I
AE
S In
t
er
na
t
io
na
l J
o
urna
l o
f
Art
if
icia
l In
t
ellig
ence
(
I
J
-
AI
)
Vo
l.
1
5
,
No
.
1
,
Feb
r
u
ar
y
2
0
2
6
,
p
p
.
707
~
724
I
SS
N:
2
2
5
2
-
8
9
3
8
,
DOI
: 1
0
.
1
1
5
9
1
/ijai.v
15
.i
1
.
p
p
7
0
7
-
7
2
4
707
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
a
i
.
ia
esco
r
e.
co
m
Autoenco
der
and
G
AN
-
a
ided pla
n
t
disea
se detec
tion i
n rice
a
nd
co
tt
o
n via
hybrid
feature
extractio
n
and decisio
n
tree
cla
ss
ificatio
n
A
na
nd
ra
dd
i N
a
d
uv
ina
m
a
ni
1
,
J
a
y
s
hri R
ud
a
g
i
2
,
M
a
llik
a
rj
un
Ana
nd
ha
lli
3
1
D
e
p
a
r
t
me
n
t
o
f
El
e
c
t
r
o
n
i
c
s a
n
d
C
o
m
mu
n
i
c
a
t
i
o
n
En
g
i
n
e
e
r
i
n
g
,
S
.
G
.
B
a
l
e
k
u
n
d
r
i
I
n
st
i
t
u
t
e
o
f
Te
c
h
n
o
l
o
g
y
,
A
f
f
i
l
i
a
t
e
d
t
o
V
i
s
v
e
s
v
a
r
a
y
a
Te
c
h
n
o
l
o
g
i
c
a
l
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
El
e
c
t
r
o
n
i
c
s a
n
d
C
o
m
mu
n
i
c
a
t
i
o
n
En
g
i
n
e
e
r
i
n
g
,
J
a
i
n
C
o
l
l
e
g
e
o
f
E
n
g
i
n
e
e
r
i
n
g
,
B
e
l
a
g
a
v
i
,
A
f
f
i
l
i
a
t
e
d
t
o
V
i
s
v
e
sv
a
r
a
y
a
Te
c
h
n
o
l
o
g
i
c
a
l
U
n
i
v
e
r
si
t
y
,
B
e
l
a
g
a
v
i
,
I
n
d
i
a
3
D
e
p
a
r
t
me
n
t
o
f
El
e
c
t
r
o
n
i
c
s a
n
d
C
o
m
mu
n
i
c
a
t
i
o
n
En
g
i
n
e
e
r
i
n
g
,
C
e
n
t
r
a
l
U
n
i
v
e
r
si
t
y
o
f
K
a
r
n
a
t
a
k
a
,
K
a
l
a
b
u
r
a
g
i
,
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
Sep
1
,
2
0
2
4
R
ev
is
ed
No
v
1
3
,
2
0
2
5
Acc
ep
ted
J
an
2
2
,
2
0
2
6
In
a
g
ricu
l
tu
re
,
c
r
o
p
d
ise
a
se
s
c
a
u
se
d
b
y
p
a
th
o
g
e
n
s,
i
n
c
lu
d
in
g
b
a
c
teria
,
v
iru
se
s,
a
n
d
fu
n
g
i,
p
o
se
a
sig
n
ifi
c
a
n
t
th
re
a
t
t
o
t
h
e
e
ffe
c
ti
v
e
n
e
ss
o
f
a
g
ricu
lt
u
ra
l
p
r
o
d
u
c
ti
v
it
y
.
S
o
m
e
m
a
jo
r
c
ro
p
s
i
n
I
n
d
ia
su
c
h
a
s
rice
a
n
d
c
o
t
to
n
a
re
a
d
v
e
rse
ly
imp
a
c
ted
,
lea
d
in
g
to
e
c
o
n
o
m
ic
lo
ss
a
n
d
l
o
ss
o
f
p
ro
d
u
c
ti
o
n
.
Ti
m
e
ly
in
terv
e
n
ti
o
n
a
n
d
s
u
sta
in
a
b
le
a
g
ricu
lt
u
re
d
e
p
e
n
d
o
n
p
r
o
p
e
r
a
n
d
e
a
rly
id
e
n
ti
fica
ti
o
n
o
f
d
ise
a
se
s.
In
th
is
p
a
p
e
r,
we
p
ro
p
o
se
a
n
o
v
e
l
p
lan
t
d
ise
a
se
d
e
tec
ti
o
n
fra
m
e
wo
rk
th
a
t
in
te
g
ra
tes
g
e
n
e
ra
ti
v
e
a
d
v
e
rsa
rial
n
e
two
r
k
(G
AN
)
-
b
a
se
d
ima
g
e
d
e
n
o
isi
n
g
wit
h
fe
a
tu
re
e
x
trac
ti
o
n
a
n
d
d
e
c
isio
n
t
re
e
(DT)
c
las
sifica
ti
o
n
.
Th
e
G
AN
m
o
d
u
le
e
ffe
c
ti
v
e
ly
re
m
o
v
e
s
n
o
ise
fro
m
a
g
ricu
lt
u
ra
l
ima
g
e
s,
e
n
h
a
n
c
in
g
q
u
a
li
t
y
a
n
d
sta
b
il
i
ty
u
n
d
e
r
c
h
a
ll
e
n
g
i
n
g
ima
g
in
g
c
o
n
d
i
ti
o
n
s.
F
o
ll
o
win
g
d
e
n
o
isin
g
,
a
c
o
m
b
i
n
a
ti
o
n
o
f
c
o
l
o
r,
te
x
tu
re
,
a
n
d
g
ra
d
ien
t
fe
a
tu
re
s
is
e
x
trac
ted
to
o
b
tain
rich
a
n
d
d
isc
rimin
a
ti
v
e
p
a
tt
e
rn
s,
wh
ich
a
re
th
e
n
u
se
d
to
t
ra
in
a
DT
c
las
sifier
fo
r
d
ise
a
se
id
e
n
ti
fica
ti
o
n
.
Ex
p
e
rime
n
ts
a
re
c
o
n
d
u
c
ted
o
n
b
e
n
c
h
m
a
rk
d
a
tas
e
ts
c
o
m
p
risin
g
rice
a
n
d
c
o
tt
o
n
lea
f
ima
g
e
s.
Th
e
p
ro
p
o
s
e
d
sy
ste
m
a
c
h
iev
e
s
su
p
e
rio
r
p
e
r
fo
rm
a
n
c
e
,
with
9
8
.
7
0
%
a
c
c
u
ra
c
y
,
9
8
.
2
0
%
p
re
c
isio
n
,
9
7
.
2
2
%
re
c
a
ll
,
a
n
d
9
8
.
5
0
%
F
1
-
sc
o
re
,
o
u
tp
e
rf
o
rm
in
g
e
x
ist
in
g
m
e
th
o
d
s.
T
h
e
se
re
su
lt
s
d
e
m
o
n
stra
t
e
th
a
t
th
e
G
A
N
-
b
a
se
d
d
e
n
o
isi
n
g
a
p
p
r
o
a
c
h
,
c
o
m
b
in
e
d
wit
h
trad
i
ti
o
n
a
l
fe
a
t
u
re
-
b
a
se
d
c
las
sifica
ti
o
n
,
o
ffe
rs
a
r
o
b
u
st,
e
f
ficie
n
t,
a
n
d
p
ra
c
ti
c
a
l
so
l
u
ti
o
n
fo
r
m
o
d
e
rn
a
g
ricu
lt
u
ra
l
d
ise
a
se
m
o
n
it
o
rin
g
sy
ste
m
s.
K
ey
w
o
r
d
s
:
Dec
is
io
n
tr
ee
Gen
er
ati
v
e
ad
v
er
s
ar
ial
n
etw
o
r
k
s
I
m
ag
e
f
ea
tu
r
e
ex
tr
ac
tio
n
Ma
ch
in
e
lear
n
in
g
Plan
t d
is
ea
s
e
d
etec
tio
n
R
ice
an
d
co
tto
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
:
An
an
d
r
ad
d
i N
ad
u
v
in
am
an
i
Dep
ar
tm
en
t o
f
E
lectr
o
n
ics an
d
C
o
m
m
u
n
icatio
n
E
n
g
in
ee
r
i
n
g
,
S.
G.
B
alek
u
n
d
r
i I
n
s
titu
te
o
f
T
ec
h
n
o
lo
g
y
Af
f
iliated
to
Vis
v
esv
ar
ay
a
T
e
ch
n
o
lo
g
ical
Un
iv
er
s
ity
B
elag
av
i,
I
n
d
ia
E
m
ail: a
n
an
d
r
e
d
d
i0
2
@
g
m
ail.
c
o
m
1.
I
NT
RO
D
UCT
I
O
N
Ag
r
icu
ltu
r
e
h
as
a
m
ajo
r
i
m
p
ac
t
in
en
s
u
r
in
g
f
o
o
d
s
e
cu
r
ity
,
p
o
v
e
r
ty
alle
v
iatio
n
,
an
d
o
v
er
all
d
ev
elo
p
m
e
n
t
[
1
]
.
I
n
th
e
co
n
tex
t
o
f
I
n
d
ia,
it
h
o
ld
s
s
ig
n
if
ican
t
im
p
o
r
tan
ce
i
n
th
e
co
u
n
tr
y
’
s
e
co
n
o
m
y
,
p
r
o
v
id
in
g
em
p
lo
y
m
e
n
t
f
o
r
a
lar
g
e
p
o
r
ti
o
n
o
f
t
h
e
p
o
p
u
latio
n
a
n
d
c
o
n
t
r
ib
u
tin
g
a
p
p
r
o
x
im
ately
1
5
%
to
th
e
n
atio
n
’
s
g
r
o
s
s
d
o
m
esti
c
p
r
o
d
u
ct
(
GDP
)
.
I
n
d
ia
is
o
n
e
o
f
th
e
wo
r
ld
’
s
lead
in
g
p
r
o
d
u
ce
r
s
o
f
v
ar
io
u
s
cr
o
p
s
,
in
clu
d
in
g
r
ice,
wh
ea
t,
s
u
g
ar
ca
n
e,
co
tto
n
,
an
d
p
u
ls
es,
f
u
r
th
e
r
em
p
h
asizin
g
i
t
i
s
im
p
o
r
tan
ce
o
n
th
e
g
lo
b
al
s
tag
e.
T
h
e
g
l
o
b
al
p
o
p
u
latio
n
is
ex
p
ec
ted
to
r
ea
c
h
9
.
7
b
illi
o
n
b
y
2
0
5
0
an
d
1
1
.
2
b
illi
o
n
b
y
t
h
e
en
d
o
f
th
is
ce
n
tu
r
y
,
lead
i
n
g
to
a
1
.
6
%
an
n
u
al
in
cr
ea
s
e
in
th
e
E
ar
th
'
s
p
o
p
u
latio
n
an
d
a
co
r
r
es
p
o
n
d
in
g
r
is
e
in
d
em
a
n
d
f
o
r
p
lan
t
p
r
o
d
u
cts
[
2
]
,
[
3
]
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
1
5
,
No
.
1
,
Feb
r
u
ar
y
2
0
2
6
:
707
-
7
2
4
708
Saf
eg
u
ar
d
in
g
cr
o
p
s
ag
ain
s
t
d
is
ea
s
es
i
s
th
er
ef
o
r
e
ess
en
tial
t
o
m
ee
t
th
e
g
r
o
win
g
d
em
an
d
f
o
r
b
o
th
th
e
q
u
ality
an
d
q
u
an
tity
o
f
f
o
o
d
.
Plan
t
d
is
ea
s
es
alo
n
e
in
c
u
r
s
ig
n
i
f
ican
t
ec
o
n
o
m
ic
lo
s
s
es,
am
o
u
n
tin
g
to
ar
o
u
n
d
US$
2
2
0
b
illi
o
n
a
n
n
u
all
y
wo
r
l
d
wid
e
[
4
]
.
I
n
th
e
co
n
te
x
t
o
f
I
n
d
ia,
th
e
I
n
d
ian
C
o
u
n
cil
o
f
Ag
r
icu
ltu
r
al
R
esear
ch
r
ep
o
r
ted
th
at
m
o
r
e
t
h
an
3
5
%
o
f
cr
o
p
p
r
o
d
u
ctio
n
is
lo
s
t
ea
c
h
y
ea
r
d
u
e
t
o
p
ests
an
d
d
is
ea
s
es
[
5
]
.
T
h
ese
o
u
tb
r
ea
k
s
n
o
t
o
n
ly
t
h
r
ea
ten
f
o
o
d
s
ec
u
r
ity
b
u
t
also
h
av
e
f
ar
-
r
ea
c
h
in
g
ec
o
n
o
m
ic,
s
o
cial,
an
d
en
v
ir
o
n
m
e
n
tal
co
n
s
eq
u
en
ce
s
.
Pr
o
tectin
g
cr
o
p
s
f
r
o
m
th
ese
th
r
ea
ts
is
v
ital
n
o
t o
n
ly
f
o
r
th
e
ag
r
icu
ltu
r
al
s
ec
to
r
b
u
t a
ls
o
f
o
r
th
e
o
v
er
all
well
-
b
ein
g
o
f
co
m
m
u
n
ities
an
d
th
e
s
u
s
tain
ab
ilit
y
o
f
th
e
en
v
ir
o
n
m
en
t.
Plan
t
d
is
ea
s
es
ar
e
co
n
s
id
er
ed
o
n
e
o
f
th
e
m
aj
o
r
co
n
ce
r
n
s
in
th
e
ag
r
icu
ltu
r
al
d
o
m
ain
b
ec
a
u
s
e
t
h
ey
lead
to
a
d
ec
lin
e
in
cr
o
p
q
u
ality
an
d
a
r
e
d
u
ctio
n
in
o
v
e
r
all
p
r
o
d
u
ctio
n
.
T
h
e
im
p
ac
ts
o
f
th
ese
d
is
ea
s
es
r
an
g
e
f
r
o
m
m
in
o
r
s
y
m
p
to
m
s
ca
u
s
i
n
g
m
in
im
al
d
am
a
g
e
to
s
ev
er
e
o
u
tb
r
ea
k
s
t
h
at
m
ay
af
f
ec
t
en
tire
r
eg
io
n
s
o
f
c
u
ltiv
ated
lan
d
.
T
h
is
d
am
a
g
e
r
esu
lt
s
in
s
u
b
s
tan
tial
f
in
an
cial
lo
s
s
es
an
d
s
ig
n
if
ican
tl
y
af
f
ec
ts
th
e
ec
o
n
o
m
y
,
p
ar
ticu
la
r
ly
in
d
e
v
elo
p
in
g
n
atio
n
s
th
at
r
ely
o
n
a
s
in
g
le
cr
o
p
o
r
ju
s
t
a
f
ew
cr
o
p
s
.
E
f
f
o
r
ts
to
p
r
ev
en
t
m
aj
o
r
ag
r
icu
ltu
r
al
lo
s
s
es
h
av
e
led
to
th
e
d
ev
el
o
p
m
en
t
o
f
v
ar
io
u
s
d
iag
n
o
s
tic
m
eth
o
d
s
f
o
r
p
lan
t
d
is
ea
s
es.
Mo
lecu
lar
b
io
lo
g
y
a
n
d
im
m
u
n
o
lo
g
y
tech
n
iq
u
es
o
f
f
er
ac
cu
r
ate
d
etec
tio
n
o
f
d
is
e
ase
-
ca
u
s
in
g
ag
en
ts
,
b
u
t
th
ese
m
eth
o
d
s
o
f
ten
r
e
q
u
ir
e
s
p
ec
ialized
k
n
o
wled
g
e
a
n
d
s
ig
n
if
ican
t
f
in
a
n
cial
r
eso
u
r
ce
s
,
m
ak
in
g
th
e
m
in
ac
ce
s
s
ib
le
to
m
an
y
f
ar
m
e
r
s
[
6
]
.
No
tab
ly
,
th
e
m
ajo
r
ity
o
f
th
e
wo
r
ld
’
s
f
a
r
m
s
ar
e
s
m
al
l,
f
am
ily
-
o
p
er
ated
v
en
tu
r
es
in
d
ev
elo
p
in
g
co
u
n
t
r
ies,
p
r
o
d
u
cin
g
f
o
o
d
f
o
r
a
s
ig
n
if
ican
t
p
o
r
tio
n
o
f
th
e
g
lo
b
al
p
o
p
u
latio
n
.
T
h
u
s
,
m
ak
in
g
s
u
ch
m
eth
o
d
s
ac
ce
s
s
i
b
le
to
th
ese
f
ar
m
e
r
s
r
em
ain
s
a
ch
allen
g
in
g
is
s
u
e.
R
i
c
e
r
e
m
a
in
s
i
n
t
h
e
t
o
p
l
i
s
t
o
f
m
o
s
t
w
i
d
e
ly
co
n
s
u
m
ed
f
o
o
d
s
g
lo
b
a
l
ly
,
w
i
th
a
t
o
t
a
l
c
o
n
s
u
m
p
t
io
n
o
f
4
9
3
.
1
3
m
i
l
l
i
o
n
m
e
t
r
ic
to
n
s
i
n
2
0
1
9
-
2
0
2
0
a
n
d
4
8
6
.
6
2
m
i
l
l
io
n
m
e
tr
i
c
to
n
s
i
n
2
0
1
8
-
2
0
1
9
[
7
]
.
T
h
e
s
e
f
ig
u
r
e
s
r
e
f
l
e
c
t
a
g
r
o
w
t
h
i
n
r
ic
e
co
n
s
u
m
p
t
i
o
n
o
v
er
th
e
y
e
ar
s
.
A
s
c
o
n
s
u
m
p
t
io
n
i
n
c
r
e
a
s
e
s
,
i
t
i
s
e
x
p
e
c
t
ed
th
a
t
p
r
o
d
u
c
t
i
o
n
r
a
t
e
s
w
i
l
l
k
ee
p
p
ac
e
.
H
o
w
ev
e
r
,
i
n
ad
e
q
u
a
t
e
m
o
n
i
t
o
r
in
g
o
f
f
a
r
m
la
n
d
h
a
s
o
f
t
en
l
e
d
t
o
s
ig
n
if
i
c
a
n
t
r
i
c
e
lo
s
s
e
s
d
u
e
t
o
d
i
s
e
a
s
e
-
r
e
l
a
t
e
d
i
s
s
u
e
s
.
V
ar
i
o
u
s
d
i
s
e
as
e
s
c
o
m
m
o
n
ly
af
f
ec
t
r
ic
e
cu
l
t
i
v
a
t
io
n
,
c
au
s
i
n
g
s
u
b
s
t
a
n
t
i
a
l
e
c
o
n
o
m
i
c
l
o
s
s
e
s
.
F
u
r
t
h
er
m
o
r
e,
t
h
e
ex
c
e
s
s
i
v
e
u
s
e
o
f
ch
e
m
i
c
a
l
s
l
ik
e
b
a
c
t
e
r
ic
id
e
s
,
a
n
d
f
u
n
g
i
c
id
e
s
t
o
c
o
m
b
a
t
p
l
an
t
d
i
s
ea
s
e
s
h
a
s
n
e
g
a
t
i
v
e
ly
i
m
p
a
c
t
ed
th
e
ag
r
o
-
e
co
s
y
s
t
e
m
[
8
]
.
F
ig
u
r
e
1
d
ep
i
c
t
s
t
h
e
t
r
en
d
s
o
f
t
o
t
a
l
r
ic
e
co
n
s
u
m
p
t
io
n
w
o
r
l
d
w
i
d
e
.
Fig
u
r
e
1
.
T
o
tal
r
ice
co
n
s
u
m
p
ti
o
n
(
wo
r
l
d
wid
e)
T
h
e
ea
r
ly
p
r
ed
ictio
n
a
n
d
f
o
r
e
ca
s
tin
g
o
f
r
ice
leaf
d
is
ea
s
es
p
lay
a
s
ig
n
if
ican
t
r
o
le
in
m
ain
t
ain
in
g
th
e
q
u
ality
an
d
q
u
an
tity
o
f
r
ice
p
r
o
d
u
ctio
n
.
T
im
ely
d
etec
tio
n
o
f
d
is
ea
s
es
allo
ws
f
o
r
ea
r
ly
in
ter
v
en
tio
n
,
wh
ic
h
h
elp
s
co
n
t
r
o
l
d
is
ea
s
e
p
r
o
g
r
ess
io
n
a
n
d
p
r
o
m
o
tes
th
e
h
ea
lth
y
g
r
o
wth
o
f
p
lan
ts
,
u
ltima
tely
i
m
p
r
o
v
i
n
g
r
ice
y
ield
an
d
s
u
p
p
ly
[
9
]
.
C
o
m
m
o
n
r
ice
d
is
ea
s
es
in
clu
d
e
s
h
ea
th
b
lig
h
t,
b
ac
ter
ial
b
lig
h
t,
an
d
r
ice
b
last
,
ea
ch
ch
ar
ac
ter
ized
b
y
d
is
tin
ct
s
y
m
p
to
m
s
r
elate
d
to
tex
tu
r
e,
co
lo
r
,
an
d
s
h
ap
e.
T
h
ese
d
is
ea
s
es
ten
d
to
s
p
r
ea
d
r
ap
i
d
ly
an
d
ar
e
ea
s
ily
tr
a
n
s
m
is
s
ib
le.
C
u
r
r
en
tly
,
ar
tific
ial
id
e
n
tific
atio
n
,
d
is
ea
s
e
m
ap
p
in
g
,
an
d
a
u
to
m
ated
d
etec
tio
n
ar
e
co
n
s
id
er
ed
k
e
y
m
eth
o
d
s
f
o
r
i
d
en
tify
in
g
r
ice
d
is
ea
s
es.
Fig
u
r
e
2
s
h
o
ws
s
am
p
le
im
ag
es
o
f
r
ice
d
is
ea
s
es
wh
er
e
r
ice
s
tack
b
u
r
n
,
lea
f
s
m
u
t,
leaf
s
ca
ld
,
f
alse
s
m
u
t,
b
last
,
s
tem
r
o
tm
wh
ite
tip
,
s
h
ea
th
r
o
t,
s
tr
ip
e
b
lig
h
t,
s
h
ea
th
b
lig
h
t,
b
ac
ter
ial
leaf
s
tr
ea
k
,
an
d
b
r
o
w
n
s
p
o
t im
ag
es a
r
e
d
ep
ic
ted
in
F
ig
u
r
e
s
2
(
a
)
to
2
(
l)
,
r
esp
ec
tiv
ely
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
A
u
to
en
co
d
er a
n
d
GA
N
-
a
id
ed
p
la
n
t d
is
ea
s
e
d
etec
tio
n
i
n
r
ice
a
n
d
co
tt
o
n
…
(
A
n
a
n
d
r
a
d
d
i Na
d
u
vin
a
ma
n
i
)
709
(
a)
(
b
)
(
c)
(
d
)
(
e)
(f)
(
g
)
(
h
)
(
i)
(
j)
(
k
)
(
l)
Fig
u
r
e
2
.
Sam
p
le
im
a
g
es o
f
r
ice
d
is
ea
s
e
of
(
a)
r
ice
s
tack
b
u
r
n
,
(
b
)
r
ice
leaf
s
m
u
t,
(
c)
r
ice
le
af
s
ca
ld
,
(
d
)
r
ice
f
alse sm
u
t,
(
e)
r
ice
b
last
,
(
f
)
r
ice
s
tem
r
o
t,
(
g
)
r
ice
w
h
ite
tip
,
(
h
)
r
ice
s
h
ea
th
r
o
t,
(
i)
r
ice
s
tr
ip
e
b
lig
h
t,
(
j)
r
ice
s
h
ea
th
b
lig
h
t,
(
k
)
b
ac
te
r
ial
leaf
s
tr
ea
k
,
an
d
(
l)
r
ice
b
r
o
w
n
s
p
o
t
Similar
ly
,
co
tto
n
is
an
im
p
o
r
tan
t
cr
o
p
g
lo
b
ally
.
I
t
is
a
v
ital
ca
s
h
cr
o
p
t
h
at
s
u
p
p
o
r
ts
th
e
liv
elih
o
o
d
s
o
f
m
illi
o
n
s
o
f
p
eo
p
le
ac
r
o
s
s
Asi
a
,
Af
r
ica,
Au
s
tr
alia,
an
d
th
e
A
m
er
icas.
I
ts
f
ib
e
r
is
a
k
ey
r
eso
u
r
ce
,
c
o
n
tr
ib
u
tin
g
to
th
e
in
co
m
e
o
f
b
o
th
f
ar
m
er
s
an
d
in
d
u
s
tr
ialis
ts
.
C
u
ltiv
ated
f
o
r
th
o
u
s
an
d
s
o
f
y
ea
r
s
,
c
o
tto
n
h
a
s
b
ee
n
an
ess
en
tial
p
ar
t
o
f
tex
tile
p
r
o
d
u
ctio
n
wo
r
ld
wid
e
an
d
is
an
in
teg
r
al
c
o
m
p
o
n
en
t
o
f
f
ar
m
i
n
g
s
y
s
tem
s
in
ap
p
r
o
x
im
ately
6
0
co
u
n
tr
ies.
T
h
e
lea
d
in
g
p
r
o
d
u
ce
r
s
o
f
co
tto
n
a
r
e
C
h
in
a,
th
e
Un
ited
States
,
an
d
I
n
d
ia.
I
n
th
e
I
n
d
ia
n
co
n
te
x
t,
2
3
% o
f
th
e
co
tto
n
p
r
o
d
u
ce
d
is
ex
p
o
r
ted
i
n
ter
n
atio
n
ally
.
Ho
wev
er
,
co
tto
n
y
ield
s
ar
e
g
r
ea
tl
y
in
f
lu
en
ce
d
b
y
cr
o
p
g
r
o
wth
,
wh
ich
ca
n
b
e
s
ig
n
if
ic
an
tly
af
f
ec
ted
b
y
v
ar
i
o
u
s
d
is
ea
s
es.
F
ig
u
r
e
3
s
h
o
ws
s
am
p
le
i
m
ag
es
o
f
co
tto
n
leaf
d
is
ea
s
es
wh
er
e
h
ea
lth
y
,
leaf
s
p
o
t,
n
u
tr
ien
t
d
e
f
icien
cy
,
p
o
wd
er
y
m
ild
ew,
tar
g
et
s
p
o
r
t,
v
etic
illi
u
m
wilt
,
an
d
leaf
cu
r
l sam
p
les ar
e
d
ep
icted
f
r
o
m
F
ig
u
r
e
s
3
(
a)
t
o
3
(
g
)
,
r
esp
ec
tiv
ely
.
(
a)
(
b
)
(
c)
(
d
)
(
e)
(f)
(
g
)
Fig
u
r
e
3
.
Sam
p
le
c
o
tto
n
leaf
im
ag
e
of
(
a)
h
ea
lth
y
,
(
b
)
leaf
s
p
o
t,
(
c)
n
u
tr
ien
t
d
ef
icien
cy
,
(
d
)
p
o
wd
er
y
m
ild
ew,
(
e)
tar
g
et
s
p
o
t,
(
f
)
v
er
icilliu
m
wilt,
an
d
(
g
)
leaf
cu
r
l
I
n
r
esp
o
n
s
e
to
th
ese
ch
allen
g
e
s
,
ex
ten
s
iv
e
r
esear
ch
h
as
f
o
cu
s
ed
o
n
cr
ea
tin
g
ac
cu
r
ate
an
d
ac
ce
s
s
ib
le
m
eth
o
d
s
f
o
r
th
e
m
ajo
r
ity
o
f
f
ar
m
er
s
.
Pre
cisi
o
n
ag
r
ic
u
ltu
r
e
,
lev
er
a
g
in
g
cu
ttin
g
-
ed
g
e
tec
h
n
o
lo
g
y
,
o
p
tim
izes
d
ec
is
io
n
-
m
ak
in
g
p
r
o
ce
s
s
es.
Mo
d
er
n
d
ig
ital
tech
n
o
lo
g
ies
en
ab
le
r
ea
l
-
tim
e
d
ata
c
o
llectio
n
,
an
d
m
ac
h
in
e
lear
n
in
g
(
ML
)
alg
o
r
ith
m
s
aid
in
p
r
o
v
id
in
g
o
p
tim
al
d
ec
is
io
n
s
,
th
er
eb
y
r
ed
u
cin
g
c
o
s
ts
.
Nev
er
th
eless
,
th
er
e
is
s
till
r
o
o
m
f
o
r
im
p
r
o
v
em
en
t,
p
ar
ticu
lar
ly
i
n
r
e
f
in
in
g
d
ec
is
io
n
-
s
u
p
p
o
r
t
s
y
s
tem
s
ca
p
a
b
le
o
f
tr
an
s
f
o
r
m
in
g
v
ast
am
o
u
n
ts
o
f
d
ata
in
to
p
r
ac
tical
r
ec
o
m
m
en
d
atio
n
s
.
T
h
e
cu
r
r
en
t
tech
n
o
lo
g
ical
ad
v
an
ce
m
en
ts
h
av
e
s
u
g
g
ested
t
o
ad
o
p
t
th
e
co
m
p
u
ter
v
is
io
n
-
b
ased
ML
b
ased
au
to
m
ated
a
p
p
r
o
ac
h
es
f
o
r
ea
r
l
y
d
etec
tio
n
o
f
p
la
n
t
d
is
ea
s
e
d
etec
tio
n
.
Sev
er
al
m
eth
o
d
s
h
av
e
b
ee
n
d
ev
elo
p
e
d
b
ased
o
n
th
is
co
n
ce
p
t
o
f
im
ag
e
p
r
o
ce
s
s
in
g
[
1
0
]
.
I
m
ag
e
p
r
o
ce
s
s
in
g
is
a
s
p
ec
i
alize
d
f
ield
with
in
s
ig
n
al
p
r
o
ce
s
s
in
g
th
at
f
o
cu
s
e
s
o
n
ex
tr
ac
tin
g
v
alu
a
b
le
in
f
o
r
m
atio
n
f
r
o
m
im
ag
es.
ML
,
a
s
u
b
s
et
o
f
ar
tific
ial
in
tellig
en
ce
(
AI
)
,
e
n
ab
les
au
to
m
atio
n
an
d
p
r
o
v
id
es
in
s
tr
u
cti
o
n
s
to
p
er
f
o
r
m
s
p
ec
if
ic
task
s
[
1
1
]
.
T
h
e
m
ain
aim
o
f
ML
is
to
r
ea
lize
tr
ain
in
g
d
ata
an
d
c
r
ea
te
m
o
d
els
th
at
ca
n
ass
is
t
p
eo
p
le
b
y
f
ac
ilit
atin
g
s
o
u
n
d
d
ec
is
io
n
-
m
ak
in
g
an
d
p
r
ed
icti
n
g
ac
cu
r
ate
o
u
tco
m
es
b
ased
o
n
ex
ten
s
iv
e
tr
ain
in
g
d
ata.
I
n
th
e
co
n
tex
t
o
f
p
lan
t
h
ea
lth
,
v
ar
i
o
u
s
im
ag
e
p
r
o
p
er
ti
es
s
u
ch
as
leaf
co
lo
r
,
ex
ten
t
o
f
d
am
a
g
e,
leaf
ar
ea
,
a
n
d
tex
t
u
r
e
p
ar
a
m
eter
s
ar
e
u
tili
ze
d
f
o
r
class
if
icatio
n
p
u
r
p
o
s
es.
I
n
o
u
r
p
r
o
ject,
we
h
av
e
m
eticu
lo
u
s
ly
an
aly
ze
d
d
if
f
er
e
n
t
im
ag
e
p
ar
am
eter
s
an
d
f
ea
t
u
r
es
to
id
en
tif
y
v
a
r
io
u
s
d
is
ea
s
es
af
f
ec
tin
g
p
lan
t
le
av
es,
aim
in
g
f
o
r
th
e
h
ig
h
est
ac
cu
r
ac
y
p
o
s
s
ib
le.
T
r
ad
itio
n
ally
,
p
lan
t
d
is
ea
s
e
d
etec
tio
n
r
elied
o
n
v
is
u
al
in
s
p
ec
tio
n
s
o
f
leav
es
o
r
in
v
o
lv
ed
ch
em
ical
p
r
o
ce
s
s
es
co
n
d
u
cte
d
b
y
ex
p
er
ts
.
Ho
wev
er
,
th
is
ap
p
r
o
ac
h
d
em
an
d
ed
a
lar
g
e
team
o
f
s
p
ec
ialis
ts
a
n
d
co
n
tin
u
o
u
s
p
lan
t
o
b
s
er
v
atio
n
,
m
a
k
in
g
it
co
s
tly
,
esp
ec
ially
f
o
r
lar
g
e
f
ar
m
s
.
I
n
s
u
ch
s
ce
n
ar
io
s
,
o
u
r
p
r
o
p
o
s
ed
s
y
s
tem
p
r
o
v
es
in
v
alu
ab
le
f
o
r
m
o
n
ito
r
in
g
v
ast
ag
r
icu
ltu
r
al
f
ield
s
.
B
y
au
to
m
atica
lly
d
etec
tin
g
d
is
ea
s
es
b
ased
o
n
o
b
s
er
v
a
b
le
s
y
m
p
to
m
s
o
n
p
lan
t
leav
es,
th
is
m
eth
o
d
b
ec
o
m
es
n
o
t
o
n
ly
m
o
r
e
s
tr
aig
h
tf
o
r
war
d
b
u
t
also
s
ig
n
if
ican
tly
m
o
r
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
1
5
,
No
.
1
,
Feb
r
u
ar
y
2
0
2
6
:
707
-
7
2
4
710
co
s
t
-
ef
f
ec
tiv
e.
T
h
er
ef
o
r
e,
we
a
d
o
p
t
t
h
e
ML
b
ased
ap
p
r
o
ac
h
f
o
r
p
lan
t
d
is
ea
s
e
d
etec
tio
n
.
T
h
e
m
ain
co
n
tr
ib
u
tio
n
o
f
th
is
wo
r
k
ar
e
as
f
o
llo
ws:
f
ir
s
t
o
f
all,
we
f
o
cu
s
o
n
im
ag
e
p
r
e
-
p
r
o
ce
s
s
in
g
s
tag
e
an
d
p
r
esen
t
g
en
er
ativ
e
ad
v
er
s
ar
ial
n
etwo
r
k
(
GAN
)
b
ased
m
o
d
el
f
o
r
im
ag
e
d
e
n
o
is
in
g
to
im
p
r
o
v
e
t
h
e
im
ag
e
q
u
ality
.
I
n
n
e
x
t
s
tag
e
we
f
o
cu
s
o
n
f
ea
tu
r
e
ex
tr
ac
tio
n
p
h
ase
wh
er
e
co
lo
r
,
s
h
ap
e
a
n
d
g
r
ad
ien
t
b
ased
f
ea
t
u
r
es
ar
e
e
x
tr
a
cted
an
d
c
o
m
b
in
e
d
to
g
eth
er
to
f
o
r
m
u
late
th
e
f
in
al
f
ea
tu
r
e
v
ec
to
r
.
Fin
ally
,
we
ad
o
p
t d
ec
is
io
n
tr
ee
(
DT
)
class
if
ier
m
o
d
el
to
p
er
f
o
r
m
th
e
class
if
icatio
n
o
f
r
ice
an
d
c
o
tto
n
leaf
d
is
ea
s
e.
2.
L
I
T
E
R
AT
U
RE
SU
RVE
Y
T
h
is
s
ec
tio
n
p
r
esen
ts
th
e
o
v
er
v
iew
o
f
ex
is
tin
g
m
eth
o
d
s
an
d
f
o
cu
s
o
n
i
d
en
tify
in
g
th
e
d
r
awb
ac
k
s
an
d
ch
allen
g
es
in
th
e
e
x
is
tin
g
s
ch
e
m
es.
Vis
h
n
o
i
et
a
l.
[
1
2
]
r
ep
o
r
ted
th
at
p
lan
t
leav
es
ar
e
u
s
ed
t
o
d
etec
t
th
e
v
ar
io
u
s
in
f
ec
tio
n
s
in
th
e
p
lan
ts
.
Au
t
h
o
r
s
u
s
ed
co
m
p
u
ter
v
is
io
n
with
s
o
f
t
co
m
p
u
tin
g
m
eth
o
d
s
an
d
s
u
g
g
ested
to
in
co
r
p
o
r
ate
ef
f
icien
t
f
ea
t
u
r
e
e
x
tr
ac
tio
n
m
eth
o
d
.
I
n
th
is
lin
e
o
f
r
esear
ch
,
Sah
u
an
d
Pan
d
ey
[
1
3
]
p
r
esen
ted
a
n
o
p
tim
ized
ap
p
r
o
ac
h
b
y
u
s
in
g
h
y
b
r
id
m
u
lticlas
s
s
u
p
p
o
r
t
v
ec
to
r
m
ac
h
in
e
(
SVM
)
m
o
d
el
f
o
r
p
lan
t
leaf
d
is
ea
s
e
d
etec
tio
n
.
T
h
e
class
if
icatio
n
m
o
d
el
also
u
s
es
r
a
n
d
o
m
f
o
r
est
(
R
F)
m
o
d
el
to
p
r
o
d
u
c
e
th
e
f
i
n
al
h
y
b
r
id
class
if
icatio
n
.
Prio
r
to
f
ea
tu
r
e
ex
tr
ac
tio
n
,
s
p
atial
f
u
zz
y
c
-
m
e
an
s
is
em
p
lo
y
ed
to
o
b
tain
t
h
e
s
eg
m
en
tatio
n
.
T
h
e
s
eg
m
en
ted
r
e
g
io
n
o
f
in
ter
est
(
R
OI
)
is
u
s
ed
f
u
r
th
e
r
f
o
r
f
ea
tu
r
e
ex
tr
ac
tio
n
.
As
d
is
c
u
s
s
ed
b
e
f
o
r
e,
f
e
at
u
r
e
ex
t
r
ac
ti
o
n
is
an
im
p
o
r
ta
n
t
p
h
a
s
e
in
th
ese
ty
p
es
o
f
c
o
m
p
u
t
e
r
v
is
i
o
n
tas
k
s
t
h
er
ef
o
r
e
,
A
h
m
a
d
e
t
a
l.
[
1
4
]
f
o
c
u
s
e
d
o
n
d
e
v
el
o
p
i
n
g
a
n
ew
f
ea
t
u
r
e
e
x
t
r
a
cti
o
n
m
et
h
o
d
a
n
d
i
n
t
r
o
d
u
c
ed
a
n
e
w
f
e
at
u
r
e
e
x
t
r
a
cti
o
n
m
e
th
o
d
k
n
o
w
n
as l
o
ca
l
tr
i
an
g
u
l
ar
-
t
er
n
a
r
y
p
att
er
n
(
L
T
r
iTP
)
.
I
n
i
m
p
le
m
e
n
ti
n
g
th
e
te
r
n
a
r
y
p
a
tte
r
n
ap
p
r
o
ac
h
,
a
d
y
n
a
m
ic
t
h
r
es
h
o
ld
b
as
ed
o
n
t
h
e
a
b
s
o
l
u
te
m
ea
n
v
a
lu
e
is
ca
l
c
u
la
te
d
.
T
h
is
m
et
h
o
d
e
n
a
b
les
a
s
e
n
s
i
ti
v
e
an
al
y
s
is
o
f
te
x
t
u
r
e
i
n
f
o
r
m
a
ti
o
n
i
n
p
la
n
t
l
ea
f
im
ag
es,
g
e
n
e
r
a
ti
n
g
a
wi
d
e
r
a
n
g
e
o
f
h
i
g
h
l
y
p
er
t
in
e
n
t
v
al
u
es
.
As p
la
n
t d
is
e
ases
c
an
m
a
n
i
f
est
a
t v
a
r
i
o
u
s
o
r
ie
n
t
ati
o
n
s
o
n
a
l
ea
f
im
a
g
e
,
a
h
is
t
o
g
r
am
o
f
t
h
e
g
r
ad
ie
n
t
is
c
o
m
p
u
t
e
d
in
f
o
u
r
d
i
r
ec
ti
o
n
s
(
0
°,
4
5
°,
9
0
°,
a
n
d
1
3
5
°
)
wit
h
i
n
e
ac
h
t
r
ia
n
g
le
.
T
h
is
p
r
o
ce
s
s
h
el
p
s
i
d
en
tif
y
t
h
e
g
r
ad
ie
n
t
c
h
a
n
g
es
i
n
in
f
ec
t
ed
r
e
g
i
o
n
s
c
o
m
p
ar
ed
t
o
h
ea
l
th
y
ar
ea
s
.
Kar
tik
e
y
an
a
n
d
Sh
r
i
v
a
s
tav
a
[
1
5
]
s
u
g
g
este
d
t
o
a
d
o
p
t
i
m
a
g
e
p
r
e
-
p
r
o
ce
s
s
in
g
a
p
p
r
o
a
ch
a
n
d
i
n
tr
o
d
u
c
ed
h
y
b
r
i
d
f
e
at
u
r
e
e
x
t
r
a
cti
o
n
a
p
p
r
o
ac
h
w
h
er
e
d
is
c
r
e
te
w
av
el
et
tr
a
n
s
f
o
r
m
(
DW
T
)
a
n
d
g
r
a
y
lev
el
co
-
o
c
cu
r
r
e
n
c
e
m
at
r
i
x
(
G
L
C
M
)
at
tr
ib
u
t
es a
r
e
f
u
s
e
d
t
o
p
r
o
d
u
c
e
t
h
e
f
in
al
r
o
b
u
s
t f
ea
t
u
r
e
v
ec
t
o
r
.
Fi
n
al
ly
,
t
h
e
S
VM
is
u
s
e
d
t
o
t
r
ai
n
th
e
m
o
d
el
b
ase
d
o
n
e
x
tr
ac
te
d
f
ea
t
u
r
es
a
n
d
l
ab
ell
e
d
i
m
a
g
es.
Ku
lk
ar
n
i
et
a
l.
[
16
]
p
r
esen
te
d
ML
b
ased
a
p
p
r
o
ac
h
wh
ic
h
is
ca
r
r
ied
o
u
t
in
to
m
u
ltip
le
s
tag
es:
f
ir
s
t
s
tag
e
d
em
o
n
s
tr
ates
th
e
p
r
e
-
p
r
o
ce
s
s
in
g
p
h
ase
wh
er
e
g
r
ay
s
ca
le
co
n
v
er
s
io
n
,
f
ilter
in
g
an
d
th
r
esh
o
ld
in
g
s
tep
s
ar
e
em
p
lo
y
ed
.
T
h
e
o
b
tain
ed
p
r
e
-
p
r
o
ce
s
s
ed
im
ag
e
is
th
en
p
r
o
c
ess
ed
th
r
o
u
g
h
th
e
f
ea
tu
r
e
e
x
tr
ac
tio
n
p
h
ase
wh
er
e
m
o
r
p
h
o
lo
g
ical,
GL
C
M
an
d
c
o
lo
r
f
ea
tu
r
es
ar
e
ex
tr
ac
te
d
a
n
d
p
r
o
ce
s
s
ed
th
r
o
u
g
h
th
e
f
ea
tu
r
e
s
elec
tio
n
p
r
o
ce
s
s
.
Fin
ally
,
a
class
if
icatio
n
m
o
d
el
is
d
ep
lo
y
ed
to
tr
ain
th
e
m
o
d
el
b
y
u
s
in
g
th
e
s
elec
ted
f
ea
tu
r
es
.
Ku
m
ar
et
a
l.
[
1
7
]
in
tr
o
d
u
ce
d
ML
b
ased
ap
p
r
o
ac
h
f
o
r
p
lan
t
d
is
ea
s
e
d
etec
tio
n
wh
er
e
im
ag
e
co
n
tr
ast
en
h
an
ce
m
en
t,
s
eg
m
en
tatio
n
an
d
f
ea
t
u
r
e
e
x
tr
ac
tio
n
s
tep
s
ar
e
em
p
lo
y
e
d
p
r
io
r
to
p
e
r
f
o
r
m
t
h
e
class
if
icatio
n
.
I
m
ag
e
s
eg
m
en
tatio
n
is
d
o
n
e
b
y
u
s
in
g
K
-
m
ea
n
s
clu
s
ter
in
g
wh
er
e
as
GL
C
M
m
o
d
el
is
u
s
ed
f
o
r
f
ea
t
u
r
e
e
x
tr
ac
tio
n
.
Fin
ally
,
SVM
class
if
ier
is
u
s
ed
to
p
er
f
o
r
m
th
e
class
if
icatio
n
task
.
Alag
u
m
ar
iap
p
a
n
et
a
l.
[
1
8
]
p
r
esen
ted
a
ML
b
ase
d
ap
p
r
o
ac
h
wh
ich
co
n
s
id
er
s
Hu
m
o
m
en
ts
an
d
Har
alick
tex
tu
r
e
f
ea
tu
r
es f
o
r
f
ea
tu
r
e
an
aly
s
is
an
d
later
th
ese
f
e
atu
r
es a
r
e
f
ed
to
th
e
ex
tr
em
e
lear
n
in
g
m
ac
h
in
e
(
E
L
M)
an
d
SVM
class
if
icatio
n
.
T
h
e
SVM
u
s
es
lin
ea
r
an
d
p
o
ly
n
o
m
ial
k
er
n
el
f
u
n
ctio
n
s
to
o
b
tain
th
e
f
in
al
o
u
tco
m
e.
Pallath
ad
k
a
et
a
l.
[
1
9
]
u
s
ed
h
is
to
g
r
am
-
b
ased
im
ag
e
eq
u
aliza
tio
n
as
p
r
e
-
p
r
o
ce
s
s
in
g
an
d
p
r
in
cip
al
c
o
m
p
o
n
en
t
an
aly
s
is
(
PC
A
)
f
o
r
f
ea
tu
r
e
ex
tr
ac
tio
n
a
n
d
f
in
ally
d
if
f
er
en
t
class
if
icatio
n
m
eth
o
d
s
ar
e
u
s
ed
f
o
r
class
if
icatio
n
.
Sar
an
g
d
h
ar
an
d
Pawar
[
2
0
]
u
s
ed
SVM
b
ased
r
eg
r
ess
io
n
m
eth
o
d
f
o
r
p
lan
t
d
is
ea
s
e
d
etec
tio
n
an
d
d
ev
elo
p
ed
a
n
an
d
r
o
i
d
ap
p
licatio
n
c
o
m
b
in
ed
with
I
o
T
f
ac
ilit
ies.
J
aisak
th
i
et
a
l.
[
21
]
p
r
esen
ted
a
ML
b
ased
d
is
ea
s
e
d
etec
tio
n
ap
p
r
o
ac
h
f
o
r
d
is
ea
s
e
d
etec
tio
n
in
g
r
ap
es.
I
n
f
ir
s
t
p
h
ase,
th
is
ap
p
r
o
ac
h
u
s
es
g
lo
b
al
th
r
esh
o
ld
in
g
an
d
s
em
i
-
s
u
p
er
v
is
ed
m
o
d
el
f
o
r
s
e
g
m
en
tatio
n
to
id
en
tify
th
e
R
OI
.
Fin
ally
,
SVM,
ad
ap
tiv
e
b
o
o
s
tin
g
(
A
d
aBo
o
s
t)
an
d
RF
tr
ee
class
if
ier
s
ar
e
u
s
ed
to
o
b
tain
t
h
e
class
if
icatio
n
o
u
tco
m
e.
R
o
y
et
a
l.
[
2
2
]
r
ep
o
r
te
d
th
at
ex
is
tin
g
m
eth
o
d
s
s
u
f
f
er
f
r
o
m
s
ev
er
al
is
s
u
e
s
an
d
s
u
g
g
ested
to
in
co
r
p
o
r
ate
d
im
en
s
io
n
ality
r
ed
u
ctio
n
m
et
h
o
d
s
to
im
p
r
o
v
e
th
e
class
if
icati
o
n
p
er
f
o
r
m
a
n
ce
.
I
s
in
k
ay
e
et
a
l.
[
2
3
]
p
r
esen
ted
a
co
m
b
in
atio
n
o
f
th
e
v
ar
i
atio
n
al
au
to
en
co
d
e
r
(
VAE
)
an
d
v
is
io
n
tr
an
s
f
o
r
m
er
(
ViT
)
f
r
am
ew
o
r
k
th
at
class
if
ies m
u
lti
-
cla
s
s
p
lan
t d
is
ea
s
es.
T
h
e
VAE
im
p
lem
en
ted
d
im
en
s
io
n
ality
r
ed
u
ctio
n
i
n
im
ag
es
m
ain
tain
in
g
im
p
o
r
tan
t
f
ea
tu
r
es,
an
d
ViT
allo
wed
ex
tr
ac
tin
g
g
lo
b
al
f
ea
tu
r
es.
On
th
e
d
ataset
Plan
tVil
lag
e
(
co
r
n
,
p
o
tato
,
an
d
to
m
ato
)
,
th
eir
m
o
d
el
p
er
f
o
r
m
s
with
a
9
3
.
2
%
ac
cu
r
ac
y
r
ate,
wh
ich
s
h
o
ws an
ad
d
ed
b
en
ef
it o
f
r
o
b
u
s
tn
ess
an
d
lar
g
er
s
ca
le
in
d
is
ea
s
e
class
if
icat
io
n
.
Pra
s
an
n
ak
u
m
ar
an
d
L
ath
a
[
2
4
]
d
ev
elo
p
e
d
co
n
tex
tu
al
m
ask
a
u
to
en
co
d
er
(
C
MA
E
)
wh
o
s
e
o
p
tim
izatio
n
was
d
esig
n
ed
o
n
th
e
b
asis
o
f
th
e
d
y
n
am
ic
d
i
f
f
er
en
tial
an
n
ea
le
d
o
p
tim
izatio
n
al
g
o
r
ith
m
(
DDAO
A)
.
T
h
is
p
lan
t
d
is
ea
s
e
id
en
tific
atio
n
(
PDI
)
-
C
MA
E
-
DDAO
A
m
o
d
el
u
s
in
g
th
is
co
m
b
in
atio
n
o
f
ad
ap
ti
v
e
n
o
is
e
f
ilter
in
g
,
s
tatis
tical
f
ea
tu
r
e
ex
t
r
ac
tio
n
,
an
d
co
s
in
e
s
im
ilar
ity
en
a
b
led
h
id
d
en
Ma
r
k
o
v
m
o
d
el
(
HM
M
)
to
o
u
t
p
er
f
o
r
m
m
o
s
t
cu
r
r
en
t
m
o
d
els
in
ter
m
s
o
f
ac
cu
r
ac
y
,
F1
-
s
co
r
e,
an
d
ev
e
n
s
p
ec
if
icity
.
T
h
is
s
tr
ateg
y
wa
s
also
f
o
cu
s
ed
o
n
th
e
ess
en
ce
o
f
o
p
tim
izatio
n
to
ac
cu
r
ately
ca
teg
o
r
ize
th
e
p
lan
t
d
is
ea
s
e.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
A
u
to
en
co
d
er a
n
d
GA
N
-
a
id
ed
p
la
n
t d
is
ea
s
e
d
etec
tio
n
i
n
r
ice
a
n
d
co
tt
o
n
…
(
A
n
a
n
d
r
a
d
d
i Na
d
u
vin
a
ma
n
i
)
711
C
u
i
et
a
l.
[
2
5
]
h
a
v
e
im
p
r
o
v
e
d
th
e
m
aize
d
is
ea
s
e
id
en
tific
ati
o
n
with
a
c
o
n
v
o
lu
tio
n
al
b
lo
ck
atten
tio
n
m
o
d
u
le
(
C
B
AM
)
in
teg
r
ated
lig
h
tweig
h
t
au
to
en
c
o
d
er
.
Pre
p
r
o
ce
s
s
in
g
was
d
o
n
e
u
s
in
g
DW
T
wh
er
ea
s
C
B
A
M
en
h
an
ce
d
s
p
atial
an
d
c
h
an
n
el
atten
tio
n
o
f
f
ea
tu
r
es.
T
h
e
m
o
d
el
ac
h
iev
ed
9
9
.
4
4
%
ac
cu
r
a
cy
,
an
d
it
s
u
r
p
ass
ed
th
e
p
er
f
o
r
m
an
ce
o
f
s
o
m
e
d
ee
p
co
n
v
o
lu
tio
n
al
n
eu
r
al
n
e
two
r
k
(
C
NN
)
s
tr
u
ctu
r
es
(
e
.
g
.
,
R
esNet
-
50
an
d
Den
s
eNe
t2
0
1
)
,
th
u
s
b
ein
g
u
n
ex
p
ec
ted
ly
ef
f
icien
t
a
n
d
in
ter
p
r
etab
le.
I
n
s
eg
m
e
n
tatio
n
an
d
class
if
icatio
n
o
f
p
lan
t
leaf
d
is
ea
s
es,
Ab
in
ay
a
et
a
l.
[
2
6
]
d
esig
n
ed
ca
s
ca
d
in
g
au
t
o
en
co
d
er
with
atten
tio
n
r
esid
u
al
U
-
Net
(
C
AAR
-
UNe
t)
.
I
t
is
tr
ain
ed
with
cu
s
to
m
d
ataset
an
d
r
ea
ch
ed
th
e
ac
cu
r
ac
y
o
f
9
5
.
2
6
%
o
f
m
ea
n
p
ix
el
an
d
0
.
7
4
5
1
o
f
m
ea
n
in
ter
s
ec
tio
n
o
v
er
u
n
io
n
(
I
o
U
)
,
in
d
icatin
g
h
i
g
h
s
u
cc
ess
in
th
e
lo
ca
tio
n
a
n
d
b
o
u
n
d
ar
y
d
ef
in
itio
n
o
f
d
is
ea
s
e
wh
ich
is
cr
itical
to
ea
r
ly
d
iag
n
o
s
is
.
3.
P
RO
P
O
SE
D
M
O
D
E
L
T
h
is
s
ec
tio
n
p
r
esen
ts
th
e
co
m
p
lete
p
r
o
p
o
s
ed
ap
p
r
o
ac
h
wh
e
r
e
we
h
av
e
u
s
ed
im
ag
e
p
r
e
-
p
r
o
ce
s
s
in
g
,
f
ea
tu
r
e
e
x
tr
ac
tio
n
an
d
s
u
p
e
r
v
is
ed
class
if
icatio
n
to
o
b
tain
t
h
e
f
in
al
r
esu
lt.
T
h
e
p
r
e
-
p
r
o
ce
s
s
in
g
m
o
d
u
le
u
s
es
GAN
ap
p
r
o
ac
h
to
en
h
a
n
ce
th
e
im
ag
e
q
u
ality
b
ef
o
r
e
p
r
o
ce
s
s
in
g
it
th
r
o
u
g
h
th
e
f
ea
tu
r
e
ex
t
r
ac
tio
n
m
o
d
el.
T
h
is
o
b
tin
ae
d
im
a
g
e
is
p
r
o
ce
s
s
ed
th
r
o
u
g
h
th
e
f
ea
tu
r
e
ex
tr
ac
tio
n
p
h
ase
wh
er
e
c
o
lo
r
,
tex
tu
r
e
a
n
d
g
r
a
d
ien
t
f
ea
tu
r
es
ar
e
ex
tr
ac
ted
to
f
o
r
m
u
late
th
e
f
in
al
f
ea
tu
r
e
v
ec
to
r
.
Fin
ally
,
a
clas
s
if
icatio
n
m
o
d
el
is
p
r
e
s
en
ted
to
lear
n
th
e
p
atter
n
f
r
o
m
th
e
attr
ib
u
te
an
d
tr
ain
th
e
m
o
d
el
to
class
if
y
th
e
t
est d
ata.
3
.
1
.
Co
m
preheiv
e
o
v
er
v
iew
o
f
pro
po
s
ed
m
o
del
T
h
is
wo
r
k
p
r
esen
ts
a
d
e
n
o
is
in
g
m
o
d
el
p
r
o
p
o
s
ed
as
s
p
a
r
s
e
a
u
to
en
co
d
er
-
b
ased
g
e
n
er
ativ
e
a
d
v
er
s
ar
ial
n
etwo
r
k
(
Sp
ar
s
e
-
GAN)
to
im
p
r
o
v
e
th
e
im
a
g
e
q
u
ality
wh
il
e
m
ain
tain
in
g
th
e
n
ec
ess
ar
y
s
tr
u
ctu
r
al
f
ea
t
u
r
es
in
th
e
p
r
o
ce
s
s
.
T
h
e
g
e
n
er
ato
r
o
f
th
is
f
r
am
ewo
r
k
is
r
ep
r
esen
ted
as
a
co
n
v
o
lu
tio
n
al
s
p
ar
s
e
a
u
to
en
co
d
er
f
o
r
m
ed
with
an
e
n
co
d
er
-
d
ec
o
d
er
m
o
d
ality
.
T
h
e
en
co
d
er
co
n
s
is
ts
o
f
s
tack
co
n
v
o
l
u
tio
n
al
la
y
er
s
an
d
p
o
o
lin
g
lay
er
s
to
g
r
ad
u
ally
lear
n
h
ig
h
er
lev
el
f
e
atu
r
es
an
d
r
ed
u
ce
th
e
s
p
atial
d
im
en
s
io
n
s
o
f
th
e
in
p
u
t
to
a
m
o
r
e
s
tr
u
ctu
r
al
laten
t
r
ep
r
esen
tatio
n
.
T
h
e
im
p
o
r
tan
t e
lem
en
t
is
th
e
s
p
ar
s
e
f
u
lly
co
n
n
ec
ted
laten
t
lay
er
,
wh
ic
h
ad
d
s
L
1
r
eg
u
lar
izatio
n
p
en
alty
o
n
t
h
e
co
n
d
itio
n
th
at
o
n
ly
a
p
a
r
t
o
f
th
e
n
eu
r
o
n
s
is
ac
tiv
ated
.
T
h
is
s
p
ar
s
ity
s
c
h
em
e
is
u
s
ef
u
l
in
r
em
o
v
in
g
u
s
eless
o
r
n
o
is
y
f
ea
t
u
r
es
b
u
t
allo
ws
r
ete
n
tio
n
o
f
i
m
p
o
r
tan
t v
is
u
al
in
f
o
r
m
atio
n
a
n
d
h
e
n
ce
it
p
e
r
f
o
r
m
s
well
o
n
d
en
o
is
in
g
p
r
o
b
lem
s
.
T
h
e
im
ag
e
is
s
y
m
m
etr
icall
y
r
ec
o
n
s
tr
u
cte
d
b
y
t
h
e
d
ec
o
d
er
co
m
p
o
s
ed
o
f
u
p
s
am
p
lin
g
a
n
d
c
o
n
v
o
lu
tio
n
a
l
lay
er
s
with
a
g
o
al
to
r
ec
o
v
er
f
in
e
-
g
r
ain
ed
d
etails
b
ased
o
n
th
e
c
o
m
p
r
ess
ed
s
p
ar
s
e
r
ep
r
esen
tatio
n
.
Dis
cr
im
in
ato
r
is
a
co
n
v
o
lu
tio
n
al
b
i
n
ar
y
class
if
ier
th
at
ju
d
g
es
th
e
r
ea
lis
m
o
f
g
en
er
ate
d
im
ag
es
an
d
in
s
tr
u
cts
th
e
g
en
er
ato
r
to
g
en
er
ate
o
u
tp
u
t
th
a
t
ca
n
n
o
t
b
e
d
is
cr
im
in
ated
ag
ain
s
t
a
tr
u
e
clea
n
s
am
p
le.
T
h
is
in
co
r
p
o
r
atio
n
o
f
th
e
s
p
ar
s
e
au
to
en
co
d
er
with
i
n
th
e
GAN
m
o
d
el
en
ab
les
th
e
m
o
d
el
to
ac
h
iev
e
ac
cu
r
ac
y
at
th
e
p
ix
el
-
le
v
el
a
n
d
at
th
e
s
am
e
tim
e
p
er
ce
p
tu
al
r
ea
lis
m
.
All
th
e
elem
en
ts
o
f
th
e
o
u
tlin
ed
ar
ch
itectu
r
e
s
er
v
e
a
s
p
ec
if
ic
an
d
v
er
y
im
p
o
r
tan
t
p
u
r
p
o
s
e:
th
e
en
co
d
er
co
m
p
r
ess
es
n
o
is
y
d
ata
in
to
an
ef
f
icien
t
co
d
e,
th
e
s
p
ar
s
ity
p
e
n
alty
r
e
m
o
v
es
n
o
is
e
b
y
m
a
k
in
g
f
ea
tu
r
es
s
elec
tiv
e,
th
e
d
ec
o
d
er
r
en
d
er
s
clea
n
im
ag
es
with
r
etain
ed
s
tr
u
ctu
r
e,
a
n
d
t
h
e
ad
v
er
s
ar
ial
s
u
p
er
v
is
io
n
y
ield
s
v
is
u
al
q
u
ality
.
T
h
ese
b
lo
c
k
s
ca
n
b
e
u
s
ed
j
o
in
tly
to
ac
h
iev
e
s
tr
o
n
g
d
e
n
o
is
in
g
with
m
in
im
al
lo
s
s
o
f
s
em
an
tic
l
ev
el
to
th
e
in
p
u
t
im
ag
es
an
d
,
h
en
ce
,
r
en
d
er
th
e
m
o
d
el
ex
ce
p
tio
n
ally
well
to
d
o
wn
s
tr
ea
m
task
s
o
f
im
ag
e
a
n
a
ly
s
is
an
d
im
ag
e
class
if
icatio
n
.
3
.
2
.
P
re
-
pro
ce
s
s
ing
T
h
e
m
ain
aim
o
f
th
is
s
ec
tio
n
is
to
p
er
f
o
r
m
im
ag
e
en
h
an
ce
m
en
t
to
im
p
r
o
v
e
th
e
q
u
alit
y
o
f
in
p
u
t
im
ag
e.
Gen
er
ally
,
d
u
r
in
g
th
e
im
ag
e
ca
p
tu
r
i
n
g
p
r
o
ce
s
s
,
th
e
im
ag
es
g
et
co
n
ta
m
in
ated
d
u
e
to
s
ev
er
al
f
ac
to
r
s
s
u
ch
as
lo
w
illu
m
in
atio
n
,
c
am
er
a
s
h
ak
e,
an
d
m
o
tio
n
b
l
u
r
.
wh
ich
m
ay
im
p
ac
t
t
h
e
q
u
ality
o
f
a
n
aly
s
is
.
T
h
er
ef
o
r
e,
th
is
is
s
u
e
n
ee
d
s
to
b
e
tack
led
.
T
o
o
v
er
co
m
e
th
is
is
s
u
e,
im
ag
e
f
ilter
in
g
/d
en
o
is
in
g
is
co
n
s
id
er
ed
as
o
n
e
o
f
th
e
im
p
o
r
tan
t
asp
ec
ts
wh
er
e
th
e
u
n
wan
ted
n
o
is
y
d
a
ta
is
r
em
o
v
ed
f
r
o
m
th
e
im
a
g
e.
I
n
th
is
wo
r
k
,
we
ad
o
p
t
d
ee
p
lear
n
in
g
-
b
ased
m
ec
h
an
is
m
to
p
er
f
o
r
m
th
e
im
a
g
e
f
ilter
in
g
task
.
T
h
er
ef
o
r
e,
we
p
r
esen
t
a
s
p
ar
s
e
GAN
ar
ch
itectu
r
e
to
p
er
f
o
r
m
im
ag
e
f
ilter
in
g
.
GANs
ar
e
a
class
o
f
ML
wh
ic
h
ar
e
wid
ely
u
s
ed
in
v
ar
i
o
u
s
c
o
m
p
u
ter
v
is
io
n
r
elate
d
task
s
.
T
h
e
GAN
m
o
d
els co
n
s
is
t o
f
two
n
eu
r
al
n
etwo
r
k
s
(
NN)
as g
en
er
ato
r
an
d
d
is
cr
im
in
ato
r
wh
ich
a
r
e
tr
ain
ed
with
th
e
h
elp
o
f
ad
v
er
s
ar
ial
tr
ain
in
g
m
eth
o
d
.
W
h
ile
GAN
s
ar
e
p
r
im
ar
ily
k
n
o
wn
f
o
r
g
en
er
atin
g
n
ew
d
at
a
s
am
p
les,
th
ey
ca
n
also
b
e
u
tili
ze
d
f
o
r
task
s
lik
e
im
ag
e
d
e
n
o
is
in
g
.
T
h
e
c
o
m
p
l
ete
p
r
o
ce
s
s
o
f
GA
f
o
r
im
ag
e
d
en
o
is
in
g
task
is
as f
o
llo
ws
.
3
.
2
.
1
.
G
ener
a
t
o
r
T
h
e
g
e
n
er
ato
r
tak
es
a
n
o
is
y
i
m
ag
e
as
in
p
u
t
a
n
d
tr
ies
to
g
en
er
ate
a
d
e
n
o
is
ed
im
a
g
e
(
)
.
T
h
e
g
en
er
ato
r
ca
n
b
e
r
ep
r
esen
ted
b
y
a
f
u
n
ctio
n
:
ℝ
×
×
→
ℝ
×
×
,
wh
er
e
W
,
H
,
a
n
d
C
r
ep
r
esen
t
th
e
wid
th
,
h
eig
h
t,
an
d
n
u
m
b
e
r
o
f
ch
an
n
e
ls
o
f
th
e
im
ag
es,
r
esp
ec
tiv
ely
.
Fig
u
r
e
4
p
r
esen
ts
a
g
en
er
ato
r
ar
ch
itectu
r
e
b
ased
o
n
GAN
m
o
d
el.
T
h
e
ar
c
h
itectu
r
al
d
etails o
f
g
e
n
er
ato
r
b
l
o
ck
ar
e
p
r
esen
ted
in
T
ab
le
1
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
1
5
,
No
.
1
,
Feb
r
u
ar
y
2
0
2
6
:
707
-
7
2
4
712
Fig
u
r
e
4.
GAN
b
ased
g
en
er
ato
r
b
lo
ck
f
o
r
d
en
o
is
in
g
T
ab
le
1.
Ar
c
h
itectu
r
e
d
etails o
f
g
en
er
at
o
r
b
lo
c
k
La
y
e
r
Ty
p
e
K
e
r
n
e
l
si
z
e
S
t
r
i
d
e
A
c
t
i
v
a
t
i
o
n
O
u
t
p
u
t
s
i
z
e
1
C
o
n
v
2
D
3
×
3
1
R
e
LU
2
5
6
×
2
5
6
×
6
4
2
C
o
n
v
2
D
3
×
3
1
R
e
LU
2
5
6
×
2
5
6
×
6
4
3
C
o
n
v
2
D
3
×
3
2
R
e
LU
1
2
8
×
1
2
8
×
1
2
8
4
C
o
n
v
2
D
3
×
3
1
R
e
LU
1
2
8
×
1
2
8
×
1
2
8
5
C
o
n
v
2
D
3
×
3
2
R
e
LU
6
4
×
6
4
×
2
5
6
6
C
o
n
v
2
D
3
×
3
1
R
e
LU
6
4
×
6
4
×
2
5
6
7
Tr
a
n
s
p
o
se
d
C
o
n
v
2
D
3
×
3
2
R
e
LU
1
2
8
×
1
2
8
×
1
2
8
8
Tr
a
n
s
p
o
se
d
C
o
n
v
2
D
3
×
3
2
R
e
LU
2
5
6
×
2
5
6
×
6
4
9
C
o
n
v
2
D
3
×
3
1
Ta
n
h
2
5
6
×
2
5
6
×
C
3
.
2
.
2
.
Dis
cr
im
ina
t
o
r
S
im
ilar
ly
,
d
is
cr
im
in
ato
r
tr
ies
to
d
is
tin
g
u
is
h
b
etwe
en
r
ea
l
cl
ea
n
im
ag
es
(
)
an
d
g
e
n
er
ated
d
e
n
o
is
ed
im
ag
es
(
(
)
)
.
I
t
co
n
s
id
er
s
th
e
in
p
u
t
im
ag
e
wh
ich
is
g
en
er
ated
b
y
g
en
er
ato
r
b
lo
ck
a
n
d
f
o
cu
s
o
n
p
r
o
d
u
cin
g
th
e
d
is
cr
im
in
ated
o
u
tp
u
t
ac
c
o
r
d
in
g
to
th
e
tr
ain
in
g
p
r
o
ce
d
u
r
e
:
i
)
in
p
u
t
:
eith
e
r
a
r
ea
l
clea
n
im
ag
e
o
r
a
g
en
er
ated
d
e
n
o
is
ed
im
ag
e
(
)
an
d
ii)
o
u
tp
u
t
:
p
r
o
b
ab
ilit
y
s
co
r
e
in
d
icatin
g
th
e
au
th
en
ticity
o
f
t
h
e
in
p
u
t im
ag
e
(
r
ea
l
o
r
f
a
k
e)
.
T
h
e
d
is
cr
im
in
ato
r
ca
n
b
e
r
e
p
r
esen
ted
b
y
a
f
u
n
ctio
n
:
ℝ
×
×
→
[
0
,
1
]
wh
er
e
[
0
,
1
]
[
0
,
1
]
r
ep
r
esen
ts
th
e
p
r
o
b
ab
ilit
y
s
co
r
e
o
u
tp
u
t
.
Fig
u
r
e
5
p
r
esen
ts
th
e
d
escr
im
in
ato
r
ar
c
h
itectu
r
e
to
o
b
tain
th
e
f
i
n
al
d
en
o
is
ed
im
ag
e.
T
h
e
ar
c
h
itectu
r
al
d
etails o
f
th
is
d
is
cr
im
in
at
o
r
b
lo
c
k
ar
e
p
r
esen
ted
in
T
a
b
l
e
2
.
Fig
u
r
e
5.
Dis
cr
im
in
ato
r
f
o
r
d
is
tin
g
u
is
h
in
g
o
r
ig
in
al
clea
n
a
n
d
g
en
er
ated
d
en
o
is
ed
im
a
g
e
T
ab
le
2.
Ar
c
h
itectu
r
al
d
etails o
f
d
is
cr
im
in
ato
r
b
lo
ck
La
y
e
r
Ty
p
e
K
e
r
n
e
l
si
z
e
S
t
r
i
d
e
A
c
t
i
v
a
t
i
o
n
O
u
t
p
u
t
s
i
z
e
1
C
o
n
v
2
D
4
×
4
2
Le
a
k
y
R
e
LU
(
0
.
2
)
1
2
8
×
1
2
8
×
6
4
2
C
o
n
v
2
D
4
×
4
2
Le
a
k
y
R
e
LU
(
0
.
2
)
6
4
×
6
4
×
1
2
8
3
C
o
n
v
2
D
4
×
4
2
Le
a
k
y
R
e
LU
(
0
.
2
)
3
2
×
3
2
×
2
5
6
4
C
o
n
v
2
D
4
×
4
2
Le
a
k
y
R
e
LU
(
0
.
2
)
1
6
×
1
6
×
5
1
2
5
C
o
n
v
2
D
4
×
4
1
Le
a
k
y
R
e
LU
(
0
.
2
)
1
5
×
1
5
×
1
6
F
l
a
t
t
e
n
+
d
e
n
se
-
-
S
i
g
m
o
i
d
1
3
.
2
.
3
.
L
o
s
s
f
un
ct
io
n
L
o
s
s
f
u
n
ctio
n
estab
lis
h
es
th
e
g
o
al
wh
ich
g
u
i
d
es
th
e
tr
ai
n
in
g
o
f
b
o
t
h
th
e
d
is
cr
im
in
ato
r
n
e
two
r
k
an
d
th
e
g
en
er
ato
r
n
etwo
r
k
.
GANs
h
av
e
th
e
g
o
al
o
f
en
s
u
r
i
n
g
th
at
th
e
s
am
p
les
g
en
er
ated
b
y
th
e
g
en
e
r
ato
r
ar
e
in
d
is
tin
g
u
is
h
ab
le
to
ac
tu
al
d
at
a,
an
d
th
e
g
o
al
o
f
th
e
d
is
cr
im
in
ato
r
is
to
en
s
u
r
e
th
at
r
ea
l
a
n
d
g
en
er
ated
s
am
p
les
ar
e
d
is
tin
g
u
is
h
ab
le.
T
h
e
lo
s
s
f
u
n
ctio
n
t
h
at
is
ap
p
lied
to
th
e
g
en
er
ato
r
is
d
ef
in
e
d
in
s
u
ch
a
way
th
at
it
r
ed
u
ce
s
th
e
p
r
o
b
ab
ilit
y
o
f
th
e
d
is
cr
im
in
ato
r
to
c
o
r
r
ec
tly
r
ec
o
g
n
ize
th
e
g
en
er
ated
s
am
p
les
as
f
ak
e.
T
h
u
s
,
t
h
e
o
b
jectiv
e
o
f
th
e
g
e
n
er
ato
r
m
o
d
u
le
is
to
m
in
im
ize
th
e
lo
g
p
r
o
b
ab
ilit
y
t
h
at
th
e
d
is
cr
im
in
ato
r
m
ak
es a
m
is
tak
e
.
ℒ
(
,
)
=
~
(
)
[
l
og
(
)
]
+
~
(
)
[
l
og
(
1
−
(
(
)
)
)
]
(
1
)
W
h
er
e
(
)
is
th
e
d
is
tr
ib
u
tio
n
o
f
r
ea
l
clea
n
im
ag
es,
(
)
is
th
e
d
is
tr
i
b
u
tio
n
o
f
n
o
is
y
im
ag
es,
(
)
is
th
e
o
u
tp
u
t
o
f
th
e
d
is
cr
im
in
at
o
r
f
o
r
r
ea
l
clea
n
im
a
g
es
an
d
(
(
)
)
is
th
e
o
u
tp
u
t
o
f
th
e
d
is
cr
im
in
ato
r
f
o
r
g
en
er
ated
d
en
o
is
ed
im
ag
es.
Si
m
ilar
ly
,
d
en
o
is
in
g
lo
s
s
m
ea
s
u
r
es th
e
d
if
f
er
en
ce
b
etwe
en
th
e
g
en
er
ated
d
en
o
is
ed
im
ag
e
(
(
)
)
an
d
th
e
clea
n
tar
g
et
im
ag
e
(
)
.
W
e
u
s
e
m
ea
n
s
q
u
ar
ed
er
r
o
r
(
MSE
)
lo
s
s
f
o
r
th
is
p
u
r
p
o
s
e
wh
ich
is
ex
p
r
ess
ed
as
(
2
)
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
A
u
to
en
co
d
er a
n
d
GA
N
-
a
id
ed
p
la
n
t d
is
ea
s
e
d
etec
tio
n
i
n
r
ice
a
n
d
co
tt
o
n
…
(
A
n
a
n
d
r
a
d
d
i Na
d
u
vin
a
ma
n
i
)
713
ℒ
(
)
=
~
(
)
,
~
(
)
[
‖
(
)
−
‖
2
2
]
(
2
)
3
.
3
.
F
e
a
t
ure
ex
t
r
a
ct
io
n
T
h
is
p
r
esen
ts
th
e
d
etailed
d
is
cu
s
s
io
n
ab
o
u
t
p
r
o
p
o
s
ed
a
p
p
r
o
a
ch
f
o
r
f
ea
t
u
r
e
ex
tr
ac
tio
n
.
T
h
is
ap
p
r
o
ac
h
is
b
ased
o
n
th
r
ee
d
if
f
er
en
t
f
e
atu
r
e
ex
tr
a
ctio
n
m
eth
o
d
s
an
d
co
m
b
in
ed
th
e
o
b
tain
e
d
f
ea
tu
r
es
to
f
o
r
m
u
late
th
e
f
in
al
f
ea
tu
r
e
v
ec
to
r
.
T
h
e
f
e
a
t
u
r
e
e
x
t
r
a
c
ti
o
n
i
n
c
l
u
d
e
s
c
o
lo
r
,
t
e
x
t
u
r
e
a
n
d
i
m
p
r
o
v
e
d
s
c
a
l
e
-
i
n
v
a
r
i
a
n
t
f
e
at
u
r
e
t
r
a
n
s
f
o
r
m
(
S
I
F
T
)
a
n
d
h
i
s
t
o
g
r
am
o
f
o
r
i
e
n
t
e
d
g
r
a
d
i
e
n
t
s
(
HO
G
)
f
e
a
t
u
r
e
e
x
t
r
a
c
ti
o
n
b
y
u
s
i
n
g
g
r
a
d
i
e
n
t
c
o
r
r
e
l
a
ti
o
n
s
.
3
.
3
.
1
.
Co
lo
r
f
e
a
t
ure
T
h
is
s
u
b
s
ec
tio
n
p
r
esen
ts
th
e
m
o
r
e
ad
v
an
ce
d
m
ath
em
atic
al
m
o
d
el
f
o
r
co
lo
r
f
ea
tu
r
e
ex
tr
ac
tio
n
in
v
o
lv
es
th
e
u
s
e
o
f
c
o
lo
r
m
o
m
en
ts
.
C
o
lo
r
m
o
m
e
n
ts
f
e
atu
r
es
r
ep
r
esen
t
th
e
s
tatis
tical
m
ea
s
u
r
es
o
f
th
e
co
n
s
id
er
ed
im
ag
e
an
d
u
s
ed
to
d
escr
ib
e
th
e
d
is
tr
ib
u
tio
n
o
f
co
l
o
r
s
in
th
e
in
p
u
t
i
m
ag
e.
A
lg
o
r
ith
m
1
d
em
o
n
s
tr
ates th
e
p
r
o
ce
s
s
to
co
m
p
u
te
th
e
c
o
lo
r
f
ea
t
u
r
e
ex
tr
ac
t
io
n
u
s
in
g
c
o
lo
r
m
o
m
en
ts
.
Alg
o
r
ith
m
1
.
C
o
lo
r
f
ea
tu
r
e
ex
t
r
ac
tio
n
u
s
in
g
c
o
lo
r
m
o
m
en
ts
Step
1
: im
ag
e
p
r
e
-
p
r
o
ce
s
s
in
g
1
:
I
n
p
u
t: c
o
n
s
id
er
th
e
p
lan
t le
a
f
im
ag
e
as in
p
u
t im
ag
e
with
d
im
en
s
io
n
×
.
2
:
Pre
-
p
r
o
ce
s
s
: p
er
f
o
r
m
d
if
f
er
en
t step
s
s
u
ch
as r
esize,
n
o
r
m
a
lize
an
d
C
NN
b
ased
d
en
o
is
in
g
m
o
d
el
.
Step
2
: c
o
lo
r
m
o
m
en
t c
o
m
p
u
t
atio
n
1
:
C
alcu
late
th
e
co
lo
r
m
o
m
en
t
s
f
o
r
ea
ch
c
h
an
n
el.
T
h
e
co
l
o
r
m
o
m
en
ts
ar
e
ca
lcu
lated
as f
o
llo
ws:
–
First o
r
d
er
m
o
m
e
n
t (
m
ea
n
)
:
=
1
∑
∑
(
,
)
=
1
=
1
(
3
)
–
Seco
n
d
o
r
d
er
m
o
m
en
ts
(
v
a
r
ian
ce
)
:
2
=
1
∑
∑
[
(
,
)
−
1
]
=
1
=
1
(
4
)
–
T
h
ir
d
o
r
d
er
m
o
m
en
t
:
=
1
∑
∑
[
(
,
)
−
1
]
=
1
=
1
(
5
)
w
h
e
r
e
r
e
p
r
e
s
e
n
t
s
t
h
e
m
o
m
e
n
t
o
r
d
e
r
(
1
r
e
p
r
e
s
e
n
t
s
m
e
a
n
2
r
e
p
r
e
s
e
n
ts
v
a
r
i
a
n
c
e
a
n
d
3
r
e
p
r
e
s
e
n
ts
t
h
e
s
k
e
w
n
es
s
)
.
Step
3
: f
ea
tu
r
e
v
ec
to
r
f
o
r
m
u
lat
io
n
–
I
n
o
r
d
er
to
c
r
ea
te
th
e
f
in
al
f
ea
t
u
r
e
v
ec
to
r
,
we
ar
r
a
n
g
e
all
f
ea
t
u
r
es a
s
(
6
)
.
=
[
,
2
,
,
,
2
,
,
,
2
,
]
(
6
)
3
.
3
.
2
.
T
ex
t
ure
f
ea
t
ure
ex
t
ra
c
t
io
n
T
ex
tu
r
e
f
ea
tu
r
e
ex
t
r
ac
tio
n
f
r
o
m
im
ag
es
o
f
ten
in
v
o
lv
es
an
aly
zin
g
p
atter
n
s
an
d
v
ar
iatio
n
s
in
p
ix
el
in
ten
s
ities
.
On
e
wid
ely
u
s
ed
m
eth
o
d
f
o
r
tex
tu
r
e
f
ea
tu
r
e
ex
tr
ac
tio
n
is
t
h
e
GL
C
M.
GL
C
M
ca
lcu
lates
th
e
o
cc
u
r
r
e
n
ce
s
o
f
p
ix
el
p
air
s
with
s
p
ec
if
ic
v
alu
es
an
d
d
is
tan
ce
s
in
a
n
im
a
g
e,
ca
p
tu
r
i
n
g
tex
t
u
r
e
p
atter
n
s
.
A
lg
o
r
ith
m
2
s
h
o
ws th
e
p
r
o
ce
s
s
to
ex
tr
ac
t th
e
tex
tu
r
e
f
ea
tu
r
e
v
ec
to
r
.
Alg
o
r
ith
m
2
.
T
ex
tu
r
e
f
ea
tu
r
e
e
x
tr
ac
tio
n
u
s
in
g
GL
C
M
Step
1
:
im
ag
e
p
r
e
-
p
r
o
ce
s
s
in
g
1
:
I
n
p
u
t: c
o
n
s
id
er
th
e
p
lan
t le
a
f
im
ag
e
as in
p
u
t im
ag
e
with
d
im
en
s
io
n
×
.
2
:
Pre
-
p
r
o
ce
s
s
: p
er
f
o
r
m
d
if
f
er
en
t step
s
s
u
ch
as r
esize,
n
o
r
m
a
lize
an
d
C
NN
b
ased
d
en
o
is
in
g
m
o
d
el
.
Step
2
: G
L
C
M
co
m
p
u
tatio
n
–
C
h
o
o
s
e
a
s
et
o
f
d
is
p
lace
m
en
t
v
ec
to
r
s
(
,
)
to
d
e
f
in
e
p
ix
el
p
air
s
'
d
is
tan
ce
s
an
d
an
g
les (
e.
g
.
,
(
1
,
0
)
(
1
,
0
)
f
o
r
h
o
r
izo
n
tal,
(
1
,
1
)
(
1
,
1
)
f
o
r
d
iag
o
n
al)
.
–
Fo
r
ea
ch
p
ix
el
(
,
)
in
th
e
im
ag
e,
ca
lcu
late
th
e
GL
C
M
v
a
lu
es
b
y
co
n
s
id
er
in
g
th
e
p
ix
el
p
air
s
(
(
,
)
,
(
+
,
+
)
)
with
in
th
e
s
p
ec
if
ied
d
is
p
lace
m
en
t.
–
C
o
u
n
t th
e
o
cc
u
r
r
en
ce
s
o
f
th
es
e
p
ix
el
p
air
s
an
d
co
n
s
tr
u
ct
a
GL
C
M
f
o
r
ea
ch
d
is
p
lace
m
e
n
t v
ec
to
r
.
–
No
r
m
alize
th
e
GL
C
M
m
atr
ices to
o
b
tain
p
r
o
b
a
b
ilit
ies o
f
o
cc
u
r
r
en
ce
.
Step
3
:
f
ea
tu
r
e
e
x
tr
ac
tio
n
‒
C
o
n
tr
ast:
m
ea
s
u
r
es th
e
lo
ca
l v
ar
iatio
n
s
in
th
e
im
ag
e
∑
∑
(
−
)
2
×
(
,
)
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
1
5
,
No
.
1
,
Feb
r
u
ar
y
2
0
2
6
:
707
-
7
2
4
714
‒
E
n
tr
o
p
y
: m
ea
s
u
r
es th
e
r
a
n
d
o
m
n
ess
o
r
co
m
p
lex
ity
o
f
t
h
e
tex
t
u
r
e
:
=
−
∑
∑
(
,
)
×
l
og
(
(
,
)
)
(
7
)
‒
Ho
m
o
g
en
eity
:
m
ea
s
u
r
es
th
e
clo
s
en
ess
o
f
th
e
d
is
tr
ib
u
tio
n
o
f
elem
en
ts
in
th
e
GL
C
M
to
th
e
GL
C
M
d
iag
o
n
al
:
=
∑
∑
(
,
)
1
+
|
−
|
(
8
)
‒
C
o
r
r
elatio
n
: m
ea
s
u
r
es th
e
lin
e
ar
d
ep
en
d
en
cy
b
etwe
en
th
e
p
i
x
el
p
air
s
in
th
e
GL
C
M
:
=
∑
∑
(
−
)
(
−
)
×
(
,
)
2
(
9
)
wh
er
e
is
th
e
m
ea
n
an
d
is
th
e
s
tan
d
ar
d
d
ev
iatio
n
o
f
th
e
GL
C
M.
3
.
3
.
3
.
G
ra
dient
ba
s
ed
f
ea
t
ur
e
ex
t
ra
ct
i
o
n f
o
r
i
m
pro
v
ing
t
he
H
O
G
a
nd
SI
F
T
f
ea
t
ure
T
h
is
s
ec
tio
n
d
escr
ib
es
th
e
p
r
o
p
o
s
ed
m
ec
h
a
n
is
m
f
o
r
im
p
r
o
v
ed
HOG
an
d
SIFT
f
ea
t
u
r
e
ex
t
r
ac
tio
n
f
o
r
p
lan
t d
is
ea
s
e
d
etec
tio
n
b
y
u
s
in
g
ML
ap
p
r
o
ac
h
.
T
h
is
m
o
d
el
ca
n
b
e
co
n
s
id
er
ed
as a
n
ex
ten
s
io
n
o
f
Ho
G
o
r
SIFT
f
r
o
m
1
st
to
2
nd
o
r
d
e
r
s
tatis
tics
.
Acc
o
r
d
in
g
t
o
th
is
ap
p
r
o
ac
h
th
e
g
r
ad
ien
ts
o
f
im
a
g
e
ar
e
r
e
p
r
esen
ted
s
p
ar
s
ely
co
r
r
esp
o
n
d
in
g
to
th
eir
m
a
g
n
it
u
d
e
an
d
o
r
ie
n
tatio
n
s
.
L
et
u
s
co
n
s
id
er
th
at
im
ag
e
r
eg
io
n
is
p
r
esen
ted
as
an
d
it
s
p
o
s
itio
n
v
ec
to
r
is
r
ep
r
esen
ted
as
=
(
,
)
.
T
h
e
g
r
ad
ien
t o
f
th
is
im
ag
e
is
r
ep
r
esen
ted
as
(
,
)
wh
ich
ca
n
b
e
ex
p
r
ess
ed
in
ter
m
s
o
f
m
ag
n
itu
d
e
at
ea
ch
p
ix
el
as
=
√
2
+
2
an
d
it
s
co
r
r
esp
o
n
d
in
g
o
r
ien
tatio
n
a
n
g
le
ca
n
b
e
ex
p
r
ess
ed
as
=
a
r
c
ta
n
(
,
)
.
I
n
th
i
s
ap
p
r
o
ac
h
,
t
h
e
o
b
tain
e
d
o
r
i
en
tatio
n
s
is
co
d
ed
in
to
o
r
ien
tatio
n
b
in
s
with
th
e
h
elp
o
f
ass
ig
n
in
g
weig
h
ts
to
th
e
n
ea
r
est
b
in
s
.
T
h
is
i
s
d
ef
in
ed
as
s
p
ar
s
e
v
ec
to
r
wh
ich
is
ex
p
r
ess
ed
as
(
∈
)
an
d
k
n
o
wn
a
s
g
r
ad
ien
t
o
r
ien
tatio
n
v
ec
to
r
.
W
ith
th
e
h
elp
o
f
g
r
ad
ien
t
v
ec
to
r
an
d
m
ag
n
itu
d
e
,
th
e
ℎ
o
r
d
er
au
to
co
r
r
elatio
n
g
r
a
d
ien
t f
u
n
ctio
n
in
lo
ca
l n
eig
h
b
o
u
r
ca
n
b
e
e
x
p
r
ess
ed
as
(
1
0
)
.
(
0
,
…
,
,
1
,
…
.
,
)
=
∫
[
(
)
,
(
+
1
)
,
…
,
(
+
)
]
0
(
)
1
(
+
1
)
…
(
+
)
(
1
0
)
W
h
er
e
r
ep
r
esen
ts
th
e
d
is
p
la
ce
m
en
t
v
ec
to
r
f
r
o
m
t
h
e
co
n
s
i
d
er
ed
r
e
f
er
en
ce
p
o
in
t
,
r
ep
r
e
s
en
ts
th
e
ℎ
elem
en
t
o
f
an
d
is
th
e
s
ca
lar
weig
h
tin
g
f
u
n
ctio
n
.
T
h
is
m
ath
em
atica
l
ex
p
r
ess
io
n
d
em
o
n
s
tr
ates
two
ty
p
es
o
f
g
r
ad
ien
t
c
o
r
r
elatio
n
wh
ich
ar
e
s
p
atial
an
d
o
r
ie
n
tatio
n
al
co
r
r
e
latio
n
.
T
h
e
s
p
atial
co
r
r
elatio
n
is
d
er
iv
ed
f
r
o
m
th
e
d
is
p
lace
m
en
t
v
ec
to
r
a
n
d
o
r
ien
tatio
n
al
co
r
r
elatio
n
is
o
b
tai
n
e
d
f
r
o
m
p
r
o
d
u
ct
o
f
elem
en
ts
o
f
.
B
ased
o
n
th
is
,
th
e
g
r
ad
ien
t
b
ased
lo
ca
l c
o
r
r
el
atio
n
ar
e
co
m
p
u
ted
as
(
1
1
)
an
d
(
1
2
)
.
0
ℎ
=
0
(
0
)
=
∑
(
)
0
(
)
∈
(
1
1
)
1
=
1
(
0
,
1
,
1
)
=
∑
min
[
(
)
,
(
+
)
]
0
(
)
1
(
+
)
∈
(
1
2
)
T
h
e
ab
o
v
e
-
m
e
n
tio
n
ed
0
ℎ
o
r
d
er
g
r
ad
ie
n
t
r
ep
r
esen
ts
th
e
h
is
t
o
g
r
am
o
f
g
r
ad
ie
n
ts
u
s
ed
in
SIFT
an
d
HOG
wh
er
ea
s
th
e
1
st
o
r
d
er
co
r
r
elatio
n
ca
n
b
e
co
n
s
id
er
e
d
as
jo
in
t
h
is
to
g
r
am
o
f
o
r
ie
n
tatio
n
p
air
s
.
Mo
r
eo
v
er
,
th
e
1
st
o
r
d
er
o
r
ien
tatio
n
f
ea
tu
r
es a
ls
o
ch
ar
ac
ter
ize
th
e
im
ag
e
co
n
to
u
r
s
.
3
.
4
.
Cla
s
s
if
ica
t
io
n
m
o
del
T
h
is
wo
r
k
h
as
ad
o
p
ted
th
e
s
u
p
er
v
is
ed
class
if
icatio
n
ap
p
r
o
ac
h
to
o
b
tain
th
e
f
in
al
class
if
icatio
n
.
I
n
th
is
wo
r
k
,
we
h
av
e
ad
o
p
te
d
DT
class
if
ica
tio
n
f
o
r
m
u
lticlas
s
s
u
p
er
v
is
ed
class
if
icatio
n
.
A
lg
o
r
ith
m
3
d
em
o
n
s
tr
ates
th
e
class
if
icatio
n
alg
o
r
ith
m
.
Acc
o
r
d
in
g
t
o
th
i
s
ap
p
r
o
ac
h
,
is
co
n
s
id
er
ed
as
a
f
ea
tu
r
e
m
atr
ix
with
s
am
p
les
an
d
f
ea
tu
r
es
wh
ich
is
r
e
p
r
esen
ted
as
=
[
1
,
2
,
…
]
wh
er
e
is
th
e
f
ea
tu
r
e
v
ec
t
o
r
o
f
ℎ
s
am
p
le.
Similar
ly
,
is
th
e
v
ec
to
r
o
f
lab
els
with
elem
en
ts
wh
ich
is
ex
p
r
ess
ed
as
=
[
1
,
2
,
…
,
]
wh
er
e
is
th
e
lab
el
o
f
ℎ
s
am
p
le.
T
h
ese
v
ec
to
r
s
ar
e
ar
r
an
g
e
d
as
=
(
,
)
to
f
o
r
m
u
late
th
e
d
ataset
an
d
is
t
h
e
tr
ee
m
o
d
el.
A
lg
o
r
ith
m
3
.
DT
alg
o
r
ith
m
Step
1
: in
itializatio
n
: c
r
ea
te
a
n
o
d
e
at
th
e
r
o
o
t o
f
t
h
e
tr
ee
.
Step
2
:
s
p
litt
in
g
cr
iter
ia:
s
ele
ct
th
e
b
est
f
ea
tu
r
e
to
s
p
lit
t
h
e
d
ata
at
n
o
d
e
.
T
h
is
is
d
o
n
e
b
y
e
v
alu
atin
g
a
s
p
litt
in
g
cr
iter
io
n
s
u
ch
as Gin
i
im
p
u
r
ity
o
r
E
n
t
r
o
p
y
.
L
et
b
e
th
e
s
et
o
f
p
o
s
s
ib
le
f
ea
tu
r
e
v
alu
es f
o
r
f
ea
tu
r
e
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
A
u
to
en
co
d
er a
n
d
GA
N
-
a
id
ed
p
la
n
t d
is
ea
s
e
d
etec
tio
n
i
n
r
ice
a
n
d
co
tt
o
n
…
(
A
n
a
n
d
r
a
d
d
i Na
d
u
vin
a
ma
n
i
)
715
Step
3
: sp
lit th
e
d
ata:
p
ar
titi
o
n
th
e
d
ataset
in
to
s
u
b
s
ets
1
,
2
,
…
,
b
ased
o
n
th
e
v
alu
es in
.
Step
4
:
r
ec
u
r
s
iv
e
s
p
litt
in
g
:
f
o
r
ea
ch
s
u
b
s
et
,
cr
ea
te
a
ch
ild
n
o
d
e
an
d
r
e
p
ea
t
s
tep
s
2
-
3
u
n
til
a
s
to
p
p
in
g
cr
iter
io
n
is
m
et
(
e.
g
.
,
m
ax
im
u
m
d
ep
th
a
n
d
m
i
n
im
u
m
s
am
p
le
s
p
er
leaf
)
.
Step
4
:
ass
ig
n
lab
els:
if
a
s
to
p
p
in
g
c
r
iter
io
n
is
m
et,
ass
ig
n
a
lab
el
to
n
o
d
e
b
ased
o
n
th
e
m
ajo
r
ity
class
in
.
I
f
all
s
am
p
les in
h
av
e
th
e
s
a
m
e
class
,
b
ec
o
m
es a
leaf
n
o
d
e.
Step
5
: p
r
ed
ictio
n
: t
o
class
if
y
a
n
ew
s
am
p
le
,
tr
av
er
s
e
th
e
tr
e
e
f
r
o
m
t
h
e
r
o
o
t,
f
o
llo
win
g
th
e
ap
p
r
o
p
r
iate
b
r
an
ch
es b
ased
o
n
th
e
f
ea
tu
r
e
v
alu
es o
f
.
T
h
is
p
r
o
ce
s
s
is
r
ep
ea
ted
u
n
til th
e
leaf
n
o
d
e
is
r
ea
ch
ed
.
Fin
ally
,
th
e
leaf
n
o
d
e
class
is
ass
ig
n
ed
as p
r
ed
icted
class
o
f
.
T
h
e
Gin
i
in
d
ex
is
u
s
ed
in
t
h
is
wo
r
k
as
th
e
s
p
litt
in
g
c
r
iter
io
n
in
t
h
e
d
ec
is
io
n
-
m
ak
in
g
p
r
o
c
ess
o
f
th
e
m
o
d
el.
T
h
e
Gin
i
in
d
e
x
is
u
s
ed
to
m
ea
s
u
r
e
th
e
im
p
u
r
ity
o
f
a
d
ataset
an
d
it
is
al
s
o
u
s
ed
t
o
d
eter
m
in
e
th
e
b
est
s
p
lit at
ea
ch
n
o
d
e
I
t c
an
b
e
ex
p
r
ess
ed
as
(
1
3
)
.
(
)
=
1
−
∑
(
)
2
=
1
(
1
3
)
W
h
er
e
,
is
th
e
n
u
m
b
er
o
f
clas
s
es,
an
d
is
th
e
p
r
o
p
o
r
tio
n
o
f
s
am
p
les
in
class
in
d
ataset
.
Similalr
y
,
th
e
en
tr
o
p
y
m
ea
s
u
r
e
is
ap
p
lied
as
a
clas
s
if
icatio
n
cr
iter
io
n
in
th
is
wo
r
k
.
E
n
tr
o
p
y
m
ea
s
u
r
es
th
e
u
n
ce
r
tain
ty
o
r
d
is
o
r
d
er
in
t
h
e
d
ataset
an
d
ca
n
b
e
ap
p
lied
t
o
d
eter
m
in
e
th
e
b
est
attr
ib
u
te
t
o
s
p
lit
u
s
in
g
t
h
e
o
n
e
th
at
o
f
f
er
s
th
e
g
r
ea
test
in
f
o
r
m
atio
n
g
ain
.
I
t c
an
b
e
ex
p
r
ess
ed
as
(
1
4
)
.
(
)
=
−
∑
l
og
2
(
)
=
1
(
1
4
)
W
h
er
e
,
is
th
e
n
u
m
b
er
o
f
class
es,
an
d
is
th
e
p
r
o
p
o
r
tio
n
o
f
s
am
p
les in
class
in
d
ataset
.
4.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
N
T
h
is
s
ec
tio
n
p
r
esen
ts
th
e
o
u
tc
o
m
e
o
f
p
r
o
p
o
s
ed
ap
p
r
o
ac
h
a
n
d
co
m
p
a
r
es
it
p
er
f
o
r
m
a
n
ce
wi
th
ex
is
tin
g
ap
p
r
o
ac
h
es
to
s
h
o
ws
th
e
r
o
b
u
s
tn
ess
o
f
p
r
o
p
o
s
ed
m
o
d
el.
First
o
f
all,
we
d
escr
ib
e
th
e
ex
p
er
iem
n
tal
s
etu
p
f
o
llo
wed
b
y
t
h
e
d
ataset
d
et
ails
an
d
later
p
er
f
o
r
m
an
ce
m
ea
s
u
r
em
en
t
p
ar
am
eter
s
a
r
e
d
escr
ib
ed
.
Fin
ally
,
co
m
p
ar
ativ
e
a
n
aly
s
is
is
p
r
esen
ted
.
4
.
1
.
E
x
perim
ent
a
l set
up
E
x
p
er
im
en
ts
wer
e
p
e
r
f
o
r
m
ed
in
th
e
s
am
e
f
r
am
ewo
r
k
o
n
a
PC
p
o
wer
ed
b
y
W
in
d
o
ws
1
1
,
eq
u
ip
p
ed
with
I
n
tel
C
o
r
e
i7
,
1
6
GB
o
f
R
AM
an
d
NVI
DI
A
R
T
X
3
0
6
0
(
6
GB
VR
AM
)
.
I
t
w
as
im
p
lem
en
ted
o
n
Py
th
o
n
3
.
1
0
,
a
n
d
th
e
m
ain
d
e
ep
lear
n
i
n
g
f
r
am
ewo
r
k
s
c
o
m
p
r
is
e
T
en
s
o
r
Flo
w
2
.
1
1
an
d
Ke
r
as
in
b
u
ild
in
g
a
n
d
tr
ain
in
g
th
e
m
o
d
els.
Oth
er
lib
r
ar
ies
th
at
wer
e
u
tili
ze
d
in
clu
d
ed
Nu
m
Py
,
Pan
d
as,
Op
e
n
C
V,
Scik
it
-
lear
n
,
Ma
tp
lo
tlib
,
an
d
Alb
u
m
e
n
tatio
n
s
to
m
an
i
p
u
late
d
ata,
p
e
r
f
o
r
m
d
ata
a
u
g
m
en
tatio
n
,
a
n
d
v
is
u
alize
th
e
d
ata.
I
n
p
u
t
im
ag
es
wer
e
s
ca
led
to
t
h
e
s
ize
o
f
2
5
6
×
2
5
6
p
ix
els
an
d
th
e
m
o
d
els
wer
e
tr
ain
ed
u
p
to
1
0
0
e
p
o
ch
s
with
a
b
atch
s
ize
o
f
8
u
s
in
g
Ad
am
o
p
tim
izatio
n
alg
o
r
ith
m
with
an
in
i
tial
lear
n
in
g
r
ate
o
f
0
.
0
0
1
w
h
ich
was
ad
ju
s
ted
d
y
n
am
ically
th
r
o
u
g
h
lear
n
in
g
r
ate
s
ch
ed
u
ler
.
T
h
e
lo
s
s
f
u
n
c
tio
n
was
ca
teg
o
r
ical
cr
o
s
s
-
en
tr
o
p
y
.
T
o
av
o
id
th
e
is
s
u
e
o
f
o
v
er
f
itti
n
g
,
ea
r
ly
s
to
p
p
in
g
was p
er
f
o
r
m
ed
,
wh
ich
o
c
cu
r
r
ed
i
n
th
e
r
a
n
g
e
o
f
6
0
-
8
0
e
p
o
ch
s
.
T
h
e
p
r
o
p
o
s
ed
GAN
-
b
ased
d
e
n
o
is
in
g
ar
ch
itectu
r
e
in
clu
d
es
a
g
en
er
ato
r
an
d
a
d
is
cr
im
in
at
o
r
aim
ed
at
im
p
r
o
v
in
g
n
o
is
y
ag
r
icu
ltu
r
al
i
m
ag
es.
T
h
e
g
e
n
er
ato
r
tr
ain
s
t
h
e
n
o
is
y
im
a
g
es
to
th
e
d
e
n
o
is
ed
r
esu
lts
u
s
in
g
s
ix
co
n
v
o
l
u
tio
n
al
lay
er
s
with
r
e
ctif
ied
lin
ea
r
u
n
it
(
R
eL
U
)
a
ctiv
atio
n
,
s
k
ip
co
n
n
ec
tio
n
s
,
an
d
two
tr
an
s
p
o
s
ed
co
n
v
o
l
u
tio
n
lay
er
s
to
u
p
s
am
p
l
e
im
ag
es,
an
d
th
en
T
an
h
ac
tiv
atio
n
is
u
s
ed
to
o
b
tain
th
e
d
en
o
is
ed
im
ag
e
in
th
e
f
in
al
lay
er
.
T
h
e
d
is
cr
im
in
ato
r
in
clu
d
es
f
iv
e
co
n
v
o
lu
tio
n
al
b
lo
ck
s
with
a
L
ea
k
y
R
eL
U
ac
tiv
atio
n
an
d
a
b
atch
n
o
r
m
aliza
tio
n
f
o
llo
wed
b
y
a
s
ig
m
o
id
o
u
tp
u
t
t
o
class
if
y
b
et
wee
n
r
ea
l
an
d
g
en
e
r
ated
im
ag
e,
an
d
d
r
o
p
o
u
t
(
0
.
3
)
is
ad
d
ed
to
av
o
id
o
v
er
f
itti
n
g
.
T
h
e
n
etwo
r
k
s
ar
e
tr
ain
ed
,
with
th
e
h
elp
o
f
b
o
th
a
d
v
er
s
ar
ial
lo
s
s
an
d
MSE
lo
s
s
,
o
p
tim
ized
u
s
in
g
Ad
am
o
p
tim
izer
(
lear
n
in
g
r
ate=
0
.
0
0
0
2
,
b
a
tch
s
ize=
1
6
)
th
r
o
u
g
h
1
0
0
ep
o
ch
s
.
I
m
ag
e
s
ize
is
r
ed
u
ce
d
to
2
5
6
×
2
5
6
p
ix
els
d
u
r
in
g
d
ataset
p
r
ep
r
o
ce
s
s
in
g
,
wh
ich
is
th
en
au
g
m
en
te
d
b
y
u
s
in
g
r
o
tatio
n
,
f
lip
p
in
g
,
an
d
b
r
i
g
h
tn
ess
m
an
ip
u
latio
n
t
o
en
h
a
n
ce
g
en
e
r
aliza
tio
n
.
I
n
o
r
d
er
t
o
ac
h
iev
e
th
e
b
lan
ce
b
etwe
en
ac
cu
r
ac
y
a
n
d
p
e
r
ce
p
tu
al
f
id
elity
,
th
e
o
v
er
all
o
b
jectiv
e
f
u
n
ctio
n
o
f
p
r
o
p
o
s
ed
m
o
d
el
is
f
o
r
m
ed
as
weig
h
te
co
m
b
in
atio
n
o
f
ad
v
e
r
s
ar
ial
lo
s
s
a
n
d
MSE
lo
s
s
.
T
h
e
ad
v
er
s
ar
ial
lo
s
s
h
elp
s
to
en
co
u
r
ag
e
th
e
g
en
er
ato
r
to
p
r
o
d
u
c
e
r
ea
lis
tic
d
en
o
is
ed
im
ag
e
wh
i
le
MSE
lo
s
s
en
s
u
r
es
th
e
s
tr
u
ctu
r
al
co
n
s
is
ten
cy
an
d
p
ix
el
lev
el
s
im
ilalr
tiy
.
ℒ
(
,
)
=
ℒ
(
,
)
+
ℒ
(
)
(
1
5
)
W
h
er
e
ℒ
(
,
)
is
th
e
ad
v
er
s
ar
ial
lo
s
s
wh
ich
is
d
ef
in
ed
as
(
1
6
)
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
1
5
,
No
.
1
,
Feb
r
u
ar
y
2
0
2
6
:
707
-
7
2
4
716
ℒ
(
,
)
=
~
(
)
[
l
og
(
)
]
+
~
(
)
[
l
og
(
1
−
(
(
)
)
)
]
(
1
6
)
ℒ
(
)
r
ep
r
esen
ts
th
e
m
ea
n
s
q
u
ar
ed
r
ec
o
n
s
tr
u
ctio
n
lo
s
s
wh
ich
is
r
e
p
r
esen
ted
as
(
1
7
)
.
ℒ
(
)
=
,
~
[
‖
(
)
−
‖
2
2
]
(
1
7
)
an
d
r
ep
r
esen
ts
th
e
weig
h
tin
g
co
ef
f
icen
ts
wh
ich
ar
e
u
s
ed
t
o
b
alan
ce
th
e
p
er
ce
p
tu
al
an
d
p
ix
el
-
wis
e
lo
s
s
es.
4
.
2
.
Da
t
a
s
et
det
a
ils
T
h
is
s
ec
tio
n
p
r
esen
ts
th
e
b
r
ief
d
is
cu
s
s
io
n
ab
o
u
t
th
e
d
ataset
u
s
ed
in
th
is
wo
r
k
wh
er
e
we
h
av
e
u
s
ed
r
ice
leaf
d
is
ea
s
e
an
d
co
tto
n
leaf
d
is
ea
s
e
d
ataset.
T
h
e
r
ice
d
a
taset
co
n
s
is
ts
o
f
im
ag
es
o
f
r
ic
e
leav
es
wh
ich
ar
e
af
f
ec
ted
b
y
s
ev
er
al
d
is
ea
s
es
s
u
ch
as
b
ac
ter
ial
leaf
b
lig
h
t,
b
r
o
wn
s
p
o
t,
an
d
o
th
e
r
s
.
Simila
r
ly
,
th
e
co
tto
n
leaf
d
ata
in
clu
d
e
leaf
cu
r
l,
b
ac
ter
ia
l b
lig
h
t,
an
d
m
o
r
e
.
4
.
2
.
1
.
Rice
pla
nt
da
t
a
s
et
T
h
e
r
ice
leaf
d
is
ea
s
e
d
ataset
is
o
b
tain
ed
b
y
co
m
p
r
eh
en
s
iv
e
co
m
p
ilatio
n
o
f
in
f
o
r
m
atio
n
b
a
s
ed
o
n
th
e
th
r
ee
co
m
m
o
n
r
ice
d
is
ea
s
e
wh
ich
co
n
s
id
er
s
b
ac
ter
ial
b
lig
h
t
,
b
r
o
wn
s
p
o
t
,
an
d
leaf
s
m
u
t
.
T
h
is
d
ataset
is
d
esig
n
ed
to
s
u
p
p
o
r
t
r
esear
c
h
er
s
,
ag
r
o
n
o
m
is
ts
,
an
d
ML
ex
p
er
ts
in
an
aly
s
in
g
,
d
iag
n
o
s
in
g
,
an
d
p
o
s
s
ib
ly
f
o
r
ec
asti
n
g
t
h
e
a
p
p
ea
r
a
n
ce
o
f
th
ese
d
is
ea
s
es
b
y
u
tili
zin
g
a
r
a
n
g
e
o
f
attr
ib
u
tes
an
d
p
ar
am
ete
r
s
.
T
h
e
d
ataset
was
co
m
p
iled
f
r
o
m
v
a
r
io
u
s
s
ec
o
n
d
ar
y
s
o
u
r
ce
s
,
in
clu
d
in
g
s
ev
e
r
al
well
-
k
n
o
wn
o
n
lin
e
r
ep
o
s
ito
r
ies.
T
ab
le
3
p
r
o
v
id
e
s
a
d
etailed
o
v
e
r
v
iew
o
f
th
ese
s
o
u
r
ce
s
,
wh
ich
i
n
clu
d
e
Me
n
d
el
ey
[
27
]
,
Kag
g
le
[
28
]
,
UC
I
[
29
]
,
an
d
GitHu
b
[
30
]
.
Sp
ec
if
ically
,
1
,
5
8
4
,
4
0
,
4
0
,
a
n
d
1
9
2
im
a
g
es
o
f
b
ac
ter
ial
b
l
ig
h
t
wer
e
s
o
u
r
ce
d
f
r
o
m
Me
n
d
eley
,
Kag
g
le,
UC
I
,
an
d
GitHu
b
,
r
esp
ec
tiv
ely
.
Fo
r
leaf
s
m
u
t,
4
0
im
ag
es
wer
e
o
b
tain
ed
f
r
o
m
Kag
g
le
a
n
d
an
o
th
er
4
0
f
r
o
m
UC
I
.
Ad
d
itio
n
ally
,
im
ag
es
o
f
r
ice
b
last
in
f
ec
tio
n
wer
e
g
ath
er
e
d
f
r
o
m
Me
n
d
eley
an
d
GitH
u
b
.
Alto
g
eth
er
,
t
h
e
d
ataset
co
m
p
r
is
es
3
,
5
3
5
im
a
g
es
o
f
d
is
ea
s
ed
r
ice
leav
es.
I
m
ag
es
wer
e
co
llected
u
n
d
er
d
iv
er
s
e
lig
h
tin
g
co
n
d
itio
n
s
,
v
a
r
y
in
g
b
ac
k
g
r
o
u
n
d
s
,
an
d
with
d
if
f
er
en
t
ca
m
er
a
d
ev
ices.
T
h
e
im
ag
es
wer
e
r
esized
to
a
u
n
if
o
r
m
s
ize
o
f
2
2
4
×2
2
4
d
u
r
i
n
g
p
r
ep
r
o
ce
s
s
in
g
.
T
o
en
s
u
r
e
ef
f
ec
tiv
e
tr
ain
in
g
a
n
d
u
n
b
iased
ev
alu
ati
o
n
,
t
h
e
d
ataset
was
r
an
d
o
m
l
y
p
ar
titi
o
n
ed
in
to
:
7
0
%
tr
ain
in
g
,
1
5
%
v
alid
atio
n
,
a
n
d
1
5
%
test
in
g
.
T
h
is
s
tr
atif
ie
d
s
p
lit
was
ap
p
lied
p
er
class
to
m
ain
tain
class
b
alan
ce
ac
r
o
s
s
all
s
u
b
s
ets.
T
ab
le
3.
Data
s
et
d
etails f
o
r
r
ic
e
p
lan
t
D
i
sea
s
e
t
y
p
e
O
n
l
i
n
e
r
e
p
o
s
i
t
o
r
y
G
i
t
h
H
u
b
K
a
g
g
l
e
U
C
I
M
e
n
d
e
l
e
y
B
a
c
t
e
r
i
a
l
b
l
i
g
h
t
1
9
2
40
40
1
,
5
8
4
Le
a
f
sm
u
t
40
40
R
i
c
e
b
l
a
st
1
5
9
1
,
4
4
0
4
.
2
.
2
.
Co
t
t
o
n lea
f
dis
ea
s
e
A
d
ataset
f
ea
tu
r
in
g
im
ag
es
o
f
b
o
th
h
ea
lth
y
an
d
d
is
ea
s
ed
co
tto
n
leav
es
an
d
p
lan
ts
h
as
b
ee
n
s
o
u
r
ce
d
f
r
o
m
Kag
g
le.
c
o
m
,
s
p
ec
if
ically
f
r
o
m
a
co
m
p
etitio
n
o
r
g
an
ized
b
y
D3
V
[
31
]
.
T
h
is
d
at
aset
in
clu
d
es
f
o
u
r
ca
teg
o
r
ies
o
f
im
ag
es:
f
r
esh
co
tto
n
leav
es,
f
r
esh
co
tto
n
p
la
n
ts
,
d
is
ea
s
ed
co
tto
n
leav
es,
a
n
d
d
is
ea
s
ed
co
tto
n
p
lan
ts
,
to
talin
g
2
,
3
1
0
im
ag
es.
Fo
r
in
s
tan
ce
,
im
ag
es
lab
eled
“
2
”
an
d
“
3
”
r
e
p
r
esen
t
f
r
esh
an
d
d
is
ea
s
ed
s
am
p
les,
r
esp
ec
tiv
ely
,
with
in
th
is
d
atas
et.
I
m
ag
e
r
eso
lu
tio
n
s
is
o
b
tain
ed
as
2
5
6
×
2
5
6
p
ix
els,
an
d
im
ag
es
wer
e
tak
en
in
n
atu
r
al
lig
h
tin
g
ac
r
o
s
s
d
if
f
er
en
t
tim
es
o
f
d
ay
,
in
tr
o
d
u
cin
g
v
ar
iatio
n
in
illu
m
in
atio
n
an
d
b
ac
k
g
r
o
u
n
d
.
T
h
e
d
ataset
was p
ar
titi
o
n
ed
as f
o
llo
ws:
7
0
%
tr
ain
in
g
,
1
5
% v
alid
atio
n
,
an
d
1
5
% testi
n
g
.
4
.
3
.
P
er
f
o
r
m
a
nce
ev
a
lua
t
io
n pa
ra
m
et
er
s
T
h
is
wo
r
k
m
ain
ly
f
o
c
u
s
ed
o
n
p
lan
t
im
ag
e
d
en
o
is
e
an
d
p
lan
t
leaf
d
is
ea
s
e
cla
s
s
if
icatio
n
b
y
u
s
in
g
co
m
p
u
ter
v
is
io
n
a
n
d
ML
b
ase
d
s
o
lu
tio
n
s
.
T
h
er
ef
o
r
e,
we
d
is
cu
s
s
im
ag
e
d
en
o
s
in
g
p
er
f
o
r
m
an
ce
ev
alu
atio
n
an
d
class
if
icatio
n
p
er
f
o
r
m
an
ce
m
ea
s
u
r
em
en
t
m
eth
o
d
s
in
th
is
s
ec
tio
n
.
T
h
e
im
ag
e
d
en
o
is
in
g
p
er
f
o
r
m
an
ce
is
m
ea
s
u
r
ed
in
ter
m
s
o
f
p
ea
k
s
ig
n
al
-
to
-
n
o
is
e
r
atio
(
PS
NR
)
an
d
MSE
.
On
th
e
o
th
er
h
a
n
d
,
th
e
class
if
icatio
n
p
er
f
o
r
m
an
ce
is
m
ea
s
u
r
e
d
in
t
er
m
s
o
f
ac
c
u
r
ac
y
,
s
en
s
itiv
ity
,
s
p
ec
if
icity
,
an
d
F1
-
s
co
r
e.
T
h
e
f
ir
s
t
s
u
b
s
ec
tio
n
d
escr
ib
es
th
e
m
eth
o
d
s
u
s
ed
to
m
ea
s
u
r
e
th
e
d
en
o
is
in
g
p
er
f
o
r
m
an
ce
an
d
n
ex
t
s
u
b
s
e
ctio
n
d
escr
ib
es
th
e
p
ar
am
eter
s
u
s
ed
to
e
v
alu
ate
th
e
class
if
icatio
n
p
er
f
o
r
m
a
n
ce
.
4
.
3
.
1
.
I
m
a
g
e
deno
is
ing
perf
o
rm
a
nce
m
e
a
s
urem
ent
pa
ra
m
et
er
s
T
h
e
im
ag
e
d
e
n
o
is
in
g
p
er
f
o
r
m
an
ce
is
m
ea
s
u
r
ed
b
y
c
o
m
p
ar
i
n
g
th
e
f
ilter
ed
im
ag
e
with
t
h
e
a
ctu
al
in
p
u
t
im
ag
e
g
iv
en
to
th
e
f
ilter
in
g
m
o
d
u
le.
T
h
e
PS
NR
is
th
e
co
m
m
o
n
m
etr
ic
u
s
ed
to
ass
ess
th
e
p
er
f
o
r
m
a
n
ce
o
f
th
ese
Evaluation Warning : The document was created with Spire.PDF for Python.