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.
15
,
No
.
3
,
J
u
n
e
20
26
,
p
p
.
2521
~
2
5
3
2
I
SS
N:
2
2
5
2
-
8
9
3
8
,
DOI
: 1
0
.
1
1
5
9
1
/ijai.v
15
.i
3
.
p
p
2
5
2
1
-
2
5
3
2
2521
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
a
i
.
ia
esco
r
e.
co
m
To
ma
to
lea
f
disea
se cla
ss
ificatio
n u
sing
DenseNe
t
-
1
2
1
wit
h
da
ta
a
ug
menta
tion a
nd f
ine
-
tuni
ng
Su
f
a
j
a
r
B
utsia
nto
,
Ang
g
i Mu
ha
m
m
a
d Rif
a
’
i
D
e
p
a
r
t
me
n
t
o
f
I
n
f
o
r
m
a
t
i
c
s
En
g
i
n
e
e
r
i
n
g
,
F
a
c
u
l
t
y
o
f
En
g
i
n
e
e
r
i
n
g
,
P
e
l
i
t
a
B
a
n
g
sa
U
n
i
v
e
r
si
t
y
,
B
e
k
a
si
,
I
n
d
o
n
e
si
a
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Dec
1
,
2
0
2
4
R
ev
is
ed
May
3
,
2
0
2
6
Acc
ep
ted
May
11
,
2
0
2
6
In
re
c
e
n
t
y
e
a
rs,
a
c
c
u
ra
te
c
las
sifica
ti
o
n
o
f
a
g
ricu
lt
u
ra
l
ima
g
e
s
h
a
s
b
e
c
o
m
e
in
c
re
a
sin
g
ly
imp
o
rtan
t
to
su
p
p
o
rt
p
re
c
isio
n
fa
rm
in
g
a
n
d
c
ro
p
d
ise
a
se
m
o
n
it
o
r
in
g
.
Ho
we
v
e
r,
a
c
h
iev
i
n
g
re
li
a
b
le
p
e
rfo
rm
a
n
c
e
re
m
a
in
s
c
h
a
ll
e
n
g
i
n
g
d
u
e
to
v
isu
a
l
sim
il
a
rit
y
b
e
twe
e
n
d
ise
a
se
c
las
se
s
a
n
d
d
a
tas
e
t
v
a
riab
il
it
y
.
T
h
is
stu
d
y
p
re
se
n
ts
a
n
a
p
p
li
e
d
e
v
a
l
u
a
ti
o
n
o
f
De
n
se
Ne
t
-
1
2
1
c
o
m
b
i
n
e
d
with
d
a
ta
a
u
g
m
e
n
tatio
n
a
n
d
fi
n
e
-
tu
n
in
g
fo
r
m
u
lt
i
-
c
las
s
to
m
a
to
lea
f
d
ise
a
se
c
las
sifica
ti
o
n
.
Ex
p
e
rime
n
ts
we
re
c
o
n
d
u
c
te
d
u
si
n
g
a
p
u
b
li
c
ly
a
v
a
il
a
b
le
to
m
a
to
lea
f
ima
g
e
d
a
tas
e
t
c
o
n
sist
in
g
o
f
5
,
0
0
0
ima
g
e
s
a
c
ro
ss
1
0
c
l
a
ss
e
s.
All
ima
g
e
s
we
re
re
siz
e
d
to
6
4
×
6
4
p
i
x
e
ls
a
n
d
s
p
li
t
in
t
o
8
0
%
train
i
n
g
a
n
d
2
0
%
tes
ti
n
g
se
ts
u
sin
g
a
stra
ti
fied
stra
teg
y
.
Da
ta
a
u
g
m
e
n
tatio
n
wa
s
a
p
p
li
e
d
e
x
c
lu
siv
e
ly
to
t
h
e
train
i
n
g
d
a
ta
to
imp
ro
v
e
g
e
n
e
ra
li
z
a
ti
o
n
.
Th
e
e
x
p
e
rime
n
tal
re
su
lt
s
sh
o
w
a
p
ro
g
re
ss
iv
e
p
e
rf
o
r
m
a
n
c
e
imp
ro
v
e
m
e
n
t
a
c
ro
ss
train
i
n
g
sta
g
e
s,
a
c
h
iev
in
g
a
fi
n
a
l
c
las
sifica
ti
o
n
a
c
c
u
ra
c
y
o
f
9
8
.
4
4
%
wit
h
a
lo
ss
o
f
4
.
7
2
%
a
fter
fin
e
-
tu
n
i
n
g
.
P
e
r
-
c
las
s
e
v
a
lu
a
ti
o
n
i
n
d
ica
tes
stro
n
g
p
e
rfo
rm
a
n
c
e
a
c
ro
ss
m
o
st
d
ise
a
se
c
a
teg
o
ries
,
with
m
in
o
r
c
o
n
fu
si
o
n
o
b
se
rv
e
d
a
m
o
n
g
v
isu
a
ll
y
sim
il
a
r
c
las
s
e
s.
Wh
il
e
th
e
re
su
lt
s
d
e
m
o
n
stra
te
th
e
e
ffe
c
ti
v
e
n
e
ss
o
f
th
e
p
ro
p
o
se
d
train
in
g
stra
teg
y
u
n
d
e
r
c
o
n
tro
ll
e
d
e
x
p
e
rime
n
tal
c
o
n
d
i
ti
o
n
s,
fu
rt
h
e
r
v
a
li
d
a
ti
o
n
u
si
n
g
re
a
l
-
field
ima
g
e
s
is
re
q
u
ired
.
Ov
e
ra
ll
,
th
is
stu
d
y
sh
o
ws
t
h
e
p
o
ten
ti
a
l
o
f
De
n
se
Ne
t
-
1
2
1
wi
th
tran
sfe
r
lea
rn
i
n
g
t
o
su
p
p
o
rt
to
m
a
to
lea
f
d
ise
a
se
c
las
sifica
ti
o
n
in
p
re
c
isio
n
a
g
ricu
lt
u
re
a
p
p
li
c
a
ti
o
n
s.
K
ey
w
o
r
d
s
:
Ag
r
icu
ltu
r
al
im
ag
es
Data
au
g
m
en
tatio
n
Den
s
eNe
t
-
121
Fin
e
-
tu
n
in
g
T
o
m
ato
leaf
d
is
ea
s
e
d
etec
tio
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
:
Su
f
ajar
B
u
ts
ian
to
Dep
ar
tm
en
t o
f
I
n
f
o
r
m
atics E
n
g
in
ee
r
in
g
,
Facu
lty
o
f
E
n
g
in
ee
r
in
g
,
Pelita Ban
g
s
a
Un
iv
er
s
ity
So
u
th
C
ik
ar
an
g
,
B
ek
asi,
W
est J
av
a,
I
n
d
o
n
esia
E
m
ail: su
f
ajar
@
p
elitab
an
g
s
a.
ac
.
id
1.
I
NT
RO
D
UCT
I
O
N
T
h
e
ag
r
icu
ltu
r
al
s
ec
to
r
is
r
ap
id
ly
ev
o
lv
in
g
with
tech
n
o
lo
g
ical
ad
v
an
ce
m
en
ts
aim
ed
at
im
p
r
o
v
i
n
g
ef
f
icien
cy
an
d
s
u
s
tain
ab
ilit
y
,
y
et
ac
cu
r
ate
ag
r
icu
ltu
r
al
im
ag
e
class
if
icatio
n
r
em
ain
s
a
m
aj
o
r
ch
allen
g
e
[
1
]
–
[
4
]
.
Var
iatio
n
s
in
en
v
ir
o
n
m
en
tal
co
n
d
itio
n
s
,
cr
o
p
d
iv
er
s
ity
,
im
ag
e
q
u
ality
,
an
d
s
u
b
tle
d
is
ea
s
e
s
y
m
p
to
m
s
s
ig
n
if
ican
tly
co
m
p
licate
class
if
icatio
n
task
s
,
o
f
ten
lea
d
in
g
to
m
is
class
if
icatio
n
th
at
ca
n
r
esu
lt
in
in
c
o
r
r
ec
t
tr
ea
tm
en
t
d
ec
is
io
n
s
,
ec
o
n
o
m
i
c
lo
s
s
es,
an
d
r
ed
u
ce
d
tr
u
s
t
in
p
r
ec
is
io
n
ag
r
icu
ltu
r
e
s
y
s
tem
s
[
5
]
–
[
1
0
]
.
T
h
ese
ch
allen
g
es
ar
e
f
u
r
t
h
er
in
ten
s
if
ied
b
y
n
o
is
y
,
r
ed
u
n
d
an
t,
a
n
d
l
im
ited
d
atasets
,
h
ig
h
lig
h
tin
g
t
h
e
n
ee
d
f
o
r
r
o
b
u
s
t
an
d
ad
ap
tab
le
class
if
icatio
n
a
p
p
r
o
ac
h
es.
Dee
p
lear
n
in
g
m
o
d
els,
p
ar
ticu
lar
ly
co
n
v
o
lu
tio
n
al
n
eu
r
al
n
etwo
r
k
s
(
C
NN)
s
u
ch
as
r
esid
u
al
n
etwo
r
k
(
R
esNet
)
,
v
is
u
al
g
eo
m
etr
y
g
r
o
u
p
(
VGG
)
,
an
d
I
n
ce
p
tio
n
,
h
av
e
d
em
o
n
s
tr
ated
s
tr
o
n
g
p
er
f
o
r
m
a
n
ce
in
im
a
g
e
class
if
icatio
n
b
y
au
to
m
ati
n
g
f
ea
tu
r
e
ex
tr
ac
tio
n
[
1
1
]
–
[
1
3
]
.
Ho
we
v
er
,
th
eir
ef
f
ec
tiv
en
ess
o
f
ten
d
ec
lin
es
w
h
en
ap
p
lied
to
h
ig
h
ly
v
a
r
iab
le
ag
r
icu
ltu
r
al
d
atasets
d
u
e
to
f
ea
tu
r
e
r
ed
u
n
d
an
c
y
,
class
o
v
er
lap
,
lim
ited
lab
eled
d
ata,
an
d
o
v
er
f
itti
n
g
is
s
u
es
[
1
4
]
–
[
1
6
]
.
Den
s
eNe
t
-
1
2
1
ad
d
r
ess
es th
ese
l
im
itatio
n
s
th
r
o
u
g
h
e
f
f
icien
t
f
ea
t
u
r
e
r
eu
s
e,
im
p
r
o
v
ed
g
r
ad
ien
t
f
lo
w,
an
d
r
ed
u
ce
d
p
ar
am
eter
co
m
p
lex
ity
,
m
ak
in
g
it
well
-
s
u
ited
f
o
r
h
eter
o
g
en
e
o
u
s
ag
r
icu
ltu
r
al
d
ata
[
1
7
]
–
[
1
9
]
.
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.
15
,
No
.
3
,
J
u
n
e
20
26
:
2
5
2
1
-
2
5
3
2
2522
T
h
is
s
tu
d
y
e
v
alu
ates
th
e
ap
p
licatio
n
o
f
Den
s
eNe
t
-
1
2
1
c
o
m
b
in
ed
with
d
ata
au
g
m
en
t
atio
n
an
d
f
in
e
-
tu
n
in
g
s
tr
ateg
ies
f
o
r
t
o
m
ato
leaf
d
is
ea
s
e
class
if
icatio
n
u
s
in
g
th
e
Far
m
Villag
e
d
ataset
[
2
0
]
–
[
2
2
]
.
R
ath
er
th
an
p
r
o
p
o
s
in
g
a
n
ew
ar
c
h
itectu
r
e,
th
e
r
esear
c
h
f
o
c
u
s
es
o
n
em
p
ir
ically
a
n
aly
zin
g
h
o
w
p
r
ep
r
o
ce
s
s
in
g
an
d
tr
an
s
f
er
lear
n
in
g
s
tag
es
en
h
an
ce
class
if
icat
io
n
p
er
f
o
r
m
an
ce
u
n
d
er
r
ea
lis
tic
ag
r
icu
ltu
r
al
co
n
d
itio
n
s
.
T
h
e
r
esu
lts
d
em
o
n
s
tr
ate
s
tr
o
n
g
ad
a
p
tab
ilit
y
to
d
ataset
v
a
r
iab
ilit
y
an
d
i
n
ter
-
class
s
im
ilar
ity
,
alth
o
u
g
h
f
u
r
th
e
r
v
alid
atio
n
u
s
in
g
r
ea
l
-
f
ield
im
a
g
es is
n
ec
ess
ar
y
to
co
n
f
ir
m
r
o
b
u
s
tn
ess
in
p
r
ac
tical
f
ar
m
in
g
en
v
ir
o
n
m
e
n
ts
.
2.
M
E
T
H
O
D
T
h
e
p
r
o
p
o
s
ed
m
eth
o
d
o
l
o
g
y
f
o
r
to
m
ato
lea
f
d
is
ea
s
e
d
etec
tio
n
u
s
in
g
Den
s
eNe
t
-
1
2
1
with
d
ata
au
g
m
en
tatio
n
an
d
f
in
e
-
tu
n
in
g
co
n
s
is
ts
o
f
f
iv
e
m
ain
s
tag
es,
as
illu
s
tr
ated
in
Fig
u
r
e
1
,
d
ata
c
o
llectio
n
,
ex
p
lo
r
ato
r
y
d
ata
an
aly
s
is
(
E
DA)
,
d
ata
p
r
ep
r
o
ce
s
s
in
g
,
m
o
d
el
d
ev
elo
p
m
e
n
t,
an
d
m
o
d
el
ev
alu
atio
n
[
2
3
]
.
T
h
e
p
r
o
ce
s
s
b
eg
in
s
with
s
y
s
tem
atic
d
ata
co
llectio
n
,
f
o
llo
wed
b
y
d
ataset
an
aly
s
is
to
d
ete
r
m
in
e
ap
p
r
o
p
r
iate
p
r
ep
r
o
ce
s
s
in
g
an
d
m
o
d
elin
g
s
tr
ateg
ies.
Data
p
r
ep
r
o
ce
s
s
in
g
i
s
th
en
ap
p
lied
to
e
n
s
u
r
e
c
o
n
s
is
ten
cy
an
d
q
u
ality
.
Su
b
s
eq
u
en
tly
,
th
e
Den
s
eNe
t
-
1
2
1
m
o
d
el
is
tr
ain
ed
a
n
d
test
ed
,
an
d
th
e
r
esu
lts
ar
e
c
o
m
p
r
e
h
en
s
iv
ely
ev
alu
ate
d
to
ass
ess
d
etec
tio
n
p
er
f
o
r
m
an
ce
an
d
m
o
d
el
r
o
b
u
s
tn
ess
.
T
h
is
s
tr
u
ctu
r
ed
wo
r
k
f
lo
w
en
s
u
r
es
s
y
s
tem
atic
ex
p
er
im
en
tatio
n
an
d
en
ab
les r
ep
r
o
d
u
cib
ilit
y
o
f
t
h
e
p
r
o
p
o
s
ed
ap
p
r
o
ac
h
.
Fig
u
r
e
1
.
Me
th
o
d
o
lo
g
y
o
f
th
e
p
r
o
p
o
s
ed
Den
s
eNe
t
-
121
with
d
ata
au
g
m
e
n
tatio
n
2
.
1
.
Da
t
a
c
o
llect
io
n
T
h
e
d
ataset
u
tili
ze
d
in
th
is
r
esear
ch
was
s
o
u
r
ce
d
f
r
o
m
[
2
4
]
,
th
e
Plan
tVillag
e
r
ep
o
s
ito
r
y
s
h
o
w
in
Fig
u
r
e
2
,
wh
ich
is
wid
ely
r
ec
o
g
n
ized
f
o
r
its
co
m
p
r
eh
en
s
iv
e
an
d
h
ig
h
-
q
u
ality
p
lan
t
d
is
ea
s
e
im
ag
e
c
o
llectio
n
s
.
Fig
u
r
e
2
(
a
)
s
h
o
ws
to
m
ato
tar
g
et
s
p
o
t,
Fig
u
r
e
2
(
b
)
s
h
o
ws
to
m
ato
m
o
za
ic
v
ir
u
s
,
Fig
u
r
e
2
(
c)
s
h
o
ws
to
m
at
o
y
ello
w
leaf
c
u
r
l
v
ir
u
s
,
Fig
u
r
e
2
(
d
)
s
h
o
ws
t
o
m
ato
b
ac
ter
i
al
s
p
o
t,
Fig
u
r
e
2
(
e)
s
h
o
ws
t
o
m
ato
ea
r
ly
b
lig
h
t,
Fig
u
r
e
2
(
f
)
s
h
o
ws
to
m
ato
h
ea
l
th
y
,
Fig
u
r
e
2
(
g
)
s
h
o
ws
to
m
ato
late
b
lig
h
t,
Fig
u
r
e
2
(
h
)
s
h
o
ws
to
m
ato
leaf
m
o
ld
,
Fig
u
r
e
2
(
i)
s
h
o
ws
to
m
ato
s
ep
to
r
ia
leaf
s
p
o
t,
an
d
Fig
u
r
e
2
(
j
)
s
h
o
ws
to
m
ato
s
p
id
er
m
ites
two
s
p
o
tted
s
p
id
er
m
ite.
E
ac
h
ca
teg
o
r
y
ex
h
ib
it
s
d
is
tin
ct
v
is
u
al
ch
ar
ac
ter
is
t
ics
s
u
ch
as
lesi
o
n
s
,
d
is
co
lo
r
atio
n
,
d
ef
o
r
m
atio
n
p
atter
n
s
,
o
r
th
e
a
b
s
en
ce
o
f
s
y
m
p
to
m
s
in
h
ea
lth
y
leav
es.
T
h
i
s
d
etailed
ca
teg
o
r
izatio
n
en
s
u
r
es d
ataset
s
u
itab
ilit
y
f
o
r
m
u
lti
class
class
if
icatio
n
t
ask
s
an
d
en
ab
les
th
e
m
o
d
el
t
o
lear
n
s
u
b
tle
in
ter
-
class
d
if
f
er
en
ce
s
ef
f
ec
tiv
ely
.
Sp
ec
if
ically
,
th
is
r
esear
ch
f
o
cu
s
ed
o
n
a
s
u
b
s
et
o
f
th
e
d
ataset
co
n
tain
in
g
5
,
0
0
0
lab
eled
im
ag
es
o
f
to
m
ato
leav
es
with
s
ize
2
5
6
×
2
5
6
p
ix
e
l.
T
h
ese
im
ag
es
ar
e
d
is
tr
ib
u
ted
ac
r
o
s
s
1
0
d
is
tin
ct
ca
teg
o
r
ies,
ea
ch
r
ep
r
esen
tin
g
a
u
n
iq
u
e
co
n
d
itio
n
o
r
d
is
ea
s
e
af
f
ec
tin
g
to
m
ato
p
la
n
ts
.
T
h
e
d
ataset
in
clu
d
es
b
o
th
d
is
ea
s
ed
an
d
h
ea
lth
y
leav
es,
p
r
o
v
id
i
n
g
a
b
alan
ce
d
r
ep
r
esen
tatio
n
o
f
th
e
c
h
allen
g
es
f
ac
ed
in
to
m
ato
cr
o
p
m
an
ag
em
e
n
t.
T
h
e
i
m
ag
es
ar
e
ca
p
tu
r
ed
in
v
a
r
io
u
s
lig
h
tin
g
c
o
n
d
itio
n
s
an
d
o
r
ien
tatio
n
s
,
ad
d
in
g
d
iv
er
s
ity
an
d
r
ea
lis
m
to
th
e
d
ataset,
wh
ich
is
cr
u
cial
f
o
r
t
r
ain
in
g
a
r
o
b
u
s
t c
lass
if
icatio
n
m
o
d
el.
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
To
ma
to
lea
f d
is
ea
s
e
cla
s
s
ifica
tio
n
u
s
in
g
Den
s
eNet
-
1
2
1
w
ith
Da
ta
a
u
g
me
n
ta
tio
n
…
(
S
u
fa
j
a
r
B
u
ts
ia
n
to
)
2523
(
a)
(
b
)
(
c)
(
d
)
(
e)
(f)
(
g
)
(
h
)
(
i)
(
j)
Fig
u
r
e
2
.
Sam
p
le
d
ataset
to
m
a
to
leaf
d
is
ea
s
e
of
(
a)
to
m
ato
ta
r
g
et
s
p
o
t,
(
b
)
to
m
at
o
m
o
za
ic
v
ir
u
s
,
(
c)
to
m
at
o
y
ello
w
leaf
cu
r
l
v
ir
u
s
,
(
d
)
to
m
ato
b
ac
ter
ial
s
p
o
t,
(
e
)
to
m
ato
e
ar
ly
b
lig
h
t,
(
f
)
t
o
m
ato
h
ea
lth
y
,
(
g
)
to
m
ato
late
b
lig
h
t
,
(
h
)
to
m
ato
leaf
m
o
l
d
,
(
i
)
to
m
ato
s
ep
to
r
ia
lea
f
s
p
o
t,
an
d
(
j)
to
m
at
o
s
p
id
er
m
ites
two
s
p
o
tted
s
p
id
er
m
ite
2
.
2
.
E
x
plo
ra
t
i
o
n da
t
a
a
na
ly
s
is
T
h
e
E
DA
p
h
ase
was
co
n
d
u
cte
d
to
g
ai
n
d
ee
p
er
in
s
ig
h
ts
in
to
t
h
e
d
is
tr
ib
u
tio
n
an
d
ch
ar
ac
ter
is
tics
o
f
th
e
d
ataset
[
2
5
]
.
On
e
o
f
th
e
k
ey
an
aly
s
es
in
v
o
lv
ed
e
x
am
in
in
g
th
e
n
u
m
b
er
o
f
im
ag
es
av
ailab
le
f
o
r
ea
ch
d
is
ea
s
e
ca
teg
o
r
y
.
T
h
is
an
aly
s
is
r
ev
ea
led
s
lig
h
t
v
ar
iatio
n
s
in
th
e
d
a
taset
clas
s
d
is
tr
ib
u
tio
n
,
as
d
ep
icted
in
Fig
u
r
e
3
.
W
h
ile
th
e
d
ataset
is
g
en
er
ally
well
-
b
alan
ce
d
,
s
o
m
e
ca
teg
o
r
i
es
h
av
e
m
a
r
g
in
ally
f
ewe
r
s
a
m
p
les,
wh
ich
c
o
u
ld
in
f
lu
en
ce
th
e
m
o
d
el
lear
n
in
g
p
r
o
ce
s
s
an
d
o
v
er
all
p
er
f
o
r
m
a
n
ce
.
Un
d
e
r
s
tan
d
in
g
t
h
ese
d
is
cr
ep
an
cies
is
cr
u
cial
to
im
p
lem
en
tin
g
s
tr
ateg
ies
lik
e
d
ata
au
g
m
en
tatio
n
o
r
class
weig
h
tin
g
to
m
itig
ate
p
o
ten
tial
b
iases
in
th
e
m
o
d
el.
T
h
e
d
is
tr
ib
u
tio
n
o
f
im
ag
es
ac
r
o
s
s
d
is
ea
s
e
ca
teg
o
r
ies
is
illu
s
tr
ated
in
Fig
u
r
e
3
.
T
h
e
d
ata
s
et
s
h
o
ws
g
en
er
ally
b
alan
ce
d
class
f
r
eq
u
en
cies,
wh
er
e
m
o
s
t
ca
teg
o
r
i
es,
in
clu
d
in
g
leaf
m
o
ld
as
s
h
o
w
in
Fig
u
r
e
3
(
a
)
,
tar
g
et
s
p
o
t
as
s
h
o
w
n
in
Fig
u
r
e
3
(
b
)
,
h
ea
lth
y
lea
v
es
as
s
h
o
wn
in
Fig
u
r
e
3
(
c)
,
a
n
d
late
b
lig
h
t
as
s
h
o
wn
in
Fig
u
r
e
3
(
d
)
co
n
tain
a
p
p
r
o
x
im
ately
5
1
3
–
5
1
5
im
a
g
es.
Ad
d
iti
o
n
al
ca
teg
o
r
ies
s
u
ch
as
s
p
id
er
m
ites
as
s
h
o
wn
in
Fig
u
r
e
3
(
e)
,
ea
r
ly
b
lig
h
t
as
s
h
o
wn
in
Fig
u
r
e
3
(
f
)
,
b
ac
ter
ial
s
p
o
t
as
s
h
o
wn
in
Fig
u
r
e
3
(
g
)
,
s
ep
to
r
ia
leaf
s
p
o
t
as
s
h
o
wn
in
Fig
u
r
e
3
(
h
)
,
an
d
y
ello
w
leaf
cu
r
l
v
ir
u
s
as
s
h
o
wn
in
Fig
u
r
e
3
(
i)
also
p
r
esen
t
co
m
p
a
r
ab
le
d
is
tr
ib
u
tio
n
s
.
T
h
e
to
m
at
o
m
o
za
ic
v
ir
u
s
class
as
s
h
o
wn
in
Fig
u
r
e
3
(
j)
c
o
n
tain
s
f
ewe
r
s
a
m
p
les
3
7
3
im
ag
es,
in
d
icatin
g
m
ild
class
im
b
alan
ce
th
at
m
o
tiv
ates
th
e
u
s
e
o
f
p
r
ep
r
o
ce
s
s
in
g
an
d
d
ata
au
g
m
en
tatio
n
to
s
u
p
p
o
r
t
r
o
b
u
s
t le
ar
n
in
g
an
d
g
en
e
r
aliza
tio
n
.
Fig
u
r
e
3
.
Fre
q
u
en
cy
o
f
ea
c
h
to
m
ato
d
is
ea
s
e
ty
p
e
f
o
r
(
a)
t
o
m
a
to
leaf
m
o
ld
,
(
b
)
to
m
ato
tar
g
et
s
p
o
t,
(
c)
to
m
ato
h
ea
lth
y
,
(
d
)
to
m
at
o
late
b
lig
h
t,
(
e)
to
m
ato
s
p
id
er
m
ites
two
s
p
o
tted
s
p
id
er
m
ite,
(
f
)
t
o
m
ato
ea
r
ly
b
lig
h
t,
(
g
)
to
m
at
o
b
ac
ter
ial
s
p
o
t,
(
h
)
t
o
m
ato
s
ep
to
r
ia
leaf
s
p
o
t,
(
i)
t
o
m
ato
y
ello
w
leaf
cu
r
l v
ir
u
s
,
a
n
d
(
j)
to
m
ato
m
o
za
ic
v
ir
u
s
T
ab
le
1
p
r
esen
ts
th
e
f
in
al
d
ata
s
et
s
tr
u
ctu
r
e,
co
n
s
is
tin
g
o
f
th
r
ee
co
lu
m
n
s
:
f
ile
,
Dis
ea
s
eI
D,
a
n
d
d
is
ea
s
e
ty
p
e
.
T
h
e
f
ile
co
lu
m
n
co
n
tain
s
r
elativ
e
im
ag
e
p
ath
s
,
th
e
Dis
e
aseI
D
co
lu
m
n
ass
ig
n
s
n
u
m
er
ical
lab
els
(
0
–
9
)
f
o
r
ef
f
icien
t
m
o
d
el
tr
ain
in
g
,
an
d
th
e
d
is
ea
s
e
ty
p
e
c
o
lu
m
n
p
r
o
v
id
es
in
ter
p
r
etab
le
class
n
a
m
es.
Sam
p
le
en
tr
ies
s
h
o
wn
in
T
ab
le
1
d
em
o
n
s
tr
ate
im
ag
es
f
r
o
m
th
e
to
m
ato
h
ea
lt
h
y
ca
teg
o
r
y
,
illu
s
tr
atin
g
a
wel
l
-
o
r
g
an
ize
d
d
ataset
f
o
r
m
at
th
at
s
u
p
p
o
r
ts
s
ea
m
less
in
teg
r
atio
n
in
to
th
e
m
ac
h
in
e
le
ar
n
in
g
wo
r
k
f
lo
w.
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.
15
,
No
.
3
,
J
u
n
e
20
26
:
2
5
2
1
-
2
5
3
2
2524
T
ab
le
1
.
Sam
p
le
f
in
al
d
ataset
f
o
r
m
at
No
.
F
il
e
Dise
a
se
ID
Dise
a
se
ty
p
e
4
9
9
5
To
m
a
to
_
He
a
lt
h
y
/8
c
c
1
1
f2
4
-
0
b
7
9
-
4
e
d
a
-
8
a
0
9
-
9
c
7
7
6
e
0
.
.
.
9
To
m
a
to
h
e
a
lt
h
y
4
9
9
6
To
m
a
to
_
He
a
lt
h
y
/8
d
7
8
a
3
4
f
-
1
7
7
7
-
4
b
d
8
-
a
9
2
5
-
7
9
5
c
5
7
c
.
.
.
9
To
m
a
to
h
e
a
lt
h
y
4
9
9
7
To
m
a
to
_
He
a
lt
h
y
/9
b
8
f
1
0
8
a
-
0
1
0
b
-
4
b
8
b
-
9
8
e
a
-
b
3
0
c
b
f
6
.
.
.
9
To
m
a
to
h
e
a
lt
h
y
4
9
9
8
To
m
a
to
_
He
a
lt
h
y
/7
b
e
d
a
8
6
2
-
3
9
6
4
-
4
8
3
9
-
9
d
0
8
-
b
9
b
3
fa
2
.
.
.
9
To
m
a
to
h
e
a
lt
h
y
4
9
9
9
To
m
a
to
_
He
a
lt
h
y
/7
d
7
6
a
a
9
6
-
b
2
9
e
-
4
a
6
0
-
a
3
e
7
-
b
d
4
8
9
f
5
.
.
.
9
To
m
a
to
h
e
a
lt
h
y
2
.
3
.
Da
t
a
prep
ro
ce
s
s
ing
T
h
e
im
a
g
e
r
e
s
i
z
in
g
p
r
o
c
e
s
s
i
s
i
l
l
u
s
t
r
a
t
ed
i
n
F
i
g
u
r
e
4
.
T
h
e
o
r
ig
i
n
a
l
i
m
ag
e
r
e
s
o
l
u
t
io
n
o
f
2
5
6
×
2
5
6
p
ix
e
l
s
s
h
o
w
n
in
F
ig
u
r
e
4
(
a)
w
a
s
r
ed
u
c
e
d
t
o
6
4
×
6
4
p
i
x
e
l
s
a
s
p
r
e
s
en
t
e
d
in
F
i
g
u
r
e
4
(
b
)
.
T
h
i
s
r
e
s
i
z
i
n
g
w
a
s
a
d
o
p
t
ed
a
s
a
d
el
i
b
e
r
a
t
e
tr
a
d
e
-
o
f
f
to
r
ed
u
ce
co
m
p
u
t
a
t
io
n
a
l
co
m
p
l
e
x
i
ty
,
G
PU
m
e
m
o
r
y
u
s
a
g
e
,
a
n
d
t
r
a
i
n
in
g
t
i
m
e
wh
i
l
e
m
a
in
t
a
i
n
in
g
e
s
s
e
n
t
i
a
l
v
i
s
u
a
l
ch
ar
a
c
t
e
r
i
s
t
i
c
s
r
eq
u
ir
e
d
f
o
r
c
l
a
s
s
i
f
i
c
a
t
i
o
n
.
U
n
if
o
r
m
r
e
s
i
z
i
n
g
en
s
u
r
e
s
c
o
n
s
i
s
t
e
n
t
in
p
u
t
d
i
m
en
s
i
o
n
s
a
n
d
s
t
ab
l
e
d
ee
p
l
e
ar
n
in
g
o
p
t
i
m
i
z
a
t
io
n
.
(
a)
(
b
)
Fig
u
r
e
4
.
R
esizin
g
im
ag
e
o
f
(
a
)
b
ef
o
r
e
r
esize
an
d
(
b
)
a
f
ter
r
e
s
ize
Alth
o
u
g
h
Den
s
eNe
t
-
1
2
1
was
o
r
ig
in
ally
p
r
e
-
tr
ain
ed
o
n
I
m
ag
eNe
t
u
s
in
g
2
2
4
×
2
2
4
in
p
u
ts
,
lo
wer
-
r
eso
lu
tio
n
i
n
p
u
ts
ca
n
r
em
ain
ef
f
ec
tiv
e
wh
e
n
d
is
ea
s
e
ch
ar
ac
ter
is
tics
ar
e
p
r
im
ar
ily
d
ef
in
ed
b
y
g
lo
b
al
co
lo
r
d
is
tr
ib
u
tio
n
,
lesi
o
n
p
atter
n
s
,
an
d
s
tr
u
ctu
r
al
v
a
r
iatio
n
s
r
ath
er
th
an
ex
t
r
em
ely
f
in
e
-
g
r
a
in
ed
tex
tu
r
es
[
2
6
]
.
Un
d
er
co
n
tr
o
lled
d
ataset
co
n
d
itio
n
s
,
th
e
r
esized
im
ag
es
p
r
eser
v
e
k
ey
d
is
cr
im
in
ativ
e
v
is
u
al
cu
es
r
eq
u
ir
ed
f
o
r
to
m
ato
leaf
d
is
ea
s
e
class
if
icat
io
n
,
wh
ile
s
ig
n
if
ican
tly
r
ed
u
c
in
g
co
m
p
u
tatio
n
al
co
s
t.
I
t
is
ac
k
n
o
wled
g
e
d
th
a
t
r
ed
u
cin
g
s
p
atial
r
eso
lu
tio
n
m
ay
lim
it
th
e
r
ep
r
esen
tatio
n
o
f
s
u
b
tle
tex
tu
r
e
-
lev
el
f
ea
tu
r
es.
T
h
er
ef
o
r
e,
th
e
s
elec
ted
r
eso
lu
tio
n
s
h
o
u
ld
b
e
in
ter
p
r
eted
as
a
co
m
p
u
tatio
n
a
lly
ef
f
icien
t
co
n
f
ig
u
r
atio
n
r
at
h
er
th
an
an
o
p
tim
al
u
n
iv
er
s
al
s
ettin
g
.
A
co
m
p
ar
ati
v
e
ev
alu
ati
o
n
u
s
in
g
h
ig
h
e
r
in
p
u
t
r
eso
lu
tio
n
s
1
2
8
×1
2
8
an
d
2
2
4
×2
2
4
is
p
lan
n
e
d
as
f
u
tu
r
e
w
o
r
k
to
q
u
an
titativ
ely
ass
ess
th
e
tr
ad
e
-
o
f
f
b
etw
ee
n
co
m
p
u
tatio
n
al
e
f
f
icien
cy
an
d
f
in
e
-
g
r
ain
ed
f
ea
tu
r
e
p
r
eser
v
atio
n
.
T
h
is
cl
ar
if
icatio
n
em
p
h
asizes
th
at
th
e
p
r
esen
t
s
tu
d
y
f
o
cu
s
es
o
n
ev
alu
atin
g
tr
ain
in
g
s
tr
ateg
y
ef
f
ec
tiv
en
ess
r
ath
er
t
h
an
r
eso
lu
tio
n
o
p
tim
izatio
n
.
T
h
e
d
a
t
a
s
e
t
,
c
o
n
s
i
s
t
i
n
g
o
f
5
,
0
0
0
t
o
m
a
t
o
l
e
a
f
i
m
a
g
e
s
d
is
t
r
i
b
u
te
d
a
c
r
o
s
s
1
0
c
l
a
s
s
es
,
w
as
p
a
r
t
it
i
o
n
e
d
i
n
t
o
t
r
a
i
n
i
n
g
a
n
d
t
e
s
t
i
n
g
s
e
ts
u
s
i
n
g
a
s
t
r
a
t
i
f
ie
d
8
0
:
2
0
s
p
l
i
t
,
y
i
e
l
d
i
n
g
4
,
0
0
0
t
r
a
i
n
i
n
g
s
a
m
p
l
es
a
n
d
1
,
0
0
0
t
e
s
ti
n
g
s
a
m
p
l
e
s
.
T
h
i
s
s
t
r
a
ti
f
i
c
at
i
o
n
p
r
e
s
e
r
v
e
d
t
h
e
c
l
as
s
d
is
t
r
i
b
u
t
i
o
n
i
n
b
o
t
h
s
u
b
s
e
ts
,
e
n
s
u
r
i
n
g
a
f
a
i
r
a
n
d
u
n
b
i
ase
d
e
v
a
l
u
a
t
i
o
n
o
f
t
h
e
m
o
d
e
l
.
T
o
e
n
h
a
n
c
e
d
a
t
a
s
e
t
d
i
v
e
r
s
i
t
y
a
n
d
m
i
t
i
g
a
te
o
v
e
r
f
i
t
t
i
n
g
,
d
a
t
a
a
u
g
m
e
n
t
a
t
i
o
n
w
a
s
a
p
p
l
i
ed
e
x
c
l
u
s
i
v
e
l
y
t
o
t
h
e
t
r
a
i
n
i
n
g
s
et
u
s
i
n
g
t
h
e
I
m
a
g
eD
a
t
a
G
e
n
e
r
a
t
o
r
f
r
a
m
e
w
o
r
k
[
2
7
]
.
T
h
e
a
p
p
l
i
e
d
t
r
a
n
s
f
o
r
m
a
t
i
o
n
s
i
n
c
l
u
d
e
d
r
a
n
d
o
m
r
o
t
a
t
i
o
n
s
u
p
t
o
2
0
d
e
g
r
ee
s
,
h
o
r
i
z
o
n
t
a
l
a
n
d
v
e
r
t
i
ca
l
s
h
i
f
ts
u
p
t
o
2
0
%
,
s
h
e
a
r
i
n
g
,
z
o
o
m
i
n
g
,
a
n
d
h
o
r
i
z
o
n
t
a
l
f
li
p
p
i
n
g
.
T
h
e
s
e
a
u
g
m
e
n
t
a
t
i
o
n
s
e
m
u
l
a
t
e
r
e
a
l
-
w
o
r
l
d
v
a
r
i
a
t
i
o
n
s
i
n
l
e
a
f
a
p
p
e
a
r
a
n
c
e
,
s
u
c
h
a
s
c
h
a
n
g
e
s
i
n
o
r
ie
n
t
a
t
i
o
n
,
s
c
al
e
,
a
n
d
v
i
e
w
p
o
i
n
t
,
as
i
ll
u
s
t
r
at
e
d
i
n
Fi
g
u
r
e
5
,
t
h
e
r
e
b
y
i
m
p
r
o
v
i
n
g
t
h
e
m
o
d
e
l
’
s
a
b
i
li
t
y
t
o
g
e
n
e
r
a
li
z
e
.
T
h
e
p
ar
a
m
eter
f
ill_
m
o
d
e=
‘
n
e
ar
est’
was
u
s
ed
to
p
r
eser
v
e
v
i
s
u
al
co
n
tin
u
ity
in
tr
a
n
s
f
o
r
m
ed
im
ag
es
b
y
f
illi
n
g
n
ewly
g
en
e
r
ated
p
ix
el
r
eg
io
n
s
u
s
in
g
n
ei
g
h
b
o
r
in
g
p
ix
el
v
alu
es.
T
h
is
p
r
ev
en
ts
ar
t
if
icial
ar
tifa
cts
th
at
co
u
ld
n
e
g
ativ
ely
af
f
ec
t
f
ea
t
u
r
e
lear
n
in
g
.
Data
leak
ag
e
was
av
o
id
ed
b
y
p
er
f
o
r
m
i
n
g
d
atase
t
s
p
litt
in
g
p
r
io
r
to
au
g
m
en
tatio
n
an
d
en
s
u
r
in
g
th
at
n
o
id
en
tical
o
r
n
ea
r
-
d
u
p
licate
im
ag
es
ap
p
ea
r
ed
in
b
o
th
tr
ai
n
in
g
an
d
test
in
g
s
ets
[
2
8
]
.
T
h
e
test
in
g
d
ataset
r
em
ain
e
d
u
n
ch
a
n
g
ed
to
p
r
o
v
id
e
a
n
u
n
b
iased
e
s
tim
atio
n
o
f
m
o
d
el
g
en
er
aliza
tio
n
p
er
f
o
r
m
an
ce
.
2
.
4
.
M
o
del
Du
r
in
g
th
e
m
o
d
el
d
ev
elo
p
m
e
n
t
p
h
ase,
th
e
tr
ain
in
g
p
r
o
ce
s
s
was
in
itiated
u
s
in
g
th
e
Den
s
eNe
t
-
1
2
1
ar
ch
itectu
r
e,
as
s
h
o
wn
in
Fig
u
r
e
6
,
wh
ich
r
ep
r
esen
ts
a
m
o
d
e
r
n
C
NN.
T
h
is
ar
ch
itectu
r
e
is
d
is
tin
g
u
is
h
ed
b
y
its
d
en
s
e
co
n
n
ec
tiv
ity
m
ec
h
an
is
m
,
wh
er
e
ea
ch
lay
er
is
lin
k
ed
to
all
f
o
llo
win
g
lay
er
s
,
e
n
ab
l
in
g
ef
f
icie
n
t
f
ea
tu
r
e
p
r
o
p
a
g
atio
n
a
n
d
r
eu
s
e
[
2
9
]
.
T
h
is
d
esig
n
p
r
o
m
o
tes
ef
f
icien
t
f
ea
tu
r
e
r
eu
s
e
b
y
allo
win
g
f
ea
tu
r
es
lear
n
e
d
in
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
To
ma
to
lea
f d
is
ea
s
e
cla
s
s
ifica
tio
n
u
s
in
g
Den
s
eNet
-
1
2
1
w
ith
Da
ta
a
u
g
me
n
ta
tio
n
…
(
S
u
fa
j
a
r
B
u
ts
ia
n
to
)
2525
ea
r
lier
lay
er
s
to
b
e
p
r
o
p
ag
a
ted
th
r
o
u
g
h
o
u
t
th
e
n
etwo
r
k
,
th
er
eb
y
r
ed
u
cin
g
r
e
d
u
n
d
an
cy
an
d
im
p
r
o
v
in
g
p
ar
am
eter
ef
f
icien
c
y
[
3
0
]
.
Fu
r
th
er
m
o
r
e
,
d
en
s
e
co
n
n
ec
tio
n
s
f
ac
ilit
ate
g
r
ad
ien
t
p
r
o
p
a
g
atio
n
d
u
r
in
g
b
ac
k
p
r
o
p
ag
atio
n
,
e
f
f
ec
tiv
ely
m
itig
atin
g
th
e
v
an
is
h
in
g
g
r
a
d
ien
t
p
r
o
b
lem
co
m
m
o
n
ly
en
co
u
n
ter
ed
in
d
ee
p
n
eu
r
al
n
etwo
r
k
s
[
3
1
]
.
Fig
u
r
e
5
.
R
esu
lt o
f
au
g
m
en
tati
o
n
Fig
u
r
e
6
.
Den
s
eNe
t
-
1
2
1
f
lo
w
p
r
o
ce
s
s
T
h
e
ar
ch
itectu
r
e
co
n
s
is
ts
o
f
1
2
1
lay
e
r
s
o
r
g
a
n
ized
in
t
o
f
o
u
r
d
e
n
s
e
b
lo
ck
s
s
ep
ar
ated
b
y
tr
an
s
itio
n
lay
er
s
co
n
tain
in
g
co
n
v
o
lu
tio
n
,
p
o
o
li
n
g
,
a
n
d
b
atch
n
o
r
m
ali
za
tio
n
o
p
e
r
atio
n
s
.
Ma
th
em
ati
ca
lly
,
th
e
in
p
u
t
to
lay
er
ca
n
b
e
r
ep
r
esen
ted
as
(
1
)
[
3
2
]
.
1
=
1
(
[
0
,
1
,
…
,
−
1
]
)
(
1
)
W
h
er
e
1
is
th
e
tr
an
s
f
o
r
m
ati
o
n
(
b
atc
h
n
o
r
m
aliza
tio
n
,
r
e
ctif
ied
lin
ea
r
u
n
it
(
R
eL
U
)
ac
tiv
atio
n
,
an
d
co
n
v
o
l
u
tio
n
,
an
d
[
0
,
1
,
…
,
−
1
]
r
ep
r
esen
t
s
th
e
co
n
ca
ten
ated
o
u
t
p
u
ts
o
f
all
p
r
ec
ed
in
g
lay
e
r
s
.
T
h
is
m
ec
h
an
is
m
allo
ws th
e
m
o
d
el
t
o
ef
f
icien
tly
lear
n
b
o
t
h
lo
w
-
le
v
el
an
d
h
ig
h
-
lev
el
f
ea
tu
r
es si
m
u
ltan
eo
u
s
ly
.
T
o
ac
ce
ler
ate
co
n
v
er
g
e
n
ce
an
d
lev
er
ag
e
p
r
io
r
v
is
u
al
k
n
o
wl
ed
g
e,
Den
s
eNe
t
-
1
2
1
p
r
e
-
tr
ain
ed
weig
h
ts
f
r
o
m
th
e
I
m
ag
eNe
t
d
ataset
wer
e
em
p
lo
y
e
d
[
3
3
]
.
T
r
an
s
f
er
lear
n
in
g
p
r
o
v
id
es
a
s
tr
o
n
g
in
itializatio
n
b
y
in
co
r
p
o
r
atin
g
g
en
e
r
ic
v
is
u
al
f
ea
tu
r
es
s
u
ch
as
ed
g
es,
tex
tu
r
es,
an
d
s
h
ap
es
co
m
m
o
n
ly
s
h
ar
ed
ac
r
o
s
s
im
ag
e
d
atasets
[
3
4
]
.
T
h
e
p
r
e
-
tr
ain
ed
n
etwo
r
k
was
s
u
b
s
eq
u
e
n
t
ly
f
in
e
-
tu
n
e
d
to
a
d
ap
t
f
ea
t
u
r
e
r
ep
r
esen
tatio
n
s
s
p
ec
if
ically
to
to
m
ato
leaf
d
is
ea
s
e
ch
ar
ac
ter
is
tics
wh
ile
p
r
ese
r
v
in
g
g
en
er
aliza
b
le
v
is
u
al
p
atter
n
s
[
3
5
]
.
All
ex
p
er
im
en
ts
wer
e
co
n
d
u
cted
u
s
in
g
T
en
s
o
r
Flo
w
v
2
.
1
7
.
0
with
th
e
Ker
as
API
v
3
.
4
.
1
in
a
Py
th
o
n
3
.
1
0
e
n
v
ir
o
n
m
en
t.
M
o
d
el
tr
ain
in
g
was
p
er
f
o
r
m
e
d
o
n
Go
o
g
le
C
o
lab
with
an
NVI
D
I
A
T
esla
T
4
GPU.
T
o
en
s
u
r
e
r
ep
r
o
d
u
cib
ilit
y
,
a
f
ix
ed
r
an
d
o
m
s
ee
d
was
co
n
s
is
ten
tly
ap
p
lied
ac
r
o
s
s
Nu
m
Py
,
T
en
s
o
r
Flo
w,
an
d
Py
th
o
n
’
s
r
an
d
o
m
m
o
d
u
le
d
u
r
i
n
g
d
ata
s
p
litt
in
g
an
d
m
o
d
el
in
i
tializatio
n
.
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.
15
,
No
.
3
,
J
u
n
e
20
26
:
2
5
2
1
-
2
5
3
2
2526
T
h
e
Ad
am
o
p
tim
izer
was
u
tili
ze
d
f
o
r
its
ad
ap
tiv
e
lear
n
in
g
r
ate
ca
p
ab
ilit
y
an
d
m
o
m
en
tu
m
-
b
ased
u
p
d
ate
m
ec
h
an
is
m
,
wh
ich
f
ac
i
litate
s
tab
le
an
d
ef
f
icien
t
m
o
d
el
co
n
v
er
g
e
n
ce
[
3
6
]
.
T
h
e
in
itial
lear
n
in
g
r
ate
was
s
et
to
1
×1
0
-
3
,
b
alan
cin
g
co
n
v
e
r
g
en
ce
s
p
ee
d
an
d
tr
ain
in
g
s
tab
ilit
y
wh
ile
p
r
ev
e
n
tin
g
a
b
r
u
p
t
p
ar
am
eter
u
p
d
ates.
T
h
e
d
etailed
tr
ain
in
g
co
n
f
ig
u
r
atio
n
is
s
u
m
m
ar
ized
in
T
a
b
le
2
.
T
o
en
s
u
r
e
u
n
b
iased
ev
alu
at
io
n
,
th
e
d
ataset
was
s
p
lit
u
s
in
g
a
s
tr
atif
ied
8
0
:2
0
t
r
ain
–
test
ap
p
r
o
ac
h
,
with
a
v
al
id
atio
n
s
u
b
s
et
d
er
i
v
ed
f
r
o
m
th
e
tr
ain
in
g
d
ata
to
m
o
n
ito
r
lear
n
in
g
a
n
d
p
r
ev
e
n
t
o
v
e
r
f
itti
n
g
.
Alth
o
u
g
h
k
-
f
o
ld
cr
o
s
s
-
v
alid
atio
n
was
n
o
t
ap
p
lied
d
u
e
t
o
co
m
p
u
tatio
n
al
c
o
n
s
tr
ain
ts
,
co
n
s
is
ten
t
in
itializatio
n
,
an
d
co
n
tr
o
lled
ex
p
er
im
en
tal
s
ettin
g
s
wer
e
u
s
ed
to
m
ain
tain
r
esu
lt
r
eliab
ilit
y
.
T
h
e
t
r
ai
n
i
n
g
p
r
o
ce
s
s
w
as
c
o
n
d
u
ct
e
d
in
t
h
r
e
e
s
ta
g
es:
b
as
eli
n
e
t
r
a
i
n
i
n
g
f
o
r
3
0
e
p
o
c
h
s
,
f
o
llo
wed
b
y
2
0
ep
o
ch
s
with
d
ata
au
g
m
e
n
tatio
n
to
im
p
r
o
v
e
g
en
er
aliza
tio
n
,
an
d
a
f
i
n
al
2
0
-
ep
o
c
h
f
in
e
-
tu
n
in
g
s
tag
e
b
y
u
n
f
r
ee
zin
g
all
lay
er
s
,
r
esu
ltin
g
in
im
p
r
o
v
ed
r
o
b
u
s
tn
ess
an
d
class
if
icatio
n
p
er
f
o
r
m
an
ce
.
T
ab
le
2
.
T
r
ai
n
in
g
h
y
p
e
r
p
ar
am
eter
s
an
d
ex
p
e
r
im
en
tal
s
etu
p
P
a
r
a
me
t
e
r
V
a
l
u
e
I
n
p
u
t
i
m
a
g
e
s
i
z
e
6
4
×
6
4
p
i
x
e
l
s
B
a
t
c
h
si
z
e
32
O
p
t
i
mi
z
e
r
A
d
a
m
I
n
i
t
i
a
l
l
e
a
r
n
i
n
g
r
a
t
e
1
,
0
0
E
-
03
Le
a
r
n
i
n
g
r
a
t
e
s
t
r
a
t
e
g
y
C
o
n
st
a
n
t
Ep
o
c
h
s (
b
a
se
l
i
n
e
t
r
a
i
n
i
n
g
)
30
Ep
o
c
h
s (w
i
t
h
a
u
g
me
n
t
a
t
i
o
n
)
20
Ep
o
c
h
s (f
i
n
e
-
t
u
n
i
n
g
)
20
Lo
ss f
u
n
c
t
i
o
n
C
a
t
e
g
o
r
i
c
a
l
c
r
o
ss
-
e
n
t
r
o
p
y
D
a
t
a
a
u
g
m
e
n
t
a
t
i
o
n
R
o
t
a
t
i
o
n
(
±
2
0
°
)
,
w
i
d
t
h
/
h
e
i
g
h
t
s
h
i
f
t
(
2
0
%),
z
o
o
m
,
s
h
e
a
r
,
a
n
d
h
o
r
i
z
o
n
t
a
l
f
l
i
p
A
u
g
m
e
n
t
a
t
i
o
n
a
p
p
l
i
e
d
t
o
Tr
a
i
n
i
n
g
se
t
o
n
l
y
2
.
5
.
M
o
del e
v
a
lua
t
i
o
n
Mo
d
el
ev
alu
atio
n
was
in
itiated
b
y
m
ea
s
u
r
in
g
p
er
f
o
r
m
an
ce
o
n
th
e
tr
ain
in
g
d
ataset
u
s
in
g
m
etr
ics
s
u
ch
as
ac
cu
r
ac
y
,
p
r
ec
is
io
n
,
r
ec
all,
an
d
F1
-
s
co
r
e
t
o
en
s
u
r
e
ef
f
e
ctiv
e
f
ea
tu
r
e
lear
n
in
g
ac
r
o
s
s
all
ten
to
m
ato
leaf
d
is
ea
s
e
clas
s
es.
T
r
ain
in
g
lo
s
s
an
d
ac
cu
r
ac
y
tr
en
d
s
wer
e
m
o
n
ito
r
ed
t
o
en
s
u
r
e
s
tab
le
lear
n
in
g
an
d
to
d
etec
t
p
o
ten
tial
u
n
d
er
f
itti
n
g
o
r
o
v
er
f
itti
n
g
.
Su
b
s
eq
u
en
tly
,
m
o
d
el
g
en
er
aliza
tio
n
was
ev
alu
ated
o
n
th
e
test
in
g
d
ataset
co
n
tain
in
g
u
n
s
ee
n
s
am
p
les.
T
h
is
ev
alu
atio
n
an
aly
ze
d
p
er
f
o
r
m
an
ce
g
ap
s
b
etwe
en
tr
ain
in
g
an
d
test
in
g
r
esu
lts
to
id
en
tif
y
o
v
er
f
itti
n
g
o
r
u
n
d
er
f
itti
n
g
an
d
e
m
p
h
asized
class
-
wis
e
ac
cu
r
ac
y
to
en
s
u
r
e
b
al
an
ce
d
class
if
icatio
n
ac
r
o
s
s
all
d
is
ea
s
e
ca
teg
o
r
ies.
B
ased
o
n
th
ese
ev
alu
atio
n
s
,
r
etr
ain
in
g
was
p
er
f
o
r
m
e
d
u
s
in
g
Den
s
eNe
t
-
121
with
d
ata
au
g
m
en
tatio
n
an
d
f
in
e
-
tu
n
in
g
to
e
n
h
an
ce
r
o
b
u
s
tn
ess
an
d
ad
ap
ta
b
ilit
y
[
1
4
]
.
Data
au
g
m
en
tatio
n
was
u
s
ed
to
s
im
u
late
r
ea
lis
tic
v
ar
iatio
n
s
in
t
h
e
d
ata,
wh
il
e
f
in
e
-
tu
n
in
g
en
ab
led
th
e
m
o
d
el
to
b
etter
lear
n
d
o
m
ain
-
s
p
ec
if
ic
d
is
ea
s
e
ch
ar
ac
ter
is
tics
.
A
co
n
f
u
s
io
n
m
atr
ix
was
th
en
u
tili
ze
d
f
o
r
d
etailed
class
-
lev
el
ev
alu
atio
n
b
y
s
u
m
m
ar
izin
g
tr
u
e
p
o
s
itiv
es,
tr
u
e
n
e
g
ativ
es,
f
al
s
e
p
o
s
itiv
es,
an
d
f
alse
n
e
g
ativ
es
,
wh
ich
s
er
v
e
as
th
e
b
a
s
is
f
o
r
ca
lcu
latin
g
p
r
ec
is
io
n
,
r
ec
all,
an
d
F1
-
s
co
r
e
,
as we
ll a
s
f
o
r
id
en
tify
in
g
s
tr
e
n
g
th
s
an
d
wea
k
n
ess
es a
cr
o
s
s
class
es
[
1
4
]
.
T
h
e
co
n
f
u
s
io
n
m
atr
ix
also
s
er
v
ed
as
a
d
iag
n
o
s
tic
to
o
l
to
u
n
co
v
er
p
atter
n
s
o
f
m
is
class
if
ic
atio
n
.
Fo
r
in
s
tan
ce
,
f
r
eq
u
e
n
t
m
is
class
if
i
ca
tio
n
b
etwe
en
s
im
ilar
class
e
s
,
s
u
ch
as
to
m
ato
late
b
lig
h
t
an
d
to
m
ato
ea
r
ly
b
lig
h
t,
m
ig
h
t
in
d
icate
o
v
er
la
p
p
in
g
v
is
u
al
f
ea
tu
r
es
th
at
r
eq
u
ir
e
ad
d
itio
n
al
au
g
m
en
tati
o
n
o
r
f
in
e
-
tu
n
in
g
.
Fu
r
th
e
r
m
o
r
e
,
a
n
o
r
m
ali
ze
d
c
o
n
f
u
s
io
n
m
atr
ix
w
as
u
s
e
d
t
o
v
is
u
aliz
e
t
h
e
c
lass
i
f
i
ca
t
io
n
ac
c
u
r
ac
y
f
o
r
ea
c
h
c
ate
g
o
r
y
as
p
e
r
c
en
ta
g
es
,
m
ak
in
g
it
e
as
ier
t
o
i
d
e
n
ti
f
y
u
n
d
e
r
p
e
r
f
o
r
m
i
n
g
class
es.
T
h
is
m
a
tr
i
x
,
wh
e
n
co
m
b
i
n
e
d
wit
h
th
e
d
e
r
i
v
e
d
m
et
r
ics
,
li
n
k
e
d
t
h
e
q
u
a
litat
iv
e
in
s
ig
h
ts
w
it
h
q
u
a
n
ti
tat
i
v
e
e
v
i
d
e
n
ce
,
an
d
o
f
f
e
r
i
n
g
ac
ti
o
n
a
b
l
e
s
t
r
ate
g
i
es
f
o
r
im
p
r
o
v
e
m
e
n
t
.
I
t
als
o
v
al
id
ate
d
t
h
e
m
o
d
el'
s
r
o
b
u
s
t
n
ess
b
y
d
e
m
o
n
s
tr
ati
n
g
c
o
n
s
is
t
en
t
p
e
r
f
o
r
m
a
n
ce
ac
r
o
s
s
al
l
d
is
e
ase
c
ate
g
o
r
ies
,
ess
e
n
ti
al
f
o
r
en
s
u
r
i
n
g
r
e
lia
b
il
it
y
in
r
ea
l
-
w
o
r
ld
t
o
m
at
o
le
af
d
is
e
ase
d
et
ec
t
i
o
n
s
c
e
n
a
r
i
o
s
.
3.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
N
Fig
u
r
e
7
s
h
o
ws
t
h
e
r
esu
lts
d
e
m
o
n
s
tr
ate
th
e
i
n
f
lu
en
ce
o
f
p
r
e
p
r
o
ce
s
s
in
g
an
d
tr
ain
i
n
g
s
tr
ate
g
ies
o
n
th
e
p
er
f
o
r
m
an
ce
o
f
Den
s
eNe
t
-
1
2
1
f
o
r
t
o
m
ato
leaf
d
is
ea
s
e
class
if
icatio
n
.
I
n
itially
,
im
ag
es
wer
e
r
esized
f
r
o
m
2
5
6
×2
5
6
to
6
4
×6
4
p
ix
els
to
r
ed
u
ce
co
m
p
u
tatio
n
al
co
s
t
wh
ile
r
etain
in
g
k
ey
v
is
u
al
f
ea
tu
r
es.
Usi
n
g
th
e
Ad
am
o
p
tim
izer
with
a
m
i
n
im
u
m
lear
n
in
g
r
ate
o
f
1
×1
0
-
3
,
th
e
ac
cu
r
ac
y
c
u
r
v
e
s
h
o
wn
in
Fig
u
r
e
7
(
a)
,
in
d
icate
s
th
at
th
e
m
o
d
el
ac
h
iev
e
d
a
n
in
itial
v
alid
atio
n
ac
c
u
r
ac
y
o
f
7
5
.
5
2
%,
wh
ile
th
e
lo
s
s
p
r
o
g
r
es
s
io
n
in
Fig
u
r
e
7
(
b
)
s
h
o
ws
a
co
r
r
esp
o
n
d
in
g
v
alid
atio
n
lo
s
s
o
f
3
7
.
7
7
%.
T
h
is
b
aselin
e
p
er
f
o
r
m
an
ce
d
em
o
n
s
t
r
ates
th
e
ab
ilit
y
o
f
Den
s
eNe
t
-
121
to
ex
t
r
ac
t f
ea
tu
r
es e
f
f
ec
tiv
ely
ev
e
n
f
r
o
m
r
esiz
ed
im
ag
es,
alb
eit
with
r
o
o
m
f
o
r
im
p
r
o
v
em
en
t.
T
o
en
h
an
ce
m
o
d
el
p
er
f
o
r
m
a
n
ce
an
d
r
ed
u
ce
o
v
er
f
itti
n
g
,
d
at
a
au
g
m
e
n
tatio
n
was
i
n
co
r
p
o
r
a
ted
d
u
r
in
g
tr
ain
in
g
.
T
h
e
ap
p
lied
tec
h
n
iq
u
es
in
clu
d
ed
r
a
n
d
o
m
r
o
tatio
n
s
,
tr
an
s
latio
n
s
,
s
h
ea
r
in
g
,
zo
o
m
in
g
,
an
d
h
o
r
izo
n
tal
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
To
ma
to
lea
f d
is
ea
s
e
cla
s
s
ifica
tio
n
u
s
in
g
Den
s
eNet
-
1
2
1
w
ith
Da
ta
a
u
g
me
n
ta
tio
n
…
(
S
u
fa
j
a
r
B
u
ts
ia
n
to
)
2527
f
lip
p
in
g
.
T
h
ese
tr
an
s
f
o
r
m
atio
n
s
in
tr
o
d
u
ce
d
v
ar
iab
ilit
y
in
to
th
e
d
ataset,
m
im
ick
in
g
r
ea
l
-
w
o
r
ld
co
n
d
itio
n
s
an
d
im
p
r
o
v
in
g
th
e
m
o
d
el’
s
ab
ilit
y
to
g
en
e
r
alize
to
u
n
s
ee
n
d
ata.
Fig
u
r
e
8
s
h
o
ws
th
e
r
es
u
lt
af
ter
ad
d
d
ata
au
g
m
en
tatio
n
.
T
h
e
ac
cu
r
ac
y
c
u
r
v
e
in
Fig
u
r
e
8
(
a)
s
h
o
ws
a
s
ig
n
if
ican
t
im
p
r
o
v
em
e
n
t
to
9
5
.
3
1
%,
wh
ile
th
e
lo
s
s
cu
r
v
e
i
n
Fig
u
r
e
8
(
b
)
d
ec
r
ea
s
es
to
1
1
.
7
4
%.
T
h
is
m
ar
k
ed
i
m
p
r
o
v
e
m
en
t
e
m
p
h
asizes
th
e
im
p
o
r
tan
ce
o
f
d
ata
d
iv
er
s
ity
in
en
h
an
cin
g
th
e
m
o
d
el'
s
ab
ilit
y
to
g
en
er
alize
.
I
n
th
e
f
in
al
s
tag
e,
f
i
n
e
-
tu
n
i
n
g
was
p
er
f
o
r
m
ed
b
y
u
n
f
r
ee
z
in
g
s
elec
ted
lay
er
s
o
f
th
e
p
r
e
-
tr
ain
ed
Den
s
eNe
t
-
1
2
1
m
o
d
el
an
d
r
et
r
ain
in
g
it
u
s
in
g
th
e
au
g
m
en
t
ed
d
ataset.
T
h
is
p
r
o
ce
s
s
en
a
b
led
th
e
n
etwo
r
k
to
ad
ju
s
t
its
lear
n
ed
r
e
p
r
esen
tatio
n
s
to
th
e
s
p
ec
if
ic
ch
a
r
ac
ter
i
s
tics
o
f
to
m
ato
leaf
d
is
ea
s
es,
r
esu
ltin
g
in
m
o
r
e
d
is
cr
im
in
ativ
e
an
d
task
-
r
ele
v
a
n
t
f
ea
tu
r
es.
Fig
u
r
e
9
s
h
o
ws
th
e
r
esu
lt
af
ter
f
in
e
-
tu
n
in
g
.
T
h
e
ac
cu
r
ac
y
c
u
r
v
e
in
Fig
u
r
e
9
(
a)
d
em
o
n
s
tr
ates
f
u
r
th
er
im
p
r
o
v
e
m
en
t,
r
ea
ch
in
g
9
8
.
4
4
%,
wh
ile
th
e
l
o
s
s
cu
r
v
e
in
Fig
u
r
e
9
(
b
)
d
ec
r
ea
s
es
to
4
.
7
2
%.
T
h
e
s
u
b
s
tan
tial
im
p
r
o
v
em
e
n
t
in
b
o
th
ac
cu
r
ac
y
an
d
lo
s
s
u
n
d
e
r
s
co
r
es
th
e
ef
f
ec
tiv
en
ess
o
f
tr
an
s
f
er
lear
n
in
g
an
d
th
e
cr
itic
al
r
o
le
o
f
f
in
e
-
tu
n
in
g
in
ac
h
iev
in
g
o
p
tim
al
p
er
f
o
r
m
an
ce
.
An
aly
s
is
o
f
th
e
tr
ain
in
g
an
d
v
alid
atio
n
c
u
r
v
es
in
d
icate
s
th
at
d
ata
au
g
m
en
tatio
n
an
d
f
in
e
-
tu
n
in
g
s
ig
n
if
ican
tly
r
ed
u
ce
d
o
v
er
f
itt
in
g
.
T
h
e
b
aselin
e
m
o
d
el
s
h
o
wed
a
n
o
ticea
b
le
g
ap
b
etwe
en
tr
ain
in
g
a
n
d
v
alid
atio
n
ac
cu
r
ac
y
,
s
u
g
g
esti
n
g
lim
ited
g
en
er
aliza
tio
n
p
er
f
o
r
m
an
ce
.
T
h
is
g
ap
p
r
o
g
r
ess
iv
ely
n
ar
r
o
wed
a
f
ter
d
ata
au
g
m
en
tatio
n
,
ac
co
m
p
an
ied
b
y
s
m
o
o
th
e
r
lo
s
s
co
n
v
er
g
e
n
ce
,
an
d
f
u
r
th
er
d
im
in
is
h
ed
f
o
llo
win
g
f
in
e
-
tu
n
in
g
,
wh
er
e
b
o
th
cu
r
v
es
co
n
v
er
g
e
d
m
o
r
e
clo
s
ely
,
d
em
o
n
s
tr
atin
g
im
p
r
o
v
ed
tr
ai
n
in
g
s
tab
ilit
y
an
d
g
en
er
aliza
tio
n
p
er
f
o
r
m
an
ce
.
C
o
m
p
ar
ativ
e
r
esu
lts
ac
r
o
s
s
t
h
e
th
r
ee
tr
ain
in
g
s
tag
es
co
n
f
ir
m
th
e
c
u
m
u
lativ
e
b
en
ef
its
o
f
e
ac
h
e
n
h
an
ce
m
e
n
t
.
Mo
d
el
ac
c
u
r
ac
y
in
cr
ea
s
ed
f
r
o
m
7
5
.
5
2
%
–
9
8
.
4
4
%,
wh
ile
lo
s
s
d
ec
r
ea
s
ed
f
r
o
m
3
7
.
7
7
%
–
4
.
7
2
%,
h
ig
h
lig
h
tin
g
th
e
ef
f
ec
tiv
en
ess
o
f
d
ata
au
g
m
en
tatio
n
a
n
d
f
in
e
-
t
u
n
in
g
in
s
tr
en
g
th
en
in
g
Den
s
eNe
t
-
1
2
1
f
ea
tu
r
e
lear
n
in
g
.
(
a)
(
b
)
Fig
u
r
e
7
.
R
esu
lt
u
s
e
Den
s
eNe
t
-
121
of
(
a
)
ac
cu
r
ac
y
an
d
(
b
)
lo
s
s
(
a)
(
b
)
Fig
u
r
e
8
.
R
esu
lt
af
ter
ad
d
d
ata
au
g
m
en
tatio
n
o
f
(
a)
ac
cu
r
ac
y
an
d
(
b
)
lo
s
s
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.
15
,
No
.
3
,
J
u
n
e
20
26
:
2
5
2
1
-
2
5
3
2
2528
(
a)
(
b
)
Fig
u
r
e
9
.
R
esu
lt
af
ter
f
in
e
-
tu
n
i
n
g
o
f
(
a)
ac
c
u
r
ac
y
a
n
d
(
b
)
lo
s
s
Fig
u
r
e
1
0
p
r
esen
ts
th
e
ac
cu
r
ac
y
r
esu
lts
o
b
tain
ed
f
r
o
m
ea
ch
f
o
ld
in
th
e
5
-
f
o
ld
s
tr
atif
ied
cr
o
s
s
-
v
alid
atio
n
.
T
h
e
m
o
d
el
d
em
o
n
s
tr
ates
co
n
s
is
ten
tly
h
ig
h
p
er
f
o
r
m
a
n
ce
ac
r
o
s
s
all
f
o
ld
s
,
with
ac
cu
r
ac
y
v
alu
es
o
f
9
8
.
3
8
%
-
f
o
ld
1
,
9
8
.
5
1
%
-
f
o
ld
2
,
9
8
.
4
7
%
-
f
o
ld
3
,
9
8
.
3
7
%
-
f
o
ld
4
,
a
n
d
9
8
.
4
7
%
-
f
o
ld
5
.
T
h
e
v
ar
iatio
n
b
etwe
en
f
o
l
d
s
is
m
in
im
al,
i
n
d
icatin
g
s
tab
le
m
o
d
el
b
eh
a
v
io
r
a
n
d
s
tr
o
n
g
g
en
er
aliza
tio
n
ca
p
ab
ilit
y
ac
r
o
s
s
d
if
f
er
en
t
d
ata
p
ar
titi
o
n
s
.
T
h
e
av
er
ag
e
ac
cu
r
ac
y
o
f
9
8
.
4
4
%,
as
in
d
icate
d
b
y
th
e
d
ash
e
d
lin
e,
c
o
n
f
ir
m
s
th
e
r
o
b
u
s
tn
ess
an
d
r
eliab
ilit
y
o
f
th
e
p
r
o
p
o
s
ed
Den
s
eNe
t
-
121
m
o
d
el
with
d
ata
au
g
m
en
tatio
n
a
n
d
f
in
e
-
tu
n
in
g
.
Fig
u
r
e
1
0
.
Acc
u
r
ac
y
p
er
f
o
ld
T
h
e
r
ep
o
r
ted
p
er
f
o
r
m
an
ce
wa
s
o
b
tain
ed
u
n
d
er
a
co
n
tr
o
lled
ex
p
er
im
e
n
tal
co
n
f
i
g
u
r
atio
n
u
s
in
g
f
ix
ed
in
itializatio
n
s
ettin
g
s
an
d
id
en
tical
tr
ain
in
g
co
n
d
itio
n
s
.
A
f
ix
ed
r
an
d
o
m
s
ee
d
was
ap
p
lied
d
u
r
in
g
d
ataset
s
p
litt
in
g
an
d
m
o
d
el
in
itializa
tio
n
to
r
ed
u
ce
s
to
ch
asti
c
v
ar
iab
ilit
y
d
u
r
in
g
o
p
tim
izatio
n
.
C
o
n
s
eq
u
en
tly
,
th
e
r
ep
o
r
ted
ac
c
u
r
ac
y
r
ef
lects
a
r
ep
r
o
d
u
cib
le
o
u
tco
m
e
u
n
d
e
r
th
e
d
ef
in
e
d
ex
p
e
r
im
en
tal
s
etu
p
r
ath
er
t
h
an
a
s
to
ch
asti
c
b
est
-
ca
s
e
o
b
s
er
v
atio
n
.
Alth
o
u
g
h
r
ep
ea
ted
r
u
n
s
tatis
ti
ca
l
an
aly
s
is
an
d
co
n
f
id
en
c
e
in
ter
v
al
esti
m
atio
n
co
u
l
d
f
u
r
th
e
r
q
u
a
n
tify
p
er
f
o
r
m
an
ce
v
ar
iab
ilit
y
,
th
e
p
r
im
ar
y
o
b
jectiv
e
o
f
th
is
s
tu
d
y
is
to
e
v
alu
ate
t
h
e
e
f
f
ec
tiv
en
e
s
s
o
f
a
p
r
o
g
r
ess
iv
e
tr
ain
in
g
s
tr
ateg
y
with
in
a
s
in
g
le
co
n
s
is
ten
t
ex
p
er
im
en
tal
p
ip
elin
e.
T
h
e
s
tag
ed
im
p
r
o
v
em
e
n
ts
o
b
s
er
v
ed
ac
r
o
s
s
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
To
ma
to
lea
f d
is
ea
s
e
cla
s
s
ifica
tio
n
u
s
in
g
Den
s
eNet
-
1
2
1
w
ith
Da
ta
a
u
g
me
n
ta
tio
n
…
(
S
u
fa
j
a
r
B
u
ts
ia
n
to
)
2529
b
aselin
e
tr
ain
in
g
,
au
g
m
en
tati
o
n
,
an
d
f
in
e
-
tu
n
in
g
p
r
o
v
i
d
e
in
ter
n
al
co
m
p
ar
ativ
e
ev
i
d
en
ce
in
d
icatin
g
th
at
th
e
o
b
s
er
v
ed
p
er
f
o
r
m
an
ce
g
ain
s
a
r
e
p
r
im
ar
ily
ass
o
ciate
d
with
th
e
p
r
o
p
o
s
ed
tr
ain
in
g
p
r
o
ce
d
u
r
e.
T
h
ese
r
esu
lts
in
d
icate
th
at
a
s
tr
u
ctu
r
ed
p
ip
elin
e
in
teg
r
atin
g
p
r
ep
r
o
ce
s
s
in
g
,
d
ata
a
u
g
m
en
tatio
n
,
an
d
f
in
e
-
tu
n
in
g
is
h
ig
h
ly
ef
f
ec
tiv
e
f
o
r
to
m
ato
leaf
d
is
ea
s
e
class
if
i
ca
tio
n
.
T
h
e
ac
h
iev
e
d
h
ig
h
ac
c
u
r
ac
y
an
d
l
o
w
lo
s
s
d
em
o
n
s
tr
ate
th
e
m
o
d
el’
s
r
eli
ab
ilit
y
f
o
r
p
r
ac
tical
a
g
r
icu
ltu
r
al
ap
p
licatio
n
s
.
Fu
t
u
r
e
w
o
r
k
m
ay
in
co
r
p
o
r
ate
ad
v
an
ce
d
ap
p
r
o
ac
h
es,
s
u
ch
as
atten
tio
n
m
ec
h
an
is
m
s
o
r
e
n
s
em
b
le
m
eth
o
d
s
,
to
f
u
r
th
er
en
h
an
ce
p
er
f
o
r
m
an
c
e
an
d
r
o
b
u
s
tn
ess
u
n
d
er
v
ar
y
i
n
g
co
n
d
itio
n
s
.
T
h
e
p
e
r
-
class
ev
alu
atio
n
r
e
s
u
lts
in
T
ab
le
3
d
em
o
n
s
tr
ate
th
at
th
e
p
r
o
p
o
s
ed
m
o
d
el
ac
h
iev
es
co
n
s
is
ten
tly
h
ig
h
p
er
f
o
r
m
a
n
c
e
ac
r
o
s
s
m
o
s
t
to
m
ato
leaf
d
is
ea
s
e
ca
teg
o
r
ies,
with
an
o
v
er
all
ac
cu
r
ac
y
o
f
9
8
.
4
4
%
af
ter
f
in
e
-
t
u
n
in
g
.
T
o
m
ato
m
o
za
ic
v
ir
u
s
ac
h
iev
es
p
er
f
ec
t
r
ec
all
1
0
0
%
d
esp
ite
h
av
in
g
f
ewe
r
s
am
p
les,
in
d
icatin
g
th
at
its
v
is
u
al
c
h
ar
a
cter
is
tics
ar
e
h
ig
h
ly
d
is
tin
g
u
is
h
ab
le
b
y
t
h
e
m
o
d
el.
Ho
wev
er
,
lo
wer
p
er
f
o
r
m
an
ce
is
o
b
s
er
v
ed
in
to
m
ato
s
p
id
er
m
ites
an
d
to
m
ato
tar
g
et
s
p
o
t
,
as
r
ef
lecte
d
in
th
ei
r
r
ed
u
ce
d
p
r
ec
is
io
n
an
d
r
ec
all
v
alu
es.
T
h
is
b
eh
av
io
r
is
f
u
r
th
er
s
u
p
p
o
r
ted
b
y
t
h
e
co
n
f
u
s
io
n
m
atr
ix
,
wh
er
e
t
h
ese
class
es
ex
h
ib
it
h
ig
h
er
m
is
class
if
icatio
n
r
ates.
T
h
e
p
r
im
ar
y
r
ea
s
o
n
lies
in
th
e
s
t
r
o
n
g
v
is
u
al
s
im
ilar
ity
b
etwe
en
th
ese
ca
teg
o
r
ies,
p
ar
ticu
lar
ly
in
ter
m
s
o
f
tex
tu
r
e
p
atter
n
s
,
lesi
o
n
d
is
tr
ib
u
tio
n
,
an
d
co
lo
r
v
a
r
iatio
n
s
,
wh
ich
o
v
er
lap
s
ig
n
if
ican
tly
an
d
m
ak
e
d
is
cr
im
in
atio
n
m
o
r
e
ch
allen
g
in
g
.
I
n
co
n
tr
ast,
cla
s
s
es
s
u
ch
as
to
m
ato
h
ea
lth
y
an
d
to
m
ato
y
ello
w
leaf
cu
r
l
v
ir
u
s
s
h
o
w
n
ea
r
-
p
er
f
ec
t
p
er
f
o
r
m
an
ce
d
u
e
t
o
m
o
r
e
d
is
tin
ct
v
is
u
al
f
ea
tu
r
es.
T
h
ese
f
in
d
i
n
g
s
h
i
g
h
lig
h
t
th
at
wh
ile
th
e
m
o
d
el
is
r
o
b
u
s
t
o
v
er
all,
class
if
icatio
n
er
r
o
r
s
ar
e
m
ain
ly
co
n
ce
n
tr
ate
d
in
v
is
u
a
lly
s
im
ilar
d
is
ea
s
e
ca
teg
o
r
ies,
s
u
g
g
esti
n
g
th
at
a
d
d
itio
n
al
f
ea
tu
r
e
en
h
an
ce
m
e
n
t
o
r
h
i
g
h
er
-
r
eso
lu
tio
n
in
p
u
ts
c
o
u
ld
f
u
r
th
er
im
p
r
o
v
e
class
-
lev
el
d
is
cr
im
in
atio
n
.
Data
au
g
m
en
tatio
n
an
d
f
in
e
-
tu
n
in
g
p
ar
ticu
lar
l
y
b
en
ef
ited
cla
s
s
es
with
h
ig
h
v
is
u
al
v
ar
iab
ilit
y
,
s
u
ch
as
to
m
ato
ea
r
ly
b
lig
h
t
an
d
to
m
ato
s
ep
to
r
ia
leaf
s
p
o
t,
wh
er
e
im
p
r
o
v
e
d
r
ec
all
in
d
icate
s
en
h
an
ce
d
g
e
n
er
aliza
tio
n
.
C
o
n
v
er
s
ely
,
m
is
class
if
icatio
n
r
em
ain
ed
m
o
r
e
f
r
e
q
u
en
t
b
etwe
en
v
is
u
ally
s
im
ilar
d
is
ea
s
e
p
air
s
,
n
o
tab
ly
ea
r
ly
b
lig
h
t
v
er
s
u
s
late
b
lig
h
t
a
n
d
ta
r
g
et
s
p
o
t
v
er
s
u
s
s
p
id
er
m
ites
.
Alth
o
u
g
h
Den
s
eNe
t
-
1
2
1
was
o
r
ig
in
ally
d
esig
n
ed
f
o
r
2
2
4
×2
2
4
in
p
u
ts
,
r
esizin
g
im
ag
es
to
6
4
×6
4
was
ad
o
p
ted
to
r
ed
u
ce
co
m
p
u
tatio
n
al
co
s
t
an
d
s
u
p
p
o
r
t
d
ep
lo
y
m
e
n
t
o
n
r
eso
u
r
ce
-
c
o
n
s
tr
ain
ed
p
latf
o
r
m
s
.
Desp
ite
p
o
ten
tial
lo
s
s
o
f
f
in
e
-
g
r
ain
ed
d
etails,
th
e
m
o
d
el
m
ain
tain
ed
h
ig
h
p
r
ec
is
io
n
an
d
r
ec
all,
in
d
icatin
g
th
at
g
lo
b
al
co
lo
r
an
d
s
tr
u
ctu
r
al
f
ea
tu
r
es
r
em
ain
in
f
o
r
m
ati
v
e
u
n
d
er
co
n
tr
o
lled
co
n
d
itio
n
s
.
Ho
wev
er
,
f
u
r
th
e
r
v
alid
atio
n
u
s
in
g
r
ea
l
-
f
ield
d
ata
is
n
ee
d
ed
to
en
s
u
r
e
r
o
b
u
s
tn
ess
in
d
iv
er
s
e
en
v
ir
o
n
m
e
n
ts
.
T
ab
le
3
.
Per
-
class
p
er
f
o
r
m
an
c
e
o
f
Den
s
eNe
t
-
1
2
1
D
i
sea
s
e
c
l
a
ss
P
r
e
c
i
s
i
o
n
(
%)
R
e
c
a
l
l
(
%)
F1
-
sc
o
r
e
(
%)
S
u
p
p
o
r
t
To
ma
t
o
b
a
c
t
e
r
i
a
l
s
p
o
t
95
93
94
87
To
ma
t
o
e
a
r
l
y
b
l
i
g
h
t
91
97
94
96
To
ma
t
o
l
a
t
e
b
l
i
g
h
t
1
0
0
93
97
1
2
2
To
ma
t
o
l
e
a
f
mo
l
d
1
0
0
90
95
92
To
ma
t
o
se
p
t
o
r
i
a
l
e
a
f
sp
o
t
87
98
92
94
To
ma
t
o
s
p
i
d
e
r
m
i
t
e
s (
t
w
o
-
s
p
o
t
t
e
d
sp
i
d
e
r
m
i
t
e
)
99
81
89
1
0
8
To
ma
t
o
t
a
r
g
e
t
sp
o
t
81
96
88
1
0
8
To
ma
t
o
y
e
l
l
o
w
l
e
a
f
c
u
r
l
v
i
r
u
s
99
94
97
1
0
9
To
ma
t
o
m
o
z
a
i
c
v
i
r
u
s
99
1
0
0
99
82
To
ma
t
o
h
e
a
l
t
h
y
95
1
0
0
98
1
0
2
T
h
is
s
tu
d
y
f
o
cu
s
es
o
n
an
aly
z
in
g
th
e
im
p
ac
t
o
f
p
r
ep
r
o
ce
s
s
in
g
,
d
ata
au
g
m
en
tatio
n
,
an
d
f
i
n
e
-
tu
n
in
g
with
in
a
f
ix
ed
Den
s
eNe
t
-
1
2
1
b
ac
k
b
o
n
e
r
ath
er
th
an
b
en
ch
m
ar
k
in
g
m
u
ltip
le
C
NN
ar
ch
itectu
r
es.
T
h
e
b
aselin
e
co
n
f
ig
u
r
atio
n
,
Den
s
eNe
t
-
1
2
1
tr
ain
ed
with
o
u
t
au
g
m
en
tatio
n
an
d
f
in
e
-
t
u
n
in
g
s
er
v
ed
as
an
in
ter
n
al
r
ef
er
e
n
ce
,
en
ab
lin
g
is
o
latio
n
o
f
p
er
f
o
r
m
a
n
ce
im
p
r
o
v
em
e
n
ts
attr
ib
u
tab
le
to
th
e
p
r
o
p
o
s
ed
tr
ain
i
n
g
s
tr
at
eg
y
.
I
t
s
h
o
u
ld
also
b
e
n
o
te
d
th
at
m
o
d
el
g
en
e
r
a
lizatio
n
was
ev
alu
ated
o
n
ly
with
in
a
s
in
g
le
b
en
c
h
m
ar
k
d
ataset,
r
ef
lectin
g
in
tr
a
-
d
ataset
p
er
f
o
r
m
an
ce
r
at
h
er
th
an
cr
o
s
s
-
d
ataset
r
o
b
u
s
tn
ess
.
I
n
ad
d
itio
n
,
in
ter
p
r
etab
ilit
y
an
aly
s
is
wa
s
b
ey
o
n
d
th
e
s
co
p
e
o
f
t
h
is
s
tu
d
y
.
T
ab
le
4
p
r
esen
ts
a
co
m
p
ar
is
o
n
b
etwe
en
th
e
p
r
o
p
o
s
ed
ap
p
r
o
a
ch
an
d
s
ev
er
al
p
r
ev
io
u
s
s
tu
d
ie
s
o
n
p
la
n
t
leaf
d
is
ea
s
e
clas
s
if
icatio
n
u
s
i
n
g
d
if
f
e
r
en
t
d
atasets
,
p
r
ep
r
o
c
ess
in
g
s
tr
ateg
ies,
an
d
d
ee
p
lear
n
in
g
ar
c
h
itectu
r
es.
Pri
o
r
wo
r
k
s
h
a
v
e
e
x
p
lo
r
e
d
a
w
id
e
r
an
g
e
o
f
m
o
d
els
,
i
n
cl
u
d
i
n
g
lig
h
t
wei
g
h
t a
r
ch
ite
ct
u
r
es
s
u
c
h
as
m
i
c
r
o
co
n
t
r
o
ll
e
r
u
n
it
n
et
wo
r
k
(
MCUNe
t
)
[
2
8
]
,
r
e
g
i
o
n
-
b
ase
d
a
p
p
r
o
ac
h
es
l
ik
e
r
e
g
io
n
-
b
ase
d
c
o
n
v
o
lu
ti
o
n
al
n
e
u
r
al
n
et
wo
r
k
(
R
-
C
NN
)
[
2
6
]
a
n
d
m
as
k
r
eg
i
o
n
-
b
as
ed
c
o
n
v
o
l
u
ti
o
n
a
l
n
e
u
r
al
n
et
wo
r
k
(
MRC
NN
)
c
o
m
b
i
n
e
d
wit
h
o
p
ti
m
i
za
ti
o
n
tec
h
n
i
q
u
es
[
3
7
]
,
as
wel
l
as
c
o
n
v
e
n
ti
o
n
al
C
NN
in
cl
u
d
in
g
V
GG
-
1
6
,
VGG
-
1
9
[
3
8
]
,
Ale
x
N
et
[
2
2
]
,
a
n
d
h
y
b
r
id
R
esNet
-
50
–
s
u
p
p
o
r
t v
ec
t
o
r
m
a
ch
in
e
(
SVM
)
f
r
am
ewo
r
k
s
[
2
1
]
.
R
ep
o
r
te
d
a
cc
u
r
ac
i
es
v
ar
y
f
r
o
m
9
1
.
3
2
%
–
9
8
.
0
0
%
,
d
e
p
e
n
d
i
n
g
o
n
d
a
tas
et
c
o
m
p
l
ex
i
ty
,
n
u
m
b
e
r
o
f
cl
ass
es
,
a
n
d
p
r
ep
r
o
ce
s
s
i
n
g
c
o
n
f
i
g
u
r
at
io
n
s
.
Mo
s
t
ex
is
tin
g
s
tu
d
ies
em
p
lo
y
im
ag
e
r
esizin
g
an
d
d
ata
au
g
m
en
tatio
n
as
ess
en
tial
p
r
ep
r
o
ce
s
s
in
g
s
tep
s
to
en
h
an
ce
m
o
d
el
g
en
er
aliza
tio
n
.
Ho
wev
er
,
p
er
f
o
r
m
a
n
ce
d
if
f
er
en
ce
s
ar
e
s
tr
o
n
g
ly
in
f
lu
en
ce
d
b
y
d
ataset
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.
15
,
No
.
3
,
J
u
n
e
20
26
:
2
5
2
1
-
2
5
3
2
2530
ch
ar
ac
ter
is
tics
an
d
class
if
icatio
n
co
m
p
lex
ity
,
p
ar
ticu
lar
l
y
t
h
e
n
u
m
b
er
o
f
d
is
ea
s
e
ca
teg
o
r
ies.
Fo
r
ex
a
m
p
le,
s
ev
er
al
s
tu
d
ies
ac
h
iev
in
g
h
ig
h
ac
cu
r
ac
y
wer
e
co
n
d
u
cte
d
o
n
d
atasets
co
n
tain
in
g
f
e
wer
class
es,
wh
ich
g
en
er
ally
r
e
d
u
ce
s
class
if
icatio
n
d
if
f
icu
lty
.
I
n
co
m
p
ar
is
o
n
,
th
e
p
r
o
p
o
s
ed
Den
s
eNe
t
-
121
-
b
ased
a
p
p
r
o
ac
h
ac
h
iev
es
an
ac
c
u
r
ac
y
o
f
9
8
.
4
4
%
o
n
th
e
Plan
tVillag
e
d
ataset
with
ten
d
is
ea
s
e
class
es,
in
co
r
p
o
r
a
tin
g
im
ag
e
r
esizin
g
,
d
ata
a
u
g
m
en
tatio
n
,
an
d
f
in
e
-
tu
n
in
g
with
in
a
tr
a
n
s
f
er
lear
n
in
g
f
r
am
ewo
r
k
.
T
h
e
r
esu
lts
d
em
o
n
s
tr
ate
th
at
o
p
ti
m
izin
g
th
e
tr
ain
in
g
s
tr
ateg
y
an
d
lev
e
r
ag
in
g
d
en
s
e
f
ea
tu
r
e
r
e
u
s
e
ca
n
ac
h
iev
e
c
o
m
p
etitiv
e
p
er
f
o
r
m
an
ce
e
v
en
with
in
cr
ea
s
ed
class
co
m
p
lex
ity
.
T
h
ese
f
in
d
in
g
s
s
u
g
g
est
th
at
m
eth
o
d
o
lo
g
ical
r
ef
in
em
en
t
th
r
o
u
g
h
p
r
o
g
r
ess
iv
e
tr
ain
in
g
s
tag
es
ca
n
b
e
as
in
f
lu
en
tial
as
ar
ch
itectu
r
al
v
ar
iatio
n
in
im
p
r
o
v
i
n
g
class
if
icatio
n
p
er
f
o
r
m
a
n
ce
f
o
r
p
lan
t
d
is
ea
s
e
d
etec
tio
n
task
s
.
T
ab
le
4
.
C
o
m
p
a
r
is
o
n
r
esu
lt
A
u
t
h
o
r
D
a
t
a
s
e
t
N
u
mb
e
r
o
f
c
l
a
ss
e
s
Pre
-
p
r
o
c
e
ssi
n
g
M
o
d
e
l
a
r
c
h
i
t
e
c
t
u
r
e
A
c
c
u
r
a
c
y
(
%)
[
2
8
]
P
l
a
n
t
V
i
l
l
a
g
e
4
I
mag
e
r
e
si
z
i
n
g
,
a
n
d
d
a
t
a
a
u
g
me
n
t
a
t
i
o
n
M
C
U
N
e
t
9
1
.
3
2
[
3
7
]
A
I
c
h
a
l
l
e
n
g
e
r
8
N
o
i
se
r
e
m
o
v
a
l
,
d
a
t
a
a
n
n
o
t
a
t
i
o
n
,
d
a
t
a
a
u
g
me
n
t
a
t
i
o
n
,
a
n
d
d
a
t
a
r
e
s
i
z
e
M
R
C
N
N
,
A
n
t
c
o
l
o
n
y
o
p
t
i
m
i
z
e
r
9
7
.
6
6
[
2
6
]
P
l
a
n
t
V
i
l
l
a
g
e
6
G
r
a
y
sc
a
l
e
i
m
a
g
e
a
n
d
i
ma
g
e
r
e
si
z
i
n
g
R
-
C
N
N
9
6
.
7
3
[
3
8
]
K
a
g
g
l
e
11
D
a
t
a
a
u
g
m
e
n
t
a
t
i
o
n
VGG
-
1
6
,
V
G
G
-
19
9
5
.
0
0
[
2
2
]
K
a
g
g
l
e
10
-
A
l
e
x
N
e
t
9
8
.
0
0
[
2
1
]
P
l
a
n
t
V
i
l
l
a
g
e
4
I
mag
e
r
e
si
z
i
n
g
R
e
sN
e
t
-
5
0
,
S
V
M
9
3
.
3
3
O
u
r
p
r
o
p
o
se
d
P
l
a
n
t
V
i
l
l
a
g
e
10
I
mag
e
r
e
si
z
i
n
g
,
d
a
t
a
a
u
g
me
n
t
a
t
i
o
n
,
a
n
d
f
i
n
e
-
t
u
n
i
n
g
D
e
n
seN
e
t
-
1
2
1
9
8
.
4
4
4.
CO
NCLU
SI
O
N
T
h
is
s
tu
d
y
d
e
m
o
n
s
tr
ates
th
at
in
teg
r
atin
g
im
a
g
e
p
r
ep
r
o
ce
s
s
in
g
,
d
ata
au
g
m
en
tatio
n
,
an
d
f
i
n
e
-
tu
n
in
g
ef
f
ec
tiv
ely
en
h
a
n
ce
s
Den
s
eNe
t
-
121
f
o
r
to
m
at
o
leaf
d
is
ea
s
e
class
if
icatio
n
.
R
esiz
in
g
im
ag
es
to
6
4
×6
4
p
ix
els
p
r
o
v
id
e
d
a
co
m
p
u
tatio
n
ally
e
f
f
icien
t
b
aselin
e
wh
ile
p
r
eser
v
in
g
ess
en
tial
v
is
u
al
c
h
ar
ac
te
r
is
tics
,
ac
h
iev
in
g
an
in
itial
ac
cu
r
ac
y
o
f
7
5
.
5
2
%.
T
h
e
in
tr
o
d
u
ctio
n
o
f
d
ata
au
g
m
en
tatio
n
im
p
r
o
v
ed
g
e
n
er
aliza
t
io
n
p
er
f
o
r
m
an
ce
t
o
9
5
.
3
1
%,
an
d
s
u
b
s
eq
u
e
n
t
f
i
n
e
-
t
u
n
in
g
f
u
r
th
er
i
n
cr
ea
s
ed
ac
c
u
r
a
cy
to
9
8
.
4
4
%
with
r
ed
u
ce
d
lo
s
s
,
h
ig
h
lig
h
tin
g
th
e
cu
m
u
lativ
e
im
p
ac
t
o
f
p
r
o
g
r
es
s
iv
e
tr
ain
in
g
s
tag
es.
T
h
ese
r
e
s
u
lts
in
d
icate
th
at
o
p
tim
izin
g
tr
ain
in
g
s
tr
ateg
ies
with
in
a
tr
a
n
s
f
er
lea
r
n
in
g
f
r
a
m
ewo
r
k
ca
n
p
r
o
d
u
ce
a
s
tab
le
an
d
h
ig
h
-
p
e
r
f
o
r
m
in
g
class
if
icatio
n
m
o
d
el
u
n
d
e
r
co
n
tr
o
lled
e
x
p
er
im
e
n
tal
co
n
d
itio
n
s
.
Alth
o
u
g
h
th
e
f
i
n
d
in
g
s
s
h
o
w
p
r
o
m
is
in
g
p
o
ten
tial
f
o
r
au
to
m
ated
p
la
n
t
d
is
ea
s
e
an
aly
s
is
,
f
u
r
th
er
v
alid
atio
n
u
s
in
g
r
ea
l
-
f
ield
d
atasets
is
r
eq
u
ir
ed
to
co
n
f
ir
m
r
o
b
u
s
tn
ess
in
p
r
ac
tical
en
v
ir
o
n
m
en
ts
.
Fu
tu
r
e
wo
r
k
will in
v
esti
g
ate
h
ig
h
er
in
p
u
t r
eso
lu
tio
n
s
,
alter
n
ativ
e
ar
ch
itectu
r
es,
an
d
ex
p
lain
ab
le
AI
ap
p
r
o
ac
h
es to
f
u
r
th
er
im
p
r
o
v
e
g
e
n
er
aliza
tio
n
ca
p
a
b
ilit
y
a
n
d
r
ea
l
-
wo
r
ld
ap
p
licab
ilit
y
.
F
UNDING
I
NF
O
R
M
A
T
I
O
N
T
h
is
wo
r
k
was
s
u
p
p
o
r
ted
b
y
Dep
ar
tm
en
t
o
f
R
esear
ch
a
n
d
C
o
m
m
u
n
ity
Ser
v
ice,
Pelita
B
an
g
s
a
Un
iv
er
s
ity
,
f
o
r
s
u
p
p
o
r
tin
g
a
n
d
f
u
n
d
i
n
g
th
is
r
esear
ch
.
T
h
is
r
esear
ch
r
ec
eiv
ed
n
o
s
p
ec
if
ic
g
r
an
t
f
r
o
m
an
y
f
u
n
d
in
g
ag
e
n
cy
in
t
h
e
p
u
b
lic,
co
m
m
er
cial,
o
r
n
o
t
-
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
Su
f
ajar
B
u
ts
ian
to
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
An
g
g
i M
u
h
a
m
m
ad
R
if
a’
i
✓
✓
✓
✓
✓
✓
✓
✓
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
Evaluation Warning : The document was created with Spire.PDF for Python.