I
nte
rna
t
io
na
l J
o
urna
l o
f
E
lect
rica
l a
nd
Co
m
pu
t
er
E
ng
ineering
(
I
J
E
CE
)
Vo
l.
15
,
No
.
2
,
A
p
r
il
20
25
,
p
p
.
2
4
1
6
~
2
4
2
6
I
SS
N:
2088
-
8
7
0
8
,
DOI
: 1
0
.
1
1
5
9
1
/ijece.
v
15
i
2
.
pp
2
4
1
6
-
2
4
2
6
2416
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
ec
e.
ia
esco
r
e.
co
m
O
ptimi
zing
conv
o
lutiona
l neural n
e
tworks
-
ba
sed
ens
emble
lea
rning
f
o
r e
ff
ec
t
iv
e herbal lea
f
dis
ea
se detec
tion
Ni
L
uh
Wiwik
Sri R
a
ha
y
u G
ina
ntr
a
,
Chris
t
ina
P
urna
m
a
Ya
nti,
M
a
de
Su
ci
Aria
nti
ni
D
e
p
a
r
t
me
n
t
o
f
I
n
f
o
r
m
a
t
i
c
s
En
g
i
n
e
e
r
i
n
g
,
F
a
c
u
l
t
y
o
f
Te
c
h
n
o
l
o
g
y
a
n
d
I
n
f
o
r
m
a
t
i
c
s
,
I
n
st
i
t
u
t
B
i
sn
i
s
d
a
n
T
e
k
n
o
l
o
g
i
I
n
d
o
n
e
si
a
,
D
e
n
p
a
sar
,
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
J
u
n
2
6
,
2
0
2
4
R
ev
is
ed
Dec
9
,
2
0
2
4
Acc
ep
ted
Dec
1
9
,
2
0
2
4
Th
is
stu
d
y
a
ims
to
o
p
ti
m
ize
c
o
n
v
o
lu
ti
o
n
a
l
n
e
u
ra
l
n
e
two
r
k
s
(C
NN
)
-
b
a
se
d
e
n
se
m
b
le
lea
rn
in
g
m
o
d
e
ls
to
e
n
h
a
n
c
e
a
c
c
u
ra
c
y
a
n
d
sta
b
il
it
y
in
d
e
tec
ti
n
g
h
e
rb
a
l
lea
f
d
ise
a
se
s.
Th
e
d
a
tas
e
t
u
se
d
in
th
is
st
u
d
y
is
so
u
rc
e
d
fro
m
t
h
e
“
Lo
n
tar
Taru
P
ra
m
a
n
a
”
c
o
ll
e
c
ti
o
n
,
wh
ic
h
in
c
l
u
d
e
s
v
a
ri
o
u
s
ima
g
e
s
o
f
h
e
rb
a
l
lea
v
e
s
a
ffe
c
ted
b
y
d
iffere
n
t
d
ise
a
se
s
su
c
h
a
s
a
n
c
a
k
b
a
c
teria
l
sp
o
t,
d
a
p
d
a
p
m
o
sa
ic
v
iru
s,
a
n
d
k
e
lo
r
p
o
wd
e
r
y
m
il
d
e
w
.
S
e
v
e
ra
l
CNN
m
o
d
e
ls,
in
c
l
u
d
i
n
g
VG
G
1
6
,
A
l
e
x
N
e
t
,
R
e
s
Ne
t
5
0
,
D
e
n
s
e
N
e
t
1
2
1
,
M
o
b
i
l
e
N
e
t
V
2
,
a
n
d
I
n
c
e
p
t
i
o
n
V
2
,
we
re
e
v
a
lu
a
ted
.
Am
o
n
g
th
e
se
,
t
h
e
e
n
se
m
b
le
m
o
d
e
ls
c
o
m
b
i
n
in
g
VG
G
1
6
,
De
n
se
Ne
t1
2
1
,
a
n
d
M
o
b
il
e
Ne
tV
2
we
re
se
lec
ted
d
u
e
t
o
th
e
ir
su
p
e
ri
o
r
p
e
rfo
rm
a
n
c
e
.
Th
e
e
n
se
m
b
le
m
o
d
e
l
a
c
h
iev
e
d
p
re
c
isio
n
sc
o
re
s
o
f
0
.
8
1
f
o
r
c
las
s
1
,
0
.
7
6
f
o
r
c
las
s
2
,
a
n
d
0
.
7
8
fo
r
c
las
s
3
,
wi
th
c
o
rre
sp
o
n
d
i
n
g
re
c
a
ll
sc
o
re
s
o
f
0
.
8
1
6
7
,
0
.
7
4
,
a
n
d
0
.
7
6
3
3
,
a
n
d
F
1
-
sc
o
re
s
o
f
0
.
8
1
3
3
,
0
.
7
5
,
a
n
d
0
.
7
7
1
7
re
sp
e
c
ti
v
e
l
y
.
T
h
e
se
re
su
lt
s
i
n
d
ica
te
sig
n
ifi
c
a
n
t
imp
ro
v
e
m
e
n
ts
in
m
o
d
e
l
p
e
rf
o
rm
a
n
c
e
a
n
d
r
o
b
u
stn
e
s
s.
K
ey
w
o
r
d
s
:
Den
s
eNe
t1
2
1
E
n
s
em
b
le
lear
n
in
g
Her
b
al
leaf
d
is
ea
s
e
d
etec
tio
n
Mo
b
ileNetV2
VGG1
6
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
:
Ni
L
u
h
W
iwik
Sri
R
ah
ay
u
Gin
an
tr
a
Dep
ar
tm
en
t o
f
I
n
f
o
r
m
atics E
n
g
in
ee
r
in
g
,
Facu
lty
o
f
T
ec
h
n
o
lo
g
y
an
d
I
n
f
o
r
m
atics,
I
n
s
titu
t Bi
s
n
is
d
an
T
ek
n
o
lo
g
i I
n
d
o
n
esia
J
alan
T
u
k
ad
Pak
er
is
an
No
.
9
7
,
Pan
jer
,
So
u
th
Den
p
asar
,
Den
p
asar
C
ity
,
B
ali,
8
0
2
2
5
,
I
n
d
o
n
esia
E
m
ail: w
iwik
@
in
s
tik
i.a
c.
id
1.
I
NT
RO
D
UCT
I
O
N
B
ali
b
o
asts
an
in
v
alu
a
b
le
we
alth
o
f
lo
ca
l
wis
d
o
m
,
o
n
e
o
f
wh
ich
is
tr
ad
itio
n
al
m
ed
icin
e
k
n
o
wn
as
Usad
a
B
ali
[
1
]
.
T
h
e
L
o
n
tar
T
ar
u
Pra
m
an
a
is
a
p
r
o
m
in
en
t
tr
ad
itio
n
al
m
ed
ical
tex
t
th
at
ca
talo
g
s
1
6
8
ty
p
es
o
f
h
er
b
al
p
la
n
ts
u
s
ed
in
v
ar
io
u
s
h
ea
lth
th
er
ap
ies
[
2
]
.
H
o
wev
er
,
m
o
d
e
r
n
izatio
n
a
n
d
e
n
v
ir
o
n
m
en
tal
ch
an
g
es
h
av
e
s
h
if
ted
tr
ad
itio
n
al
m
ed
icin
e
t
o
war
d
s
co
n
v
en
tio
n
al
tr
ea
tm
e
n
ts
,
co
u
p
le
d
with
p
lan
t
d
is
ea
s
es
th
at
th
r
ea
ten
th
e
q
u
ality
an
d
av
ailab
ilit
y
o
f
t
h
ese
h
er
b
al
p
lan
ts
.
L
ea
f
d
is
ea
s
es
af
f
ec
tin
g
h
er
b
al
p
lan
ts
s
u
ch
as
th
o
s
e
d
o
cu
m
en
te
d
in
th
e
L
o
n
tar
T
ar
u
Pra
m
an
a
p
o
s
e
a
s
er
io
u
s
t
h
r
ea
t
to
th
e
p
r
eser
v
atio
n
o
f
th
is
tr
a
d
itio
n
al
m
ed
icin
e,
m
ak
in
g
ef
f
ec
tiv
e
d
is
ea
s
e
d
etec
tio
n
an
d
co
n
tr
o
l
cr
u
cial
[
3
]
,
[
4
]
.
I
n
r
ec
en
t
y
ea
r
s
,
co
n
v
o
lu
tio
n
al
n
e
u
r
al
n
etwo
r
k
(
C
NN)
h
av
e
b
ec
o
m
e
th
e
lead
i
n
g
m
eth
o
d
f
o
r
p
la
n
t d
is
ea
s
e
d
etec
tio
n
an
d
class
if
icatio
n
d
u
e
to
th
eir
ab
ilit
y
to
r
ec
o
g
n
ize
p
atter
n
s
in
im
ag
es
with
h
ig
h
ac
cu
r
ac
y
[
5
]
–
[
9
]
.
Var
io
u
s
C
NN
m
o
d
els
s
u
ch
as
Alex
Net
[
1
0
]
–
[
1
2
]
,
VGGN
et
[
1
3
]
–
[
1
6
]
,
Go
o
g
L
eNe
t
[
1
7
]
,
R
esNet
[
1
8
]
–
[
2
0
]
,
I
n
ce
p
tio
n
V3
[
2
1
]
,
[
2
2
]
,
Den
s
eNe
t
[
2
3
]
,
[
2
4
]
,
Mo
b
ileNet
[
2
5
]
,
E
f
f
icien
tNet
[
2
6
]
,
NASNet
[
2
7
]
,
a
n
d
Xce
p
tio
n
[
2
8
]
h
a
v
e
b
ee
n
a
p
p
lied
t
o
a
r
an
g
e
o
f
im
ag
e
class
if
icatio
n
task
s
.
Ad
d
itio
n
ally
,
e
n
s
em
b
l
e
lear
n
in
g
m
eth
o
d
s
,
wh
ich
c
o
m
b
in
e
m
u
ltip
le
C
NN
m
o
d
e
ls
,
h
av
e
also
b
ee
n
u
tili
ze
d
to
en
h
a
n
ce
th
e
ac
cu
r
a
cy
an
d
r
o
b
u
s
tn
ess
o
f
m
o
d
els i
n
p
lan
t d
is
ea
s
e
d
etec
tio
n
[
2
9
]
,
[
3
0
]
.
Alth
o
u
g
h
C
NN
h
av
e
t
h
e
ad
v
an
tag
e
o
f
h
ig
h
ac
cu
r
ac
y
in
im
ag
e
class
if
icatio
n
s
u
ch
a
s
r
esear
ch
L
i
et
a
l.
[
1
2
]
,
co
m
b
i
n
ed
Alex
Net
with
I
n
ce
p
tio
n
V4
f
o
r
p
l
an
t
d
is
ea
s
e
d
iag
n
o
s
is
,
ac
h
iev
in
g
an
ac
cu
r
ac
y
o
f
9
4
.
6
%.
Me
an
w
h
ile,
Nex
t
r
es
ea
r
ch
Alataw
i
et
a
l.
[
1
3
]
u
s
ed
an
ar
tific
ial
in
tellig
en
ce
(
AI
)
b
ased
VGG
-
1
6
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8
7
0
8
Op
timiz
in
g
co
n
vo
lu
tio
n
a
l n
eu
r
a
l n
etw
o
r
ks
-
b
a
s
ed
en
s
emb
le
lea
r
n
in
g
…
(
N
i Lu
h
Wiw
ik
S
r
i R
a
h
a
yu
Gin
a
n
tr
a
)
2417
m
o
d
el
to
d
etec
t
p
la
n
t
d
is
ea
s
es
,
r
ea
ch
in
g
a
n
ac
cu
r
ac
y
o
f
9
1
.
8
%.
Nex
t
r
esear
ch
by
Pay
m
o
d
e
an
d
Ma
lo
d
e
[
1
4
]
ac
h
iev
ed
8
7
.
5
%
ac
c
u
r
ac
y
in
class
if
y
in
g
m
u
lti
-
cr
o
p
leaf
d
i
s
ea
s
e
im
ag
es
u
s
in
g
tr
an
s
f
er
lear
n
in
g
o
n
a
C
NN
VGG
m
o
d
el.
T
h
e
im
p
lem
e
n
tatio
n
o
f
a
p
r
etr
ain
ed
VGG1
6
m
o
d
el
b
y
Su
s
en
o
et
a
l.
[
1
5
]
f
o
r
r
ice
lea
f
d
is
ea
s
e
class
if
icatio
n
u
s
in
g
im
ag
e
s
eg
m
en
tatio
n
s
h
o
wed
a
n
ac
cu
r
a
cy
o
f
9
3
.
2
%.
I
n
a
c
o
m
p
ar
is
o
n
b
etwe
en
Alex
Net
an
d
VGG1
6
,
T
o
m
y
et
a
l.
[
1
6
]
f
o
u
n
d
th
at
VGG1
6
o
u
tp
e
r
f
o
r
m
ed
Alex
Net
with
an
ac
c
u
r
ac
y
o
f
9
0
%
co
m
p
ar
ed
to
8
8
%.
Nex
t
r
esear
ch
by
Ya
n
g
et
a
l.
[
1
7
]
id
e
n
tifie
d
r
ice
le
af
d
is
ea
s
es
u
s
in
g
Go
o
g
L
eNe
t
b
ased
o
n
a
r
esid
u
al
n
etwo
r
k
a
n
d
atten
tio
n
m
ec
h
a
n
is
m
,
ac
h
iev
in
g
an
ac
cu
r
ac
y
o
f
9
5
.
4
%.
R
esNet
-
b
ased
ap
p
r
o
ac
h
es
also
y
ield
e
d
s
ig
n
if
ican
t
r
esu
lts
.
Ku
m
ar
et
a
l.
[
1
8
]
r
ep
o
r
ted
an
ac
cu
r
ac
y
o
f
9
2
.
6
%,
an
d
R
ed
d
y
et
a
l.
[
1
9
]
u
s
ed
a
m
o
d
if
ied
r
ed
d
ee
r
o
p
tim
izatio
n
with
a
d
ee
p
lear
n
in
g
b
ased
co
n
v
o
lu
tio
n
al
n
e
u
r
al
n
etwo
r
k
s
(
DL
C
NN)
clas
s
if
ier
,
ac
h
iev
in
g
an
ac
cu
r
ac
y
o
f
9
4
.
3
%.
B
alav
an
i
et
a
l.
[
2
0
]
o
p
tim
ized
a
p
lan
t
d
is
ea
s
e
clas
s
if
i
ca
tio
n
s
y
s
tem
b
ased
o
n
R
esNet
-
5
0
ar
ch
itectu
r
e
an
d
tr
an
s
f
er
lear
n
in
g
,
r
esu
ltin
g
in
an
ac
cu
r
ac
y
o
f
9
3
.
5
%.
Oth
er
s
tu
d
ies
u
s
ed
I
n
ce
p
tio
n
V3
an
d
De
n
s
eNe
t
m
o
d
els.
Sam
ala
et
a
l.
[
2
1
]
ap
p
l
ied
I
n
ce
p
tio
n
V3
to
id
en
tif
y
to
m
ato
leaf
d
is
ea
s
es,
ac
h
iev
in
g
9
0
.
7
%
ac
cu
r
ac
y
,
w
h
ile
Qian
g
et
a
l.
[
2
2
]
ac
h
iev
e
d
8
9
.
5
%
ac
cu
r
ac
y
u
s
in
g
tr
an
s
f
e
r
lear
n
in
g
an
d
f
in
e
-
tu
n
in
g
o
n
I
n
ce
p
tio
n
V3
.
Nex
t
r
esear
ch
Pil
lai
et
a
l.
[
2
3
]
an
d
B
ak
r
et
a
l.
[
2
4
]
u
s
ed
Den
s
eNe
t
f
o
r
p
lan
t
d
is
ea
s
e
class
if
icatio
n
,
ac
h
iev
in
g
ac
cu
r
ac
ies
o
f
9
1
.
2
%
an
d
9
3
%,
r
esp
ec
tiv
ely
.
T
h
e
u
s
e
o
f
Mo
b
ileN
etV2
b
y
Z
ak
i
et
a
l.
[
2
5
]
f
o
r
class
if
y
in
g
to
m
ato
l
ea
f
d
is
ea
s
es
s
h
o
wed
an
ac
c
u
r
ac
y
o
f
8
8
.
3
%,
an
d
Atila
et
a
l.
[
2
6
]
u
s
ed
th
e
E
f
f
icien
tNet
d
ee
p
lear
n
i
n
g
m
o
d
el
f
o
r
p
lan
t
leaf
d
is
ea
s
e
class
if
icatio
n
,
ac
h
iev
in
g
9
4
%
ac
cu
r
ac
y
.
Ad
e
d
o
ja
et
a
l.
[
2
7
]
d
ev
elo
p
ed
a
n
in
tellig
en
t
m
o
b
ile
p
lan
t
d
is
ea
s
e
d
iag
n
o
s
tic
s
y
s
tem
u
s
in
g
NASNet
-
M
o
b
ile,
ac
h
iev
i
n
g
a
n
ac
cu
r
ac
y
o
f
9
2
.
5
%.
Fin
ally
,
Mo
id
an
d
C
h
au
r
asia
[
2
8
]
d
e
v
elo
p
ed
a
t
r
an
s
f
er
lear
n
in
g
-
b
ased
p
lan
t
d
is
ea
s
e
d
etec
tio
n
an
d
d
iag
n
o
s
is
s
y
s
te
m
u
s
in
g
Xce
p
tio
n
,
ac
h
iev
in
g
an
ac
cu
r
ac
y
o
f
9
0
.
8
%.
Ho
wev
e
r
,
th
e
ap
p
licatio
n
o
f
en
s
em
b
le
lear
n
in
g
f
o
r
h
e
r
b
al
p
lan
t
d
is
ea
s
es
r
em
ain
s
r
ar
e.
T
h
is
s
ca
r
city
h
as
r
esu
lted
in
s
u
b
o
p
tim
al
o
p
tim
izatio
n
an
d
ef
f
icien
c
y
o
f
m
o
d
els
in
th
is
s
p
ec
if
ic
co
n
tex
t.
Ad
d
itio
n
ally
,
C
NN
m
o
d
els
r
eq
u
ir
e
lar
g
e
d
atasets
an
d
h
ig
h
co
m
p
u
tatio
n
al
r
eso
u
r
ce
s
[
3
1
]
–
[
3
3
]
.
Fu
r
t
h
er
m
o
r
e
,
p
r
ev
io
u
s
r
esear
ch
h
as
n
o
t
ex
ten
s
iv
ely
f
o
cu
s
ed
o
n
o
p
tim
izin
g
C
NN
m
o
d
els
th
r
o
u
g
h
en
s
em
b
le
lea
r
n
in
g
,
p
ar
ticu
lar
ly
f
o
r
th
e
leaf
d
is
ea
s
es
o
f
h
er
b
al
p
lan
ts
d
o
cu
m
e
n
ted
in
t
h
e
L
o
n
tar
T
ar
u
Pra
m
an
a.
T
o
ad
d
r
ess
th
ese
is
s
u
es,
th
is
r
esear
ch
will
d
ev
elo
p
an
d
o
p
tim
ize
a
C
NN
-
b
ased
en
s
em
b
le
lear
n
i
n
g
m
o
d
el.
T
h
is
ap
p
r
o
ac
h
in
v
o
lv
es
co
m
b
in
i
n
g
m
u
ltip
le
C
NN
ar
ch
itectu
r
es
an
d
u
tili
zin
g
h
y
p
er
p
ar
am
eter
o
p
tim
izatio
n
tec
h
n
iq
u
es
an
d
d
ata
au
g
m
en
tatio
n
to
o
v
er
c
o
m
e
d
ata
lim
itatio
n
s
an
d
en
h
a
n
ce
m
o
d
e
l
p
er
f
o
r
m
an
ce
.
T
h
e
m
o
d
el
wil
l
b
e
th
o
r
o
u
g
h
ly
e
v
alu
ated
t
o
e
n
s
u
r
e
its
ac
cu
r
ac
y
an
d
ef
f
ec
tiv
e
n
ess
in
r
ea
l
-
wo
r
ld
co
n
d
itio
n
s
.
T
h
r
o
u
g
h
th
is
o
p
t
im
izatio
n
,
th
e
aim
is
to
p
r
o
d
u
ce
a
m
o
d
el
th
at
n
o
t
o
n
ly
ac
cu
r
ately
id
en
tifie
s
leaf
d
is
ea
s
es b
u
t a
ls
o
p
r
o
v
id
es e
f
f
e
ctiv
e
co
n
tr
o
l so
lu
tio
n
s
.
T
h
e
aim
o
f
th
is
r
esear
ch
is
t
o
d
ev
elo
p
a
n
d
o
p
tim
ize
an
e
f
f
ec
tiv
e
C
NN
-
b
ased
en
s
em
b
le
lear
n
in
g
m
o
d
el
f
o
r
d
etec
tin
g
leaf
d
is
ea
s
es
in
th
e
h
er
b
al
p
lan
ts
d
o
cu
m
en
ted
in
th
e
L
o
n
ta
r
T
a
r
u
Pra
m
a
n
a.
T
h
is
o
p
tim
izatio
n
is
ex
p
ec
ted
to
e
n
h
an
ce
th
e
ac
cu
r
ac
y
an
d
ef
f
i
cien
cy
o
f
lea
f
d
is
ea
s
e
id
en
tifi
ca
tio
n
an
d
p
r
o
v
id
e
tim
ely
an
d
ef
f
ec
tiv
e
co
n
tr
o
l
s
o
lu
tio
n
s
.
T
h
is
r
esear
ch
is
an
ti
cip
ated
to
o
f
f
er
s
ev
er
al
b
en
ef
i
ts
,
in
clu
d
in
g
aid
in
g
in
th
e
p
r
eser
v
atio
n
o
f
L
o
n
ta
r
T
ar
u
Pra
m
an
a
h
er
b
al
p
lan
t
s
as
a
f
u
n
d
am
e
n
tal
co
m
p
o
n
en
t
o
f
Usad
a
B
ali
tr
ad
itio
n
al
m
ed
icin
e,
im
p
r
o
v
i
n
g
f
ar
m
er
s
'
k
n
o
wled
g
e
an
d
s
k
ills
in
id
en
tify
in
g
an
d
m
a
n
ag
in
g
h
er
b
al
leaf
d
is
ea
s
es,
s
u
p
p
o
r
tin
g
th
e
p
r
eser
v
atio
n
o
f
B
ali'
s
lo
ca
l
cu
ltu
r
e,
an
d
p
r
o
v
id
in
g
a
leaf
d
is
ea
s
e
d
etec
tio
n
m
o
d
el
th
a
t
ca
n
b
e
ad
o
p
ted
f
o
r
o
th
er
h
er
b
al
p
lan
ts
.
T
h
is
r
esear
c
h
co
n
tr
i
b
u
tes
to
th
e
f
ield
s
o
f
c
o
m
p
u
t
atio
n
al
tech
n
o
lo
g
y
an
d
ag
r
icu
ltu
r
e,
p
ar
ticu
lar
l
y
in
th
e
ap
p
licatio
n
o
f
C
NN
-
b
ased
en
s
em
b
le
lear
n
in
g
m
o
d
el
s
f
o
r
p
lan
t
d
is
ea
s
e
d
etec
tio
n
.
Sp
ec
if
ic
c
o
n
tr
ib
u
tio
n
s
in
clu
d
e
th
e
d
ev
el
o
p
m
en
t
o
f
an
o
p
tim
ized
C
NN
en
s
em
b
le
lear
n
in
g
m
o
d
el
f
o
r
d
etec
tin
g
h
e
r
b
al
lea
f
d
is
ea
s
es,
th
e
ap
p
licatio
n
o
f
t
h
is
tech
n
o
lo
g
y
i
n
p
r
eser
v
in
g
tr
ad
i
tio
n
al
Usad
a
B
ali
m
ed
icin
e,
an
d
th
e
p
r
o
v
is
io
n
o
f
p
r
ac
tical
an
d
ef
f
ec
tiv
e
s
o
lu
ti
o
n
s
f
o
r
leaf
d
is
ea
s
e
co
n
tr
o
l,
wh
ich
ca
n
en
h
an
ce
th
e
p
r
o
d
u
ctiv
ity
an
d
q
u
ality
o
f
h
er
b
al
p
la
n
ts
.
2.
M
E
T
H
O
D
T
h
e
d
ataset
u
s
ed
in
th
is
s
tu
d
y
co
m
p
r
is
es
im
ag
es
o
f
h
er
b
al
leav
es
af
f
ec
ted
b
y
v
ar
io
u
s
d
is
ea
s
es,
s
o
u
r
ce
d
f
r
o
m
th
e
L
o
n
ta
r
T
a
r
u
Pra
m
an
a
co
llectio
n
.
T
h
is
d
ataset
co
n
tain
s
a
to
tal
o
f
9
3
i
m
ag
es,
d
iv
i
d
ed
i
n
to
1
3
d
is
tin
ct
class
es,
ea
ch
r
ep
r
esen
tin
g
d
if
f
er
e
n
t
d
is
ea
s
e
ca
teg
o
r
ies,
s
u
ch
as
an
ca
k
b
ac
ter
ial
s
p
o
t,
an
ca
k
leaf
s
p
o
t,
d
ap
d
a
p
b
ac
ter
ial
s
p
o
t,
d
ap
d
ap
leaf
s
p
o
t,
d
ap
d
a
p
m
o
s
a
ic
v
ir
u
s
,
k
elo
r
leaf
s
p
o
t,
k
elo
r
p
o
wd
er
y
m
ild
ew,
n
an
g
k
a
leaf
s
p
o
t,
n
a
n
g
k
a
p
o
wd
er
y
m
ild
ew,
n
a
n
g
k
a
tar
g
et
s
p
o
t,
s
ir
s
ak
b
ac
ter
ial
s
p
o
t,
s
ir
s
ak
leaf
s
p
o
t,
an
d
s
ir
s
ak
p
o
wd
er
y
m
ild
ew
.
E
ac
h
f
o
ld
er
with
in
th
e
d
ataset
co
n
tain
s
m
u
ltip
le
im
ag
es
s
p
ec
if
ic
to
th
e
d
is
ea
s
e
m
en
tio
n
ed
,
co
llected
u
n
d
e
r
v
ar
io
u
s
co
n
d
itio
n
s
to
ca
p
tu
r
e
d
iv
er
s
e
s
y
m
p
t
o
m
s
.
T
h
e
d
a
taset
was
s
p
lit
in
to
tr
ain
in
g
an
d
v
alid
atio
n
s
u
b
s
ets
u
s
in
g
an
8
0
-
2
0
s
p
lit,
r
esu
ltin
g
in
6
8
im
ag
es
b
ein
g
allo
ca
t
ed
f
o
r
tr
ain
in
g
a
n
d
1
3
im
ag
es
r
eser
v
e
d
f
o
r
v
alid
atio
n
,
wh
ile
m
ain
tain
in
g
th
e
class
d
is
tr
ib
u
tio
n
.
T
o
im
p
r
o
v
e
th
e
d
iv
er
s
ity
a
n
d
r
o
b
u
s
tn
ess
o
f
th
e
tr
ai
n
in
g
d
ata,
I
ma
g
eD
a
ta
Gen
era
to
r
was
u
s
ed
f
o
r
d
ata
a
u
g
m
en
tatio
n
,
ap
p
ly
in
g
tr
an
s
f
o
r
m
atio
n
s
s
u
c
h
as
r
escal
in
g
,
s
h
ea
r
,
zo
o
m
,
a
n
d
h
o
r
izo
n
tal
f
lip
p
in
g
.
T
h
is
ap
p
r
o
ac
h
en
h
an
ce
s
th
e
m
o
d
el'
s
ab
ilit
y
to
g
e
n
er
alize
ef
f
ec
tiv
e
ly
,
en
s
u
r
in
g
ac
cu
r
ate
class
if
icatio
n
o
f
u
n
s
ee
n
d
ata
a
n
d
r
e
d
u
cin
g
th
e
r
is
k
o
f
o
v
er
f
itti
n
g
[
3
4
]
.
As
s
h
o
wn
in
Fig
u
r
e
1
,
th
e
s
am
p
le
im
a
g
es
o
f
h
er
b
al
leaf
d
is
ea
s
es
ar
e
d
e
r
iv
ed
f
r
o
m
th
e
L
o
n
tar
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
15
,
No
.
2
,
Ap
r
il
20
25
:
2
4
1
6
-
2
4
2
6
2418
T
ar
u
Pra
m
an
a
co
llectio
n
,
f
o
r
m
in
g
a
cr
u
cial
p
ar
t
o
f
th
e
t
r
ain
in
g
d
ataset.
T
h
ese
im
a
g
es
s
er
v
e
as
a
f
o
u
n
d
atio
n
al
r
eso
u
r
ce
to
c
r
ea
te
a
m
o
r
e
d
iv
e
r
s
e
an
d
r
o
b
u
s
t d
ataset,
en
h
a
n
cin
g
th
e
m
o
d
el'
s
class
if
icatio
n
p
er
f
o
r
m
a
n
ce
.
T
h
e
r
esear
ch
f
r
am
ew
o
r
k
o
u
tli
n
es
th
e
s
tep
s
in
v
o
lv
ed
i
n
d
ete
ctin
g
h
er
b
al
leaf
d
is
ea
s
es
u
s
in
g
en
s
em
b
le
lear
n
in
g
b
ased
o
n
C
NN.
T
h
e
f
lo
wch
ar
t
p
r
o
v
id
es
a
d
etai
led
o
v
er
v
iew
o
f
th
e
e
n
tire
p
r
o
ce
s
s
,
f
r
o
m
d
ata
co
llectio
n
to
th
e
f
in
al
v
is
u
aliz
atio
n
o
f
r
esu
lts
.
Key
s
tag
es in
clu
d
e
m
o
d
el
s
elec
tio
n
,
tr
ain
i
n
g
,
an
d
ev
al
u
atio
n
,
all
aim
ed
at
en
s
u
r
in
g
th
e
m
o
d
el
ac
h
iev
es
o
p
tim
al
ac
cu
r
ac
y
.
T
h
e
d
iag
r
am
clar
if
ies
th
e
m
eth
o
d
o
lo
g
y
b
y
illu
s
tr
atin
g
h
o
w
d
ata
is
p
r
o
ce
s
s
ed
,
h
o
w
th
e
m
o
d
el
is
tr
ain
ed
,
h
o
w
p
r
ed
ictio
n
s
ar
e
co
m
b
in
e
d
th
r
o
u
g
h
en
s
em
b
l
e
lear
n
in
g
,
an
d
h
o
w
th
e
p
er
f
o
r
m
an
ce
o
f
th
e
en
s
em
b
le
m
o
d
el
is
ev
alu
ated
.
T
h
is
s
tr
u
ctu
r
ed
a
p
p
r
o
ac
h
en
s
u
r
es
th
e
ef
f
ec
tiv
e
d
etec
tio
n
o
f
h
er
b
al
l
ea
f
d
is
ea
s
es,
p
r
o
v
id
es
a
co
m
p
r
eh
en
s
iv
e
u
n
d
er
s
tan
d
in
g
o
f
th
e
r
esear
ch
p
r
o
ce
s
s
,
an
d
v
is
u
ally
s
u
m
m
a
r
izes th
e
wo
r
k
f
lo
w
in
Fig
u
r
e
2
.
Fig
u
r
e
1
.
Sam
p
le
o
f
h
er
b
al
lea
f
d
is
ea
s
es f
r
o
m
L
o
n
ta
r
T
ar
u
Pra
m
an
a
T
h
e
s
t
u
d
y
b
eg
i
n
s
w
i
t
h
th
e
c
o
l
l
e
c
t
io
n
a
n
d
p
r
ep
r
o
c
e
s
s
in
g
o
f
i
m
ag
e
s
o
f
h
e
r
b
a
l
l
ea
v
e
s
,
ea
c
h
e
x
h
ib
i
t
i
n
g
v
a
r
io
u
s
d
i
s
e
a
s
e
s
in
F
i
g
u
r
e
2
.
F
o
l
l
o
w
in
g
t
h
i
s
,
d
i
f
f
e
r
en
t
C
N
N
a
r
ch
i
t
e
c
tu
r
e
s
,
i
n
c
l
u
d
i
n
g
V
G
G1
6
,
A
l
e
x
N
e
t
,
R
e
s
N
e
t
5
0
,
D
en
s
e
Ne
t
1
2
1
,
M
o
b
i
l
e
N
e
t
V2
,
an
d
I
n
ce
p
t
i
o
n
V
2
,
a
r
e
s
e
l
e
c
te
d
f
o
r
tr
ai
n
i
n
g
.
E
a
ch
m
o
d
e
l
a
r
c
h
i
t
ec
t
u
r
e
i
s
f
ir
s
t
d
ef
i
n
ed
,
c
o
m
p
i
l
e
d
,
a
n
d
th
e
n
tr
a
i
n
e
d
u
s
i
n
g
t
h
e
p
r
o
c
e
s
s
e
d
d
a
ta
s
e
t
.
T
h
e
p
er
f
o
r
m
an
c
e
o
f
t
h
e
s
e
m
o
d
e
l
s
i
s
e
v
a
lu
a
t
ed
u
s
i
n
g
m
e
tr
i
c
s
s
u
c
h
a
s
p
r
ec
i
s
i
o
n
,
r
e
ca
l
l
,
a
n
d
F1
-
s
co
r
e
,
a
s
s
h
o
w
n
i
n
th
e
c
l
a
s
s
i
f
i
c
a
t
i
o
n
r
e
p
o
r
t
s
.
T
h
e
s
e
r
e
p
o
r
t
s
p
r
o
v
id
e
a
n
in
i
t
i
a
l
p
er
f
o
r
m
a
n
c
e
s
n
a
p
s
h
o
t
f
o
r
e
a
ch
m
o
d
e
l
b
e
f
o
r
e
an
d
a
f
t
e
r
o
p
t
im
i
z
a
t
i
o
n
.
Fo
r
ex
am
p
l
e
,
th
e
o
p
t
i
m
iz
e
d
V
G
G1
6
m
o
d
e
l
s
h
o
w
s
s
i
g
n
if
i
c
an
t
i
m
p
r
o
v
em
e
n
t
s
in
p
r
e
c
i
s
io
n
,
r
e
c
a
l
l,
a
n
d
F
1
-
s
co
r
e
c
o
m
p
a
r
ed
t
o
th
e
s
t
a
n
d
a
r
d
v
e
r
s
i
o
n
.
N
e
x
t,
e
n
s
e
m
b
l
e
l
e
ar
n
in
g
i
s
e
m
p
l
o
y
e
d
to
c
o
m
b
i
n
e
t
h
e
p
r
ed
i
c
t
io
n
s
f
r
o
m
e
ac
h
t
r
a
in
ed
m
o
d
e
l
.
T
h
e
g
o
a
l
o
f
en
s
e
m
b
l
e
le
a
r
n
i
n
g
i
s
to
en
h
a
n
ce
th
e
a
c
c
u
r
a
c
y
a
n
d
r
e
l
i
ab
i
l
i
t
y
o
f
d
is
e
a
s
e
d
e
t
ec
t
i
o
n
b
y
l
e
v
er
a
g
in
g
t
h
e
s
t
r
en
g
t
h
s
o
f
i
n
d
i
v
id
u
a
l
m
o
d
e
l
s
.
T
h
e
f
in
a
l
e
n
s
e
m
b
le
p
r
ed
i
c
t
io
n
i
s
o
b
ta
i
n
e
d
b
y
av
er
a
g
in
g
t
h
e
p
r
e
d
i
c
t
io
n
s
f
r
o
m
a
l
l
m
o
d
e
l
s
.
T
h
i
s
t
e
ch
n
i
q
u
e
e
n
s
u
r
e
s
th
a
t
t
h
e
w
e
ak
n
e
s
s
e
s
o
f
o
n
e
m
o
d
e
l
a
r
e
co
m
p
e
n
s
a
te
d
b
y
th
e
s
tr
en
g
t
h
s
o
f
a
n
o
t
h
er
,
le
a
d
in
g
t
o
m
o
r
e
a
c
cu
r
a
t
e
a
n
d
r
e
l
i
a
b
l
e
p
r
e
d
i
c
t
io
n
s
.
T
h
e
p
e
r
f
o
r
m
an
c
e
o
f
th
e
e
n
s
em
b
l
e
m
o
d
e
l
i
s
t
h
en
e
v
a
lu
a
t
ed
b
y
c
o
m
p
a
r
in
g
i
t
s
m
e
tr
i
c
s
w
i
t
h
t
h
o
s
e
o
f
t
h
e
i
n
d
i
v
id
u
al
m
o
d
e
l
s
.
T
h
e
f
o
l
l
o
w
in
g
i
s
g
e
n
er
a
l
f
o
r
m
u
la
f
o
r
th
e
en
s
e
m
b
l
e
p
r
e
d
i
c
t
io
n
(
)
u
s
i
n
g
t
h
e
s
i
m
p
le
av
e
r
ag
i
n
g
m
e
t
h
o
d
.
=
1
∑
=
1
(
1
)
wh
er
e
is
th
e
p
r
ed
ictio
n
f
r
o
m
t
h
e
−
ℎ
m
o
d
el
an
d
is
th
e
to
tal
n
u
m
b
er
o
f
m
o
d
els
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8
7
0
8
Op
timiz
in
g
co
n
vo
lu
tio
n
a
l n
eu
r
a
l n
etw
o
r
ks
-
b
a
s
ed
en
s
emb
le
lea
r
n
in
g
…
(
N
i Lu
h
Wiw
ik
S
r
i R
a
h
a
yu
Gin
a
n
tr
a
)
2419
Fig
u
r
e
2
.
R
esear
ch
f
r
am
ewo
r
k
3.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
N
Af
ter
co
m
p
letin
g
th
e
v
a
r
io
u
s
s
tep
s
in
th
e
r
esear
ch
p
r
o
ce
s
s
,
wh
ich
in
clu
d
e
d
ata
co
llectio
n
,
p
r
ep
r
o
ce
s
s
in
g
,
m
o
d
el
s
elec
tio
n
,
an
d
tr
ain
i
n
g
,
ea
ch
C
NN
m
o
d
el
was
in
d
ep
en
d
e
n
tly
tr
ain
e
d
u
s
in
g
a
d
ataset
o
f
im
ag
es
co
n
tain
in
g
h
er
b
al
leav
es
in
f
ec
ted
with
v
ar
io
u
s
d
is
ea
s
es.
T
h
is
d
ata
s
et
i
s
cr
u
cial
in
t
h
e
tr
ain
in
g
p
r
o
ce
s
s
as
it
p
r
o
v
i
d
es
a
r
a
n
g
e
o
f
e
x
am
p
les
th
at
allo
w
ea
ch
m
o
d
el
to
i
d
en
tify
a
n
d
u
n
d
e
r
s
tan
d
v
is
u
al
p
atter
n
s
ass
o
ciate
d
with
th
e
s
p
ec
if
ic
s
y
m
p
to
m
s
o
f
ea
ch
d
is
ea
s
e.
C
NN
m
o
d
els
lik
e
VGG1
6
,
Alex
Net,
R
esNet5
0
,
Den
s
eNe
t1
2
1
,
Mo
b
ileNetV2
,
an
d
I
n
ce
p
tio
n
V2
ar
e
tr
ain
e
d
in
d
iv
id
u
ally
t
o
s
tr
en
g
th
e
n
th
eir
ab
ilit
y
to
class
if
y
an
d
d
etec
t
d
is
ea
s
es
with
h
ig
h
ac
cu
r
ac
y
.
Du
r
in
g
t
h
e
tr
ain
in
g
p
r
o
ce
s
s
,
ea
ch
m
o
d
el
r
ec
eiv
es
a
s
er
ie
s
o
f
p
r
e
p
r
o
ce
s
s
ed
im
ag
es,
en
s
u
r
in
g
t
h
at
th
e
d
ata
b
ein
g
p
r
o
ce
s
s
ed
is
in
an
o
p
tim
al
f
o
r
m
at
f
o
r
an
aly
s
is
b
y
t
h
e
m
o
d
el.
T
h
r
o
u
g
h
th
is
tr
ain
in
g
,
th
e
m
o
d
els
ad
ju
s
t
th
eir
in
ter
n
al
p
ar
am
eter
s
,
s
u
ch
as
weig
h
ts
an
d
b
iases
,
to
p
r
o
v
id
e
m
o
r
e
ac
cu
r
ate
p
r
ed
ictio
n
s
in
t
h
e
task
o
f
d
is
ea
s
e
d
etec
tio
n
.
Af
ter
all
in
d
iv
id
u
al
m
o
d
els
ar
e
tr
ain
ed
,
th
e
n
ex
t
s
tep
is
to
b
u
ild
a
n
en
s
em
b
le
m
o
d
el.
T
h
is
en
s
em
b
le
m
o
d
el
is
a
co
m
b
in
atio
n
o
f
th
e
p
r
ed
ictio
n
s
g
en
er
ate
d
b
y
all
in
d
iv
id
u
al
C
NN
m
o
d
els,
allo
win
g
th
is
ap
p
r
o
ac
h
to
c
o
llectiv
ely
lev
e
r
ag
e
t
h
e
s
tr
en
g
th
s
o
f
ea
ch
m
o
d
el.
T
h
e
en
s
em
b
le
ap
p
r
o
ac
h
is
ex
p
ec
ted
to
p
r
o
d
u
ce
a
m
o
r
e
r
eliab
le
an
d
ac
cu
r
ate
s
y
s
tem
co
m
p
ar
ed
to
u
s
in
g
a
s
in
g
l
e
m
o
d
el,
as
it
ca
n
b
alan
ce
th
e
wea
k
n
ess
es
o
f
o
n
e
m
o
d
el
with
th
e
s
tr
en
g
th
s
o
f
an
o
th
e
r
.
T
h
e
r
esu
lts
o
f
th
e
e
n
s
em
b
le
m
o
d
el
ar
e
th
en
an
aly
ze
d
a
n
d
p
r
esen
ted
t
h
r
o
u
g
h
v
ar
i
o
u
s
p
er
f
o
r
m
an
ce
m
etr
ics,
d
em
o
n
s
tr
atin
g
h
o
w
ef
f
ec
tiv
e
th
is
ap
p
r
o
ac
h
is
in
im
p
r
o
v
in
g
t
h
e
o
v
er
all
ac
cu
r
ac
y
an
d
r
eliab
ilit
y
o
f
h
er
b
al
leaf
d
is
ea
s
e
d
etec
tio
n
.
T
h
e
u
s
e
o
f
an
en
s
em
b
le
m
o
d
el
n
o
t
o
n
l
y
en
h
an
ce
s
p
r
e
d
ictiv
e
ca
p
ab
ilit
ies
b
u
t
also
s
tr
en
g
th
en
s
th
e
s
y
s
tem
'
s
r
esil
ie
n
ce
to
v
a
r
iatio
n
s
i
n
d
ata
n
o
t p
r
ev
io
u
s
ly
e
n
co
u
n
ter
ed
.
3
.
1
.
Resul
t
s
I
n
th
is
s
tu
d
y
,
we
u
s
ed
s
p
ec
if
ic
h
y
p
er
p
ar
a
m
eter
s
to
o
p
tim
ize
th
e
tr
ain
i
n
g
o
f
C
NN
m
o
d
els.
T
h
e
v
alid
atio
n
s
p
lit
was
s
et
to
0
.
2
,
u
s
in
g
2
0
%
o
f
th
e
tr
ai
n
in
g
d
ata
f
o
r
v
alid
atio
n
to
m
o
n
it
o
r
p
e
r
f
o
r
m
an
ce
a
n
d
p
r
ev
en
t
o
v
er
f
itti
n
g
.
A
lear
n
in
g
r
ate
o
f
0
.
0
0
0
1
e
n
s
u
r
ed
s
m
all
,
s
tab
le
weig
h
t
u
p
d
ates.
T
h
e
Ad
am
o
p
tim
izer
was
ch
o
s
en
f
o
r
its
ef
f
icien
c
y
in
d
e
ep
lear
n
in
g
,
a
n
d
t
h
e
m
o
d
el
wa
s
tr
ain
ed
f
o
r
1
0
0
ep
o
c
h
s
to
th
o
r
o
u
g
h
ly
lea
r
n
d
ata
p
atter
n
s
.
T
h
ese
h
y
p
er
p
ar
am
et
er
s
wer
e
s
elec
ted
b
ased
o
n
p
r
elim
in
ar
y
ex
p
e
r
im
en
ts
an
d
e
x
is
tin
g
liter
atu
r
e
to
ac
h
iev
e
th
e
b
est
r
esu
lts
in
d
etec
tin
g
h
er
b
al
leaf
d
is
ea
s
es.
Her
e
is
a
d
etailed
an
aly
s
is
o
f
th
e
ac
cu
r
ac
y
an
d
lo
s
s
cu
r
v
es o
f
v
ar
io
u
s
d
ee
p
lear
n
in
g
m
o
d
els
in
Fig
u
r
e
3
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
15
,
No
.
2
,
Ap
r
il
20
25
:
2
4
1
6
-
2
4
2
6
2420
V
G
G
1
6
A
l
e
x
N
e
t
R
e
sN
e
t
5
0
D
e
n
seN
e
t
1
2
1
M
o
b
i
l
e
N
e
t
V
2
I
n
c
e
p
t
i
o
n
V
2
Fig
u
r
e
3
.
An
al
y
s
is
o
f
m
o
d
el
p
er
f
o
r
m
a
n
ce
b
ased
o
n
ac
cu
r
ac
y
an
d
lo
s
s
cu
r
v
es
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8
7
0
8
Op
timiz
in
g
co
n
vo
lu
tio
n
a
l n
eu
r
a
l n
etw
o
r
ks
-
b
a
s
ed
en
s
emb
le
lea
r
n
in
g
…
(
N
i Lu
h
Wiw
ik
S
r
i R
a
h
a
yu
Gin
a
n
tr
a
)
2421
T
h
e
VGG1
6
m
o
d
el
d
em
o
n
s
tr
ates
a
co
n
s
is
ten
t
in
cr
ea
s
e
in
ac
cu
r
ac
y
f
o
r
b
o
th
s
in
g
le
an
d
en
s
em
b
le
v
er
s
io
n
s
,
with
th
e
en
s
em
b
le
m
o
d
el
ac
h
iev
in
g
s
lig
h
tly
h
ig
h
er
ac
cu
r
ac
y
an
d
l
o
wer
tr
ain
in
g
lo
s
s
,
in
d
icatin
g
b
etter
g
en
er
aliza
tio
n
an
d
f
aster
co
n
v
e
r
g
en
ce
.
T
h
e
Alex
Net
m
o
d
el,
h
o
wev
er
,
s
h
o
ws
s
ig
n
if
ican
t
f
lu
ctu
atio
n
s
i
n
b
o
th
ac
c
u
r
ac
y
an
d
lo
s
s
,
h
ig
h
lig
h
tin
g
i
n
s
tab
ilit
y
.
Desp
ite
th
is
,
th
e
e
n
s
em
b
le
m
o
d
el
g
e
n
er
all
y
o
u
t
p
er
f
o
r
m
s
th
e
s
in
g
le
m
o
d
el
in
ter
m
s
o
f
ac
cu
r
ac
y
an
d
lo
s
s
,
s
u
g
g
esti
n
g
b
etter
o
v
er
all
p
e
r
f
o
r
m
an
ce
d
esp
ite
th
e
o
b
s
er
v
ed
v
ar
iab
ilit
y
[
3
5
]
.
R
esNet5
0
d
is
p
lay
s
a
s
tead
y
in
cr
ea
s
e
in
ac
c
u
r
ac
y
f
o
r
b
o
th
m
o
d
els,
with
t
h
e
en
s
em
b
le
m
o
d
el
p
er
f
o
r
m
in
g
s
lig
h
tly
b
etter
a
n
d
s
h
o
win
g
a
m
ar
g
in
ally
lo
wer
l
o
s
s
,
in
d
icatin
g
im
p
r
o
v
ed
p
er
f
o
r
m
an
ce
.
Similar
ly
,
th
e
Den
s
eNe
t1
2
1
m
o
d
el
ex
h
i
b
its
co
n
s
is
ten
t
an
d
s
tab
le
im
p
r
o
v
em
en
ts
i
n
ac
cu
r
ac
y
an
d
a
s
tead
y
d
ec
r
ea
s
e
in
lo
s
s
o
v
er
tim
e.
T
h
e
en
s
em
b
le
m
o
d
el
ac
h
iev
es
h
ig
h
er
ac
c
u
r
a
cy
an
d
lo
wer
lo
s
s
co
m
p
ar
ed
t
o
th
e
s
in
g
le
m
o
d
el,
in
d
icatin
g
s
u
p
er
io
r
g
e
n
er
aliza
tio
n
an
d
p
er
f
o
r
m
an
ce
.
Mo
b
ileNetV2
s
h
o
ws
a
s
tab
le
in
cr
e
ase
in
ac
cu
r
ac
y
f
o
r
b
o
th
m
o
d
els,
with
th
e
en
s
em
b
le
m
o
d
el
ac
h
iev
in
g
s
lig
h
tly
h
ig
h
er
ac
c
u
r
ac
y
a
n
d
lo
we
r
lo
s
s
,
s
u
g
g
esti
n
g
b
etter
p
er
f
o
r
m
an
ce
.
I
n
ce
p
tio
n
V
2
f
o
ll
o
ws
a
s
im
ilar
tr
en
d
,
with
b
o
th
m
o
d
els
s
h
o
win
g
s
tead
y
ac
cu
r
ac
y
im
p
r
o
v
em
en
ts
an
d
a
co
n
s
is
ten
t
d
ec
r
ea
s
e
in
l
o
s
s
.
T
h
e
en
s
em
b
le
m
o
d
el
ag
ai
n
ac
h
iev
es
s
lig
h
tly
b
etter
r
esu
lts
in
b
o
th
m
etr
ics.
T
h
e
en
s
em
b
le
m
o
d
els
g
en
er
a
lly
o
u
tp
e
r
f
o
r
m
s
in
g
le
m
o
d
els
in
ter
m
s
o
f
b
o
th
ac
c
u
r
ac
y
a
n
d
lo
s
s
,
in
d
icatin
g
b
etter
g
en
er
aliza
tio
n
an
d
s
tab
ilit
y
d
u
r
in
g
tr
ain
in
g
.
Mo
d
els
lik
e
Den
s
eNe
t1
2
1
,
Mo
b
ileNetV2
,
an
d
I
n
ce
p
tio
n
V
2
d
em
o
n
s
tr
ate
co
n
s
is
ten
t
an
d
s
tab
le
p
er
f
o
r
m
an
ce
im
p
r
o
v
em
en
ts
,
m
ak
in
g
th
em
r
elia
b
le
c
h
o
ices
f
o
r
f
u
r
t
h
er
d
ev
elo
p
m
e
n
t.
C
o
n
v
er
s
ely
,
Alex
Net
r
eq
u
ir
es
f
u
r
th
er
tu
n
i
n
g
an
d
o
p
tim
izatio
n
to
s
tab
ilize
its
p
er
f
o
r
m
an
ce
d
u
e
to
th
e
s
ig
n
if
ican
t
f
lu
ctu
atio
n
s
o
b
s
er
v
ed
.
Am
o
n
g
th
e
m
o
d
els
an
aly
ze
d
,
VGG1
6
,
Den
s
eNe
t1
2
1
,
an
d
Mo
b
ileNetV2
s
tan
d
o
u
t
as
to
p
p
er
f
o
r
m
e
r
s
,
s
h
o
win
g
s
ig
n
if
ic
an
t
im
p
r
o
v
e
m
en
ts
an
d
s
tab
ilit
y
,
p
ar
ticu
lar
ly
with
en
s
em
b
le
m
eth
o
d
s
.
B
ased
o
n
th
e
p
er
f
o
r
m
an
ce
c
o
m
p
ar
is
o
n
g
r
ap
h
s
o
f
d
if
f
er
e
n
t
d
ee
p
lear
n
in
g
m
o
d
els
b
ef
o
r
e
an
d
af
ter
o
p
tim
izatio
n
in
T
ab
le
1
an
d
Fig
u
r
e
4
,
t
h
e
VGG1
6
m
o
d
el
ex
h
ib
ited
s
u
b
s
tan
tial
im
p
r
o
v
em
e
n
t
ac
r
o
s
s
all
m
etr
ics
p
o
s
t
-
o
p
tim
izatio
n
,
with
th
e
m
icr
o
F1
-
s
co
r
e
i
n
cr
ea
s
in
g
f
r
o
m
0
.
5
4
to
0
.
6
2
,
th
e
m
ac
r
o
F1
-
s
co
r
e
f
r
o
m
0
.
4
1
to
0
.
5
1
,
an
d
th
e
weig
h
ted
F1
-
s
co
r
e
f
r
o
m
0
.
4
9
to
0
.
5
4
.
T
h
is
i
n
d
icate
s
th
at
th
e
o
p
tim
izatio
n
ap
p
lied
was
h
ig
h
l
y
ef
f
ec
tiv
e
in
en
h
a
n
cin
g
th
e
m
o
d
el’
s
p
er
f
o
r
m
an
ce
.
I
n
co
n
tr
ast,
th
e
Alex
Net
m
o
d
el
s
h
o
wed
a
n
o
ticea
b
le
d
ec
lin
e
in
p
er
f
o
r
m
an
ce
af
te
r
o
p
tim
iza
tio
n
.
T
h
e
m
icr
o
F1
-
s
co
r
e
d
r
o
p
p
ed
f
r
o
m
0
.
3
1
to
0
.
2
3
,
t
h
e
m
a
cr
o
F1
-
s
co
r
e
f
r
o
m
0
.
1
5
to
0
.
1
3
,
a
n
d
th
e
weig
h
te
d
F1
-
s
co
r
e
f
r
o
m
0
.
1
8
to
0
.
1
3
,
s
u
g
g
esti
n
g
th
at
th
e
o
p
tim
izatio
n
s
tr
ateg
y
u
s
ed
was
eith
er
u
n
s
u
itab
le
o
r
d
etr
im
en
tal
to
th
is
m
o
d
el.
Fo
r
R
esNet5
0
,
th
e
r
e
was
n
o
ch
a
n
g
e
in
p
er
f
o
r
m
an
ce
p
o
s
t
-
o
p
tim
izatio
n
,
with
all
m
etr
ics
r
em
ain
in
g
co
n
s
tan
t.
T
h
is
in
d
icate
s
th
at
th
e
o
p
tim
izatio
n
h
a
d
n
o
im
p
ac
t
o
n
th
e
m
o
d
el.
Similar
ly
,
Den
s
eNe
t1
2
1
s
h
o
wed
a
s
ig
n
if
ica
n
t
d
ec
r
ea
s
e
in
p
er
f
o
r
m
a
n
ce
f
o
llo
win
g
o
p
tim
izatio
n
.
T
h
e
m
icr
o
F1
-
s
co
r
e
d
ec
r
ea
s
ed
f
r
o
m
0
.
5
4
to
0
.
3
8
,
th
e
m
ac
r
o
F1
-
s
co
r
e
f
r
o
m
0
.
4
8
to
0
.
3
8
,
an
d
th
e
weig
h
te
d
F1
-
s
co
r
e
f
r
o
m
0
.
5
1
to
0
.
3
8
,
h
ig
h
lig
h
tin
g
th
at
th
e
a
p
p
lie
d
o
p
tim
izatio
n
n
eg
ativ
ely
a
f
f
ec
ted
th
e
m
o
d
el’
s
p
er
f
o
r
m
an
ce
.
B
o
th
Mo
b
ileN
etV2
an
d
I
n
ce
p
tio
n
V2
m
ai
n
tain
ed
s
tab
le
p
e
r
f
o
r
m
an
ce
b
ef
o
r
e
an
d
af
ter
o
p
tim
izatio
n
,
with
n
o
s
ig
n
if
ic
an
t
ch
an
g
es
in
th
eir
m
etr
ics.
T
h
is
s
tab
ilit
y
s
u
g
g
ests
th
at
th
e
o
p
tim
izatio
n
h
ad
n
o
ef
f
ec
t o
n
t
h
ese
m
o
d
els.
B
ased
o
n
a
d
etailed
an
aly
s
is
o
f
ac
c
u
r
ac
y
an
d
lo
s
s
co
m
p
ar
i
s
o
n
s
,
th
e
m
o
s
t
s
u
itab
le
e
n
s
em
b
le
m
o
d
els
f
o
r
h
e
r
b
al
leaf
d
is
ea
s
e
d
etec
t
io
n
ar
e
VGG1
6
,
De
n
s
eNe
t1
2
1
,
an
d
M
o
b
ileNetV2
.
T
h
e
e
n
s
em
b
le
m
o
d
els
o
f
VGG1
6
,
Den
s
eNe
t1
2
1
,
an
d
M
o
b
ileNetV2
co
n
s
is
ten
tly
o
u
tp
e
r
f
o
r
m
e
d
th
ei
r
s
in
g
le
m
o
d
el
co
u
n
ter
p
ar
ts
i
n
ter
m
s
o
f
ac
cu
r
ac
y
an
d
lo
s
s
.
T
h
is
in
d
icate
s
th
at
th
e
en
s
em
b
le
ap
p
r
o
ac
h
en
h
an
ce
s
b
o
th
p
er
f
o
r
m
a
n
ce
an
d
r
o
b
u
s
tn
ess
.
B
elo
w
is
th
e
ca
lcu
latio
n
r
ef
er
r
in
g
to
(
1
)
.
V
G
G
1
6
P
r
e
c
i
s
i
o
n
:
16
=
(
0
.
8
;
0
.
75
;
0
.
78
)
R
e
c
a
l
:
16
=
(
0
.
82
;
0
.
73
;
0
.
76
)
F1
-
S
c
o
r
e
:
1
16
=
(
0
.
81
;
0
.
74
;
0
.
77
)
D
e
n
seN
e
t
1
2
1
P
r
e
c
i
s
i
o
n
:
16
=
(
0
.
82
;
0
.
77
;
0
.
79
)
R
e
c
a
l
:
16
=
(
0
.
83
;
0
.
75
;
0
.
78
)
F1
-
S
c
o
r
e
:
1
16
=
(
0
.
82
;
0
.
76
;
0
.
78
)
M
o
b
i
l
e
N
e
t
V
2
P
r
e
c
i
s
i
o
n
:
16
=
(
0
.
81
;
0
.
76
;
0
.
77
)
R
e
c
a
l
:
16
=
(
0
.
80
;
0
.
74
;
0
.
75
)
F1
-
S
c
o
r
e
:
1
16
=
(
0
.
80
;
0
.
75
;
0
.
76
)
P
r
e
c
is
ion
F
or
e
a
c
h
c
las
s
(
)
=
16
(
)
+
121
(
)
+
2
(
)
3
C
las
s
1:
(
1
)
=
0
.
8
+
0
.
82
+
0
.
81
3
=
0
.
81
C
las
s
2:
(
2
)
=
0
.
75
+
0
.
77
+
0
.
76
3
=
0
.
76
C
las
s
3:
(
3
)
=
0
.
78
+
0
.
79
+
0
.
77
3
=
0
.
78
R
e
c
a
ll
F
or
e
a
c
h
c
las
s
(
)
=
16
(
)
+
121
(
)
+
2
(
)
3
C
las
s
1:
(
1
)
=
0
.
82
+
0
.
83
+
0
.
80
3
=
0
.
8167
C
las
s
2:
(
2
)
=
0
.
73
+
0
.
75
+
0
.
74
3
=
0
.
74
C
las
s
3:
(
3
)
=
0
.
76
+
0
.
78
+
0
.
75
3
=
0
.
7633
F1
-
S
c
or
e
F
or
e
a
c
h
c
las
s
1
(
)
=
1
16
(
)
+
1
121
(
)
+
1
2
(
)
3
C
las
s
1:
1
(
1
)
=
0
.
81
+
0
.
82
+
0
.
80
3
=
0
.
8133
C
las
s
2:
1
(
2
)
=
0
.
74
+
0
.
76
+
0
.
75
3
=
0
.
75
C
las
s
3:
1
(
3
)
=
0
.
77
+
0
.
78
+
0
.
76
3
=
0
.
7717
T
h
e
r
esu
lts
o
f
th
e
ca
lcu
latio
n
s
ar
e
p
r
esen
ted
in
T
ab
le
2
f
o
r
th
e
e
n
s
em
b
le
m
o
d
els
o
f
VGG1
6
,
Den
s
eNe
t1
2
1
,
an
d
M
o
b
ileNetV2
.
T
h
e
co
m
p
ar
is
o
n
o
f
i
n
d
iv
id
u
al
m
o
d
els
(
VGG1
6
,
Den
s
eNe
t1
2
1
,
an
d
Mo
b
ileNetV2
)
an
d
th
e
en
s
em
b
le
m
o
d
el
h
ig
h
li
g
h
ts
th
e
ef
f
ec
tiv
en
ess
o
f
th
e
en
s
em
b
le
ap
p
r
o
ac
h
in
im
p
r
o
v
in
g
th
e
p
er
f
o
r
m
an
ce
o
f
h
er
b
al
leaf
d
is
ea
s
e
d
etec
tio
n
.
T
h
e
en
s
em
b
le
m
o
d
el
ac
h
iev
ed
a
p
r
e
cisi
o
n
o
f
0
.
8
1
f
o
r
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
15
,
No
.
2
,
Ap
r
il
20
25
:
2
4
1
6
-
2
4
2
6
2422
class
1
,
0
.
7
6
f
o
r
class
2
,
a
n
d
0
.
7
8
f
o
r
class
3
,
d
em
o
n
s
tr
atin
g
b
alan
ce
d
ac
c
u
r
ac
y
ac
r
o
s
s
all
c
lass
es.
I
n
ter
m
s
o
f
r
ec
all,
th
e
en
s
em
b
le
m
o
d
el
p
e
r
f
o
r
m
e
d
b
etter
th
a
n
s
o
m
e
in
d
i
v
id
u
al
m
o
d
els,
with
s
co
r
es
o
f
0
.
8
1
6
7
f
o
r
class
1
,
0
.
7
4
f
o
r
class
2
,
an
d
0
.
7
6
3
3
f
o
r
class
3
,
in
d
icatin
g
its
ef
f
ec
tiv
en
ess
in
co
r
r
ec
tly
id
e
n
tify
in
g
t
r
u
e
p
o
s
itiv
e
ca
s
es.
T
h
e
F1
-
s
co
r
es
o
f
th
e
e
n
s
em
b
le
m
o
d
el
wer
e
also
im
p
r
ess
iv
e,
with
0
.
8
1
3
3
f
o
r
class
1
,
0
.
7
5
f
o
r
class
2
,
an
d
0
.
7
7
1
7
f
o
r
class
3
,
r
ef
lec
tin
g
a
well
-
b
alan
ce
d
p
er
f
o
r
m
a
n
ce
b
etwe
en
p
r
ec
is
io
n
an
d
r
e
ca
ll.
T
h
is
b
alan
ce
d
p
er
f
o
r
m
an
ce
is
cr
u
cial
f
o
r
p
r
ac
tical
ap
p
licatio
n
s
wh
er
e
b
o
th
f
alse
p
o
s
itiv
es
an
d
f
alse
n
eg
ativ
es
ca
n
h
av
e
s
ig
n
if
ican
t
im
p
licatio
n
s
.
B
y
co
m
b
in
in
g
th
e
s
tr
en
g
th
s
o
f
VGG1
6
,
Den
s
eNe
t1
2
1
,
an
d
Mo
b
ileNetV2
,
th
e
en
s
em
b
le
m
o
d
el
o
f
f
er
s
a
m
o
r
e
r
o
b
u
s
t a
n
d
ac
cu
r
ate
p
r
ed
ictio
n
s
y
s
tem
.
T
ab
le
1
.
Mo
d
el
p
er
f
o
r
m
a
n
ce
c
o
m
p
ar
is
o
n
M
o
d
e
l
M
i
c
r
o
F
1
(
S
t
a
n
d
a
r
d
)
M
i
c
r
o
F
1
(
O
p
t
i
m
i
z
a
t
i
o
n
)
M
a
c
r
o
F
1
(
S
t
a
n
d
a
r
d
)
M
a
c
r
o
F
1
(
O
p
t
i
m
i
z
a
t
i
o
n
)
W
e
i
g
h
t
e
d
F
1
(
S
t
a
n
d
a
r
d
)
W
e
i
g
h
t
e
d
F
1
(
O
p
t
i
m
i
z
a
t
i
o
n
)
V
G
G
1
6
0
.
5
4
0
.
6
2
0
.
4
1
0
.
5
1
0
.
4
9
0
.
5
4
A
l
e
x
N
e
t
0
.
3
1
0
.
2
3
0
.
1
5
0
.
1
3
0
.
1
8
0
.
1
3
R
e
sN
e
t
5
0
0
.
3
1
0
.
3
1
0
.
1
5
0
.
1
5
0
.
1
7
0
.
1
7
D
e
n
seN
e
t
1
2
1
0
.
5
4
0
.
3
8
0
.
4
8
0
.
3
8
0
.
5
1
0
.
3
8
M
o
b
i
l
e
N
e
t
V
2
0
.
5
4
0
.
5
4
0
.
4
8
0
.
4
8
0
.
4
9
0
.
4
9
I
n
c
e
p
t
i
o
n
V
2
0
.
5
4
0
.
5
4
0
.
4
8
0
.
4
8
0
.
4
9
0
.
4
9
Fig
u
r
e
4
.
An
al
y
s
is
o
f
m
o
d
el
p
er
f
o
r
m
a
n
ce
T
ab
le
2
.
th
e
c
o
m
p
a
r
is
o
n
tab
le
o
f
in
d
iv
i
d
u
al
m
o
d
els an
d
th
e
e
n
s
em
b
le
m
o
d
el
M
o
d
e
l
P
r
e
c
i
s
i
o
n
(
C
l
a
ss
1
)
P
r
e
c
i
s
i
o
n
(
C
l
a
ss
2
)
P
r
e
c
i
s
i
o
n
(
C
l
a
ss
3
)
R
e
c
a
l
l
(
C
l
a
ss
1
)
R
e
c
a
l
l
(
C
l
a
ss
2
)
R
e
c
a
l
l
(
C
l
a
ss
3
)
F1
-
S
c
o
r
e
(
C
l
a
ss
1
)
F1
-
S
c
o
r
e
(
C
l
a
ss
2
)
F1
-
S
c
o
r
e
(
C
l
a
ss
3
)
V
G
G
1
6
0
.
8
0
0
.
7
5
0
.
7
8
0
.
8
2
0
.
7
3
0
.
7
6
0
.
8
1
0
.
7
4
0
.
7
7
D
e
n
seN
e
t
1
2
1
0
.
8
2
0
.
7
7
0
.
7
9
0
.
8
3
0
.
7
5
0
.
7
8
0
.
8
2
5
0
.
7
6
0
.
7
8
5
M
o
b
i
l
e
N
e
t
V
2
0
.
8
1
0
.
7
6
0
.
7
7
0
.
8
0
0
.
7
4
0
.
7
5
0
.
8
0
5
0
.
7
5
0
.
7
6
En
se
mb
l
e
0
.
8
1
0
.
7
6
0
.
7
8
0
.
8
1
6
7
0
.
7
4
0
.
7
6
3
3
0
.
8
1
3
3
0
.
7
5
0
.
7
7
1
7
3
.
2
.
Dis
cu
s
s
io
n
T
h
is
s
tu
d
y
aim
e
d
to
o
p
tim
i
ze
C
NN
-
b
ased
en
s
em
b
le
lea
r
n
in
g
f
o
r
ef
f
ec
tiv
e
h
e
r
b
al
le
af
d
is
ea
s
e
d
etec
tio
n
.
B
ased
o
n
th
e
co
m
p
r
eh
en
s
iv
e
a
n
aly
s
is
,
th
e
o
p
ti
m
izatio
n
s
ap
p
lied
to
C
NN
m
o
d
els,
p
ar
ticu
lar
l
y
th
r
o
u
g
h
en
s
em
b
le
lear
n
in
g
m
eth
o
d
s
,
s
ig
n
if
ican
tly
im
p
r
o
v
ed
p
er
f
o
r
m
an
ce
.
Mo
d
els
s
u
ch
as
VGG1
6
,
Den
s
eNe
t1
2
1
,
an
d
Mo
b
ileNet
V2
s
h
o
wed
n
o
tab
le
en
h
a
n
ce
m
en
ts
af
ter
o
p
tim
izatio
n
,
with
en
s
em
b
le
lear
n
in
g
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8
7
0
8
Op
timiz
in
g
co
n
vo
lu
tio
n
a
l n
eu
r
a
l n
etw
o
r
ks
-
b
a
s
ed
en
s
emb
le
lea
r
n
in
g
…
(
N
i Lu
h
Wiw
ik
S
r
i R
a
h
a
yu
Gin
a
n
tr
a
)
2423
p
lay
in
g
a
cr
u
cial
r
o
le
i
n
ac
h
ie
v
in
g
b
etter
r
esu
lts
.
T
h
e
p
r
im
a
r
y
ad
v
a
n
tag
e
o
f
e
n
s
em
b
le
lear
n
in
g
is
its
ab
ilit
y
to
co
m
b
in
e
th
e
s
tr
en
g
th
s
o
f
m
u
l
tip
le
m
o
d
els
to
p
r
o
d
u
ce
m
o
r
e
s
tab
le
an
d
ac
cu
r
ate
p
er
f
o
r
m
a
n
ce
.
I
n
th
is
s
tu
d
y
,
en
s
em
b
le
m
o
d
els
co
n
s
is
ten
tly
o
u
tp
e
r
f
o
r
m
ed
s
in
g
le
m
o
d
els
i
n
ter
m
s
o
f
b
o
t
h
ac
c
u
r
ac
y
an
d
lo
s
s
,
d
em
o
n
s
tr
atin
g
th
e
ef
f
ec
tiv
en
ess
o
f
th
is
ap
p
r
o
ac
h
in
d
etec
tin
g
d
is
ea
s
es
in
h
er
b
al
leav
es.
E
n
s
em
b
le
lear
n
in
g
h
elp
s
r
e
d
u
ce
o
v
er
f
itti
n
g
an
d
im
p
r
o
v
es
th
e
m
o
d
el'
s
g
en
er
aliza
tio
n
ca
p
a
b
ilit
ies,
wh
ich
is
cr
itical
in
r
ea
l
-
wo
r
ld
ap
p
licatio
n
s
with
h
ig
h
d
ata
v
ar
iab
ilit
y
.
T
h
e
p
r
ac
tical
im
p
licatio
n
s
o
f
th
is
r
esear
ch
ar
e
s
ig
n
if
ican
t.
B
y
en
h
an
cin
g
m
o
d
el
p
er
f
o
r
m
an
ce
th
r
o
u
g
h
o
p
tim
izatio
n
an
d
th
e
u
s
e
o
f
e
n
s
em
b
le
lear
n
in
g
,
th
e
s
y
s
tem
f
o
r
d
etec
tin
g
h
er
b
al
leaf
d
is
ea
s
es
b
ec
o
m
es
m
o
r
e
ac
cu
r
ate
an
d
r
eliab
le.
T
h
is
is
cr
u
cial
f
o
r
f
ar
m
e
r
s
an
d
ag
r
icu
ltu
r
al
p
r
ac
titi
o
n
er
s
wh
o
d
e
p
en
d
o
n
tech
n
o
l
o
g
y
f
o
r
q
u
ick
a
n
d
ac
cu
r
ate
d
iag
n
o
s
is
to
im
p
r
o
v
e
cr
o
p
y
ield
a
n
d
q
u
ality
.
Ho
wev
er
,
th
er
e
ar
e
s
o
m
e
lim
itatio
n
s
in
th
is
s
tu
d
y
th
at
s
h
o
u
ld
b
e
n
o
ted
.
Fo
r
ex
am
p
le,
th
e
d
ataset
u
s
ed
was lim
ited
in
s
ize
an
d
v
ar
iety
,
wh
ich
ca
n
af
f
ec
t th
e
m
o
d
el'
s
ab
ilit
y
to
g
en
er
alize
t
o
v
ar
io
u
s
f
ield
co
n
d
itio
n
s
.
Fo
r
f
u
tu
r
e
r
esear
ch
,
it
is
r
ec
o
m
m
en
d
ed
to
ex
p
lo
r
e
ad
d
itio
n
al
o
p
tim
izatio
n
tech
n
i
q
u
es
an
d
m
o
r
e
co
m
p
lex
en
s
e
m
b
le
ar
ch
itectu
r
es.
I
n
teg
r
atin
g
tech
n
iq
u
es
s
u
ch
as
tr
an
s
f
er
lear
n
in
g
co
u
l
d
f
u
r
th
er
en
h
an
ce
m
o
d
el
p
e
r
f
o
r
m
a
n
ce
.
T
h
is
s
tu
d
y
s
u
cc
ess
f
u
lly
d
em
o
n
s
tr
ates
th
at
o
p
tim
izin
g
C
NN
-
b
ased
e
n
s
em
b
le
lear
n
in
g
ca
n
s
ig
n
if
ican
tl
y
im
p
r
o
v
e
t
h
e
p
er
f
o
r
m
a
n
ce
o
f
h
er
b
al
leaf
d
is
ea
s
e
d
etec
tio
n
.
B
y
u
s
in
g
en
s
em
b
le
lear
n
in
g
m
eth
o
d
s
,
m
o
d
els
s
u
ch
as
VGG1
6
,
Den
s
eNe
t1
2
1
,
an
d
Mo
b
ileNetV2
ac
h
iev
ed
h
ig
h
er
ac
cu
r
ac
y
an
d
m
o
r
e
s
tab
le
p
er
f
o
r
m
an
ce
,
m
a
k
in
g
th
em
ef
f
ec
tiv
e
to
o
ls
f
o
r
d
e
tectin
g
h
er
b
al
leaf
d
is
ea
s
es.
Her
e
ar
e
th
e
p
r
ed
icti
o
n
r
esu
lts
f
r
o
m
t
h
e
th
r
ee
en
s
e
m
b
le
m
o
d
els,
n
am
ely
VGG1
6
,
Den
s
eNe
t1
2
1
,
an
d
Mo
b
ileNetV2
,
as sh
o
wn
in
Fi
g
u
r
e
5
(
a)
to
(
c)
.
(
a
)
(
b
)
(
c
)
Fig
u
r
e
5
.
Pre
d
ictio
n
r
esu
lts
b
a
s
ed
o
n
th
e
b
est m
o
d
el
,
(
a
)
VGG1
6
en
s
em
b
le
lear
n
in
g
,
(
b
)
D
en
s
eNe
t1
2
1
en
s
em
b
le
lear
n
in
g
,
an
d
(
c)
Mo
b
ileNetV2
en
s
em
b
le
lear
n
in
g
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
15
,
No
.
2
,
Ap
r
il
20
25
:
2
4
1
6
-
2
4
2
6
2424
T
h
e
Fig
u
r
e
5
illu
s
tr
ates
th
e
p
r
ed
ictio
n
r
esu
lts
o
f
C
NN
-
b
ased
en
s
em
b
le
lear
n
in
g
m
o
d
els
f
o
r
d
etec
tin
g
d
is
ea
s
es
in
h
er
b
al
leav
es,
u
s
in
g
th
r
ee
ar
c
h
itectu
r
es:
VGG1
6
,
Den
s
eNe
t1
2
1
,
a
n
d
Mo
b
ileNetV2
.
C
o
r
r
ec
t
p
r
ed
ictio
n
s
ar
e
m
ar
k
e
d
in
g
r
ee
n
tex
t,
wh
ile
in
co
r
r
ec
t
o
n
es
ar
e
in
r
ed
.
T
h
e
VGG1
6
m
o
d
el
m
ak
es
s
o
m
e
co
r
r
ec
t
p
r
ed
ictio
n
s
b
u
t
s
till
h
as
er
r
o
r
s
,
in
d
icatin
g
a
n
ee
d
f
o
r
im
p
r
o
v
ed
ac
cu
r
ac
y
.
Den
s
eNe
t1
2
1
p
e
r
f
o
r
m
s
b
etter
,
with
m
o
r
e
co
r
r
ec
t
p
r
ed
ictio
n
s
d
u
e
t
o
its
ab
ilit
y
to
h
an
d
le
co
m
p
lex
f
ea
tu
r
es.
Mo
b
ileNetV2
also
s
h
o
ws
g
o
o
d
r
esu
lts
an
d
is
ef
f
icien
t
f
o
r
d
e
v
ices
with
lim
ited
r
eso
u
r
ce
s
.
T
h
e
ar
ticl
e
“
Op
tim
izin
g
C
NN
-
b
ased
en
s
em
b
le
lear
n
in
g
f
o
r
ef
f
ec
tiv
e
h
er
b
al
leaf
d
is
ea
s
e
d
etec
tio
n
”
h
ig
h
lig
h
ts
th
e
im
p
o
r
tan
ce
o
f
s
elec
tin
g
th
e
r
ig
h
t
co
n
v
o
lu
tio
n
al
n
eu
r
al
n
etwo
r
k
ar
ch
itectu
r
e
to
en
h
a
n
ce
d
etec
tio
n
ac
cu
r
ac
y
.
T
h
is
co
m
p
ar
is
o
n
h
elp
s
r
esear
ch
e
r
s
u
n
d
er
s
tan
d
th
e
s
tr
en
g
th
s
an
d
wea
k
n
ess
es
o
f
ea
ch
m
o
d
el
a
n
d
id
en
tify
w
ay
s
to
im
p
r
o
v
e
o
v
e
r
all
p
er
f
o
r
m
an
ce
,
wh
ich
is
b
en
ef
icial
f
o
r
s
u
s
tain
ab
le
ag
r
i
cu
ltu
r
e.
4.
CO
NCLU
SI
O
N
T
h
e
r
esear
ch
s
u
cc
ess
f
u
lly
d
em
o
n
s
tr
ated
th
at
o
p
tim
izin
g
C
NN
-
b
ased
en
s
em
b
le
lear
n
in
g
m
o
d
els
s
ig
n
if
ican
tly
im
p
r
o
v
es
t
h
e
p
er
f
o
r
m
a
n
ce
o
f
h
er
b
al
leaf
d
is
ea
s
e
d
etec
tio
n
.
Am
o
n
g
t
h
e
m
o
d
els
a
n
aly
ze
d
,
VGG1
6
,
Den
s
eNe
t1
2
1
,
an
d
Mo
b
ileNetV2
,
wh
en
u
s
ed
i
n
a
n
en
s
em
b
le,
p
r
o
v
id
ed
s
u
p
er
i
o
r
r
esu
lts
co
m
p
ar
e
d
to
th
eir
in
d
i
v
id
u
al
c
o
u
n
te
r
p
ar
ts
.
T
h
e
en
s
em
b
le
ap
p
r
o
ac
h
led
to
h
ig
h
er
p
r
ec
is
io
n
,
r
ec
all,
an
d
F1
-
s
co
r
es,
in
d
icatin
g
b
etter
ac
cu
r
ac
y
an
d
r
o
b
u
s
tn
e
s
s
.
T
h
is
en
h
an
ce
m
e
n
t
is
cr
u
c
ial
f
o
r
p
r
ac
tical
ag
r
icu
ltu
r
al
ap
p
licatio
n
s
wh
er
e
r
eliab
le
d
is
ea
s
e
d
etec
tio
n
ca
n
lead
to
b
etter
cr
o
p
m
an
ag
em
e
n
t
an
d
y
ield
im
p
r
o
v
e
m
en
t.
Fu
tu
r
e
r
esear
ch
co
u
l
d
ex
p
lo
r
e
ad
d
itio
n
al
o
p
tim
izatio
n
tech
n
iq
u
es
an
d
m
o
r
e
co
m
p
lex
en
s
em
b
le
ar
c
h
itectu
r
e
s
to
f
u
r
th
er
en
h
an
ce
m
o
d
el
p
er
f
o
r
m
a
n
ce
an
d
s
tab
i
lity
.
Ou
r
s
tu
d
y
u
n
d
er
s
co
r
es
t
h
e
p
o
ten
tial
o
f
e
n
s
em
b
le
lear
n
in
g
in
d
e
v
elo
p
in
g
ef
f
ec
tiv
e
an
d
r
eliab
le
s
y
s
tem
s
f
o
r
th
e
d
etec
tio
n
o
f
h
er
b
a
l
leaf
d
is
ea
s
es,
th
er
eb
y
s
u
p
p
o
r
tin
g
s
u
s
tain
ab
le
ag
r
icu
ltu
r
al
p
r
ac
tices.
ACK
NO
WL
E
DG
E
M
E
NT
S
T
h
e
r
esear
ch
er
s
wo
u
l
d
lik
e
t
o
th
an
k
t
h
e
Dir
ec
to
r
ate
Gen
e
r
al
o
f
Hig
h
er
E
d
u
ca
tio
n
,
R
esear
ch
,
an
d
T
ec
h
n
o
lo
g
y
o
f
th
e
Min
is
tr
y
o
f
E
d
u
ca
tio
n
,
C
u
ltu
r
e,
R
esear
ch
,
an
d
T
ec
h
n
o
lo
g
y
f
o
r
f
u
n
d
in
g
th
is
r
esear
ch
u
n
d
er
th
e
Fu
n
d
am
e
n
tal
B
asic Re
s
ea
r
ch
s
ch
em
e
in
2
0
2
4
.
RE
F
E
R
E
NC
E
S
[
1
]
P
.
S
.
A
n
d
i
l
a
,
I
.
G
.
Ti
r
t
a
,
T.
W
a
r
se
n
o
,
a
n
d
S
u
t
o
m
o
,
“
M
e
d
i
c
i
n
a
l
p
l
a
n
t
s
d
i
v
e
r
si
t
y
u
se
d
b
y
B
a
l
i
n
e
se
i
n
B
u
l
e
l
e
n
g
R
e
g
e
n
c
y
,
B
a
l
i
,
”
J
o
u
rn
a
l
o
f
T
r
o
p
i
c
a
l
B
i
o
d
i
v
e
r
si
t
y
a
n
d
Bi
o
t
e
c
h
n
o
l
o
g
y
,
v
o
l
.
8
,
n
o
.
1
,
p
p
.
1
–
1
8
,
2
0
2
3
,
d
o
i
:
1
0
.
2
2
1
4
6
/
j
t
b
b
.
7
3
3
0
3
.
[
2
]
I
.
G
.
L.
W
i
r
a
t
m
a
a
n
d
I
.
A
.
A
.
Y
u
l
i
a
mi
a
s
t
u
t
i
,
“
Et
h
n
o
c
h
e
mi
s
t
r
y
p
o
t
e
n
t
i
a
l
o
f
v
i
n
e
s
c
o
n
t
a
i
n
e
d
i
n
l
o
n
t
a
r
U
sa
d
a
Ta
r
u
P
r
a
ma
n
a
on
st
u
d
e
n
t
s’
sc
i
e
n
t
i
f
i
c
e
x
p
l
a
n
a
t
i
o
n
s
k
i
l
l
s
t
h
r
o
u
g
h
t
a
s
k
-
b
a
se
d
l
e
a
r
n
i
n
g
,
”
J
u
r
n
a
l
Pe
n
d
i
d
i
k
a
n
I
PA
I
n
d
o
n
e
s
i
a
,
v
o
l
.
1
2
,
n
o
.
2
,
p
p
.
2
0
8
–
2
2
0
,
2
0
2
3
,
d
o
i
:
1
0
.
1
5
2
9
4
/
j
p
i
i
.
v
1
2
i
2
.
4
2
8
2
6
.
[
3
]
H
.
W
a
n
g
,
Y
.
C
h
e
n
,
L.
W
a
n
g
,
Q
.
Li
u
,
S
.
Y
a
n
g
,
a
n
d
C
.
W
a
n
g
,
“
A
d
v
a
n
c
i
n
g
h
e
r
b
a
l
me
d
i
c
i
n
e
:
e
n
h
a
n
c
i
n
g
p
r
o
d
u
c
t
q
u
a
l
i
t
y
a
n
d
s
a
f
e
t
y
t
h
r
o
u
g
h
r
o
b
u
st
q
u
a
l
i
t
y
c
o
n
t
r
o
l
p
r
a
c
t
i
c
e
s
,
”
Fro
n
t
i
e
rs
i
n
P
h
a
rm
a
c
o
l
o
g
y
,
v
o
l
.
1
4
,
n
o
.
S
e
p
t
e
m
b
e
r
,
p
p
.
1
–
1
6
,
2
0
2
3
,
d
o
i
:
1
0
.
3
3
8
9
/
f
p
h
a
r
.
2
0
2
3
.
1
2
6
5
1
7
8
.
[
4
]
M
.
E
k
o
r
,
“
Th
e
g
r
o
w
i
n
g
u
s
e
o
f
h
e
r
b
a
l
me
d
i
c
i
n
e
s
:
i
ss
u
e
s
r
e
l
a
t
i
n
g
t
o
a
d
v
e
r
se
r
e
a
c
t
i
o
n
s
a
n
d
c
h
a
l
l
e
n
g
e
s
i
n
mo
n
i
t
o
r
i
n
g
safe
t
y
,
”
Fro
n
t
i
e
rs
i
n
N
e
u
ro
l
o
g
y
,
v
o
l
.
4
,
n
o
.
Ja
n
u
a
r
y
,
p
p
.
1
–
1
0
,
2
0
1
4
,
d
o
i
:
1
0
.
3
3
8
9
/
f
p
h
a
r
.
2
0
1
3
.
0
0
1
7
7
.
[
5
]
J.
B
o
u
l
e
n
t
,
S
.
F
o
u
c
h
e
r
,
J.
Th
é
a
u
,
a
n
d
P
.
L
.
S
t
-
C
h
a
r
l
e
s,
“
C
o
n
v
o
l
u
t
i
o
n
a
l
n
e
u
r
a
l
n
e
t
w
o
r
k
s
f
o
r
t
h
e
a
u
t
o
ma
t
i
c
i
d
e
n
t
i
f
i
c
a
t
i
o
n
o
f
p
l
a
n
t
d
i
s
e
a
se
s,
”
Fr
o
n
t
i
e
rs
i
n
Pl
a
n
t
S
c
i
e
n
c
e
,
v
o
l
.
1
0
,
n
o
.
Ju
l
y
,
2
0
1
9
,
d
o
i
:
1
0
.
3
3
8
9
/
f
p
l
s
.
2
0
1
9
.
0
0
9
4
1
.
[
6
]
B
.
T
u
g
r
u
l
,
E
.
El
f
a
t
i
mi
,
a
n
d
R
.
Er
y
i
g
i
t
,
“
C
o
n
v
o
l
u
t
i
o
n
a
l
n
e
u
r
a
l
n
e
t
w
o
r
k
s
i
n
d
e
t
e
c
t
i
o
n
o
f
p
l
a
n
t
l
e
a
f
d
i
se
a
ses
:
a
r
e
v
i
e
w
,
”
A
g
r
i
c
u
l
t
u
r
e
(
S
w
i
t
zer
l
a
n
d
)
,
v
o
l
.
1
2
,
n
o
.
8
,
2
0
2
2
,
d
o
i
:
1
0
.
3
3
9
0
/
a
g
r
i
c
u
l
t
u
r
e
1
2
0
8
1
1
9
2
.
[
7
]
Y
.
T
o
d
a
a
n
d
F
.
O
k
u
r
a
,
“
H
o
w
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
d
i
a
g
n
o
se
p
l
a
n
t
d
i
s
e
a
s
e
,
”
P
l
a
n
t
Ph
e
n
o
m
i
c
s
,
v
o
l
.
2
0
1
9
,
2
0
1
9
,
d
o
i
:
1
0
.
3
4
1
3
3
/
2
0
1
9
/
9
2
3
7
1
3
6
.
[
8
]
M
.
B
a
g
g
a
a
n
d
S
.
G
o
y
a
l
,
“
I
mag
e
-
b
a
s
e
d
d
e
t
e
c
t
i
o
n
a
n
d
c
l
a
ssi
f
i
c
a
t
i
o
n
o
f
p
l
a
n
t
d
i
sea
ses
u
s
i
n
g
d
e
e
p
l
e
a
r
n
i
n
g
:
S
t
a
t
e
-
of
-
t
h
e
-
a
r
t
r
e
v
i
e
w
,
”
U
rb
a
n
Ag
r
i
c
u
l
t
u
r
e
a
n
d
Re
g
i
o
n
a
l
F
o
o
d
S
y
s
t
e
m
s
,
v
o
l
.
9
,
n
o
.
1
,
2
0
2
4
,
d
o
i
:
1
0
.
1
0
0
2
/
u
a
r
2
.
2
0
0
5
3
.
[
9
]
M
.
I
f
t
i
k
h
a
r
,
I
.
A
.
K
a
n
d
h
r
o
,
N
.
K
a
u
sar
,
A
.
K
e
h
a
r
,
M
.
U
d
d
i
n
,
a
n
d
A
.
D
a
n
d
o
u
sh
,
“
P
l
a
n
t
d
i
sea
s
e
ma
n
a
g
e
me
n
t
:
a
f
i
n
e
-
t
u
n
e
d
e
n
h
a
n
c
e
d
C
N
N
a
p
p
r
o
a
c
h
w
i
t
h
m
o
b
i
l
e
a
p
p
i
n
t
e
g
r
a
t
i
o
n
f
o
r
e
a
r
l
y
d
e
t
e
c
t
i
o
n
a
n
d
c
l
a
ssi
f
i
c
a
t
i
o
n
,
”
Ar
t
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
R
e
v
i
e
w
,
v
o
l
.
5
7
,
n
o
.
7
,
2
0
2
4
,
d
o
i
:
1
0
.
1
0
0
7
/
s1
0
4
6
2
-
0
2
4
-
1
0
8
0
9
-
z.
[
1
0
]
M
.
H
e
m
a
,
T
.
S
a
r
a
sw
a
t
h
i
,
L.
G
o
p
i
n
a
t
h
,
A
.
K
o
ma
t
e
sw
a
r
a
n
,
K
.
La
k
s
h
mi
,
a
n
d
S
.
Y
o
g
e
s
h
w
a
r
i
,
“
P
l
a
n
t
d
i
sea
s
e
a
n
d
p
e
s
t
i
d
e
n
t
i
f
i
c
a
t
i
o
n
u
si
n
g
A
l
e
x
N
e
t
,
”
I
n
t
e
r
n
a
t
i
o
n
a
l
J
o
u
r
n
a
l
o
f
I
n
t
e
l
l
i
g
e
n
t
S
y
st
e
m
s
a
n
d
A
p
p
l
i
c
a
t
i
o
n
s
i
n
E
n
g
i
n
e
e
r
i
n
g
,
v
o
l
.
1
2
,
n
o
.
3
,
p
p
.
9
2
2
–
9
2
9
,
2
0
2
4
.
[
1
1
]
J.
K
h
a
t
i
b
S
u
l
a
i
ma
n
D
a
l
a
m
N
o
,
M
.
M
i
c
h
e
n
i
,
R
.
B
i
r
i
t
h
i
a
,
C
.
M
u
g
a
m
b
i
,
B
.
T
o
o
,
a
n
d
M
.
K
i
n
y
u
a
,
“
I
d
e
n
t
i
f
i
c
a
t
i
o
n
o
f
ma
i
z
e
l
e
a
f
d
i
s
e
a
se
s
u
si
n
g
s
u
p
p
o
r
t
v
e
c
t
o
r
ma
c
h
i
n
e
a
n
d
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
u
r
a
l
n
e
t
w
o
r
k
s
A
l
e
x
N
e
t
a
n
d
R
e
sN
e
t
5
0
,
”
I
n
d
o
n
e
si
a
n
J
o
u
rn
a
l
o
f
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
At
t
ri
b
u
t
i
o
n
,
v
o
l
.
1
2
,
n
o
.
4
,
p
p
.
2
0
2
3
–
1
5
1
8
,
2
0
2
3
.
[
1
2
]
Z.
Li
e
t
a
l
.
,
“
I
mp
r
o
v
e
d
A
l
e
x
N
e
t
w
i
t
h
I
n
c
e
p
t
i
o
n
-
V
4
f
o
r
p
l
a
n
t
d
i
s
e
a
s
e
d
i
a
g
n
o
s
i
s,
”
C
o
m
p
u
t
a
t
i
o
n
a
l
I
n
t
e
l
l
i
g
e
n
c
e
a
n
d
N
e
u
ro
s
c
i
e
n
c
e
,
v
o
l
.
2
0
2
2
,
2
0
2
2
,
d
o
i
:
1
0
.
1
1
5
5
/
2
0
2
2
/
5
8
6
2
6
0
0
.
[
1
3
]
A
.
A
.
A
l
a
t
a
w
i
,
S
.
M
.
A
l
o
ma
n
i
,
N
.
I
.
A
l
h
a
w
i
t
i
,
a
n
d
M
.
A
y
a
z
,
“
P
l
a
n
t
d
i
s
e
a
se
d
e
t
e
c
t
i
o
n
u
si
n
g
A
I
b
a
se
d
V
G
G
-
1
6
mo
d
e
l
,
”
I
n
t
e
r
n
a
t
i
o
n
a
l
J
o
u
r
n
a
l
o
f
A
d
v
a
n
c
e
d
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
a
n
d
A
p
p
l
i
c
a
t
i
o
n
s
,
v
o
l
.
1
3
,
n
o
.
4
,
p
p
.
7
1
8
–
7
2
7
,
2
0
2
2
,
d
o
i
:
1
0
.
1
4
5
6
9
/
I
JA
C
S
A
.
2
0
2
2
.
0
1
3
0
4
8
4
.
[
1
4
]
A
.
S
.
P
a
y
mo
d
e
a
n
d
V
.
B
.
M
a
l
o
d
e
,
“
Tr
a
n
sf
e
r
l
e
a
r
n
i
n
g
f
o
r
m
u
l
t
i
-
c
r
o
p
l
e
a
f
d
i
s
e
a
se
i
ma
g
e
c
l
a
ssi
f
i
c
a
t
i
o
n
u
s
i
n
g
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
u
r
a
l
n
e
t
w
o
r
k
V
G
G
,
”
Art
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
i
n
A
g
ri
c
u
l
t
u
re
,
v
o
l
.
6
,
p
p
.
2
3
–
3
3
,
2
0
2
2
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
a
i
i
a
.
2
0
2
1
.
1
2
.
0
0
2
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8
7
0
8
Op
timiz
in
g
co
n
vo
lu
tio
n
a
l n
eu
r
a
l n
etw
o
r
ks
-
b
a
s
ed
en
s
emb
le
lea
r
n
in
g
…
(
N
i Lu
h
Wiw
ik
S
r
i R
a
h
a
yu
Gin
a
n
tr
a
)
2425
[
1
5
]
J.
R
.
K
.
S
u
se
n
o
,
A
.
E.
M
i
n
a
r
n
o
,
a
n
d
Y
.
A
z
h
a
r
,
“
I
mp
l
e
m
e
n
t
a
t
i
o
n
o
f
p
r
e
t
r
a
i
n
e
d
V
G
G
1
6
mo
d
e
l
f
o
r
r
i
c
e
l
e
a
f
d
i
se
a
se
c
l
a
ssi
f
i
c
a
t
i
o
n
u
si
n
g
i
m
a
g
e
se
g
m
e
n
t
a
t
i
o
n
,
”
K
i
n
e
t
i
k
:
G
a
m
e
T
e
c
h
n
o
l
o
g
y
,
I
n
f
o
rm
a
t
i
o
n
S
y
s
t
e
m
,
C
o
m
p
u
t
e
r
N
e
t
w
o
rk
,
C
o
m
p
u
t
i
n
g
,
El
e
c
t
r
o
n
i
c
s
,
a
n
d
C
o
n
t
r
o
l
,
v
o
l
.
4
,
n
o
.
1
,
2
0
2
3
,
d
o
i
:
1
0
.
2
2
2
1
9
/
k
i
n
e
t
i
k
.
v
8
i
1
.
1
5
9
2
.
[
1
6
]
D
.
M
.
T
o
my
,
A
.
J
a
i
s
o
n
,
A
.
C
h
r
i
st
o
p
h
e
r
,
A
.
T
o
m
y
,
J.
Ja
c
o
b
,
a
n
d
A
.
H
a
r
sh
a
,
“
C
o
m
p
a
r
i
so
n
o
f
A
LEX
N
E
T
a
n
d
V
G
G
1
6
f
o
r
a
n
a
l
y
s
i
s
o
f
p
l
a
n
t
l
e
a
f
d
i
se
a
se,
”
i
n
2
0
2
1
I
n
t
e
rn
a
t
i
o
n
a
l
C
o
n
f
e
r
e
n
c
e
o
n
Ad
v
a
n
c
e
s
i
n
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
(
I
C
A
C
C
)
,
2
0
2
1
,
p
p
.
1
–
6
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
A
C
C
-
2
0
2
1
5
2
7
1
9
.
2
0
2
1
.
9
7
0
8
0
7
5
.
[
1
7
]
L.
Y
a
n
g
e
t
a
l
.
,
“
G
o
o
g
Le
N
e
t
b
a
se
d
o
n
r
e
si
d
u
a
l
n
e
t
w
o
r
k
a
n
d
a
t
t
e
n
t
i
o
n
m
e
c
h
a
n
i
sm
i
d
e
n
t
i
f
i
c
a
t
i
o
n
o
f
r
i
c
e
l
e
a
f
d
i
sea
ses,
”
C
o
m
p
u
t
e
rs
a
n
d
El
e
c
t
r
o
n
i
c
s i
n
Ag
r
i
c
u
l
t
u
re
,
v
o
l
.
2
0
4
,
p
.
1
0
7
5
4
3
,
2
0
2
3
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
c
o
mp
a
g
.
2
0
2
2
.
1
0
7
5
4
3
.
[
1
8
]
V
.
K
u
mar,
H
.
A
r
o
r
a
,
H
a
r
s
h
,
a
n
d
J.
S
i
so
d
i
a
,
“
R
e
sN
e
t
-
b
a
se
d
a
p
p
r
o
a
c
h
f
o
r
d
e
t
e
c
t
i
o
n
a
n
d
c
l
a
s
si
f
i
c
a
t
i
o
n
o
f
p
l
a
n
t
l
e
a
f
d
i
s
e
a
ses
,
”
i
n
2
0
2
0
I
n
t
e
rn
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
El
e
c
t
ro
n
i
c
s
a
n
d
S
u
st
a
i
n
a
b
l
e
C
o
m
m
u
n
i
c
a
t
i
o
n
S
y
st
e
m
s
(
I
C
ES
C
)
,
2
0
2
0
,
p
p
.
4
9
5
–
5
0
2
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
ESC
4
8
9
1
5
.
2
0
2
0
.
9
1
5
5
5
8
5
.
[
1
9
]
S
.
R
.
G
.
R
e
d
d
y
,
G
.
P
.
S
.
V
a
r
ma,
a
n
d
R
.
L.
D
a
v
u
l
u
r
i
,
“
R
e
sN
e
t
-
b
a
se
d
m
o
d
i
f
i
e
d
r
e
d
d
e
e
r
o
p
t
i
m
i
z
a
t
i
o
n
w
i
t
h
D
L
C
N
N
c
l
a
s
si
f
i
e
r
f
o
r
p
l
a
n
t
d
i
s
e
a
s
e
i
d
e
n
t
i
f
i
c
a
t
i
o
n
a
n
d
c
l
a
ssi
f
i
c
a
t
i
o
n
,
”
C
o
m
p
u
t
e
rs
a
n
d
El
e
c
t
r
i
c
a
l
E
n
g
i
n
e
e
ri
n
g
,
v
o
l
.
1
0
5
,
p
.
1
0
8
4
9
2
,
2
0
2
3
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
c
o
mp
e
l
e
c
e
n
g
.
2
0
2
2
.
1
0
8
4
9
2
.
[
2
0
]
K
.
B
a
l
a
v
a
n
i
,
D
.
S
r
i
r
a
m,
M
.
B
.
S
h
a
n
k
a
r
,
a
n
d
D
.
S
.
C
h
a
r
a
n
,
“
A
n
o
p
t
i
m
i
z
e
d
p
l
a
n
t
d
i
se
a
se
c
l
a
ss
i
f
i
c
a
t
i
o
n
s
y
st
e
m
b
a
sed
o
n
R
e
sN
e
t
-
5
0
a
r
c
h
i
t
e
c
t
u
r
e
a
n
d
t
r
a
n
sf
e
r
l
e
a
r
n
i
n
g
,
”
i
n
2
0
2
3
4
t
h
I
n
t
e
rn
a
t
i
o
n
a
l
C
o
n
f
e
r
e
n
c
e
f
o
r
Em
e
r
g
i
n
g
T
e
c
h
n
o
l
o
g
y
(
I
N
C
ET)
,
2
0
2
3
,
p
p
.
1
–
5
,
d
o
i
:
1
0
.
1
1
0
9
/
I
N
C
ET5
7
9
7
2
.
2
0
2
3
.
1
0
1
7
0
3
6
8
.
[
2
1
]
S
.
S
a
ma
l
a
,
N
.
B
h
a
v
i
t
h
,
R
.
B
a
n
g
,
D
.
K
.
R
a
o
,
C
.
R
.
P
r
a
sa
d
,
a
n
d
S
.
Y
a
l
a
b
a
k
a
,
“
D
i
s
e
a
s
e
i
d
e
n
t
i
f
i
c
a
t
i
o
n
i
n
t
o
m
a
t
o
l
e
a
v
e
s
u
si
n
g
I
n
c
e
p
t
i
o
n
V
3
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,
”
i
n
2
0
2
3
7
t
h
I
n
t
e
rn
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
T
re
n
d
s
i
n
E
l
e
c
t
r
o
n
i
c
s
a
n
d
I
n
f
o
rm
a
t
i
c
s
(
I
C
O
EI
)
,
2
0
2
3
,
p
p
.
8
6
5
–
8
7
0
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
O
EI
5
6
7
6
5
.
2
0
2
3
.
1
0
1
2
5
7
5
8
.
[
2
2
]
Z.
Q
i
a
n
g
,
L.
H
e
,
a
n
d
F
.
D
a
i
,
“
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
s
e
a
s
e
s
b
a
s
e
d
o
n
I
n
c
e
p
t
i
o
n
V
3
t
r
a
n
sf
e
r
l
e
a
r
n
i
n
g
a
n
d
f
i
n
e
-
t
u
n
i
n
g
,
”
i
n
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
m
a
r
t
C
i
t
y
a
n
d
I
n
f
o
rm
a
t
i
za
t
i
o
n
,
2
0
1
9
,
p
p
.
1
1
8
–
1
2
7
.
[
2
3
]
R
.
P
i
l
l
a
i
,
N
.
S
h
a
r
ma,
R
.
G
u
p
t
a
,
a
n
d
A
.
S
h
a
r
m
a
,
“
C
l
a
ssi
f
i
c
a
t
i
o
n
o
f
p
l
a
n
t
d
i
s
e
a
ses
u
s
i
n
g
D
e
n
seN
e
t
1
2
1
t
r
a
n
sf
e
r
l
e
a
r
n
i
n
g
mo
d
e
l
,
”
i
n
2
0
2
3
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
A
d
v
a
n
c
e
s
i
n
C
o
m
p
u
t
i
n
g
,
C
o
m
m
u
n
i
c
a
t
i
o
n
a
n
d
A
p
p
l
i
e
d
I
n
f
o
rm
a
t
i
c
s
(
AC
C
AI
)
,
2
0
2
3
,
p
p
.
1
–
6
,
d
o
i
:
1
0
.
1
1
0
9
/
A
C
C
A
I
5
8
2
2
1
.
2
0
2
3
.
1
0
2
0
0
4
0
1
.
[
2
4
]
M
.
B
a
k
r
,
S
.
A
b
d
e
l
-
G
a
b
e
r
,
M
.
N
a
sr
,
a
n
d
M
.
H
a
z
ma
n
,
“
D
e
n
seN
e
t
b
a
se
d
m
o
d
e
l
f
o
r
p
l
a
n
t
d
i
se
a
ses
d
i
a
g
n
o
s
i
s,
”
E
u
r
o
p
e
a
n
J
o
u
r
n
a
l
o
f
El
e
c
t
r
i
c
a
l
En
g
i
n
e
e
r
i
n
g
a
n
d
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
,
v
o
l
.
6
,
n
o
.
5
,
p
p
.
1
–
9
,
2
0
2
2
,
d
o
i
:
1
0
.
2
4
0
1
8
/
e
j
e
c
e
.
2
0
2
2
.
6
.
5
.
4
5
8
.
[
2
5
]
S
.
Z.
M
.
Za
k
i
,
M
.
A
s
y
r
a
f
Zu
l
k
i
f
l
e
y
,
M
.
M
o
h
d
S
t
o
f
a
,
N
.
A
.
M
.
K
a
mar
i
,
a
n
d
N
.
A
y
u
n
i
M
o
h
a
m
e
d
,
“
C
l
a
ss
i
f
i
c
a
t
i
o
n
o
f
t
o
m
a
t
o
l
e
a
f
d
i
s
e
a
se
s
u
s
i
n
g
M
o
b
i
l
e
N
e
t
v
2
,
”
I
AE
S
I
n
t
e
rn
a
t
i
o
n
a
l
J
o
u
r
n
a
l
o
f
Art
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
(
I
J
-
AI
)
,
v
o
l
.
9
,
n
o
.
2
,
p
p
.
2
9
0
–
2
9
6
,
J
u
n
.
2
0
2
0
,
d
o
i
:
1
0
.
1
1
5
9
1
/
i
j
a
i
.
v
9
.
i
2
.
p
p
2
9
0
-
2
9
6
.
[
2
6
]
Ü
.
A
t
i
l
a
,
M
.
U
ç
a
r
,
K
.
A
k
y
o
l
,
a
n
d
E.
U
ç
a
r
,
“
P
l
a
n
t
l
e
a
f
d
i
se
a
se
c
l
a
ss
i
f
i
c
a
t
i
o
n
u
si
n
g
E
f
f
i
c
i
e
n
t
N
e
t
d
e
e
p
l
e
a
r
n
i
n
g
m
o
d
e
l
,
”
Ec
o
l
o
g
i
c
a
l
I
n
f
o
rm
a
t
i
c
s
,
v
o
l
.
6
1
,
M
a
r
.
2
0
2
1
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
e
c
o
i
n
f
.
2
0
2
0
.
1
0
1
1
8
2
.
[
2
7
]
A
.
O
.
A
d
e
d
o
j
a
,
P
.
A
.
O
w
o
l
a
w
i
,
T
.
M
a
p
a
y
i
,
a
n
d
C
.
T
u
,
“
I
n
t
e
l
l
i
g
e
n
t
mo
b
i
l
e
p
l
a
n
t
d
i
s
e
a
s
e
d
i
a
g
n
o
st
i
c
sy
s
t
e
m
u
si
n
g
N
A
S
N
e
t
-
mo
b
i
l
e
d
e
e
p
l
e
a
r
n
i
n
g
,
”
I
AE
N
G
I
n
t
e
rn
a
t
i
o
n
a
l
J
o
u
rn
a
l
o
f
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
,
v
o
l
.
4
9
,
n
o
.
1
,
p
p
.
2
1
6
–
2
3
1
,
2
0
2
2
.
[
2
8
]
M
.
A
.
M
o
i
d
a
n
d
M
.
A
.
C
h
a
u
r
a
si
a
,
“
Tr
a
n
sf
e
r
l
e
a
r
n
i
n
g
-
b
a
se
d
p
l
a
n
t
d
i
s
e
a
s
e
d
e
t
e
c
t
i
o
n
a
n
d
d
i
a
g
n
o
si
s s
y
st
e
m
u
si
n
g
X
c
e
p
t
i
o
n
,
”
i
n
2
0
2
1
Fi
f
t
h
I
n
t
e
rn
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
I
-
S
MA
C
(
I
o
T
i
n
S
o
c
i
a
l
,
M
o
b
i
l
e
,
A
n
a
l
y
t
i
c
s
a
n
d
C
l
o
u
d
)
(
I
-
S
MA
C
)
,
2
0
2
1
,
p
p
.
1
–
5
,
d
o
i
:
1
0
.
1
1
0
9
/
I
-
S
M
A
C
5
2
3
3
0
.
2
0
2
1
.
9
6
4
0
6
9
4
.
[
2
9
]
B
.
B
u
a
k
u
m
e
t
a
l
.
,
“
Tw
o
-
st
a
g
e
e
n
s
e
mb
l
e
d
e
e
p
l
e
a
r
n
i
n
g
m
o
d
e
l
f
o
r
p
r
e
c
i
se
l
e
a
f
a
b
n
o
r
m
a
l
i
t
y
d
e
t
e
c
t
i
o
n
i
n
c
e
n
t
e
l
l
a
a
s
i
a
t
i
c
a
,
”
Ag
ri
En
g
i
n
e
e
r
i
n
g
,
v
o
l
.
6
,
n
o
.
1
,
p
p
.
6
2
0
–
6
4
4
,
2
0
2
4
,
d
o
i
:
1
0
.
3
3
9
0
/
a
g
r
i
e
n
g
i
n
e
e
r
i
n
g
6
0
1
0
0
3
7
.
[
3
0
]
S
.
P
u
d
u
m
a
l
a
r
a
n
d
S
.
M
u
t
h
u
r
a
mal
i
n
g
a
m,
“
H
y
d
r
a
:
A
n
e
n
s
e
m
b
l
e
d
e
e
p
l
e
a
r
n
i
n
g
r
e
c
o
g
n
i
t
i
o
n
m
o
d
e
l
f
o
r
p
l
a
n
t
d
i
s
e
a
ses
,
”
J
o
u
r
n
a
l
o
f
En
g
i
n
e
e
ri
n
g
Re
s
e
a
rc
h
(
K
u
w
a
i
t
)
,
n
o
.
S
e
p
t
e
m
b
e
r
,
2
0
2
4
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
j
e
r
.
2
0
2
3
.
0
9
.
0
3
3
.
[
3
1
]
L.
A
l
z
u
b
a
i
d
i
e
t
a
l
.
,
“
R
e
v
i
e
w
o
f
d
e
e
p
l
e
a
r
n
i
n
g
:
c
o
n
c
e
p
t
s,
C
N
N
a
r
c
h
i
t
e
c
t
u
r
e
s
,
c
h
a
l
l
e
n
g
e
s,
a
p
p
l
i
c
a
t
i
o
n
s
,
f
u
t
u
r
e
d
i
r
e
c
t
i
o
n
s,
”
J
o
u
rn
a
l
o
f
Bi
g
D
a
t
a
,
v
o
l
.
8
,
n
o
.
1
,
D
e
c
.
2
0
2
1
,
d
o
i
:
1
0
.
1
1
8
6
/
s4
0
5
3
7
-
021
-
0
0
4
4
4
-
8.
[
3
2
]
H
.
L
.
D
a
w
so
n
,
O
.
D
u
b
r
u
l
e
,
a
n
d
C
.
M
.
J
o
h
n
,
“
I
mp
a
c
t
o
f
d
a
t
a
set
s
i
z
e
a
n
d
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
u
r
a
l
n
e
t
w
o
r
k
a
r
c
h
i
t
e
c
t
u
r
e
o
n
t
r
a
n
sf
e
r
l
e
a
r
n
i
n
g
f
o
r
c
a
r
b
o
n
a
t
e
r
o
c
k
c
l
a
ss
i
f
i
c
a
t
i
o
n
,
”
C
o
m
p
u
t
e
rs
a
n
d
G
e
o
s
c
i
e
n
c
e
s
,
v
o
l
.
1
7
1
,
n
o
.
D
e
c
e
m
b
e
r
2
0
2
2
,
2
0
2
3
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
c
a
g
e
o
.
2
0
2
2
.
1
0
5
2
8
4
.
[
3
3
]
C
.
R
.
E
d
w
i
n
S
e
l
v
a
R
e
x
,
J.
A
n
n
r
o
se,
a
n
d
J
.
J
e
n
i
f
e
r
J
o
se
,
“
C
o
m
p
a
r
a
t
i
v
e
a
n
a
l
y
si
s
o
f
d
e
e
p
c
o
n
v
o
l
u
t
i
o
n
n
e
u
r
a
l
n
e
t
w
o
r
k
mo
d
e
l
s
o
n
smal
l
sc
a
l
e
d
a
t
a
se
t
s,
”
O
p
t
i
k
,
v
o
l
.
2
7
1
,
p
.
1
7
0
2
3
8
,
2
0
2
2
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
i
j
l
e
o
.
2
0
2
2
.
1
7
0
2
3
8
.
[
3
4
]
H
.
H
a
ss
a
n
e
t
a
l
.
,
“
R
e
v
i
e
w
a
n
d
c
l
a
ss
i
f
i
c
a
t
i
o
n
o
f
A
I
-
e
n
a
b
l
e
d
C
O
V
I
D
-
1
9
C
T
i
ma
g
i
n
g
mo
d
e
l
s
b
a
se
d
o
n
c
o
mp
u
t
e
r
v
i
s
i
o
n
t
a
sk
s
,
”
C
o
m
p
u
t
e
rs
i
n
B
i
o
l
o
g
y
a
n
d
Me
d
i
c
i
n
e
,
v
o
l
.
1
4
1
,
n
o
.
S
e
p
t
e
m
b
e
r
2
0
2
1
,
p
.
1
0
5
1
2
3
,
2
0
2
2
,
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
5
1
2
3
.
[
3
5
]
H.
-
K
.
L
e
,
V
.
-
S
.
D
o
a
n
,
a
n
d
V
.
-
P
.
H
o
a
n
g
,
“
E
n
sem
b
l
e
o
f
c
o
n
v
o
l
u
t
i
o
n
n
e
u
r
a
l
n
e
t
w
o
r
k
s
f
o
r
i
m
p
r
o
v
i
n
g
a
u
t
o
mat
i
c
m
o
d
u
l
a
t
i
o
n
c
l
a
ss
i
f
i
c
a
t
i
o
n
p
e
r
f
o
r
ma
n
c
e
,
”
J
o
u
rn
a
l
o
f
S
c
i
e
n
c
e
a
n
d
T
e
c
h
n
o
l
o
g
y
I
ss
u
e
o
n
I
n
f
o
rm
a
t
i
o
n
a
n
d
C
o
m
m
u
n
i
c
a
t
i
o
n
s
T
e
c
h
n
o
l
o
g
y
,
v
o
l
.
2
0
,
n
o
.
6
,
p
p
.
2
5
–
3
2
,
2
0
2
2
,
d
o
i
:
1
0
.
3
1
1
3
0
/
u
d
-
j
st
.
2
0
2
2
.
2
9
3
e
.
B
I
O
G
RAP
H
I
E
S O
F
AUTH
O
RS
Ni
Lu
h
Wiwik
S
r
i
Ra
h
a
y
u
G
i
n
a
n
tr
a
b
o
rn
in
Ba
n
g
li
o
n
M
a
rc
h
2
9
1
9
8
6
.
S
h
e
c
o
m
p
lete
d
h
e
r
m
a
ste
rs
in
c
o
m
p
u
ter
sc
ien
c
e
a
t
P
e
n
d
id
i
k
a
n
G
a
n
e
sh
a
Un
iv
e
rsity
S
in
g
a
ra
ja
in
2
0
1
4
.
T
h
e
a
u
t
h
o
r
h
a
s
b
e
e
n
a
n
a
c
ti
v
e
lec
tu
re
r
a
t
In
stit
u
t
Bisn
is
d
a
n
Tek
n
o
lo
g
i
In
d
o
n
e
sia
sin
c
e
2
0
1
6
,
tea
c
h
in
g
in
th
e
c
o
m
p
u
te
r
s
c
ien
c
e
.
He
r
re
se
a
rc
h
fo
c
u
se
s
o
n
a
rti
ficia
l
in
telli
g
e
n
c
e
(d
e
c
isio
n
su
p
p
o
r
t
sy
ste
m
s,
e
x
p
e
r
t
sy
ste
m
s,
d
a
ta
m
in
in
g
,
n
e
u
ra
l
n
e
two
rk
s,
f
u
z
z
y
l
o
g
ic,
d
e
e
p
lea
rn
in
g
,
a
n
d
g
e
n
e
ti
c
a
l
g
o
ri
th
m
s).
Ad
d
it
io
n
a
ll
y
,
t
h
e
a
u
th
o
r
h
a
s
se
rv
e
d
a
s
a
re
v
iew
e
r
fo
r
v
a
ri
o
u
s
n
a
ti
o
n
a
ll
y
a
c
c
re
d
it
e
d
jo
u
rn
a
ls
(
S
INTA
3
-
S
INTA
6
).
Sh
e
h
a
s
w
o
n
m
u
lt
i
p
le
re
se
a
rc
h
g
ra
n
t
p
ro
p
o
sa
ls
fr
o
m
DIK
TI
(twice
i
n
2
0
2
0
-
2
0
2
2
),
a
n
d
a
c
o
-
f
o
u
n
d
e
r
o
f
t
h
e
Ya
y
a
sa
n
Ad
witi
y
a
Ba
su
ra
ta
In
o
v
a
si
(Ya
y
a
sa
n
Ab
iv
a
si)
fo
u
n
d
a
ti
o
n
wit
h
fe
ll
o
w
p
r
o
fe
ss
o
rs.
S
h
e
c
a
n
b
e
c
o
n
tac
te
d
a
t
e
m
a
il
:
wiwik
@in
sti
k
i.
a
c
.
id
.
Evaluation Warning : The document was created with Spire.PDF for Python.