I
nte
rna
t
io
na
l J
o
urna
l o
f
E
lect
rica
l a
nd
Co
m
pu
t
er
E
ng
ineering
(
I
J
E
CE
)
Vo
l.
15
,
No
.
6
,
Decem
b
er
20
25
,
p
p
.
5
6
6
7
~
5
6
7
8
I
SS
N:
2088
-
8
7
0
8
,
DOI
: 1
0
.
1
1
5
9
1
/ijece.
v
15
i
6
.
pp
5
6
6
7
-
5
6
7
8
5667
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
ec
e.
ia
esco
r
e.
co
m
Fine
-
tuning p
re
-
t
ra
ined deep
learn
ing
mo
dels for c
r
o
p
prediction
using
s
o
il conditio
ns in s
ma
rt
a
g
ricultur
e
P
ra
v
ee
n P
a
wa
s
ka
r
1
,2
,
Yo
g
is
h H
K
1
,
P
a
k
rudd
in
B
2
,
Dee
pa
Yo
g
is
h
3
1
D
e
p
a
r
t
me
n
t
o
f
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
a
n
d
E
n
g
i
n
e
e
r
i
n
g
,
M
S
R
a
m
a
i
a
h
I
n
st
i
t
u
t
e
o
f
T
e
c
h
n
o
l
o
g
y
,
B
e
n
g
a
l
u
r
u
,
V
i
sv
e
sv
a
r
a
y
a
Te
c
h
n
o
l
o
g
y
U
n
i
v
e
r
si
t
y
,
B
e
l
a
g
a
v
i
,
I
n
d
i
a
2
D
e
p
a
r
t
me
n
t
o
f
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
a
n
d
E
n
g
i
n
e
e
r
i
n
g
,
P
r
e
si
d
e
n
c
y
U
n
i
v
e
r
si
t
y
,
B
e
n
g
a
l
u
r
u
,
I
n
d
i
a
3
D
e
p
a
r
t
me
n
t
o
f
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
a
n
d
E
n
g
i
n
e
e
r
i
n
g
,
C
h
r
i
s
t
U
n
i
v
e
r
s
i
t
y
,
B
e
n
g
a
l
u
r
u
,
I
n
d
i
a
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Ma
y
2
2
,
2
0
2
5
R
ev
is
ed
J
u
l 2
1
,
2
0
2
5
Acc
ep
ted
Sep
1
4
,
2
0
2
5
Ag
ricu
lt
u
re
is
t
h
e
b
a
c
k
b
o
n
e
o
f
t
h
e
In
d
ian
e
c
o
n
o
m
y
,
wit
h
s
o
il
q
u
a
li
ty
p
lay
in
g
a
c
ru
c
ial
ro
le
in
c
ro
p
p
r
o
d
u
c
ti
v
it
y
.
F
a
rm
e
rs
o
fte
n
str
u
g
g
le
t
o
se
lec
t
th
e
a
p
p
ro
p
riate
c
ro
p
b
a
se
d
o
n
so
il
ty
p
e
,
lea
d
in
g
t
o
si
g
n
if
ica
n
t
l
o
ss
e
s
in
y
iel
d
a
n
d
p
ro
d
u
c
ti
v
it
y
.
T
o
a
d
d
re
ss
th
is
c
h
a
l
len
g
e
,
d
e
e
p
lea
rn
in
g
tec
h
n
i
q
u
e
s
p
ro
v
i
d
e
a
n
e
fficie
n
t
so
l
u
ti
o
n
fo
r
a
u
t
o
m
a
ted
s
o
il
c
las
sifica
ti
o
n
.
In
th
is
stu
d
y
,
a
d
a
tas
e
t
o
f
7
8
1
o
ri
g
in
a
l
s
o
il
ima
g
e
s,
i
n
c
lu
d
in
g
c
lay
so
il
,
a
ll
u
v
ial
so
il
,
re
d
s
o
il
,
a
n
d
b
lac
k
so
il
,
wa
s
c
o
l
lec
ted
fro
m
Ka
g
g
le
a
n
d
a
u
g
m
e
n
ted
to
3
,
7
0
2
ima
g
e
s
t
o
e
n
h
a
n
c
e
m
o
d
e
l
train
in
g
.
S
e
v
e
ra
l
d
e
e
p
l
e
a
rn
in
g
m
o
d
e
ls
we
re
e
m
p
lo
y
e
d
fo
r
so
il
c
las
sifica
ti
o
n
,
in
c
lu
d
i
n
g
p
re
trai
n
e
d
a
rc
h
it
e
c
tu
re
s
a
n
d
a
p
ro
p
o
se
d
m
o
d
e
l,
S
o
i
lNe
t.
E
x
p
e
rim
e
n
ta
l
re
s
u
l
ts
d
e
m
o
n
s
tra
te
d
t
h
a
t
De
n
se
Ne
t
2
0
1
a
c
h
ie
v
e
d
1
0
0
%
v
a
li
d
a
ti
o
n
a
c
c
u
ra
c
y
,
Re
sN
e
t5
0
V
2
9
8
%
,
VG
G
1
6
9
9
%
,
M
o
b
i
leN
e
tV2
9
9
%
,
a
n
d
t
h
e
p
r
o
p
o
se
d
S
o
il
Ne
t
m
o
d
e
l
9
7
%
.
T
h
e
p
r
o
p
o
se
d
a
p
p
r
o
a
c
h
o
u
t
p
e
rfo
rm
e
d
e
x
isti
n
g
w
o
rk
b
y
su
r
p
a
ss
in
g
9
5
%
a
c
c
u
ra
c
y
.
Ad
d
it
i
o
n
a
ll
y
,
m
o
d
e
l
p
e
rfo
rm
a
n
c
e
wa
s
e
v
a
lu
a
ted
u
si
n
g
p
re
c
isio
n
,
re
c
a
ll
,
a
n
d
F
1
-
sc
o
re
,
e
n
su
rin
g
a
c
o
m
p
re
h
e
n
siv
e
a
n
a
ly
sis
o
f
c
la
ss
ifi
c
a
ti
o
n
e
ffe
c
ti
v
e
n
e
ss
.
T
h
e
se
fin
d
in
g
s
h
ig
h
li
g
h
t
t
h
e
p
o
ten
t
ial
o
f
d
e
e
p
lea
rn
in
g
in
imp
ro
v
i
n
g
s
o
il
c
la
ss
ifi
c
a
ti
o
n
a
c
c
u
ra
c
y
,
a
id
in
g
fa
rm
e
rs i
n
m
a
k
in
g
in
f
o
rm
e
d
c
ro
p
se
lec
ti
o
n
d
e
c
isio
n
s.
K
ey
w
o
r
d
s
:
Ag
r
icu
ltu
r
e
Dee
p
lear
n
in
g
Featu
r
e
ex
tr
ac
tio
n
I
m
ag
e
p
r
o
ce
s
s
in
g
Pre
tr
ain
ed
m
o
d
els
So
il ty
p
e
p
r
ed
ictio
n
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
:
Pra
v
ee
n
Pawask
ar
Dep
ar
tm
en
t o
f
C
o
m
p
u
ter
Scie
n
ce
an
d
E
n
g
in
ee
r
in
g
,
M
S R
a
m
aiah
I
n
s
titu
te
o
f
T
ec
h
n
o
l
o
g
y
,
B
en
g
alu
r
u
,
Vis
v
esv
ar
ay
a
T
ec
h
n
o
l
o
g
ical
Un
iv
er
s
ity
B
elag
av
i
–
5
9
0
0
1
8
,
I
n
d
ia
Dep
ar
tm
en
t o
f
C
o
m
p
u
ter
Scie
n
ce
an
d
E
n
g
in
ee
r
in
g
,
Pre
s
id
en
cy
Un
iv
er
s
ity
B
en
g
alu
r
u
–
5
6
0
0
6
4
,
I
n
d
ia
E
m
ail:
p
r
av
ee
n
p
awa
s
k
ar
5
5
5
@
g
m
ail.
co
m
1.
I
NT
RO
D
UCT
I
O
N
T
h
e
f
o
u
n
d
atio
n
o
f
wo
r
ld
wid
e
f
o
o
d
p
r
o
d
u
ctio
n
is
a
g
r
icu
ltu
r
e
,
an
d
cr
o
p
y
ield
is
h
ea
v
ily
in
f
l
u
en
ce
d
b
y
th
e
q
u
ality
o
f
th
e
s
o
il.
So
il
ty
p
es
—
lik
e
s
an
d
y
,
clay
ey
,
lo
a
m
y
,
an
d
s
ilt
—
d
if
f
er
in
tex
tu
r
e,
wate
r
r
eten
tio
n
ca
p
ac
ity
,
an
d
n
u
t
r
ien
t
av
aila
b
ilit
y
,
wh
ich
d
i
r
ec
tly
af
f
ec
ts
p
lan
t
g
r
o
wth
.
Far
m
e
r
s
n
ee
d
to
u
n
d
er
s
tan
d
s
o
il
p
r
o
p
er
ties
in
o
r
d
e
r
to
m
a
k
e
in
f
o
r
m
ed
ch
o
ices
ab
o
u
t
cr
o
p
s
el
ec
tio
n
,
ir
r
ig
atio
n
,
an
d
f
er
tili
za
tio
n
.
No
n
eth
eless
,
co
n
v
en
tio
n
al
m
eth
o
d
s
o
f
s
o
il a
n
aly
s
is
ar
e
f
r
eq
u
en
tly
c
h
ar
ac
t
er
ized
b
y
h
i
g
h
co
s
ts
an
d
tim
e
d
em
an
d
s
,
as we
ll a
s
a
lack
o
f
ac
ce
s
s
ib
ilit
y
f
o
r
n
u
m
er
o
u
s
f
a
r
m
er
s
—
esp
ec
ially
th
o
s
e
in
is
o
lated
r
eg
io
n
s
.
T
h
e
p
r
o
g
r
ess
m
ad
e
i
n
ar
tific
ial
in
tellig
en
ce
(
AI
)
an
d
d
ee
p
lear
n
in
g
h
as
s
p
ar
k
ed
in
cr
ea
s
ed
cu
r
io
s
ity
ab
o
u
t
u
s
in
g
tech
n
o
lo
g
y
to
en
h
an
ce
a
g
r
icu
ltu
r
al
p
r
ac
tices
[
1
]
.
A
s
ig
n
if
ican
t
ch
alle
n
g
e
f
o
r
c
o
n
tem
p
o
r
ar
y
ag
r
ic
u
ltu
r
e
is
th
at
f
lu
ctu
atio
n
s
i
n
s
o
il
a
n
d
en
v
ir
o
n
m
e
n
tal
co
n
d
itio
n
s
ca
n
lead
to
u
n
p
r
ed
i
ctab
le
cr
o
p
p
e
r
f
o
r
m
an
ce
.
Far
m
er
s
f
r
eq
u
en
tly
d
ep
e
n
d
o
n
tr
a
d
itio
n
al
m
eth
o
d
s
o
r
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
15
,
No
.
6
,
Decem
b
e
r
20
25
:
5
6
6
7
-
5
6
7
8
5668
m
an
u
al
test
in
g
to
ev
alu
ate
s
o
il
s
u
itab
ilit
y
,
wh
ich
ca
n
r
esu
l
t
in
in
ef
f
icien
t
r
eso
u
r
ce
allo
c
atio
n
an
d
r
ed
u
ce
d
p
r
o
d
u
ctiv
ity
[
2
]
.
Fu
r
t
h
er
m
o
r
e
,
th
e
d
ec
is
io
n
-
m
ak
in
g
p
r
o
ce
s
s
is
m
ad
e
m
o
r
e
co
m
p
lex
b
y
c
lim
ate
ch
an
g
e
a
n
d
s
o
il
d
eg
r
ad
atio
n
,
w
h
ich
h
in
d
er
s
th
e
attain
m
en
t
o
f
s
u
s
tain
ab
le
f
ar
m
in
g
t
h
at
also
y
ield
s
h
ig
h
p
r
o
d
u
ctio
n
.
Pre
-
tr
ain
ed
d
ee
p
lea
r
n
in
g
m
o
d
els
lik
e
VGG
an
d
R
esNet
h
av
e
p
r
o
v
en
ef
f
ec
ti
v
e
in
im
ag
e
class
if
icatio
n
an
d
p
atter
n
r
ec
o
g
n
itio
n
,
b
u
t
th
ei
r
d
ir
ec
t
u
s
e
in
ag
r
ic
u
ltu
r
e
n
ec
e
s
s
itate
s
f
in
e
-
tu
n
in
g
to
s
u
it
s
p
ec
if
ic
s
o
il
an
d
cr
o
p
co
n
d
itio
n
s
[
3
]
.
T
h
is
s
tu
d
y
aim
s
to
m
ee
t
th
ese
ch
allen
g
es
b
y
f
o
c
u
s
in
g
o
n
th
e
f
in
e
-
tu
n
in
g
o
f
p
r
e
-
tr
ain
ed
d
ee
p
lear
n
in
g
m
o
d
els
f
o
r
th
e
p
u
r
p
o
s
e
o
f
ac
cu
r
ate
cr
o
p
p
r
ed
ictio
n
b
ased
o
n
s
o
il
co
n
d
itio
n
s
[
4
]
.
T
h
ese
m
o
d
els
ca
n
ev
alu
ate
s
o
il
tex
tu
r
e,
m
o
is
tu
r
e
lev
els,
a
n
d
n
u
tr
ien
t
co
n
ten
t
to
s
u
g
g
est
th
e
m
o
s
t
ap
p
r
o
p
r
iate
c
r
o
p
s
f
o
r
a
s
p
ec
if
ic
ar
ea
b
y
u
s
in
g
s
o
il
im
a
g
es
an
d
r
elate
d
d
ata.
AI
-
d
r
iv
en
s
m
ar
t
a
g
r
i
cu
ltu
r
e
s
o
lu
tio
n
s
ca
n
g
r
ea
tly
im
p
r
o
v
e
d
ec
is
io
n
-
m
ak
in
g
,
m
in
im
ize
tr
ial
-
an
d
-
er
r
o
r
f
a
r
m
in
g
m
eth
o
d
s
,
a
n
d
b
o
o
s
t
o
v
er
all
y
ield
.
T
h
is
r
esear
c
h
s
ee
k
s
to
co
n
n
ec
t
th
e
f
ield
s
o
f
d
ee
p
lear
n
in
g
an
d
p
r
ec
is
io
n
ag
r
icu
ltu
r
e
,
o
f
f
er
i
n
g
f
ar
m
er
s
a
tr
u
s
two
r
th
y
,
d
at
a
-
in
f
o
r
m
e
d
m
eth
o
d
f
o
r
o
p
tim
izin
g
p
r
o
d
u
ctiv
ity
a
n
d
s
u
s
tain
ab
ilit
y
.
T
h
is
r
esear
ch
im
p
r
o
v
es
p
r
ec
is
io
n
f
ar
m
i
n
g
th
r
o
u
g
h
th
e
u
s
e
o
f
Den
s
eNe
t1
2
1
an
d
R
esNet5
0
f
o
r
ac
cu
r
ate
cr
o
p
class
if
icati
o
n
f
r
o
m
s
atellite
im
ag
es,
f
ac
ilit
atin
g
im
p
r
o
v
ed
d
ec
is
io
n
-
m
ak
in
g
an
d
r
eso
u
r
ce
m
an
ag
em
en
t.
Sectio
n
2
o
u
tlin
es
r
elate
d
w
o
r
k
,
Sectio
n
3
o
u
tlin
es
th
e
p
r
o
p
o
s
ed
m
eth
o
d
o
lo
g
y
,
in
clu
d
in
g
d
ata
ac
q
u
is
itio
n
,
p
r
ep
r
o
ce
s
s
in
g
,
m
o
d
el
f
in
e
-
tu
n
in
g
,
a
n
d
ev
alu
ati
o
n
.
Sectio
n
4
d
is
cu
s
s
es
th
e
e
x
p
er
im
en
tal
s
etu
p
,
r
esu
lts
,
an
d
an
aly
s
is
.
Sectio
n
5
co
n
clu
d
es
th
e
s
tu
d
y
with
k
ey
f
in
d
in
g
s
,
lim
itatio
n
s
,
an
d
d
ir
ec
tio
n
s
f
o
r
f
u
tu
r
e
wo
r
k
.
2.
RE
L
AT
E
D
WO
RK
J
asv
an
th
an
d
Fre
d
r
ik
[
5
]
p
r
o
p
o
s
e
d
a
co
n
v
o
l
u
tio
n
al
n
eu
r
al
n
etwo
r
k
(
C
NN
)
-
b
ased
m
eth
o
d
f
o
r
class
if
y
in
g
s
o
il
im
ag
es
an
d
r
ec
o
m
m
e
n
d
in
g
cr
o
p
s
,
th
er
eb
y
im
p
r
o
v
in
g
p
r
ec
is
io
n
ag
r
icu
ltu
r
e.
Usi
n
g
p
r
ep
r
o
ce
s
s
in
g
tech
n
iq
u
es
to
s
tan
d
ar
d
ize
in
p
u
t
d
ata,
th
e
m
o
d
el
is
tr
ain
ed
o
n
an
aly
s
es
d
atas
et
o
f
d
if
f
er
e
n
t
s
o
il
ty
p
es
[
6
]
.
Fo
llo
win
g
class
if
icatio
n
,
th
e
s
y
s
tem
r
ec
o
m
m
en
d
s
ap
p
r
o
p
r
iate
cr
o
p
s
,
o
f
f
er
in
g
an
au
to
m
ated
an
d
ef
f
ec
tiv
e
m
ea
n
s
f
o
r
m
ak
in
g
i
n
f
o
r
m
e
d
ag
r
ic
u
ltu
r
al
d
ec
is
io
n
s
.
T
h
e
p
u
r
p
o
s
e
o
f
r
esear
c
h
[
7
]
is
to
h
elp
f
ar
m
e
r
s
ch
o
o
s
e
ap
p
r
o
p
r
iate
cr
o
p
s
b
y
ex
am
in
in
g
th
e
ch
ar
ac
ter
is
tics
o
f
lan
d
an
d
s
o
il
th
r
o
u
g
h
g
eo
s
p
atial
m
eth
o
d
s
.
T
o
ass
es
s
th
e
ap
p
r
o
p
r
iaten
ess
o
f
cr
o
p
s
,
elem
en
ts
s
u
ch
as
s
o
il
tex
tu
r
e
an
d
m
o
is
tu
r
e
lev
els,
n
u
tr
ien
t
co
n
ten
t,
a
n
d
s
lo
p
e
ar
e
an
aly
s
es.
A
web
-
b
ased
m
o
d
el
th
at
p
r
o
ce
s
s
es
d
y
n
am
ic
d
ata
f
ac
ilit
ates
im
p
r
o
v
ed
p
la
n
n
in
g
an
d
en
h
an
ce
s
th
e
y
ield
p
er
h
ec
tar
e
.
R
ef
er
en
ce
[
8
]
p
u
ts
f
o
r
war
d
a
s
u
p
er
v
is
ed
lear
n
i
n
g
m
o
d
el
b
ased
o
n
d
ec
is
io
n
tr
ee
s
to
im
p
r
o
v
e
t
h
e
ac
cu
r
ac
y
o
f
cr
o
p
y
ield
p
r
ed
ic
tio
n
s
u
s
in
g
s
o
il
m
o
is
tu
r
e
p
ar
am
eter
s
an
d
to
d
ec
r
ea
s
e
er
r
o
r
r
ates.
I
t
ex
am
i
n
es
cu
r
r
en
t
m
ac
h
in
e
lear
n
in
g
(
M
L
)
alg
o
r
ith
m
s
,
elab
o
r
ates
o
n
th
e
p
r
o
p
o
s
ed
ap
p
r
o
ac
h
,
ev
alu
ates
o
u
tco
m
es,
an
d
co
n
s
id
er
s
p
o
te
n
tial
en
h
a
n
ce
m
en
ts
,
p
r
o
v
id
in
g
u
s
ef
u
l
p
e
r
s
p
ec
tiv
es
f
o
r
r
esear
ch
e
r
s
in
a
g
r
icu
ltu
r
al
a
r
tifi
cial
in
tellig
en
ce
(
AI
)
.
Usi
n
g
d
ee
p
lear
n
in
g
,
au
t
h
o
r
s
p
r
o
p
o
s
ed
C
NN
-
b
ased
m
eth
o
d
a
n
aly
s
es
s
o
il
ch
ar
ac
ter
is
tics
[
9
]
an
d
f
o
r
ec
asts
ap
p
r
o
p
r
iate
cr
o
p
s
,
g
u
ar
an
teein
g
a
s
o
lu
tio
n
r
o
o
ted
in
d
ata.
T
h
r
o
u
g
h
co
m
p
r
eh
en
s
iv
e
test
in
g
o
n
ac
tu
al
d
atasets
,
h
ig
h
ac
cu
r
ac
y
an
d
ef
f
icien
c
y
h
av
e
b
ee
n
s
h
o
wn
,
p
r
o
m
o
tin
g
p
r
ec
is
io
n
ag
r
icu
ltu
r
e
f
o
r
im
p
r
o
v
ed
s
o
il
class
if
icatio
n
an
d
cr
o
p
f
o
r
ec
asti
n
g
[
4
]
.
Ah
m
ed
et
a
l
.
[
1
0
]
u
tili
ze
s
m
ac
h
in
e
lea
r
n
in
g
to
f
o
r
ec
ast
s
ig
n
if
ican
t
cr
o
p
p
in
g
p
atter
n
s
in
B
an
g
lad
e
s
h
,
d
r
awin
g
o
n
lan
d
,
s
o
il,
an
d
clim
ate
d
ata
f
r
o
m
5
2
Up
az
ilas
.
Mo
d
els s
u
ch
as k
-
n
ea
r
est
n
eig
h
b
o
r
s
(
KNN)
,
d
e
cisi
o
n
tr
ee
(
DT
)
,
r
an
d
o
m
f
o
r
est
class
if
ier
(
R
FC
)
,
ex
tr
em
e
g
r
ad
ien
t
b
o
o
s
tin
g
(
XGBo
o
s
t)
,
an
d
s
u
p
p
o
r
t
v
ec
t
o
r
m
ac
h
in
e
(
SVM)
ar
e
ca
p
a
b
le
o
f
m
an
a
g
in
g
m
i
x
ed
d
ata
an
d
v
ar
io
u
s
cr
o
p
class
e
s
with
an
ac
cu
r
ac
y
e
x
ce
ed
in
g
9
5
%.
Ad
d
itio
n
ally
,
a
s
y
s
tem
th
at
is
ea
s
y
to
u
s
e
was
cr
ea
ted
f
o
r
s
tr
aig
h
tf
o
r
war
d
p
r
ed
ictio
n
d
e
p
lo
y
m
e
n
t.
Mittal
an
d
B
h
an
ja
[
1
1
]
d
e
v
elo
p
ed
an
ML
m
o
d
el
th
at
r
ec
o
m
m
en
d
s
o
p
tim
al
cr
o
p
s
b
ased
o
n
s
o
il,
clim
ate,
an
d
r
eso
u
r
ce
s
.
Usi
n
g
n
atu
r
al
lan
g
u
a
g
e
p
r
o
ce
s
s
in
g
(
NL
P)
to
ex
tr
ac
t
in
s
ig
h
ts
f
r
o
m
cr
o
p
d
ata,
th
e
m
o
d
el
p
r
e
d
icts
s
u
itab
le
cr
o
p
s
an
d
is
d
ep
lo
y
e
d
as a
web
s
er
v
ice
f
o
r
ea
s
y
ac
ce
s
s
.
Allu
v
ial
s
o
il,
lo
ca
ted
i
n
r
i
v
e
r
p
lain
s
s
u
ch
as
th
o
s
e
o
f
th
e
Gan
g
es,
B
r
ah
m
a
p
u
tr
a,
an
d
I
n
d
u
s
,
is
ex
tr
em
ely
f
e
r
tile
an
d
m
i
n
er
al
-
r
ich
,
m
a
k
in
g
it
p
er
f
ec
t
f
o
r
f
ar
m
in
g
.
I
t
is
co
n
d
u
civ
e
to
th
e
f
ar
m
in
g
o
f
s
tap
le
cr
o
p
s
lik
e
r
ice,
wh
ea
t,
s
u
g
ar
c
an
e,
p
u
ls
es,
an
d
o
ils
ee
d
s
d
u
e
to
its
ex
ce
llen
t
d
r
ain
ag
e
an
d
m
o
is
tu
r
e
r
ete
n
tio
n
ca
p
ab
ilit
ies.
Du
e
to
its
n
u
tr
ien
t
-
r
ich
co
m
p
o
s
itio
n
,
it
en
s
u
r
e
s
h
ig
h
y
ield
s
an
d
is
am
o
n
g
th
e
m
o
s
t
p
r
o
d
u
ctiv
e
s
o
il
ty
p
es
f
o
r
f
ar
m
in
g
[
1
2
]
.
Als
o
r
ef
er
r
ed
to
as
r
eg
u
r
s
o
il,
b
lack
s
o
il
is
v
er
y
f
er
tile
an
d
id
ea
l
f
o
r
g
r
o
win
g
co
tto
n
,
s
o
y
b
ea
n
,
s
u
n
f
lo
wer
,
m
aize
,
an
d
p
u
ls
es.
I
t
is
m
ai
n
ly
lo
ca
ted
i
n
Ma
h
ar
ash
tr
a
,
G
u
jar
at,
an
d
Ma
d
h
y
a
Pra
d
esh
,
an
d
it r
etain
s
m
o
is
tu
r
e
well,
m
ak
in
g
it
s
u
itab
le
f
o
r
d
r
y
lan
d
ag
r
icu
ltu
r
e.
B
lack
s
o
il
,
wh
ich
is
ab
u
n
d
an
t
in
ca
lciu
m
an
d
m
ag
n
esiu
m
,
p
r
o
m
o
tes
n
u
tr
ie
n
t
u
p
ta
k
e
an
d
g
u
ar
an
tees
r
o
b
u
s
t
cr
o
p
g
r
o
wth
.
I
ts
ab
ilit
y
to
s
elf
-
p
lo
w
d
im
in
is
h
es
th
e
n
ec
ess
it
y
f
o
r
r
e
g
u
lar
till
in
g
,
th
u
s
b
o
o
s
tin
g
ag
r
icu
ltu
r
al
p
r
o
d
u
ctiv
ity
[
1
3
]
.
R
ed
s
o
il,
lo
ca
ted
in
ar
ea
s
s
u
ch
as
T
am
il
Nad
u
,
Kar
n
atak
a,
a
n
d
Od
is
h
a,
h
as
g
o
o
d
d
r
ain
ag
e
a
n
d
is
h
i
g
h
in
ir
o
n
co
n
ten
t.
Ho
wev
er
,
its
n
atu
r
al
f
er
tili
ty
is
lo
w,
n
ec
ess
ita
tin
g
th
e
u
s
e
o
f
f
er
tili
ze
r
s
f
o
r
id
ea
l
cr
o
p
d
ev
elo
p
m
e
n
t.
I
t
is
ap
p
r
o
p
r
iate
f
o
r
th
e
c
u
ltiv
atio
n
o
f
g
r
o
u
n
d
n
u
t,
m
illets
,
p
u
l
s
es,
co
tto
n
,
r
ice,
a
n
d
v
ar
io
u
s
v
eg
etab
les.
W
ith
ap
p
r
o
p
r
iate
s
o
il
m
a
n
ag
em
e
n
t
an
d
f
e
r
tili
za
tio
n
,
r
e
d
s
o
il
ca
n
s
u
s
tain
ag
r
icu
ltu
r
e
an
d
im
p
r
o
v
e
cr
o
p
y
ield
[
1
4
]
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8
7
0
8
F
in
e
-
tu
n
in
g
p
r
e
-
tr
a
in
ed
d
ee
p
lea
r
n
in
g
mo
d
els fo
r
cro
p
p
r
ed
ictio
n
u
s
in
g
…
(
P
r
a
ve
e
n
P
a
w
a
s
ka
r
)
5669
W
ater
-
in
ten
s
iv
e
cr
o
p
s
lik
e
p
ad
d
y
,
wh
ea
t,
b
ar
le
y
,
an
d
v
eg
et
ab
les
th
r
iv
e
in
clay
ey
s
o
il
th
at
is
n
u
tr
ien
t
-
r
ich
an
d
r
etain
s
m
o
is
tu
r
e
ef
f
ec
tiv
ely
.
T
h
is
s
o
il
ty
p
e,
m
ain
ly
lo
ca
ted
in
ar
ea
s
s
u
ch
as
As
s
am
an
d
W
est
B
en
g
al,
f
ac
ilit
ates
h
ig
h
-
y
ield
ag
r
icu
ltu
r
e
b
u
t
n
ec
ess
itates
ef
f
ec
tiv
e
d
r
ain
ag
e
m
an
ag
e
m
en
t
to
av
er
t
wate
r
lo
g
g
in
g
.
Du
e
to
its
f
er
tile
ch
ar
ac
ter
is
tics
,
it
is
id
ea
l
f
o
r
s
u
s
tain
ab
le
cr
o
p
p
r
o
d
u
ctio
n
wh
en
p
r
o
p
er
i
r
r
i
g
atio
n
m
eth
o
d
s
ar
e
ap
p
lied
an
d
s
o
il a
er
atio
n
is
s
u
f
f
icien
t
[
1
5
]
.
G
r
o
u
n
d
n
u
t
i
s
a
v
i
t
al
f
o
o
d
a
n
d
o
i
l
s
e
e
d
c
r
o
p
i
n
W
e
s
t
A
f
r
i
ca
,
c
o
n
t
r
i
b
u
t
i
n
g
s
i
g
n
i
f
i
c
a
n
t
l
y
to
f
o
o
d
a
n
d
n
u
t
r
i
t
i
o
n
a
l
s
e
c
u
r
i
t
y
.
T
h
i
s
s
t
u
d
y
a
i
m
e
d
t
o
as
s
es
s
t
h
e
i
m
p
a
ct
o
f
d
i
f
f
e
r
e
n
t
s
o
i
l
t
y
p
e
s
o
n
t
h
e
n
u
t
r
i
t
i
o
n
a
l
q
u
a
li
t
y
o
f
g
r
o
u
n
d
n
u
t
i
n
L
e
b
d
a
v
i
l
l
a
g
e
,
C
e
n
t
r
e
-
N
o
r
t
h
B
u
r
k
i
n
a
F
as
o
.
G
r
o
u
n
d
n
u
t
s
e
e
d
s
(
S
H
4
7
0
P
v
a
r
i
e
t
y
)
w
e
r
e
c
o
l
l
e
c
t
ed
f
r
o
m
f
o
u
r
t
e
e
n
f
a
r
m
e
r
s
ac
r
o
s
s
t
h
r
e
e
s
o
il
t
y
p
es
,
a
n
d
t
h
e
i
r
m
ac
r
o
n
u
t
r
i
e
n
t
a
n
d
m
i
n
e
r
al
c
o
n
t
en
t
s
we
r
e
a
n
a
l
y
ze
d
.
V
a
r
i
a
n
c
e
a
n
a
l
y
s
is
r
e
v
e
a
l
e
d
s
i
g
n
i
f
i
c
a
n
t
d
i
f
f
e
r
e
n
c
es
:
c
l
a
y
s
o
i
l
s
y
i
e
l
d
e
d
s
e
e
d
s
wi
t
h
h
i
g
h
e
r
f
a
t
c
o
n
t
e
n
t
(
4
6
.
6
%
±
6
.
3
g
/
1
0
0
g
d
r
y
m
a
t
t
e
r
)
,
w
h
i
l
e
g
r
a
v
e
l
l
y
s
o
i
l
s
p
r
o
d
u
c
e
d
s
e
e
d
s
r
i
c
h
e
r
i
n
c
a
r
b
o
h
y
d
r
a
t
e
s
(
1
8
.
8
±
1
.
9
g
/
1
0
0
g
d
r
y
m
a
t
t
e
r
)
.
I
r
o
n
c
o
n
t
e
n
t
r
a
n
g
e
d
f
r
o
m
1
.
9
±
0
.
5
m
g
/
1
0
0
g
o
n
s
a
n
d
y
s
o
i
ls
t
o
2
.
4
6
±
0
.
3
9
m
g
/
1
0
0
g
o
n
c
l
a
y
s
o
il
s
[
1
6
]
,
[
1
7
]
.
A
tw
o
-
y
e
a
r
f
i
el
d
s
t
u
d
y
at
H
i
m
a
c
h
a
l
P
r
a
d
es
h
A
g
r
i
c
u
l
t
u
r
a
l
U
n
i
v
e
r
s
it
y
,
Pa
l
a
m
p
u
r
,
as
s
es
s
e
d
t
h
e
i
m
p
a
c
t
o
f
v
e
r
m
i
c
o
m
p
o
s
t
a
n
d
s
p
li
t
-
a
p
p
l
i
ed
n
i
t
r
o
g
e
n
o
n
p
o
l
e
F
r
e
n
c
h
b
e
a
n
.
T
w
e
l
v
e
t
r
e
at
m
e
n
t
c
o
m
b
i
n
a
t
i
o
n
s
we
r
e
t
es
t
e
d
,
v
a
r
y
i
n
g
o
r
g
a
n
i
c
m
a
n
u
r
e
s
,
n
i
t
r
o
g
e
n
l
e
v
e
l
s
,
a
n
d
a
p
p
l
i
c
a
t
i
o
n
m
e
t
h
o
d
s
.
T
h
e
c
o
m
b
i
n
a
t
i
o
n
o
f
v
e
r
m
i
c
o
m
p
o
s
t
w
i
t
h
1
2
5
%
r
e
c
o
m
m
e
n
d
e
d
n
i
t
r
o
g
e
n
a
p
p
l
i
e
d
i
n
s
p
l
i
ts
a
c
h
i
e
v
e
d
t
h
e
h
i
g
h
e
s
t
s
e
e
d
y
i
el
d
o
f
1
0
.
4
3
q
/
h
a
a
n
d
i
m
p
r
o
v
e
d
n
u
t
r
i
e
n
t
u
p
t
a
k
e
.
V
e
r
m
i
c
o
m
p
o
s
t
w
it
h
7
5
%
n
i
t
r
o
g
e
n
a
l
s
o
m
a
t
c
h
e
d
f
u
l
l
-
d
o
s
e
y
i
e
l
d
s
,
e
n
a
b
l
in
g
a
2
5
%
f
e
r
t
i
l
i
z
e
r
s
a
v
i
n
g
.
S
p
li
t
a
p
p
li
c
a
t
i
o
n
at
1
2
5
%
n
i
t
r
o
g
e
n
i
n
c
r
e
as
e
d
y
i
el
d
s
b
y
5
0
%
o
v
e
r
b
a
s
a
l
a
p
p
li
c
a
ti
o
n
,
h
i
g
h
l
i
g
h
t
i
n
g
t
h
at
i
n
t
e
g
r
a
t
i
n
g
v
e
r
m
i
c
o
m
p
o
s
t
w
it
h
s
p
l
i
t
n
i
t
r
o
g
e
n
a
p
p
l
i
c
at
i
o
n
b
o
o
s
ts
p
r
o
d
u
c
t
i
v
i
t
y
a
n
d
s
u
p
p
o
r
ts
s
o
il
h
e
a
l
t
h
[
1
8
]
.
Field
ex
p
er
im
en
ts
wer
e
co
n
d
u
cted
d
u
r
in
g
th
e
s
u
m
m
er
,
k
h
ar
if
,
an
d
r
ab
i
s
ea
s
o
n
s
o
f
2
0
1
6
–
20
1
7
an
d
2017
–
20
1
8
at
AC
&
R
I
,
T
NA
U,
Ma
d
u
r
ai
t
o
ass
ess
th
e
im
p
a
ct
o
f
n
u
tr
ien
t
m
an
ag
em
e
n
t
an
d
s
o
il
am
en
d
m
en
ts
o
n
g
r
o
u
n
d
n
u
t
p
r
o
d
u
ctiv
ity
.
T
h
e
s
tu
d
y
test
ed
th
r
ee
ir
r
ig
atio
n
lev
els
(
I
1
:
0
.
8
I
W
/C
PE,
I
2
:
0
.
6
I
W
/C
PE,
I
3
:
0
.
6
I
W
/C
PE)
an
d
f
o
u
r
n
u
tr
ien
t
m
an
ag
em
en
t
p
r
ac
tices
(
N1
–
N4
)
in
v
o
l
v
in
g
v
a
r
y
in
g
f
er
tili
ze
r
r
ates,
ch
ar
r
ed
r
ic
e
h
u
s
k
,
an
d
Ar
b
u
s
cu
lar
m
y
co
r
r
h
izae
s
ee
d
tr
ea
tm
en
ts
.
R
esu
lts
r
ev
ea
led
th
at
ap
p
ly
in
g
7
5
%
o
f
th
e
r
ec
o
m
m
en
d
ed
f
er
tili
ze
r
alo
n
g
with
5
t
o
f
ch
ar
r
ed
r
ice
h
u
s
k
an
d
Ar
b
u
s
cu
lar
m
y
co
r
r
h
izae
s
ig
n
if
ica
n
tly
en
h
an
ce
d
p
lan
t
g
r
o
wth
,
d
r
y
m
atter
p
r
o
d
u
ctio
n
,
leaf
a
r
ea
in
d
ex
,
SP
AD
v
alu
e,
n
u
tr
ie
n
t
u
p
tak
e,
s
o
il
en
z
y
m
e
ac
tiv
ity
,
an
d
y
ield
s
.
T
h
e
h
ig
h
est
p
o
d
y
ield
s
(
1
7
8
3
,
1
9
3
5
,
a
n
d
1
8
5
4
k
g
/h
a)
a
n
d
h
a
u
lm
y
ield
s
(
4
7
4
3
,
4
2
7
2
,
an
d
4
3
3
8
k
g
/h
a)
wer
e
ac
h
iev
ed
d
u
r
in
g
s
u
m
m
er
,
k
h
a
r
if
,
an
d
r
ab
i
2
0
1
7
,
r
esp
ec
tiv
ely
,
u
n
d
er
t
h
is
tr
ea
tm
en
t
[
1
9
]
.
A
1
0
-
y
ea
r
s
tu
d
y
o
n
o
r
g
an
ic,
in
teg
r
ated
,
an
d
in
o
r
g
a
n
ic
n
u
tr
ien
t
m
an
a
g
em
en
t
s
y
s
te
m
s
ass
es
s
ed
th
eir
im
p
ac
t o
n
s
o
il
m
icr
o
b
io
l
o
g
ical
p
r
o
p
er
ties
.
R
esu
lts
s
h
o
wed
a
C
m
in
er
aliza
tio
n
r
ate
o
f
6
.
8
m
g
/k
g
s
o
il
an
d
a
p
o
ten
tially
m
i
n
er
aliza
b
le
n
itro
g
en
lev
el
o
f
4
1
.
5
m
g
/
k
g
s
o
il.
Ar
g
in
in
e
am
m
o
n
if
icatio
n
an
d
n
it
r
if
icatio
n
ac
tiv
ities
m
ea
s
u
r
ed
0
.
8
8
µg
NH₄⁺
-
N/g
s
o
il/h
an
d
5
6
.
0
µg
NO₃⁻
-
N/g
/
d
ay
,
r
esp
ec
tiv
el
y
.
Mic
r
o
b
ial
b
io
m
ass
C
,
N,
an
d
P
wer
e
3
2
0
,
4
0
,
an
d
1
2
m
g
/k
g
s
o
il.
T
h
e
h
ig
h
est
ac
tiv
ities
o
f
alk
alin
e
p
h
o
s
p
h
atase,
u
r
ea
s
e,
an
d
ce
llu
lase
wer
e
o
b
s
er
v
ed
with
v
er
m
ico
m
p
o
s
t
ap
p
licatio
n
at
1
5
t/h
a
[
2
0
]
.
Acc
o
r
d
in
g
to
Gh
an
i
et
a
l
.
[
2
1
]
,
wh
e
n
it
co
m
es
to
f
o
r
ec
asti
n
g
s
o
il
liq
u
ef
ac
tio
n
,
lo
n
g
s
h
o
r
t
-
ter
m
m
em
o
r
y
n
etwo
r
k
(
L
STM
)
o
u
t
p
er
f
o
r
m
s
C
NN,
XGB,
an
d
C
atB
.
I
ts
ac
cu
r
ac
y
is
0
.
9
6
,
an
d
its
F1
-
s
co
r
e
is
0
.
9
5
.
Ad
d
itio
n
ally
,
it
s
h
o
ws
th
at
th
e
s
o
il
with
th
e
lar
g
est
liq
u
ef
ac
tio
n
r
is
k
is
SM
-
SP
,
p
r
o
v
id
in
g
im
p
o
r
ta
n
t
in
f
o
r
m
atio
n
f
o
r
g
eo
tec
h
n
ical
e
n
g
in
ee
r
s
.
3.
T
H
E
P
RO
P
O
SE
D
M
E
T
H
O
D
3
.
1
.
Da
t
a
s
et
u
s
ed
W
e
h
av
e
co
llected
d
ataset
f
r
o
m
Kag
g
le,
wh
ich
in
clu
d
es
3
,
7
0
2
en
h
a
n
ce
d
p
h
o
to
g
r
ap
h
s
in
a
d
d
itio
n
to
7
8
1
o
r
ig
in
al
d
ir
t
p
h
o
t
o
s
as
s
h
o
wn
in
Fig
u
r
e
1
.
T
wen
t
y
p
er
c
en
t
o
f
th
e
d
ata
is
u
s
ed
f
o
r
test
in
g
d
u
r
in
g
tr
ain
i
n
g
,
an
d
eig
h
ty
p
e
r
ce
n
t
is
u
s
ed
f
o
r
tr
ain
in
g
.
Fo
u
r
s
o
il
g
r
o
u
p
s
ar
e
r
ep
r
esen
ted
in
th
e
d
ataset:
clay
s
o
il
(
9
9
5
p
h
o
to
s
)
,
r
ed
s
o
il (
9
1
0
im
ag
es),
b
lack
s
o
il (
9
8
5
im
a
g
es),
an
d
allu
v
ial
s
o
il (
8
1
2
im
ag
es).
3
.
2
.
P
re
-
pro
ce
s
s
ing
a
nd
da
t
a
a
ug
m
ent
a
t
io
n
T
h
e
p
r
ep
a
r
ato
r
y
s
tag
es
in
th
e
co
d
e
in
clu
d
e
f
etch
in
g
p
h
o
to
s
f
r
o
m
s
u
b
f
o
ld
er
s
,
u
s
in
g
cv
2
.
r
esize
(
)
to
r
esize
th
em
to
2
2
4
×
2
2
4
p
ix
el
s
,
an
d
u
s
in
g
(
)
to
en
c
o
d
e
class
lab
els
b
ef
o
r
e
o
n
e
-
h
o
t
en
c
o
d
in
g
with
_
(
)
.
T
o
i
n
cr
ea
s
e
tr
ain
i
n
g
s
ta
b
ilit
y
,
im
ag
es
a
r
e
n
o
r
m
alize
d
b
y
s
ca
lin
g
p
ix
el
v
alu
es
to
t
h
e
[
0
,
1
]
r
an
g
e.
T
o
en
s
u
r
e
co
r
r
e
ct
m
o
d
el
ev
alu
atio
n
,
_
_
(
)
d
iv
id
es
th
e
d
ataset
in
to
8
0
%
tr
ai
n
in
g
an
d
2
0
%
v
alid
atio
n
.
A
d
r
o
p
o
u
t
lay
er
(
0
.
5
p
r
o
b
ab
ilit
y
)
is
in
c
o
r
p
o
r
ate
d
to
av
o
id
o
v
e
r
f
itti
n
g
.
Alth
o
u
g
h
f
lip
p
i
n
g
an
d
r
o
tatio
n
ar
e
n
o
t u
s
ed
e
x
p
li
citly
,
(
)
ca
n
b
e
u
s
ed
to
in
c
o
r
p
o
r
ate
th
em
.
3
.
3
.
P
er
f
o
r
m
a
nce
m
e
t
rics
T
h
e
ac
cu
r
ac
y
,
p
r
ec
is
io
n
,
r
ec
al
l,
F1
-
s
co
r
e,
an
d
s
u
p
p
o
r
t
m
etr
ics
ar
e
u
s
ed
to
as
s
es
s
th
e
p
er
f
o
r
m
an
ce
o
f
d
ee
p
lear
n
in
g
m
o
d
els.
W
h
ile
th
e
co
n
f
u
s
io
n
m
atr
ix
tr
u
e
p
o
s
itiv
es
(
T
P),
tr
u
e
n
eg
ativ
es
(
T
N)
,
f
alse
p
o
s
itiv
es
(
FP
)
,
an
d
f
alse
n
eg
ativ
es
(
FN
)
aid
in
ev
alu
atin
g
th
e
ef
f
icac
y
o
f
class
if
icatio
n
,
s
u
p
p
o
r
t
s
h
o
ws
th
e
d
is
tr
ib
u
tio
n
o
f
class
es
[
2
2
]
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
15
,
No
.
6
,
Decem
b
e
r
20
25
:
5
6
6
7
-
5
6
7
8
5670
Fig
u
r
e
1
.
Sam
p
le
im
a
g
es f
r
o
m
d
ataset
=
+
(
1
)
=
+
(
2
)
1
−
=
2
+
(
3
)
=
+
+
+
+
(
4
)
3
.
4
.
P
r
o
po
s
ed
m
o
del a
rc
hite
ct
ure
A
s
tr
u
ctu
r
ed
d
ee
p
lear
n
in
g
p
i
p
elin
e
is
u
s
ed
b
y
th
e
s
u
g
g
este
d
s
o
il
ca
teg
o
r
izatio
n
m
o
d
el
as
s
h
o
wn
in
Fig
u
r
e
2
.
T
o
im
p
r
o
v
e
m
o
d
el
p
er
f
o
r
m
an
ce
,
it
s
tar
ts
with
an
in
p
u
t
d
ataset
o
f
s
o
il
im
ag
es
th
at
is
p
r
e
-
p
r
o
ce
s
s
ed
an
d
en
h
an
ce
d
.
Af
ter
th
at,
th
e
d
ataset
is
d
iv
id
ed
in
t
o
2
0
%
test
in
g
an
d
8
0
%
t
r
ain
in
g
.
Usi
n
g
ac
cu
r
ac
y
,
p
r
ec
is
io
n
,
r
ec
all,
an
d
F1
-
s
co
r
e,
s
ev
e
r
al
d
ee
p
lear
n
in
g
m
o
d
els
—
So
il
Net,
VGG1
6
,
R
esNet5
0
V2
,
Den
s
eNe
t2
0
1
,
an
d
Mo
b
ileNetV2
—
ar
e
tr
ain
ed
an
d
ass
ess
ed
.
Allu
v
ial
s
o
il,
b
lac
k
s
o
il,
clay
s
o
il,
a
n
d
r
ed
s
o
il
a
r
e
d
is
tin
g
u
is
h
ed
b
y
th
e
ca
teg
o
r
izatio
n
s
y
s
tem
.
I
n
o
r
d
er
to
m
ax
im
ize
m
o
d
el
p
er
f
o
r
m
an
ce
an
d
g
u
ar
an
tee
ac
cu
r
ate
an
d
r
eliab
le
s
o
il
ca
teg
o
r
izatio
n
,
h
y
p
e
r
p
ar
a
m
eter
ad
ju
s
tm
en
t is in
clu
d
ed
[
2
3
]
.
Fig
u
r
e
2
.
Pro
p
o
s
ed
m
o
d
el
ar
c
h
itectu
r
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8
7
0
8
F
in
e
-
tu
n
in
g
p
r
e
-
tr
a
in
ed
d
ee
p
lea
r
n
in
g
mo
d
els fo
r
cro
p
p
r
ed
ictio
n
u
s
in
g
…
(
P
r
a
ve
e
n
P
a
w
a
s
ka
r
)
5671
3
.
5
.
F
e
a
t
ure
ex
t
r
a
ct
io
n by
S
o
ilNet
CNN
m
o
del
T
h
e
f
o
llo
win
g
is
a
m
ath
em
atica
l
r
ep
r
esen
tatio
n
o
f
th
e
f
ea
tu
r
e
ex
tr
ac
tio
n
p
r
o
ce
d
u
r
e
f
o
r
th
e
s
u
g
g
ested
So
ilNet
m
o
d
el:
a.
C
o
n
v
o
lu
tio
n
al
lay
e
r
o
p
er
atio
n
s
F
l
=
σ
(
W
l
*
X
l
-
1
+
b
l
)
(
5
)
w
h
er
e
F
l
is
th
e
f
ea
tu
r
e
m
ap
at
lay
er
l,
W
l
an
d
b
l
ar
e
th
e
lea
r
n
ed
f
ilter
s
an
d
b
iases
,
X
l
-
1
is
th
e
in
p
u
t
f
r
o
m
th
e
p
r
ev
io
u
s
lay
e
r
,
∗
d
e
n
o
tes th
e
c
o
n
v
o
lu
ti
o
n
o
p
er
atio
n
,
σ
is
th
e
ac
tiv
atio
n
f
u
n
cti
o
n
(
R
eL
U)
.
b.
Max
-
p
o
o
lin
g
f
o
r
d
o
wn
s
am
p
li
n
g
P
l
=
m
ax
(
F
l
)
(
6
)
P
l
r
ep
r
esen
ts
th
e
p
o
o
le
d
f
ea
tu
r
e
m
ap
,
m
a
x
(
)
d
en
o
tes m
ax
p
o
o
lin
g
with
a
2
×
2
f
ilter
.
c.
Flatten
in
g
an
d
f
u
lly
co
n
n
ec
ted
lay
er
s
Z
=
Flatten
(
P
l
)
(
7
)
H
=
σ
(
W
fc
⋅
Z
+
b
fc
)
(
8
)
W
h
er
e
Z
is
th
e
f
latten
ed
v
ec
to
r
,
H
is
th
e
o
u
tp
u
t
o
f
th
e
d
e
n
s
e
lay
er
,
W
fc
an
d
b
fc
ar
e
t
h
e
weig
h
t
m
atr
ix
an
d
b
ias
o
f
th
e
d
e
n
s
e
lay
er
.
d.
So
f
tMa
x
c
lass
if
icatio
n
̂
=
∑
=
1
(
9
)
̂
is
th
e
p
r
o
b
ab
ilit
y
o
f
class
i,
N
is
th
e
to
tal
n
u
m
b
er
o
f
class
es
(
4
s
o
il
ty
p
es
)
,
Z
i
is
t
h
e
ac
ti
v
atio
n
o
u
tp
u
t
f
o
r
class
i.
T
o
g
eth
er
,
th
ese
p
r
o
ce
d
u
r
es
m
ak
e
u
p
th
e
So
ilNet
m
o
d
els
f
ea
tu
r
e
ex
tr
ac
tio
n
p
ip
elin
e,
wh
ich
allo
ws
it
to
id
en
tify
p
atter
n
s
in
s
o
il tex
tu
r
e
f
o
r
ca
teg
o
r
izatio
n
.
4.
E
XP
E
R
I
M
E
N
T
A
L
RE
SUL
T
S
T
h
e
co
d
e
f
o
r
all
e
x
p
er
im
en
ts
was
im
p
lem
en
ted
an
d
e
x
ec
u
te
d
in
Go
o
g
le
C
o
lab
,
u
s
in
g
t
h
e
m
o
s
t
r
ec
en
t
v
er
s
io
n
s
o
f
Ker
as
an
d
T
en
s
o
r
Flo
w.
T
ab
le
1
p
r
o
v
i
d
es
a
d
etailed
ac
co
u
n
t
o
f
th
e
h
ar
d
war
e
s
p
ec
if
icatio
n
s
u
s
ed
d
u
r
in
g
th
e
tr
ials
[
2
4
]
.
T
ab
le
1
.
Har
d
war
e
r
eq
u
ir
em
en
ts
f
o
r
ex
p
er
i
m
en
ts
H
a
r
d
w
a
r
e
c
o
mp
o
n
e
n
t
S
p
e
c
i
f
i
c
a
t
i
o
n
G
P
U
Te
sl
a
T
4
,
H
i
g
h
R
A
M
S
y
st
e
m
R
A
M
5
0
.
9
9
G
B
D
i
sk
s
t
o
r
a
g
e
2
3
8
.
6
8
G
B
T
r
ain
ed
m
o
d
els’
p
e
r
f
o
r
m
an
ce
with
o
u
t
d
ata
au
g
m
e
n
tatio
n
f
o
r
s
o
il
class
if
icatio
n
is
s
u
m
m
ar
ized
in
T
ab
le
2
.
T
h
e
h
ig
h
est
ac
cu
r
ac
y
(
9
2
%)
was
ac
h
iev
ed
b
y
De
n
s
eNe
t2
0
1
,
with
F1
-
s
co
r
es
r
an
g
in
g
f
r
o
m
0
.
9
1
to
0
.
9
3
.
At
9
1
%,
R
esNet5
0
V2
ca
m
e
in
s
ec
o
n
d
p
lace
,
d
em
o
n
s
tr
atin
g
ex
ce
p
tio
n
al
p
e
r
f
o
r
m
an
ce
in
B
lack
So
il
(
F1
-
s
co
r
e:
0
.
9
5
)
.
B
o
th
VGG1
6
an
d
Mo
b
ileNetV2
ac
h
iev
e
d
9
0
%,
with
Mo
b
ileNetV2
d
em
o
n
s
tr
atin
g
a
h
ig
h
p
r
ec
is
io
n
o
f
0
.
9
6
f
o
r
Allu
v
ial
So
il.
W
h
ile
th
e
So
ilNet
m
o
d
el
r
ea
ch
ed
a
n
ac
cu
r
ac
y
o
f
8
3
%,
it
ex
ce
lled
in
R
ed
So
il
(
F1
-
s
co
r
e:
0
.
9
9
)
b
u
t
h
ad
d
if
f
icu
lties
in
C
lay
So
il
(
F1
-
s
co
r
e:
0
.
6
7
)
,
h
ig
h
lig
h
t
in
g
th
e
n
ee
d
f
o
r
en
h
an
ce
m
e
n
ts
.
As
p
er
th
e
p
er
f
o
r
m
an
ce
ass
ess
m
en
t
o
f
tr
ain
ed
m
o
d
els
wi
th
au
g
m
en
tatio
n
as
s
h
o
wn
in
T
ab
le
3
,
Den
s
eNe
t2
0
1
r
ea
c
h
ed
th
e
h
i
g
h
est
ac
cu
r
ac
y
(
1
0
0
%),
f
o
llo
wed
b
y
VGG1
6
an
d
Mo
b
ileNetV2
(
9
9
%)
an
d
R
esNet5
0
V2
(
9
8
%).
T
h
e
p
r
o
p
o
s
ed
So
ilNet
m
o
d
el
ac
h
iev
e
d
an
ac
c
u
r
ac
y
o
f
9
7
%,
d
e
m
o
n
s
tr
atin
g
s
u
p
er
i
o
r
p
er
f
o
r
m
an
ce
in
class
if
y
in
g
b
lack
an
d
clay
s
o
ils
.
Alth
o
u
g
h
Den
s
eNe
t2
0
1
s
h
o
wed
t
h
e
b
est
class
if
icat
io
n
r
esu
lts
,
all
m
o
d
els ex
h
ib
ited
c
o
m
p
etitiv
en
ess
with
s
lig
h
t d
if
f
er
en
ce
s
in
m
etr
ics.
T
h
e
p
r
o
p
o
s
ed
m
o
d
els’
ac
cu
r
a
cy
,
lo
s
s
,
an
d
co
m
p
u
tatio
n
al
p
e
r
f
o
r
m
a
n
ce
wer
e
ex
am
in
e
d
b
o
t
h
with
an
d
with
o
u
t
au
g
m
e
n
tatio
n
in
T
ab
l
e
s
4
an
d
5
.
W
ith
o
u
t
a
u
g
m
en
t
atio
n
,
Den
s
eNe
t2
0
1
an
d
R
esNet5
0
V2
attain
ed
a
f
lawless
tr
ain
in
g
ac
cu
r
ac
y
o
f
1
0
0
%,
y
et
t
h
eir
test
in
g
ac
cu
r
a
cies
f
ell
to
9
1
.
8
1
%
an
d
9
0
.
6
4
%,
r
esp
ec
tiv
ely
.
T
h
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
15
,
No
.
6
,
Decem
b
e
r
20
25
:
5
6
6
7
-
5
6
7
8
5672
tr
ain
in
g
ac
c
u
r
ac
y
o
f
VGG1
6
an
d
Mo
b
ileNetV2
was
s
o
m
e
wh
at
lo
wer
,
at
9
2
.
8
1
%
a
n
d
1
0
0
%,
r
esp
ec
tiv
ely
,
wh
ile
th
e
test
in
g
ac
c
u
r
ac
y
h
o
v
er
ed
ar
o
u
n
d
9
0
%.
T
h
e
p
r
o
p
o
s
ed
So
ilNet
m
o
d
el
d
em
o
n
s
tr
ated
a
tr
ain
i
n
g
ac
cu
r
ac
y
o
f
9
5
.
9
8
%,
b
u
t
it
h
a
d
th
e
lo
west
test
in
g
ac
cu
r
ac
y
(
8
3
.
8
3
%)
an
d
h
i
g
h
er
test
in
g
lo
s
s
.
Den
s
eNe
t2
0
1
ac
h
iev
ed
a
1
0
0
%
test
in
g
ac
cu
r
ac
y
an
d
m
in
im
al
lo
s
s
(
0
.
0
0
3
2
)
with
au
g
m
en
tatio
n
,
o
u
tp
er
f
o
r
m
in
g
all
m
o
d
els
.
W
ith
ac
cu
r
ac
ies
o
f
9
9
.
4
6
%
a
n
d
9
8
.
7
9
%,
VGG1
6
an
d
Mo
b
il
eNe
tV2
wer
e
n
o
t
f
ar
b
eh
in
d
.
W
ith
ac
cu
r
ac
ies
o
f
9
7
.
9
8
%
a
n
d
9
7
.
5
7
%,
r
esp
ec
tiv
ely
,
R
esNet5
0
V2
an
d
th
e
p
r
o
p
o
s
ed
So
ilNet
m
o
d
el
e
x
h
ib
ited
en
h
an
ce
d
g
en
er
aliza
tio
n
in
c
o
m
p
ar
is
o
n
t
o
tr
ain
in
g
with
o
u
t
a
u
g
m
en
tati
o
n
.
O
v
er
all,
au
g
m
en
tatio
n
co
n
s
id
er
ab
ly
im
p
r
o
v
e
d
m
o
d
el
p
er
f
o
r
m
a
n
ce
,
less
en
in
g
o
v
er
f
itti
n
g
a
n
d
in
c
r
ea
s
in
g
test
in
g
ac
cu
r
ac
y
.
T
ab
le
2
.
Per
f
o
r
m
an
ce
ev
alu
ati
o
n
o
f
tr
ain
ed
m
o
d
els with
o
u
t a
u
g
m
en
tatio
n
M
o
d
e
l
s
C
l
a
s
ses
P
r
e
c
i
s
i
o
n
R
e
c
a
l
l
F1
-
sc
o
r
e
A
c
c
u
r
a
c
y
D
e
n
seN
e
t
2
0
1
A
l
l
u
v
i
a
l
s
o
i
l
0
.
9
4
0
.
8
8
0
.
9
1
9
2
%
B
l
a
c
k
S
o
i
l
0
.
8
8
0
.
9
5
0
.
9
1
C
l
a
y
s
o
i
l
0
.
9
5
0
.
9
0
0
.
9
3
R
e
d
s
o
i
l
0
.
9
2
0
.
9
4
0
.
9
3
R
e
sN
e
t
5
0
V
2
A
l
l
u
v
i
a
l
s
o
i
l
0
.
9
1
0
.
8
4
0
.
8
8
9
1
%
B
l
a
c
k
S
o
i
l
0
.
9
1
0
.
9
8
0
.
9
5
C
l
a
y
s
o
i
l
0
.
8
6
0
.
8
6
0
.
8
6
R
e
d
s
o
i
l
0
.
9
3
0
.
9
4
0
.
9
3
V
G
G
1
6
A
l
l
u
v
i
a
l
s
o
i
l
0
.
9
3
0
.
8
6
0
.
8
9
9
0
%
B
l
a
c
k
S
o
i
l
0
.
9
1
0
.
9
5
0
.
9
3
C
l
a
y
s
o
i
l
0
.
9
0
0
.
8
6
0
.
8
8
R
e
d
s
o
i
l
0
.
8
6
0
.
9
2
0
.
8
9
M
o
i
l
e
N
e
t
V
2
A
l
l
u
v
i
a
l
s
o
i
l
0
.
9
6
0
.
8
3
0
.
8
9
9
0
%
B
l
a
c
k
S
o
i
l
0
.
8
9
0
.
9
5
0
.
9
2
C
l
a
y
s
o
i
l
0
.
8
6
0
.
9
0
0
.
8
8
R
e
d
s
o
i
l
0
.
8
7
0
.
9
4
0
.
9
0
P
r
o
p
o
se
d
S
o
i
l
N
e
t
M
o
d
e
l
A
l
l
u
v
i
a
l
s
o
i
l
0
.
8
3
0
.
7
0
0
.
7
6
8
3
%
B
l
a
c
k
S
o
i
l
0
.
9
7
0
.
7
1
0
.
8
2
C
l
a
y
s
o
i
l
0
.
5
1
0
.
9
5
0
.
6
7
R
e
d
s
o
i
l
0
.
9
8
1
.
0
0
0
.
9
9
T
ab
le
3
.
Per
f
o
r
m
an
ce
ev
alu
ati
o
n
o
f
tr
ain
ed
m
o
d
els with
au
g
m
en
tatio
n
M
o
d
e
l
s
C
l
a
s
ses
P
r
e
c
i
s
i
o
n
R
e
c
a
l
l
F1
-
sc
o
r
e
A
c
c
u
r
a
c
y
D
e
n
seN
e
t
2
0
1
A
l
l
u
v
i
a
l
s
o
i
l
1
.
0
0
1
.
0
0
1
.
0
0
1
0
0
%
B
l
a
c
k
S
o
i
l
1
.
0
0
1
.
0
0
1
.
0
0
C
l
a
y
s
o
i
l
1
.
0
0
1
.
0
0
1
.
0
0
R
e
d
s
o
i
l
1
.
0
0
1
.
0
0
1
.
0
0
R
e
sN
e
t
5
0
V
2
A
l
l
u
v
i
a
l
s
o
i
l
1
.
0
0
0
.
9
7
0
.
9
9
9
8
%
B
l
a
c
k
S
o
i
l
0
.
9
6
0
.
9
9
0
.
9
7
C
l
a
y
s
o
i
l
0
.
9
6
0
.
9
6
0
.
9
6
R
e
d
s
o
i
l
1
.
0
0
1
.
0
0
1
.
0
0
V
G
G
1
6
A
l
l
u
v
i
a
l
s
o
i
l
1
.
0
0
0
.
9
8
0
.
9
9
9
9
%
B
l
a
c
k
S
o
i
l
1
.
0
0
1
.
0
0
1
.
0
0
C
l
a
y
s
o
i
l
0
.
9
9
0
.
9
9
0
.
9
9
R
e
d
s
o
i
l
0
.
9
9
1
.
0
0
0
.
9
9
M
o
i
l
e
N
e
t
V
2
A
l
l
u
v
i
a
l
s
o
i
l
0
.
9
9
0
.
9
9
0
.
9
9
9
9
%
B
l
a
c
k
S
o
i
l
0
.
9
8
0
.
9
8
0
.
9
8
C
l
a
y
s
o
i
l
0
.
9
8
0
.
9
8
0
.
9
8
R
e
d
s
o
i
l
1
.
0
0
1
.
0
0
1
.
0
0
P
r
o
p
o
se
d
S
o
i
l
N
e
t
M
o
d
e
l
A
l
l
u
v
i
a
l
s
o
i
l
0
.
9
3
0
.
9
9
0
.
9
6
9
7
%
B
l
a
c
k
S
o
i
l
1
.
0
0
0
.
9
7
0
.
9
8
C
l
a
y
s
o
i
l
0
.
9
7
1
.
0
0
0
.
9
8
R
e
d
s
o
i
l
0
.
9
9
0
.
9
2
0
.
9
5
T
ab
le
4
.
Acc
u
r
ac
y
,
L
o
s
s
an
d
tim
e
co
m
p
u
tin
g
o
f
p
r
o
p
o
s
ed
m
o
d
els with
o
u
t a
u
g
m
e
n
tatio
n
Pre
-
Tr
a
i
n
e
d
mo
d
e
l
Tr
a
i
n
i
n
g
A
c
c
u
r
a
c
y
(
%)
Tr
a
i
n
i
n
g
Lo
ss (
%)
Te
st
i
n
g
A
c
c
u
r
a
c
y
(
%)
Te
st
i
n
g
Lo
ss (
%)
D
e
n
seN
e
t
2
0
1
1
0
0
0
.
0
0
5
1
9
1
.
8
1
0
.
2
1
3
9
R
e
sN
e
t
5
0
V
2
1
0
0
0
.
0
0
9
1
9
0
.
6
4
0
.
3
2
4
1
V
G
G
1
6
9
2
.
8
1
0
.
2
4
2
9
9
0
.
0
6
0
.
2
9
2
6
M
o
i
l
e
N
e
t
V
2
1
0
0
0
.
0
0
5
0
9
0
.
0
6
0
.
3
1
0
6
P
r
o
p
o
se
d
S
o
i
l
N
e
t
M
o
d
e
l
9
5
.
9
8
0
.
0
9
7
0
8
3
.
8
3
0
.
5
0
1
6
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8
7
0
8
F
in
e
-
tu
n
in
g
p
r
e
-
tr
a
in
ed
d
ee
p
lea
r
n
in
g
mo
d
els fo
r
cro
p
p
r
ed
ictio
n
u
s
in
g
…
(
P
r
a
ve
e
n
P
a
w
a
s
ka
r
)
5673
T
ab
le
5
.
Acc
u
r
ac
y
,
l
o
s
s
an
d
tim
e
co
m
p
u
tin
g
o
f
p
r
o
p
o
s
ed
m
o
d
els with
au
g
m
en
tatio
n
Pre
-
Tr
a
i
n
e
d
mo
d
e
l
Tr
a
i
n
i
n
g
A
c
c
u
r
a
c
y
(
%)
Tr
a
i
n
i
n
g
Lo
ss (
%)
Te
st
i
n
g
A
c
c
u
r
a
c
y
(
%)
Te
st
i
n
g
Lo
ss (
%)
D
e
n
seN
e
t
2
0
1
9
9
.
9
8
0
.
0
0
1
2
1
0
0
0
.
0
0
3
2
R
e
sN
e
t
5
0
V
2
9
9
.
9
3
0
.
0
0
2
0
9
7
.
9
8
0
.
0
5
4
0
V
G
G
1
6
9
9
.
7
8
0
.
0
1
2
9
9
9
.
4
6
0
.
0
1
7
9
M
o
i
l
e
N
e
t
V
2
1
0
0
0
.
0
0
3
5
9
8
.
7
9
0
.
0
2
2
4
P
r
o
p
o
se
d
S
o
i
l
N
e
t
M
o
d
e
l
9
9
.
5
5
0
.
0
1
2
6
9
7
.
5
7
0
.
1
9
8
2
All
m
o
d
els’
tr
ain
in
g
an
d
v
ali
d
atio
n
ac
cu
r
ac
y
/lo
s
s
cu
r
v
es
s
h
o
w
a
co
n
s
is
ten
t
r
is
e
in
ac
cu
r
ac
y
o
v
er
ep
o
ch
s
as
lo
s
s
d
ec
r
ea
s
es.
So
i
lNet’
s
ex
ce
p
tio
n
al
p
e
r
f
o
r
m
an
ce
in
s
o
il
class
if
icatio
n
is
co
n
f
ir
m
ed
b
y
its
n
ea
r
-
p
er
f
ec
t
ac
c
u
r
ac
y
.
Hig
h
ac
c
u
r
ac
y
is
also
d
e
m
o
n
s
tr
ated
b
y
R
esNet5
0
V2
,
VGG1
6
,
Den
s
eNe
t2
0
1
,
an
d
Mo
b
ileNetV2
,
with
s
lig
h
t
v
ar
iatio
n
s
in
v
alid
atio
n
lo
s
s
s
u
g
g
esti
n
g
s
o
m
e
v
o
latilit
y
b
u
t
o
v
er
all
g
o
o
d
g
en
er
aliza
tio
n
.
All
m
o
d
els
o
p
er
ate
well,
b
u
t
So
ilNet
is
th
e
m
o
s
t
d
ep
en
d
ab
le
f
o
r
s
o
il
ca
te
g
o
r
izatio
n
s
in
ce
it
p
er
f
o
r
m
s
b
etter
th
an
th
e
o
th
er
s
in
ter
m
s
o
f
ac
cu
r
ac
y
a
n
d
s
tab
ilit
y
.
Fig
u
r
e
3
s
h
o
ws
tr
ain
in
g
,
test
in
g
ac
cu
r
ac
y
an
d
lo
s
s
cu
r
v
es.
Den
s
eNe
t2
0
1
R
esNet5
0
V2
VGG1
6
Mo
b
ileNetV2
P
r
o
p
o
s
ed
So
ilNet
m
o
d
el
Fig
u
r
e
3
.
T
r
ain
in
g
a
n
d
test
in
g
ac
cu
r
ac
y
an
d
lo
s
s
an
aly
s
is
I
n
s
o
il
class
if
icatio
n
,
th
e
c
o
n
f
u
s
io
n
m
atr
ices
o
f
d
if
f
er
en
t
m
o
d
els
s
h
o
w
h
o
w
e
f
f
ec
tiv
e
th
ey
ar
e.
Den
s
eNe
t2
0
1
’
s
r
o
b
u
s
tn
ess
is
d
em
o
n
s
tr
ated
b
y
its
p
er
f
ec
t
cl
ass
if
icatio
n
with
ze
r
o
m
is
class
if
icatio
n
s
ac
r
o
s
s
all
f
o
u
r
s
o
il
ty
p
es
(
Allu
v
ial,
B
la
ck
,
C
lay
,
an
d
R
ed
)
.
Hig
h
ac
cu
r
ac
y
is
d
em
o
n
s
tr
ated
b
y
R
esNet5
0
V2
,
wh
ich
class
if
ies
th
e
m
ajo
r
ity
o
f
s
am
p
les
ac
cu
r
ately
with
f
ew
er
r
o
r
s
.
I
t
p
r
im
ar
ily
co
n
f
u
s
es
clay
s
o
il
with
b
lack
s
o
il
an
d
allu
v
ial
s
o
il
with
o
th
er
ca
teg
o
r
ies.
W
ith
a
f
ew
s
m
all
m
is
class
if
ica
tio
n
s
in
clay
an
d
allu
v
ial
s
o
il
b
u
t
ex
ce
llen
t
o
v
er
all
ac
cu
r
ac
y
,
VGG1
6
p
er
f
o
r
m
s
ad
m
ir
ab
ly
as
well.
W
ith
v
er
y
f
ew
in
co
r
r
ec
t
class
if
icatio
n
s
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
15
,
No
.
6
,
Decem
b
e
r
20
25
:
5
6
6
7
-
5
6
7
8
5674
in
clu
d
in
g
C
lay
an
d
B
lack
So
il,
Mo
b
ileNetV2
co
n
tin
u
e
s
to
ex
h
ib
it
s
tr
o
n
g
class
if
ic
atio
n
p
er
f
o
r
m
an
ce
.
Alth
o
u
g
h
1
3
R
ed
So
il
s
am
p
l
es
wer
e
in
co
r
r
ec
tly
class
if
ied
as
Allu
v
ial
So
il,
th
e
s
u
g
g
es
ted
So
ilNet
m
o
d
el
s
h
o
ws
r
em
ar
k
ab
le
ac
cu
r
ac
y
,
esp
ec
ially
in
Allu
v
ial,
B
lack
,
an
d
C
lay
s
o
ils
.
Nev
er
th
eless
,
So
ilNet
p
er
f
o
r
m
s
b
etter
th
an
o
t
h
er
m
o
d
els,
m
ak
i
n
g
it a
v
er
y
u
s
ef
u
l so
il c
lass
if
icatio
n
as sh
o
wn
in
Fig
u
r
e
4
.
Den
s
eNe
t2
0
1
R
esNet5
0
V2
VGG1
6
Mo
b
ileNetV2
P
r
o
p
o
s
ed
So
ilNet
m
o
d
el
Fig
u
r
e
4
.
C
o
n
f
u
s
io
n
m
atr
i
x
o
f
p
r
o
p
o
s
ed
m
o
d
els
B
y
g
r
ap
h
i
n
g
t
h
e
tr
u
e
p
o
s
itiv
e
r
ate
(
T
PR
)
v
er
s
u
s
th
e
f
alse
p
o
s
itiv
e
r
ate
(
FP
R
)
,
th
e
R
OC
cu
r
v
es
ass
es
s
d
ee
p
lear
n
in
g
m
o
d
els
f
o
r
m
u
lti
-
class
s
o
il
clas
s
if
icatio
n
an
d
d
em
o
n
s
tr
ate
th
eir
ca
p
ac
ity
f
o
r
d
is
cr
im
in
atio
n
.
T
h
e
ar
ea
u
n
d
er
t
h
e
cu
r
v
e
(
AUC)
is
u
s
ed
to
ev
alu
ate
m
o
d
els s
u
ch
as So
ilNet,
R
esNet5
0
V2
,
VG
G1
6
,
Den
s
eNe
t2
0
1
,
an
d
Mo
b
ileNetV2
.
Valu
es
n
ea
r
1
.
0
in
d
icate
g
o
o
d
class
if
icatio
n
p
er
f
o
r
m
a
n
ce
.
R
o
b
u
s
t
g
en
er
aliza
tio
n
is
co
n
f
ir
m
e
d
b
y
h
ig
h
er
AUC
ac
r
o
s
s
all
s
o
il
cla
s
s
e
s
;
So
il
Net
m
o
s
t
lik
ely
ac
h
iev
es
t
h
e
g
r
ea
test
AUC,
d
em
o
n
s
tr
atin
g
its
im
p
r
o
v
ed
ac
cu
r
ac
y
.
R
OC
cu
r
v
e
co
m
p
a
r
is
o
n
aid
s
i
n
id
e
n
tify
in
g
th
e
b
est
ac
cu
r
ate
m
o
d
el
f
o
r
s
o
il c
lass
if
icat
io
n
as sh
o
wn
in
Fig
u
r
e
5
.
T
a
b
le
6
p
r
o
p
o
s
ed
m
o
d
el
co
m
p
ar
is
o
n
with
o
t
h
er
s
tu
d
ies.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8
7
0
8
F
in
e
-
tu
n
in
g
p
r
e
-
tr
a
in
ed
d
ee
p
lea
r
n
in
g
mo
d
els fo
r
cro
p
p
r
ed
ictio
n
u
s
in
g
…
(
P
r
a
ve
e
n
P
a
w
a
s
ka
r
)
5675
Den
s
eNe
t2
0
1
R
esNet5
0
V2
VGG1
6
Mo
b
ileNetV2
P
r
o
p
o
s
ed
So
ilNet
m
o
d
el
Fig
u
r
e
5
.
R
OC
cu
r
v
es o
f
p
r
o
p
o
s
ed
m
o
d
els
T
ab
le
6
.
Pro
p
o
s
ed
m
o
d
el
co
m
p
ar
is
o
n
with
o
th
e
r
s
tu
d
ies
R
e
f
e
r
e
n
c
e
s
Y
e
a
r
D
a
t
a
s
e
t
u
s
e
d
A
c
c
u
r
a
c
y
(
%)
P
r
o
p
o
se
d
w
o
r
k
2
0
2
5
K
a
g
g
l
e
9
7
%
[
1
0
]
2
0
2
2
K
a
g
g
l
e
9
5
%
5.
CO
NCLU
SI
O
N
I
n
o
r
d
er
to
f
ac
ilit
ate
in
f
o
r
m
e
d
cr
o
p
s
elec
tio
n
in
p
r
ec
is
io
n
ag
r
icu
ltu
r
e,
th
is
p
a
p
er
o
f
f
e
r
s
a
th
o
r
o
u
g
h
ass
es
s
m
en
t
o
f
d
ee
p
lear
n
in
g
m
eth
o
d
s
f
o
r
s
o
il
ca
teg
o
r
izatio
n
.
T
h
e
p
er
f
o
r
m
a
n
ce
o
f
f
o
u
r
p
o
p
u
lar
p
r
e
-
tr
ain
ed
m
o
d
els
—
Den
s
eNe
t2
0
1
,
R
esN
et5
0
V2
,
VGG1
6
,
an
d
Mo
b
ile
NetV2
—
was
co
m
p
ar
ed
to
th
e
s
u
g
g
ested
So
ilNet
C
NN
m
o
d
el
u
s
in
g
a
ca
r
ef
u
l
ly
s
elec
ted
d
ataset
o
f
3
,
7
0
2
s
o
il
p
ictu
r
es
(
o
r
ig
in
al
an
d
en
h
an
ce
d
)
.
T
h
ese
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
15
,
No
.
6
,
Decem
b
e
r
20
25
:
5
6
6
7
-
5
6
7
8
5676
d
em
o
n
s
tr
ated
th
e
r
o
b
u
s
tn
ess
o
f
d
ee
p
lear
n
in
g
in
s
o
il
-
ty
p
e
p
r
ed
ictio
n
,
with
Den
s
eNe
t2
0
1
ac
h
iev
in
g
th
e
g
r
ea
test
class
if
icatio
n
ac
cu
r
ac
y
(
1
0
0
%),
f
o
llo
wed
b
y
VGG1
6
an
d
Mo
b
ileNetV2
(
9
9
%),
R
esNet5
0
V2
(
9
8
%),
an
d
th
e
s
u
g
g
ested
So
ilNet
(
9
7
%).
Ou
r
m
o
d
els,
esp
ec
ially
Den
s
eNe
t2
0
1
,
s
h
o
wed
b
etter
g
en
er
aliza
tio
n
a
n
d
class
if
icatio
n
p
r
ec
is
io
n
th
an
p
r
ev
io
u
s
s
tu
d
ies
th
at
in
d
icate
d
a
m
a
x
im
u
m
o
f
9
5
%
ac
cu
r
a
cy
,
esp
ec
ially
af
ter
ad
d
in
g
d
ata
au
g
m
en
tatio
n
.
No
tab
ly
,
th
e
So
ilNet
m
o
d
el
d
em
o
n
s
tr
ated
e
x
ce
p
tio
n
al
class
-
s
p
ec
if
ic
p
r
ec
is
io
n
,
esp
ec
ially
f
o
r
r
ed
an
d
cla
y
s
o
ils
,
h
ig
h
lig
h
tin
g
its
p
o
ten
tial
in
s
p
ec
ialized
clas
s
if
icatio
n
task
s
,
alth
o
u
g
h
m
ar
g
in
ally
lag
g
i
n
g
b
e
h
in
d
in
ag
g
r
eg
ate
p
e
r
f
o
r
m
an
ce
.
T
h
i
s
s
tu
d
y
h
as
s
ig
n
if
ican
t
r
am
if
icatio
n
s
f
o
r
s
m
ar
t
ag
r
icu
ltu
r
e
s
in
ce
it
ca
n
r
ep
lace
lab
o
r
-
in
ten
s
iv
e
m
an
u
al
s
o
il
t
esti
n
g
with
au
to
m
ated
s
o
il
cla
s
s
if
icatio
n
b
ased
o
n
im
ag
e
d
ata,
allo
win
g
f
o
r
lo
ca
t
io
n
-
awa
r
e
cr
o
p
s
u
g
g
esti
o
n
s
in
r
ea
l
tim
e.
Ou
r
r
esu
lt
s
h
ig
h
lig
h
t
h
o
w
im
p
o
r
tan
t
it
is
to
in
co
r
p
o
r
ate
d
ee
p
lea
r
n
in
g
in
to
ag
r
icu
ltu
r
al
s
y
s
tem
s
in
o
r
d
er
t
o
ac
h
iev
e
h
ig
h
-
y
ield
,
s
u
s
tain
ab
le
f
ar
m
in
g
.
W
e
in
ten
d
to
b
u
ild
o
n
th
is
r
esear
ch
in
th
e
f
u
tu
r
e
b
y
a
d
d
i
n
g
m
u
ltimo
d
al
s
o
il
ch
ar
ac
ter
i
s
tics
(
s
u
ch
as
p
H,
m
o
is
tu
r
e,
an
d
n
u
tr
ien
t
co
n
ten
t
)
,
r
ef
in
in
g
m
o
d
els
f
o
r
m
o
b
ile
r
ea
l
-
tim
e
ap
p
licatio
n
s
,
an
d
co
n
f
ir
m
in
g
r
esu
lts
o
n
s
o
il
s
am
p
les
at
th
e
f
ield
lev
el.
T
h
ese
d
ev
elo
p
m
en
ts
will
im
p
r
o
v
e
th
e
u
s
e
o
f
AI
-
p
o
wer
ed
p
r
ec
is
io
n
ag
r
icu
ltu
r
e
in
s
tr
u
m
en
ts
in
a
v
ar
iety
o
f
e
n
v
ir
o
n
m
en
tal
s
ettin
g
s
.
F
UNDING
I
NF
O
R
M
A
T
I
O
N
T
h
is
r
esear
ch
d
id
n
o
t
r
ec
eiv
e
an
y
s
p
ec
if
ic
g
r
an
t
f
r
o
m
f
u
n
d
i
n
g
ag
en
cies
in
th
e
p
u
b
lic,
co
m
m
er
cial,
o
r
not
-
f
o
r
-
p
r
o
f
it secto
r
s
.
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
Pra
v
ee
n
Pawask
ar
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
Yo
g
is
h
H
K
✓
✓
✓
✓
✓
Pak
r
u
d
d
in
B
✓
✓
✓
✓
✓
✓
✓
Dee
p
a
Yo
g
is
h
✓
✓
✓
✓
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
T
h
e
au
t
h
o
r
s
d
ec
lar
e
th
at
th
e
y
h
av
e
n
o
k
n
o
wn
c
o
m
p
etin
g
f
in
an
cial
in
ter
ests
o
r
p
er
s
o
n
al
r
el
atio
n
s
h
ip
s
th
at
co
u
ld
h
av
e
ap
p
ea
r
ed
t
o
in
f
lu
en
ce
th
e
wo
r
k
r
e
p
o
r
te
d
in
t
h
is
p
ap
er
.
DATA AV
AI
L
AB
I
L
I
T
Y
T
h
e
d
ata
th
at
s
u
p
p
o
r
t th
e
f
in
d
i
n
g
s
o
f
th
is
s
tu
d
y
a
r
e
o
p
en
ly
a
v
ailab
le
in
Kag
g
le
[
2
5
]
.
RE
F
E
R
E
NC
E
S
[
1
]
J.
P
.
S
a
r
v
a
i
y
a
,
A
.
P
.
C
h
a
u
d
h
a
r
i
,
a
n
d
J.
P
.
V
e
r
m
a
,
“
M
o
n
i
t
o
r
i
n
g
a
g
r
i
c
u
l
t
u
r
a
l
e
s
se
n
t
i
a
l
s
,
”
i
n
A
p
p
l
i
c
a
t
i
o
n
o
f
M
a
c
h
i
n
e
L
e
a
r
n
i
n
g
i
n
Ag
ri
c
u
l
t
u
r
e
,
El
se
v
i
e
r
,
2
0
2
2
,
p
p
.
2
9
–
4
6
.
d
o
i
:
1
0
.
1
0
1
6
/
B
9
7
8
-
0
-
323
-
9
0
5
5
0
-
3
.
0
0
0
0
4
-
7.
[
2
]
V
.
K
a
k
a
n
i
,
V
.
H
.
N
g
u
y
e
n
,
B
.
P
.
K
u
mar,
H
.
K
i
m,
a
n
d
V
.
R
.
P
a
s
u
p
u
l
e
t
i
,
“
A
c
r
i
t
i
c
a
l
r
e
v
i
e
w
o
n
c
o
m
p
u
t
e
r
v
i
s
i
o
n
a
n
d
a
r
t
i
f
i
c
i
a
l
i
n
t
e
l
l
i
g
e
n
c
e
i
n
f
o
o
d
i
n
d
u
s
t
r
y
,
”
J
o
u
r
n
a
l
o
f
A
g
r
i
c
u
l
t
u
r
e
a
n
d
Fo
o
d
Re
se
a
rc
h
,
v
o
l
.
2
,
p
.
1
0
0
0
3
3
,
D
e
c
.
2
0
2
0
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
j
a
f
r
.
2
0
2
0
.
1
0
0
0
3
3
.
[
3
]
M
.
P
e
n
g
e
t
a
l
.
,
“
C
r
o
p
m
o
n
i
t
o
r
i
n
g
u
s
i
n
g
r
e
m
o
t
e
s
e
n
s
i
n
g
l
a
n
d
u
se
a
n
d
l
a
n
d
c
h
a
n
g
e
d
a
t
a
:
C
o
m
p
a
r
a
t
i
v
e
a
n
a
l
y
si
s
o
f
d
e
e
p
l
e
a
r
n
i
n
g
met
h
o
d
s
u
s
i
n
g
p
r
e
-
t
r
a
i
n
e
d
C
N
N
m
o
d
e
l
s,”
Bi
g
D
a
t
a
Re
se
a
r
c
h
,
v
o
l
.
3
6
,
p
.
1
0
0
4
4
8
,
M
a
y
2
0
2
4
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
b
d
r
.
2
0
2
4
.
1
0
0
4
4
8
.
[
4
]
G
.
R
.
S
r
i
a
n
d
A
.
V
i
k
r
a
m,
“
S
o
i
l
t
y
p
e
d
e
t
e
c
t
i
o
n
a
n
d
c
r
o
p
p
r
e
d
i
c
t
i
o
n
u
si
n
g
d
e
e
p
l
e
a
r
n
i
n
g
me
t
h
o
d
C
N
N
,
”
i
n
2
0
2
4
I
EEE
3
rd
W
o
rl
d
C
o
n
f
e
re
n
c
e
o
n
A
p
p
l
i
e
d
I
n
t
e
l
l
i
g
e
n
c
e
a
n
d
C
o
m
p
u
t
i
n
g
(
AI
C
)
,
Ju
l
.
2
0
2
4
,
p
p
.
1
0
9
0
–
1
0
9
4
.
d
o
i
:
1
0
.
1
1
0
9
/
A
I
C
6
1
6
6
8
.
2
0
2
4
.
1
0
7
3
0
8
5
8
.
[
5
]
S
.
M
.
J
a
sv
a
n
t
h
a
n
d
E.
J.
T.
F
r
e
d
r
i
k
,
“
S
o
i
l
i
m
a
g
e
c
l
a
ss
i
f
i
c
a
t
i
o
n
a
n
d
c
r
o
p
r
e
c
o
mm
e
n
d
a
t
i
o
n
u
s
i
n
g
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
u
r
a
l
n
e
t
w
o
r
k
s,
”
i
n
2
0
2
4
2
n
d
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
r
e
n
c
e
o
n
C
o
m
p
u
t
i
n
g
a
n
d
D
a
t
a
An
a
l
y
t
i
c
s
(
I
C
C
D
A)
,
N
o
v
.
2
0
2
4
,
p
p
.
1
–
6
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
C
D
A
6
4
8
8
7
.
2
0
2
4
.
1
0
8
6
7
3
2
5
.
[
6
]
D
.
N
.
K
i
r
a
n
P
a
n
d
i
r
i
,
R
.
M
u
r
u
g
a
n
,
a
n
d
T.
G
o
e
l
,
“
S
mar
t
s
o
i
l
i
m
a
g
e
c
l
a
ss
i
f
i
c
a
t
i
o
n
s
y
s
t
e
m
u
si
n
g
l
i
g
h
t
w
e
i
g
h
t
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
u
r
a
l
n
e
t
w
o
r
k
,
”
Ex
p
e
r
t
S
y
s
t
e
m
s
w
i
t
h
A
p
p
l
i
c
a
t
i
o
n
s
,
v
o
l
.
2
3
8
,
p
.
1
2
2
1
8
5
,
M
a
r
.
2
0
2
4
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
e
sw
a
.
2
0
2
3
.
1
2
2
1
8
5
.
Evaluation Warning : The document was created with Spire.PDF for Python.