
rc
e
re
q
u
irem
e
n
ts,
p
a
v
in
g
th
e
w
a
y
f
o
r
m
o
re
e
ff
e
c
ti
v
e
a
n
d
sc
a
lab
le
so
lu
ti
o
n
s
i
n
a
g
ricu
lt
u
ra
l
tec
h
n
o
lo
g
y
.
K
ey
w
o
r
d
s
:
C
NN
H
y
p
er
p
ar
am
e
ter
Op
ti
m
izatio
n
T
A
L
O
S
T
o
m
ato
leaf
d
is
ea
s
e
T
h
is i
s
a
n
o
p
e
n
a
c
c
e
ss
a
rticle
u
n
d
e
r th
e
CC B
Y
-
SA
li
c
e
n
se
.
C
o
r
r
e
s
p
o
nd
ing
A
uth
o
r
:
Nav
ee
n
B
etta
h
alli
Dep
ar
t
m
en
t o
f
E
lectr
o
n
ics a
n
d
C
o
m
m
u
n
icat
io
n
E
n
g
i
n
ee
r
in
g
,
B
GS I
n
s
tit
u
te
o
f
T
ec
h
n
o
lo
g
y
A
d
ic
h
u
n
c
h
an
a
g
ir
i U
n
i
v
er
s
it
y
B
.
G.
Nag
ar
a,
Kar
n
atak
a,
I
n
d
ia
E
m
ail:
n
av
ee
n
b
@
b
g
s
it.a
c.
i
n
1.
I
NT
RO
D
UCT
I
O
N
J
ay
Ko
r
d
ich
’
s
s
ta
te
m
e
n
t,
“A
ll
lif
e
on
ea
r
th
e
m
an
a
tes
f
r
o
m
th
e
g
r
ee
n
of
th
e
p
lan
t,”
h
i
g
h
li
g
h
t
s
th
e
cr
itical
r
o
le
of
p
lan
ts
as
th
e
p
r
im
ar
y
s
o
u
r
ce
of
o
x
y
g
e
n
p
r
o
d
u
ctio
n
,
s
u
p
p
o
r
tin
g
ae
r
o
b
ic
lif
e
f
o
r
m
s
’
s
u
r
v
i
v
al.
A
d
d
itio
n
al
l
y
,
t
h
e
y
p
la
y
a
cr
u
cial
r
o
le
in
m
ain
ta
in
i
n
g
ec
o
lo
g
ical
b
alan
ce
,
r
e
g
u
la
tin
g
t
h
e
ea
r
th
’
s
cli
m
ate,
a
n
d
s
u
p
p
o
r
tin
g
o
u
r
p
lan
et
’
s
in
tr
ic
ate
w
eb
of
lif
e.
D
u
e
to
t
h
e
en
d
em
ic
d
is
ea
s
e
s
i
n
p
lan
t
s
,
n
u
m
er
o
u
s
p
lan
t
s
ar
e
on
th
e
v
er
g
e
of
b
ec
o
m
in
g
e
x
ti
n
c
t
[
1
]
,
[
2
]
.
T
h
e
ch
allen
g
e
o
f
a
cc
u
r
atel
y
id
en
ti
f
y
in
g
t
h
e
p
lan
t
d
is
ea
s
es
i
s
cr
u
cia
l
d
u
e
to
th
e
s
ig
n
i
f
ica
n
t
i
m
p
ac
t
o
f
th
e
s
e
d
is
ea
s
es
ca
n
h
a
v
e
o
n
ag
r
icu
l
tu
r
e.
T
r
ad
itio
n
al
m
et
h
o
d
s
,
w
h
ile
u
s
e
f
u
l
i
n
s
p
ec
if
ic
co
n
t
e
x
ts
,
ar
e
o
f
te
n
li
m
ited
b
y
t
h
eir
m
a
n
u
al,
lab
o
r
-
in
te
n
s
i
v
e
n
a
tu
r
e,
an
d
d
ep
en
d
en
c
y
o
n
ex
p
er
t
k
n
o
w
led
g
e.
Vis
u
al
i
n
s
p
ec
tio
n
s
ar
e
s
u
b
j
ec
tiv
e
a
n
d
in
co
n
s
i
s
t
en
t,
m
icr
o
s
co
p
y
r
eq
u
ir
es
s
p
ec
ialized
s
k
i
lls
a
n
d
is
ti
m
e
co
n
s
u
m
i
n
g
,
an
d
cu
lt
u
r
in
g
is
n
o
t
ap
p
licab
le
to
all
p
a
th
o
g
e
n
s
.
Mo
r
eo
v
er
,
tr
ad
itio
n
al
m
ac
h
i
n
e
lear
n
in
g
(
ML
)
ap
p
r
o
ac
h
es,
w
h
ile
au
to
m
ated
,
o
f
ten
f
a
il
to
h
an
d
le
th
e
co
m
p
lex
it
y
a
n
d
v
ar
iab
ilit
y
o
f
d
is
ea
s
e
s
y
m
p
to
m
s
ef
f
ec
tiv
e
l
y
[
3
]
-
[
5
]
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
d
o
n
esia
n
J
E
lec
E
n
g
&
C
o
m
p
Sci
I
SS
N:
2
5
0
2
-
4752
TALOS
:
o
p
timiz
a
tio
n
o
f th
e
C
N
N
fo
r
th
e
d
etec
tio
n
…
(
S
h
r
u
th
i Ki
kk
eri S
u
b
r
a
ma
n
ya
)
293
Dee
p
lear
n
in
g
(
DL
)
is
e
m
er
g
i
n
g
as
a
p
o
w
er
f
u
l
tech
n
o
lo
g
y
,
esp
ec
iall
y
co
n
v
o
lu
tio
n
al
n
eu
r
a
l
n
et
w
o
r
k
s
(
C
NN’
s
)
,
w
h
ic
h
h
a
v
e
b
ee
n
in
cr
ea
s
in
g
l
y
u
tili
ze
d
.
C
N
N
ca
n
au
to
m
at
icall
y
e
x
tr
ac
t
r
elev
a
n
t
f
ea
t
u
r
es
f
r
o
m
lar
g
e
an
d
co
m
p
lex
d
ataset
s
,
eli
m
i
n
ati
n
g
th
e
n
ee
d
f
o
r
m
a
n
u
al
f
ea
t
u
r
e
ex
tr
ac
tio
n
(
FE)
[
6
]
-
[
8
]
.
Ho
w
e
v
er
,
th
e
ap
p
licatio
n
o
f
C
N
Ns
i
s
n
o
t
with
o
u
t
c
h
alle
n
g
es.
T
h
e
p
er
f
o
r
m
an
ce
o
f
t
h
e
DL
m
o
d
el
i
s
cr
u
cial
f
o
r
ac
h
ie
v
in
g
o
p
tim
a
l
p
er
f
o
r
m
an
ce
w
h
ic
h
m
ai
n
l
y
d
ep
en
d
s
o
n
t
h
e
q
u
a
n
t
it
y
an
d
q
u
ali
t
y
o
f
t
h
e
i
m
ag
e
s
in
th
e
d
ataset,
t
h
e
r
o
b
u
s
t
d
esig
n
o
f
th
e
m
o
d
els,
an
d
th
e
o
p
ti
m
izat
io
n
o
f
t
h
e
h
y
p
er
p
ar
a
m
eter
s
.
Firstl
y
,
a
s
ig
n
i
f
ican
t
ch
alle
n
g
e
i
n
tr
ain
i
n
g
th
e
D
L
m
o
d
els
is
g
ettin
g
a
h
ig
h
-
q
u
alit
y
d
ata
s
et
th
at
ar
e
lar
g
e,
d
iv
er
s
e,
ac
cu
r
ate,
an
d
w
el
l
p
r
e
-
p
r
o
ce
s
s
ed
[
9
]
,
w
it
h
b
ala
n
ce
d
class
es
to
p
r
ev
en
t
b
ias.
T
h
is
p
r
o
ce
s
s
is
cr
u
cia
l
b
u
t
co
m
p
u
t
atio
n
all
y
ex
p
e
n
s
i
v
e
an
d
it
m
a
y
ca
u
s
e
o
v
er
f
itti
n
g
[
1
0
]
w
h
er
e
t
h
e
m
o
d
el
p
er
f
o
r
m
s
b
etter
o
n
t
h
e
tr
ai
n
in
g
d
ata
co
m
p
ar
ed
to
v
alid
atio
n
/
test
d
ata.
Ma
th
e
m
at
icall
y
,
t
h
e
o
v
er
f
it
tin
g
ca
n
b
e
r
ep
r
esen
ted
as f
o
llo
w
s
:
E
_
tr
ain
<<
E
_
test
(
1
)
w
h
er
e
E
_
tr
ain
is
t
h
e
tr
ain
in
g
d
ataset
er
r
o
r
,
an
d
E
_
test
is
th
e
e
r
r
o
r
o
n
test
o
r
v
alid
atio
n
d
atas
ets.
Seco
n
d
,
th
e
d
esig
n
in
g
a
r
o
b
u
s
t
m
o
d
el
[
1
1
]
in
v
o
lv
e
s
ch
o
o
s
in
g
a
n
ap
p
r
o
p
r
iate
a
r
ch
itect
u
r
e,
co
n
f
i
g
u
r
in
g
t
h
e
la
y
er
s
ef
f
ec
ti
v
el
y
,
s
e
lectin
g
th
e
p
r
o
p
er
ac
tiv
atio
n
f
u
n
ctio
n
,
an
d
in
co
r
p
o
r
atin
g
r
eg
u
lar
izatio
n
to
i
m
p
r
o
v
e
s
tab
ilit
y
an
d
g
e
n
er
ali
za
tio
n
o
f
th
e
m
o
d
els.
L
a
s
tl
y
,
h
y
p
er
p
ar
a
m
eter
T
u
n
in
g
is
e
s
s
e
n
tial to
o
p
ti
m
ize
th
e
m
o
d
el
’
s
p
er
f
o
r
m
a
n
ce
w
h
ic
h
i
n
v
o
l
v
es
ad
j
u
s
t
m
en
t
o
f
t
h
e
p
ar
a
m
eter
s
s
u
c
h
as
b
atch
s
ize,
n
u
m
b
er
o
f
ep
o
ch
s
,
lear
n
in
g
r
ate,
an
d
t
h
e
ch
o
ice
o
f
o
p
ti
m
izer
[
1
2
]
.
T
r
ad
itio
n
al
m
et
h
o
d
s
lik
e
g
r
id
s
ea
r
ch
[
1
3
]
,
w
h
il
e
e
x
h
a
u
s
tiv
e
l
y
,
ar
e
co
m
p
u
tatio
n
all
y
ex
p
e
n
s
i
v
e
as
t
h
e
m
o
d
el
co
m
p
le
x
it
y
i
n
cr
ea
s
es;
r
an
d
o
m
s
ea
r
c
h
[
1
4
]
,
w
h
ile
m
o
r
e
e
f
f
icien
t
f
o
r
lar
g
e
d
ata
s
ets,
lac
k
s
t
h
e
ce
r
tain
t
y
o
f
f
in
d
i
n
g
t
h
e
b
est
co
n
f
ig
u
r
atio
n
.
A
ls
o
,
t
h
e
m
a
n
u
a
l
t
u
n
i
n
g
,
th
o
u
g
h
p
r
o
v
id
in
g
i
n
s
i
g
h
ts
,
is
s
u
b
j
ec
tiv
e,
ti
m
e
-
co
n
s
u
m
i
n
g
,
an
d
p
r
o
n
e
to
e
r
r
o
r
s
,
p
ar
ticu
lar
l
y
w
it
h
co
m
p
lex
m
o
d
el
s
[
1
5
]
.
T
h
er
ef
o
r
e,
f
in
d
in
g
a
n
ef
f
icie
n
t
a
n
d
ef
f
ec
tiv
e
tu
n
i
n
g
s
tr
ate
g
y
f
o
r
lar
g
e
an
d
in
tr
icate
m
o
d
el
s
r
e
m
ain
s
a
s
i
g
n
if
ica
n
t c
h
alle
n
g
e
in
m
o
d
el
o
p
ti
m
izat
io
n
[
1
6
]
.
R
ev
ie
w
in
g
th
e
r
ele
v
a
n
t
liter
at
u
r
e
h
elp
s
to
id
en
tify
m
aj
o
r
c
o
n
tr
ib
u
to
r
s
’
w
o
r
k
an
d
f
i
n
d
in
g
s
,
g
u
id
i
n
g
p
o
ten
tial
ad
v
an
ce
m
e
n
ts
i
n
th
e
f
ield
b
y
s
u
m
m
ar
izi
n
g
t
h
e
r
ec
en
t
p
r
o
g
r
ess
io
n
s
in
h
y
p
er
p
ar
am
eter
t
u
n
i
n
g
,
in
cl
u
d
in
g
alg
o
r
it
h
m
s
lik
e
g
r
id
s
ea
r
ch
,
r
an
d
o
m
s
ea
r
ch
,
an
d
B
ay
esia
n
o
p
ti
m
iz
atio
n
,
w
h
ich
ai
m
to
en
h
an
c
e
o
p
tim
izatio
n
e
f
f
icie
n
c
y
a
n
d
p
er
f
o
r
m
an
ce
.
T
h
e
w
o
r
k
p
r
o
p
o
s
ed
in
[
1
7
]
h
ig
h
lig
h
t
s
th
e
w
h
ale
o
p
ti
m
iza
t
io
n
alg
o
r
ith
m
(
W
O
A
)
,
to
o
p
t
i
m
ize
t
h
e
h
y
p
er
p
ar
a
m
eter
s
i
n
n
e
u
r
al
n
et
w
o
r
k
s
.
I
t
ac
h
iev
ed
a
n
o
tab
le
ac
cu
r
ac
y
o
f
8
0
.
6
0
%
an
d
8
9
.
8
5
%
o
n
r
e
u
ter
s
d
atasets
a
n
d
f
as
h
io
n
MN
I
S
T
,
r
esp
ec
tiv
ely
.
T
h
e
r
esear
c
h
p
r
o
p
o
s
ed
in
[
1
8
]
is
a
1
4
-
la
y
er
ed
d
ee
p
C
NN
(
1
4
-
DC
NN)
to
id
e
n
ti
f
y
d
is
ea
s
es
f
r
o
m
a
d
ataset
o
f
1
4
7
,
5
0
0
i
m
a
g
es
o
f
5
8
p
lan
t
leaf
cla
s
s
es.
T
h
e
m
o
d
el
is
tr
ain
ed
f
o
r
1
,
0
0
0
ep
o
ch
s
an
d
th
en
o
p
ti
m
ized
u
s
i
n
g
r
an
d
o
m
s
ea
r
ch
w
i
th
co
ar
s
e
-
to
-
f
i
n
e
h
y
p
er
p
ar
a
m
eter
s
s
ea
r
ch
i
n
g
.
T
h
e
s
tated
D
C
N
N
m
o
d
el
p
r
o
v
id
es
a
9
9
.
9
6
5
5
%h
i
g
h
ac
c
u
r
ac
y
,
9
9
.
7
9
6
6
%
r
ec
all,
9
9
.
7
9
9
9
%
w
ei
g
h
ted
av
er
ag
e
p
r
ec
is
io
n
,
a
n
d
9
9
.
7
9
6
8
F1
-
s
co
r
e.
T
h
e
m
et
h
o
d
o
lo
g
y
in
[
1
9
]
em
p
lo
y
s
a
W
OA
w
it
h
h
y
b
r
id
p
r
in
cip
al
co
m
p
o
n
e
n
t
an
al
y
s
i
s
(
P
C
A
)
to
id
en
tify
d
is
ea
s
es
f
r
o
m
a
d
ataset
o
f
1
8
1
5
9
o
f
1
0
class
es
o
f
to
m
a
to
leav
es
f
r
o
m
t
h
e
P
lan
tVilla
g
e
d
ataset.
Gr
id
s
ea
r
ch
is
ad
o
p
te
d
to
tu
n
e
an
d
f
in
d
th
e
o
p
ti
m
a
l
h
y
p
er
p
ar
a
m
eter
s
,
w
h
ic
h
en
h
a
n
ce
s
m
o
d
el
p
er
f
o
r
m
an
ce
.
T
h
e
m
o
d
el
p
r
o
v
id
es
9
9
%
o
f
tr
ain
in
g
ac
c
u
r
ac
y
a
n
d
8
6
%
test
i
n
g
ac
c
u
r
ac
y
at
th
e
1
5
th
e
p
o
ch
.
P
an
d
ian
et
a
l.
[
2
0
]
th
e
DC
NN
m
o
d
el
w
it
h
f
i
v
e
co
n
v
o
lu
tio
n
al
la
y
er
s
is
tr
ain
ed
w
it
h
th
e
au
g
m
e
n
ted
d
ataset
o
f
th
e
P
lan
tVilla
g
e
d
ataset,
w
h
ich
o
r
ig
i
n
all
y
co
n
tai
n
s
i
m
ag
e
s
o
f
5
5
4
4
8
w
it
h
3
9
d
if
f
er
e
n
t
class
es
i
s
s
u
b
j
ec
ted
to
d
ee
p
co
n
v
o
lu
tio
n
a
l
g
e
n
er
ativ
e
ad
v
er
s
ar
ial
n
et
w
o
r
k
s
(
DC
G
A
N)
au
g
m
e
n
tatio
n
tech
n
iq
u
es
r
es
u
lti
n
g
in
a
s
et
o
f
2
4
0
,
0
0
0
im
ag
e
s
.
H
y
p
er
p
ar
am
eter
s
ar
e
s
et
u
s
i
n
g
a
r
an
d
o
m
s
ea
r
ch
,
r
esu
lti
n
g
i
n
an
ac
cu
r
ac
y
o
f
9
8
.
4
1
% o
n
th
e
test
d
ataset.
T
h
e
p
r
o
p
o
s
ed
w
o
r
k
i
n
[
2
1
]
is
a
co
n
tex
tu
al
m
a
s
k
au
to
-
e
n
co
d
er
o
p
tim
ized
w
it
h
a
d
y
n
a
m
ic
d
if
f
er
e
n
tia
l
an
n
ea
led
o
p
ti
m
izatio
n
al
g
o
r
ith
m
(
P
DI
-
C
M
A
E
-
DD
A
O
A
)
f
o
r
t
h
e
ea
r
l
y
d
etec
tio
n
o
f
p
lan
t
d
is
ea
s
e
s
.
P
DI
-
C
M
A
E
-
DD
A
O
A
ac
h
iev
e
s
h
ig
h
er
ac
cu
r
ac
y
2
3
.
3
4
%,
3
4
.
3
3
%,
an
d
3
2
.
0
7
%,
F1
-
s
co
r
e
4
6
.
6
7
%,
5
7
.
5
6
%,
s
en
s
iti
v
it
y
3
6
.
6
7
%,
3
6
.
3
3
%,
an
d
2
3
.
2
1
%,
an
d
4
3
.
2
1
%,
an
d
s
p
ec
i
f
icit
y
5
6
.
6
7
%,
6
7
.
5
6
%,
an
d
2
3
.
2
1
%
co
m
p
ar
ed
to
th
ese
e
x
is
tin
g
m
o
d
el
s
s
u
ch
a
s
P
DI
-
DE
NN,
P
DI
-
C
A
E
-
C
NN,
a
n
d
P
DI
-
EN
-
C
NN,
r
esp
ec
tiv
el
y
.
Du
d
i
an
d
R
aj
esh
[
2
2
]
p
r
esen
t
a
m
et
h
o
d
th
at
u
s
es
th
e
s
h
a
r
k
s
m
ll
-
b
ased
-
W
O
A
(
SS
-
W
O
A
)
to
o
p
tim
ize
th
e
C
NN
’
s
ac
t
iv
at
io
n
f
u
n
ct
io
n
f
o
r
m
a
x
i
m
u
m
c
lass
if
ica
tio
n
ac
c
u
r
ac
y
.
C
o
m
p
ar
ed
to
NB
an
d
S
VM
,
th
e
ac
c
u
r
ac
y
of
th
e
s
u
p
p
lied
SS
-
W
O
A
-
C
NN
is
7
.
1
4
%
an
d
5
.
6
3
%
h
i
g
h
er
r
es
p
ec
tiv
el
y
.
Hali
m
et
a
l.
[
2
3
]
co
n
s
id
er
ed
th
e
C
NN
ar
ch
itect
u
r
es
s
u
c
h
as
Xce
p
tio
n
an
d
De
n
s
eNe
t
w
it
h
t
h
e
A
i
Sar
a
tu
n
i
n
g
al
g
o
r
ith
m
,
r
esu
lt
in
g
i
n
a
2
3
%
i
m
p
r
o
v
e
m
en
t
i
n
ac
cu
r
ac
y
co
m
p
ar
ed
to
t
y
p
ical
tu
n
i
n
g
tec
h
n
iq
u
es.
T
h
e
m
o
d
els
o
n
th
e
P
lan
tVillag
e
an
d
P
lan
tDo
c
d
atasets
ar
e
ev
alu
at
ed
in
th
e
s
t
u
d
y
;
De
n
s
eNe
t1
2
1
an
d
Xce
p
tio
n
o
b
tain
ed
ac
cu
r
ac
y
o
f
8
9
.
6
0
%
an
d
8
5
.
9
4
%
o
n
P
lan
tVillag
e
an
d
8
1
.
5
1
%
o
n
P
lan
tDo
c,
r
es
p
ec
tiv
el
y
,
w
it
h
o
u
t
h
y
p
er
p
ar
am
eter
ad
j
u
s
t
m
e
n
t
.
A
cc
u
r
ac
y
i
n
cr
ea
s
ed
to
9
4
.
7
5
%
an
d
9
1
.
0
3
%
o
n
P
lan
tVilla
g
e
an
d
8
4
.
8
4
%
an
d
8
7
.
6
6
%
o
n
P
lan
tDo
c
w
it
h
A
i
Sar
a
t
w
ea
k
in
g
.
A
k
k
u
ş
et
a
l
.
[
2
4
]
ev
alu
ates
t
h
e
e
f
f
ec
ti
v
en
ess
o
f
t
w
o
C
N
N
m
o
d
els,
R
es
Net1
8
an
d
Alex
Ne
t
in
d
etec
tin
g
th
e
s
e
v
er
it
y
o
f
s
ar
io
p
s
is
leaf
s
p
o
t
d
is
ea
s
e
in
g
r
ap
e
leav
es
w
h
ich
ar
e
g
ath
er
ed
f
r
o
m
t
h
e
P
lan
tVilla
g
e
d
ataset.
T
h
e
s
t
u
d
y
ex
p
lo
r
es
t
h
e
i
m
p
ac
t
o
f
t
u
n
i
n
g
h
y
p
er
p
ar
a
m
eter
s
,
in
c
lu
d
in
g
ep
o
ch
s
,
d
ata
au
g
m
e
n
tatio
n
,
an
d
m
in
i
-
b
atc
h
s
ize.
R
esNet1
8
r
ea
ch
ed
an
ac
cu
r
ac
y
o
f
8
7
.
6
%
w
ith
a
m
i
n
i
-
b
atch
s
ize
o
f
6
4
,
1
0
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N:
2
5
0
2
-
4752
I
n
d
o
n
esia
n
J
E
lec
E
n
g
&
C
o
m
p
Sci,
Vo
l.
3
8
,
No
.
1
,
A
p
r
il 2
0
2
5
: 2
9
2
-
302
294
ep
o
ch
s
,
an
d
w
it
h
d
ata
au
g
m
e
n
tatio
n
.
A
le
x
Net
ac
h
iev
ed
t
h
e
h
ig
h
est
cla
s
s
i
f
icatio
n
ac
cu
r
ac
y
o
f
9
0
.
3
1
%
w
i
th
a
m
i
n
i
-
b
atch
s
ize
o
f
3
2
,
an
d
5
0
ep
o
ch
s
,
an
d
w
ith
o
u
t d
ata
au
g
m
en
tatio
n
.
A
s
u
cc
ess
f
u
l
i
m
p
le
m
e
n
tatio
n
o
f
D
L
f
o
r
t
h
e
id
en
tific
atio
n
o
f
p
lan
t
d
is
ea
s
es
d
ep
en
d
s
o
n
s
e
v
er
a
l
f
ac
to
r
s
,
in
cl
u
d
in
g
d
ataset
’
s
q
u
alit
y
,
th
e
ar
ch
i
tectu
r
e
o
f
t
h
e
m
o
d
el,
an
d
th
e
o
p
ti
m
izatio
n
o
f
h
y
p
er
p
ar
a
m
et
er
s
.
Desp
ite
u
s
in
g
tec
h
n
iq
u
es
s
u
c
h
as
W
OA
,
P
C
A
,
FO
A
[
2
5
]
an
d
r
an
d
o
m
s
ea
r
c
h
,
ch
alle
n
g
es
r
e
m
ai
n
in
ad
d
r
ess
i
n
g
th
ese
in
ter
d
ep
en
d
e
n
cies
e
f
f
i
cien
tl
y
.
A
h
o
lis
t
ic
ap
p
r
o
ac
h
in
co
r
p
o
r
atin
g
r
o
b
u
s
t
d
ata
s
et
,
ad
v
an
ce
d
m
o
d
e
l
ar
ch
itect
u
r
es,
an
d
e
f
f
icie
n
t
h
y
p
er
p
ar
am
eter
o
p
ti
m
izat
io
n
i
s
ess
e
n
tial.
T
h
e
p
r
o
p
o
s
e
d
m
e
t
h
o
d
t
a
c
k
l
e
s
t
h
e
s
e
c
h
a
l
l
e
n
g
e
s
b
y
c
o
n
s
i
d
e
r
i
n
g
t
h
e
f
o
l
l
o
w
i
n
g
s
e
q
u
e
n
t
i
a
l
s
t
e
p
s
:
T
h
is
s
tu
d
y
p
r
o
p
o
s
es
a
d
is
ea
s
e
id
en
ti
f
icatio
n
s
y
s
te
m
f
o
r
to
m
ato
p
lan
ts
u
tili
zi
n
g
a
cu
s
to
m
ized
C
N
N
ar
ch
itect
u
r
e
d
esig
n
ed
to
p
r
o
v
id
e
a
r
ed
u
ce
d
p
a
r
am
eter
.
T
h
e
t
r
a
i
n
i
n
g
o
f
a
C
N
N
m
o
d
e
l
i
s
p
e
r
f
o
r
m
e
d
w
i
t
h
a
l
i
m
i
t
e
d
d
a
t
a
s
e
t
o
f
t
o
m
a
t
o
l
e
a
f
i
m
a
g
e
s
.
T
h
e
s
tu
d
y
e
m
p
lo
y
s
t
u
n
ab
le
au
to
m
ated
h
y
p
er
p
ar
a
m
eter
lear
n
in
g
o
p
ti
m
izatio
n
s
y
s
te
m
(
T
A
L
OS)
f
o
r
h
y
p
er
p
ar
a
m
eter
t
u
n
in
g
,
r
es
u
lt
i
n
g
in
ef
f
ic
ien
tl
y
o
p
ti
m
izi
n
g
m
o
d
el
p
er
f
o
r
m
a
n
ce
w
h
ile
m
i
n
i
m
izi
n
g
m
a
n
u
al
in
ter
v
e
n
tio
n
an
d
co
m
p
u
tatio
n
a
l r
eso
u
r
ce
s
,
th
u
s
m
ee
ti
n
g
its
p
r
i
m
ar
y
o
b
j
ec
tiv
e.
T
h
e
r
em
ai
n
i
n
g
s
ec
tio
n
of
t
h
e
r
esear
ch
w
o
r
k
is
o
r
g
a
n
ize
d
ac
co
r
d
in
g
to
th
e
p
r
ec
ed
in
g
s
tr
u
ctu
r
e.
Sectio
n
2
p
r
esen
ts
a
p
r
o
p
o
s
ed
m
o
d
el
f
o
r
a
class
i
f
icatio
n
s
y
s
te
m
b
ased
o
n
C
NN
a
n
d
T
AL
O
S
o
p
ti
m
izatio
n
.
Sectio
n
3
s
h
o
w
s
t
h
e
r
esu
lts
an
d
d
is
cu
s
s
io
n
.
Sectio
n
4
c
o
n
c
l
u
d
e
t
h
e
w
o
r
k
.
2.
M
E
T
H
O
D
A
c
u
s
to
m
ized
C
NN
i
s
u
s
ed
w
it
h
a
h
y
p
er
p
ar
a
m
eter
o
p
ti
m
i
za
tio
n
to
o
l
to
id
en
tify
to
m
a
to
p
lan
t
lea
f
d
is
ea
s
es
u
s
i
n
g
a
m
i
n
i
m
a
l
n
u
m
b
er
o
f
i
m
a
g
e
s
.
T
A
L
O
S
is
a
h
ier
ar
ch
ical
o
p
ti
m
i
ze
r
th
at
tu
n
es
t
h
e
h
y
p
er
p
ar
a
m
eter
s
w
it
h
o
u
t
m
an
u
all
y
te
s
ti
n
g
each
v
a
lu
e.
A
n
i
llu
s
tr
atio
n
of
th
e
m
o
d
el
f
o
r
ca
teg
o
r
izin
g
to
m
ato
leaf
d
is
e
ase
s
is
s
h
o
w
n
in
Fig
u
r
e
1
.
T
h
e
m
e
th
o
d
o
lo
g
y
i
s
p
er
f
o
r
m
ed
in
t
h
e
f
o
llo
w
i
n
g
s
tep
s
:
a.
Data
s
et
co
llectio
n
a
n
d
p
r
e
-
p
r
o
ce
s
s
in
g
b.
T
A
L
O
S set
u
p
c.
Def
i
n
e
th
e
C
NN
ar
ch
itect
u
r
e
d.
T
A
L
O
S in
te
g
r
atio
n
e.
T
A
L
O
S e
x
p
er
i
m
e
n
t
f.
Selectin
g
o
p
ti
m
a
l p
ar
a
m
eter
s
Fig
u
r
e
1
.
T
A
L
OS
-
C
NN
b
ased
to
m
a
to
leaf
d
is
e
as
e
clas
s
if
icat
io
n
s
y
s
te
m
2
.
1
.
Da
t
a
c
o
llect
io
n
a
nd
pre
-
pro
ce
s
s
ing
T
h
e
to
m
ato
lea
v
es
d
ataset
e
m
p
lo
y
ed
i
n
t
h
is
s
t
u
d
y
i
s
o
b
tain
e
d
f
r
o
m
t
h
e
P
lan
tVi
llag
e
d
ata
s
e
t.
T
h
e
late
b
lig
h
t,
y
ello
w
lea
f
c
u
r
l,
an
d
h
ea
lt
h
y
leaf
o
f
t
h
e
to
m
a
to
ar
e
co
n
s
id
er
ed
f
o
r
ex
p
er
i
m
e
n
ta
tio
n
,
as
s
h
o
w
n
i
n
Fig
u
r
e
2
.
E
ac
h
cla
s
s
co
n
tai
n
s
an
i
m
a
g
e
o
f
9
0
0
in
t
h
e
tr
ai
n
ed
s
et
an
d
1
1
0
in
th
e
test
i
n
g
s
et,
b
alan
ce
d
a
n
d
in
d
icati
n
g
a
n
ev
e
n
d
is
tr
i
b
u
ti
o
n
o
f
attr
ib
u
te
s
ac
r
o
s
s
all
c
ateg
o
r
ies
as
it
p
r
ev
e
n
ts
a
n
y
s
in
g
le
clas
s
f
r
o
m
d
o
m
i
n
ati
n
g
.
T
h
e
d
ataset’
s
i
m
ag
es
w
er
e
p
r
e
-
p
r
o
ce
s
s
ed
s
o
t
h
e
s
u
g
g
e
s
ted
m
o
d
el
co
u
ld
e
x
t
r
ac
t
th
eir
n
ec
e
s
s
ar
y
attr
ib
u
tes.
T
h
e
i
m
a
g
es
w
er
e
n
o
r
m
al
ized
an
d
r
esized
to
2
5
6
×
2
5
6
p
ix
els in
th
e
f
ir
s
t
s
tag
e.
Fig
u
r
e
2
.
Sa
m
p
le
leaf
i
m
a
g
es
o
f
th
e
to
m
ato
leav
es
f
r
o
m
t
h
e
t
r
ain
in
g
d
ataset
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
d
o
n
esia
n
J
E
lec
E
n
g
&
C
o
m
p
Sci
I
SS
N:
2
5
0
2
-
4752
TALOS
:
o
p
timiz
a
tio
n
o
f th
e
C
N
N
fo
r
th
e
d
etec
tio
n
…
(
S
h
r
u
th
i Ki
kk
eri S
u
b
r
a
ma
n
ya
)
295
2
.
2
.
Def
ini
ng
t
he
CNN
a
rc
hite
ct
ure
T
h
is
s
tu
d
y
in
co
r
p
o
r
ates
a
C
NN
as
a
d
is
ea
s
e
d
etec
tio
n
ap
p
r
o
ac
h
.
T
h
e
C
NN
ar
ch
itectu
r
e
co
m
p
r
is
e
s
:
th
e
FE
s
tag
e
a
n
d
clas
s
i
f
icati
o
n
s
ta
g
e.
T
h
e
p
r
o
p
o
s
ed
C
NN
ar
ch
itect
u
r
e
co
n
s
i
s
ts
o
f
la
y
er
s
s
u
c
h
as
b
atc
h
n
o
r
m
aliza
t
io
n
,
C
o
n
v
2
D,
Ma
x
P
o
o
lin
g
,
an
d
d
r
o
p
o
u
t
la
y
er
s
in
th
e
FE
p
ar
t,
f
o
llo
w
ed
b
y
d
en
s
e
an
d
d
r
o
p
o
u
t
la
y
er
s
in
t
h
e
clas
s
if
icatio
n
p
ar
t.
T
h
e
cu
s
to
m
ized
C
NN
ar
ch
itectu
r
e
is
s
h
o
w
n
i
n
Fig
u
r
e
3
,
w
it
h
th
e
co
r
r
esp
o
n
d
in
g
p
ar
a
m
eter
s
co
r
r
elate
d
to
th
e
la
y
er
s
d
is
p
la
y
ed
i
n
T
ab
le
1
.
T
h
e
ar
ch
itectu
r
e
in
t
eg
r
ates
f
o
u
r
b
atch
n
o
r
m
aliza
t
io
n
la
y
er
s
an
d
f
o
u
r
co
n
v
o
lu
tio
n
la
y
er
s
w
it
h
a
s
tr
id
e
o
f
(
1
,
1
)
,
p
r
eser
v
in
g
f
i
n
e
s
p
a
tial f
ea
t
u
r
es c
r
u
c
ial
f
o
r
ef
f
ec
ti
v
e
FE.
P
o
o
lin
g
la
y
er
s
d
ec
r
ea
s
e
o
u
tp
u
t
d
i
m
e
n
s
i
o
n
alit
y
,
r
ed
u
c
in
g
co
m
p
u
tatio
n
al
co
m
p
le
x
it
y
f
o
r
s
u
b
s
eq
u
en
t
la
y
er
s
;
e
m
p
lo
y
i
n
g
m
ax
-
p
o
o
lin
g
w
it
h
a
2
×2
p
o
o
l
s
ize
o
p
ti
m
ize
s
th
i
s
p
r
o
ce
s
s
.
Desp
ite
th
e
d
e
n
s
e
la
y
er
’
s
s
i
g
n
i
f
ican
t p
ar
a
m
eter
c
o
u
n
t (
9
2
4
8
)
,
its
r
o
le
in
ca
p
tu
r
i
n
g
co
m
p
lex
f
ea
t
u
r
e
r
elatio
n
s
h
i
p
s
r
e
m
ai
n
s
p
i
v
o
tal.
Ho
w
e
v
er
,
th
i
s
p
ar
a
m
eter
co
u
n
t
is
co
m
p
ar
ativ
e
l
y
m
o
d
est.
T
h
e
So
f
tMa
x
ac
tiv
at
io
n
f
u
n
ct
io
n
,
ap
p
lied
to
th
e
d
en
s
e
la
y
er
’
s
o
u
tp
u
t,
tr
an
s
f
o
r
m
s
it
in
to
a
p
r
o
b
ab
ilit
y
d
is
tr
ib
u
tio
n
.
T
h
is
ar
ch
itectu
r
e
b
alan
ce
s
FE,
r
eg
u
lar
izatio
n
,
an
d
p
ar
a
m
eter
ef
f
icien
c
y
,
o
f
f
er
in
g
r
o
b
u
s
t
p
er
f
o
r
m
a
n
ce
in
to
m
ato
leaf
d
is
ea
s
e
d
etec
tio
n
w
h
il
e
m
iti
g
ati
n
g
co
m
p
u
ta
tio
n
al
o
v
er
h
ea
d
.
T
h
e
co
n
f
ig
u
r
atio
n
o
f
a
n
eu
r
al
n
et
w
o
r
k
m
o
d
el
is
d
ep
ict
ed
in
T
ab
le
2
.
Fig
u
r
e
3
.
C
NN
m
o
d
el
a
r
ch
it
ec
tu
r
e
f
o
r
t
o
m
ato
lea
f
d
is
ea
s
e
r
e
co
g
n
itio
n
T
ab
le
1
.
Key
p
ar
a
m
eter
s
f
o
r
C
NN’
s
m
o
d
el
S
l
.
n
o
.
F
a
c
t
o
r
s
D
e
f
i
n
e
d
v
a
l
u
e
1
F
i
l
t
e
r
si
z
e
3
×
3
2
N
u
mb
e
r
o
f
f
i
l
t
e
r
s
32
3
P
o
o
l
i
n
g
M
a
x
P
o
o
l
i
n
g
w
i
t
h
a
2
×
2
p
o
o
l
i
n
g
w
i
n
d
o
w
a
n
d
st
r
i
d
e
s o
f
1
×
1
4
A
c
t
i
v
a
t
i
o
n
f
u
n
c
t
i
o
n
R
e
L
U
(
r
e
c
t
i
f
i
e
d
l
i
n
e
a
r
u
n
i
t
)
a
n
d
EL
U
(
e
x
p
o
n
e
n
t
i
a
l
l
i
n
e
a
r
u
n
i
t
)
5
L
o
ss fu
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
T
ab
le
2
.
Net
w
o
r
k
s
u
m
m
ar
y
S
l
.
n
o
.
L
a
y
e
r
s
A
c
t
i
v
a
t
i
o
n
f
u
n
c
t
i
o
n
s
T
r
a
i
n
a
b
l
e
p
a
r
a
me
t
e
r
s
1.
b
a
t
c
h
n
o
r
m
a
l
i
z
a
t
i
o
n
2
5
6
*
2
5
6
*
3
12
c
o
n
v
2
d
_
1
2
5
6
,
2
5
6
,
3
2
8
9
6
max
_
p
o
o
l
i
n
g
2
d
2
5
5
*
2
5
5
*
3
0
2.
b
a
t
c
h
n
o
r
mal
i
z
a
t
i
o
n
2
5
5
*
2
5
5
*
3
1
2
8
c
o
n
v
2
d
_
2
2
5
5
*
2
5
5
*
3
9
2
4
8
max
_
p
o
o
l
i
n
g
2
d
_
1
2
5
4
*
2
5
4
*
3
0
3.
b
a
t
c
h
n
o
r
mal
i
z
a
t
i
o
n
2
5
4
*
2
5
4
*
3
1
2
8
c
o
n
v
2
d
_
3
2
5
4
*
2
5
4
*
3
9
2
4
8
max
_
p
o
o
l
i
n
g
2
d
_
2
2
5
3
*
2
5
3
*
3
0
4.
b
a
t
c
h
n
o
r
mal
i
z
a
t
i
o
n
2
5
3
*
2
5
3
*
3
1
2
8
c
o
n
v
2
d
_
4
2
5
3
*
2
5
3
*
3
9
2
4
8
max
_
p
o
o
l
i
n
g
2
d
_
3
2
5
2
*
2
5
2
*
3
0
FC
1
2
8
4
2
2
4
D
e
n
se
1
2
8
3
8
7
T
o
t
a
l
p
a
r
a
m
e
t
e
r
s
:
3
3
,
6
4
7
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N:
2
5
0
2
-
4752
I
n
d
o
n
esia
n
J
E
lec
E
n
g
&
C
o
m
p
Sci,
Vo
l.
3
8
,
No
.
1
,
A
p
r
il 2
0
2
5
: 2
9
2
-
302
296
2
.
3
.
T
AL
O
S
inte
g
ra
t
i
o
n
w
it
h
CNN’
s
Fin
e
-
t
u
n
in
g
a
C
NN
m
o
d
el
t
y
p
icall
y
i
n
v
o
l
v
es
m
a
n
u
all
y
s
e
t
tin
g
h
y
p
er
p
ar
a
m
eter
s
f
o
llo
w
e
d
b
y
g
r
id
o
r
r
an
d
o
m
s
ea
r
c
h
ap
p
r
o
ac
h
es
to
ex
a
m
i
n
e
alter
n
at
iv
e
co
n
f
i
g
u
r
atio
n
s
s
y
s
te
m
atica
ll
y
.
T
h
is
m
et
h
o
d
ca
n
b
e
co
m
p
u
tatio
n
all
y
d
e
m
a
n
d
i
n
g
,
esp
ec
iall
y
w
h
e
n
d
ea
lin
g
w
i
th
co
m
p
lex
m
o
d
els
an
d
en
o
r
m
o
u
s
d
atasets
[
2
6
]
.
Af
ter
d
ef
i
n
in
g
t
h
e
C
NN
m
o
d
el,
T
A
L
O
S,
a
P
y
t
h
o
n
m
o
d
u
le
d
es
ig
n
ed
to
o
p
tim
ize
h
y
p
er
p
ar
am
eter
s
,
is
an
ass
is
tan
t
t
h
at
ex
p
lo
r
es
d
if
f
er
e
n
t
h
y
p
er
p
ar
a
m
e
ter
co
m
b
i
n
ati
o
n
s
f
o
r
th
e
m
o
d
el
is
e
m
p
lo
y
e
d
.
Fo
r
th
e
r
an
g
e
o
f
p
er
m
i
s
s
ib
le
v
al
u
es
f
o
r
ea
c
h
h
y
p
er
p
ar
a
m
eter
,
T
AL
O
S
tes
t
s
t
h
e
v
ar
io
u
s
co
m
b
in
a
tio
n
s
o
f
t
h
ese
v
al
u
es
an
d
p
r
io
r
iti
ze
s
th
o
s
e
th
at
s
ee
m
p
r
o
m
is
in
g
f
o
r
i
m
p
r
o
v
in
g
C
NN
’
s
p
er
f
o
r
m
a
n
ce
i
n
id
en
ti
f
y
i
n
g
to
m
ato
leaf
d
is
ea
s
es.
2
.
4
.
Select
ing
o
ptim
a
l pa
ra
m
et
er
s
On
ce
t
h
e
s
et
o
f
h
y
p
er
p
ar
a
m
et
er
s
an
d
th
eir
s
ea
r
c
h
s
p
ac
e
is
d
ef
in
ed
,
T
AL
OS
e
x
p
er
i
m
e
n
t
i
s
in
it
iated
,
af
ter
all
th
e
iter
atio
n
s
,
T
AL
OS
p
r
o
v
id
es
u
s
e
f
u
l
v
i
s
u
al
iza
tio
n
s
to
tr
ac
k
an
d
id
en
ti
f
ie
s
th
e
o
p
ti
m
al
s
et
o
f
h
y
p
er
p
ar
a
m
eter
s
t
h
at
lead
s
to
th
e
b
est
r
esu
lt
s
[
2
7
]
.
Fig
u
r
e
4
s
h
o
w
s
th
e
f
lo
w
d
ia
g
r
a
m
o
f
t
h
e
H
y
p
er
p
ar
a
m
eter
o
p
tim
izatio
n
tech
n
iq
u
e
u
s
i
n
g
th
e
T
AL
O
S.
T
h
e
Go
o
g
le
C
o
ll
ab
o
r
ato
r
y
p
latf
o
r
m
(
C
o
lab
)
,
b
u
ilt
ar
o
u
n
d
J
u
p
y
ter
No
teb
o
o
k
s
[
1
4
]
is
u
tili
ze
d
to
ex
ec
u
te
th
e
e
x
p
er
i
m
e
n
t.
T
h
e
i
m
p
le
m
e
n
tat
io
n
al
s
o
in
cl
u
d
es
t
h
e
s
t
u
d
y
to
e
x
a
m
in
e
h
o
w
v
ar
y
i
n
g
f
il
ter
s
izes
in
C
NN
m
o
d
els
[
2
8
]
af
f
ec
t
ac
cu
r
ac
y
in
to
m
ato
lea
f
d
is
ea
s
e
d
etec
tio
n
,
d
e
m
o
n
s
tr
at
in
g
th
e
cr
itical
r
o
le
o
f
h
y
p
er
p
ar
a
m
eter
o
p
tim
izatio
n
in
e
n
h
a
n
ci
n
g
p
er
f
o
r
m
a
n
ce
.
T
h
e
f
o
llo
w
i
n
g
s
ec
tio
n
p
r
o
v
id
es
an
alg
o
r
ith
m
f
o
r
o
p
tim
izin
g
a
C
N
N
m
o
d
el
f
o
r
d
etec
tin
g
to
m
ato
l
ea
f
d
is
ea
s
e
s
u
s
i
n
g
T
AL
O
S.
Fig
u
r
e
4
.
H
y
p
er
-
p
ar
a
m
eter
o
p
ti
m
izatio
n
tech
n
iq
u
e
u
s
in
g
t
h
e
T
A
L
O
S
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
d
o
n
esia
n
J
E
lec
E
n
g
&
C
o
m
p
Sci
I
SS
N:
2
5
0
2
-
4752
TALOS
:
o
p
timiz
a
tio
n
o
f th
e
C
N
N
fo
r
th
e
d
etec
tio
n
…
(
S
h
r
u
th
i Ki
kk
eri S
u
b
r
a
ma
n
ya
)
297
I
n
p
u
t:
T
o
m
ato
leaf
d
is
ea
s
e
d
ata
s
et
Desire
d
h
y
p
er
p
ar
a
m
eter
s
ea
r
c
h
s
p
ac
e
Ou
tp
u
t:
B
est
m
o
d
el
co
n
f
i
g
u
r
atio
n
Nu
m
b
er
o
f
r
o
u
n
d
s
i
n
t
h
e
s
ca
n
n
in
g
p
r
o
ce
s
s
(
m
n
=2
5
,
m
:
se
ar
ch
opt
i
ons,
n:
t
ot
al
num
ber
of
h
y
p
er
p
ar
a
m
eter
s
)
Hig
h
est ac
c
u
r
ac
y
g
iv
e
n
p
ar
a
m
eter
s
s
et
1.
I
n
s
tall
r
eq
u
ir
ed
lib
r
ar
ies:
‘
s
et
u
p
-
to
o
ls
’
an
d
‘
T
AL
OS
’
,
i
m
a
g
e
p
r
o
ce
s
s
in
g
lib
r
ar
ies:
Op
en
C
V,
P
I
L
,
d
ata
m
an
ip
u
latio
n
lib
r
ar
ies:
N
u
m
P
y
2.
Data
p
r
ep
r
o
ce
s
s
in
g
L
o
ad
th
e
to
m
a
to
leaf
d
is
ea
s
e
d
ataset
in
to
C
o
llab
.
P
r
e
-
p
r
o
ce
s
s
i
m
ag
e
s
.
C
o
n
v
er
t c
ateg
o
r
ical
lab
els to
o
n
e
-
h
o
t e
n
co
d
ed
f
o
r
m
at.
Sp
lit d
ata
in
to
tr
ain
i
n
g
,
v
alid
at
io
n
,
an
d
test
i
n
g
s
ets.
3.
Def
i
n
e
C
N
N
m
o
d
el
ar
ch
itect
u
r
e
C
r
ea
te
a
cu
s
to
m
C
N
N
m
o
d
el
w
it
h
b
ase
ar
ch
i
tectu
r
e
a
n
d
p
ar
a
m
eter
s
to
b
e
o
p
ti
m
ized
.
Def
i
n
e
th
e
i
n
p
u
t d
i
m
e
n
s
io
n
s
a
n
d
h
y
p
er
p
ar
a
m
eter
s
f
o
r
th
e
m
o
d
el.
4.
Def
i
n
e
th
e
s
ea
r
ch
s
p
ac
e
f
o
r
h
y
p
er
p
ar
am
eter
s
,
in
c
lu
d
i
n
g
t
h
e
r
an
g
e
s
an
d
v
a
lu
e
s
to
b
e
ex
p
lo
r
ed
.
5.
R
u
n
T
AL
OS e
x
p
er
i
m
e
n
t
E
x
ec
u
te
t
h
e
T
AL
OS
s
ca
n
to
p
er
f
o
r
m
h
y
p
er
p
ar
a
m
eter
o
p
ti
m
i
za
tio
n
o
n
th
e
C
NN
m
o
d
el.
6.
R
es
u
lt a
n
a
l
y
s
is
P
r
in
t th
e
r
esu
lts
o
f
t
h
e
T
AL
O
S scan
,
i
n
cl
u
d
in
g
th
e
to
p
-
p
er
f
o
r
m
i
n
g
co
n
f
ig
u
r
atio
n
s
.
I
d
en
tify
th
e
m
o
d
el
I
D
th
at
y
iel
d
s
th
e
b
est v
al
id
atio
n
ac
cu
r
ac
y
.
L
o
ad
th
e
b
est
m
o
d
el.
R
etr
iev
e
t
h
e
to
tal
n
u
m
b
er
o
f
r
o
u
n
d
s
co
m
p
leted
i
n
th
e
T
AL
O
S scan
a
n
d
h
i
g
h
e
s
t
v
alid
atio
n
ac
cu
r
ac
y
.
3.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
N
T
h
e
s
ec
tio
n
s
h
o
w
ca
s
es
th
e
r
e
s
u
lt
s
f
r
o
m
a
u
to
m
ated
t
u
n
in
g
ex
p
er
i
m
e
n
t
s
o
h
y
p
er
p
ar
a
m
ete
r
s
,
w
h
ic
h
ai
m
to
in
cr
ea
s
e
t
h
e
p
er
f
o
r
m
a
n
ce
o
f
th
e
C
N
N
m
o
d
el
f
o
r
p
lan
t
leaf
d
i
s
ea
s
e
id
en
t
if
ica
tio
n
.
Fil
ter
s
i
n
C
NN
s
ar
e
k
n
o
w
n
f
o
r
th
eir
ef
f
icie
n
t
d
etec
tio
n
o
f
i
m
p
o
r
tan
t
f
ea
tu
r
es
s
u
ch
as
tex
t
u
r
es,
ed
g
es,
an
d
p
atter
n
s
in
t
h
e
i
m
a
g
e
s
an
d
allo
w
th
e
m
o
d
el
to
lear
n
an
d
d
etec
t
c
o
m
p
le
x
d
etails
r
eq
u
ir
ed
f
o
r
ac
cu
r
ate
class
if
i
ca
tio
n
task
s
.
T
h
r
ee
s
ce
n
ar
io
s
o
f
w
er
e
ex
p
lo
r
ed
u
s
in
g
T
AL
O
S
to
cu
s
to
m
ize
th
e
C
NN
w
it
h
v
ar
y
in
g
f
i
lter
s
izes
an
d
d
en
s
e
n
e
u
r
o
n
s
,
ai
m
i
n
g
to
id
en
ti
f
y
t
h
e
m
o
s
t
ef
f
ec
tiv
e
co
n
f
i
g
u
r
atio
n
f
o
r
o
p
ti
m
al
ac
cu
r
ac
y
.
T
h
e
r
esu
lt
s
ar
e
s
u
m
m
ar
ized
i
n
T
ab
le
3
.
Fil
ter
s
izes
o
f
3
×3
r
eg
u
lar
l
y
o
u
t
p
er
f
o
r
m
ed
th
o
s
e
o
f
5
×5
,
m
o
s
t
li
k
el
y
b
ec
au
s
e
th
e
y
w
er
e
b
etter
a
t
ca
p
tu
r
in
g
f
ea
tu
r
e
s
.
I
n
ad
d
itio
n
to
th
ese,
th
e
m
a
in
h
y
p
er
p
ar
a
m
eter
s
u
n
d
er
in
v
est
ig
at
io
n
i
n
c
lu
d
ed
th
e
ac
tiv
a
tio
n
f
u
n
ctio
n
,
t
h
e
o
p
ti
m
izer
tec
h
n
i
q
u
e,
d
r
o
p
o
u
t,
th
e
C
o
n
v
_
d
r
o
p
o
u
t,
an
d
d
e
n
s
e
n
e
u
r
o
n
co
u
n
t,
e
ac
h
h
y
p
er
p
ar
a
m
eter
test
ed
ac
r
o
s
s
a
d
ef
in
ed
r
an
g
e.
T
ab
le
3
.
A
n
al
y
s
i
s
o
f
v
ar
y
i
n
g
f
i
lter
s
ize
an
d
d
en
s
e
n
eu
r
o
n
o
n
C
NN
F
i
l
t
e
r
si
z
e
D
e
n
se
n
e
u
r
o
n
s
A
c
c
u
r
a
c
y
T
r
a
i
n
a
b
l
e
p
a
r
a
me
t
e
r
s
5
×
5
[
3
2
,
6
4
]
8
5
%
5
6
2
7
1
3
×
3
[
5
1
2
]
8
7
%
4
7
2
7
1
3
×
3
[
1
2
8
,
2
5
6
]
9
4
.
5
%
3
7
6
1
5
T
ab
le
4
s
h
o
w
s
th
e
lis
t
o
f
h
y
p
er
p
ar
am
eter
s
an
d
th
eir
s
p
ac
e
co
v
er
ag
e
f
o
r
an
al
y
zi
n
g
th
e
co
n
s
tr
ai
n
ts
o
f
th
e
T
AL
OS
to
o
l.
On
ce
all
r
a
n
g
e
s
h
a
v
e
b
ee
n
d
ef
i
n
ed
,
th
e
T
A
L
O
S
to
o
l
ex
p
er
i
m
e
n
ts
co
m
b
in
e
all
p
o
s
s
ib
le
3
2
co
m
b
i
n
atio
n
s
a
n
d
f
o
r
ea
ch
c
o
n
f
i
g
u
r
atio
n
,
T
AL
OS
tr
ai
n
s
a
m
o
d
el
an
d
ev
al
u
ates
i
ts
p
er
f
o
r
m
an
ce
o
n
th
e
d
ataset.
B
ased
o
n
th
e
e
v
al
u
a
tio
n
r
es
u
lts
,
T
AL
O
S
id
en
t
if
i
es
th
e
h
y
p
er
p
ar
a
m
e
ter
s
et
wh
ich
r
esu
lts
i
n
b
est
p
er
f
o
r
m
a
n
ce
.
G
e
n
er
al
l
y
,
t
h
e
c
o
n
f
i
g
u
r
at
io
n
w
i
t
h
m
a
x
i
m
u
m
v
al
id
a
tio
n
ac
c
u
r
a
c
y
a
n
d
m
i
n
i
m
u
m
v
al
id
a
t
io
n
lo
s
s
is
p
r
e
f
er
r
ed
.
H
e
n
ce
,
Mo
d
el
ID
25
is
co
n
s
id
er
ed
as
th
e
b
est
m
o
d
el
co
n
f
i
g
u
r
atio
n
w
it
h
9
4
.
5
%
ac
cu
r
ac
y
a
n
d
9
4
.
7
%
v
alid
atio
n
ac
cu
r
ac
y
.
T
he
h
y
p
er
p
ar
a
m
eter
s
t
h
at
r
es
u
l
ted
in
h
i
g
h
ac
cu
r
ac
y
ca
n
b
e
f
o
u
n
d
in
th
e
o
u
tp
u
t
co
lu
m
n
o
f
T
ab
le
4
.
T
h
is
ap
p
r
o
ac
h
d
r
asti
ca
ll
y
r
ed
u
ce
s
co
m
p
u
tatio
n
al
ti
m
e
b
y
tr
ac
k
in
g
t
h
e
p
er
f
o
r
m
an
ce
o
f
ea
ch
tr
ial.
T
h
u
s
,
T
A
L
O
S
is
ab
le
to
p
r
o
v
id
es
th
e
in
s
i
g
h
ts
i
n
to
th
e
ef
f
ec
t
o
f
d
i
f
f
er
en
t
h
y
p
er
p
ar
a
m
eter
s
etti
n
g
s
o
n
m
o
d
el
p
er
f
o
r
m
an
ce
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N:
2
5
0
2
-
4752
I
n
d
o
n
esia
n
J
E
lec
E
n
g
&
C
o
m
p
Sci,
Vo
l.
3
8
,
No
.
1
,
A
p
r
il 2
0
2
5
: 2
9
2
-
302
298
T
ab
le
4
.
Def
in
itio
n
o
f
h
y
p
er
p
a
r
a
m
eter
s
s
p
ac
e
an
d
t
h
e
co
r
r
esp
o
n
d
in
g
r
esu
lt
H
y
p
e
r
p
a
r
a
me
t
e
r
s
R
a
n
g
e
O
u
t
p
u
t
A
c
t
i
v
a
t
i
o
n
f
u
n
c
t
i
o
n
[
R
e
L
U
,
E
L
U
]
R
e
L
U
O
p
t
i
mi
z
e
r
[
A
d
a
m,
S
G
D
]
A
d
a
m
d
r
o
p
o
u
t
,
[
0
.
2
5
,
0
.
5
]
0
.
2
5
C
o
n
v
_
d
r
o
p
o
u
t
[
0
.
2
5
,
0
.
5
]
0
.
2
5
D
e
n
se
n
e
u
r
o
n
[
1
2
8
,
2
5
6
]
1
2
8
T
h
e
p
r
o
p
o
s
ed
C
NN’
s
p
er
f
o
r
m
an
ce
i
s
v
i
s
u
alize
d
in
Fi
g
u
r
e
5
an
d
Fig
u
r
e
6
w
h
ic
h
in
d
icate
s
th
e
co
n
f
u
s
io
n
m
a
tr
ix
w
h
ile
F
ig
u
r
es
6
(
a)
an
d
6
(
b
)
d
ep
ict
th
e
v
alid
atio
n
ac
cu
r
ac
y
a
n
d
v
a
lid
atio
n
lo
s
s
o
f
C
N
N
r
esp
ec
tiv
el
y
.
T
ab
les
5
-
7
co
m
p
ar
e
th
e
p
r
o
p
o
s
ed
T
A
L
OS
-
C
NN’
s
m
o
d
el
p
er
f
o
r
m
an
ce
w
it
h
th
e
s
ta
te
-
of
-
th
e
-
ar
t
m
o
d
el
d
e
v
elo
p
ed
i
n
th
e
w
o
r
k
[
1
4
]
an
d
p
r
ed
ef
in
ed
m
o
d
els,
r
ev
ea
ls
s
a
tis
f
ac
to
r
y
p
er
f
o
r
m
an
ce
.
Alth
o
u
g
h
it
s
ac
cu
r
ac
y
i
s
s
li
g
h
tl
y
lo
w
er
,
th
e
m
o
d
e
l
d
em
o
n
s
tr
a
tes
c
o
m
p
eti
t
iv
e
r
esu
l
ts
in
te
r
m
s
o
f
to
tal
p
a
r
a
m
eter
s
a
n
d
to
tal
elap
s
ed
ti
m
e.
T
A
L
O
S
ca
n
g
en
er
ate
th
e
v
i
s
u
a
lizatio
n
to
p
r
o
v
id
e
in
s
i
g
h
t
s
in
to
th
e
o
p
ti
m
izatio
n
p
r
o
ce
s
s
as
d
ep
icted
in
th
e
Fi
g
u
r
e
7
.
Fig
u
r
e
5
.
C
o
n
f
u
s
io
n
m
atr
ix
(
a)
(
b
)
Fig
u
r
e
6
.
Va
li
d
at
io
n
p
er
f
o
r
m
an
ce
o
f
th
e
C
NN
d
u
r
in
g
tr
ain
in
g
(
a
)
v
alid
atio
n
ac
cu
r
ac
y
c
u
r
v
e
an
d
(
b
)
v
alid
atio
n
lo
s
s
c
u
r
v
e
-
Dis
cu
s
s
io
n
T
h
e
f
in
d
in
g
s
o
f
th
i
s
s
tu
d
y
r
ev
ea
l
th
at
h
y
p
er
p
ar
a
m
eter
t
u
n
i
n
g
is
cr
itical
f
o
r
o
p
tim
iz
in
g
C
NN
p
er
f
o
r
m
a
n
ce
i
n
to
m
ato
lea
f
d
is
ea
s
e
d
etec
tio
n
.
I
t
ex
p
lo
r
ed
t
h
r
ee
s
ce
n
ar
io
s
u
s
i
n
g
T
AL
OS,
ea
ch
w
it
h
v
ar
y
in
g
f
ilter
s
ize
s
an
d
d
en
s
e
n
e
u
r
o
n
c
o
n
f
i
g
u
r
atio
n
s
.
T
h
e
b
est
-
p
er
f
o
r
m
i
n
g
m
o
d
el,
ac
h
ie
v
ed
an
i
m
p
r
ess
iv
e
ac
c
u
r
ac
y
o
f
9
4
.
5
%
w
it
h
a
3
×3
f
ilter
a
n
d
d
en
s
e
n
e
u
r
o
n
s
in
th
e
r
an
g
e
o
f
[
1
2
8
,
2
5
6
]
.
T
h
is
r
esu
l
t
u
n
d
er
s
c
o
r
es
th
e
i
m
p
o
r
tan
c
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
d
o
n
esia
n
J
E
lec
E
n
g
&
C
o
m
p
Sci
I
SS
N:
2
5
0
2
-
4752
TALOS
:
o
p
timiz
a
tio
n
o
f th
e
C
N
N
fo
r
th
e
d
etec
tio
n
…
(
S
h
r
u
th
i Ki
kk
eri S
u
b
r
a
ma
n
ya
)
299
o
f
ca
r
ef
u
l
h
y
p
er
p
ar
a
m
eter
ad
ju
s
t
m
e
n
t,
as
f
ilter
s
izes
o
f
3
×3
co
n
s
is
te
n
tl
y
o
u
tp
er
f
o
r
m
ed
th
e
5
×5
f
ilter
s
,
lik
el
y
d
u
e
to
th
eir
s
u
p
er
io
r
ab
ilit
y
to
ca
p
tu
r
e
r
elev
an
t
f
ea
t
u
r
es.
T
h
e
T
A
L
O
S
to
o
l
f
ac
ilit
ated
th
is
o
p
ti
m
izat
io
n
p
r
o
ce
s
s
b
y
ef
f
icie
n
tl
y
e
v
al
u
ati
n
g
3
2
d
if
f
er
en
t
co
m
b
in
at
io
n
s
o
f
h
y
p
er
p
ar
am
eter
s
,
w
it
h
Mo
d
el
I
D
2
5
ac
h
ie
v
i
n
g
a
9
4
.
5
%
ac
cu
r
ac
y
a
n
d
a
9
4
.
7
% v
alid
atio
n
ac
cu
r
ac
y
.
W
h
en
co
m
p
ar
in
g
t
h
e
r
esu
lt
s
w
it
h
t
h
o
s
e
o
f
p
r
ev
io
u
s
s
tu
d
ie
s
[
2
9
]
,
[
3
0
]
h
av
e
s
i
m
i
lar
l
y
wh
ich
al
s
o
h
ig
h
li
g
h
t
th
e
i
m
p
o
r
tan
ce
o
f
h
y
p
er
p
ar
a
m
eter
tu
n
i
n
g
i
n
ac
h
iev
i
n
g
o
p
ti
m
a
l
C
NN
p
er
f
o
r
m
an
ce
an
d
n
o
te
th
a
t
s
m
al
ler
f
ilter
s
izes
o
f
te
n
o
u
tp
er
f
o
r
m
lar
g
er
o
n
es
in
i
m
ag
e
c
lass
i
f
icatio
n
task
s
.
Ho
w
e
v
er
,
th
is
s
t
u
d
y
u
n
iq
u
el
y
d
em
o
n
s
tr
ate
s
th
e
p
r
ac
tical
ap
p
licatio
n
o
f
th
e
T
AL
O
S
o
p
tim
izatio
n
to
o
l,
w
h
ich
s
y
s
te
m
a
ticall
y
e
x
p
lo
r
es
an
d
id
en
ti
f
ies
t
h
e
b
est
h
y
p
er
p
ar
am
eter
co
m
b
i
n
atio
n
s
.
U
n
li
k
e
p
r
ev
io
u
s
w
o
r
k
s
t
h
at
o
f
ten
r
e
lied
o
n
m
a
n
u
al
o
r
h
eu
r
i
s
tic
m
et
h
o
d
s
,
th
i
s
ap
p
r
o
ac
h
le
v
er
ag
es
au
to
m
ated
o
p
tim
izatio
n
to
ac
h
ie
v
e
s
u
p
er
io
r
r
esu
lt
s
w
it
h
g
r
ea
ter
ef
f
icien
c
y
.
T
ab
les
5
-
7
co
m
p
ar
e
th
is
w
o
r
k
w
it
h
p
r
ev
io
u
s
s
tu
d
ies
in
ter
m
s
o
f
m
o
d
el
co
m
p
le
x
it
y
,
to
tal
p
ar
am
eter
s
,
tr
ain
i
n
g
t
i
m
e,
a
n
d
d
ataset
s
ize.
A
s
tr
en
g
t
h
o
f
o
u
r
s
t
u
d
y
li
es
in
t
h
e
co
m
p
r
eh
e
n
s
iv
e
e
v
alu
a
tio
n
o
f
d
i
f
f
er
en
t
h
y
p
er
p
ar
am
eter
co
n
f
i
g
u
r
atio
n
s
u
s
i
n
g
a
r
o
b
u
s
t
o
p
ti
m
izatio
n
to
o
l,
w
h
ic
h
co
n
tr
ast
s
w
it
h
t
h
e
m
o
r
e
li
m
ited
s
co
p
e
o
f
p
r
ev
io
u
s
r
esear
ch
th
at
t
y
p
icall
y
ex
p
lo
r
ed
f
e
w
er
co
n
f
ig
u
r
atio
n
s
.
No
n
eth
e
less
,
o
u
r
m
o
d
el’
s
ac
cu
r
ac
y
,
w
h
ile
h
i
g
h
,
is
s
lig
h
tl
y
lo
w
er
t
h
a
n
s
o
m
e
s
ta
te
-
of
-
t
h
e
-
ar
t
m
o
d
el
s
r
ep
o
r
ted
in
t
h
e
l
iter
atu
r
e.
T
h
is
s
u
g
g
ests
t
h
at
w
h
i
le
o
u
r
ap
p
r
o
ac
h
is
ef
f
ec
tiv
e,
t
h
er
e
is
r
o
o
m
f
o
r
i
m
p
r
o
v
e
m
e
n
t,
p
ar
ticu
lar
l
y
i
n
i
n
te
g
r
atin
g
m
o
r
e
s
o
p
h
is
t
icate
d
tech
n
iq
u
es
o
r
ad
d
itio
n
al
la
y
e
r
s
o
f
o
p
ti
m
izat
io
n
.
Un
e
x
p
ec
t
ed
ly
,
t
h
e
m
o
d
el
w
i
th
t
h
e
lar
g
est
d
en
s
e
n
e
u
r
o
n
co
n
f
i
g
u
r
atio
n
i
n
ca
s
e
2
d
id
n
o
t
o
u
tp
er
f
o
r
m
th
e
s
m
aller
co
n
f
ig
u
r
at
io
n
i
n
ca
s
e
3
,
in
d
icatin
g
th
at
h
i
g
h
er
m
o
d
el
co
m
p
le
x
it
y
d
o
es
n
o
t
n
ec
es
s
ar
i
l
y
tr
an
s
late
to
b
etter
p
e
r
f
o
r
m
a
n
ce
,
w
h
ic
h
alig
n
s
w
ith
f
in
d
i
n
g
s
in
s
o
m
e
p
r
ev
io
u
s
s
tu
d
ie
s
.
F
ig
u
r
e
7
.
Vis
u
al
izatio
n
o
f
t
h
e
T
A
L
O
S
ex
p
er
i
m
e
n
t
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N:
2
5
0
2
-
4752
I
n
d
o
n
esia
n
J
E
lec
E
n
g
&
C
o
m
p
Sci,
Vo
l.
3
8
,
No
.
1
,
A
p
r
il 2
0
2
5
: 2
9
2
-
302
300
T
ab
le
5
.
Ov
er
v
ie
w
o
f
th
e
l
iter
atu
r
e
s
u
r
v
e
y
a
n
d
co
m
p
ar
is
o
n
w
it
h
p
r
o
p
o
s
ed
w
o
r
k
M
etho
d
Dat
ase
t
DL
mod
e
l
s
Opt
i
mi
ze
r/
T
ec
h
n
i
q
u
e
Accu
racy
S
u
mmary
[17
]
R
e
u
t
e
r
s (1
1
2
2
8
i
mag
e
s)
,
F
a
sh
i
o
n
M
N
I
S
T
(
7
0
0
0
0
i
mag
e
s)
C
u
st
o
mi
ze
d
d
ee
p
N
N
W
OA
8
9
.
8
5
%
(
F
a
sh
i
o
n
M
N
I
S
T)
,
8
0
.
6
0
%(
R
e
u
t
e
r
s)
C
a
n
b
e
se
n
si
t
i
v
e
t
o
i
n
i
t
i
a
l
i
z
a
t
i
o
n
,
mi
g
h
t
g
e
t
st
u
c
k
i
n
l
o
c
a
l
o
p
t
i
ma
.
[18
]
1
4
7
,
5
0
0
i
m
a
g
e
s p
l
a
n
t
l
e
a
f
c
l
a
sse
s
14
-
D
C
N
N
R
a
n
d
o
m se
a
r
c
h
9
9
.
9
7
%
I
n
e
f
f
i
c
i
e
n
t
,
o
f
t
e
n
mi
sse
s o
p
t
i
m
a
l
h
y
p
e
r
p
a
r
a
me
t
e
r
c
o
mb
i
n
a
t
i
o
n
s.
[19
]
1
8
1
5
9
i
m
a
g
e
s o
f
t
o
mat
o
l
e
a
f
D
e
e
p
N
N
W
O
A
w
i
t
h
h
y
b
r
i
d
P
C
A
86%
(t
esti
n
g
)
S
i
mi
l
a
r
t
o
W
O
A
,
c
a
n
b
e
se
n
s
i
t
i
v
e
t
o
i
n
i
t
i
a
l
i
z
a
t
i
o
n
.
[20
]
2
4
0
,
0
0
0
a
u
g
me
n
t
e
d
i
mag
e
s p
l
a
n
t
s
D
C
N
N
R
a
n
d
o
m
s
e
a
r
c
h
9
8
.
4
1
%
I
n
e
f
f
i
c
i
e
n
t
,
m
i
g
h
t
mi
s
s o
p
t
i
m
a
l
h
y
p
e
r
p
a
r
a
me
t
e
r
c
o
mb
i
n
a
t
i
o
n
s.
[21
]
5
4
3
0
9
i
m
a
g
e
s,
P
l
a
n
t
V
i
l
l
a
g
e
d
a
t
a
se
t
P
D
I
-
C
M
A
E
P
DI
-
C
M
AE
-
D
DA
OA
9
8
%
C
o
mp
l
e
x
a
l
g
o
r
i
t
h
m,
c
o
mp
u
t
a
t
i
o
n
a
l
l
y
e
x
p
e
n
si
v
e
.
[22
]
1
1
2
5
i
mag
e
s o
f
S
w
e
d
i
sh
l
e
a
f
d
a
t
a
se
t
,
4
5
0
3
i
mag
e
s
o
f
me
n
d
e
l
e
y
d
a
t
a
C
N
N
SS
-
W
OA
97%
S
p
e
c
i
f
i
c
t
o
a
c
t
i
v
a
t
i
o
n
f
u
n
c
t
i
o
n
o
p
t
i
m
i
z
a
t
i
o
n
,
l
i
mi
t
e
d
a
p
p
l
i
c
a
b
i
l
i
t
y
.
[23
]
5
4
3
0
6
i
m
a
g
e
s o
f
P
l
a
n
t
V
i
l
l
a
g
e
,
2
5
9
8
i
mag
e
s
o
f
P
l
a
n
t
D
o
c
X
c
e
p
t
i
o
n
,
D
e
n
se
N
e
t
Ai
S
ara
2
3
%i
m
p
r
o
v
e
me
n
t
S
p
e
c
i
f
i
c
t
o
c
e
r
t
a
i
n
a
r
c
h
i
t
e
c
t
u
r
e
s.
[24
]
5
4
3
0
9
i
m
a
g
e
s o
f
P
l
a
n
t
V
i
l
l
a
g
e
,
R
e
sN
e
t
1
8
,
A
l
e
x
N
e
t
M
a
n
u
a
l
f
i
n
e
-
t
u
n
i
n
g
h
y
p
e
r
p
a
r
a
me
t
e
r
s
9
0
.
3
1
%
(
A
l
e
x
N
e
t
)
,
8
7
.
6
%
(
R
e
sN
e
t
1
8
)
T
i
me
c
o
n
su
m
i
n
g
,
r
e
q
u
i
r
e
s e
x
p
e
r
t
k
n
o
w
l
e
d
g
e
.
[25
]
1
0
0
7
1
i
m
a
g
e
s,
P
l
a
n
t
V
i
l
l
a
g
e
d
a
t
a
se
t
S
V
M
f
r
u
i
t
f
l
y
(
F
O
A
)
9
1
.
1
%
C
o
n
v
e
r
g
e
s p
r
e
mat
u
r
e
l
y
t
o
su
b
o
p
t
i
mal
so
l
u
t
i
o
n
s
,
w
h
e
n
d
e
a
l
i
n
g
w
i
t
h
c
o
m
p
l
e
x
o
p
t
i
m
i
z
a
t
i
o
n
p
r
o
b
l
e
ms.
P
rop
o
sed
work
3
0
3
0
i
mag
e
s o
f
P
l
a
n
t
V
i
l
l
a
g
e
C
NN
T
AL
OS t
o
o
l
9
4
.
5
%
A
c
c
e
l
e
r
a
t
e
d
h
y
p
e
r
p
a
r
a
me
t
e
r
t
u
n
i
n
g
,
r
e
d
u
c
e
d
m
a
n
u
a
l
i
n
t
e
r
v
e
n
t
i
o
n
,
o
p
t
i
m
i
z
e
t
i
me
a
n
d
c
o
s
t
T
ab
le
6.
C
o
m
p
ar
is
o
n
of
p
r
o
p
o
s
ed
w
o
r
k
i
n
ter
m
s
o
f
f
ilter
R
e
f
e
r
e
n
c
e
K
e
r
n
e
l
si
z
e
A
c
c
u
r
a
c
y
(
%)
P
r
e
c
i
si
o
n
R
e
c
a
l
l
F1
-
sco
r
e
P
a
r
a
me
t
e
r
s
El
a
p
se
d
t
i
me
[
1
4
]
3
×
3
9
7
.
4
1
0
.
9
7
6
0
0
.
9
7
2
8
0
.
9
7
4
4
1
3
3
4
5
3
3
6hr
2
8
mi
n
2
3
se
c
[
1
4
]
5
×
5
9
7
.
6
1
0
.
9
7
6
6
0
.
9
7
5
8
0
.
9
7
6
2
8hr
2
4
mi
n
3
se
c
[
1
4
]
3
×
3
9
9
.
5
9
0
.
9
9
6
3
0
.
9
9
8
4
0
.
9
9
8
4
6hr
5
8
mi
n
2
7
se
c
[
1
4
]
5
×
5
9
9
.
6
6
0
.
9
9
6
8
0
.
9
9
6
5
0
.
9
9
6
7
8hr
3
8
mi
n
2
7
se
c
P
r
o
p
o
se
d
w
o
r
k
3
x
3
9
4
.
5
0
.
9
8
6
7
0
.
9
8
6
7
0
.
9
8
6
7
3
7
,
6
4
7
3hr
1
8
mi
n
4
5
se
c
T
ab
le
7
.
C
o
m
p
ar
is
o
n
o
f
p
r
o
p
o
s
ed
T
A
L
OS
-
C
NN
w
i
th
t
h
e
ex
i
s
tin
g
m
o
d
el
s
M
o
d
e
l
s
N
u
mb
e
r
o
f
p
a
r
a
me
t
e
r
s
Ep
o
c
h
s re
q
u
i
r
e
d
V
G
G
1
6
3
9
4
4
3
0
4
3
15
I
n
c
e
p
t
i
o
n
V
3
[
1
3
]
2
4
,
9
3
7
,
2
8
3
15
M
o
b
i
l
e
N
e
t
V
2
[
1
3
]
2
,
4
2
2
,
3
3
9
15
14
-
D
C
N
N
[
1
8
]
1
7
,
9
2
8
,
5
7
1
1
0
0
0
C
N
N
1
[
1
3
]
5
,
1
0
8
,
4
2
6
50
C
N
N
2
[
1
3
]
4
9
4
,
2
1
8
50
P
r
o
p
o
se
d
C
N
N
3
3
6
4
7
10
4.
CO
NCLU
SI
O
N
T
h
e
d
ep
l
o
y
m
e
n
t
of
cu
tt
in
g
-
e
d
g
e
d
ee
p
C
NN
m
o
d
els
f
o
r
ea
r
l
y
p
lan
t
d
is
ea
s
e
d
etec
tio
n
h
as
m
ad
e
tr
e
m
en
d
o
u
s
s
tr
id
es
in
r
ec
e
n
t
y
ea
r
s
.
A
s
a
r
e
s
u
l
t,
it
h
a
s
b
ec
o
m
e
a
p
r
o
m
i
n
e
n
t
an
d
ac
tiv
e
r
esear
c
h
f
ie
ld
,
p
r
o
m
i
s
in
g
ea
r
l
y
d
is
ea
s
e
d
etec
tio
n
an
d
i
m
p
r
o
v
ed
p
r
ec
is
io
n
in
m
an
a
g
i
n
g
ag
r
ic
u
lt
u
r
al
d
is
ea
s
es.
E
f
f
icie
n
t
id
en
ti
f
icatio
n
o
f
t
h
e
d
is
ea
s
es
r
esu
lt
s
i
n
r
ed
u
ce
d
lo
s
s
es,
b
ette
r
cr
o
p
m
a
n
ag
e
m
e
n
t
a
n
d
i
m
p
r
o
v
ed
f
o
o
d
s
ec
u
r
it
y
.
Ho
w
e
v
er
,
is
s
u
es
lik
e
m
an
u
a
l
tu
n
i
n
g
o
f
th
e
h
y
p
er
p
ar
a
m
eter
s
,
th
e
n
ec
es
s
it
y
o
f
lar
g
e
d
atasets
,
an
d
h
i
g
h
co
m
p
u
tatio
n
al
co
m
p
le
x
it
y
p
er
s
is
t.
T
h
is
r
esear
c
h
ad
d
r
ess
es
t
h
e
ea
r
lier
co
n
ce
r
n
s
b
y
i
n
tr
o
d
u
cin
g
th
e
a
u
to
m
ated
h
y
p
er
p
ar
a
m
eter
tu
n
i
n
g
tech
n
i
q
u
e
e
m
p
lo
y
ed
w
ith
a
cu
s
to
m
i
ze
d
C
NN
m
o
d
el.
T
r
ain
ed
u
s
i
n
g
a
d
ataset
o
f
3
0
3
0
to
m
ato
leaf
i
m
a
g
es.
Firs
tl
y
,
t
h
e
C
NN
m
o
d
el
is
d
ef
i
n
ed
ef
f
i
cien
tl
y
to
tr
ain
w
it
h
a
li
m
ited
n
u
m
b
er
o
f
i
m
ag
e
s
.
Nex
t,
T
AL
OS
is
s
e
t b
y
s
p
ec
i
f
y
in
g
t
h
e
h
y
p
er
p
ar
a
m
eter
s
a
n
d
r
an
g
e,
a
n
d
th
e
e
x
p
er
i
m
e
n
t i
s
s
et
f
o
r
d
if
f
er
en
t
f
ilter
s
izes,
5
×5
an
d
3
×3
,
w
it
h
v
a
r
y
in
g
d
r
o
p
o
u
t
r
ates
o
f
n
e
u
r
o
n
s
.
Ho
w
e
v
er
,
T
A
L
O
S
can
be
co
m
p
u
tatio
n
all
y
in
te
n
s
i
v
e,
t
h
e
b
en
e
f
it
s
o
v
e
r
co
m
e
th
e
d
r
a
w
b
ac
k
s
.
B
ec
au
s
e
T
AL
OS
n
ar
r
o
w
s
t
h
e
s
ea
r
ch
s
p
ac
e
o
f
h
y
p
er
p
ar
a
m
eter
s
,
w
h
ic
h
en
ab
l
es
m
o
r
e
ef
f
ec
ti
v
e
tu
n
in
g
co
m
p
ar
ed
to
m
an
u
al
m
et
h
o
d
s
an
d
ca
n
b
e
ad
ap
te
d
to
h
an
d
le
lar
g
e
m
o
d
els.
T
h
is
m
a
k
es t
h
e
T
AL
OS s
u
itab
le
f
o
r
r
ea
l
-
w
o
r
ld
ap
p
licatio
n
s
d
esp
ite
i
ts
co
m
p
u
t
in
g
n
ee
d
s
.
T
h
e
m
o
d
el
ac
h
iev
ed
an
i
m
p
r
ess
i
ve
av
er
ag
e
ac
cu
r
ac
y
of
9
5
.
9
2
%
w
it
h
3
×3
f
ilter
s
in
cla
s
s
i
f
y
in
g
h
ea
lt
h
y
a
n
d
d
is
ea
s
ed
p
lan
ts
u
s
i
n
g
to
m
ato
leaf
i
m
a
g
e
s
w
ith
co
m
p
ar
ati
v
e
p
ar
a
m
eter
s
3
3
,
6
4
7
in
a
p
er
io
d
of
3
h
o
u
r
s
2
8
m
i
n
u
tes.
T
h
e
b
atch
s
ize,
n
u
m
b
er
of
tr
ain
i
n
g
ep
o
ch
s
,
an
d
d
r
o
p
o
u
t
r
ate
s
ig
n
i
f
ican
tl
y
i
n
f
l
u
en
ce
d
th
e
m
o
d
el
’
s
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
d
o
n
esia
n
J
E
lec
E
n
g
&
C
o
m
p
Sci
I
SS
N:
2
5
0
2
-
4752
TALOS
:
o
p
timiz
a
tio
n
o
f th
e
C
N
N
fo
r
th
e
d
etec
tio
n
…
(
S
h
r
u
th
i Ki
kk
eri S
u
b
r
a
ma
n
ya
)
301
p
er
f
o
r
m
a
n
ce
.
C
o
m
p
ar
ed
to
t
h
e
p
r
ev
io
u
s
ar
ch
itect
u
r
es,
th
e
p
r
o
p
o
s
ed
d
esig
n
s
u
b
s
ta
n
tia
ll
y
r
ed
u
ce
s
t
h
e
n
u
m
b
er
of
p
ar
a
m
eter
s
.
F
u
t
u
r
e
r
esear
ch
s
h
o
u
ld
f
o
cu
s
o
n
e
x
ten
d
i
n
g
th
e
w
o
r
k
to
v
ar
io
u
s
d
ee
p
-
lea
r
n
in
g
ar
ch
i
tectu
r
e
s
w
it
h
m
o
r
e
clas
s
es
in
d
if
f
er
en
t
cr
o
p
s
t
o
im
p
r
o
v
e
t
h
e
m
o
d
els’
r
o
b
u
s
t
n
es
s
an
d
d
ep
en
d
ab
ilit
y
in
t
h
e
f
u
t
u
r
e
to
v
ar
io
u
s
d
ee
p
-
lear
n
i
n
g
ar
c
h
itec
tu
r
es
w
it
h
m
o
r
e
clas
s
es
in
d
if
f
er
en
t c
r
o
p
s
.
RE
F
E
R
E
NC
E
S
[
1
]
R
.
G
o
w
t
h
a
mi
,
N
.
S
h
a
r
ma,
R
.
P
a
n
d
e
y
,
a
n
d
A
.
A
g
r
a
w
a
l
,
“
S
t
a
t
u
s
a
n
d
c
o
n
s
o
l
i
d
a
t
e
d
l
i
st
o
f
t
h
r
e
a
t
e
n
e
d
me
d
i
c
i
n
a
l
p
l
a
n
t
s
o
f
I
n
d
i
a
,
”
G
e
n
e
t
i
c
R
e
so
u
r
c
e
s
a
n
d
C
r
o
p
E
v
o
l
u
t
i
o
n
,
v
o
l
.
6
8
,
n
o
.
6
,
p
p
.
2
2
3
5
–
2
2
6
3
,
A
u
g
.
2
0
2
1
,
d
o
i
:
1
0
.
1
0
0
7
/
s
1
0
7
2
2
-
0
2
1
-
0
1
1
9
9
-
0.
[
2
]
E.
N
i
c
L
u
g
h
a
d
h
a
e
t
a
l
.
,
“
Ex
t
i
n
c
t
i
o
n
r
i
sk
a
n
d
t
h
r
e
a
t
s
t
o
p
l
a
n
t
s
a
n
d
f
u
n
g
i
,
”
Pl
a
n
t
s
P
e
o
p
l
e
Pl
a
n
e
t
,
v
o
l
.
2
,
n
o
.
5
,
p
p
.
3
8
9
–
4
0
8
,
S
e
p
.
2
0
2
0
,
d
o
i
:
1
0
.
1
0
0
2
/
p
p
p
3
.
1
0
1
4
6
.
[
3
]
Y
.
L
i
u
,
D
.
Z
h
a
n
g
,
G
.
L
u
,
a
n
d
W
.
-
Y
.
M
a
,
“
A
su
r
v
e
y
o
f
c
o
n
t
e
n
t
-
b
a
se
d
i
m
a
g
e
r
e
t
r
i
e
v
a
l
w
i
t
h
h
i
g
h
-
l
e
v
e
l
se
man
t
i
c
s,
”
Pa
t
t
e
r
n
Re
c
o
g
n
i
t
i
o
n
,
v
o
l
.
4
0
,
n
o
.
1
,
p
p
.
2
6
2
–
2
8
2
,
Ja
n
.
2
0
0
7
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
p
a
t
c
o
g
.
2
0
0
6
.
0
4
.
0
4
5
.
[
4
]
M
.
K
.
R
.
G
a
v
h
a
l
e
a
n
d
P
.
U
.
G
a
w
a
n
d
e
,
“
A
n
o
v
e
r
v
i
e
w
o
f
t
h
e
r
e
se
a
r
c
h
o
n
p
l
a
n
t
l
e
a
v
e
s
d
i
se
a
se
d
e
t
e
c
t
i
o
n
u
s
i
n
g
i
mag
e
p
r
o
c
e
ssi
n
g
t
e
c
h
n
i
q
u
e
s,
”
I
O
S
R
J
o
u
rn
a
l
o
f
C
o
m
p
u
t
e
r E
n
g
i
n
e
e
r
i
n
g
,
v
o
l
.
1
6
,
n
o
.
1
,
p
p
.
1
0
–
1
6
,
2
0
1
4
,
d
o
i
:
1
0
.
9
7
9
0
/
0
6
6
1
-
1
6
1
5
1
0
1
6
.
[
5
]
G
.
D
h
i
n
g
r
a
,
V
.
K
u
mar
,
a
n
d
H
.
D
.
Jo
sh
i
,
“
S
t
u
d
y
o
f
d
i
g
i
t
a
l
i
mag
e
p
r
o
c
e
ssi
n
g
t
e
c
h
n
i
q
u
e
s
f
o
r
l
e
a
f
d
i
se
a
se
d
e
t
e
c
t
i
o
n
a
n
d
c
l
a
ssi
f
i
c
a
t
i
o
n
,
”
Mu
l
t
i
m
e
d
i
a
T
o
o
l
s
a
n
d
Ap
p
l
i
c
a
t
i
o
n
s
,
v
o
l
.
7
7
,
n
o
.
1
5
,
p
p
.
1
9
9
5
1
–
2
0
0
0
0
,
=
2
0
1
8
,
d
o
i
:
1
0
.
1
0
0
7
/
s1
1
0
4
2
-
0
1
7
-
5
4
4
5
-
8.
[
6
]
Y
.
L
i
,
J.
N
i
e
,
a
n
d
X
.
C
h
a
o
,
“
D
o
w
e
r
e
a
l
l
y
n
e
e
d
d
e
e
p
C
N
N
f
o
r
p
l
a
n
t
d
i
se
a
se
s
i
d
e
n
t
i
f
i
c
a
t
i
o
n
?
,
”
C
o
m
p
u
t
e
rs
a
n
d
E
l
e
c
t
r
o
n
i
c
s
i
n
A
g
r
i
c
u
l
t
u
r
e
,
v
o
l
.
1
7
8
,
p
.
1
0
5
8
0
3
,
N
o
v
.
2
0
2
0
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
c
o
mp
a
g
.
2
0
2
0
.
1
0
5
8
0
3
.
[
7
]
G
.
S
a
l
e
e
m,
M
.
A
k
h
t
a
r
,
N
.
A
h
me
d
,
a
n
d
W
.
S
.
Q
u
r
e
sh
i
,
“
A
u
t
o
mat
e
d
a
n
a
l
y
si
s
o
f
v
i
s
u
a
l
l
e
a
f
s
h
a
p
e
f
e
a
t
u
r
e
s
f
o
r
p
l
a
n
t
c
l
a
ss
i
f
i
c
a
t
i
o
n
,
”
C
o
m
p
u
t
e
rs
a
n
d
E
l
e
c
t
r
o
n
i
c
s
i
n
A
g
r
i
c
u
l
t
u
r
e
,
v
o
l
.
1
5
7
,
p
p
.
2
7
0
–
2
8
0
,
N
o
v
.
2
0
1
9
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
c
o
mp
a
g
.
2
0
1
8
.
1
2
.
0
3
8
.
[
8
]
L
.
W
a
n
g
,
J.
Z
h
a
o
,
a
n
d
R
.
M
o
r
t
i
e
r
,
“
N
e
u
r
a
l
n
e
t
w
o
r
k
,
”
i
n
P
o
w
e
r
S
y
st
e
m
s
,
v
o
l
.
2
8
,
2
0
0
7
,
p
p
.
7
5
–
1
5
9
.
[
9
]
J.
W
a
n
g
,
Z
.
M
o
,
H
.
Z
h
a
n
g
,
a
n
d
Q
.
M
i
a
o
,
“
E
n
se
mb
l
e
d
i
a
g
n
o
si
s
me
t
h
o
d
b
a
se
d
o
n
t
r
a
n
sf
e
r
l
e
a
r
n
i
n
g
a
n
d
i
n
c
r
e
me
n
t
a
l
l
e
a
r
n
i
n
g
t
o
w
a
r
d
s me
c
h
a
n
i
c
a
l
b
i
g
d
a
t
a
,
”
i
n
M
e
a
su
reme
n
t
:
J
o
u
r
n
a
l
o
f
t
h
e
I
n
t
e
r
n
a
t
i
o
n
a
l
M
e
a
s
u
r
e
m
e
n
t
C
o
n
f
e
d
e
r
a
t
i
o
n
,
v
o
l
.
1
5
5
,
2
0
2
0
.
[
1
0
]
G
.
G
e
e
t
h
a
r
a
ma
n
i
a
n
d
A
.
P
.
J
.
,
“
I
d
e
n
t
i
f
i
c
a
t
i
o
n
o
f
p
l
a
n
t
l
e
a
f
d
i
se
a
se
s
u
s
i
n
g
a
n
i
n
e
-
l
a
y
e
r
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
,
”
C
o
m
p
u
t
e
rs
a
n
d
E
l
e
c
t
r
i
c
a
l
En
g
i
n
e
e
r
i
n
g
,
v
o
l
.
7
6
,
p
p
.
3
2
3
–
3
3
8
,
J
u
n
.
2
0
1
9
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
c
o
m
p
e
l
e
c
e
n
g
.
2
0
1
9
.
0
4
.
0
1
1
.
[
1
1
]
V
.
F
e
r
r
a
r
i
,
C
.
S
mi
n
c
h
i
se
sc
u
,
M
.
H
e
b
e
r
t
,
a
n
d
Y
.
W
e
i
ss,
“
S
h
u
f
f
l
e
N
e
t
V
2
:
p
r
a
c
t
i
c
a
l
g
u
i
d
e
l
i
n
e
s
f
o
r
e
f
f
i
c
i
e
n
t
C
N
N
a
r
c
h
i
t
e
c
t
u
r
e
d
e
si
g
n
,
”
L
e
c
t
u
re
N
o
t
e
s
i
n
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
(
i
n
c
l
u
d
i
n
g
s
u
b
ser
i
e
s
L
e
c
t
u
r
e
N
o
t
e
s
i
n
Art
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
a
n
d
L
e
c
t
u
re
N
o
t
e
s
i
n
Bi
o
i
n
f
o
rm
a
t
i
c
s)
,
v
o
l
.
1
1
2
1
8
,
p
p
.
v
i
i
–
i
x
,
2
0
1
8
,
d
o
i
:
1
0
.
1
0
0
7
/
9
7
8
-
3
-
0
3
0
-
3
4
2
0
1
2
6
4
-
9.
[
1
2
]
S
.
I
.
P
r
o
t
t
a
sh
a
a
n
d
S
.
M
.
S
.
R
e
z
a
,
“
A
c
l
a
ssi
f
i
c
a
t
i
o
n
mo
d
e
l
b
a
se
d
o
n
d
e
p
t
h
w
i
se
se
p
a
r
a
b
l
e
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
t
o
i
d
e
n
t
i
f
y
r
i
c
e
p
l
a
n
t
d
i
se
a
se
s,”
I
n
t
e
rn
a
t
i
o
n
a
l
J
o
u
rn
a
l
o
f
El
e
c
t
ri
c
a
l
a
n
d
C
o
m
p
u
t
e
r
E
n
g
i
n
e
e
ri
n
g
(
I
J
E
C
E)
,
v
o
l
.
1
2
,
n
o
.
4
,
p
p
.
3
6
4
2
–
3
6
5
4
,
A
u
g
.
2
0
2
2
,
d
o
i
:
1
0
.
1
1
5
9
1
/
i
j
e
c
e
.
v
1
2
i
4
.
p
p
3
6
4
2
-
3
6
5
4
.
[
1
3
]
H
.
U
l
u
t
a
ş
a
n
d
V
.
A
sl
a
n
t
a
ş,
“
D
e
si
g
n
o
f
e
f
f
i
c
i
e
n
t
me
t
h
o
d
s fo
r
t
h
e
d
e
t
e
c
t
i
o
n
o
f
t
o
mat
o
l
e
a
f
d
i
se
a
se
u
t
i
l
i
z
i
n
g
p
r
o
p
o
se
d
e
n
se
mb
l
e
C
N
N
mo
d
e
l
,
”
El
e
c
t
ro
n
i
c
s (
S
w
i
t
z
e
rl
a
n
d
)
,
v
o
l
.
1
2
,
n
o
.
4
,
p
.
8
2
7
,
F
e
b
.
2
0
2
3
,
d
o
i
:
1
0
.
3
3
9
0
/
e
l
e
c
t
r
o
n
i
c
s
1
2
0
4
0
8
2
7
.
[
1
4
]
R
.
V
a
l
a
r
m
a
t
h
i
a
n
d
T
.
S
h
e
e
l
a
,
“
H
e
a
r
t
d
i
se
a
se
p
r
e
d
i
c
t
i
o
n
u
si
n
g
h
y
p
e
r
p
a
r
a
me
t
e
r
o
p
t
i
mi
z
a
t
i
o
n
(
H
P
O
)
t
u
n
i
n
g
,
”
Bi
o
m
e
d
i
c
a
l
S
i
g
n
a
l
Pr
o
c
e
s
si
n
g
a
n
d
C
o
n
t
r
o
l
,
v
o
l
.
7
0
,
p
.
1
0
3
0
3
3
,
S
e
p
.
2
0
2
1
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
b
s
p
c
.
2
0
2
1
.
1
0
3
0
3
3
.
[
1
5
]
X
.
L
i
u
a
n
d
C
.
W
a
n
g
,
“
A
n
e
mp
i
r
i
c
a
l
st
u
d
y
o
n
h
y
p
e
r
p
a
r
a
me
t
e
r
o
p
t
i
mi
z
a
t
i
o
n
f
o
r
f
i
n
e
-
t
u
n
i
n
g
p
r
e
-
t
r
a
i
n
e
d
l
a
n
g
u
a
g
e
mo
d
e
l
s,”
i
n
AC
L
-
I
J
C
N
L
P
2
0
2
1
-
5
9
t
h
A
n
n
u
a
l
Me
e
t
i
n
g
o
f
t
h
e
Asso
c
i
a
t
i
o
n
f
o
r
C
o
m
p
u
t
a
t
i
o
n
a
l
L
i
n
g
u
i
st
i
c
s
a
n
d
t
h
e
1
1
t
h
I
n
t
e
rn
a
t
i
o
n
a
l
J
o
i
n
t
C
o
n
f
e
re
n
c
e
o
n
N
a
t
u
ra
l
L
a
n
g
u
a
g
e
Pro
c
e
ss
i
n
g
,
Pr
o
c
e
e
d
i
n
g
s
o
f
t
h
e
C
o
n
f
e
re
n
c
e
,
2
0
2
1
,
p
p
.
2
2
8
6
–
2
3
0
0
,
d
o
i
:
1
0
.
1
8
6
5
3
/
v
1
/
2
0
2
1
.
a
c
l
-
l
o
n
g
.
1
7
8
.
[
1
6
]
B
.
Z
h
a
n
g
e
t
a
l
.
,
“
O
n
t
h
e
i
mp
o
r
t
a
n
c
e
o
f
h
y
p
e
r
p
a
r
a
me
t
e
r
o
p
t
i
mi
z
a
t
i
o
n
f
o
r
mo
d
e
l
-
b
a
se
d
r
e
i
n
f
o
r
c
e
me
n
t
l
e
a
r
n
i
n
g
,
”
Pro
c
e
e
d
i
n
g
s
o
f
Ma
c
h
i
n
e
L
e
a
r
n
i
n
g
R
e
se
a
rc
h
,
v
o
l
.
1
3
0
,
p
p
.
4
0
1
5
–
4
0
2
3
,
F
e
b
.
2
0
2
1
,
[
O
n
l
i
n
e
]
.
A
v
a
i
l
a
b
l
e
:
h
t
t
p
:
/
/
a
r
x
i
v
.
o
r
g
/
a
b
s/
2
1
0
2
.
1
3
6
5
1
.
[
1
7
]
A
.
B
r
o
d
z
i
c
k
i
,
M
.
P
i
e
k
a
r
sk
i
,
a
n
d
J
.
Ja
w
o
r
e
k
-
K
o
r
j
a
k
o
w
sk
a
,
“
T
h
e
w
h
a
l
e
o
p
t
i
mi
z
a
t
i
o
n
a
l
g
o
r
i
t
h
m
a
p
p
r
o
a
c
h
f
o
r
d
e
e
p
n
e
u
r
a
l
n
e
t
w
o
r
k
s,”
S
e
n
so
rs
,
v
o
l
.
2
1
,
n
o
.
2
3
,
p
.
8
0
0
3
,
N
o
v
.
2
0
2
1
,
d
o
i
:
1
0
.
3
3
9
0
/
s
2
1
2
3
8
0
0
3
.
[
1
8
]
J.
A
.
P
a
n
d
i
a
n
,
V
.
D
.
K
u
mar,
O
.
G
e
man
,
M
.
H
n
a
t
i
u
c
,
M
.
A
r
i
f
,
a
n
d
K
.
K
a
n
c
h
a
n
a
d
e
v
i
,
“
P
l
a
n
t
d
i
se
a
se
d
e
t
e
c
t
i
o
n
u
si
n
g
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
,
”
A
p
p
l
i
e
d
S
c
i
e
n
c
e
s
(
S
w
i
t
zer
l
a
n
d
)
,
v
o
l
.
1
2
,
n
o
.
1
4
,
p
.
6
9
8
2
,
Ju
l
.
2
0
2
2
,
d
o
i
:
1
0
.
3
3
9
0
/
a
p
p
1
2
1
4
6
9
8
2
.
[
1
9
]
T
.
R
.
G
a
d
e
k
a
l
l
u
e
t
a
l
.
,
“
A
n
o
v
e
l
P
C
A
–
w
h
a
l
e
o
p
t
i
m
i
z
a
t
i
o
n
-
b
a
se
d
d
e
e
p
n
e
u
r
a
l
n
e
t
w
o
r
k
mo
d
e
l
f
o
r
c
l
a
ssi
f
i
c
a
t
i
o
n
o
f
t
o
mat
o
p
l
a
n
t
d
i
se
a
se
s
u
si
n
g
G
P
U
,
”
J
o
u
r
n
a
l
o
f
Re
a
l
-
T
i
m
e
I
m
a
g
e
Pr
o
c
e
s
si
n
g
,
v
o
l
.
1
8
,
n
o
.
4
,
p
p
.
1
3
8
3
–
1
3
9
6
,
A
u
g
.
2
0
2
1
,
d
o
i
:
1
0
.
1
0
0
7
/
s
1
1
5
5
4
-
020
-
0
0
9
8
7
-
8.
[
2
0
]
J.
A
.
P
a
n
d
i
a
n
e
t
a
l
.
,
“
A
f
i
v
e
c
o
n
v
o
l
u
t
i
o
n
a
l
l
a
y
e
r
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
f
o
r
p
l
a
n
t
l
e
a
f
d
i
s
e
a
se
d
e
t
e
c
t
i
o
n
,
”
El
e
c
t
r
o
n
i
c
s
(
S
w
i
t
zer
l
a
n
d
)
,
v
o
l
.
1
1
,
n
o
.
8
,
p
.
1
2
6
6
,
A
p
r
.
2
0
2
2
,
d
o
i
:
1
0
.
3
3
9
0
/
e
l
e
c
t
r
o
n
i
c
s1
1
0
8
1
2
6
6
.
[
2
1
]
M
.
P
r
a
sa
n
n
a
k
u
m
a
r
a
n
d
K
.
L
a
t
h
a
,
“
P
l
a
n
t
d
i
se
a
se
i
d
e
n
t
i
f
i
c
a
t
i
o
n
u
si
n
g
c
o
n
t
e
x
t
u
a
l
mas
k
a
u
t
o
-
e
n
c
o
d
e
r
o
p
t
i
m
i
z
e
d
w
i
t
h
d
y
n
a
m
i
c
d
i
f
f
e
r
e
n
t
i
a
l
a
n
n
e
a
l
e
d
o
p
t
i
m
i
z
a
t
i
o
n
a
l
g
o
r
i
t
h
m,
”
Mi
c
r
o
sc
o
p
y
Re
s
e
a
r
c
h
a
n
d
T
e
c
h
n
i
q
u
e
,
v
o
l
.
8
7
,
n
o
.
3
,
p
p
.
4
8
4
–
4
9
4
,
M
a
r
.
2
0
2
4
,
d
o
i
:
1
0
.
1
0
0
2
/
j
e
mt
.
2
4
4
5
1
.
[
2
2
]
B
.
D
u
d
i
a
n
d
V
.
R
a
j
e
sh
,
“
O
p
t
i
m
i
z
e
d
t
h
r
e
sh
o
l
d
-
b
a
se
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
f
o
r
p
l
a
n
t
l
e
a
f
c
l
a
ssi
f
i
c
a
t
i
o
n
:
a
c
h
a
l
l
e
n
g
e
t
o
w
a
r
d
s
u
n
t
r
a
i
n
e
d
d
a
t
a
,
”
J
o
u
r
n
a
l
o
f
C
o
m
b
i
n
a
t
o
ri
a
l
O
p
t
i
m
i
za
t
i
o
n
,
v
o
l
.
4
3
,
n
o
.
2
,
p
p
.
3
1
2
–
3
4
9
,
2
0
2
2
,
d
o
i
:
1
0
.
1
0
0
7
/
s
1
0
8
7
8
-
0
2
1
-
0
0
7
7
0
-
w.
[
2
3
]
A
.
H
a
l
i
m,
C
.
C
h
o
w
,
M
.
A
mab
e
l
,
S
.
A
c
h
ma
d
,
a
n
d
R
.
S
u
t
o
y
o
,
“
T
h
e
i
mp
a
c
t
o
f
h
y
p
e
r
p
a
r
a
me
t
e
r
t
u
n
i
n
g
i
n
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
o
n
i
mag
e
c
l
a
ssi
f
i
c
a
t
i
o
n
mo
d
e
l
:
a
c
a
se
st
u
d
y
o
f
p
l
a
n
t
d
i
se
a
se
d
e
t
e
c
t
i
o
n
,
”
i
n
2
0
2
3
5
t
h
I
n
t
e
rn
a
t
i
o
n
a
l
C
o
n
f
e
r
e
n
c
e
o
n
C
y
b
e
r
n
e
t
i
c
s
a
n
d
I
n
t
e
l
l
i
g
e
n
t
S
y
st
e
m
s,
I
C
O
RI
S
2
0
2
3
,
O
c
t
.
2
0
2
3
,
p
p
.
1
–
6
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
O
R
I
S
6
0
1
1
8
.
2
0
2
3
.
1
0
3
5
2
2
0
9
.
[
2
4
]
E.
A
k
k
u
ş
,
U
.
B
a
l
,
F
.
Ö
.
K
o
ç
o
ğ
l
u
,
a
n
d
S
.
B
e
y
h
a
n
,
“
H
y
p
e
r
p
a
r
a
me
t
e
r
o
p
t
i
mi
z
a
t
i
o
n
o
f
p
r
e
-
t
r
a
i
n
e
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
u
si
n
g
a
d
o
l
e
sce
n
t
i
d
e
n
t
i
t
y
se
a
r
c
h
a
l
g
o
r
i
t
h
m
,
”
N
e
u
r
a
l
C
o
m
p
u
t
i
n
g
a
n
d
A
p
p
l
i
c
a
t
i
o
n
s
,
v
o
l
.
3
6
,
n
o
.
4
,
p
p
.
1
5
2
3
–
1
5
3
7
,
F
e
b
.
2
0
2
4
,
d
o
i
:
1
0
.
1
0
0
7
/
s
0
0
5
2
1
-
023
-
0
9
1
2
1
-
8.
[
2
5
]
E.
G
a
n
g
a
d
e
v
i
,
R
.
S
.
R
a
n
i
,
R
.
K
.
D
h
a
n
a
r
a
j
,
a
n
d
A
.
N
a
y
y
a
r
,
“
S
p
o
t
-
o
u
t
f
r
u
i
t
f
l
y
a
l
g
o
r
i
t
h
m
w
i
t
h
s
i
mu
l
a
t
e
d
a
n
n
e
a
l
i
n
g
o
p
t
i
m
i
z
e
d
S
V
M
f
o
r
d
e
t
e
c
t
i
n
g
t
o
m
a
t
o
p
l
a
n
t
d
i
se
a
se
s,”
N
e
u
r
a
l
C
o
m
p
u
t
i
n
g
a
n
d
A
p
p
l
i
c
a
t
i
o
n
s
,
v
o
l
.
3
6
,
n
o
.
8
,
p
p
.
4
3
4
9
–
4
3
7
5
,
M
a
r
.
2
0
2
4
,
d
o
i
:
1
0
.
1
0
0
7
/
s
0
0
5
2
1
-
023
-
0
9
2
9
5
-
1.
[
2
6
]
R
.
C
h
a
u
h
a
n
,
K
.
K
.
G
h
a
n
sh
a
l
a
,
a
n
d
R
.
C
.
Jo
sh
i
,
“
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
(
C
N
N
)
f
o
r
i
mag
e
d
e
t
e
c
t
i
o
n
a
n
d
r
e
c
o
g
n
i
t
i
o
n
,
”
i
n
I
C
S
C
C
C
2
0
1
8
-
1
s
t
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
S
e
c
u
re
C
y
b
e
r
C
o
m
p
u
t
i
n
g
a
n
d
C
o
m
m
u
n
i
c
a
t
i
o
n
s
,
D
e
c
.
2
0
1
8
,
p
p
.
2
7
8
–
2
8
2
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
S
C
C
C
.
2
0
1
8
.
8
7
0
3
3
1
6
.
[
2
7
]
O
.
C
.
O
ma
n
k
w
u
,
M
.
C
.
O
k
o
r
o
n
k
w
o
,
a
n
d
K
a
n
u
,
“
C
a
r
d
i
o
v
a
sc
u
l
a
r
(
h
e
a
r
t
)
d
i
se
a
se
s
p
r
e
d
i
c
t
i
o
n
u
s
i
n
g
d
e
e
p
l
e
a
r
n
i
n
g
n
e
u
r
a
l
n
e
t
w
o
r
k
mo
d
e
l
,
”
J
o
u
r
n
a
l
o
f
S
c
i
e
n
c
e
a
n
d
T
e
c
h
n
o
l
o
g
y
R
e
se
a
rc
h
,
v
o
l
.
6
,
n
o
.
1
,
p
p
.
2
0
2
4
–
2
0
5
2
,
2
0
2
4
,
[
O
n
l
i
n
e
]
.
A
v
a
i
l
a
b
l
e
:
h
t
t
p
s:
/
/
d
o
i
.
o
r
g
/
1
0
.
5
2
8
1
/
z
e
n
o
d
o
.
1
0
8
0
2
3
7
5
.
[
2
8
]
S
.
P
.
S
i
n
g
h
,
K
.
P
r
i
t
a
m
d
a
s
,
K
.
J.
D
e
v
i
,
a
n
d
S
.
D
.
D
e
v
i
,
“
C
u
s
t
o
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
f
o
r
d
e
t
e
c
t
i
o
n
a
n
d
c
l
a
ssi
f
i
c
a
t
i
o
n
o
f
r
i
c
e
p
l
a
n
t
d
i
se
a
se
s,”
Pr
o
c
e
d
i
a
C
o
m
p
u
t
e
r S
c
i
e
n
c
e
,
v
o
l
.
2
1
8
,
p
p
.
2
0
2
6
–
2
0
4
0
,
2
0
2
2
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
p
r
o
c
s.
2
0
2
3
.
0
1
.
1
7
9
.
Evaluation Warning : The document was created with Spire.PDF for Python.