I
AE
S In
t
er
na
t
io
na
l J
o
urna
l o
f
Art
if
icia
l In
t
ellig
ence
(
I
J
-
AI
)
Vo
l.
14
,
No
.
3
,
J
u
n
e
2
0
2
5
,
p
p
.
2270
~
2
2
8
1
I
SS
N:
2
2
5
2
-
8
9
3
8
,
DOI
: 1
0
.
1
1
5
9
1
/ijai.v
14
.i
3
.
p
p
2
2
7
0
-
2
2
8
1
2270
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
a
i
.
ia
esco
r
e.
co
m
An optimiz
ed t
ra
nsfer learni
ng
-
ba
sed a
ppro
a
ch f
o
r
C
ro
cid
o
lo
m
ia
pa
v
o
na
na
la
rv
a
e clas
sifica
tion
Ris
na
wa
t
i
1
,
Ro
dia
h
2
,
Sa
rif
ud
din
M
a
de
nd
a
3
,
Dia
na
T
ri
S
u
s
et
ia
nin
g
t
ia
s
4
1
D
e
p
a
r
t
me
n
t
o
f
M
a
n
a
g
e
me
n
t
,
F
a
c
u
l
t
y
o
f
Ec
o
n
o
my
,
G
u
n
a
d
a
r
ma
U
n
i
v
e
r
s
i
t
y
,
D
e
p
o
k
,
I
n
d
o
n
e
si
a
2
D
e
p
a
r
t
me
n
t
o
f
I
n
f
o
r
mat
i
c
s
,
F
a
c
u
l
t
y
o
f
I
n
d
u
st
r
y
Te
c
h
n
o
l
o
g
y
,
G
u
n
a
d
a
r
ma
U
n
i
v
e
r
s
i
t
y
,
D
e
p
o
k
,
I
n
d
o
n
e
s
i
a
3
D
o
c
t
o
r
a
l
P
r
o
g
r
a
m
,
D
e
p
a
r
t
me
n
t
o
f
I
n
f
o
r
mat
i
o
n
Te
c
h
n
o
l
o
g
y
,
G
u
n
a
d
a
r
m
a
U
n
i
v
e
r
si
t
y
,
D
e
p
o
k
,
I
n
d
o
n
e
si
a
4
D
e
p
a
r
t
m
e
n
t
o
f
C
o
m
p
u
t
e
r
S
y
s
t
e
m
,
F
a
c
u
l
t
y
o
f
I
n
f
o
r
m
a
t
i
o
n
S
y
s
t
e
m
a
n
d
I
n
f
o
r
m
a
t
i
o
n
T
e
c
h
n
o
l
o
g
y
,
G
u
n
a
d
a
r
m
a
U
n
i
v
e
r
s
i
t
y
,
D
e
p
o
k
,
I
n
d
o
n
e
s
i
a
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Au
g
3
,
2
0
2
4
R
ev
is
ed
J
an
2
9
,
2
0
2
5
Acc
ep
ted
Ma
r
1
5
,
2
0
2
5
Th
e
in
c
re
a
sin
g
d
e
m
a
n
d
f
o
r
m
u
sta
rd
g
re
e
n
s
h
a
s
d
riv
e
n
fa
rm
e
rs
to
c
o
n
ti
n
u
o
u
sly
imp
r
o
v
e
m
u
sta
rd
g
re
e
n
s
c
u
lt
iv
a
ti
o
n
.
On
e
o
f
th
e
c
h
a
ll
e
n
g
e
s
i
n
m
u
sta
rd
g
re
e
n
s
c
u
l
ti
v
a
ti
o
n
is
th
e
p
re
se
n
c
e
o
f
in
se
c
t
p
e
sts.
A
sig
n
if
ica
n
t
p
e
st
in
m
u
sta
rd
g
re
e
n
s
is
Cro
c
id
o
lo
m
ia
p
a
v
o
n
a
n
a
(
C
.
p
a
v
o
n
a
n
a
)
.
C.
p
a
v
o
n
a
n
a
d
a
m
a
g
e
s
p
lan
ts
b
y
fe
e
d
i
n
g
o
n
v
a
rio
u
s
p
a
rts,
e
sp
e
c
iall
y
t
h
e
lea
v
e
s.
Th
e
i
n
it
ia
l
ste
p
in
c
o
n
tr
o
ll
in
g
t
h
e
m
is
in
s
e
c
t
p
e
st
m
o
n
it
o
rin
g
.
M
o
n
it
o
ri
n
g
a
ims
to
e
sta
b
li
sh
th
e
c
o
n
tr
o
l
th
re
sh
o
ld
.
C
.
p
a
v
o
n
a
n
a
larv
a
e
h
a
v
e
fo
u
r
i
n
st
a
r
sta
g
e
s:
in
sta
r
1
,
2
,
3
,
a
n
d
4
.
Id
e
n
ti
fica
ti
o
n
o
f
t
h
e
in
sta
r
larv
a
l
sta
g
e
s
u
ti
li
z
e
s
d
e
e
p
c
o
n
v
o
lu
ti
o
n
a
l
n
e
u
ra
l
n
e
tw
o
r
k
(
CNN
)
to
c
las
sify
C.
Pa
v
o
n
a
n
a
larv
a
e
o
n
m
u
sta
rd
g
re
e
n
s
u
sin
g
Re
sN
e
t5
0
V2
a
n
d
De
n
se
Ne
t1
6
9
a
rc
h
it
e
c
tu
re
s
o
p
ti
m
ize
d
to
e
n
h
a
n
c
e
c
las
sifica
ti
o
n
a
c
c
u
ra
c
y
.
Th
e
c
las
sifica
ti
o
n
e
v
a
lu
a
ti
o
n
re
su
lt
s
sh
o
w
t
h
a
t
b
o
th
De
n
se
Ne
t
1
6
9
a
n
d
Re
sN
e
t5
0
V2
m
o
d
e
ls
a
c
h
iev
e
h
i
g
h
a
c
c
u
ra
c
y
,
with
De
n
se
Ne
t1
6
9
re
a
c
h
in
g
th
e
h
ig
h
e
st
a
c
c
u
ra
c
y
a
t
9
7
.
1
%
,
w
h
il
e
Re
sN
e
t5
0
V2
a
c
h
iev
e
s
a
n
a
c
c
u
ra
c
y
o
f
9
4
.
2
%
.
Th
e
l
o
we
r
lo
ss
v
a
l
u
e
s
o
n
t
h
e
tes
t
d
a
ta
c
o
m
p
a
re
d
to
th
e
v
a
li
d
a
ti
o
n
d
a
ta
in
d
ica
te
t
h
a
t
t
h
e
d
e
e
p
lea
rn
i
n
g
m
o
d
e
ls
h
a
v
e
su
c
c
e
ss
fu
ll
y
c
a
p
tu
r
e
d
th
e
p
a
tt
e
rn
s
in
C.
p
a
v
o
n
a
n
a
i
m
a
g
e
s
fo
r
c
las
sifica
ti
o
n
.
Th
is
c
las
sifica
ti
o
n
p
r
o
c
e
ss
is
e
x
p
e
c
ted
to
b
e
o
n
e
o
f
th
e
a
c
ti
v
it
ies
i
n
m
o
n
it
o
rin
g
th
e
in
sta
r
larv
a
e
to
imp
r
o
v
e
th
e
a
c
c
u
ra
c
y
o
f
in
se
c
ti
c
id
e
sp
ra
y
i
n
g
a
n
d
e
n
h
a
n
c
e
m
u
sta
rd
g
re
e
n
s
p
ro
d
u
c
ti
o
n
.
K
ey
w
o
r
d
s
:
C
las
s
if
icatio
n
C
r
o
cid
o
lo
mia
p
a
vo
n
a
n
a
Den
s
eNe
t1
6
9
Mu
s
tar
d
g
r
ee
n
s
R
esNet5
0
V2
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
:
R
o
d
iah
Dep
ar
tm
en
t o
f
I
n
f
o
r
m
atics,
Facu
lty
o
f
I
n
d
u
s
tr
y
T
ec
h
n
o
lo
g
y
,
Gu
n
ad
ar
m
a
U
n
iv
er
s
ity
Ma
r
g
o
n
d
a
R
ay
a
1
0
0
,
Po
n
d
o
k
C
in
a,
Dep
o
k
,
W
est J
av
a,
I
n
d
o
n
esia
E
m
ail: r
o
d
iah
@
s
taf
f
.
g
u
n
a
d
ar
m
a.
ac
.
id
1.
I
NT
RO
D
UCT
I
O
N
T
h
e
v
e
g
etab
les
o
f
th
e
B
r
ass
ic
ac
ea
e
f
am
ily
h
av
e
im
p
o
r
tan
t
p
o
ten
tial
f
o
r
b
o
d
y
h
ea
lth
[
1
]
.
B
ec
au
s
e
o
f
th
e
co
n
ten
t
o
f
v
itam
in
s
,
p
h
en
o
lic
co
m
p
o
u
n
d
s
,
g
lu
c
o
s
in
o
lates,
ca
r
o
ten
o
id
s
,
an
d
asco
r
b
ic
ac
id
.
So
m
e
o
f
th
e
u
s
es
o
f
th
ese
co
m
p
o
u
n
d
s
ar
e
[
2
]
i
t
ca
n
p
r
o
v
id
e
p
r
e
v
en
tio
n
ag
ain
s
t
ca
n
ce
r
,
an
tio
x
id
a
n
ts
,
an
ti
-
in
f
lam
m
ato
r
y
,
an
ti
-
d
iab
etic,
an
d
p
r
o
tectio
n
f
o
r
n
er
v
es
[
3
]
.
Sev
er
al
ty
p
es
o
f
v
eg
etab
les
in
th
e
B
r
ass
icac
ea
e
f
am
ily
in
clu
d
e
[
4
]
W
h
ite
ca
b
b
ag
e,
g
r
ee
n
ca
b
b
a
g
e
,
C
h
in
ese
b
r
o
cc
o
li,
p
a
k
ch
o
i,
a
n
d
m
u
s
tar
d
g
r
ee
n
s
.
On
e
o
f
t
h
e
im
p
o
r
ta
n
t
p
ests
in
th
e
B
r
ass
icac
ea
e
v
eg
etab
le
f
a
m
ily
is
C
r
o
cid
o
lo
mia
p
a
vo
n
a
n
a
(
C
.
p
a
vo
n
a
n
a
)
[
5
]
.
T
h
e
p
r
e
s
en
ce
o
f
p
est
in
s
ec
ts
ca
n
ca
u
s
e
d
am
a
g
e
to
ca
b
b
ag
e
v
eg
etab
le
c
r
o
p
s
with
attac
k
i
n
ten
s
ities
r
an
g
in
g
f
r
o
m
m
ild
to
s
ev
er
e,
an
d
ca
n
ev
en
lead
to
s
ig
n
if
ican
t
y
iel
d
lo
s
s
es
d
u
e
to
p
ests
s
u
ch
a
s
C
.
p
a
vo
n
a
n
a
[
6
]
.
T
h
e
m
o
n
ito
r
in
g
o
f
p
ests
in
cu
ltiv
atio
n
ar
ea
s
is
an
ac
tiv
ity
th
at
alig
n
s
with
th
e
co
n
ce
p
t
o
f
in
teg
r
ated
p
est
m
an
ag
e
m
en
t
[
7
]
.
I
n
teg
r
ated
p
est
m
an
ag
em
en
t
is
a
cu
ltiv
atio
n
co
n
ce
p
t
in
v
o
lv
in
g
th
e
m
o
n
ito
r
in
g
o
f
cr
o
p
s
to
o
b
s
er
v
e
th
e
p
r
esen
ce
o
f
in
s
ec
ts
,
ty
p
es
an
d
p
o
p
u
latio
n
s
o
f
p
est
in
s
ec
ts
,
s
y
m
p
to
m
s
o
f
d
am
a
g
e,
an
d
th
e
p
er
ce
n
tag
e
o
f
p
lan
t
d
am
ag
e
ca
u
s
ed
b
y
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
A
n
o
p
timiz
ed
tr
a
n
s
fer lea
r
n
in
g
-
b
a
s
ed
a
p
p
r
o
a
ch
fo
r
C
r
o
cid
o
lo
mia
p
a
vo
n
a
n
a
la
r
va
e
…
(
R
is
n
a
w
a
ti
)
2271
p
est
in
s
ec
t
attac
k
s
in
cu
ltiv
ated
p
lan
ts
[
8
]
.
T
h
e
ac
tiv
ity
o
f
m
o
n
ito
r
in
g
cu
ltiv
ated
p
lan
ts
is
co
n
d
u
cte
d
with
th
e
aim
o
f
estab
lis
h
in
g
a
th
r
esh
o
ld
o
f
ac
tio
n
[
9
]
.
T
h
e
ac
tio
n
th
r
esh
o
ld
f
o
r
c
o
n
tr
o
l
u
s
in
g
s
y
n
th
etic
ch
em
ical
in
s
ec
ticid
es
ag
ain
s
t
C
.
p
a
vo
n
a
n
a
lar
v
a
e
is
≥
3
eg
g
cl
u
s
ter
s
p
er
1
0
ca
b
b
ag
e
p
lan
ts
[
9
]
.
T
h
e
m
o
n
ito
r
in
g
is
also
co
n
d
u
cte
d
to
a
v
o
id
s
ch
ed
u
le
d
u
s
e
o
f
in
s
ec
ticid
es
th
at
r
es
u
lt
in
r
esid
u
es
o
n
cr
o
p
y
ield
s
[
1
0
]
a
n
d
i
n
s
ec
ts
r
esis
tan
ce
[
1
1
]
.
On
e
o
f
th
e
i
m
p
o
r
tan
t
p
ests
in
m
u
s
tar
d
g
r
ee
n
s
is
C
.
p
a
vo
n
a
n
a
,
wh
ich
d
am
ag
es
p
lan
ts
b
y
f
ee
d
in
g
o
n
th
eir
p
ar
ts
,
esp
ec
ially
th
e
leav
es
[
1
2
]
.
T
h
e
in
cid
e
n
t
r
esu
lted
in
m
in
o
r
to
s
ev
er
e
d
am
ag
es,
ca
u
s
in
g
ev
en
f
a
r
m
er
s
to
ex
p
er
ien
ce
f
ailed
h
ar
v
ests
[
1
3
]
.
T
h
is
n
ec
ess
itate
s
th
e
n
ee
d
f
o
r
co
n
tr
o
l
m
ea
s
u
r
es,
with
t
h
e
in
itial st
ep
b
ein
g
m
o
n
ito
r
in
g
t
h
e
tar
g
et
p
est in
s
ec
ts
.
Mo
n
ito
r
in
g
aim
s
to
estab
lis
h
th
e
co
n
tr
o
l th
r
esh
o
ld
.
O
n
e
o
f
th
e
m
o
n
ito
r
in
g
ac
tiv
ities
in
v
o
l
v
es
id
en
tify
i
n
g
C
.
p
a
v
o
n
a
n
a
l
ar
v
ae
.
C
.
p
a
v
o
n
a
n
a
lar
v
ae
co
n
s
is
t
o
f
in
s
tar
s
tag
es
1
,
2
,
3
,
an
d
4
[
1
4
]
w
h
ich
h
as
v
ar
iatio
n
s
in
s
ize,
co
lo
r
,
an
d
s
h
ap
e.
I
n
s
tar
1
h
as
th
e
s
m
alles
t
s
ize
co
m
p
ar
ed
to
later
in
s
tar
s
.
C
o
n
v
en
tio
n
al
id
en
tific
atio
n
h
as
lim
itatio
n
s
,
tak
es
a
lo
n
g
tim
e,
r
eq
u
ir
es
ex
p
er
ts
,
an
d
in
cu
r
s
h
ig
h
co
s
ts
[
1
5
]
.
T
h
e
ac
cu
r
ate
a
n
d
r
ap
id
i
d
en
tific
atio
n
ca
n
b
e
u
s
ed
to
h
elp
f
a
r
m
er
s
r
ec
o
g
n
ize
th
e
in
s
tar
s
tag
es
o
f
C
.
p
a
vo
n
a
n
a
p
ests
.
On
e
s
m
ar
t
tech
n
iq
u
e
f
o
r
id
en
tific
atio
n
is
u
s
in
g
d
ee
p
c
o
n
v
o
lu
ti
o
n
al
n
eu
r
al
n
etwo
r
k
s
(
C
NN)
.
Ar
ch
itectu
r
es
in
d
ee
p
C
NN
ca
p
ab
le
o
f
id
en
tify
in
g
im
ag
es,
esp
ec
ially
s
m
all
lar
v
al
f
o
r
m
s
,
in
clu
d
e
r
esid
u
al
n
etwo
r
k
(
R
esNet)
an
d
d
en
s
ely
co
n
n
ec
ted
co
n
v
o
lu
ti
o
n
al
n
et
wo
r
k
(
Den
s
eNe
t)
.
R
esNet
[
1
6
]
is
an
ar
ch
itectu
r
e
d
esig
n
ed
to
e
n
ab
le
m
o
r
e
ef
f
icien
t
f
lo
w
o
f
i
n
f
o
r
m
atio
n
th
r
o
u
g
h
d
ee
p
lay
er
s
in
a
n
etwo
r
k
.
R
esNet
ca
n
ad
d
r
ess
is
s
u
es
r
elate
d
to
d
ee
p
n
etwo
r
k
tr
ain
in
g
an
d
s
ig
n
if
ican
tly
i
m
p
r
o
v
e
im
a
g
e
p
r
o
ce
s
s
in
g
p
er
f
o
r
m
an
ce
.
Den
s
eNe
t
ar
ch
itectu
r
e
[
1
5
]
c
an
c
r
ea
te
a
d
en
s
e
n
etwo
r
k
s
tr
u
ctu
r
e
to
ac
h
iev
e
p
ar
a
m
eter
ef
f
ici
en
cy
,
esp
ec
ially
i
n
ex
tr
ac
tin
g
s
tr
o
n
g
f
ea
tu
r
es f
r
o
m
v
er
y
s
m
all
lar
v
ae
im
ag
es
.
Pattn
aik
an
d
Par
v
ath
i
[
1
7
]
c
la
s
s
if
y
in
g
to
m
ato
p
la
n
t
p
ests
u
s
in
g
Den
s
eNe
t
-
1
6
9
m
o
d
el
r
esu
lted
in
an
ac
cu
r
ac
y
o
f
8
8
.
8
3
%,
with
th
e
h
ig
h
est
ac
cu
r
ac
y
o
f
8
8
.
3
7
%
a
ch
iev
ed
u
s
in
g
t
h
e
R
esNet5
0
V2
m
o
d
el.
T
h
e
m
o
d
el
was
tr
ain
ed
f
o
r
1
0
0
ep
o
ch
s
with
a
d
ata
s
p
lit
r
atio
o
f
7
:1
:2
f
o
r
tr
ain
in
g
,
v
alid
atio
n
,
an
d
test
d
ata,
r
esp
ec
tiv
ely
.
L
i
et
a
l.
[
1
8
]
p
er
f
o
r
m
i
n
g
t
h
e
c
lass
if
icatio
n
o
f
r
ice
p
la
n
t
p
est
s
u
s
in
g
a
p
r
e
-
tr
ain
e
d
m
o
d
el
to
d
etec
t
in
s
ec
t
p
ests
b
y
s
p
litt
in
g
th
e
d
ata
i
n
to
a
r
atio
o
f
6
:2
:
2
.
T
h
is
r
esear
ch
u
tili
ze
d
atten
tiv
e
r
ec
u
r
r
e
n
t
g
en
er
ativ
e
ad
v
er
s
ar
ial
n
etwo
r
k
(
AR
GAN)
d
ata
au
g
m
en
tatio
n
to
ad
d
r
ess
th
e
im
b
alan
ce
in
th
e
d
ataset,
r
esu
ltin
g
in
a
class
if
icatio
n
ac
cu
r
ac
y
o
f
8
7
.
8
1
%.
L
i
et
a
l.
[
1
9
]
p
er
f
o
r
m
in
g
p
lan
t
p
est
class
if
icatio
n
in
im
ag
es
wi
th
n
atu
r
al
b
ac
k
g
r
o
u
n
d
s
u
s
in
g
p
r
e
-
tr
ain
ed
m
o
d
els
R
es
Net5
0
an
d
R
esNet1
5
2
,
em
p
lo
y
in
g
d
ata
au
g
m
en
tatio
n
tech
n
i
q
u
es
s
u
ch
as
m
ir
r
o
r
im
ag
e,
9
0
-
d
e
g
r
ee
r
o
tatio
n
,
ad
d
in
g
n
o
is
e
to
im
a
g
es,
an
d
cr
o
p
p
ed
im
ag
es.
Fath
im
at
h
u
l
et
a
l.
[
2
0
]
c
lass
if
y
in
g
7
5
d
if
f
er
en
t
s
p
ec
ies
o
f
b
u
tter
f
lie
s
with
a
s
p
lit
r
atio
o
f
7
:2
:
1
.
T
h
e
e
x
p
er
im
en
tal
r
esu
lts
s
h
o
w
a
class
if
icatio
n
ac
cu
r
ac
y
o
f
4
3
%
u
s
in
g
th
e
R
e
s
Net5
0
m
o
d
el.
W
u
et
a
l.
[
2
1
]
p
er
f
o
r
m
in
g
class
if
icatio
n
in
to
3
0
class
es
u
s
in
g
a
p
r
e
-
tr
ain
ed
m
o
d
el
an
d
g
r
a
d
-
C
AM
to
h
ig
h
lig
h
t
in
s
ec
t
p
est
ar
ea
s
with
a
h
ig
h
est
ac
cu
r
ac
y
o
f
9
6
.
1
%
o
n
th
e
R
esNet1
0
1
m
o
d
el.
On
s
ev
er
al
p
r
ev
i
o
u
s
r
esear
c
h
[
2
]
,
[
1
7
]
–
[
20
]
,
t
h
e
class
if
icatio
n
ac
cu
r
ac
y
a
p
p
ea
r
s
to
b
e
b
elo
w
ex
ce
llen
t
class
if
icatio
n
[
2
2
]
.
R
esear
ch
b
y
L
i
et
a
l.
[
1
9
]
d
id
n
o
t
ev
alu
ate
th
e
tr
ain
i
n
g
p
r
o
c
ess
b
y
an
aly
zin
g
th
e
ac
cu
r
ac
y
an
d
lo
s
s
o
f
th
e
m
ac
h
in
e
lear
n
in
g
m
o
d
el.
As a
r
esu
lt,
th
e
ac
cu
r
ac
y
with
th
e
R
es
-
N
et
ar
ch
itectu
r
e
o
n
ly
r
ea
ch
ed
4
3
%.
Acc
u
r
ac
y
an
d
l
o
s
s
ar
e
im
p
o
r
tan
t
m
etr
ics
an
aly
ze
d
at
th
e
en
d
o
f
ea
ch
tr
ain
in
g
ep
o
ch
to
ass
ess
th
e
m
o
d
el'
s
p
r
o
g
r
ess
d
u
r
in
g
th
e
tr
ain
in
g
p
r
o
ce
s
s
.
T
h
is
ev
alu
atio
n
h
el
p
s
in
d
etec
tin
g
o
v
er
f
itti
n
g
in
t
h
e
class
if
icatio
n
m
o
d
el
an
d
d
eter
m
in
in
g
w
h
eth
er
th
e
m
ac
h
in
e
l
ea
r
n
in
g
m
o
d
el
h
as
s
u
cc
ess
f
u
ll
y
lear
n
e
d
p
atter
n
s
f
r
o
m
th
e
d
ata.
I
n
th
is
r
esear
ch
,
o
p
tim
izatio
n
will
b
e
p
er
f
o
r
m
e
d
o
n
t
h
e
ar
ch
itectu
r
es
o
f
Den
s
eNe
t1
6
9
a
n
d
R
esNet5
0
V2
.
T
h
e
ar
c
h
itectu
r
e
o
f
th
e
m
ac
h
in
e
lear
n
i
n
g
m
o
d
el
f
o
r
m
ed
co
n
s
is
ts
o
f
a
co
m
b
in
atio
n
o
f
p
r
e
-
tr
ain
ed
m
o
d
els,
f
latten
lay
er
s
,
d
r
o
p
o
u
t
lay
e
r
s
,
d
en
s
e
lay
er
s
,
an
d
o
u
tp
u
t
la
y
er
s
.
E
ac
h
m
o
d
el
is
tr
ain
ed
f
o
r
3
0
ep
o
c
h
s
u
s
in
g
tr
ain
in
g
an
d
v
alid
atio
n
d
ata.
Af
ter
tr
ain
in
g
is
co
m
p
leted
,
an
an
aly
s
is
is
co
n
d
u
cted
o
n
th
e
lo
s
s
an
d
ac
cu
r
ac
y
g
r
ap
h
s
o
b
tain
e
d
b
y
th
e
m
ac
h
in
e
lear
n
i
n
g
m
o
d
el
d
u
r
i
n
g
th
e
tr
ain
in
g
p
r
o
ce
s
s
.
T
h
e
tr
ain
ed
m
ac
h
in
e
lear
n
in
g
m
o
d
els
ar
e
th
en
ev
alu
ated
o
n
ce
ag
ai
n
u
s
in
g
test
d
ata
to
d
eter
m
in
e
th
e
p
er
f
o
r
m
an
ce
o
f
th
e
m
o
d
el
o
n
n
ew
d
ata
[
2
3
]
.
T
h
e
s
ec
o
n
d
ev
alu
atio
n
is
im
p
o
r
tan
t
to
g
et
an
o
v
e
r
v
iew
o
f
th
e
m
o
d
el'
s
p
er
f
o
r
m
an
ce
b
ef
o
r
e
th
e
m
o
d
el
[
2
4
]
is
u
s
ed
f
o
r
class
if
y
in
g
im
ag
es
o
f
C
.
p
a
vo
n
a
n
a
lar
v
ae
in
in
s
tar
s
1
,
2
,
3
,
an
d
4
,
th
is
r
esear
ch
aim
s
to
p
r
o
d
u
ce
t
h
e
b
est
m
ac
h
in
e
lear
n
in
g
m
o
d
el
f
o
r
class
if
y
in
g
C
.
p
a
vo
n
a
n
a
with
h
ig
h
ac
cu
r
ac
y
.
T
h
e
class
if
icatio
n
r
esu
lts
ar
e
ex
p
ec
ted
to
f
ac
ilit
ate
th
e
m
o
n
ito
r
in
g
p
r
o
ce
s
s
o
f
lar
v
al
i
n
s
tar
s
,
en
h
an
cin
g
th
e
ac
cu
r
ac
y
o
f
in
s
ec
ticid
e
s
p
r
ay
in
g
[
2
5
]
.
I
n
ad
d
itio
n
,
it
is
ec
o
n
o
m
ically
e
f
f
icien
t
to
r
ed
u
c
e
co
n
tr
o
l
c
o
s
ts
an
d
m
in
im
ize
en
v
ir
o
n
m
en
tal
p
o
llu
tio
n
ca
u
s
ed
b
y
ac
tiv
e
in
s
ec
ticid
e
s
u
b
s
tan
ce
s
.
2.
M
E
T
H
O
D
T
h
is
r
esear
ch
was
co
n
d
u
cted
b
ased
o
n
f
i
v
e
s
tag
es
o
f
p
r
o
ce
s
s
es,
n
am
ely
d
ata
p
r
e
-
p
r
o
ce
s
s
in
g
s
tag
e,
d
ata
s
p
litt
in
g
s
tag
e,
m
o
d
elin
g
s
tag
e,
m
o
d
el
ev
alu
atio
n
s
tag
e,
an
d
th
e
s
tag
e
o
f
s
elec
tin
g
th
e
b
est
m
o
d
el
ac
cu
r
ac
y
.
T
h
is
s
tu
d
y
u
tili
ze
d
C
NN
to
g
en
er
ate
a
m
o
d
el
f
o
r
id
en
tify
i
n
g
C
.
p
a
v
o
n
a
n
a
lar
v
ae
.
T
h
e
f
ir
s
t
s
tag
e
in
v
o
lv
ed
d
ata
p
r
e
-
p
r
o
ce
s
s
in
g
,
co
n
s
is
tin
g
o
f
s
tep
s
to
p
r
o
ce
s
s
an
d
p
r
ep
ar
e
th
e
d
ata
f
o
r
u
s
e
in
d
ee
p
lear
n
in
g
m
o
d
el
tr
ain
in
g
.
T
h
is
s
tag
e
in
c
lu
d
ed
co
llectin
g
p
ath
l
o
ca
tio
n
d
ata
an
d
o
r
g
a
n
izin
g
it
in
to
a
tab
le
o
r
d
ataf
r
am
e,
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
14
,
No
.
3
,
J
u
n
e
2
0
2
5
:
2
2
7
0
-
2
2
8
1
2272
r
esizin
g
im
ag
es,
au
g
m
en
tin
g
d
ata
u
s
in
g
tec
h
n
iq
u
es
s
u
c
h
as
h
o
r
izo
n
tal
f
lip
,
zo
o
m
i
n
g
,
r
o
t
atio
n
,
s
h
ea
r
in
g
,
an
d
d
ata
n
o
r
m
ali
za
tio
n
.
Data
s
p
litt
in
g
was
p
er
f
o
r
m
ed
u
s
in
g
an
8
:1
:1
r
atio
.
T
h
e
s
u
b
s
eq
u
en
t
s
tag
e
in
v
o
lv
ed
th
e
im
p
lem
en
tatio
n
,
tr
ain
i
n
g
,
a
n
d
ev
alu
atio
n
o
f
d
ee
p
lea
r
n
in
g
m
o
d
els.
T
wo
d
if
f
e
r
en
t
d
ee
p
lear
n
in
g
m
o
d
els
wer
e
p
r
o
p
o
s
ed
an
d
tr
ain
e
d
in
th
is
s
tu
d
y
.
T
h
e
d
if
f
er
en
ce
b
etwe
e
n
th
e
m
o
d
els
lay
in
th
e
p
r
e
-
t
r
ain
ed
m
o
d
el
a
n
d
ad
d
itio
n
al
lay
e
r
s
u
s
ed
.
T
h
e
p
r
e
-
tr
ain
ed
R
esNet5
0
v
2
m
o
d
el
co
n
s
is
ted
o
f
1
f
latten
lay
er
,
1
d
r
o
p
o
u
t
lay
er
,
1
d
en
s
e
lay
er
,
a
n
d
1
o
u
tp
u
t
la
y
er
.
T
h
e
p
r
e
-
tr
ai
n
ed
Den
s
eNe
t
1
6
9
m
o
d
el
c
o
n
s
is
ted
o
f
1
f
lat
ten
lay
er
,
3
d
r
o
p
o
u
t
lay
er
s
,
3
d
e
n
s
e
lay
er
s
,
an
d
1
o
u
tp
u
t
la
y
er
.
E
ac
h
m
o
d
el
w
as
tr
ain
ed
f
o
r
3
0
e
p
o
ch
s
u
s
in
g
b
o
th
tr
ain
in
g
an
d
v
alid
atio
n
d
ata.
An
aly
s
is
o
f
th
e
lo
s
s
an
d
ac
cu
r
ac
y
g
r
a
p
h
s
o
b
tain
ed
f
r
o
m
t
h
e
d
ee
p
le
ar
n
in
g
m
o
d
els
was
p
er
f
o
r
m
ed
af
ter
th
e
tr
ain
in
g
p
r
o
ce
s
s
was
co
m
p
leted
.
Fu
r
th
er
m
o
r
e,
th
e
ev
alu
atio
n
o
f
d
ee
p
l
ea
r
n
in
g
m
o
d
els wa
s
co
n
d
u
cte
d
o
n
ce
m
o
r
e
u
s
in
g
te
s
tin
g
d
ata
to
ass
es
s
th
eir
p
er
f
o
r
m
an
ce
o
n
n
ew
d
ata.
T
h
is
s
ec
o
n
d
ev
alu
atio
n
was
ca
r
r
ied
o
u
t to
g
ain
an
u
n
d
er
s
ta
n
d
in
g
o
f
th
e
m
o
d
el'
s
p
er
f
o
r
m
a
n
ce
.
2
.
1
.
C
.
p
a
vo
na
na
la
r
v
a
e
da
t
a
s
et
T
h
e
d
ataset
o
f
L
ar
v
a
C
.
p
a
vo
n
a
n
a
was
o
b
tain
ed
th
r
o
u
g
h
d
i
r
ec
t
ac
q
u
is
itio
n
in
a
m
u
s
tar
d
p
lan
tatio
n
ar
ea
in
Dep
o
k
,
W
est
J
av
a.
T
h
e
d
ata
c
o
m
p
r
is
es
6
8
4
im
a
g
es
o
f
C
.
p
a
vo
n
a
n
a
,
in
cl
u
d
in
g
1
3
1
im
ag
es
o
f
i
n
s
tar
1
class
,
1
3
2
im
ag
es o
f
in
s
tar
2
c
lass
,
1
5
4
im
ag
es o
f
in
s
tar
3
cl
ass
,
an
d
2
6
7
im
ag
es o
f
in
s
tar
4
class
.
E
x
am
p
les o
f
th
e
f
o
u
r
ty
p
es o
f
d
atasets
ca
n
b
e
s
ee
n
in
Fig
u
r
e
1
.
C
.
p
a
vo
n
a
n
a
u
n
d
er
g
o
es
co
m
p
lete
m
etam
o
r
p
h
o
s
is
,
in
clu
d
in
g
th
e
s
tag
es
o
f
eg
g
,
lar
v
a
,
p
u
p
a,
an
d
im
ag
o
[
2
6
]
.
E
ac
h
p
h
ase
h
as
a
d
if
f
er
en
t
d
u
r
atio
n
o
f
th
e
s
tag
e.
T
h
e
lar
v
al
s
tag
e
in
clu
d
es
in
s
tar
s
1
,
2
,
3
,
an
d
4
[
2
7
]
.
W
h
er
e
t
h
e
lar
v
al
s
tag
e
is
a
cr
itical
p
h
ase
b
ec
au
s
e
d
u
r
i
n
g
th
is
s
tag
e,
th
er
e
is
ac
tiv
ity
o
f
ea
tin
g
v
eg
eta
b
les
th
at
ca
u
s
es d
am
ag
e.
T
h
e
lar
v
a
l stag
e
s
tar
ts
f
r
o
m
wh
en
th
e
la
r
v
a
em
er
g
es u
n
til in
s
tar
I
V.
Fig
u
r
e
1
.
C
.
p
a
vo
n
a
n
a
la
r
v
ae
i
m
ag
e
2
.
2
.
C
.
p
a
vo
na
na
la
r
v
a
e
pre
pro
ce
s
s
ing
Pre
p
r
o
ce
s
s
in
g
o
f
C
.
p
a
vo
n
a
n
a
lar
v
a
im
ag
es c
o
n
s
is
ts
o
f
s
ev
er
al
s
tag
es o
f
p
r
o
ce
s
s
es,
in
clu
d
i
n
g
:
−
Ad
ju
s
tin
g
th
e
s
ize
o
f
th
e
im
ag
e
C
.
p
a
vo
n
a
n
a
.
T
h
e
p
r
e
-
tr
ain
e
d
m
o
d
el
h
as
a
d
ef
au
lt
R
GB
co
lo
r
im
ag
e
[
2
8
]
.
T
h
e
d
ef
a
u
lt
s
ize
o
f
t
h
e
C
.
p
a
v
o
n
a
n
a
im
ag
e
v
ar
ies.
Ad
ju
s
tm
en
t
to
th
e
C
.
p
a
vo
n
a
n
a
im
ag
e
is
m
ad
e
to
m
ee
t
th
e
in
p
u
t
s
p
ec
if
icatio
n
s
o
f
ea
c
h
p
r
e
-
tr
ain
ed
m
o
d
el
[
2
9
]
.
T
h
e
s
ize
to
b
e
u
s
ed
f
o
r
th
e
im
ag
e
o
f
C
.
p
a
vo
n
a
n
a
is
2
5
6
×2
5
6
p
ix
els.
T
ab
le
1
s
h
o
ws th
e
d
etails o
f
th
e
in
p
u
t siz
e
f
o
r
ea
ch
p
r
e
-
tr
ain
e
d
m
o
d
el.
T
ab
le
1
p
r
o
v
id
es
an
ex
am
p
le
o
f
th
e
s
ize
ad
j
u
s
tm
en
t r
esu
lts
f
o
r
th
e
im
a
g
e
o
f
C
.
p
a
vo
n
a
n
a
.
T
ab
le
1
.
E
x
am
p
le
o
f
im
ag
e
a
d
ju
s
m
en
t r
esu
lts
f
o
r
C
.
p
a
v
o
n
a
n
a
Pre
-
t
r
a
i
n
e
d
m
o
d
e
l
I
n
p
u
t
si
z
e
R
e
sN
e
t
5
0
V
2
(
2
5
6
,
2
5
6
,
3
)
D
e
n
seN
e
t
1
6
9
(
2
5
6
,
2
5
6
,
3
)
−
Per
f
o
r
m
in
g
au
g
m
en
tatio
n
p
r
o
ce
s
s
es
to
in
cr
ea
s
e
th
e
am
o
u
n
t
o
f
tr
ain
in
g
d
ata
[
3
0
]
b
y
m
o
d
if
y
in
g
th
e
d
ata
o
f
C
.
p
a
vo
n
a
n
a
,
v
ar
io
u
s
tech
n
i
q
u
es
wer
e
em
p
l
o
y
ed
,
in
clu
d
in
g
h
o
r
izo
n
tal
f
lip
,
zo
o
m
in
g
,
r
o
tatio
n
,
an
d
s
h
ea
r
in
g
.
H
o
r
izo
n
tal
f
lip
d
u
p
licates
th
e
im
ag
e
b
y
f
li
p
p
in
g
it
h
o
r
izo
n
tally
.
Z
o
o
m
in
g
in
v
o
lv
es
p
r
o
p
o
r
tio
n
ally
en
lar
g
in
g
o
r
r
ed
u
cin
g
th
e
im
ag
e.
R
o
tatio
n
is
an
au
g
m
en
tatio
n
tech
n
iq
u
e
th
at
in
v
o
lv
es
r
o
tatin
g
th
e
im
ag
e
at
d
if
f
er
e
n
t
an
g
les.
Sh
ea
r
in
g
is
p
er
f
o
r
m
e
d
b
y
s
h
if
tin
g
th
e
p
o
in
ts
o
n
th
e
im
ag
e
alo
n
g
a
s
p
ec
if
ic
ax
is
to
cr
ea
te
a
d
is
to
r
tio
n
ef
f
ec
t
s
im
ilar
to
p
e
r
s
p
ec
tiv
e
d
is
to
r
tio
n
.
T
h
ese
f
o
u
r
au
g
m
en
tatio
n
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
A
n
o
p
timiz
ed
tr
a
n
s
fer lea
r
n
in
g
-
b
a
s
ed
a
p
p
r
o
a
ch
fo
r
C
r
o
cid
o
lo
mia
p
a
vo
n
a
n
a
la
r
va
e
…
(
R
is
n
a
w
a
ti
)
2273
tech
n
iq
u
es
wer
e
im
p
lem
en
te
d
u
s
in
g
th
e
p
ar
am
eter
s
o
f
t
h
e
I
m
a
g
eDa
taGe
n
er
ato
r
,
s
p
e
cif
ically
f
o
r
th
e
tr
ain
in
g
d
ata
o
n
ly
.
T
h
e
r
esu
lts
o
f
th
e
au
g
m
en
tatio
n
tech
n
iq
u
es,
n
am
ely
h
o
r
izo
n
tal
f
lip
,
zo
o
m
in
g
,
r
o
tatio
n
,
an
d
s
h
ea
r
in
g
,
ca
n
b
e
o
b
s
er
v
ed
in
Fig
u
r
e
2.
Fig
u
r
e
2
.
E
x
am
p
le
o
f
au
g
m
en
t
ed
im
ag
es o
f
C
.
p
a
vo
n
a
n
a
lar
v
ae
−
No
r
m
alizin
g
th
e
v
al
u
es
o
f
ea
ch
p
ix
el
in
th
e
im
a
g
e.
N
o
r
m
aliza
tio
n
is
an
im
p
o
r
tan
t
p
r
e
-
p
r
o
ce
s
s
in
g
tech
n
iq
u
e
t
o
en
s
u
r
e
th
at
th
e
m
o
d
el
tr
ain
i
n
g
p
r
o
ce
s
s
ca
n
r
u
n
f
aster
a
n
d
m
o
r
e
ef
f
ec
tiv
el
y
[
3
1
]
,
t
h
e
task
p
er
f
o
r
m
ed
b
y
d
iv
id
i
n
g
ea
ch
p
ix
el
v
alu
e
b
y
2
5
5
,
r
esu
ltin
g
i
n
n
ew
v
alu
es
with
in
th
e
r
a
n
g
e
o
f
0
-
1
,
ca
n
b
e
ex
p
r
ess
ed
in
p
s
eu
d
o
co
d
e
as f
o
llo
ws:
#
Image Generator
ts_gen = ImageDataGenerator(preprocessing_function= scalar, rescale=1./255
)
T
h
e
I
m
ag
eDa
taGe
n
er
ato
r
is
a
d
ec
lar
atio
n
o
f
an
im
ag
e
g
en
er
ato
r
th
at
will
ca
ll
th
e
C
.
p
a
vo
n
a
n
a
im
ag
es
f
o
r
v
alid
atio
n
an
d
test
in
g
.
T
h
e
p
ar
am
eter
r
escale=
1
.
/2
5
5
in
d
icate
s
th
e
v
alu
e
u
s
ed
f
o
r
d
at
a
n
o
r
m
aliza
tio
n
p
r
o
ce
s
s
.
2
.
3
.
C
.
p
a
vo
na
na
la
r
v
a
e
da
t
a
s
pli
t
t
ing
T
h
e
d
ata
s
p
litt
in
g
is
d
o
n
e
b
y
d
iv
id
in
g
t
h
e
d
ataset
in
to
th
r
ee
p
ar
ts
:
tr
ain
in
g
d
ata
(
tr
ai
n
_
d
f
)
,
v
alid
atio
n
d
ata
(
v
alid
_
d
f
)
,
an
d
test
d
ata
(
test
_
d
f
)
.
T
h
e
d
iv
is
io
n
r
atio
u
s
ed
f
o
r
ea
ch
d
ata
s
u
b
s
et
is
8
:1
:
1
.
T
h
e
d
ata
d
iv
is
io
n
s
tar
ts
b
y
s
p
litt
in
g
th
e
d
ata
in
to
tr
ain
_
d
f
an
d
d
u
m
m
y
_
d
f
with
an
8
:2
r
atio
.
T
h
en
,
d
u
m
m
y
_
d
f
is
f
u
r
th
er
d
iv
id
e
d
in
to
a
5
:5
r
atio
,
r
esu
ltin
g
in
v
alid
_
d
f
an
d
test
_
d
f
.
T
h
e
d
ata
d
iv
is
io
n
is
d
o
n
e
u
s
in
g
t
h
e
tr
ain
_
test
_
s
p
lit()
f
u
n
ctio
n
f
r
o
m
th
e
Scik
itLe
ar
n
lib
r
ar
y
[
3
2
]
.
T
h
e
d
iv
is
io
n
o
f
th
is
d
ata
is
d
o
n
e
in
p
r
o
p
o
r
tio
n
to
th
e
r
esp
ec
tiv
e
class
.
Stra
tifie
d
s
am
p
lin
g
m
eth
o
d
is
u
s
ed
to
o
b
tain
a
b
alan
c
ed
d
is
tr
ib
u
tio
n
o
f
s
u
b
d
ata.
T
h
e
d
ataset
u
s
ed
in
th
is
r
esear
ch
co
n
s
is
ts
o
f
3
9
%
in
s
t
ar
4
class
,
2
3
%
in
s
tar
3
class
,
1
9
%
in
s
tar
2
class
,
an
d
1
9
%
in
s
tar
1
class
.
T
h
e
s
tr
atif
ied
s
am
p
lin
g
m
et
h
o
d
c
an
p
r
o
d
u
ce
s
u
b
d
ata
with
th
e
s
am
e
p
er
ce
n
tag
e
d
is
tr
ib
u
tio
n
o
f
class
es
[
3
3
]
T
h
r
o
u
g
h
th
e
co
m
p
lete
d
ataset.
T
h
is
m
eth
o
d
is
im
p
lem
en
t
ed
b
y
m
ap
p
i
n
g
th
e
class
co
lu
m
n
to
th
e
s
tr
atif
y
p
ar
am
eter
o
f
th
e
tr
ain
_
test
_
s
p
lit()
f
u
n
ctio
n
.
2
.
4
.
O
pti
m
iza
t
io
n
o
f
C
.
p
a
v
o
na
na
la
rv
a
e
cla
s
s
if
ica
t
io
n
mo
del da
t
a
s
pli
t
t
ing
Mo
d
elin
g
is
th
e
c
o
r
e
s
tag
e
i
n
th
is
r
esear
c
h
,
w
h
ich
s
tar
ts
with
th
e
in
itiatio
n
o
f
th
e
d
ee
p
lear
n
i
n
g
m
o
d
el.
T
h
e
p
r
o
ce
s
s
in
v
o
lv
es
tr
ain
in
g
th
e
d
ee
p
lear
n
i
n
g
m
o
d
el
u
s
in
g
tr
ain
in
g
d
ata
an
d
ev
al
u
atin
g
th
e
tr
ain
in
g
p
r
o
ce
s
s
.
All
s
tep
s
in
th
is
s
ta
g
e
ar
e
r
ep
ea
te
d
u
s
in
g
d
if
f
er
e
n
t
p
r
e
-
tr
ain
e
d
m
o
d
els.
Fig
u
r
e
3
r
ep
r
esen
ts
th
e
ar
ch
itectu
r
e
o
f
th
e
p
r
e
-
tr
ain
ed
m
o
d
els an
d
th
e
f
in
e
-
tu
n
ed
m
o
d
els in
itialized
in
th
is
s
tu
d
y
.
T
h
e
ar
ch
itectu
r
e
o
f
th
is
d
ee
p
lear
n
in
g
m
o
d
el
co
n
s
is
ts
o
f
a
co
m
b
in
atio
n
o
f
p
r
e
-
tr
ain
ed
m
o
d
els
an
d
f
u
lly
co
n
n
ec
ted
lay
er
s
.
T
h
e
s
elec
tio
n
o
f
p
r
e
-
tr
ain
ed
m
o
d
el
s
is
d
o
n
e
b
ec
au
s
e
th
e
y
p
r
o
v
id
e
b
etter
r
esu
lts
in
a
s
h
o
r
ter
tr
ain
in
g
tim
e
[
3
4
]
.
T
h
is
r
esear
ch
u
tili
ze
s
two
p
r
e
-
tr
ain
ed
m
o
d
els,
n
am
ely
R
esNet5
0
V2
an
d
Den
s
eNe
t1
6
9
,
o
b
tain
e
d
th
r
o
u
g
h
th
e
tf
.
k
er
as.a
p
p
licatio
n
s
m
o
d
u
le.
T
h
e
d
ee
p
lear
n
in
g
m
o
d
el
k
n
o
wn
as th
e
f
in
e
-
tu
n
ed
m
o
d
el
was
im
p
lem
en
ted
in
th
is
r
esear
ch
.
I
n
th
is
ap
p
r
o
ac
h
,
th
e
weig
h
ts
an
d
b
iases
o
f
th
e
p
r
e
-
tr
ain
e
d
m
o
d
el
wer
e
n
o
t
f
r
o
ze
n
[
3
5
]
,
s
o
th
at
th
ey
ca
n
b
e
u
p
d
ated
d
u
r
in
g
tr
ain
in
g
.
T
h
e
ap
p
r
o
ac
h
o
f
u
s
in
g
t
h
is
f
in
e
-
tu
n
ed
m
o
d
el
aim
s
t
o
allo
w
t
h
e
m
o
d
el
[
3
6
]
u
n
d
e
r
s
tan
d
in
g
th
e
d
etails
in
ea
ch
im
a
g
e
o
f
C
.
p
a
vo
n
a
n
a
m
o
r
e
ac
cu
r
ately
.
Def
au
lt
p
r
e
-
t
r
ain
e
d
m
o
d
els
h
av
e
th
eir
r
esp
ec
tiv
e
o
u
tp
u
t
lay
er
s
.
On
av
er
a
g
e,
p
r
e
-
tr
ain
ed
m
o
d
els
av
ailab
le
in
th
e
T
en
s
o
r
Flo
w
li
b
r
ar
y
h
a
v
e
o
u
tp
u
t
lay
er
s
with
1
,
0
0
0
class
es
[
3
7
]
.
I
n
th
e
o
u
t
p
u
t
lay
er
,
it
n
ee
d
s
to
b
e
r
ep
lace
d
with
a
n
ew
f
u
lly
-
co
n
n
ec
ted
lay
er
to
g
e
n
er
ate
th
e
class
if
icatio
n
o
f
4
clas
s
es
th
at
alig
n
with
th
e
o
b
jectiv
es
o
f
th
is
r
esear
ch
.
I
n
th
e
R
esNet5
0
V2
m
o
d
el,
t
h
er
e
is
o
n
e
f
u
lly
-
co
n
n
ec
ted
lay
e
r
,
o
n
e
f
latten
lay
er
,
o
n
e
d
r
o
p
o
u
t
lay
e
r
,
o
n
e
d
e
n
s
e
lay
er
with
2
5
6
n
o
d
es,
a
n
d
o
n
e
o
u
tp
u
t
lay
er
with
4
n
o
d
es.
I
n
th
e
Den
s
eNe
t1
6
9
m
o
d
el,
th
e
r
e
is
o
n
e
f
u
lly
-
co
n
n
ec
ted
lay
er
,
o
n
e
f
latten
lay
er
,
t
h
r
ee
d
r
o
p
o
u
t
lay
er
s
,
th
r
ee
d
e
n
s
e
lay
er
s
with
2
5
6
,
1
2
8
,
6
4
n
o
d
es
r
esp
ec
tiv
ely
,
an
d
o
n
e
o
u
tp
u
t
lay
e
r
with
4
n
o
d
es.
Op
tim
izatio
n
in
th
e
f
in
e
-
tu
n
e
d
m
o
d
el
is
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
14
,
No
.
3
,
J
u
n
e
2
0
2
5
:
2
2
7
0
-
2
2
8
1
2274
p
er
f
o
r
m
ed
b
y
u
tili
zin
g
th
e
f
ea
tu
r
e
ex
tr
ac
tio
n
n
etwo
r
k
f
r
o
m
th
e
p
r
e
-
tr
ain
e
d
m
o
d
el,
co
m
b
in
ed
with
a
n
ew
f
u
lly
-
co
n
n
ec
ted
n
etwo
r
k
.
Fig
u
r
e
3
.
Ar
c
h
itectu
r
e
o
f
p
r
e
-
tr
ain
ed
m
o
d
els an
d
f
i
n
e
-
tu
n
e
d
m
o
d
els
T
h
e
f
ea
tu
r
e
e
x
tr
ac
tio
n
is
u
s
ed
to
r
ea
d
a
n
d
r
ec
o
g
n
ize
p
atter
n
s
f
r
o
m
C
.
p
a
vo
n
a
n
a
im
ag
es,
wh
ile
th
e
f
u
lly
-
co
n
n
ec
ted
n
etwo
r
k
is
r
es
p
o
n
s
ib
le
f
o
r
lear
n
in
g
an
d
d
is
tin
g
u
is
h
in
g
p
atter
n
s
f
r
o
m
ea
ch
class
.
−
R
esNet
V5
0
V2
m
o
d
el
in
itiatio
n
:
t
h
e
f
ir
s
t
p
r
e
-
tr
ain
ed
m
o
d
el
in
itialized
in
th
is
r
esear
ch
is
th
e
p
r
e
-
tr
ain
ed
R
es
-
NetV5
0
V2
m
o
d
el,
wh
ich
is
av
ailab
le
in
th
e
tf.k
era
s
.
a
p
p
lica
tio
n
s
.
r
esn
et_
v2
.
R
esN
et5
0
V
2
m
o
d
u
le.
T
h
e
in
itializatio
n
o
f
th
e
p
r
e
-
tr
ain
e
d
m
o
d
el
is
in
clu
d
ed
with
o
u
t
th
e
d
ef
au
lt
o
u
tp
u
t
lay
er
s
s
o
th
at
n
ew
o
u
tp
u
t
lay
er
s
ca
n
b
e
a
d
d
ed
[
3
8
]
u
s
ed
f
u
lly
-
co
n
n
ec
ted
lay
e
r
.
−
Den
s
eNe
t1
6
9
in
itiatio
n
m
o
d
e
l:
t
h
e
s
ec
o
n
d
p
r
e
-
tr
ain
e
d
m
o
d
el
in
itialized
in
th
is
r
esear
c
h
is
th
e
Den
s
e
-
Net1
6
9
p
r
e
-
tr
ai
n
ed
m
o
d
el
av
a
ilab
le
in
th
e
tf.k
era
s
.
a
p
p
lica
tio
n
s
.
r
esn
et_
v2
.
Den
s
eNet1
6
9
m
o
d
u
le.
T
h
e
p
r
e
-
tr
ain
ed
m
o
d
el
in
itializatio
n
is
in
clu
d
ed
with
o
u
t
th
e
d
ef
au
lt
o
u
tp
u
t
la
y
er
s
,
allo
win
g
th
e
a
d
d
itio
n
o
f
n
ew
o
u
tp
u
t la
y
e
r
s
u
s
in
g
f
u
ll
y
-
co
n
n
ec
ted
lay
er
s
.
−
Fu
lly
-
co
n
n
ec
ted
m
o
d
el
R
es
-
NetV5
0
V2
lay
er
in
itiatio
n
:
t
h
e
p
r
e
-
tr
ain
e
d
m
o
d
el
with
o
u
t
th
e
o
u
tp
u
t
la
y
er
u
s
ed
f
r
o
m
th
e
T
e
n
s
o
r
Flo
w
l
ib
r
ar
y
n
ee
d
s
to
b
e
s
u
p
p
lem
e
n
ted
with
a
f
u
lly
-
co
n
n
ec
ted
lay
er
[
3
9
]
.
T
o
g
en
er
ate
a
tr
ain
ab
le
d
ee
p
lea
r
n
in
g
m
o
d
el,
th
e
p
r
e
-
tr
ain
ed
R
esNet5
0
V2
m
o
d
el
is
p
r
o
v
id
ed
with
f
u
lly
-
co
n
n
ec
ted
lay
e
r
s
an
d
co
m
p
il
ed
u
s
in
g
o
p
tim
izatio
n
f
u
n
cti
o
n
s
an
d
tr
ain
in
g
m
etr
ics
u
s
in
g
th
e
f
o
llo
win
g
p
s
eu
d
o
co
d
e:
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
A
n
o
p
timiz
ed
tr
a
n
s
fer lea
r
n
in
g
-
b
a
s
ed
a
p
p
r
o
a
ch
fo
r
C
r
o
cid
o
lo
mia
p
a
vo
n
a
n
a
la
r
va
e
…
(
R
is
n
a
w
a
ti
)
2275
x
=
F
l
at
te
n
()
(b
as
e
_m
od
e
l.
ou
t
pu
t
)
x
=
D
r
op
ou
t
(r
at
e=
0
.2
)(
x
)
x
=
D
e
ns
e(
2
56
,
ac
t
iv
at
i
on
='
r
el
u
')
(x
)
ou
t
pu
t
s
=
D
en
se
(c
l
as
s_
c
ou
nt
,
a
c
ti
va
t
io
n=
's
o
ft
ma
x
')
(x
)
mo
d
el
=
tf
.
ke
ra
s.
m
od
el
s
.M
od
e
l(
b
as
e_
m
od
el
.i
n
pu
t,
ou
tp
u
ts
)
model.compile(Adamax(learning_rate=0.001), loss='categorical_crossentropy',
metrics=['accuracy'])
T
h
e
p
a
r
a
m
et
e
r
x
=Fl
att
en
(
)
(
b
ase
_
m
o
d
el
.
o
u
tp
u
t
)
will
a
d
d
a
f
la
tte
n
la
y
e
r
t
o
th
e
f
u
ll
y
-
c
o
n
n
ec
t
ed
la
y
e
r
a
n
d
x
=D
r
o
p
o
u
t
(
r
at
e=
0
.
2
)
(
x
)
wil
l
a
d
d
a
d
r
o
p
o
u
t
l
ay
er
t
o
t
h
e
f
u
ll
y
-
c
o
n
n
ec
t
ed
l
a
y
e
r
.
As
m
a
n
y
a
s
2
0
%
o
f
i
n
p
u
t
v
al
u
es
p
ass
i
n
g
t
h
r
o
u
g
h
th
is
la
y
e
r
a
r
e
n
u
ll
if
ie
d
.
T
h
e
n
x
=D
e
n
s
e(
2
5
6
,
a
cti
v
ati
o
n
='
r
e
lu
'
)
(
x
)
w
ill
a
d
d
a
d
e
n
s
e
lay
e
r
t
o
t
h
e
f
u
l
ly
-
c
o
n
n
e
cte
d
l
a
y
e
r
.
T
h
e
ac
t
iv
ati
o
n
f
u
n
c
ti
o
n
u
s
ed
is
r
elu
.
P
ar
am
et
er
o
u
t
p
u
ts
=D
en
s
e(
cl
ass
_
co
u
n
t
,
a
cti
v
at
io
n
='
s
o
f
tm
ax
'
)
(
x
)
wil
l
a
d
d
t
h
e
d
e
n
s
e
l
ay
er
as
an
o
u
t
p
u
t
l
ay
er
t
o
t
h
e
f
u
ll
y
-
c
o
n
n
e
ct
ed
la
y
e
r
.
C
lass
_
co
u
n
t
is
t
h
e
n
u
m
b
e
r
o
f
c
lass
es
f
o
r
t
h
e
c
lass
i
f
i
ca
t
io
n
t
as
k
o
f
th
is
r
ese
ar
ch
,
n
a
m
el
y
4
wi
th
s
o
f
t
m
a
x
as
th
e
ac
ti
v
at
io
n
f
u
n
ct
io
n
.
Us
in
g
m
o
d
el
=t
f
.
k
e
r
as
.
m
o
d
e
ls
.
Mo
d
el(
b
ase
_
m
o
d
el
.
i
n
p
u
t,
o
u
t
p
u
ts
)
as
a
d
e
e
p
lea
r
n
i
n
g
m
o
d
el
is
cr
ea
t
e
d
b
y
c
o
m
b
i
n
i
n
g
t
h
e
C
NN
l
a
y
e
r
f
r
o
m
t
h
e
p
r
e
-
t
r
ai
n
e
d
m
o
d
el
w
it
h
th
e
n
ewl
y
cr
ea
t
ed
f
u
l
ly
-
c
o
n
n
e
cte
d
l
ay
e
r
a
n
d
s
t
o
r
ed
i
n
t
h
e
m
o
d
el
v
a
r
i
a
b
le
.
T
h
is
r
ese
ar
ch
u
s
es
t
h
e
A
d
a
m
a
x
o
p
ti
m
iz
ati
o
n
f
u
n
c
ti
o
n
wit
h
a
l
ea
r
n
in
g
_
r
at
e
o
f
0
.
0
0
1
[
4
0
]
.
C
ate
g
o
r
ica
l_
cr
o
s
s
en
tr
o
p
y
u
s
ed
as
lo
s
s
f
u
n
cti
o
n
an
d
ac
c
u
r
ac
y
as
tr
ai
n
i
n
g
m
et
r
i
c.
−
Fu
lly
-
co
n
n
ec
ted
m
o
d
el
Den
s
e
Net1
6
9
lay
er
in
itiatio
n
:
th
e
p
r
e
-
tr
ain
ed
m
o
d
el
with
o
u
t
an
o
u
tp
u
t
lay
er
u
s
ed
f
r
o
m
th
e
T
e
n
s
o
r
Flo
w
lib
r
ar
y
n
ee
d
s
to
b
e
ad
d
ed
with
a
f
u
ll
y
-
co
n
n
ec
ted
lay
er
to
p
r
o
d
u
ce
a
d
ee
p
lear
n
in
g
m
o
d
el
th
at
ca
n
b
e
t
r
ain
ed
[
4
1
]
.
T
h
e
p
r
e
-
tr
ain
ed
Den
s
eNe
t1
6
9
m
o
d
el
is
g
iv
en
f
u
lly
-
c
o
n
n
ec
ted
lay
er
s
a
n
d
co
m
p
iled
with
o
p
tim
izatio
n
f
u
n
ctio
n
s
an
d
tr
ai
n
in
g
m
etr
ics.
2
.
5
.
T
ra
ini
ng
cla
s
s
if
ica
t
io
n m
o
del o
f
C
.
p
a
vo
na
na
T
h
e
m
o
d
el
tr
ai
n
in
g
s
tag
e
im
p
lem
en
ted
i
n
d
ee
p
lea
r
n
in
g
m
o
d
els
is
in
itiated
t
h
r
o
u
g
h
a
tr
ain
in
g
p
r
o
ce
s
s
ca
lled
f
in
e
-
tu
n
i
n
g
[
4
1
]
.
T
h
e
p
r
o
ce
s
s
o
f
tr
ain
in
g
th
e
d
ee
p
lear
n
in
g
m
o
d
el
in
v
o
lv
e
s
tr
ain
in
g
d
ata
an
d
v
alid
atio
n
d
ata,
wh
er
e
th
e
tr
a
in
in
g
d
ata
is
u
s
ed
to
lear
n
th
e
f
ea
tu
r
es
o
f
th
e
C
.
P
a
vo
n
a
n
a
im
ag
es,
an
d
th
e
v
alid
atio
n
d
ata
is
u
s
ed
to
as
s
e
s
s
th
e
m
o
d
el
'
s
p
er
f
o
r
m
an
ce
d
u
r
in
g
tr
ain
in
g
.
T
h
e
tr
ain
in
g
p
r
o
ce
s
s
is
ca
r
r
ied
o
u
t
f
o
r
6
×
5
ep
o
c
h
s
,
an
d
th
e
r
esu
lts
o
f
ea
ch
ep
o
ch
ar
e
m
o
n
ito
r
e
d
u
s
in
g
ca
llb
ac
k
f
u
n
ctio
n
s
.
T
h
is
r
esear
ch
u
s
es
a
b
atch
s
ize
o
f
3
2
with
1
0
0
e
p
o
ch
s
.
A
th
r
esh
o
ld
o
f
0
.
9
is
u
s
ed
in
th
is
s
tu
d
y
to
d
eter
m
in
e
th
e
v
a
r
iab
les
m
o
n
ito
r
ed
as
lo
s
s
.
I
f
th
e
tr
ain
in
g
ac
cu
r
ac
y
is
less
th
an
th
e
th
r
esh
o
ld
,
th
e
lo
s
s
will
b
e
eq
u
al
to
th
e
tr
ain
in
g
ac
cu
r
ac
y
.
I
f
it
is
g
r
ea
ter
th
an
th
e
th
r
esh
o
ld
,
th
e
l
o
s
s
will
b
e
eq
u
al
to
th
e
v
alid
atio
n
lo
s
s
.
T
h
e
tr
ain
in
g
m
o
d
el
ev
alu
atio
n
p
r
o
ce
s
s
ca
n
b
e
s
ee
n
in
Fig
u
r
e
4
.
Fig
u
r
e
4
.
C
lass
if
icatio
n
m
o
d
el
tr
ain
in
g
ev
al
u
atio
n
p
r
o
ce
s
s
3.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
N
T
h
e
r
esear
ch
r
esu
lts
co
n
s
is
t
o
f
th
e
r
esu
lts
o
f
p
r
ep
r
o
ce
s
s
in
g
o
f
th
e
C
.
p
a
v
o
n
a
n
a
im
a
g
e
d
ataset,
d
ata
s
p
litt
in
g
,
m
o
d
elin
g
,
e
v
alu
atio
n
,
r
esu
lts
o
f
th
e
b
est
m
o
d
el
ac
cu
r
ac
y
.
T
ests
wer
e
ca
r
r
ied
o
u
t
u
s
in
g
s
ev
er
al
C
.
p
a
vo
n
a
n
a
im
a
g
es
f
r
o
m
te
s
t
d
ata
to
o
b
tain
class
if
icatio
n
r
esu
lts
f
o
r
ea
ch
m
o
d
el,
a
s
well
as
th
e
b
est
ac
cu
r
ac
y
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
14
,
No
.
3
,
J
u
n
e
2
0
2
5
:
2
2
7
0
-
2
2
8
1
2276
3
.
1
.
C
.
p
a
vo
na
na
i
m
a
g
e
pro
c
ess
ing
re
s
ult
I
n
th
is
r
esear
ch
,
th
e
d
ataset
p
r
ep
r
o
ce
s
s
in
g
s
tag
e
was
ca
r
r
ied
o
u
t
t
o
ad
d
v
a
r
iatio
n
t
o
th
e
tr
a
in
in
g
d
ata
an
d
p
r
e
p
ar
e
th
e
d
ata
s
o
th
at
t
h
e
tr
ain
in
g
p
r
o
ce
s
s
co
u
ld
r
u
n
ef
f
ec
tiv
ely
.
T
h
is
s
tag
e
b
eg
in
s
with
en
ter
in
g
th
e
d
ata
p
ath
lo
ca
tio
n
in
to
t
h
e
d
at
af
r
am
e,
ad
ju
s
tin
g
th
e
im
a
g
e
s
ize,
p
er
f
o
r
m
in
g
h
o
r
izo
n
tal
f
lip
,
zo
o
m
in
g
,
r
o
tatio
n
,
s
h
ea
r
in
g
,
an
d
n
o
r
m
aliza
tio
n
.
T
h
e
s
tep
s
in
th
e
d
ata
p
r
e
-
p
r
o
ce
s
s
in
g
s
tag
e
in
th
is
r
esear
ch
wer
e
ca
r
r
ied
o
u
t
u
s
in
g
th
e
I
m
ag
eDa
taGe
n
e
r
ato
r
(
)
f
u
n
ctio
n
s
o
th
at
th
e
r
esu
lts
wer
e
o
n
ly
v
is
ib
le
wh
e
n
e
x
ec
u
tin
g
th
e
im
ag
e
ca
llin
g
co
d
e.
T
h
e
r
esu
lts
o
f
C
.
p
a
v
o
n
a
n
a
'
s
im
ag
e
p
r
ep
r
o
ce
s
s
in
g
ca
n
b
e
s
ee
n
in
Fig
u
r
e
5
.
Fig
u
r
e
5
.
Pre
p
r
o
ce
s
s
in
g
C
.
p
a
vo
n
a
n
a
r
esu
lt
3
.
2
.
Cla
s
s
if
ica
t
io
n
m
o
del initia
liza
t
io
n r
esu
lt
s
Mo
d
el
in
itiatio
n
in
th
is
r
esea
r
ch
was
ca
r
r
ied
o
u
t
b
y
co
m
b
in
in
g
a
p
r
e
-
tr
ain
e
d
m
o
d
el
with
a
f
u
lly
-
co
n
n
ec
ted
lay
e
r
to
o
b
tain
a
f
in
e
-
tu
n
ed
m
o
d
el.
T
wo
d
if
f
er
en
t
p
r
e
-
tr
ain
ed
m
o
d
els
ar
e
u
s
ed
v
ia
th
e
tf
.
k
er
as.a
p
p
licatio
n
s
lib
r
ar
y
t
o
in
itiate
two
f
in
e
-
t
u
n
ed
m
o
d
el
s
.
T
h
e
T
en
s
o
r
Flo
w
API
with
f
u
n
ctio
n
al
t
y
p
es
is
im
p
lem
en
ted
th
r
o
u
g
h
th
is
s
tag
e
s
o
th
at
th
e
ar
ch
itectu
r
al
lay
er
s
o
f
th
e
m
o
d
el
ca
n
b
e
s
ee
n
.
T
h
e
ar
ch
itectu
r
e
o
f
th
e
f
in
e
-
tu
n
ed
m
o
d
el
i
n
itiated
f
r
o
m
th
e
p
r
e
-
tr
ai
n
ed
R
esNet5
0
V2
m
o
d
el.
T
h
is
m
o
d
el
h
as
a
to
tal
o
f
2
4
,
0
9
0
,
3
7
2
p
ar
am
eter
s
an
d
2
4
,
0
4
4
,
9
3
2
p
ar
am
eter
s
th
at
ca
n
b
e
tr
ai
n
ed
.
T
h
e
ar
c
h
itectu
r
e
o
f
th
e
f
in
e
-
tu
n
ed
m
o
d
el
in
itialized
f
r
o
m
th
e
p
r
e
-
tr
ain
ed
Den
s
eNe
t1
6
9
m
o
d
el.
T
h
i
s
m
o
d
el
h
as
a
to
tal
n
u
m
b
er
o
f
p
ar
am
eter
s
o
f
1
3
,
1
1
0
,
5
3
2
an
d
1
2
,
9
5
2
,
1
3
2
p
a
r
am
eter
s
th
at
ca
n
b
e
t
r
ain
ed
.
3
.
3
.
Resul
t
s
o
f
cla
s
s
if
ica
t
io
n
m
o
del a
cc
ura
cy
e
v
a
lua
t
io
n
T
h
e
ev
alu
atio
n
o
f
th
e
tr
ai
n
in
g
m
o
d
el
is
d
o
n
e
u
s
in
g
ac
cu
r
ac
y
an
d
lo
s
s
r
ec
o
r
d
s
f
r
o
m
th
e
d
ee
p
lear
n
in
g
m
o
d
el
p
r
o
ce
s
s
.
T
r
ain
in
g
e
v
alu
atio
n
aim
s
to
ass
ess
th
e
m
o
d
el'
s
p
er
f
o
r
m
an
ce
o
n
b
o
th
tr
a
in
in
g
an
d
v
alid
atio
n
d
ata.
T
h
is
s
tag
e
is
cr
u
cial
to
d
eter
m
in
e
wh
eth
e
r
th
e
m
o
d
el
h
as
u
n
d
e
r
s
to
o
d
p
atter
n
s
in
th
e
d
ata
o
r
is
m
er
ely
m
ak
in
g
r
an
d
o
m
g
u
ess
es.
I
n
th
is
p
h
ase,
lin
e
g
r
ap
h
s
ar
e
u
s
ed
to
v
is
u
alize
ac
cu
r
ac
y
an
d
lo
s
s
v
alu
es
at
ea
ch
ep
o
ch
.
T
h
ese
lin
e
g
r
ap
h
s
in
clu
d
e
tr
ain
in
g
d
ata
ac
cu
r
ac
y
,
v
alid
atio
n
d
ata
ac
cu
r
ac
y
,
t
r
ain
in
g
d
ata
lo
s
s
,
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
A
n
o
p
timiz
ed
tr
a
n
s
fer lea
r
n
in
g
-
b
a
s
ed
a
p
p
r
o
a
ch
fo
r
C
r
o
cid
o
lo
mia
p
a
vo
n
a
n
a
la
r
va
e
…
(
R
is
n
a
w
a
ti
)
2277
v
alid
atio
n
d
ata
lo
s
s
,
t
h
e
b
est
ep
o
ch
b
ased
o
n
ac
cu
r
ac
y
an
d
lo
s
s
,
as
well
as
t
h
e
b
est
v
alu
es
f
o
r
ac
cu
r
ac
y
a
n
d
lo
s
s
.
E
ac
h
g
r
ap
h
co
n
s
is
ts
o
f
t
wo
s
u
b
p
lo
ts
d
is
p
lay
in
g
lo
s
s
a
n
d
ac
cu
r
ac
y
.
A
g
o
o
d
m
o
d
el
is
o
n
e
with
lo
w
lo
s
s
an
d
h
i
g
h
ac
c
u
r
ac
y
.
T
h
e
x
a
n
d
y
ax
is
v
alu
es
ar
e
s
et
to
t
h
e
s
a
m
e
s
ca
le
to
p
r
o
v
i
d
e
a
p
r
o
p
o
r
ti
o
n
al
r
e
p
r
esen
tatio
n
o
f
th
e
tr
ain
i
n
g
p
r
o
ce
s
s
f
o
r
ea
c
h
m
o
d
el.
Fig
u
r
e
6
s
h
o
ws
th
e
p
r
o
g
r
ess
o
f
th
e
tr
ai
n
in
g
p
r
o
ce
s
s
f
o
r
a
m
ac
h
in
e
lear
n
in
g
m
o
d
el
b
as
ed
o
n
t
h
e
p
r
e
-
tr
ain
ed
R
esNet5
0
V2
m
o
d
el
o
v
er
3
0
e
p
o
ch
s
.
T
h
e
tr
ain
i
n
g
lo
s
s
d
ata
d
em
o
n
s
tr
ates
a
s
tead
y
d
ec
r
ea
s
e
f
r
o
m
8
0
to
wa
r
d
s
ze
r
o
,
a
n
d
th
e
t
r
ain
in
g
d
ata
ac
cu
r
ac
y
s
h
o
ws
a
s
ta
b
le
in
cr
ea
s
e
t
o
war
d
s
1
0
0
%.
Valid
atio
n
lo
s
s
d
ata
s
h
o
ws
a
d
ec
r
ea
s
in
g
tr
en
d
.
Me
an
wh
ile,
v
alid
atio
n
d
ata
ac
cu
r
ac
y
h
as
an
u
p
war
d
tr
en
d
with
two
lo
w
-
v
alu
e
d
i
p
s
.
T
h
e
b
est
ep
o
ch
b
ased
o
n
lo
s
s
o
cc
u
r
s
at
ep
o
ch
2
7
with
a
l
o
s
s
v
alu
e
o
f
0
.
1
4
0
.
B
ased
o
n
ac
cu
r
ac
y
,
th
e
b
est
ep
o
ch
also
o
cc
u
r
s
at
ep
o
c
h
2
7
with
a
v
alid
atio
n
ac
cu
r
ac
y
o
f
9
4
.
1
1
%.
Fig
u
r
e
6
.
R
esNet5
0
V2
ac
cu
r
a
cy
m
o
d
el
r
esu
lt
Fig
u
r
e
7
s
h
o
ws
th
e
p
r
o
g
r
ess
o
f
th
e
m
ac
h
in
e
lear
n
i
n
g
m
o
d
el
tr
ain
in
g
p
r
o
ce
s
s
b
ased
o
n
th
e
p
r
e
-
tr
ain
ed
Den
s
eNe
t1
6
9
m
o
d
el
f
o
r
3
0
ep
o
ch
s
.
T
h
e
tr
ain
in
g
lo
s
s
d
ata
s
h
o
ws
a
s
tab
le
d
ec
r
ea
s
in
g
v
alu
e
f
r
o
m
1
.
8
to
war
d
s
ze
r
o
an
d
t
h
e
tr
ain
in
g
d
ata
ac
c
u
r
ac
y
also
s
h
o
ws
a
s
tab
le
in
cr
ea
s
in
g
tr
en
d
to
war
d
s
1
0
0
%.
Valid
atio
n
lo
s
s
d
ata
s
h
o
ws
a
d
o
wn
wa
r
d
tr
e
n
d
.
M
ea
n
wh
ile,
th
e
ac
cu
r
ac
y
o
f
th
e
v
alid
atio
n
d
ata
h
as
an
u
p
w
ar
d
tr
e
n
d
with
two
v
alley
s
with
lo
w
v
alu
es.
T
h
e
b
est
ep
o
ch
b
ased
o
n
lo
s
s
o
cc
u
r
r
ed
at
th
e
2
9
th
ep
o
c
h
with
a
l
o
s
s
v
alu
e
o
f
0
.
0
9
3
.
B
ased
o
n
ac
cu
r
ac
y
,
th
e
b
est ep
o
ch
o
cc
u
r
r
ed
in
th
e
2
6
th
e
p
o
c
h
with
a
v
alid
atio
n
ac
cu
r
ac
y
v
alu
e
o
f
9
7
.
0
5
%.
T
h
e
ev
alu
atio
n
r
esu
lts
o
f
th
e
d
ee
p
lear
n
in
g
m
o
d
el
tr
ai
n
in
g
p
r
o
ce
s
s
s
h
o
w
th
at
all
m
o
d
els
h
av
e
an
av
er
ag
e
ac
cu
r
ac
y
o
n
tr
ain
in
g
d
ata
an
d
v
alid
atio
n
d
ata
ab
o
v
e
8
0
%.
T
h
e
tr
en
d
o
f
in
cr
ea
s
i
n
g
ac
cu
r
ac
y
v
al
u
es
to
war
d
s
1
0
0
%
in
v
alid
atio
n
d
ata
in
d
icate
s
th
at
th
e
m
o
d
el
h
as
b
ee
n
ab
le
to
u
n
d
er
s
tan
d
p
a
tter
n
s
in
th
e
d
ata
i
n
m
ak
in
g
class
if
icatio
n
s
.
T
h
e
ac
cu
r
ac
y
an
d
lo
s
s
o
b
tain
e
d
at
th
is
s
tag
e
in
d
icate
th
at
th
e
m
o
d
el
ca
n
b
e
u
s
ed
f
o
r
f
u
r
th
er
e
v
alu
atio
n
u
s
in
g
test
d
ata.
T
ab
le
2
s
h
o
ws th
e
r
esu
lts
o
f
m
o
d
el
ev
alu
atio
n
u
s
in
g
th
e
ev
alu
ate(
)
f
u
n
ctio
n
.
T
ab
le
2
s
h
o
ws
th
e
ac
cu
r
a
cy
an
d
lo
s
s
o
f
ea
ch
m
o
d
el
o
n
tr
ain
in
g
d
ata,
v
alid
atio
n
d
ata
a
n
d
test
in
g
d
ata.
T
h
e
Den
s
eNe
t1
6
9
m
o
d
el
ac
h
iev
ed
th
e
h
ig
h
est
ac
cu
r
ac
y
o
n
tr
ain
i
n
g
d
ata
o
f
9
6
.
8
%
an
d
th
e
s
m
a
lles
t
lo
s
s
o
f
0
.
0
9
0
,
v
alid
atio
n
d
ata
o
f
9
6
.
8
% a
n
d
t
h
e
s
m
allest lo
s
s
o
f
0
.
1
3
3
,
test
in
g
d
ata
o
f
9
7
.
1
% with
th
e
s
m
a
lles
t lo
s
s
o
f
0
.
1
1
2
.
Fig
u
r
e
7
.
Den
s
eNe
t1
6
9
ac
cu
r
a
cy
m
o
d
el
r
esu
lt
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
14
,
No
.
3
,
J
u
n
e
2
0
2
5
:
2
2
7
0
-
2
2
8
1
2278
T
ab
le
2
.
R
esu
lts
o
f
m
o
d
el
e
v
al
u
atio
n
with
ac
cu
r
ac
y
an
d
lo
s
s
M
o
d
e
l
Tr
a
i
n
d
a
t
a
V
a
l
d
a
t
a
Te
st
d
a
t
a
A
c
c
Lo
ss
A
c
c
Lo
ss
A
c
c
Lo
ss
R
e
sN
e
t
5
0
V
2
0
.
9
3
7
0
.
1
2
1
0
.
9
0
6
0
.
1
6
0
0
.
9
4
2
0
.
1
5
6
D
e
n
seN
e
t
1
6
9
0
.
9
6
8
0
.
0
9
0
0
.
9
6
8
0
.
1
3
3
0
.
9
7
1
0
.
1
1
2
Fig
u
r
e
8
class
if
icatio
n
r
esu
lts
f
r
o
m
o
n
e
o
f
C
.
p
a
vo
n
a
n
a
'
s
im
ag
es,
test
d
ata
co
m
p
leted
with
in
s
tar
2
with
th
e
R
esNet5
0
V2
an
d
De
n
s
eNe
t1
6
9
m
o
d
els.
B
ased
o
n
T
ab
le
3
,
it
was
o
b
tain
ed
f
r
o
m
v
ar
io
u
s
test
s
o
n
th
e
Den
s
eNe
t1
6
9
m
o
d
el,
in
th
e
f
ir
s
t
test
u
s
in
g
a
b
atch
s
ize
o
f
3
2
an
d
u
s
in
g
1
a
d
d
itio
n
al
lay
e
r
,
n
am
ely
,
1
d
r
o
p
o
u
t
lay
er
an
d
1
d
en
s
e
lay
er
an
d
r
u
n
n
in
g
f
o
r
3
0
ep
o
ch
s
,
th
e
ac
cu
r
ac
y
o
n
th
e
test
d
ata
was
9
1
%
an
d
lo
s
s
o
f
0
.
2
5
.
I
n
th
e
s
ec
o
n
d
test
,
u
s
in
g
th
e
s
am
e
n
u
m
b
e
r
o
f
b
atc
h
s
izes
wi
th
3
ad
d
itio
n
al
lay
e
r
s
,
3
d
r
o
p
o
u
t
lay
er
s
,
3
d
e
n
s
e
lay
er
s
an
d
r
u
n
n
in
g
f
o
r
3
0
ep
o
ch
s
,
th
e
ac
cu
r
ac
y
o
n
th
e
test
d
ata
was
9
7
.
1
%
a
n
d
t
h
e
lo
s
s
was
0
.
1
1
.
T
h
e
th
ir
d
an
d
f
o
u
r
th
tr
ials
ca
r
r
ied
o
u
t
u
s
in
g
a
b
atch
s
ize
o
f
6
4
a
n
d
u
s
in
g
1
o
r
3
a
d
d
itio
n
al
lay
er
s
ex
p
er
ien
ce
d
e
r
r
o
r
s
wh
en
r
u
n
n
in
g
m
o
d
el
t
r
ain
in
g
.
Fig
u
r
e
8
.
C
lass
if
icatio
n
r
esu
lts
with
R
esNet5
0
V2
an
d
Den
s
eNe
t1
6
9
m
o
d
els
T
ab
le
3
.
C
o
m
p
a
r
is
o
n
r
esu
lts
o
f
ac
cu
r
ac
y
an
d
l
o
s
s
o
f
Den
s
e
N
et1
6
9
m
o
d
el
e
p
o
c
h
B
a
t
c
h
si
z
e
A
d
d
l
a
y
e
r
Tr
a
i
n
d
a
t
a
V
a
l
d
a
t
a
Te
st
d
a
t
a
A
c
c
Lo
ss
A
c
c
Lo
ss
A
c
c
Lo
ss
30
32
1
0
.
9
3
0
.
1
2
0
.
9
0
0
.
1
6
0
.
9
4
0
.
1
5
30
32
3
0
.
9
6
0
.
1
2
0
.
8
7
0
.
5
4
0
.
8
6
0
.
3
6
45
64
1
0
.
9
8
0
.
1
1
0
.
8
9
0
.
2
6
0
.
9
2
0
.
1
3
35
64
3
0
.
9
3
0
.
1
2
0
.
8
9
0
.
3
4
0
.
9
2
0
.
1
2
4.
CO
NCLU
SI
O
N
B
ased
o
n
th
e
im
p
lem
en
tatio
n
r
esu
lts
o
f
C
.
p
a
vo
n
a
n
a
lar
v
a
id
en
tific
atio
n
u
s
in
g
th
e
C
NN
m
o
d
el
as
well
as
th
e
p
r
e
-
tr
ain
ed
ar
c
h
itectu
r
es
R
esNet5
0
V2
an
d
Den
s
eNe
t1
6
9
o
n
C
.
p
a
vo
n
a
n
a
im
ag
es
in
th
is
s
tu
d
y
,
s
ev
er
al
co
n
clu
s
io
n
s
ca
n
b
e
d
r
awn
.
T
h
e
ac
cu
r
ac
y
o
f
th
e
s
ec
o
n
d
tr
ial
o
f
th
e
Den
s
eNe
t1
6
9
m
o
d
el
y
ield
ed
th
e
m
o
s
t
o
p
tim
al
ac
c
u
r
ac
y
an
d
lo
s
s
,
with
ac
cu
r
ac
y
ex
ce
e
d
in
g
9
0
%,
an
d
th
e
lo
s
s
f
ig
u
r
es
f
o
r
ea
ch
tr
ain
in
g
d
ata,
v
alid
atio
n
d
ata,
a
n
d
test
d
ata
wer
e
r
elativ
ely
clo
s
e.
T
h
is
in
d
icate
s
th
at
th
er
e
was
n
o
o
v
e
r
f
itti
n
g
in
th
e
f
ir
s
t
tr
ial.
I
n
th
e
f
ir
s
t
tr
ial,
th
e
ac
c
u
r
ac
y
was
ab
o
v
e
9
0
%,
b
u
t
th
e
l
o
s
s
f
ig
u
r
es
f
o
r
ea
c
h
tr
ain
in
g
d
ata,
v
alid
atio
n
d
ata,
an
d
test
d
ata
s
h
o
wed
s
ig
n
if
ican
t
d
if
f
er
en
ce
s
,
in
d
icatin
g
th
a
t
th
e
m
o
d
el
ex
p
er
ien
ce
d
o
v
er
f
itti
n
g
.
Me
an
wh
ile,
th
e
th
ir
d
a
n
d
f
o
u
r
th
tr
ials
co
u
l
d
n
o
t
b
e
ex
ec
u
te
d
d
u
e
to
er
r
o
r
s
.
Me
an
wh
ile,
th
e
R
esNet5
0
V2
m
o
d
el
ac
h
iev
ed
an
ac
cu
r
ac
y
o
f
9
4
.
2
%
an
d
a
l
o
s
s
o
f
0
.
1
5
6
.
All
m
o
d
els
d
e
m
o
n
s
tr
ated
g
o
o
d
p
e
r
f
o
r
m
an
ce
o
n
th
e
test
d
ata
with
ac
cu
r
ac
y
ab
o
v
e
9
0
%.
I
n
f
u
tu
r
e
r
esear
ch
,
it
is
ex
p
ec
ted
th
at
th
e
d
ev
elo
p
m
e
n
t
o
f
th
e
d
ee
p
l
ea
r
n
in
g
m
o
d
el
f
o
r
class
if
y
in
g
C
.
p
a
vo
n
a
n
a
lar
v
ae
will
b
e
ca
r
r
ied
o
u
t
u
s
in
g
o
th
er
p
r
e
-
tr
ai
n
ed
m
eth
o
d
s
to
ac
h
iev
e
ev
en
m
o
r
e
o
p
tim
al
ac
cu
r
ac
y
.
ACK
NO
WL
E
DG
E
M
E
NT
S
T
h
e
au
th
o
r
s
wo
u
ld
lik
e
to
e
x
p
r
ess
o
u
r
ap
p
r
ec
iatio
n
to
t
h
e
R
esear
ch
I
n
s
titu
tio
n
s
o
f
Gu
n
ad
ar
m
a
Un
iv
er
s
ity
f
o
r
th
eir
co
n
tin
u
ed
f
ac
ilit
atio
n
an
d
s
u
p
p
o
r
t th
r
o
u
g
h
o
u
t th
e
r
esear
ch
ac
tiv
ities
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
A
n
o
p
timiz
ed
tr
a
n
s
fer lea
r
n
in
g
-
b
a
s
ed
a
p
p
r
o
a
ch
fo
r
C
r
o
cid
o
lo
mia
p
a
vo
n
a
n
a
la
r
va
e
…
(
R
is
n
a
w
a
ti
)
2279
F
UNDING
I
NF
O
R
M
A
T
I
O
N
T
h
is
r
esear
ch
was
f
u
n
d
e
d
b
y
th
e
Fu
n
d
am
en
tal
R
esear
ch
Gr
an
t
f
r
o
m
th
e
Min
is
tr
y
o
f
R
esear
ch
,
T
ec
h
n
o
lo
g
y
,
an
d
Hig
h
e
r
E
d
u
ca
tio
n
(
Kem
en
r
is
tek
d
ik
ti)
u
n
d
er
co
n
tr
ac
t
n
u
m
b
er
L
L
DI
KT
I
3
Nu
m
b
er
7
9
5
/LL
3
/AL
.
0
4
/2
0
2
4
,
d
ated
J
u
n
e
2
6
,
2
0
2
4
.
AUTHO
R
CO
NT
RI
B
UT
I
O
NS ST
A
T
E
M
E
N
T
T
h
is
jo
u
r
n
al
u
s
es
th
e
C
o
n
tr
ib
u
to
r
R
o
les
T
ax
o
n
o
m
y
(
C
R
ed
iT)
to
r
ec
o
g
n
ize
in
d
iv
id
u
al
au
th
o
r
co
n
tr
ib
u
tio
n
s
,
r
ed
u
ce
au
th
o
r
s
h
ip
d
is
p
u
tes,
an
d
f
ac
ilit
ate
co
llab
o
r
atio
n
.
Na
m
e
o
f
Aut
ho
r
C
M
So
Va
Fo
I
R
D
O
E
Vi
Su
P
Fu
R
is
n
awa
ti
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
R
o
d
iah
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
Sar
if
u
d
d
in
Ma
d
en
d
a
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
Dian
a
T
r
i
Su
s
etian
in
g
tias
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
C
:
C
o
n
c
e
p
t
u
a
l
i
z
a
t
i
o
n
M
:
M
e
t
h
o
d
o
l
o
g
y
So
:
So
f
t
w
a
r
e
Va
:
Va
l
i
d
a
t
i
o
n
Fo
:
Fo
r
mal
a
n
a
l
y
s
i
s
I
:
I
n
v
e
s
t
i
g
a
t
i
o
n
R
:
R
e
so
u
r
c
e
s
D
:
D
a
t
a
C
u
r
a
t
i
o
n
O
:
W
r
i
t
i
n
g
-
O
r
i
g
i
n
a
l
D
r
a
f
t
E
:
W
r
i
t
i
n
g
-
R
e
v
i
e
w
&
E
d
i
t
i
n
g
Vi
:
Vi
su
a
l
i
z
a
t
i
o
n
Su
:
Su
p
e
r
v
i
s
i
o
n
P
:
P
r
o
j
e
c
t
a
d
mi
n
i
st
r
a
t
i
o
n
Fu
:
Fu
n
d
i
n
g
a
c
q
u
i
si
t
i
o
n
CO
NF
L
I
C
T
O
F
I
N
T
E
R
E
S
T
ST
A
T
E
M
E
NT
Au
th
o
r
s
s
tate
n
o
co
n
f
lict o
f
in
t
er
est
.
I
NF
O
RM
E
D
CO
NS
E
N
T
T
h
is
s
tu
d
y
d
id
n
o
t in
v
o
lv
e
h
u
m
an
p
ar
ticip
an
ts
; th
er
e
f
o
r
e,
in
f
o
r
m
ed
co
n
s
en
t w
as n
o
t
r
eq
u
ir
ed
.
E
T
H
I
CAL AP
P
RO
V
AL
T
h
is
s
tu
d
y
u
tili
ze
d
p
est
d
ataset
an
d
d
id
n
o
t
in
v
o
lv
e
h
u
m
a
n
o
r
v
er
te
b
r
ate
a
n
im
al
s
u
b
jects
;
th
er
ef
o
r
e,
eth
ical
ap
p
r
o
v
al
was n
o
t r
eq
u
i
r
ed
.
DATA AV
AI
L
AB
I
L
I
T
Y
T
h
e
d
ata
c
o
llected
f
o
r
th
is
s
tu
d
y
ar
e
p
ar
t
o
f
o
n
g
o
in
g
r
esear
ch
with
p
o
te
n
tial
c
o
m
m
er
cial
ap
p
licatio
n
s
.
T
h
er
e
f
o
r
e,
we
a
r
e
u
n
ab
le
to
m
a
k
e
th
e
d
ata
p
u
b
licly
av
ailab
le
at
t
h
is
tim
e
d
u
e
to
in
tellectu
al
p
r
o
p
er
t
y
r
estrictio
n
s
an
d
th
e
o
n
g
o
in
g
co
m
m
er
cializa
tio
n
p
r
o
ce
s
s
.
Ho
wev
er
,
r
esear
ch
er
s
in
ter
ested
in
ac
ce
s
s
in
g
th
e
d
ataset
ar
e
wel
co
m
e
to
c
o
n
tact
th
e
co
r
r
esp
o
n
d
in
g
au
th
o
r
at
r
o
d
iah
@
s
taf
f
.
g
u
n
ad
a
r
m
a.
ac
.
id
f
o
r
f
u
r
th
er
i
n
f
o
r
m
atio
n
an
d
to
d
is
cu
s
s
p
o
ten
tial d
ata
s
h
ar
in
g
u
n
d
er
ap
p
r
o
p
r
iate
co
n
d
itio
n
s
.
RE
F
E
R
E
NC
E
S
[
1
]
F
.
P
.
A
l
l
o
g
g
i
a
,
R
.
F
.
B
a
f
u
m
o
,
D
.
A
.
R
a
m
i
r
e
z
,
M
.
A
.
M
a
z
a
,
a
n
d
A
.
B
.
C
a
m
a
r
g
o
,
“
B
r
a
ss
i
c
a
c
e
a
e
m
i
c
r
o
g
r
e
e
n
s
:
A
n
o
v
e
l
a
n
d
p
r
o
m
i
s
s
o
r
y
s
o
u
r
c
e
o
f
s
u
s
t
a
i
n
a
b
l
e
b
i
o
a
c
t
i
v
e
c
o
m
p
o
u
n
d
s
,
”
C
u
rr
e
n
t
R
e
se
a
r
c
h
i
n
F
o
o
d
S
c
i
e
n
c
e
,
v
o
l
.
6
,
2
0
2
3
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
c
r
f
s
.
2
0
2
3
.
1
0
0
4
8
0
.
[
2
]
M
.
H
a
sa
n
u
z
z
a
ma
n
,
S
.
A
r
a
ú
j
o
,
a
n
d
S
.
S
.
G
i
l
l
,
T
h
e
p
l
a
n
t
f
a
m
i
l
y
f
a
b
a
c
e
a
e
:
Bi
o
l
o
g
y
a
n
d
p
h
y
s
i
o
l
o
g
i
c
a
l
res
p
o
n
s
e
s
t
o
e
n
v
i
ro
n
m
e
n
t
a
l
st
ress
e
s
,
S
p
r
i
n
g
e
r
S
i
n
g
a
p
o
r
e
,
2
0
2
0
,
d
o
i
:
1
0
.
1
0
0
7
/
9
7
8
-
981
-
15
-
4
7
5
2
-
2.
[
3
]
D
.
R
a
mi
r
e
z
,
A
.
A
b
e
l
l
á
n
-
V
i
c
t
o
r
i
o
,
V
.
B
e
r
e
t
t
a
,
A
.
C
a
marg
o
,
a
n
d
D
.
A
.
M
o
r
e
n
o
,
“
F
u
n
c
t
i
o
n
a
l
i
n
g
r
e
d
i
e
n
t
s
f
r
o
m
b
r
a
s
si
c
a
c
e
a
e
sp
e
c
i
e
s
:
O
v
e
r
v
i
e
w
a
n
d
p
e
r
sp
e
c
t
i
v
e
s,
”
I
n
t
e
r
n
a
t
i
o
n
a
l
J
o
u
r
n
a
l
o
f
M
o
l
e
c
u
l
a
r
S
c
i
e
n
c
e
s
,
v
o
l
.
2
1
,
n
o
.
6
,
2
0
2
0
,
d
o
i
:
1
0
.
3
3
9
0
/
i
j
m
s2
1
0
6
1
9
9
8
.
[
4
]
H
.
S
.
H
i
l
l
e
n
,
G
.
K
o
k
i
c
,
L
.
F
a
r
n
u
n
g
,
C
.
D
i
e
n
e
ma
n
n
,
D
.
Te
g
u
n
o
v
,
a
n
d
P
.
C
r
a
m
e
r
,
“
S
t
r
u
c
t
u
r
e
o
f
r
e
p
l
i
c
a
t
i
n
g
S
A
R
S
-
C
o
V
-
2
p
o
l
y
meras
e
,
”
N
a
t
u
r
e
,
v
o
l
.
5
8
4
,
n
o
.
7
8
1
9
,
p
p
.
1
5
4
–
1
5
6
,
2
0
2
0
,
d
o
i
:
1
0
.
1
0
3
8
/
s
4
1
5
8
6
-
0
2
0
-
2
3
6
8
-
8.
[
5
]
S
.
S
r
i
n
i
v
a
sa
n
e
t
a
l
.
,
“
S
t
r
u
c
t
u
r
a
l
g
e
n
o
mi
c
s
o
f
S
A
R
S
-
C
o
V
-
2
i
n
d
i
c
a
t
e
s,
”
V
i
r
u
ses
,
v
o
l
.
1
2
,
n
o
.
3
6
0
,
p
p
.
1
–
1
7
,
2
0
2
0
.
[
6
]
N
.
M
p
u
m
i
,
R
.
S
.
M
a
c
h
u
n
d
a
,
K
.
M
.
M
t
e
i
,
a
n
d
P
.
A
.
N
d
a
k
i
d
e
mi
,
“
S
e
l
e
c
t
e
d
i
n
sec
t
p
e
st
s
o
f
e
c
o
n
o
m
i
c
i
mp
o
r
t
a
n
c
e
t
o
B
r
a
ss
i
c
a
o
l
e
r
a
c
e
a
,
t
h
e
i
r
c
o
n
t
r
o
l
s
t
r
a
t
e
g
i
e
s
a
n
d
t
h
e
p
o
t
e
n
t
i
a
l
t
h
r
e
a
t
t
o
e
n
v
i
r
o
n
m
e
n
t
a
l
p
o
l
l
u
t
i
o
n
i
n
A
f
r
i
c
a
,
”
S
u
st
a
i
n
a
b
i
l
i
t
y
,
v
o
l
.
1
2
,
n
o
.
9
,
2
0
2
0
,
d
o
i
:
1
0
.
3
3
9
0
/
s
u
1
2
0
9
3
8
2
4
.
[
7
]
P
.
B
.
A
n
g
o
n
e
t
a
l
.
,
“
I
n
t
e
g
r
a
t
e
d
p
e
s
t
ma
n
a
g
e
me
n
t
(
I
P
M
)
i
n
a
g
r
i
c
u
l
t
u
r
e
a
n
d
i
t
s
r
o
l
e
i
n
mai
n
t
a
i
n
i
n
g
e
c
o
l
o
g
i
c
a
l
b
a
l
a
n
c
e
a
n
d
b
i
o
d
i
v
e
r
si
t
y
,
”
Ad
v
a
n
c
e
s
i
n
A
g
ri
c
u
l
t
u
r
e
,
v
o
l
.
2
0
2
3
,
2
0
2
3
,
d
o
i
:
1
0
.
1
1
5
5
/
2
0
2
3
/
5
5
4
6
3
7
3
.
[
8
]
R
.
C
o
l
l
i
e
r
,
“
P
e
st
i
n
sec
t
ma
n
a
g
e
me
n
t
i
n
v
e
g
e
t
a
b
l
e
c
r
o
p
s
g
r
o
w
n
o
u
t
d
o
o
r
s
i
n
n
o
r
t
h
e
r
n
E
u
r
o
p
e
–
a
p
p
r
o
a
c
h
e
s
a
t
t
h
e
b
o
t
t
o
m
o
f
t
h
e
I
P
M
p
y
r
a
mi
d
,
”
Fr
o
n
t
i
e
rs
i
n
H
o
r
t
i
c
u
l
t
u
re
,
v
o
l
.
2
,
p
p
.
1
–
1
1
,
A
p
r
.
2
0
2
3
,
d
o
i
:
1
0
.
3
3
8
9
/
f
h
o
r
t
.
2
0
2
3
.
1
1
5
9
3
7
5
.
Evaluation Warning : The document was created with Spire.PDF for Python.