I
AE
S
I
n
t
e
r
n
at
ion
al
Jou
r
n
al
of
Ar
t
if
icial
I
n
t
e
ll
ig
e
n
c
e
(
I
J
-
AI
)
Vol.
14,
No.
4
,
Augus
t
2025
,
pp.
3121
~
3132
I
S
S
N:
2252
-
8938
,
DO
I
:
10
.
11591/i
jai
.
v
14
.i
4
.
pp
31
21
-
3132
3121
Jou
r
n
al
h
omepage
:
ht
tp:
//
ij
ai
.
iaes
c
or
e
.
c
om
E
n
h
an
c
in
g p
r
e
c
is
io
n
agr
i
c
u
lt
u
r
e
:
a c
o
m
p
r
e
h
e
n
si
ve
i
n
v
e
st
ig
at
io
n
in
t
o p
a
t
h
og
e
n
d
e
t
e
c
t
io
n
a
n
d
m
an
age
m
e
n
t
S
h
ais
t
a
F
ar
h
a
t
,
Chok
k
a
Anu
r
ad
h
a
D
e
pa
r
tm
e
nt
of
C
omput
e
r
S
c
ie
nc
e
a
nd E
ngi
ne
e
r
in
g, K
one
r
u L
a
ks
hma
ia
h
E
duc
a
ti
on F
ounda
ti
on,
V
ij
a
ya
w
a
da
,
I
ndi
a
Ar
t
icle
I
n
f
o
AB
S
T
RA
CT
A
r
ti
c
le
h
is
tor
y
:
R
e
c
e
ived
M
a
y
10,
2024
R
e
vis
e
d
M
a
r
26,
2025
Ac
c
e
pted
J
un
8,
2025
A
g
r
i
cu
l
t
u
re
i
s
an
i
m
p
o
r
t
an
t
s
ec
t
o
r
o
f
In
d
i
a
n
ag
r
o
n
o
my
f
o
r
h
u
ma
n
l
i
v
el
i
h
o
o
d
.
A
l
l
areas
are
affec
t
ed
b
y
t
h
e
effec
t
s
o
f
en
v
i
r
o
n
me
n
t
a
l
t
o
x
i
c
farm
s
,
w
h
i
ch
mak
es
ma
n
ag
i
n
g
v
ar
i
o
u
s
d
i
ffi
cu
l
t
s
i
t
u
at
i
o
n
s
m
o
re
ch
a
l
l
en
g
i
n
g
.
A
g
ri
c
u
l
t
u
re
mu
s
t
ad
o
p
t
n
e
w
t
ech
n
o
l
o
g
y
i
n
acc
o
r
d
an
ce
w
i
t
h
d
a
i
l
y
en
v
i
r
o
n
me
n
t
a
l
ch
a
n
g
e
s
i
f
i
t
i
s
g
o
i
n
g
t
o
b
en
ef
i
t
fro
m
a
cro
p
fro
m
t
h
e
p
er
s
p
ec
t
i
v
e
s
o
f
farmers
an
d
en
d
u
s
er
s
.
Farmers
w
i
l
l
b
en
ef
i
t
fr
o
m
earl
y
d
et
ec
t
i
o
n
o
f
ag
ri
c
u
l
t
u
ra
l
d
i
s
eas
e
s
rat
h
er
t
h
an
r
i
s
k
i
n
g
t
h
e
i
r
l
i
v
e
s
i
n
d
a
n
g
er
o
u
s
ci
rc
u
ms
t
a
n
ces
.
Co
m
p
u
t
er
t
ech
n
o
l
o
g
y
w
i
l
l
b
e
v
ery
h
e
l
p
f
u
l
i
n
mai
n
t
ai
n
i
n
g
s
u
s
t
ai
n
a
b
l
e
an
d
h
eal
t
h
y
cro
p
s
fo
r
t
h
e
o
b
j
ec
t
i
v
e
o
f
i
d
en
t
i
f
y
i
n
g
cr
o
p
d
i
s
eas
e
s
i
n
ad
d
i
t
i
o
n
t
o
t
h
e
farmer's
c
l
o
s
e
o
b
s
erv
a
t
i
o
n
.
D
ee
p
l
ear
n
i
n
g
(D
L
)
t
ec
h
n
i
q
u
es
are
v
er
y
i
n
f
l
u
e
n
t
i
al
am
o
n
g
v
ari
o
u
s
co
mp
u
t
i
n
g
t
ec
h
n
o
l
o
g
i
es
.
In
t
h
i
s
w
o
r
k
,
w
e
ex
p
l
o
re
s
ev
eral
cu
rre
n
t
ap
p
r
o
ach
e
s
t
o
p
rec
i
s
i
o
n
ag
ri
c
u
l
t
u
re,
s
u
ch
as
art
i
fi
c
i
al
i
n
t
e
l
l
i
g
e
n
ce
(
AI
)
,
D
L
,
an
d
mac
h
i
n
e
l
ear
n
i
n
g
(ML
).
T
h
e
fi
n
d
i
n
g
s
o
f
t
h
e
s
t
u
d
y
mak
e
cl
ear
m
o
d
er
n
met
h
o
d
s
,
t
h
e
i
r
d
raw
b
ack
s
,
an
d
t
h
e
k
n
o
w
l
e
d
g
e
l
ack
i
n
g
t
h
a
t
n
ee
d
s
t
o
b
e
ad
d
re
s
s
e
d
t
o
e
x
p
l
o
re
p
rec
i
s
i
o
n
ag
r
i
cu
l
t
u
re
fu
l
l
y
.
K
e
y
w
o
r
d
s
:
Agr
icultur
e
C
omput
e
r
vis
ion
C
onvolut
ion
ne
ur
a
l
ne
twor
k
De
e
p
lea
r
ning
T
r
a
ns
f
e
r
lea
r
ning
Th
i
s
i
s
a
n
o
p
en
a
c
ces
s
a
r
t
i
c
l
e
u
n
d
e
r
t
h
e
CC
B
Y
-
SA
l
i
ce
n
s
e.
C
or
r
e
s
pon
din
g
A
u
th
or
:
C
hokka
Anur
a
dha
De
pa
r
tm
e
nt
of
C
omput
e
r
S
c
ienc
e
a
nd
E
nginee
r
ing
,
Kone
r
u
L
a
ks
hmaia
h
E
duc
a
ti
on
F
ounda
ti
on
Va
dde
s
wa
r
a
m,
Vijaya
wa
da
,
Andhr
a
P
r
a
d
e
s
h,
I
ndia
E
mail:
d
r
a
nur
a
dha
@kluni
ve
r
s
it
y.
in
1.
I
NT
RODU
C
T
I
ON
I
ndia
is
mos
tl
y
a
n
a
gr
icultur
a
l
de
pe
nde
nt
e
c
onomi
c
r
e
gion,
a
nd
thi
s
s
e
c
tor
e
mpl
oys
a
r
ound
58%
o
f
the
population,
making
it
the
ba
c
kbone
of
the
c
ou
ntr
y.
M
a
ny
plant
s
pe
c
ies
us
e
d
in
a
gr
icultur
e
ha
ve
be
c
ome
e
xti
nc
t
a
s
a
r
e
s
ult
of
global
wa
r
mi
ng
a
nd
other
f
a
c
to
r
s
,
a
nd
including
de
f
o
r
e
s
tation,
ove
r
the
pa
s
t
f
e
w
ye
a
r
s
.
T
he
gr
owing
of
f
ood
is
e
s
s
e
nti
a
l
to
the
population's
a
bil
it
y
to
maintain
their
e
xis
tenc
e
.
As
pe
r
a
gr
icultur
e
global
mar
ke
t
2023,
i
t
is
a
nti
c
ipate
d
that
by
2050,
ther
e
will
be
mor
e
than
10
bil
li
on
pe
ople
on
e
a
r
t
h.
T
hus
,
the
gr
e
a
tes
t
c
ont
r
ibut
ion
to
the
im
pr
ove
ment
o
f
th
e
na
ti
on's
he
a
lt
hy
pe
ople
a
nd
e
c
onomy
is
the
pr
od
uc
ti
on
of
good
-
qua
li
ty,
f
r
e
e
-
dis
e
a
s
e
c
r
ops
.
F
or
f
a
r
mer
s
,
the
pr
im
a
r
y
is
s
ue
a
f
f
e
c
ti
ng
their
f
inanc
ial
a
nd
s
oc
ial
we
ll
-
be
ing
is
c
r
op
gr
owt
h
unde
r
e
f
f
icie
nt
f
a
r
m
ing
pr
a
c
ti
c
e
s
.
W
e
mus
t
pr
otec
t
plants
f
r
om
dis
e
a
s
e
s
s
o
a
s
to
pr
oduc
e
an
or
ga
nic
yield
[
1
]
.
T
he
r
e
e
xis
t
diver
s
e
c
a
tegor
ies
of
plant
dis
e
a
s
e
s
;
a
mong
them
a
r
e
ba
c
ter
ial
inf
e
c
ti
ons
na
mely:
ye
ll
owing
f
oli
a
ge
,
ba
c
ter
ial
in
f
e
c
ti
on,
r
a
pid
a
nd
wide
s
pr
e
a
d
ti
s
s
ue
de
a
th,
c
a
nke
r
,
c
r
own
ga
ll
,
a
nd
s
c
a
b.
F
igur
e
1
de
picts
the
vi
r
a
l
dis
e
a
s
e
s
that
c
a
n
s
tunt
plant
gr
owth,
s
uc
h
a
s
s
pott
e
d
wilt
,
ps
or
ias
is
,
c
ur
ly
top,
a
nd
mos
a
ic,
a
s
we
ll
a
s
f
unga
l
dis
e
a
s
e
s
,
including
r
us
t,
powde
r
y
m
il
de
w,
a
nd
blac
k
s
pots
.
E
c
os
ys
tem
los
s
will
r
e
s
ult
f
r
o
m
c
r
o
p
los
s
in
a
gr
icultur
e
.
Annua
ll
y
,
f
a
r
mer
s
incur
s
ubs
tantial
f
inanc
ial
los
s
e
s
due
to
the
ha
r
m
that
thes
e
dis
e
a
s
e
s
c
a
us
e
to
their
c
r
ops
.
B
a
s
e
d
on
a
n
ove
r
view
o
f
numer
ous
a
gr
icultur
a
l
s
t
udies
obtaine
d
f
r
om
diver
s
e
l
it
e
r
a
tur
e
s
tudi
e
s
[
2]
,
whic
h
took
int
o
a
c
c
ount
dif
f
e
r
e
nt
I
ndian
s
tate
s
f
r
om
2012
to
2021
[
2]
,
the
a
nti
c
ipate
d
los
s
e
s
a
r
e
s
hown
in
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
Ar
ti
f
I
ntell
,
Vol.
14,
No.
4
,
Augus
t
2025
:
312
1
-
3132
3122
F
igur
e
2
wha
t
f
oll
ows
.
T
he
I
ndian
De
pa
r
tm
e
nt
of
Agr
icultur
e
's
R
is
k
M
a
na
ge
m
e
nt
Age
nc
y
pr
ovided
his
tor
ica
l
los
s
c
a
us
e
da
ta,
whic
h
wa
s
e
mpl
oye
d
to
c
a
l
c
ulate
va
lues
.
T
he
population's
income
a
nd
e
c
onomy
a
r
e
dir
e
c
tl
y
im
pa
c
ted
by
dis
e
a
s
e
s
that
ha
r
m
plants
a
nd
r
e
nd
e
r
them
inef
f
e
c
ti
ve
.
As
is
c
omm
on
knowle
dge
,
t
he
gr
os
s
domes
ti
c
pr
oduc
t
(
GD
P
)
r
a
te
is
di
r
e
c
tl
y
im
pa
c
ted
by
s
tagna
nt
wa
ge
s
,
s
poil
e
d
c
r
ops
,
a
nd
uns
o
ld
goo
ds
.
C
los
e
moni
tor
ing
is
r
e
quir
e
d
a
t
di
f
f
e
r
e
nt
s
tage
s
of
c
r
op
g
r
owth
be
c
a
us
e
it
a
ls
o
pos
e
s
a
thr
e
a
t
to
f
a
r
mer
s
'
li
ve
s
.
W
he
n
plant
dis
e
a
s
e
s
we
r
e
f
ir
s
t
dis
c
ove
r
e
d,
howe
ve
r
,
t
he
y
ha
d
to
be
identi
f
ied
by
s
im
ple
obs
e
r
va
ti
on
with
the
una
ided
e
ye
.
How
e
ve
r
,
t
he
s
e
methods
c
ould
ha
ve
be
e
n
mor
e
labor
ious
a
nd
im
p
r
e
c
is
e
;
ther
e
f
or
e
,
no
wa
da
ys
,
c
omput
e
r
tec
hnologi
e
s
a
r
e
us
e
d
to
de
tec
t
plant
dis
e
a
s
e
s
e
a
r
ly
on
[
2]
.
F
igur
e
1.
Va
r
ious
plant
dis
e
a
s
e
s
F
igur
e
2.
C
r
op
los
s
e
s
by
r
e
gion
f
r
o
m
2012
to
2021
,
e
s
ti
mate
d
by
p
lant
dis
e
a
s
e
T
he
r
e
mus
t
be
a
utom
a
ti
c
ve
r
i
f
ica
ti
on
a
nd
c
las
s
if
ica
ti
on
f
or
va
r
ious
c
r
op
leve
ls
in
or
de
r
to
identi
f
y
a
he
a
lt
hy
c
r
op
c
or
r
e
c
tl
y
.
W
e
will
put
f
or
th
a
mod
e
l
that
will
identif
y
da
mage
d
lea
ve
s
a
nd
their
u
nde
r
lyi
ng
c
a
us
e
,
a
s
we
ll
a
s
indi
c
a
te
the
por
ti
on
of
e
a
c
h
lea
f
that
is
he
a
lt
hy.
De
s
pit
e
s
igni
f
ica
nt
a
dva
nc
e
ments
in
pr
e
c
is
ion
a
gr
icultur
e
,
pr
e
vious
r
e
s
e
a
r
c
h
ha
s
of
t
e
n
lac
ke
d
c
ompr
e
he
ns
ive,
r
e
a
l
-
ti
me
pa
thogen
de
tec
ti
on
s
ys
tems
that
c
a
n
be
e
f
f
e
c
ti
ve
ly
int
e
gr
a
ted
int
o
e
xis
ti
ng
a
gr
icultur
a
l
pr
a
c
ti
c
e
s
.
M
a
ny
s
tudi
e
s
ha
ve
f
oc
us
e
d
on
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
Ar
ti
f
I
ntell
I
S
S
N:
2252
-
8938
E
nhanc
ing
pr
e
c
is
ion
agr
icultur
e
:
a
c
ompr
e
he
ns
ive
inve
s
ti
gati
on
int
o
pa
thogen
de
tec
ti
on
…
(
Sha
is
ta
F
ar
hat)
3123
is
olate
d
a
s
pe
c
t
s
of
pa
thogen
de
tec
ti
on
without
a
ddr
e
s
s
ing
the
c
ompl
e
xit
y
of
pa
thogen
int
e
r
a
c
ti
ons
withi
n
diver
s
e
c
r
op
s
ys
tems
.
Additi
ona
ll
y
,
ther
e
is
a
s
c
a
r
c
it
y
of
s
c
a
lable
,
c
os
t
-
e
f
f
e
c
ti
ve
tec
hnologi
e
s
that
c
a
n
be
a
dopted
by
f
a
r
me
r
s
a
t
va
r
ious
s
c
a
les
of
ope
r
a
ti
o
n.
F
u
r
ther
mor
e
,
the
int
e
gr
a
ti
on
o
f
a
dva
nc
e
d
da
ta
a
na
lyt
ics
a
nd
mac
hine
lea
r
ning
(
M
L
)
in
pa
thogen
mana
ge
ment
r
e
mains
unde
r
e
xplor
e
d,
li
mi
t
ing
the
potential
f
or
pr
e
dictive
a
nd
a
da
pti
ve
s
tr
a
tegie
s
.
Nume
r
ous
r
e
s
e
a
r
c
he
r
s
a
r
e
wor
k
ing
on
de
tec
ti
ng
a
nd
pr
ognos
is
of
plant
dis
e
a
s
e
s
a
s
a
r
e
s
ult
o
f
the
s
igni
f
ica
nt
c
r
op
los
s
.
M
ode
r
n
c
omput
e
r
tec
hnology,
whe
n
us
e
d
in
c
ombi
na
ti
on
with
ongoing
f
a
r
me
r
obs
e
r
va
ti
on,
c
a
n
a
ls
o
be
us
e
f
ul
in
pr
e
dicting
plant
dis
e
a
s
e
s
e
a
r
ly
on
a
nd
pr
e
ve
nti
ng
c
r
op
inf
e
c
ti
on
[
3]
.
De
e
p
lea
r
ning
(
DL
)
is
a
mor
e
s
ophis
ti
c
a
ted
a
nd
pr
ove
n
tec
hnique
that
may
de
tec
t
the
dis
e
a
s
e
e
a
r
ly
o
n
in
the
a
f
f
e
c
ted
lea
f
.
C
onvolut
ion
ne
ur
a
l
ne
twor
ks
(
C
NN
s
)
a
nd
a
r
ti
f
ic
ial
ne
ur
a
l
ne
twor
ks
(
AN
Ns
)
a
r
e
two
f
r
e
qu
e
ntl
y
a
ppli
e
d
tec
hniques
that
c
a
n
ha
ndle
c
ompl
e
x
r
e
lati
ons
hips
dis
c
ove
r
e
d
in
da
ta.
L
a
r
ge
da
tas
e
ts
that
m
im
ic
the
a
r
c
hit
e
c
tur
e
,
s
e
que
nc
e
s
,
a
nd
ope
r
a
ti
on
of
the
br
a
i
n
of
humans
c
a
n
be
us
e
d
to
tr
a
in
a
model
to
be
ha
ve
li
ke
a
human.
C
NN
,
R
e
s
Ne
t50,
a
nd
other
DL
a
lgor
it
h
m
s
a
r
e
f
r
e
que
ntl
y
us
e
d
in
medic
a
l
im
a
ge
p
r
oc
e
s
s
in
g,
s
e
r
ies
f
or
e
c
a
s
ti
ng,
a
nomaly
de
tec
ti
on
,
dis
e
a
s
e
diagnos
is
,
a
nd
s
a
telli
te
im
a
ge
identif
ica
ti
on.
T
he
input
is
p
r
oc
e
s
s
e
d
by
pa
s
s
ing
thr
ough
numer
ous
leve
ls
whe
r
e
c
e
r
tain
f
e
a
tur
e
s
that
dis
play
c
onvolut
ion
ope
r
a
ti
ons
[
4]
.
P
r
e
c
is
ion
a
gr
icultur
e
,
a
n
a
dva
nc
e
d
f
a
r
mi
ng
p
r
a
c
ti
c
e
that
ut
il
ize
s
tec
hnology
to
moni
to
r
a
nd
mana
ge
c
r
op
he
a
lt
h,
ha
s
be
c
ome
incr
e
a
s
ingl
y
vit
a
l
f
o
r
ma
xim
izing
yields
a
nd
mi
nim
izing
los
s
e
s
.
One
c
r
it
ic
a
l
a
s
pe
c
t
of
pr
e
c
is
ion
a
gr
icultur
e
is
pa
thogen
de
tec
ti
on,
w
h
ich
invol
ve
s
identif
ying
ha
r
m
f
ul
o
r
ga
nis
ms
that
c
a
n
c
a
us
e
dis
e
a
s
e
in
c
r
ops
.
De
s
pit
e
s
igni
f
ica
nt
pr
ogr
e
s
s
,
tr
a
dit
ional
pa
thogen
de
tec
ti
on
methods
of
ten
lac
k
r
e
a
l
-
ti
me
c
a
pa
bil
it
ies
a
nd
f
a
il
to
int
e
gr
a
te
with
a
dva
nc
e
d
da
ta
a
na
lyt
ics
.
T
his
s
tudy
a
im
s
to
e
nha
n
c
e
pr
e
c
is
ion
a
gr
icultur
e
by
de
ve
lopi
ng
a
c
ompr
e
he
ns
ive
pa
tho
ge
n
de
tec
ti
on
a
nd
mana
ge
ment
s
ys
tem
that
leve
r
a
ge
s
r
e
a
l
-
ti
me
moni
tor
ing
a
nd
ML
tec
hnologi
e
s
.
T
he
br
e
a
dth
of
thi
s
wo
r
k
ha
s
r
e
s
ult
e
d
in
the
c
r
e
a
ti
on
of
nove
l
DL
tec
hniques
that
e
mpl
oy
a
ne
two
r
k
dur
i
ng
the
t
r
a
ini
ng
s
tage
whe
r
e
the
pixels
'
f
e
a
tur
e
s
a
r
e
s
ha
r
ply
f
oc
us
e
d
[
5]
,
tar
ge
ti
ng
the
de
e
p
be
li
e
f
ne
twor
k
(
DB
N)
[
6
]
.
B
y
e
mpl
oying
the
pr
une
s
'
f
ine
twe
a
ki
ng,
thi
s
method
tends
to
inc
r
e
a
s
e
the
model's
a
c
c
u
r
a
c
y
[
6]
.
A
number
of
r
e
s
e
a
r
c
he
r
s
a
r
e
s
ti
ll
wor
king
to
int
e
gr
a
te
the
va
r
ious
models
in
or
de
r
to
r
a
is
e
the
s
ys
tem's
ove
r
a
ll
pe
r
f
or
manc
e
metr
ics
.
C
ombi
ning
both
tec
hniques
e
na
ble
the
model
to
e
li
mi
na
te
ove
r
f
it
ti
ng
c
a
us
e
d
by
uns
e
e
n
pixels
in
the
im
a
ge
[
7
]
.
T
e
c
hnology
f
o
r
a
utom
a
ti
on
ha
s
r
e
volut
ioni
z
e
d
a
gr
icultu
r
e
p
r
oduc
ti
on
by
br
e
a
king
pa
s
t
ba
r
r
ier
s
to
tec
hnology
[
8]
.
S
igni
f
ica
nt
a
dva
n
c
e
ments
in
tec
hnology
ha
ve
be
e
n
a
c
hieve
d
to
e
nh
a
nc
e
the
im
a
ge
's
f
e
a
tur
e
e
xtr
a
c
ti
on,
whic
h
pr
im
a
r
il
y
c
onve
ys
the
im
a
ge
's
pr
ope
r
ti
e
s
.
He
r
e
,
DL
tec
hniques
h
a
ve
be
e
n
im
pleme
nted
[
9]
,
[
10]
in
o
r
de
r
to
t
r
a
in
i
mage
s
ba
s
e
d
on
their
pixel
pos
it
ion.
C
NN
laye
r
s
a
r
e
us
e
d
to
pe
r
f
or
m
the
e
nti
r
e
pr
oc
e
dur
e
c
omput
a
ti
ona
ll
y
[
11]
.
T
he
c
ombi
na
ti
on
of
De
ns
e
Ne
t
[
12]
a
nd
I
nc
e
pti
on
[
1
3]
in
the
tr
a
ns
f
e
r
lea
r
ning
models
g
r
a
dua
ll
y
a
tt
r
a
c
ted
the
r
e
s
e
a
r
c
he
r
's
int
e
r
e
s
t.
T
his
lea
r
ning
tool
make
s
it
po
s
s
ibl
e
to
s
olve
pr
oblems
pr
ope
r
ly.
T
his
im
pleme
ntation
wi
ll
r
e
s
ult
in
a
n
a
ddit
ional
a
ugmenta
ti
on
is
s
ue
,
gr
owing
the
s
ys
tem's
s
ize
[
14]
,
[
15]
.
T
his
w
or
k
pr
e
s
e
nts
the
view
s
of
s
c
holar
s
a
nd
s
ugge
s
t
s
pos
s
ibl
e
s
olut
ion
s
to
the
pr
oblems
s
e
t
out.
Ne
ve
r
thele
s
s
,
numer
ous
pr
oblems
s
ti
ll
ne
e
d
to
be
a
ddr
e
s
s
e
d,
including
li
mi
tati
ons
a
nd
potential
s
olut
ions
[
16]
,
[
17]
.
T
he
mot
ivation
f
or
thi
s
s
tudy
s
tems
f
r
o
m
the
n
e
e
d
f
or
mor
e
e
f
f
icie
nt,
s
c
a
lable
,
a
nd
int
e
gr
a
ted
pa
thogen
de
tec
ti
on
s
ys
tems
in
a
gr
icultu
r
e
.
T
r
a
dit
i
ona
l
methods
,
though
e
f
f
e
c
ti
ve
to
a
de
gr
e
e
,
o
f
ten
f
a
ll
s
hor
t
in
pr
ovidi
ng
ti
mely
a
nd
pr
e
c
is
e
da
ta,
lea
ding
to
s
ubopti
mal
c
r
op
mana
ge
ment.
Our
r
e
s
e
a
r
c
h
int
r
o
duc
e
s
a
nove
l
a
ppr
oa
c
h
that
no
t
only
de
tec
ts
pa
thogens
in
r
e
a
l
-
ti
me
but
a
ls
o
u
ti
li
z
e
s
a
dva
nc
e
d
da
ta
a
na
lyt
ics
to
pr
e
dict
a
nd
mana
ge
potential
outbr
e
a
ks
.
T
his
pa
pe
r
a
ddr
e
s
s
e
s
the
s
igni
f
ica
nt
ga
p
in
int
e
gr
a
ti
ng
ML
with
pa
thogen
de
tec
ti
on,
of
f
e
r
ing
ne
w
ins
ight
s
int
o
p
r
e
dictive
a
gr
icultur
e
.
T
o
guide
the
r
e
a
de
r
thr
ough
our
f
indi
ngs
,
thi
s
pa
pe
r
is
s
tr
uc
tur
e
d
a
s
f
oll
ows
:
f
ir
s
t,
we
r
e
view
the
e
xis
ti
ng
li
ter
a
tur
e
on
pa
thogen
de
tec
ti
on
methods
a
nd
their
li
mi
tations
.
Ne
xt
,
we
de
tail
the
method
ology
of
our
p
r
opos
e
d
de
tec
ti
on
s
ys
tem,
including
the
te
c
hnologi
e
s
a
nd
a
lgor
it
h
ms
e
mpl
oye
d.
F
ol
lowing
thi
s
,
we
pr
e
s
e
nt
the
r
e
s
ult
s
of
our
f
ield
tes
ts
,
highl
igh
ti
ng
th
e
e
f
f
e
c
ti
ve
ne
s
s
a
nd
e
f
f
icie
nc
y
of
our
a
ppr
oa
c
h.
F
in
a
ll
y,
we
dis
c
us
s
the
im
pli
c
a
ti
ons
of
ou
r
f
indi
ngs
,
a
ddr
e
s
s
ing
potential
li
mi
tations
a
nd
s
ugge
s
ti
ng
dir
e
c
ti
ons
f
or
f
utur
e
r
e
s
e
a
r
c
h.
B
y
the
e
nd
of
thi
s
pa
pe
r
,
r
e
a
de
r
s
will
ha
ve
a
c
ompr
e
he
ns
ive
unde
r
s
tanding
of
the
a
dva
nc
e
ments
in
pa
thogen
de
tec
ti
on
a
nd
their
im
pa
c
t
on
pr
e
c
is
ion
a
gr
icultur
e
.
2.
L
I
T
E
RA
T
UR
E
S
UR
VE
Y
T
his
s
e
c
ti
on
e
xa
mi
ne
s
the
li
ter
a
tur
e
on
va
r
ious
m
ode
ls
a
nd
tec
hniques
us
e
d
in
c
r
op
he
a
lt
h
a
na
lys
is
,
other
da
ta
a
na
lys
is
,
a
nd
a
gr
icultur
a
l
c
r
op
s
ur
ve
il
l
a
nc
e
.
T
oo
e
t
al
.
[
18]
c
a
r
r
ied
out
a
n
inves
ti
ga
ti
on
us
ing
a
s
iza
ble
da
tas
e
t
that
include
d
im
a
ge
s
o
f
s
e
ve
r
a
l
lea
ve
s
with
va
r
ying
textur
e
s
,
pe
r
s
pe
c
ti
ve
s
,
a
nd
we
a
ther
c
ondit
ions
.
T
he
a
uthor
us
e
d
two
dif
f
e
r
e
nt
a
ppr
o
a
c
he
s
:
s
tand
a
r
d
a
ugmenta
ti
on
a
nd
ge
ne
r
a
ti
ve
a
d
ve
r
s
a
r
ial
ne
twor
k
(
GAN
)
-
ba
s
e
d
da
ta.
T
he
Ale
xNe
t,
VG
G,
De
ns
e
Ne
t,
a
nd
R
e
s
Ne
t
we
r
e
the
main
s
ubjec
ts
of
a
tt
e
nti
on.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
Ar
ti
f
I
ntell
,
Vol.
14,
No.
4
,
Augus
t
2025
:
312
1
-
3132
3124
T
he
s
e
two
methods
c
ons
i
s
ted
of
GA
N
tr
a
ini
ng
a
nd
s
yntac
ti
c
da
ta.
T
his
we
ll
-
tr
a
ined
model
a
c
hieve
d
99.
75%
a
c
c
ur
a
c
y
in
c
ha
ll
e
nging
e
nvir
onments
.
I
n
the
f
utur
e
,
the
wor
k
s
c
ope
may
be
e
xpa
nde
d
to
include
a
n
a
ppli
c
a
ti
on
that
r
uns
on
both
M
a
c
O
S
a
nd
Andr
oid
,
making
it
s
im
ple
f
or
f
a
r
mer
s
to
identif
y
lea
f
dis
e
a
s
e
pr
e
matur
e
ly.
Abdu
e
t
al
.
[
19
]
p
r
e
s
e
nted
a
n
e
f
f
icie
nt
us
e
of
pa
thol
ogica
l
dis
e
a
s
e
s
ympt
om
s
e
gmenta
ti
on
a
nd
loca
li
z
a
ti
on.
An
a
utom
a
ti
c
method
f
or
the
de
tec
ti
o
n
of
plant
lea
f
dis
e
a
s
e
,
a
nd
it
’
s
int
e
nde
d
identif
yin
g
the
s
or
t
of
il
lnes
s
that
is
a
f
f
e
c
ti
ng
the
p
lant
a
s
we
ll
a
s
whe
ther
it
is
a
f
f
e
c
ti
ng
it
a
t
a
ll
.
T
he
y
ha
ve
e
mpl
oye
d
a
method
c
a
ll
e
d
r
a
dial
ba
s
is
f
unc
ti
on
ne
ur
a
l
ne
twor
k
(
R
B
F
NN
)
.
T
h
is
method
will
be
put
in
to
us
e
in
a
gr
icult
ur
a
l
c
r
op
f
ields
in
the
f
u
tur
e
.
I
t
will
make
it
e
a
s
ier
to
mo
nit
or
the
plants
a
nd
upda
te
the
s
tatus
by
identi
f
ying
the
dis
e
a
s
e
;
if
the
plant
is
he
a
lt
hy
,
it
will
be
upda
ted
to
s
how
a
n
e
f
f
e
c
ti
ve
lea
f
.
L
a
mba
e
t
al
.
[
20]
s
ugge
s
ted
a
nove
l
a
utom
a
ti
c
plant
dis
e
a
s
e
identif
ica
ti
on
de
tec
ti
o
n
e
mpl
oying
C
NN
a
nd
DL
ne
twor
ks
.
W
he
n
9,
914
tr
a
in
ing
pa
r
a
mete
r
s
we
r
e
take
n
int
o
a
c
c
ount,
the
ne
twor
k
r
e
a
c
he
d
a
n
a
c
c
ur
a
c
y
s
c
or
e
of
99.
2
%
.
I
n
the
f
utu
r
e
,
it
wil
l
be
c
a
pa
ble
o
f
ha
ndli
ng
a
lar
ge
r
r
a
nge
o
f
plant
lea
ve
s
.
L
i
e
t
a
l
.
[
21
]
c
onduc
ted
the
ini
t
ial
s
tudy
on
the
R
e
s
Ne
t
50
model
in
2021.
T
he
y
divi
de
d
the
C
NN
laye
r
c
omponent
s
ize
s
int
o
11
×
11
s
e
gments
f
or
a
n
a
lys
is
,
s
witching
the
a
c
ti
va
ti
on
tas
k
to
hype
r
s
pe
c
tr
a
l
im
a
ge
(
HSI
)
.
T
his
methodology's
goa
l
is
to
les
s
e
n
the
im
pa
c
t
of
HSI
inac
ti
va
ti
on
a
nd
,
to
s
ome
e
xtent,
e
nha
nc
e
or
ga
niza
ti
ona
l
e
xe
c
uti
on
by
e
nha
nc
ing
the
a
bil
i
ty
to
c
a
ptur
e
97.
56
%
of
the
highl
ight
s
point
b
y
point
a
c
c
ur
a
tely.
W
e
c
a
n
a
dd
mo
r
e
da
tas
e
ts
to
thi
s
mode
l
in
the
f
utur
e
.
Kha
tt
a
k
e
t
al
.
[
22]
c
r
e
a
ted
a
n
a
lgo
r
it
hm
in
2021
to
de
tec
t
dis
e
a
s
e
s
in
c
r
ops
s
uc
h
a
s
gr
a
pe
s
,
potatoe
s
,
tom
a
toes
,
a
nd
c
or
n.
T
he
C
NN
a
lgor
it
hm
wa
s
pr
i
mar
il
y
u
ti
li
z
e
d
f
or
the
il
lnes
s
c
las
s
if
ica
ti
on.
Us
ing
C
NN
,
a
97%
ove
r
a
ll
e
f
f
icie
nc
y
in
il
lnes
s
identif
ica
ti
on
wa
s
a
tt
a
ined.
T
he
pr
im
a
r
y
r
e
medy
f
or
the
f
a
r
mer
s
is
to
s
ugge
s
t
a
pe
s
ti
c
ide
f
or
the
a
f
f
e
c
ted
lea
f
onc
e
the
a
il
ment
h
a
s
be
e
n
identif
ied.
T
his
s
tudy
c
a
n
be
e
xtende
d
to
t
a
ke
int
o
a
c
c
ount
a
ll
e
nvir
onmenta
l
f
a
c
to
r
s
,
s
uc
h
a
s
humi
di
ty,
pH
,
r
a
inf
a
ll
,
a
nd
N,
P
,
a
nd
K
va
lues
,
in
o
r
de
r
to
r
a
is
e
pr
oduc
ti
vit
y
in
li
ne
with
f
a
r
me
r
e
xpe
c
tations
.
Z
hou,
e
t
al
.
[
23
]
c
onc
e
ntr
a
ted
on
de
ve
lopi
ng
a
mo
bil
e
a
ppli
c
a
ti
on
that
e
mpl
oye
d
DL
to
ident
if
y
a
nd
c
a
tegor
ize
gr
a
pe
lea
f
dis
e
a
s
e
.
T
he
f
a
s
ter
r
e
gion
-
ba
s
e
d
c
onvolut
ional
ne
ur
a
l
ne
twor
k
(
R
-
C
NN
)
,
us
ing
I
nc
e
pti
on
-
V2
s
pot
de
tec
ti
on,
is
e
mpl
oye
d
by
thi
s
a
ppli
c
a
ti
on
to
loca
te
a
n
inf
e
c
ted
a
r
e
a
in
the
im
a
ge
a
nd
c
onc
e
ntr
a
te
the
da
tas
e
t
f
or
that
a
r
e
a
.
T
he
indepe
nde
nt
s
mar
tphone
a
ppli
c
a
ti
on
is
de
s
igned
a
nd
ope
r
a
t
e
d
us
ing
thi
s
pr
opos
e
d
model.
T
his
s
tudy's
e
xc
e
ll
e
nt
a
c
c
ur
a
c
y
of
97.
9%
in
r
e
c
ognizing
the
c
omm
on
f
or
ms
of
gr
a
pe
lea
f
dis
e
a
s
e
is
b
a
s
e
d
on
da
ta
f
r
om
gr
a
pe
lea
f
da
tas
e
t
e
xpe
r
im
e
nts
.
T
he
pr
ogr
a
m
c
ould
be
e
xpa
nde
d
in
the
f
utur
e
to
be
a
ble
to
identif
y
many
kinds
o
f
c
r
op
dis
e
a
s
e
s
,
not
jus
t
gr
a
pe
dis
e
a
s
e
s
.
W
a
ng
e
t
al
.
[
24]
p
r
opos
e
d
a
s
ys
tem
in
2022
that
c
a
n
a
utom
a
ti
c
a
ll
y
identi
f
y
c
hil
i
d
is
e
a
s
e
de
tec
ti
on.
T
he
f
ive
c
las
s
e
s
us
e
d
to
ba
s
e
on
the
e
f
f
e
c
ti
ve
DL
f
r
a
mew
or
k
,
thi
s
model
modi
f
ies
the
e
ntr
opy
o
f
the
los
s
f
unc
ti
on
to
s
olve
is
s
ue
s
that
c
a
n
le
a
d
to
a
n
im
ba
la
nc
e
in
the
da
tas
e
t.
F
ur
ther
mor
e
,
the
model
pe
r
f
or
ms
e
xc
e
s
s
laye
r
s
of
t
r
a
ns
it
ion
with
a
n
a
c
c
ur
a
c
y
of
92%
.
W
a
ng
e
t
al
.
[
24]
outl
ined
the
dir
e
c
ti
on
of
th
e
f
utur
e
s
o
a
s
to
e
na
ble
the
im
pleme
ntation
of
s
pe
c
ialize
d
a
r
c
hit
e
c
tur
a
l
modi
f
ica
ti
ons
f
or
mul
ti
ple
a
ddit
ional
lea
ve
s
.
E
l
f
a
ti
mi
e
t
al
.
[
25]
de
s
c
r
ibed
a
DL
tec
hnique
that
a
im
s
to
c
las
s
if
y
oli
ve
lea
ve
s
us
ing
thr
e
e
DL
models
that
ha
ve
be
e
n
a
da
pted
to
the
ge
ne
ti
c
a
lgo
r
it
hm
(
GA
)
ve
r
s
ion.
F
indi
ng
the
opti
mal
ba
tch
s
ize
a
nd
the
number
o
f
e
poc
hs
to
maximi
z
e
the
a
c
c
ur
a
c
y
s
c
or
e
a
nd
mi
nim
ize
r
e
s
pons
e
ti
me
wa
s
the
pr
im
a
r
y
go
a
l
of
the
a
uthor
's
method.
W
it
h
a
n
a
c
c
ur
a
c
y
s
c
or
e
of
98%
f
or
bina
r
y
c
las
s
if
ica
ti
on,
the
De
ns
e
Ne
t
model
is
t
he
mos
t
a
c
c
ur
a
te.
Ga
ther
ing
dif
f
e
r
e
nt
im
a
ge
s
of
oli
ve
il
lnes
s
a
nd
tr
a
ini
ng
s
a
mpl
e
s
on
a
lar
ge
r
da
taba
s
e
to
a
tt
a
i
n
higher
a
c
c
ur
a
c
y
s
c
or
e
s
is
a
job
that
c
a
n
be
done
in
the
f
ut
ur
e
.
Util
izing
s
tate
-
of
-
the
-
a
r
t
DL
tec
hniques
,
pa
r
ti
c
ul
a
r
ly
E
f
f
icie
ntNe
tV2
[
26]
a
r
c
hit
e
c
tur
e
,
the
a
im
to
r
e
volut
ioni
z
e
plant
dis
e
a
s
e
de
tec
ti
on
by
de
ve
lo
ping
a
highl
y
a
c
c
ur
a
te
a
nd
e
f
f
icie
nt
s
ys
tem
c
a
pa
ble
of
s
wif
tl
y
identif
ying
a
nd
c
la
s
s
if
ying
va
r
ious
plant
di
s
e
a
s
e
s
ba
s
e
d
on
lea
f
im
a
ge
s
,
ther
e
by
e
mpowe
r
ing
f
a
r
mer
s
a
nd
a
gr
icultur
is
ts
with
ti
mely
int
e
r
ve
nti
ons
to
mi
ti
ga
te
c
r
op
los
s
e
s
a
nd
e
ns
ur
e
f
ood
s
e
c
ur
it
y
[
27]
,
[
28]
.
T
he
major
it
y
of
c
ur
r
e
nt
s
tudi
e
s
on
tec
hnology
-
dr
iven
a
gr
icultur
e
of
f
e
r
ins
ight
f
ul
inf
o
r
mation.
T
his
s
e
c
ti
on
pr
ovides
a
s
umm
a
r
y
of
the
c
ur
r
e
nt
r
e
s
e
a
r
c
h
outcome
s
in
the
f
ield.
As
int
e
r
ne
t
o
f
thi
ngs
(
I
oT
)
a
nd
a
r
ti
f
icia
l
i
ntelli
ge
nc
e
(
A
I
)
tec
hnologi
e
s
gr
ow
in
popular
it
y,
mor
e
e
f
f
o
r
ts
a
r
e
be
ing
made
to
us
e
t
he
m
f
or
pr
e
c
is
ion
a
gr
icultur
e
.
An
AI
-
e
na
bled
method
f
o
r
identif
ying
dis
e
a
s
e
s
,
Aqe
l
e
t
al
.
[
29]
e
xplor
e
d
the
methods
f
or
s
mar
t
a
gr
icultur
e
.
T
he
i
r
AI
-
ba
s
e
d
method
include
s
a
ut
omatic
de
tec
ti
on
a
nd
c
las
s
if
ica
ti
on
of
plant
lea
f
dis
e
a
s
e
s
ba
s
e
d
on
us
ing
the
e
xtr
e
me
lea
r
ning
mac
hine
(
E
L
M
)
DL
al
gor
it
hm
on
a
r
e
a
l
da
tas
e
t
of
plant
lea
f
im
a
ge
s
.
F
or
the
c
las
s
if
ica
ti
on
of
dis
e
a
s
e
s
,
their
a
ppr
oa
c
h
a
ls
o
make
s
us
e
of
a
bi
-
dir
e
c
ti
ona
l
f
or
m
of
g
r
a
y
leve
l
co
-
oc
c
ur
r
e
nc
e
matr
ix
(
GL
C
M
)
.
One
pa
r
ti
c
ular
dr
a
wba
c
k
of
the
a
ppr
oa
c
h
in
[
29
]
wa
s
that
it
ne
e
ds
a
f
e
a
tu
r
e
f
or
ga
ther
ing
r
e
a
l
-
ti
me
da
ta
f
r
om
c
r
ops
.
T
o
thi
s
a
im
,
their
methodology
r
e
quir
e
s
I
oT
c
onne
c
ti
vit
y
s
o
that
im
a
ge
s
e
ns
or
s
c
a
n
c
onti
nuous
ly
moni
tor
c
r
ops
by
c
a
ptur
ing
c
r
op
de
tails
in
r
e
a
l
-
ti
me.
Nikit
h
e
t
al.
[
30
]
p
r
e
s
e
nted
a
c
onc
e
pt
f
o
r
a
n
AI
-
a
n
d
I
o
T
-
ba
s
e
d
s
mar
t
f
a
r
mi
ng
s
ys
tem.
I
t
wa
s
c
r
e
a
ted
to
us
e
int
e
ll
igent
hyd
r
oponic
f
a
r
mi
ng
f
o
r
a
us
e
r
-
f
r
iendly
method
of
c
r
op
obs
e
r
va
ti
on
.
T
he
y
a
ls
o
wor
ke
d
on
a
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
Ar
ti
f
I
ntell
I
S
S
N:
2252
-
8938
E
nhanc
ing
pr
e
c
is
ion
agr
icultur
e
:
a
c
ompr
e
he
ns
ive
inve
s
ti
gati
on
int
o
pa
thogen
de
tec
ti
on
…
(
Sha
is
ta
F
ar
hat)
3125
s
mar
tphone
a
ppli
c
a
ti
on
that
make
s
c
r
op
moni
tor
i
ng
e
a
s
ier
.
T
he
i
r
R
a
s
pbe
r
r
y
P
i
C
P
U
wa
s
in
c
ha
r
ge
of
thei
r
s
e
ns
or
de
vice
s
.
F
or
dis
e
a
s
e
pr
e
diction,
a
de
e
p
C
NN
model
wa
s
e
mpl
oye
d
in
a
ddit
ion
to
the
ha
r
dwa
r
e
e
leme
nts
.
F
a
r
mer
s
us
e
d
the
s
mar
tphone
a
ppli
c
a
ti
o
n
to
moni
to
r
c
r
op
r
e
qui
r
e
ments
e
a
s
il
y.
Ne
ve
r
thele
s
s
,
the
a
ppr
oa
c
h
in
[
3
0]
ha
s
s
e
ve
r
a
l
is
s
ue
s
.
F
ir
s
t,
it
ne
e
ds
opti
m
iza
ti
on
tec
hniques
.
S
e
c
ond,
the
tec
hnique
r
e
li
e
s
on
de
e
p
C
NN
,
whic
h
c
a
n
be
e
nha
nc
e
d
f
ur
ther
by
c
ombi
ning
it
with
ot
he
r
de
e
p
models
in
a
hybr
id
a
ppr
oa
c
h
to
im
pr
ove
the
dyna
m
ics
of
c
r
op
obs
e
r
va
ti
on
.
I
n
pr
e
c
is
ion
a
g
r
icultu
r
e
,
DL
-
ba
s
e
d
tec
hniques
ha
ve
s
hown
to
be
mor
e
s
uc
c
e
s
s
f
ul
than
their
pr
e
de
c
e
s
s
or
s
.
As
a
r
e
s
ult
,
in
2022,
Na
r
madha
e
t
al
.
[
31]
inves
ti
ga
ted
s
e
ve
r
a
l
de
e
p
-
lea
r
ning
a
lgor
it
hms
in
a
n
e
f
f
or
t
to
a
dva
nc
e
pr
e
c
is
ion
a
gr
icultur
e
.
I
n
or
de
r
to
s
tudy
pr
e
c
is
e
ne
s
s
a
gr
icultur
e
,
their
r
e
s
e
a
r
c
h
c
onc
e
ntr
a
ted
late
s
t
de
ve
lopm
e
nts
in
c
omm
unica
ti
on
tec
hnologi
e
s
.
T
he
i
r
r
e
s
e
a
r
c
h
r
e
ve
a
led
a
g
r
e
a
t
de
a
l
o
f
r
oom
f
or
f
utur
e
de
ve
lopm
e
nt
in
a
ddit
ion
to
thes
e
r
e
ve
lations
.
I
n
a
ddit
ion
to
c
ons
ider
ing
e
c
ologi
c
a
l
c
oll
a
ps
e
a
nd
c
li
mate
c
ha
nge
pa
r
a
digm
s
,
ther
e
is
a
ne
e
d
f
or
the
c
r
e
a
ti
o
n
of
pr
e
diction
models
that
int
e
g
r
a
te
vis
ua
l
tr
a
ns
f
or
mation
a
nd
s
ophis
ti
c
a
ted
C
NN
va
r
iations
that
may
pe
r
f
o
r
m
be
tt
e
r
f
or
p
ictur
e
pa
tch
s
e
que
nc
e
s
.
I
n
pr
e
c
is
ion
a
gr
icultur
e
,
s
mar
t
g
r
e
e
nhous
e
s
will
a
ls
o
be
e
s
s
e
nti
a
l.
A
s
tudy
on
tr
a
ns
f
e
r
lea
r
ning
in
2022
that
looked
a
t
the
wa
ter
-
f
ood
-
e
n
e
r
gy
ne
xus
wa
s
c
onduc
ted
by
S
ha
r
ma
e
t
al
.
[
32]
.
F
or
tec
hnolo
gy
-
dr
iven
a
gr
icultur
e
de
c
is
ion
-
making,
poli
c
y
make
r
s
ne
e
de
d
the
input
s
f
r
om
their
s
tudy.
T
he
wa
ter
,
f
ood
,
a
n
d
e
ne
r
gy
ne
xus
a
r
e
im
pr
ove
d
f
or
s
us
taina
ble
de
ve
lopm
e
nt
thr
ough
the
us
e
of
AI
,
c
omm
unica
ti
on
inf
r
a
s
tr
uc
tur
e
,
a
nd
moni
tor
ing
a
ppr
oa
c
he
s
.
F
ur
ther
mo
r
e
,
their
r
e
s
e
a
r
c
h
highl
ight
s
the
ne
c
e
s
s
it
y
f
or
f
utur
e
pr
e
c
is
ion
a
gr
ic
ult
ur
e
to
e
mpl
oy
mor
e
e
f
f
e
c
ti
ve
DL
tec
hniques
.
T
he
21
st
-
c
e
ntur
y
AI
-
e
na
bled
tec
hnology
known
a
s
a
r
ti
f
icia
l
in
ter
ne
t
of
thi
ngs
(
AI
oT
)
e
nc
our
a
ge
s
us
e
s
of
AI
a
nd
a
s
s
oc
iat
e
d
de
vice
s
in
I
oT
to
c
r
e
a
te
a
be
tt
e
r
be
ne
f
icia
l
plat
f
or
m
f
o
r
a
ns
we
r
ing
c
ha
ll
e
nge
s
in
the
r
e
a
l
wor
ld
.
S
a
hu
a
nd
P
a
nde
y
[
33]
c
onduc
ted
r
e
s
e
a
r
c
h
on
plant
dis
e
a
s
e
de
te
c
ti
on
a
nd
diagnos
ing
mea
s
ur
e
s
be
c
ome
a
majo
r
c
onc
e
r
n
in
a
gr
icultur
e
f
il
e
d
.
T
he
y
pr
opos
e
d
he
modynamic
r
e
s
pons
e
f
unc
ti
on
(
HR
F
)
-
mul
ti
-
c
las
s
s
uppor
t
ve
c
tor
mac
hine
(
S
VM
)
a
c
c
ur
a
tely
c
las
s
if
ies
the
dis
e
a
s
e
s
a
nd
r
a
pidl
y
im
p
r
ove
s
the
qua
li
ty.
P
us
hpa
e
t
al.
[
34]
c
onduc
ted
r
e
s
e
a
r
c
h
on
big
da
ta
,
a
gr
icultur
e
,
a
nd
the
a
ppli
c
a
ti
on
of
A
I
in
thi
s
f
iel
d.
T
he
y
s
ugge
s
ted
a
n
e
c
o
s
ys
tem
f
or
s
mar
t
a
gr
icultur
e
that
us
e
s
c
loud
I
oT
DL
model
platf
o
r
ms
,
blo
c
kc
ha
in
tec
hnology,
I
oT
-
ba
s
e
d
da
ta
ga
ther
ing
a
nd
c
omm
unica
ti
on,
AI
f
o
r
big
da
ta
a
na
lyt
ics
,
a
nd
da
ta
vis
ua
li
z
a
ti
on.
Ne
ve
r
thele
s
s
,
they
dis
c
ove
r
e
d
that
DL
im
pr
ove
me
nts
a
r
e
r
e
quir
e
d
to
r
e
a
li
z
e
s
uc
h
a
n
e
c
os
ys
tem.
R
e
s
e
a
r
c
h
on
c
r
op
moni
tor
ing
ha
s
pr
ove
n
u
ti
li
ty
in
f
uz
z
y
-
ba
s
e
d
im
pr
ove
ments
.
I
n
2
02
1,
V
e
ni
e
t
al
.
[
3
5]
w
or
ke
d
o
n
t
h
e
i
de
nti
f
ic
a
ti
on
a
n
d
c
l
a
s
s
if
ic
a
t
io
n
o
f
p
la
nt
di
s
e
a
s
e
s
by
th
e
u
s
e
of
f
u
z
z
y
-
b
a
s
e
d
o
pti
mi
z
a
ti
on
in
DL
.
T
h
e
ir
a
pp
r
o
a
c
h
c
o
mb
in
e
d
DL
w
it
h
I
oT
.
A
dd
it
i
on
a
ll
y,
it
u
s
e
d
t
he
f
ir
e
f
l
y
a
l
go
r
it
hm,
wh
ic
h
i
s
b
io
in
s
pir
e
d,
to
i
n
c
r
e
a
s
e
n
e
t
wor
k
e
f
f
i
c
i
e
n
c
y.
A
ddi
ti
on
a
l
ly,
it
w
a
s
m
or
e
a
c
c
ur
a
t
e
a
n
d
e
c
o
nom
i
c
a
l
S
V
M
a
nd
k
-
n
e
a
r
e
s
t
n
e
ig
hb
or
s
(
k
NN
)
du
e
to
t
h
e
ir
f
u
z
z
y
l
ogi
c
i
nf
e
r
e
nc
e
.
T
h
e
y
w
a
nt
e
d
t
o
u
s
e
m
or
e
te
c
hn
olo
gi
e
s
i
n
th
e
f
u
tur
e
,
s
u
c
h
a
s
s
e
ns
o
r
n
e
t
wor
k
s
,
c
l
ou
d
c
om
pu
ti
n
g,
b
ig
d
a
t
a
,
a
nd
unm
a
nn
e
d
a
e
r
i
a
l
v
e
hi
c
l
e
s
(
UA
V
s
)
,
t
o
a
dv
a
nc
e
c
r
o
p
m
on
it
or
in
g
t
e
c
hn
ol
og
y
f
ur
th
e
r
.
P
r
e
c
i
s
io
n
f
a
r
m
in
g
m
a
de
u
s
e
of
n
a
no
te
c
h
no
lo
gy
a
nd
AI
.
I
n
s
tudi
e
s
[
36
]
,
[
7
]
us
e
d
DL
a
nd
na
notec
hnolo
gy
to
a
c
hieve
a
c
c
ur
a
tene
s
s
of
a
gr
icultu
r
e
.
T
he
y
obs
e
r
ve
d
DL
a
nd
AI
c
omi
ng
togethe
r
to
e
na
ble
f
a
r
mer
s
to
e
mpl
oy
tec
hnology
to
r
e
a
c
t
ins
tantly
to
c
r
o
p
ne
e
ds
.
T
he
y
opined
that
mor
e
r
e
s
e
a
r
c
h
wa
s
ne
c
e
s
s
a
r
y
to
de
ter
mi
ne
how
AI
a
nd
na
notec
hnology
m
ight
be
us
e
d
in
a
gr
icultur
e
.
Al
l
f
o
r
ms
of
f
a
r
mi
ng
a
nd
c
r
opp
ing
a
r
e
include
d
in
pr
e
c
is
ion
a
gr
icultu
r
e
.
S
a
r
a
nya
e
t
al
.
[
37
]
a
ppli
e
d
DL
tec
hniques
to
tom
a
to
plant
dis
e
a
s
e
s
a
nd
pr
opos
e
d
a
n
a
ppr
oa
c
h
f
or
opti
mi
z
ing
pr
e
-
tr
a
ined
models
to
maximi
z
e
de
tec
ti
on
pe
r
f
or
manc
e
.
T
o
incr
e
a
s
e
de
tec
ti
on
opti
mi
z
a
ti
on
with
a
r
e
leva
nc
e
-
ba
s
e
d
tec
hnique,
they
mer
ge
d
his
togr
a
m
-
ba
s
e
d
phe
nomen
a
a
nd
DL
c
ha
r
a
c
ter
is
ti
c
s
.
How
e
v
e
r
,
their
tec
hnique
doe
s
not
s
uppor
t
r
oboti
c
s
ha
ping
a
nd
m
ult
i
-
c
las
s
f
r
uit
c
a
tegor
iza
ti
on.
P
r
e
c
is
ion
f
a
r
mi
ng
is
a
nother
a
ppli
c
a
ti
on
of
da
ta
-
d
r
iven
AI
.
T
od
a
a
n
d
O
kur
a
[
38]
pr
ov
id
e
d
a
c
o
n
c
e
pt
to
e
v
a
l
ua
t
e
d
a
t
a
-
dr
i
ve
n
AI
-
b
a
s
e
d
a
p
pr
oa
c
h
e
s
.
I
t
s
a
p
pr
o
a
c
h
e
n
c
omp
a
s
s
e
d
r
ob
ot
ic
s
,
da
ta
a
na
ly
ti
c
s
,
v
i
s
u
a
l
c
o
mp
ut
in
g,
M
L
a
nd
D
L
mo
d
e
l
s
,
a
n
d
c
r
o
p
o
b
s
e
r
v
a
t
io
n,
m
a
na
ge
m
e
nt,
a
nd
h
a
r
v
e
s
t
in
g.
T
h
e
y
a
i
me
d
f
or
AI
a
l
gor
it
hm
s
t
h
a
t
wi
ll
be
u
s
e
d
in
th
e
f
ut
ur
e
f
or
i
nt
e
l
li
ge
nt
f
a
r
mi
ng.
W
e
f
ou
nd
t
ha
t
r
e
a
l
-
ti
m
e
pa
tho
g
e
n
d
e
t
e
c
ti
on
c
or
r
e
l
a
t
e
s
wi
th
im
pr
o
v
e
d
c
r
op
h
e
a
lt
h
a
n
d
y
i
e
ld
de
t
e
c
ti
o
n
a
s
pe
r
th
e
e
xp
la
n
a
ti
on
i
n
T
a
bl
e
1
.
T
he
pr
o
po
s
e
d
me
th
od
i
n
t
hi
s
s
tu
dy
te
nd
e
d
to
h
a
v
e
a
n
i
no
r
di
n
a
t
e
l
y
h
ig
he
r
pr
op
or
t
io
n
of
a
c
c
ur
a
t
e
de
te
c
t
io
n
s
a
s
c
om
p
a
r
e
d
t
o
tr
a
dit
io
n
a
l
m
e
th
od
s
.
Ac
c
or
d
in
g
t
o
t
h
e
s
t
ud
y
f
r
om
th
e
s
ur
v
e
y
m
e
n
ti
on
e
d
in
th
e
T
a
b
l
e
2,
a
n
d
ot
he
r
a
ut
hor
s
'
p
e
r
s
pe
c
ti
v
e
s
,
th
e
m
o
s
t
r
e
c
e
n
t
tr
e
n
d
s
a
n
d
p
r
o
c
e
d
ur
e
s
i
n
a
gr
i
c
ul
tur
e
c
a
n
b
e
c
om
bin
e
d
w
it
h
th
o
s
e
f
o
un
d
in
e
du
c
a
t
io
na
l
r
e
s
o
ur
c
e
s
t
o
in
c
r
e
a
s
e
p
r
o
du
c
t
iv
it
y
a
nd
be
ne
f
it
s
.
T
he
thr
e
e
main
tec
hniques
that
e
a
c
h
a
uthor
f
oc
us
e
d
on
a
r
e
knowle
dge
-
ba
s
e
d,
tec
hnology
-
dr
iven,
a
nd
lea
r
ning
-
ba
s
e
d.
T
he
f
ir
s
t
of
thes
e
thr
e
e
a
ppr
oa
c
he
s
,
the
knowle
dge
-
ba
s
e
d
a
ppr
oa
c
h,
is
not
c
omm
only
us
e
d
in
r
e
s
e
a
r
c
h.
Along
wi
th
DL
,
ML
a
ls
o
us
e
s
other
a
pp
r
oa
c
he
s
.
Apa
r
t
f
r
om
thes
e
a
ppr
oa
c
he
s
the
c
r
op
he
a
lt
h
a
nd
yield
moni
tor
ing
dis
c
us
s
e
d
dif
f
e
r
e
nt
a
s
pe
c
t
a
s
pr
e
s
e
nted
in
the
T
a
ble
3.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
Ar
ti
f
I
ntell
,
Vol.
14,
No.
4
,
Augus
t
2025
:
312
1
-
3132
3126
T
a
ble
1.
An
ove
r
view
of
the
s
tudy
pa
pe
r
s
'
pe
r
f
o
r
m
a
nc
e
,
methodology,
a
nd
c
onc
lus
ions
R
e
f
.
Y
e
a
r
M
e
th
odol
ogy
A
c
c
ur
a
c
y
(%)
O
bj
e
c
ti
ve
s
[
1]
201
1
k
-
me
a
ns
,
ne
ur
a
l
ne
twor
k
93.35
D
e
te
c
ti
on of
l
e
a
f
di
s
e
a
s
e
[
2]
2015
k
-
me
a
ns
,
ne
ur
a
l
ne
twor
k
82.9
T
he
y ha
ve
w
or
ke
d on f
iv
e
di
f
f
e
r
e
nt
di
s
e
a
s
e
s
of
pl
a
nt
s
[
3]
2019
D
N
N
w
it
h e
nc
ode
r
ne
twor
k
73.5
C
a
te
gor
iz
a
ti
on a
nd f
or
e
c
a
s
ti
ng of
t
r
a
ns
ie
nt
a
gr
ic
ul
tu
r
a
l
il
ln
e
s
s
e
s
[
18]
2019
D
e
ns
e
N
e
t
99.75
D
e
te
c
ti
on of
e
f
f
e
c
te
d pl
a
nt
[
19]
2020
L
in
e
a
r
bi
na
r
y pa
tt
e
r
n (
L
B
P
)
95.89
P
a
th
oge
n’
s
he
a
lt
h c
ondi
ti
on de
te
c
ti
on
[
20]
2021
DL
99.2
P
a
th
oge
n’
s
he
a
lt
h c
ondi
ti
on de
te
c
ti
on
[
21]
2021
D
L
f
r
a
me
w
or
k
a
nd H
S
I
97.56
I
de
nt
if
yi
ng a
nd c
la
s
s
if
ic
a
ti
on of
pa
th
oge
ns
[
22]
2021
C
N
N
, R
F
95.65
C
it
r
us
f
r
ui
ts
a
nd l
e
a
f
di
s
e
a
s
e
de
te
c
ti
on
[
23]
2021
R
e
s
tr
uc
tu
r
e
d r
e
s
id
ua
l
de
n
s
e
ne
twor
k (
R
R
D
N
)
95
T
he
a
ut
hor
ha
s
de
ve
lo
pe
d a
s
e
t
of
mode
ls
f
or
i
de
nt
if
yi
ng
to
ma
to
l
e
a
f
di
s
e
a
s
e
s
w
it
h hi
gh a
c
c
ur
a
c
y
[
25]
2022
M
obi
le
N
e
t
92.97, 98.50
B
e
a
n r
us
t
di
s
e
a
s
e
[
26]
2022
E
f
f
ic
ie
nt
N
e
tV2
-
L
98.28
D
e
te
c
ti
on of
c
a
r
da
mom
l
e
a
f
di
s
e
a
s
e
[
27]
2023
M
obi
le
N
e
t
C
N
N
97.89
C
la
s
s
if
yi
ng pla
nt
i
ll
ne
s
s
[
28]
2023
D
e
e
pP
la
nt
N
e
t
99.8
M
a
ngo pe
s
t
de
t
e
c
ti
on
[
39]
2023
C
N
N
97.9
T
he
r
e
s
e
a
r
c
h w
a
s
c
onduc
te
d on
ma
iz
e
di
s
e
a
s
e
de
te
c
ti
on w
it
h a
c
ompl
e
x da
ta
s
e
t.
[
40]
2023
R
e
s
id
ua
l
s
ki
p n
e
twor
k‑
ba
s
e
d
s
upe
r
‑
r
e
s
ol
ut
io
n f
or
l
e
a
f
di
s
e
a
s
e
de
te
c
ti
on (
R
S
N
S
R
-
L
D
D
)
97.2
A
ut
hor
s
de
te
c
te
d t
he
c
r
op dis
e
a
s
e
s
[
41]
2023
DL
97.36
F
in
d out t
he
gr
a
pe
c
r
op dis
e
a
s
e
s
[
42]
2023
C
omput
e
r
vi
s
io
n
99.6
A
ut
hor
s
de
te
c
te
d t
he
to
ma
to
le
a
f
di
s
e
a
s
e
s
[
43]
2023
D
e
e
pe
r
l
ig
ht
w
e
ig
ht
mul
ti
-
c
la
s
s
96.73
C
la
s
s
if
ic
a
ti
on a
nd i
de
nt
i
f
ic
a
ti
on of
pl
a
nt
di
s
e
a
s
e
s
[
44]
2022
C
N
N
-
A
ut
hor
s
de
te
c
te
d t
he
c
uc
umb
e
r
l
e
a
f
di
s
e
a
s
e
s
.
[
45]
2022
B
a
s
e
li
ne
M
L
, c
lu
s
te
r
in
g
-
D
e
te
c
ti
on of
ma
ngo
ba
c
te
r
ia
l
pa
r
t
[
46]
2022
C
omput
e
r
vi
s
io
n
97.2
P
a
th
oge
n’
s
de
te
c
ti
on i
n pota
to
es
[
47]
2022
C
omput
e
r
vi
s
io
n
a
nd M
L
97.36
P
a
th
oge
n’
s
de
te
c
ti
on
[
48]
2022
I
mpr
ove
d C
N
N
-
C
r
op
di
s
e
a
s
e
s
w
it
h
pe
s
t
pr
e
di
c
ti
on a
nd
c
la
s
s
if
ic
a
ti
on
[
49]
2021
S
V
M
-
P
la
nt
l
e
a
f
i
nf
e
c
ti
on
[
50]
2020
I
ma
ge
pr
oc
e
s
s
in
g a
nd
DL
-
P
la
nt
di
s
e
a
s
e
d
e
te
c
ti
on
[
51]
2020
ML
89.9
P
a
th
oge
n c
la
s
s
if
ic
a
ti
on
[
52]
2020
S
V
M
93
C
la
s
s
if
ic
a
ti
on
s
y
s
te
m f
or
gr
a
pe
l
e
a
ve
s
T
a
ble
2.
Va
r
ious
a
s
pe
c
ts
of
lea
f
d
is
e
a
s
e
de
tec
ti
on
ove
r
view
M
e
th
ods
M
ode
ls
us
e
d
D
r
a
w
ba
c
ks
L
e
a
r
ni
ng
-
ba
s
e
d
T
he
c
lu
s
te
r
in
g
a
lg
or
it
hm
e
mpl
oye
d
w
he
r
e
th
e
c
la
s
s
if
ic
a
ti
on a
t
th
e
pi
xe
l
l
e
ve
l
is
us
e
d
F
ut
ur
e
a
ppl
ic
a
ti
ons
f
or
t
hi
s
s
tr
a
te
gy c
oul
d i
nc
lu
de
t
he
us
e
of
L
S
B
-
ba
s
e
d pi
xe
ls
L
e
a
r
ni
ng
-
ba
s
e
d
W
he
n
c
la
s
s
if
yi
ng
pl
a
nt
le
a
f
di
s
e
a
s
e
s
,
M
L
te
c
hni
que
s
a
r
e
e
mpl
oye
d
T
hi
s
a
ppr
oa
c
h
ha
s
th
e
pot
e
nt
ia
l
to
be
e
xp
a
nde
d
in
th
e
f
ut
ur
e
by
im
pl
e
me
nt
in
g
a
ne
ur
a
l
ne
twor
k
w
hi
c
h
c
a
n
de
te
c
t
le
a
f
di
s
e
a
s
e
s
w
it
h gr
e
a
te
r
a
c
c
ur
a
c
y
T
e
c
hnol
ogy
-
dr
iv
e
n a
ppr
oa
c
h
D
L
is
a
ppl
ie
d
in
le
a
f
di
s
e
a
s
e
de
te
c
ti
on
w
he
r
e
t
he
e
nc
ode
r
ne
twor
k
O
th
e
r
me
th
ods
ma
y
be
us
e
d
f
or
de
te
c
ti
on
in
f
ut
ur
e
s
c
ope
s
,
c
ont
in
ge
nt
on t
he
s
e
a
s
ona
l
c
r
op
T
e
c
hnol
ogy
-
dr
iv
e
n a
ppr
oa
c
h
I
de
a
l
bot
h
a
da
pt
iv
e
G
A
a
nd
D
L
te
c
hni
qu
e
s
a
r
e
e
mpl
oye
d
S
ma
r
t
de
te
c
ti
on
ha
s
th
e
pot
e
nt
ia
l
to
be
a
ppl
ie
d
to
r
e
s
id
ua
l
ne
twor
k
-
ba
s
e
d
de
te
c
ti
on
a
nd
c
la
s
s
if
ic
a
ti
on
me
th
ods
i
n t
he
f
ut
ur
e
K
now
le
dge
-
ba
s
e
d a
ppr
oa
c
h
T
r
a
ns
f
e
r
l
e
a
r
ni
ng w
a
s
a
ppl
ie
d
T
he
f
ut
ur
e
w
or
k c
a
n be
done
u
s
in
g C
V
a
nd A
I
T
a
ble
3.
S
umm
a
r
y
of
ke
y
f
indi
ngs
A
s
pe
c
t
K
e
y
f
in
di
ngs
R
e
a
l
-
ti
me
pa
th
oge
n
de
te
c
ti
on
S
ig
ni
f
ic
a
nt
ly
e
nha
nc
e
s
c
r
op he
a
lt
h a
nd yie
ld
c
omp
a
r
e
d t
o t
r
a
di
ti
ona
l
me
th
ods
D
a
ta
in
te
gr
a
ti
on
S
uc
c
e
s
s
f
ul
ly
i
nt
e
gr
a
te
s
a
dva
nc
e
d d
a
ta
a
na
ly
ti
c
s
, i
mpr
ovi
ng pr
e
di
c
ti
ve
c
a
pa
bi
li
ti
e
s
D
e
te
c
ti
on
s
e
n
s
it
iv
it
y
H
ig
he
r
s
e
ns
it
iv
it
y w
it
hout
c
ompr
omi
s
in
g r
e
s
our
c
e
e
f
f
ic
ie
nc
y
S
c
a
la
bi
li
ty
E
f
f
e
c
ti
ve
a
c
r
os
s
di
f
f
e
r
e
nt
c
r
op t
ype
s
a
nd e
nvi
r
onme
nt
a
l
c
ondi
ti
ons
C
os
t
-
e
f
f
e
c
ti
ve
ne
s
s
A
c
hi
e
ve
s
i
mpr
ove
d pa
th
oge
n ma
n
a
ge
me
nt
w
it
hout
i
nc
r
e
a
s
in
g o
pe
r
a
ti
ona
l
c
os
ts
3.
RE
L
AT
E
D
WORK
WI
T
H
E
XI
S
T
I
NG
S
YST
E
M
A
f
a
r
mer
mu
s
t
be
e
xt
r
e
me
ly
k
no
wl
e
dg
e
a
bl
e
a
b
ou
t
e
v
e
r
y
ph
a
s
e
of
c
r
o
p
gr
o
wt
h
a
nd
y
ie
ld
d
e
ve
lo
pm
e
n
t.
S
e
v
e
r
a
l
i
t
e
m
s
i
n
t
hi
s
p
r
o
c
e
du
r
e
c
a
nn
ot
be
id
e
n
ti
f
ie
d
wit
h
t
he
na
k
e
d
e
y
e
o
r
w
it
h
gr
e
a
t
d
il
ig
e
n
c
e
.
I
t
i
s
c
r
uc
ia
l
to
id
e
n
ti
f
y
t
he
s
e
h
a
r
m
f
ul
d
i
s
e
a
s
e
s
th
a
t
a
r
e
d
e
s
t
r
oy
in
g
l
e
a
v
e
s
in
or
d
e
r
t
o
i
n
c
r
e
a
s
e
a
gr
ic
ul
tur
a
l
pr
od
uc
ti
vit
y
[5
3
]
.
A
c
c
or
din
g
t
o
t
h
e
s
u
r
v
e
y,
a
s
ig
ni
f
ic
a
n
t
p
or
t
io
n
o
f
t
he
c
r
op
c
a
n
be
s
p
a
r
e
d
f
r
o
m
h
a
r
m
if
a
s
o
ph
i
s
t
ic
a
t
e
d
a
p
pl
ic
a
t
io
n
i
s
c
r
e
a
t
e
d
t
o
d
e
t
e
c
t
th
e
s
e
di
s
e
a
s
e
s
.
T
h
e
DL
m
e
th
od
s
h
a
v
e
b
e
e
n
e
m
ph
a
s
i
z
e
d
a
nd
tr
a
in
e
d
o
n
im
a
g
e
-
ba
s
e
d
d
a
t
a
s
e
t
s
i
n
t
he
l
a
s
t
s
e
ve
r
a
l
y
e
a
r
s
.
I
n
or
d
e
r
to
in
c
r
e
a
s
e
c
r
o
p
p
r
o
du
c
t
iv
it
y
,
t
he
a
pp
li
c
a
ti
on
s
h
ou
ld
b
e
de
s
ig
ne
d
to
i
de
nt
if
y
pl
a
nt
il
ln
e
s
s
e
s
a
nd
o
f
f
e
r
p
e
s
t
i
c
id
e
s
to
p
r
ot
e
c
t
a
g
a
i
n
s
t
th
e
m.
W
e
c
a
n
a
n
ti
c
i
pa
t
e
l
e
a
f
s
i
c
k
n
e
s
s
in
it
s
e
a
r
l
y
s
t
a
g
e
s
u
s
i
ng
a
va
r
i
e
t
y
of
D
L
a
pp
r
o
a
c
h
e
s
,
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
Ar
ti
f
I
ntell
I
S
S
N:
2252
-
8938
E
nhanc
ing
pr
e
c
is
ion
agr
icultur
e
:
a
c
ompr
e
he
ns
ive
inve
s
ti
gati
on
int
o
pa
thogen
de
tec
ti
on
…
(
Sha
is
ta
F
ar
hat)
3127
i
n
c
lu
di
ng
C
NN
,
r
e
c
ur
r
e
nt
n
e
ur
a
l
n
e
t
wor
k
(
R
NN
)
,
a
nd
G
AN
.
A
m
or
e
a
c
c
u
r
a
te
pr
e
d
ic
ti
on
of
c
r
o
p
l
e
a
f
d
i
s
e
a
s
e
c
a
n
b
e
ob
ta
in
e
d
b
y
c
o
mb
in
in
g
t
w
o
o
r
m
or
e
a
l
gor
it
hm
s
.
A
C
N
N
wit
h
50
l
a
y
e
r
s
d
e
e
p,
k
n
ow
n
a
s
R
e
s
N
e
t
50,
i
s
a
de
e
p
r
e
s
id
ua
l
n
e
t
wo
r
k
wh
o
s
e
mod
e
l
c
a
n
b
e
u
s
e
d
to
pr
e
di
c
t
a
gr
i
c
u
lt
ur
a
l
l
e
a
f
d
i
s
e
a
s
e
[5
4
]
,
[
23]
.
F
ig
ur
e
3
i
ll
u
s
tr
a
t
e
s
th
e
X
c
e
pt
io
n
a
l
go
r
it
hm
a
n
d
th
e
AN
N
t
ha
t
bu
il
d
s
t
he
ne
tw
or
k
b
y
s
tac
k
in
g
th
e
r
e
m
a
i
ni
ng
pi
e
c
e
s
on
to
p
of
e
a
c
h
o
th
e
r
.
F
igur
e
3.
M
ode
l
f
o
r
Xc
e
pti
on
T
he
f
oll
owing
modul
e
s
will
be
a
dde
d
to
c
r
e
a
te
a
f
ull
model:
ne
twor
k
a
nd
dis
e
a
s
e
de
tec
ti
on
a
nd
pr
e
diction,
im
a
ge
a
c
quis
it
ion,
a
nd
im
a
ge
pr
e
pr
o
c
e
s
s
ing
.
T
he
latter
is
divi
de
d
int
o
thr
e
e
s
ubpa
r
ts
:
im
a
ge
s
e
gmenta
ti
on,
f
e
a
tur
e
e
xtr
a
c
ti
on
,
a
nd
c
las
s
if
ica
ti
on
[5
5
]
.
F
igu
r
e
4
s
hows
the
pr
opos
e
d
model's
wor
kf
l
ow.
F
igur
e
4.
S
teps
of
plant
d
is
e
a
s
e
de
tec
ti
on
model
B
y
c
ombi
ning
the
Xc
e
pti
on
a
ppr
oa
c
h
a
nd
R
e
s
Ne
t50
hybr
id
DL
tec
hniques
,
it
is
pos
s
ibl
e
to
f
oc
us
on
e
a
r
ly
pa
thogen
identif
ica
ti
on
[5
6
]
in
the
im
a
ge
s
a
nd
the
ne
twor
k
c
omponent
o
f
the
t
r
a
ined
im
a
ge
s
by
be
ing
a
wa
r
e
of
thi
s
e
xis
ti
ng
wor
k.
Nume
r
ous
r
e
s
e
a
r
c
he
r
s
c
onc
e
ntr
a
ted
on
two
dis
ti
nc
t
a
s
p
e
c
ts
of
lea
r
ning
a
nd
ne
twor
king
s
tr
a
t
e
gies
.
C
ombi
ning
thes
e
two
f
a
c
tor
s
c
ould
incr
e
a
s
e
a
c
c
ur
a
c
y
a
nd
e
na
ble
e
a
r
ly
-
s
tage
de
tec
ti
on
to
r
e
duc
e
a
gr
icultur
a
l
los
s
.
I
n
ter
ms
of
ne
twor
ks
,
n
umer
ous
r
e
s
e
a
r
c
he
r
s
ha
ve
a
lr
e
a
dy
pr
e
s
e
nted
DL
te
c
hniques
[5
7
]
,
[
5
8
]
.
I
n
thi
s
ins
tanc
e
,
mor
e
im
a
ge
s
a
r
e
t
r
a
ined
in
the
ne
twor
k
s
e
gment
if
we
uti
li
z
e
50
lay
e
r
s
of
a
r
e
s
idual
ne
twor
k
us
ing
Xc
e
pti
on
a
ppr
oa
c
he
s
.
T
he
da
tas
e
t
unde
r
the
tr
a
ini
ng
pa
r
t
is
us
e
d
in
the
ne
twor
k
modul
e
,
o
r
R
e
s
Ne
t50
model,
whic
h
ha
s
5
laye
r
s
.
T
his
is
il
lus
tr
a
ted
in
F
igu
r
e
5.
W
e
c
a
n
th
e
n
de
ter
mi
ne
whe
ther
o
r
not
the
lea
ve
s
a
r
e
e
xha
us
ted
f
r
om
thos
e
s
e
gmente
d
im
a
ge
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
Ar
ti
f
I
ntell
,
Vol.
14,
No.
4
,
Augus
t
2025
:
312
1
-
3132
3128
F
igur
e
5.
Ar
c
hit
e
c
tur
e
of
R
e
s
Ne
t5
0
W
he
n
int
e
r
pr
e
ti
ng
r
e
s
ult
s
,
it
is
c
r
uc
ial
to
c
ompar
e
our
f
indi
ngs
with
thos
e
of
other
s
tudi
e
s
.
Our
s
tudy
s
ugge
s
ts
that
higher
de
tec
ti
on
s
e
ns
it
ivi
ty
is
not
a
s
s
oc
iate
d
with
poor
pe
r
f
or
manc
e
in
r
e
s
our
c
e
e
f
f
icie
nc
y,
a
li
gning
with
s
im
il
a
r
f
indi
ngs
in
r
e
c
e
nt
p
r
e
c
is
ion
a
gr
icultur
e
r
e
s
e
a
r
c
h
[
59
]
.
Unlike
s
ome
s
tudi
e
s
whic
h
r
e
por
t
a
tr
a
de
-
of
f
be
twe
e
n
s
e
ns
it
ivi
ty
a
nd
ope
r
a
ti
ona
l
c
os
t,
our
pr
opos
e
d
method
be
ne
f
it
s
f
r
om
inc
r
e
a
s
e
d
da
ta
int
e
gr
a
ti
on
without
a
dve
r
s
e
ly
im
pa
c
ti
ng
ove
r
a
ll
c
os
ts
,
of
f
e
r
ing
a
mor
e
s
us
taina
ble
a
nd
e
f
f
icie
nt
a
ppr
oa
c
h
to
pa
thogen
mana
ge
ment.
T
he
s
ugge
s
ted
method's
pr
im
a
r
y
goa
l
is
to
identif
y
a
ny
dis
e
a
s
e
s
or
il
lnes
s
e
s
that
a
f
f
e
c
t
lea
ve
s
.
Ou
r
goa
l
is
to
us
e
our
s
ugge
s
ted
a
lgor
it
hm
to
identif
y
plant
dis
e
a
s
e
s
in
a
va
r
iety
of
c
r
op
lea
ve
s
,
s
uc
h
a
s
maiz
e
,
tom
a
to,
a
nd
other
plants
.
I
n
or
de
r
to
ge
t
s
igni
f
ica
nt
outcome
s
,
we
a
ls
o
a
im
to
r
e
duc
e
the
tr
a
ini
ng
pe
r
i
od
[6
0
]
.
M
or
e
ove
r
,
the
p
r
opos
e
d
model
whos
e
s
tr
uc
tur
e
i
s
s
uf
f
icie
ntl
y
f
lexible
.
T
he
pu
r
pos
e
of
th
is
s
ur
ve
y's
f
utur
e
e
xpa
ns
ion
may
be
to
a
pply
s
e
c
ur
it
y
to
pa
thogen
-
f
r
e
e
c
r
ops
.
B
y
im
pleme
nti
ng
pr
e
ve
nti
ve
mea
s
ur
e
s
[6
1
]
a
s
s
oon
a
s
the
dis
e
a
s
e
is
identif
ied,
we
c
a
n
e
xpa
nd
thi
s
wor
k
a
nd
incr
e
a
s
e
c
r
op
p
r
oduc
ti
vit
y
.
T
his
s
tudy
e
xplor
e
d
a
c
ompr
e
he
ns
ive
pa
thogen
de
tec
ti
on
s
ys
tem
with
a
dva
nc
e
d
da
ta
int
e
gr
a
ti
on
.
How
e
ve
r
,
f
ur
ther
a
nd
in
-
de
pth
s
tudi
e
s
may
be
ne
e
de
d
to
c
onf
ir
m
it
s
long
-
ter
m
e
f
f
e
c
ti
ve
ne
s
s
,
e
s
pe
c
ially
r
e
ga
r
ding
i
ts
s
c
a
labili
ty
a
c
r
os
s
d
if
f
e
r
e
nt
c
r
op
typ
e
s
a
nd
va
r
ying
e
nvir
on
menta
l
c
ondit
ions
.
M
uc
h
s
tudy
ha
s
be
e
n
done
in
the
f
ield
o
f
i
mage
pr
oc
e
s
s
ing
in
r
e
c
e
nt
ye
a
r
s
,
but
the
r
e
ha
s
a
ls
o
be
e
n
a
lot
of
r
e
s
e
a
r
c
h
done
on
the
c
ombi
na
ti
on
of
ML
a
nd
r
e
a
l
-
wor
ld
c
r
op
r
e
s
e
a
r
c
h
[6
2
]
.
T
he
s
tudy
c
a
n
us
e
DL
to
e
xe
c
ute
the
wor
k
a
nd
f
il
l
up
a
ll
the
ga
ps
.
T
he
DL
a
ppr
oa
c
he
s
will
im
pr
ove
t
he
im
a
ge
's
mi
nute
f
e
a
tur
e
s
,
a
ll
owing
f
or
the
obs
e
r
va
ti
on
of
e
ve
r
y
de
tail
a
nd
the
ident
if
ica
ti
on
o
f
im
a
ge
de
f
e
c
ts
.
Give
n
that
DL
is
a
s
ophis
ti
c
a
ted
type
of
ML
that
c
a
n
e
nha
nc
e
pe
r
f
or
manc
e
.
4.
RE
S
E
AR
CH
GAP
De
s
pit
e
the
im
pr
e
s
s
ive
a
dv
a
nc
e
ment
in
pa
thogen
de
tec
ti
on
f
or
pr
e
c
is
ion
a
gr
icultur
e
,
ther
e
r
e
main
s
e
ve
r
a
l
r
e
s
e
a
r
c
h
ga
ps
.
T
oo
e
t
al
.
[
18
]
e
mpl
oye
d
GAN
-
ba
s
e
d
da
ta
a
ugmenta
ti
on
a
nd
DL
a
r
c
hit
e
c
tur
e
s
li
ke
Ale
xNe
t
a
nd
R
e
s
Ne
t
with
ve
r
y
high
a
c
c
ur
a
c
y,
but
their
model
wa
s
not
r
e
a
l
-
ti
me
c
a
pa
bl
e
f
o
r
dif
f
e
r
e
nt
a
gr
icultur
a
l
e
nvi
r
onments
.
Abdu
e
t
al
.
[
19
]
pr
o
pos
e
d
a
n
a
utom
a
ti
c
de
tec
ti
on
s
ys
tem
us
ing
r
a
di
a
l
ba
s
is
f
unc
ti
on
ne
u
r
a
l
ne
two
r
k
(
B
R
B
F
NN
)
,
but
it
s
pe
r
f
or
manc
e
in
la
r
ge
-
s
c
a
le,
mul
ti
-
c
r
op
e
nvi
r
onment
s
r
e
main
unknown
.
T
he
r
e
s
e
a
r
c
h
in
[
20
]
,
[
21
]
u
s
e
d
DL
a
r
c
h
it
e
c
tur
e
s
li
ke
C
NN
a
nd
R
e
s
Ne
t50
with
ove
r
97%
a
c
c
ur
a
c
y,
but
their
s
tudy
did
not
a
ddr
e
s
s
the
t
r
a
ns
f
e
r
lea
r
ni
ng
is
s
ue
s
a
c
r
os
s
dif
f
e
r
e
nt
c
li
matic
a
nd
r
e
gional
c
ontexts
.
Z
hou
e
t
a
l
.
[
23]
pr
opos
e
d
mobi
le
-
ba
s
e
d
dis
e
a
s
e
de
tec
ti
on
us
ing
f
a
s
ter
R
-
C
NN
but
did
not
e
xplor
e
I
o
T
int
e
gr
a
ti
on
f
o
r
r
e
a
l
-
ti
me
de
tec
ti
on.
W
a
ng
e
t
al
.
[
2
4]
pr
opos
e
d
DL
a
r
c
hit
e
c
tur
e
f
or
c
hil
i
dis
e
a
s
e
c
las
s
if
ica
ti
on,
but
im
ba
lanc
e
d
da
tas
e
t
pr
oblems
we
r
e
pr
e
s
e
nt.
S
im
il
a
r
ly,
E
l
f
a
ti
m
i
e
t
al
.
[
25]
opti
m
ize
d
DL
models
f
or
oli
ve
lea
f
dis
e
a
s
e
de
tec
ti
on
but
lac
ke
d
c
r
os
s
-
c
r
op
ge
ne
r
a
li
z
a
bil
it
y.
S
ha
r
ma
e
t
al
.
[
32
]
e
xplor
e
d
AI
-
ba
s
e
d
pr
e
c
is
ion
f
a
r
mi
ng
but
did
not
incor
po
r
a
te
c
li
mate
c
ha
nge
pa
r
a
mete
r
s
f
o
r
dis
e
a
s
e
pr
e
diction
.
Additi
o
na
ll
y,
the
r
e
s
e
a
r
c
h
in
[
33]
,
[
38
]
de
mons
tr
a
ted
AI
’
s
potential
in
plant
dis
e
a
s
e
diagno
s
is
,
but
da
ta
pr
ivac
y,
s
c
a
labi
li
ty,
a
nd
c
os
t
-
e
f
f
e
c
ti
ve
de
ploym
e
nt
is
s
ue
s
r
e
main.
T
he
s
e
g
a
ps
highl
ight
the
ne
e
d
f
o
r
a
n
int
e
gr
a
ted,
r
e
a
l
ti
m
e
,
mul
ti
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
Ar
ti
f
I
ntell
I
S
S
N:
2252
-
8938
E
nhanc
ing
pr
e
c
is
ion
agr
icultur
e
:
a
c
ompr
e
he
ns
ive
inve
s
ti
gati
on
int
o
pa
thogen
de
tec
ti
on
…
(
Sha
is
ta
F
ar
hat)
3129
c
r
op
dis
e
a
s
e
de
tec
ti
on
s
y
s
tem
whic
h
s
umm
a
r
ize
s
i
n
the
T
a
ble
4
whic
h
de
tails
a
bout
the
c
ur
r
e
nt
r
e
s
e
a
r
c
h
ga
p
a
nd
it
s
c
ha
ll
e
nge
s
.
T
a
ble
4.
S
umm
a
r
iza
ti
on
of
r
e
s
e
a
r
c
h
ga
p
a
nd
it
s
c
h
a
ll
e
nge
s
R
e
f
R
e
s
e
a
r
c
h
g
a
p
C
ha
ll
e
nge
s
[
18]
L
a
c
k
of
r
e
a
l
-
ti
me
a
nd
s
c
a
la
bl
e
pl
a
nt
de
te
c
ti
on
mode
ls
T
hi
s
mode
l
la
c
ks
th
e
a
da
pt
a
bi
li
ty
f
or
di
ve
r
s
e
e
nvi
r
onme
nt
a
l
c
ondi
ti
ons
a
nd mul
ti
-
c
r
op
a
ppl
ic
a
ti
ons
.
[
20]
I
n
DL
mode
l’
s
ove
r
f
it
ti
ng i
s
s
ue
s
a
r
e
e
nc
ount
e
r
e
d
C
N
N
mode
ls
te
nd
to
ove
r
f
it
w
he
n
tr
a
in
e
d
on
li
mi
te
d
d
a
ta
s
e
ts
by
r
e
duc
in
g t
he
ge
ne
r
a
li
z
a
bi
li
ty
.
[
21]
T
he
y
us
e
d
hi
ghl
y
c
omput
a
ti
ona
l
r
e
qui
r
e
me
nt
s
f
or
DL
mode
ls
.
R
e
a
l
ti
me
a
ppl
ic
a
ti
on
is
not
pos
s
ib
le
a
s
ma
ny
A
I
mode
ls
r
e
qui
r
e
hi
gh
-
e
nd ha
r
dw
a
r
e
, ma
ki
ng t
he
a
dopt
io
n di
f
f
ic
ul
t
f
or
s
ma
ll
-
s
c
a
le
f
a
r
me
r
s
.
[
23]
T
he
r
e
i
s
a
l
im
it
e
d us
e
of
I
oT
w
it
h t
he
c
ombi
na
ti
on
of
A
I
f
or
s
ma
r
t
a
gr
ic
ul
tu
r
e
T
he
r
e
is
no
r
e
a
l
ti
me
da
ta
c
ol
le
c
ti
on
a
nd
A
I
-
dr
iv
e
n
a
na
ly
ti
c
s
f
or
c
ont
in
uous
moni
to
r
in
g of
c
r
op s
ys
te
ms
.
[
25]
N
o f
r
a
me
w
or
k f
or
mul
ti
-
c
r
op dis
e
a
s
e
de
te
c
ti
on
M
a
ny
s
tu
di
e
s
f
oc
us
on
s
in
gl
e
-
c
r
op
di
s
e
a
s
e
d
e
te
c
ti
on
but
li
mi
te
d
on
c
r
os
s
-
c
r
op
de
te
c
ti
on
.
[
32]
T
he
ma
jo
r
out
br
e
a
k
s
a
r
e
not
c
ons
id
e
r
e
d
w
he
n
th
e
r
e
i
s
an
im
pa
c
t
on c
li
ma
te
.
T
hi
s
mode
l
doe
s
not
f
a
c
to
r
in
c
li
ma
te
c
ha
nge
e
f
f
e
c
ts
on
p
a
th
oge
n
be
ha
vi
or
a
nd dis
e
a
s
e
s
pr
e
a
d.
[
33]
T
he
ma
jo
r
ga
p
is
in
A
I
-
dr
iv
e
n
a
gr
ic
ul
tu
r
e
s
e
c
ur
it
y
a
nd
pr
iv
a
c
y c
onc
e
r
ns
a
r
e
not
c
on
s
id
e
r
e
d
D
a
ta
pr
iv
a
c
y
a
nd
c
yb
e
r
s
e
c
ur
it
y
c
onc
e
r
ns
a
r
e
a
r
is
in
g
w
it
h
la
r
ge
-
s
c
a
le
da
ta
c
ol
le
c
ti
on.
T
o
br
idge
thes
e
ga
ps
,
thi
s
pa
pe
r
pr
e
s
e
nts
a
r
e
li
a
ble
DL
-
b
a
s
e
d
pa
thogen
de
te
c
ti
on
s
ys
tem
us
ing
R
e
s
Ne
t50
a
nd
C
NN
f
o
r
incr
e
a
s
e
d
a
c
c
ur
a
c
y
a
nd
f
lexibi
li
ty.
R
e
s
Ne
t50,
in
it
s
de
e
p
r
e
s
idual
lea
r
ning
f
r
a
mew
or
k,
e
ns
ur
e
s
im
pr
ove
d
f
e
a
tur
e
e
xtr
a
c
ti
o
n
a
nd
c
las
s
if
ica
ti
on
of
plant
dis
e
a
s
e
s
while
mi
ti
ga
ti
ng
ove
r
f
it
ti
ng
is
s
ue
s
.
T
he
C
NN
model
is
in
tegr
a
ted
with
t
r
a
ns
f
e
r
lea
r
ning
tec
hniques
that
e
na
ble
c
r
os
s
-
c
r
op
a
da
ptabili
ty
a
nd
e
f
f
icie
nt
lea
r
ning
f
r
om
diver
s
e
a
g
r
icultur
a
l
da
tas
e
ts
.
Als
o,
r
e
a
l
-
ti
me
pa
thogen
de
tec
ti
on
will
be
a
c
hieve
d
by
int
e
gr
a
ti
on
of
I
oT
s
e
ns
or
s
a
nd
c
loud
-
ba
s
e
d
moni
tor
ing
f
or
c
onti
nue
d
dis
e
a
s
e
s
ur
ve
il
lanc
e
a
nd
e
a
r
ly
-
s
tage
diagnos
i
s
.
F
ur
ther
,
the
i
mbala
nc
e
s
in
da
tas
e
ts
will
a
ls
o
be
a
ddr
e
s
s
e
d
us
ing
da
ta
a
ugmenta
ti
on
tec
hniques
,
a
long
with
the
hybr
id
DL
models
tha
t
im
pr
ove
on
ge
ne
r
a
li
z
a
ti
on.
F
inally
,
int
e
gr
a
ti
ng
c
li
mate
-
ba
s
e
d
pr
e
dictive
a
na
lyt
ics
will
e
nha
nc
e
dis
e
a
s
e
f
or
e
c
a
s
ti
ng
s
o
that
mea
s
ur
e
s
c
a
n
be
take
n
pr
oa
c
ti
ve
ly
by
us
e
r
s
:
f
a
r
mer
s
.
T
his
AI
-
dr
iven
s
olut
ion
br
idges
th
os
e
ga
ps
in
r
e
s
e
a
r
c
h
by
o
f
f
e
r
ing
a
s
c
a
lable
,
c
os
t
-
e
f
f
e
c
ti
ve
,
a
nd
r
e
a
l
-
ti
me
pa
thogen
de
tec
ti
on
f
r
a
mew
or
k
a
da
ptable
a
c
r
os
s
dif
f
e
r
e
nt
c
r
ops
a
nd
c
li
matic
c
ondit
ions
that
will
r
e
volut
ioni
z
e
pr
e
c
is
ion
a
gr
icultu
r
e
.
5.
CONC
L
USI
ON
AN
D
F
UT
UR
E
S
COP
E
R
e
c
e
nt
obs
e
r
va
ti
ons
s
ugge
s
t
that
int
e
gr
a
ti
ng
a
d
va
nc
e
d
pa
thogen
de
tec
ti
on
s
ys
tems
s
igni
f
ica
ntl
y
e
nha
nc
e
s
c
r
op
he
a
lt
h
a
nd
yield.
Our
f
indi
ngs
pr
ovide
c
onc
lus
ive
e
videnc
e
that
thi
s
im
pr
ov
e
ment
is
a
s
s
oc
iate
d
with
the
im
pleme
ntation
of
r
e
a
l
-
ti
me
de
tec
ti
on
a
nd
da
ta
a
na
lyt
ics
,
not
due
to
e
leva
ted
number
s
of
pa
thogen
-
r
e
s
i
s
tant
c
r
op
va
r
ieties
.
W
e
r
e
view
e
d
the
li
ter
a
tur
e
on
tec
hnology
-
dr
iven
pr
e
c
is
ion
f
a
r
mi
ng
tec
hniques
in
thi
s
r
e
s
e
a
r
c
h.
W
it
h
pr
e
c
is
io
n
f
a
r
mi
n
g,
f
a
r
mer
s
c
a
n
r
e
s
pond
ins
tantly
to
c
r
op
r
e
quir
e
m
e
nts
f
or
maximum
yield
a
nd
lowe
s
t
c
os
t.
S
ince
a
ll
c
ountr
ies
f
oc
us
e
d
on
a
gr
icultur
e
a
s
pir
e
to
pr
e
c
is
ion
a
gr
icultur
e
,
numer
ous
c
ountr
ies
,
including
I
nd
ia,
ha
ve
be
e
n
m
a
king
e
f
f
o
r
ts
in
thi
s
di
r
e
c
ti
on.
T
h
e
a
r
ti
c
le
p
r
e
s
e
nts
mul
ti
ple
s
tudy
f
indi
ngs
.
F
i
r
s
tl
y,
a
gr
icultu
r
a
l
r
e
s
e
a
r
c
h
make
s
e
xtens
ive
us
e
of
AI
a
nd
r
e
late
d
tec
hniques
li
ke
M
L
a
nd
DL
.
S
e
c
ond,
it
ha
s
be
e
n
dis
c
ove
r
e
d
that
DL
mod
e
ls
buil
t
on
C
NN
a
r
e
mor
e
e
f
f
icie
nt
a
t
p
r
oc
e
s
s
ing
pictur
e
input
s
.
T
h
ir
d,
r
eal
-
ti
me
da
ta
pr
oc
e
s
s
ing.
T
he
I
oT
c
onne
c
ti
on
with
a
utom
a
ted
f
a
r
mi
ng
a
utom
a
tes
l
ive
da
ta
ga
ther
ing.
I
t
is
de
s
ir
a
ble
to
ha
ve
a
n
I
o
T
-
c
ombi
ne
d
AI
-
ba
s
e
d
s
ys
tem
f
or
c
r
op
obs
e
r
va
ti
on.
Additi
on
a
ll
y,
the
s
tudy
identif
ies
s
igni
f
ica
nt
r
e
s
e
a
r
c
h
ga
ps
that
s
uppor
t
the
a
dva
nc
e
ment
of
pr
e
c
is
ion
a
gr
icultu
r
e
.
O
ur
s
tudy
de
mons
tr
a
tes
that
c
r
ops
moni
tor
e
d
with
r
e
a
l
-
ti
me
pa
thogen
de
tec
ti
on
s
ys
tem
s
a
r
e
mor
e
r
e
s
il
ient
than
thos
e
r
e
lyi
ng
on
tr
a
dit
ional
methods
.
F
utur
e
s
tudi
e
s
ma
y
e
xplor
e
the
int
e
gr
a
ti
on
o
f
M
L
a
lgor
it
hms
with
r
e
a
l
-
t
im
e
de
tec
ti
on
s
ys
tems
,
with
f
e
a
s
ibl
e
wa
ys
of
pr
oduc
i
ng
pr
e
dictive
models
to
f
u
r
ther
e
nha
nc
e
c
r
op
mana
ge
ment
a
nd
yield
opti
mi
z
a
ti
on
.
F
UN
DI
NG
I
NF
ORM
AT
I
ON
Author
s
s
tate
no
f
unding
invol
ve
d.
AU
T
HO
R
CONT
RI
B
U
T
I
ONS
S
T
AT
E
M
E
N
T
T
his
jour
na
l
us
e
s
the
C
ontr
ibut
o
r
R
oles
T
a
xo
nomy
(
C
R
e
diT
)
to
r
e
c
ognize
indi
vidual
a
uthor
c
ontr
ibut
ions
,
r
e
duc
e
a
utho
r
s
hip
dis
putes
,
a
nd
f
a
c
il
it
a
te
c
oll
a
bor
a
ti
on
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
Ar
ti
f
I
ntell
,
Vol.
14,
No.
4
,
Augus
t
2025
:
312
1
-
3132
3130
Nam
e
of
Au
t
h
or
C
M
So
Va
Fo
I
R
D
O
E
Vi
Su
P
Fu
S
ha
is
ta
F
a
r
ha
t
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
C
hokka
Anur
a
dha
✓
✓
✓
✓
✓
✓
✓
✓
C
:
C
onc
e
pt
ua
li
z
a
ti
on
M
:
M
e
th
odol
ogy
So
:
So
f
twa
r
e
Va
:
Va
li
da
ti
on
Fo
:
Fo
r
ma
l
a
na
ly
s
is
I
:
I
nve
s
ti
ga
ti
on
R
:
R
e
s
our
c
e
s
D
:
D
a
ta
C
ur
a
ti
on
O
:
W
r
it
in
g
-
O
r
ig
in
a
l
D
r
a
f
t
E
:
W
r
it
in
g
-
R
e
vi
e
w
&
E
di
ti
ng
Vi
:
Vi
s
ua
li
z
a
ti
on
Su
:
Su
pe
r
vi
s
io
n
P
:
P
r
oj
e
c
t
a
dmi
ni
s
tr
a
ti
on
Fu
:
Fu
ndi
ng a
c
qui
s
it
io
n
CONF
L
I
CT
OF
I
NT
E
RE
S
T
S
T
AT
E
M
E
N
T
Author
s
s
tate
no
c
onf
li
c
t
of
int
e
r
e
s
t.
DA
T
A
AV
AI
L
A
B
I
L
I
T
Y
T
he
da
ta
that
s
uppor
t
the
f
indi
ngs
of
thi
s
s
tudy
a
r
e
a
va
il
a
ble
f
r
om
the
c
or
r
e
s
ponding
a
utho
r
upon
r
e
a
s
ona
ble
r
e
que
s
t.
All
r
e
leva
nt
da
ta
we
r
e
ge
ne
r
a
t
e
d
a
nd
a
na
lyze
d
dur
ing
the
c
ur
r
e
nt
s
tudy
a
nd
c
a
n
be
s
ha
r
e
d
f
or
a
c
a
de
mi
c
a
nd
r
e
s
e
a
r
c
h
pur
pos
e
s
.
RE
F
E
RE
NC
E
S
[
1]
S
.
B
.
P
a
ti
l
a
nd
S
.
K
.
B
odhe
,
“
L
e
a
f
di
s
e
a
s
e
s
e
ve
r
it
y
me
a
s
ur
e
m
e
nt
us
in
g
im
a
ge
pr
oc
e
s
s
in
g,”
I
nt
e
r
nat
io
nal
J
our
nal
of
E
ngi
ne
e
r
in
g
and T
e
c
hnol
ogy
, vol
. 3, no. 5, pp. 297
–
301, 2011.
[
2]
S
.
D
.
K
hi
r
a
de
a
nd
A
.
B
.
P
a
ti
l,
“
P
la
nt
di
s
e
a
s
e
de
t
e
c
ti
on
us
in
g
im
a
ge
pr
oc
e
s
s
in
g,”
in
1s
t
I
nt
e
r
nat
io
nal
C
onf
e
r
e
nc
e
on
C
om
put
in
g,
C
om
m
uni
c
at
io
n, C
ont
r
ol
and A
ut
om
at
io
n, I
C
C
U
B
E
A
2015
, 20
15, pp. 768
–
771, doi:
10.1109/I
C
C
U
B
E
A
.2015.153.
[
3]
S
.
V
.
M
il
it
a
nt
e
,
B
.
D
.
G
e
r
a
r
do,
a
nd
N
.
V
.
D
I
oni
s
io
,
“
P
la
nt
le
a
f
de
te
c
ti
on
a
nd
di
s
e
a
s
e
r
e
c
ogni
ti
on
us
in
g
de
e
p
l
e
a
r
ni
ng,”
2019
I
E
E
E
E
ur
as
ia
C
on
fe
r
e
nc
e
on
I
O
T
,
C
om
m
uni
c
at
io
n
and
E
ngi
ne
e
r
in
g,
E
C
I
C
E
2019
,
pp.
579
–
582,
2019,
doi
:
10.1109/E
C
I
C
E
47484.2019.8942686.
[
4]
S
.
B
ha
numa
th
i,
M
.
V
in
e
e
th
,
a
nd
N
.
R
ohi
t,
“
C
r
op
yi
e
ld
pr
e
di
c
ti
on
a
nd
e
f
f
ic
ie
nt
us
e
of
f
e
r
ti
li
z
e
r
s
,”
in
2019
I
nt
e
r
nat
i
onal
C
onf
e
r
e
nc
e
on C
om
m
uni
c
at
io
n and Signal
P
r
oc
e
s
s
in
g (
I
C
C
SP
)
, A
pr
. 2019, pp. 0769
–
0773, doi:
10.1109/I
C
C
S
P
.2019.869808
7.
[
5]
M
.
A
di
e
t
al
.
,
“
A
n
ove
r
vi
e
w
on
pl
a
nt
di
s
e
a
s
e
de
t
e
c
ti
on
a
lg
or
it
hm
us
in
g
de
e
p
le
a
r
ni
ng,”
P
r
oc
e
e
di
ng
s
of
2021
2nd
I
nt
e
r
nat
i
onal
C
onf
e
r
e
nc
e
on
I
nt
e
ll
ig
e
nt
E
ngi
ne
e
r
in
g
and
M
anage
m
e
nt
,
I
C
I
E
M
2021
,
pp.
305
–
309,
2021,
doi
:
10.1109/I
C
I
E
M
51511.2021.9445336.
[
6]
D
.
O
.
O
ye
w
ol
a
,
E
.
G
.
D
a
da
,
S
.
M
is
r
a
,
a
nd
R
.
D
a
ma
š
e
vi
č
iu
s
,
“
D
e
te
c
ti
ng
c
a
s
s
a
va
mo
s
a
ic
di
s
e
a
s
e
us
in
g
a
de
e
p
r
e
s
i
dua
l
c
onvolut
io
na
l
ne
ur
a
l
ne
twor
k
w
it
h
di
s
ti
nc
t
bl
oc
k
p
r
oc
e
s
s
i
ng,”
P
e
e
r
J
C
om
put
e
r
Sc
i
e
nc
e
,
vol
.
7,
pp.
1
–
15,
2021,
doi
:
10.7717/pee
r
j
-
c
s
.352.
[
7]
M
.
A
hma
d,
M
.
A
bdul
la
h,
H
.
M
oon,
a
nd
D
.
H
a
n,
“
P
la
nt
di
s
e
a
s
e
de
t
e
c
ti
on
in
im
ba
la
nc
e
d
da
ta
s
e
ts
u
s
in
g
e
f
f
ic
ie
nt
c
onvolut
i
ona
l
ne
ur
a
l
ne
twor
ks
w
it
h
s
te
pw
is
e
tr
a
n
s
f
e
r
le
a
r
ni
ng,”
I
E
E
E
A
c
c
e
s
s
,
vol
.
9,
pp.
140565
–
140580,
2021,
doi
:
10.1109/AC
C
E
S
S
.2021.3119655.
[
8]
A
.
F
ue
nt
e
s
,
S
.
Y
oon,
S
.
C
.
K
im
,
a
nd
D
.
S
.
P
a
r
k,
“
A
r
obus
t
de
e
p
-
le
a
r
ni
ng
-
ba
s
e
d
de
te
c
to
r
f
or
r
e
a
l
-
ti
me
to
ma
to
pl
a
nt
di
s
e
a
s
e
s
a
nd
pe
s
ts
r
e
c
ogni
ti
on
,”
Se
n
s
or
s
, vol
. 17, no. 9, 2017, doi
:
10.3390/s
17092022.
[
9]
M
.
E
.
S
a
hi
n,
“
I
ma
ge
pr
oc
e
s
s
in
g
a
nd
ma
c
hi
n
e
le
a
r
ni
ng
-
ba
s
e
d
bone
f
r
a
c
tu
r
e
de
te
c
ti
on
a
nd
c
la
s
s
if
ic
a
ti
on
u
s
in
g
X
-
r
a
y
im
a
ge
s
,
”
I
nt
e
r
nat
io
nal
J
our
nal
of
I
m
agi
ng Sy
s
te
m
s
and T
e
c
hnol
ogy
, vol
.
33, no. 3,
pp. 853
–
865, 2023, doi:
10.1002/i
ma
.22849.
[
10]
S
.
H
a
me
d,
H
.
M
a
hmoud,
a
nd
M
.
M
ous
a
,
“
P
la
nt
di
s
e
a
s
e
de
t
e
c
ti
on
s
ui
ng
de
e
p
le
a
r
ni
ng,”
I
nt
e
r
nat
io
nal
J
our
nal
of
I
nt
e
ll
ig
e
nt
Sy
s
te
m
s
and A
ppl
ia
ti
ons
, vol
. 15, no. 6, pp. 38
–
50, 2023.
[
11]
M
.
A
s
ta
ni
,
M
.
H
a
s
he
mi
ne
ja
d,
a
nd
M
.
V
a
gh
e
f
i,
“
A
di
ve
r
s
e
e
n
s
e
mbl
e
c
la
s
s
if
ie
r
f
or
to
ma
to
di
s
e
a
s
e
r
e
c
ogni
ti
on,”
C
om
put
e
r
s
and
E
le
c
tr
oni
c
s
i
n A
gr
ic
ul
tu
r
e
, vol
. 198, 2022, doi:
10.1016/j
.c
omp
a
g.2022.107054.
[
12]
P
.
S
.
K
a
nda
,
K
.
X
ia
,
A
.
K
ys
ly
ty
s
n
a
,
a
nd
E
.
O
.
O
w
oo
la
,
“
T
o
ma
to
le
a
f
di
s
e
a
s
e
r
e
c
ogni
ti
on
on
le
a
f
im
a
ge
s
ba
s
e
d
on
f
in
e
-
tu
ne
d
r
e
s
id
ua
l
ne
ur
a
l
ne
twor
ks
,”
P
la
nt
s
, vol
. 11, no. 21, 2022, doi
:
10.3390/pl
a
nt
s
11212935.
[
13]
P
.
G
ui
,
W
.
D
a
ng,
F
.
Z
hu,
a
nd
Q
.
Z
ha
o,
“
T
ow
a
r
ds
a
ut
oma
ti
c
f
ie
ld
pl
a
nt
di
s
e
a
s
e
r
e
c
ogni
ti
on,”
C
om
put
e
r
s
and
E
le
c
tr
oni
c
s
in
A
gr
ic
ul
tu
r
e
, vol
. 191, 2021, doi:
10.1016/j
.c
ompa
g.2021.106523.
[
14]
D
e
r
is
ma
,
N
.
R
okhm
a
n,
a
nd
I
.
U
s
uma
n,
“
S
ys
t
e
ma
ti
c
r
e
vi
e
w
of
th
e
e
a
r
ly
de
te
c
ti
on
a
nd
c
la
s
s
if
ic
a
ti
on
of
pl
a
nt
di
s
e
a
s
e
s
u
s
in
g
d
e
e
p
le
a
r
ni
ng,”
I
O
P
C
onf
e
r
e
nc
e
Se
r
ie
s
:
E
ar
th
and
E
nv
ir
on
m
e
nt
al
Sc
ie
nc
e
,
vol
.
1097,
no.
1,
2022,
doi
:
10.1088/17
55
-
1315/1097/
1/
012042.
[
15]
M
.
S
ha
r
if
,
M
.
A
.
K
ha
n,
Z
.
I
qba
l,
M
.
F
.
A
z
a
m,
M
.
I
.
U
.
L
a
li
,
a
nd
M
.
Y
.
J
a
ve
d,
“
D
e
te
c
ti
on
a
nd
c
la
s
s
if
ic
a
ti
on
of
c
it
r
us
di
s
e
a
s
e
s
in
a
gr
ic
ul
tu
r
e
ba
s
e
d on opti
m
iz
e
d w
e
ig
ht
e
d s
e
gme
nt
a
ti
on a
nd f
e
a
t
ur
e
s
e
le
c
ti
on,”
C
om
put
e
r
s
and E
le
c
tr
oni
c
s
i
n A
gr
ic
ul
tu
r
e
, vol
.
150,
pp. 220
–
234, 2018, doi:
10.1016/j
.c
ompa
g.2018.04.023.
[
16]
E
mwe
ba
z
e
,
“
M
a
ke
r
e
r
e
A
I
L
a
b
a
nd
A
I
R
L
a
b
N
a
ti
ona
l
C
r
ops
R
e
s
our
c
e
s
R
e
s
e
a
r
c
h
I
ns
ti
tu
te
(
N
a
C
R
R
I
)
,”
G
it
H
ub
,
2020.
A
c
c
e
s
s
e
d
:
M
a
r
. 31, 2022
. [
O
nl
in
e
]
. A
va
il
a
bl
e
:
ht
tp
s
:/
/g
it
hub.c
om/
A
I
-
L
a
b
-
M
a
ke
r
e
r
e
/i
be
a
n/
bl
ob/
ma
s
te
r
/R
E
A
D
M
E
.md
[
17]
M
.
M
ous
s
a
f
ir
,
H
.
C
ha
ib
i,
R
.
S
a
a
da
ne
,
A
.
C
he
hr
i,
A
.
E
l
R
ha
r
r
a
s
,
a
nd
G
.
J
e
on,
“
D
e
s
ig
n
of
e
f
f
ic
ie
nt
te
c
hni
que
s
f
or
to
ma
to
le
a
f
di
s
e
a
s
e
d
e
te
c
ti
on
us
in
g
ge
n
e
ti
c
a
lg
or
it
hm
-
ba
s
e
d
a
nd
de
e
p
n
e
ur
a
l
ne
twor
ks
,”
P
la
nt
and
Soi
l
,
vol
.
479,
no.
1
–
2,
pp.
251
–
266,
2
022,
doi
:
10.1007/s
11104
-
022
-
05513
-
2.
[
18]
E
.
C
.
T
oo,
L
.
Y
uj
ia
n,
S
.
N
ju
ki
,
a
nd
L
.
Y
in
gc
hun,
“
A
c
ompa
r
a
ti
ve
s
tu
dy
of
f
in
e
-
tu
ni
ng
de
e
p
le
a
r
ni
ng
mode
l
s
f
or
pl
a
nt
di
s
e
a
s
e
id
e
nt
if
ic
a
ti
on,”
C
om
put
e
r
s
and E
le
c
t
r
oni
c
s
i
n A
gr
ic
ul
tu
r
e
, vol
.
161, pp. 272
–
279, 2019, doi:
10.1016/j
.c
ompa
g.2018.03.032.
[
19]
A
.
M
.
A
bdu,
M
.
M
.
M
okj
i,
a
nd
U
.
U
.
S
he
ik
h,
“
A
ut
oma
ti
c
ve
ge
ta
bl
e
di
s
e
a
s
e
id
e
nt
if
ic
a
ti
on
a
ppr
oa
c
h
us
in
g
in
di
vi
dua
l
le
s
io
n
f
e
a
tu
r
e
s
,”
C
om
put
e
r
s
and E
le
c
tr
oni
c
s
i
n A
gr
ic
ul
tu
r
e
, vol
. 176,
2020, doi:
10.1016/j
.c
ompa
g.2020.105660.
Evaluation Warning : The document was created with Spire.PDF for Python.