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
4
,
No
.
6
,
Dec
em
b
er
2
0
2
5
,
p
p
.
5
1
4
0
~
5
1
5
6
I
SS
N:
2
2
5
2
-
8
9
3
8
,
DOI
: 1
0
.
1
1
5
9
1
/ijai.v
14
.i
6
.
p
p
5
1
4
0
-
5
1
5
6
5140
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
a
i
.
ia
esco
r
e.
co
m
Cla
ss
ificatio
n of s
ing
le orig
in Indo
n
esia
n coffee
beans
using
co
nv
o
lutiona
l neural network
Achm
a
d P
ra
t
a
m
a
Rif
a
i
1
,
Wa
ng
i P
a
nd
a
n Sa
ri
1
,
H
a
ida
r
Ra
bb
a
ni
1
,
T
a
ri
H
a
rdia
ni Sa
f
it
ri
1
,
M
a
k
bu
l H
a
j
a
d
2
,
E
di Su
t
o
y
o
3
,
H
uu
-
T
ho
Ng
uy
en
4
1
D
e
p
a
r
t
me
n
t
o
f
M
e
c
h
a
n
i
c
a
l
a
n
d
I
n
d
u
s
t
r
i
a
l
E
n
g
i
n
e
e
r
i
n
g
,
U
n
i
v
e
r
si
t
a
s G
a
d
j
a
h
M
a
d
a
,
Y
o
g
y
a
k
a
r
t
a
,
I
n
d
o
n
e
s
i
a
2
F
a
c
u
l
t
y
o
f
A
g
r
i
c
u
l
t
u
r
a
l
Te
c
h
n
o
l
o
g
y
,
U
n
i
v
e
r
si
t
a
s
G
a
d
j
a
h
M
a
d
a
,
Y
o
g
y
a
k
a
r
t
a
,
I
n
d
o
n
e
si
a
3
D
e
p
a
r
t
me
n
t
o
f
I
n
f
o
r
mat
i
o
n
S
y
s
t
e
ms
,
Te
l
k
o
m U
n
i
v
e
r
si
t
y
,
B
a
n
d
u
n
g
,
I
n
d
o
n
e
si
a
4
F
a
c
u
l
t
y
o
f
En
g
i
n
e
e
r
i
n
g
a
n
d
T
e
c
h
n
o
l
o
g
y
,
N
g
u
y
e
n
Ta
t
T
h
a
n
h
U
n
i
v
e
r
si
t
y
,
H
o
C
h
i
M
i
n
h
C
i
t
y
,
V
i
e
t
n
a
m
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
No
v
9
,
2
0
2
4
R
ev
is
ed
Oct
2
4
,
2
0
2
5
Acc
ep
ted
No
v
8
,
2
0
2
5
Th
is
re
se
a
rc
h
a
ims
to
d
e
v
e
lo
p
a
c
o
ffe
e
b
e
a
n
ty
p
e
d
e
tec
ti
o
n
m
o
d
e
l
u
sin
g
c
o
n
v
o
lu
ti
o
n
a
l
n
e
u
ra
l
n
e
two
rk
s
(C
NN
),
lev
e
ra
g
in
g
a
d
a
tas
e
t
o
f
1
4
,
5
2
5
ima
g
e
s
fro
m
1
1
6
ty
p
e
s
o
f
In
d
o
n
e
sia
n
c
o
ffe
e
b
e
a
n
s.
P
re
-
p
ro
c
e
ss
in
g
ste
p
s
in
c
lu
d
in
g
re
siz
in
g
,
re
sc
a
li
n
g
,
a
n
d
a
u
g
m
e
n
t
a
ti
o
n
we
re
a
p
p
li
e
d
to
imp
r
o
v
e
t
h
e
d
a
tas
e
t
q
u
a
li
t
y
.
Th
e
d
a
tas
e
t
wa
s
sp
li
t
in
to
train
i
n
g
,
v
a
li
d
a
ti
o
n
,
a
n
d
tes
ti
n
g
se
ts
with
p
ro
p
o
rti
o
n
s
o
f
8
0
%
,
1
0
%
,
a
n
d
1
0
%
,
re
sp
e
c
ti
v
e
l
y
.
Two
m
o
d
e
l
d
e
v
e
lo
p
m
e
n
t
a
p
p
ro
a
c
h
e
s
we
re
u
se
d
:
tran
sfe
r
lea
rn
in
g
wit
h
In
c
e
p
ti
o
n
V
3
in
two
sc
e
n
a
rio
s
a
n
d
a
m
o
d
e
l
b
u
il
t
fro
m
sc
ra
tch
.
T
h
e
tran
sfe
r
lea
rn
i
n
g
I
n
c
e
p
ti
o
n
V
3
m
o
d
e
l
i
n
s
c
e
n
a
rio
1
a
c
h
iev
e
d
t
h
e
b
e
st
p
e
rf
o
rm
a
n
c
e
,
with
a
tes
t
a
c
c
u
ra
c
y
o
f
0
.
8
7
a
n
d
o
p
ti
m
a
l
e
v
a
l
u
a
ti
o
n
m
e
tri
c
s
a
c
ro
ss
p
re
c
isio
n
,
re
c
a
ll
,
a
n
d
F
1
-
sc
o
re
.
T
h
is
m
o
d
e
l
wa
s
fin
e
-
tu
n
e
d
u
sin
g
p
re
train
e
d
we
ig
h
ts,
a
ll
o
win
g
it
t
o
a
d
a
p
t
e
ffe
c
ti
v
e
ly
to
th
e
c
o
ffe
e
b
e
a
n
d
a
tas
e
t.
Th
e
re
su
lt
s
h
ig
h
li
g
h
t
th
a
t
tran
sfe
r
lea
rn
in
g
,
e
sp
e
c
ially
with
In
c
e
p
ti
o
n
V
3
,
p
ro
v
id
e
s
a
ro
b
u
st
m
e
th
o
d
f
o
r
c
las
sify
in
g
c
o
ffe
e
b
e
a
n
s,
o
ffe
ri
n
g
p
o
ten
ti
a
l
a
p
p
li
c
a
ti
o
n
s
i
n
t
h
e
c
o
ffe
e
i
n
d
u
stry
f
o
r
imp
ro
v
i
n
g
c
las
sifica
ti
o
n
e
fficie
n
c
y
a
n
d
a
c
c
u
ra
c
y
.
Th
e
stu
d
y
d
e
m
o
n
stra
tes
h
o
w
d
e
e
p
lea
rn
in
g
c
a
n
e
n
h
a
n
c
e
th
e
o
b
jec
ti
v
it
y
a
n
d
p
re
c
isio
n
o
f
c
o
ffe
e
b
e
a
n
c
las
sifica
ti
o
n
,
c
o
n
tri
b
u
t
in
g
t
o
g
r
e
a
ter
c
o
n
siste
n
c
y
in
p
ro
d
u
c
t
so
rti
n
g
a
n
d
q
u
a
li
t
y
a
ss
e
ss
m
e
n
t.
K
ey
w
o
r
d
s
:
C
o
f
f
ee
b
ea
n
class
if
icatio
n
C
o
n
v
o
lu
tio
n
al
n
eu
r
al
n
etwo
r
k
Dee
p
lear
n
in
g
I
n
d
o
n
esia sin
g
le
o
r
ig
in
c
o
f
f
ee
T
r
an
s
f
er
lear
n
i
n
g
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
:
Ach
m
ad
Pra
tam
a
R
if
ai
Dep
ar
tm
en
t o
f
Me
ch
an
ical
an
d
I
n
d
u
s
tr
ial
E
n
g
in
ee
r
i
n
g
,
Un
i
v
er
s
itas
Gad
jah
Ma
d
a
Yo
g
y
ak
ar
ta,
I
n
d
o
n
esia
E
m
ail: a
ch
m
ad
.
p
.
r
if
ai@
u
g
m
.
a
c.
id
1.
I
NT
RO
D
UCT
I
O
N
I
n
d
o
n
esia,
as
th
e
f
o
u
r
t
h
-
lar
g
est
co
f
f
ee
p
r
o
d
u
ce
r
in
t
h
e
wo
r
ld
,
was
ca
p
ab
le
o
f
p
r
o
d
u
cin
g
7
1
1
.
3
th
o
u
s
an
d
to
n
s
o
f
co
f
f
e
e
b
ea
n
s
an
n
u
ally
b
ef
o
r
e
2
0
2
0
[
1
]
.
C
o
f
f
ee
p
la
y
s
a
v
ital
r
o
le
as
a
p
lan
tatio
n
co
m
m
o
d
ity
,
co
n
tr
i
b
u
tin
g
s
ig
n
if
ican
tly
to
th
e
n
atio
n
al
ec
o
n
o
m
y
th
r
o
u
g
h
ex
p
o
r
t
o
p
p
o
r
t
u
n
ities
an
d
a
s
tr
o
n
g
d
o
m
esti
c
m
ar
k
et
[
2
]
.
T
h
ese
o
p
p
o
r
tu
n
ities
ar
e
b
o
ls
ter
ed
b
y
I
n
d
o
n
esia'
s
s
tead
ily
in
cr
ea
s
in
g
co
f
f
ee
p
r
o
d
u
ctio
n
ea
ch
y
ea
r
,
f
r
o
m
2
0
2
0
to
2
0
2
2
.
I
n
2
0
2
2
alo
n
e,
c
o
f
f
ee
p
r
o
d
u
ctio
n
r
o
s
e
b
y
1
.
1
%
f
r
o
m
th
e
p
r
ev
io
u
s
y
ea
r
,
r
ea
ch
in
g
7
9
4
.
8
th
o
u
s
an
d
to
n
s
[
3
]
.
So
m
e
o
f
I
n
d
o
n
esia'
s
r
en
o
wn
ed
c
o
f
f
ee
b
ea
n
v
a
r
ieties
in
clu
d
e
Ar
ab
ica,
R
o
b
u
s
ta,
L
ib
er
ica,
E
x
ce
ls
a,
a
n
d
th
e
f
am
o
u
s
K
o
p
i
Lu
w
a
k
.
C
o
f
f
ee
v
ar
ieties
ar
e
o
f
te
n
in
f
lu
e
n
ce
d
b
y
th
e
r
eg
io
n
o
r
p
r
o
v
in
ce
o
f
p
r
o
d
u
ctio
n
,
lea
d
in
g
to
m
o
r
e
s
p
ec
if
ic
ty
p
es
o
f
co
f
f
ee
.
Ad
d
itio
n
ally
,
d
r
ied
co
f
f
ee
b
ea
n
s
p
o
s
s
ess
v
ar
y
in
g
q
u
ality
ch
ar
ac
ter
is
tics
d
ep
en
d
in
g
o
n
th
eir
p
r
o
ce
s
s
in
g
[
4
]
.
W
ith
th
is
wea
lth
o
f
co
f
f
ee
ty
p
es,
class
if
y
in
g
co
f
f
ee
b
ea
n
s
b
y
v
ar
iety
an
d
q
u
ality
is
cr
u
cial
f
o
r
b
o
th
en
t
r
ep
r
en
eu
r
s
an
d
f
ar
m
er
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
C
la
s
s
i
fica
tio
n
o
f sin
g
le
o
r
ig
in
I
n
d
o
n
esia
n
co
ffee
b
e
a
n
s
u
s
in
g
co
n
vo
lu
tio
n
a
l …
(
A
c
h
ma
d
P
r
a
ta
ma
R
ifa
i)
5141
C
u
r
r
en
tly
,
m
an
y
co
f
f
ee
e
n
tr
e
p
r
en
eu
r
s
an
d
f
ar
m
er
s
s
till
m
an
u
ally
class
if
y
co
f
f
ee
b
ea
n
s
b
y
ty
p
e
an
d
q
u
ality
th
r
o
u
g
h
wo
r
k
e
r
s
o
r
ex
p
er
t
ev
alu
ato
r
s
(
cu
p
p
er
)
.
T
h
is
clas
s
if
icatio
n
is
b
ased
o
n
v
is
u
al
in
s
p
ec
tio
n
o
r
s
en
s
o
r
y
test
s
,
wh
ich
ar
e
s
u
b
jectiv
e
[
5
]
.
T
h
is
s
u
b
jectiv
ity
lea
d
s
to
in
co
n
s
is
ten
cies
in
co
f
f
ee
b
ea
n
class
if
icatio
n
.
On
th
e
o
th
er
h
an
d
,
ac
cu
r
ate
cl
ass
if
icatio
n
is
es
s
en
tial
as
it
af
f
ec
ts
th
e
p
r
o
d
u
ctiv
ity
o
f
t
h
e
c
o
f
f
ee
b
ea
n
s
,
b
o
th
i
n
ter
m
s
o
f
f
lav
o
r
a
n
d
q
u
ality
,
alig
n
in
g
with
d
iv
er
s
e
co
n
s
u
m
er
d
em
an
d
s
[
6
]
,
[
7
]
.
T
h
e
r
e
f
o
r
e,
d
e
v
elo
p
in
g
a
class
if
icatio
n
m
eth
o
d
th
at
ca
n
ev
alu
ate
th
e
ch
a
r
ac
ter
is
tics
o
f
co
f
f
ee
b
ea
n
ty
p
es
an
d
q
u
ality
m
o
r
e
ef
f
ec
tiv
ely
an
d
ef
f
icien
tly
is
n
ec
ess
ar
y
.
Su
ch
a
m
eth
o
d
m
u
s
t
also
b
e
o
b
jectiv
e
to
en
s
u
r
e
g
r
e
ater
ac
cu
r
ac
y
an
d
co
n
s
is
ten
cy
,
ad
d
r
ess
in
g
th
e
s
h
o
r
tco
m
in
g
s
o
f
m
an
u
al
class
if
icatio
n
,
wh
ich
is
p
r
o
n
e
to
h
ig
h
s
u
b
jectiv
ity
.
I
n
th
is
r
eg
ar
d
,
ar
tific
ial
in
tellig
en
ce
(
AI
)
tech
n
o
lo
g
ies,
s
u
ch
as
m
ac
h
in
e
lear
n
in
g
an
d
f
u
zz
y
lo
g
ic
,
ca
n
b
e
u
tili
ze
d
to
en
h
an
ce
th
e
ef
f
icien
cy
o
f
co
f
f
ee
b
ea
n
class
if
icatio
n
,
as
d
em
o
n
s
tr
ated
in
s
tu
d
ies
b
y
L
iv
io
an
d
Ho
d
h
o
d
[
8
]
an
d
San
to
s
et
a
l
.
[
9
]
.
AI
is
u
s
ed
to
im
p
r
o
v
e
b
o
th
th
e
ac
cu
r
ac
y
an
d
f
in
an
cial
an
d
tim
e
ef
f
icien
cy
in
id
en
tify
in
g
co
f
f
ee
b
ea
n
ty
p
e
s
an
d
q
u
ality
.
Ma
ch
in
e
l
ea
r
n
in
g
m
eth
o
d
s
s
u
ch
as
ar
tific
i
al
n
eu
r
al
n
etwo
r
k
s
(
ANN)
[
1
0
]
,
k
-
n
ea
r
est n
eig
h
b
o
r
(
KNN)
[
1
1
]
,
a
n
d
s
u
p
p
o
r
t v
e
cto
r
m
ac
h
in
e
(
SVM)
[
1
2
]
,
[
1
3
]
h
av
e
b
e
en
ap
p
lied
in
co
f
f
ee
b
ea
n
class
if
icatio
n
with
h
ig
h
ac
cu
r
ac
y
.
Ho
we
v
er
,
th
e
co
m
p
lex
d
ata
p
r
o
ce
s
s
in
g
r
eq
u
ir
ed
in
c
o
f
f
ee
b
ea
n
class
if
icatio
n
m
ak
es
m
ac
h
in
e
lear
n
in
g
less
s
u
itab
le
in
s
o
m
e
ca
s
es.
On
e
a
s
p
ec
t
o
f
m
ac
h
in
e
lear
n
in
g
th
at
ca
n
ad
d
r
ess
th
is
is
s
u
e
is
d
ee
p
lear
n
in
g
.
Dee
p
lea
r
n
in
g
ca
n
ac
co
m
m
o
d
ate
v
a
r
io
u
s
v
ar
iab
les
in
v
o
lv
ed
in
co
f
f
ee
class
if
icatio
n
,
s
u
ch
as c
o
f
f
ee
b
ea
n
ty
p
e,
r
o
ast lev
el,
g
e
o
g
r
ap
h
ical
o
r
ig
in
,
an
d
p
r
o
ce
s
s
in
g
m
e
th
o
d
s
.
D
e
e
p
le
a
r
n
i
n
g
i
s
a
b
r
a
n
c
h
o
f
A
I
t
h
a
t
u
s
e
s
n
e
u
r
a
l
n
e
t
w
o
r
k
s
c
o
n
s
i
s
t
i
n
g
o
f
m
a
n
y
l
a
y
e
r
s
c
a
p
a
b
l
e
o
f
r
e
c
o
g
n
i
z
i
n
g
c
o
m
p
l
e
x
p
a
t
t
e
r
n
s
i
n
d
a
t
a
.
A
d
d
i
t
i
o
n
a
l
l
y
,
d
e
e
p
l
e
a
r
n
i
n
g
c
a
n
a
u
t
o
m
a
t
i
c
a
l
l
y
e
x
t
r
a
c
t
i
m
p
o
r
t
a
n
t
f
e
a
t
u
r
e
s
f
r
o
m
c
o
m
p
l
e
x
,
h
i
g
h
-
d
i
m
e
n
s
i
o
n
a
l
d
a
t
a
.
T
h
e
r
e
f
o
r
e
,
i
t
i
s
w
e
l
l
-
s
u
i
t
e
d
f
o
r
c
o
f
f
e
e
c
l
a
s
s
i
f
i
c
a
t
i
o
n
p
r
o
c
e
s
s
e
s
i
n
v
o
l
v
i
n
g
i
n
t
r
i
c
a
t
e
p
a
t
t
e
r
n
s
,
s
u
c
h
a
s
a
r
o
m
a
,
t
a
s
t
e
,
c
o
l
o
r
,
s
i
z
e
,
s
h
a
p
e
,
a
n
d
t
e
x
t
u
r
e
c
h
a
r
a
c
t
e
r
i
s
t
i
c
s
o
f
c
o
f
f
e
e
b
e
a
n
s
.
D
e
e
p
l
e
a
r
n
i
n
g
i
n
c
l
u
d
e
s
s
e
v
e
r
a
l
a
l
g
o
r
i
t
h
m
i
c
m
e
t
h
o
d
s
t
h
a
t
h
a
v
e
p
r
o
v
e
n
a
c
c
u
r
a
t
e
i
n
c
o
f
f
e
e
c
l
a
s
s
i
f
i
c
a
t
i
o
n
,
s
u
c
h
a
s
m
u
l
t
i
l
a
y
e
r
p
e
r
c
e
p
t
r
o
n
(
M
L
P
)
,
p
r
i
n
c
i
p
a
l
c
o
m
p
o
n
e
n
t
a
n
a
l
y
s
i
s
(
P
C
A
)
,
A
N
N
,
a
n
d
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
u
r
a
l
n
e
t
w
o
r
k
s
(
C
N
N
)
[
1
4
]
–
[
1
6
]
.
A
m
o
n
g
t
h
e
s
e
m
e
t
h
o
d
s
,
C
N
N
i
s
t
h
e
m
o
s
t
f
r
e
q
u
e
n
t
l
y
u
s
e
d
f
o
r
c
l
a
s
s
i
f
i
c
a
t
i
o
n
,
e
s
p
e
c
i
a
l
l
y
w
h
e
n
t
h
e
p
r
o
c
e
s
s
i
n
v
o
l
v
e
s
i
m
a
g
e
d
a
t
a
,
a
s
C
N
N
o
f
f
e
r
s
h
i
g
h
a
c
c
u
r
a
c
y
a
n
d
s
u
p
e
r
i
o
r
p
e
r
f
o
r
m
a
n
c
e
i
n
i
m
a
g
e
c
l
a
s
s
i
f
i
c
a
t
i
o
n
[
1
7
]
.
Ho
wev
er
,
in
r
ea
lity
,
th
e
u
s
e
o
f
AI
m
eth
o
d
s
f
o
r
c
o
f
f
ee
b
ea
n
class
if
icatio
n
in
I
n
d
o
n
esia
is
s
till
r
ar
e,
r
esu
ltin
g
in
s
u
b
o
p
tim
al
ef
f
ici
en
cy
.
T
h
is
s
tu
d
y
s
ee
k
s
to
ad
d
r
ess
th
is
g
ap
,
wh
ich
is
p
ar
tl
y
d
u
e
to
th
e
lim
ited
am
o
u
n
t o
f
p
r
ev
io
u
s
r
esear
ch
o
n
AI
-
b
ased
c
o
f
f
ee
class
if
icatio
n
in
I
n
d
o
n
esia.
So
f
ar
,
a
f
ew
s
t
u
d
ies
in
I
n
d
o
n
esia
h
av
e
b
ee
n
co
n
d
u
cted
,
s
u
c
h
a
s
Nasu
tio
n
an
d
An
d
ay
a
n
i’
s
[
1
8
]
s
tu
d
y
o
n
r
o
asti
n
g
lev
el
c
lass
if
icatio
n
u
s
in
g
im
ag
e
p
r
o
ce
s
s
in
g
an
d
ANN,
p
o
wd
er
c
o
f
f
ee
class
if
icatio
n
u
s
in
g
ANN,
an
d
Su
h
an
d
i
an
d
Yu
lia’
s
[
1
9
]
s
tu
d
y
o
n
Ar
ab
ica
co
f
f
ee
p
o
wd
e
r
class
if
icatio
n
u
s
in
g
u
ltra
v
io
let
-
v
is
ib
le
(
UV
-
Vis
)
s
p
ec
tr
o
s
co
p
y
tech
n
o
lo
g
y
.
T
h
is
r
esear
ch
ap
p
lies
C
NN
in
th
e
class
if
icatio
n
o
f
I
n
d
o
n
esian
co
f
f
ee
b
ea
n
ty
p
es.
I
m
p
lem
e
n
tin
g
C
NN
ca
n
im
p
r
o
v
e
p
r
o
d
u
ctio
n
ef
f
ici
en
cy
an
d
en
h
a
n
ce
th
e
v
al
u
e
o
f
I
n
d
o
n
esian
co
f
f
ee
co
m
m
o
d
it
ies
b
y
r
ed
u
cin
g
th
e
tim
e
an
d
co
s
ts
ass
o
ciate
d
with
th
e
s
o
r
tin
g
p
r
o
ce
s
s
.
Mo
r
eo
v
er
,
th
is
r
esear
ch
ca
n
ass
is
t
co
f
f
ee
p
r
o
d
u
ce
r
s
i
n
p
r
o
ce
s
s
in
g
d
ata
r
elate
d
t
o
co
f
f
ee
b
ea
n
ty
p
es
a
n
d
q
u
ality
f
r
o
m
v
ar
io
u
s
s
o
u
r
ce
s
,
lea
d
in
g
to
m
o
r
e
ac
cu
r
ate
an
d
b
etter
class
if
icatio
n
o
u
tco
m
es.
T
h
is
s
tu
d
y
also
co
n
tr
ib
u
tes
a
n
o
v
el
d
ataset
co
m
p
r
is
in
g
1
4
,
5
2
5
im
a
g
es
f
r
o
m
1
1
6
s
in
g
le
-
o
r
ig
in
I
n
d
o
n
esian
co
f
f
ee
b
ea
n
ty
p
es,
wh
ich
,
to
o
u
r
k
n
o
wled
g
e,
r
e
p
r
esen
ts
th
e
m
o
s
t
ex
ten
s
iv
e
d
ataset
o
f
its
k
in
d
.
T
h
e
im
ag
es
wer
e
co
llected
u
n
d
er
s
tan
d
ar
d
ized
lig
h
tin
g
a
n
d
i
m
ag
in
g
c
o
n
d
itio
n
s
,
e
n
s
u
r
in
g
co
n
s
is
ten
cy
wh
ile
ca
p
tu
r
in
g
th
e
wid
e
m
o
r
p
h
o
l
o
g
ical
an
d
c
o
lo
r
v
ar
iatio
n
s
in
h
er
en
t
to
I
n
d
o
n
esia’
s
d
iv
er
s
e
co
f
f
ee
r
eg
io
n
s
.
B
ey
o
n
d
s
u
p
p
o
r
tin
g
th
is
s
tu
d
y
,
th
e
d
at
aset
o
f
f
er
s
s
u
b
s
tan
tial
p
o
ten
ti
al
as
a
b
en
ch
m
ar
k
f
o
r
f
u
tu
r
e
co
f
f
ee
class
if
icatio
n
r
esear
ch
,
f
ac
ilit
atin
g
cr
o
s
s
-
co
m
p
ar
is
o
n
o
f
d
ee
p
lea
r
n
in
g
m
o
d
els
an
d
ad
v
a
n
cin
g
th
e
d
ev
elo
p
m
en
t
o
f
m
o
r
e
g
en
er
alize
d
AI
s
y
s
tem
s
f
o
r
ag
r
icu
ltu
r
al
p
r
o
d
u
ct
r
ec
o
g
n
itio
n
.
2.
L
I
T
E
R
AT
U
RE
R
E
VI
E
W
T
h
e
class
if
icat
io
n
p
r
o
c
e
s
s
f
o
r
c
o
f
f
e
e
b
e
a
n
s
c
a
n
b
e
c
a
r
r
i
e
d
o
u
t b
y
o
b
s
e
r
v
i
n
g
t
h
e
c
o
l
o
r
,
s
h
a
p
e
,
s
i
z
e
,
t
a
s
t
e
,
a
n
d
a
r
o
m
a
c
h
a
r
a
c
t
e
r
i
s
ti
c
s
o
f
t
h
e
b
e
a
n
s
a
f
t
e
r
t
h
e
d
r
y
i
n
g
p
r
o
c
e
s
s
[
2
0
]
,
[
2
1
]
.
H
o
w
e
v
e
r
,
t
h
i
s
p
r
o
c
e
s
s
r
e
q
u
i
r
es
p
r
e
c
i
s
i
o
n
a
n
d
c
o
n
s
is
t
e
n
c
y
,
w
h
ic
h
p
r
e
s
e
n
ts
c
h
al
l
e
n
g
es
w
h
e
n
p
er
f
o
r
m
e
d
m
a
n
u
a
l
l
y
b
y
w
o
r
k
e
r
s
.
A
s
a
r
es
u
l
t
,
v
a
r
i
o
u
s
s
u
p
p
o
r
t
i
n
g
m
e
t
h
o
d
s
h
a
v
e
b
e
e
n
d
e
v
e
l
o
p
e
d
t
o
a
u
t
o
m
a
t
e
t
h
e
c
o
f
f
e
e
b
e
a
n
c
l
a
s
s
i
f
i
c
at
i
o
n
p
r
o
c
e
s
s
,
o
n
e
o
f
w
h
i
c
h
i
s
t
h
e
u
s
e
o
f
A
I
t
h
r
o
u
g
h
d
e
e
p
l
e
a
r
n
in
g
m
e
t
h
o
d
s
.
D
e
e
p
le
a
r
n
i
n
g
i
s
a
b
r
a
n
c
h
o
f
m
a
c
h
i
n
e
l
e
a
r
n
i
n
g
c
a
p
a
b
l
e
o
f
s
o
l
v
i
n
g
h
i
g
h
l
y
c
o
m
p
l
e
x
p
r
o
b
lem
s
b
y
m
i
m
i
c
k
i
n
g
t
h
e
w
a
y
t
h
e
h
u
m
a
n
b
r
a
i
n
w
o
r
k
s
.
C
o
m
p
o
n
e
n
t
s
o
f
d
e
e
p
l
e
a
r
n
i
n
g
i
n
c
l
u
d
e
i
n
p
u
t
,
c
o
n
v
o
l
u
t
i
o
n
,
p
o
o
l
i
n
g
,
f
u
l
l
y
c
o
n
n
e
c
t
e
d
(
d
e
n
s
e
)
la
y
e
r
s
,
an
d
a
c
t
i
v
a
t
i
o
n
f
u
n
c
t
i
o
n
s
,
d
e
p
e
n
d
i
n
g
o
n
t
h
e
n
e
t
w
o
r
k
u
s
e
d
[
2
2
]
.
D
e
e
p
le
a
r
n
i
n
g
m
a
k
e
s
p
r
e
d
i
c
t
i
o
n
s
b
y
l
e
a
r
n
i
n
g
f
r
o
m
i
n
p
u
t
d
a
t
a
t
o
g
e
n
e
r
a
t
e
t
h
e
d
e
s
i
r
e
d
o
u
t
p
u
t
,
i
d
e
n
t
i
f
y
i
n
g
t
h
e
c
o
r
r
e
c
t
i
t
e
m
o
r
a
c
t
i
o
n
[
2
3
]
.
I
n
t
h
i
s
s
t
u
d
y
,
t
h
e
d
a
t
a
c
o
n
s
is
t
s
o
f
i
m
a
g
e
s
o
f
c
o
f
f
e
e
b
e
a
n
s
,
w
h
i
c
h
a
r
e
u
s
e
d
t
o
p
r
o
d
u
c
e
o
u
t
p
u
t
i
n
t
h
e
f
o
r
m
o
f
t
h
e
c
o
r
r
e
c
t
t
y
p
e
a
n
d
q
u
a
l
i
t
y
c
l
ass
i
f
i
c
at
i
o
n
o
f
t
h
e
b
e
a
n
s
.
D
e
e
p
l
e
a
r
n
i
n
g
h
a
s
b
e
e
n
w
i
d
e
l
y
u
s
e
d
i
n
v
a
r
i
o
u
s
c
l
a
s
s
i
f
i
c
a
t
i
o
n
p
r
o
c
e
s
s
e
s
,
e
s
p
e
c
i
a
l
l
y
i
n
t
h
e
a
g
r
i
c
u
l
t
u
r
a
l
f
i
e
l
d
[
2
4
]
.
I
d
e
n
t
i
f
y
i
n
g
a
n
d
c
l
a
s
s
i
f
y
i
n
g
p
a
r
t
s
o
f
p
l
a
n
t
s
t
o
e
n
s
u
r
e
t
h
e
e
x
p
e
c
t
e
d
q
u
a
l
i
t
y
h
a
s
b
e
e
n
a
f
o
c
u
s
o
f
m
a
n
y
p
r
e
v
i
o
u
s
s
t
u
d
i
e
s
.
I
n
a
s
t
u
d
y
b
y
N
a
s
i
r
e
t
a
l
.
[
2
5
]
,
a
C
N
N
m
o
d
e
l
u
s
e
d
t
o
d
e
t
e
c
t
d
i
s
e
a
s
e
s
i
n
f
r
u
i
t
s
b
a
s
e
d
o
n
c
o
l
o
r
a
n
d
s
h
a
p
e
c
h
a
r
a
c
t
e
r
i
s
t
i
c
s
a
c
h
i
e
v
e
d
h
i
g
h
d
e
t
e
c
t
i
o
n
a
c
c
u
r
a
c
y
.
B
e
y
o
n
d
f
r
u
i
t
d
i
s
e
a
s
e
d
e
t
e
c
t
i
o
n
,
C
N
N
m
o
d
e
l
s
h
a
v
e
a
l
s
o
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
4
,
No
.
6
,
Dec
em
b
er
2
0
2
5
:
5
1
4
0
-
5
1
5
6
5142
b
e
e
n
a
p
p
l
i
e
d
t
o
c
l
a
s
s
i
f
y
f
r
u
i
t
t
y
p
e
s
w
i
t
h
a
n
a
c
c
u
r
a
c
y
r
a
t
e
o
f
9
3
%
[
1
]
.
E
a
r
l
y
d
i
s
e
a
s
e
i
d
e
n
t
i
f
i
c
a
t
i
o
n
i
n
p
l
a
n
t
s
t
h
r
o
u
g
h
l
e
a
f
a
n
a
l
y
s
i
s
h
a
s
a
l
s
o
b
e
e
n
c
o
n
d
u
c
t
e
d
u
s
i
n
g
C
N
N
a
n
d
i
m
a
g
e
r
e
c
o
g
n
i
t
i
o
n
t
e
c
h
n
i
q
u
e
s
t
o
a
c
h
i
e
v
e
m
o
r
e
a
c
c
u
r
a
t
e
a
n
d
e
f
f
i
c
i
e
n
t
d
e
t
e
c
t
i
o
n
[
2
6
]
.
M
o
r
e
o
v
e
r
,
d
e
e
p
l
e
a
r
n
i
n
g
h
a
s
b
e
e
n
u
s
e
d
t
o
i
d
e
n
t
i
f
y
s
e
e
d
s
,
s
u
c
h
a
s
i
n
t
h
e
s
t
u
d
y
b
y
J
u
m
a
r
l
i
s
e
t
a
l
.
[
2
7
]
,
w
h
e
r
e
K
N
N
w
a
s
u
s
e
d
t
o
d
e
t
e
c
t
d
e
f
e
c
t
s
i
n
c
o
f
f
e
e
b
e
a
n
s
.
T
h
u
s
,
d
e
e
p
l
e
a
r
n
i
n
g
i
s
e
m
e
r
g
i
n
g
a
s
a
p
r
o
m
i
s
i
n
g
t
o
o
l
f
o
r
t
h
e
i
n
s
p
e
c
t
i
o
n
a
n
d
c
l
a
s
s
i
f
i
c
a
t
i
o
n
o
f
f
o
o
d
q
u
a
l
i
t
y
a
n
d
s
a
f
e
t
y
i
n
t
h
e
a
g
r
i
c
u
l
t
u
r
a
l
s
e
c
t
o
r
[
2
8
]
.
On
e
o
f
th
e
m
o
s
t
co
m
m
o
n
ly
u
s
ed
d
ee
p
lear
n
in
g
alg
o
r
ith
m
s
f
o
r
class
if
icatio
n
an
d
id
e
n
tific
atio
n
p
r
o
ce
s
s
es
is
C
NN.
C
NN
s
p
ec
ializes
in
p
atter
n
r
ec
o
g
n
itio
n
an
d
u
s
es
co
n
n
ec
ted
lay
e
r
s
o
f
p
o
o
lin
g
a
n
d
co
n
v
o
l
u
tio
n
to
tr
an
s
f
o
r
m
an
i
n
p
u
t
im
ag
e
in
to
a
s
er
ies
o
f
f
ea
tu
r
e
m
ap
s
,
wh
ich
ar
e
th
en
class
if
ied
b
y
f
u
lly
co
n
n
ec
ted
lay
er
s
b
ased
o
n
th
e
im
ag
e’
s
f
ea
tu
r
es
[
2
9
]
.
T
h
is
m
ec
h
an
is
m
allo
ws
C
NN
to
ex
ce
l
in
im
ag
e
r
ec
o
g
n
itio
n
p
r
o
ce
s
s
es,
ef
f
ec
tiv
ely
ex
tr
ac
tin
g
c
o
lo
r
an
d
s
h
ap
e
f
ea
tu
r
es
f
r
o
m
im
a
g
es
[
3
0
]
.
C
NN
is
r
elativ
ely
ea
s
y
to
tr
ain
an
d
d
ev
elo
p
,
d
ep
en
d
in
g
o
n
th
e
b
r
ea
d
th
a
n
d
d
ep
th
o
f
th
e
d
ataset
u
s
ed
d
u
r
i
n
g
m
o
d
el
tr
ai
n
in
g
,
an
d
ca
n
ac
cu
r
ately
p
r
e
d
ict
im
ag
e
p
r
o
p
er
ties
[
3
1
]
.
Ho
wev
e
r
,
t
o
ac
h
iev
e
h
ig
h
ac
cu
r
ac
y
,
C
NN
r
eq
u
ir
es
lar
g
e
d
atasets
o
f
im
ag
e
s
am
p
les
d
u
r
in
g
th
e
tr
ain
i
n
g
p
h
ase
[
3
2
]
.
T
h
e
im
p
o
r
ta
n
ce
o
f
p
r
e
-
tr
ain
i
n
g
to
ac
h
ie
v
e
h
ig
h
ac
cu
r
ac
y
h
as
b
ee
n
d
em
o
n
s
tr
at
ed
b
y
U
n
al
et
a
l
.
[
3
3
]
,
wh
o
e
m
p
lo
y
ed
f
o
u
r
C
NN
m
o
d
els
to
d
etec
t
co
f
f
ee
b
ea
n
s
,
all
o
f
wh
ich
ac
h
ie
v
ed
h
i
g
h
ac
c
u
r
ac
y
af
ter
p
r
e
-
t
r
ain
in
g
.
Giv
en
its
ad
v
a
n
tag
es,
C
NN
h
as
b
ee
n
wid
ely
d
e
v
elo
p
ed
an
d
im
p
lem
en
ted
in
c
o
f
f
ee
b
ea
n
class
if
icatio
n
p
r
o
ce
s
s
es.
Pre
v
io
u
s
r
esear
ch
h
as
d
em
o
n
s
tr
at
ed
th
e
h
ig
h
ac
c
u
r
ac
y
o
f
C
NN
in
class
if
icatio
n
task
s
,
in
clu
d
in
g
th
o
s
e
in
v
o
lv
in
g
ty
p
e,
q
u
ality
,
a
n
d
d
e
f
ec
t
d
etec
tio
n
.
B
u
o
n
o
co
r
e
et
a
l
.
[
3
4
]
u
s
ed
C
NN
to
au
to
m
atica
lly
class
if
y
m
ix
ed
Ar
ab
ica
an
d
R
o
b
u
s
ta
co
f
f
ee
b
ea
n
s
,
d
is
tin
g
u
is
h
in
g
th
em
b
ased
o
n
s
h
ap
e,
co
lo
r
,
an
d
s
ize
f
r
o
m
ca
p
tu
r
e
d
im
ag
es
,
with
h
ig
h
ly
ac
cu
r
ate
r
esu
lts
.
W
an
g
et
a
l
.
[
3
5
]
a
n
d
Prz
y
b
y
ł
et
a
l
.
[
3
6
]
also
d
em
o
n
s
tr
ated
th
at
co
m
b
in
in
g
C
NN
with
co
m
p
u
ter
v
is
io
n
ca
n
ac
c
u
r
ately
i
d
en
tify
co
f
f
ee
b
ea
n
q
u
ality
.
Mic
ar
aseth
et
a
l
.
[
3
7
]
em
p
lo
y
ed
th
r
e
e
C
NN
m
o
d
els
f
o
r
co
f
f
ee
d
ef
ec
t
d
etec
tio
n
,
ac
h
ie
v
in
g
h
ig
h
ac
cu
r
ac
y
an
d
s
h
o
r
t
d
etec
tio
n
tim
es
with
all
th
r
ee
m
o
d
els.
I
n
a
d
d
itio
n
to
s
h
o
r
ten
i
n
g
p
r
o
ce
s
s
in
g
tim
es,
C
h
en
et
a
l
.
[
3
8
]
d
em
o
n
s
tr
a
ted
th
at
a
m
u
ltimo
d
al
r
ea
l
-
tim
e
alg
o
r
it
h
m
u
s
in
g
C
NN
an
d
a
n
ea
r
-
in
f
r
ar
ed
s
n
a
p
s
h
o
t
h
y
p
er
s
p
ec
tr
al
s
en
s
o
r
f
o
r
co
f
f
ee
b
ea
n
d
etec
tio
n
c
o
u
ld
r
ed
u
ce
lab
o
r
c
o
s
ts
wh
ile
m
ain
tain
in
g
h
ig
h
ac
c
u
r
ac
y
.
T
h
e
u
s
e
o
f
C
NN
to
im
p
r
o
v
e
d
ef
ec
t
d
etec
tio
n
ac
cu
r
ac
y
in
c
o
f
f
ee
b
ea
n
s
h
as
also
b
ee
n
p
r
o
v
en
b
y
Pin
to
e
t
a
l
.
[
3
9
]
,
w
h
o
class
if
ied
co
f
f
ee
b
ea
n
s
in
to
s
ix
ty
p
es
o
f
d
ef
ec
ts
,
in
cr
ea
s
in
g
ac
cu
r
ac
y
f
r
o
m
7
2
t
o
9
8
%.
I
n
t
h
is
s
tu
d
y
,
th
e
a
u
to
m
atic
d
etec
tio
n
o
f
I
n
d
o
n
esian
c
o
f
f
ee
b
ea
n
ty
p
es
is
co
n
d
u
cte
d
b
y
d
ev
elo
p
in
g
an
d
ap
p
ly
i
n
g
a
v
is
u
al
-
b
ased
d
ee
p
lea
r
n
in
g
m
e
th
o
d
u
s
in
g
C
NN.
T
h
is
s
tu
d
y
m
ak
es
s
ev
er
al
k
ey
co
n
tr
ib
u
tio
n
s
to
th
e
f
ield
o
f
co
f
f
ee
b
ea
n
class
if
icatio
n
.
First,
it
u
tili
ze
s
a
d
ataset
co
n
s
is
tin
g
o
f
1
1
6
ty
p
es
o
f
s
in
g
le
-
o
r
ig
in
co
f
f
ee
,
w
h
ich
,
to
th
e
b
est
o
f
o
u
r
k
n
o
wl
ed
g
e,
is
th
e
lar
g
est
d
ataset
o
f
I
n
d
o
n
esian
c
o
f
f
ee
v
ar
ieties
co
m
p
iled
to
d
ate.
T
h
is
ex
ten
s
iv
e
d
ataset
p
r
o
v
id
e
s
a
co
m
p
r
eh
en
s
iv
e
f
o
u
n
d
atio
n
f
o
r
a
n
aly
zin
g
th
e
d
iv
er
s
e
ch
ar
ac
te
r
is
tics
o
f
I
n
d
o
n
esian
co
f
f
ee
b
ea
n
s
.
Seco
n
d
,
th
e
d
ataset
was
ca
p
tu
r
ed
u
s
in
g
a
s
tan
d
ar
d
d
i
g
ital
ca
m
er
a,
av
o
id
in
g
th
e
n
ee
d
f
o
r
co
m
p
lex
an
d
co
s
tly
im
ag
in
g
s
etu
p
s
,
m
ak
in
g
th
e
m
eth
o
d
m
o
r
e
ac
ce
s
s
ib
le
a
n
d
r
ep
licab
le
f
o
r
o
th
er
r
esear
c
h
er
s
an
d
p
r
ac
titi
o
n
er
s
.
Fin
ally
,
th
is
s
tu
d
y
in
v
o
lv
es
th
e
cr
ea
tio
n
an
d
d
ev
el
o
p
m
en
t
o
f
C
NN
m
o
d
el
s
s
p
ec
if
ically
d
esig
n
ed
f
o
r
th
e
v
is
u
al
p
r
ed
ic
tio
n
o
f
co
f
f
ee
b
ea
n
ty
p
es
u
s
in
g
t
h
is
d
ataset.
T
h
e
s
e
C
NN
m
o
d
el
s
ar
e
o
p
tim
i
ze
d
to
d
eliv
er
ac
cu
r
ate
class
if
icatio
n
s
,
th
er
eb
y
ad
v
an
cin
g
th
e
u
s
e
o
f
AI
in
t
h
e
co
f
f
ee
in
d
u
s
tr
y
an
d
c
o
n
tr
ib
u
tin
g
to
th
e
b
r
o
ad
e
r
g
o
al
o
f
im
p
r
o
v
i
n
g
ef
f
icien
c
y
an
d
p
r
ec
is
io
n
in
co
f
f
ee
b
ea
n
s
o
r
tin
g
p
r
o
ce
s
s
es.
3.
M
E
T
H
O
DS
T
h
is
s
tu
d
y
d
ev
el
o
p
s
p
r
e
d
ictio
n
m
o
d
els
u
s
in
g
two
d
if
f
er
en
t
m
eth
o
d
s
.
T
h
e
f
ir
s
t
m
eth
o
d
i
s
tr
an
s
f
er
lear
n
in
g
,
wh
ich
u
tili
ze
s
th
e
p
r
etr
ain
ed
I
n
ce
p
tio
n
V3
m
o
d
el
.
T
h
e
s
ec
o
n
d
m
eth
o
d
in
v
o
l
v
e
s
b
u
ild
in
g
a
m
o
d
el
f
r
o
m
s
cr
atch
,
with
p
r
ep
r
o
ce
s
s
in
g
ap
p
lie
d
in
b
o
th
to
alig
n
th
e
d
ataset
with
th
e
r
esear
ch
o
b
jec
tiv
es.
3
.
1
.
Da
t
a
prepro
ce
s
s
ing
Sev
er
al
p
r
ep
r
o
ce
s
s
in
g
s
tep
s
w
er
e
ca
r
r
ied
o
u
t,
as
s
h
o
wn
in
Fi
g
u
r
e
1
,
in
cl
u
d
in
g
s
h
u
f
f
lin
g
t
h
e
d
ata
an
d
s
p
litt
in
g
it
in
to
tr
ain
in
g
,
v
alid
atio
n
,
an
d
test
in
g
s
ets.
Af
ter
war
d
,
r
esizin
g
,
r
escalin
g
,
an
d
d
ata
a
u
g
m
e
n
tatio
n
wer
e
p
er
f
o
r
m
ed
.
T
h
e
d
ata
s
h
u
f
f
lin
g
p
r
o
ce
s
s
was
d
o
n
e
r
an
d
o
m
ly
ac
r
o
s
s
th
e
d
ataset.
T
h
is
was
f
o
llo
wed
b
y
r
esizin
g
an
d
r
escalin
g
th
e
d
ata.
Ad
d
itio
n
ally
,
au
g
m
en
tatio
n
was
o
n
ly
a
p
p
lied
t
o
th
e
tr
ain
i
n
g
d
ata
to
in
tr
o
d
u
ce
m
o
r
e
v
a
r
iatio
n
an
d
im
p
r
o
v
e
th
e
m
o
d
el'
s
p
er
f
o
r
m
a
n
ce
d
u
r
in
g
tr
ain
in
g
.
3.
2
.
T
ra
ns
f
er
lea
rning
m
o
de
l
T
h
e
m
o
d
el
d
ev
elo
p
m
en
t
p
r
o
c
ess
in
v
o
lv
es
tr
an
s
f
er
lear
n
i
n
g
,
u
tili
zin
g
th
e
p
r
etr
ain
ed
I
n
ce
p
tio
n
V3
m
o
d
el
alo
n
g
with
weig
h
ts
o
b
tain
ed
f
r
o
m
p
r
ev
io
u
s
tr
ain
in
g
o
n
d
if
f
er
e
n
t
d
atasets
.
T
wo
f
in
e
-
tu
n
in
g
s
ce
n
ar
io
s
ar
e
ap
p
lied
d
u
r
in
g
tr
ain
in
g
: in
th
e
f
ir
s
t scen
ar
io
,
h
alf
o
f
th
e
f
ea
tu
r
e
ex
tr
ac
tio
n
lay
er
s
ar
e
u
n
f
r
o
ze
n
,
wh
ile
in
th
e
s
ec
o
n
d
s
ce
n
ar
io
,
all
th
e
lay
er
s
ar
e
u
n
f
r
o
ze
n
.
T
h
e
I
n
ce
p
tio
n
V3
m
o
d
el
was
s
elec
ted
d
u
e
t
o
its
h
ig
h
ac
cu
r
ac
y
wh
en
tr
ain
ed
o
n
th
e
I
m
a
g
eNe
t
d
ataset
an
d
its
co
m
p
lex
ar
ch
itectu
r
e,
wh
ich
in
clu
d
es
2
3
.
9
m
illi
o
n
p
ar
am
eter
s
.
T
h
e
d
etails o
f
th
e
tr
ai
n
in
g
s
ce
n
ar
io
s
ar
e
as f
o
llo
ws
.
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
C
la
s
s
i
fica
tio
n
o
f sin
g
le
o
r
ig
in
I
n
d
o
n
esia
n
co
ffee
b
e
a
n
s
u
s
in
g
co
n
vo
lu
tio
n
a
l …
(
A
c
h
ma
d
P
r
a
ta
ma
R
ifa
i)
5143
Scen
ar
io
1
.
I
n
th
e
f
ir
s
t
s
ce
n
ar
io
,
th
e
m
o
d
el
is
tr
ain
ed
b
y
f
r
ee
zin
g
all
f
ea
tu
r
e
ex
tr
ac
tio
n
l
ay
er
s
an
d
r
ep
lacin
g
th
e
class
if
icatio
n
la
y
er
to
s
u
it
th
e
s
p
ec
if
ic
task
o
f
c
o
f
f
ee
class
if
icatio
n
.
C
o
n
s
eq
u
e
n
tly
,
th
e
weig
h
ts
i
n
th
e
f
ea
tu
r
e
ex
t
r
ac
tio
n
la
y
er
s
r
em
ain
u
n
ch
an
g
ed
.
T
h
is
s
tep
a
im
s
to
tr
ain
th
e
class
if
icatio
n
lay
er
to
alig
n
with
th
e
o
b
jectiv
es
o
f
th
is
m
o
d
el
d
ev
elo
p
m
en
t.
T
h
e
tr
ain
in
g
p
r
o
c
ess
is
ca
r
r
ied
o
u
t
f
o
r
2
5
ep
o
c
h
s
,
r
ep
r
esen
tin
g
h
alf
o
f
th
e
to
tal
ep
o
c
h
s
p
lan
n
e
d
f
o
r
m
o
d
el
tr
ain
in
g
.
Af
ter
war
d
,
h
alf
o
f
th
e
f
ea
tu
r
e
ex
tr
ac
tio
n
l
ay
er
s
ar
e
u
n
f
r
o
ze
n
,
allo
win
g
f
o
r
weig
h
t
u
p
d
ates
in
th
ese
lay
er
s
.
T
h
is
f
in
e
-
tu
n
in
g
s
tep
is
p
e
r
f
o
r
m
ed
t
o
ad
j
u
s
t
th
e
p
r
e
v
io
u
s
ly
p
r
etr
ain
ed
weig
h
ts
to
th
e
cu
r
r
en
t
d
ataset,
with
an
o
th
er
2
5
ep
o
ch
s
o
f
tr
ain
in
g
.
T
h
is
ap
p
r
o
ac
h
o
p
tim
izes
th
e
m
o
d
el
f
o
r
th
e
s
p
ec
if
ic
class
if
icatio
n
task
in
th
is
r
esear
ch
.
Scen
ar
io
2
.
I
n
th
e
s
ec
o
n
d
s
ce
n
ar
io
,
all
f
ea
tu
r
e
ex
tr
ac
ti
o
n
la
y
er
s
ar
e
u
n
f
r
o
ze
n
,
an
d
th
e
cla
s
s
if
icatio
n
lay
er
is
r
ep
lace
d
to
s
u
it
th
e
co
f
f
ee
class
if
icatio
n
task
.
T
h
is
ad
ju
s
tm
en
t
allo
ws
th
e
en
tire
m
o
d
el
to
lear
n
f
r
o
m
s
cr
atch
,
ac
co
u
n
tin
g
f
o
r
p
o
ten
ti
al
d
if
f
er
en
ce
s
in
th
e
d
atasets
u
s
ed
in
p
r
ev
io
u
s
tr
ain
in
g
.
As
a
r
esu
lt,
th
e
weig
h
ts
will
b
e
u
p
d
ated
at
ea
ch
s
tag
e
o
f
th
e
tr
ain
in
g
p
r
o
ce
s
s
,
wh
ich
is
co
n
d
u
cted
o
v
e
r
5
0
ep
o
ch
s
.
I
n
ce
p
tio
n
V3
.
T
h
is
p
r
etr
ain
ed
m
o
d
el
is
a
n
a
d
v
a
n
ce
m
en
t
b
y
Szeg
ed
y
et
a
l
.
[
4
0
]
o
v
er
th
e
p
r
ev
io
u
s
I
n
ce
p
t
io
n
m
o
d
el
a
n
d
was
tr
ain
ed
o
n
th
e
I
m
ag
eNe
t
d
ata
s
et.
I
n
ce
p
tio
n
V
3
co
m
p
r
is
es
co
n
v
o
l
u
tio
n
al
b
l
o
ck
s
,
I
n
ce
p
tio
n
m
o
d
u
les,
an
d
a
class
if
ier
,
with
a
to
tal
o
f
4
8
la
y
er
s
.
T
h
e
ar
c
h
itectu
r
e
o
f
th
is
m
o
d
el
is
s
h
o
wn
in
Fig
u
r
e
2
.
Fig
u
r
e
1
.
Data
p
r
ep
r
o
ce
s
s
in
g
s
tag
es
Fig
u
r
e
2
.
Ar
c
h
itectu
r
e
o
f
I
n
ce
p
tio
n
V3
[
4
1
]
3.
3
.
T
ra
ini
ng
f
ro
m
s
cr
a
t
ch
m
o
del
T
h
e
b
u
il
d
-
f
r
o
m
-
s
c
r
a
tc
h
a
p
p
r
o
a
ch
i
n
v
o
l
v
es
d
es
ig
n
i
n
g
th
e
m
o
d
el
ar
ch
ite
ct
u
r
e
f
r
o
m
t
h
e
g
r
o
u
n
d
u
p
.
T
h
e
d
e
v
e
lo
p
ed
s
tr
u
c
tu
r
e
c
o
n
s
is
ts
o
f
f
e
at
u
r
e
e
x
t
r
a
cti
o
n
la
y
e
r
s
,
w
h
e
r
e
ea
c
h
co
n
v
o
l
u
ti
o
n
al
la
y
e
r
h
a
s
a
s
p
ec
i
f
i
c
n
u
m
b
e
r
an
d
s
ize
o
f
f
ilte
r
s
,
f
o
llo
wed
b
y
a
m
a
x
p
o
o
li
n
g
l
ay
e
r
.
T
o
o
p
ti
m
iz
e
t
h
e
m
o
d
e
l’
s
a
r
c
h
i
tec
tu
r
e
,
th
e
T
a
g
u
ch
i m
et
h
o
d
is
ap
p
l
ie
d
u
s
i
n
g
a
d
esi
g
n
o
f
e
x
p
e
r
i
m
e
n
t
(
D
o
E
)
a
p
p
r
o
ac
h
,
ai
m
in
g
t
o
f
in
d
a
r
o
b
u
s
t
c
o
m
b
i
n
a
ti
o
n
o
f
t
h
e
n
u
m
b
e
r
o
f
f
e
at
u
r
e
ex
tr
ac
t
io
n
la
y
er
s
a
n
d
tr
ain
in
g
e
p
o
c
h
s
.
T
h
e
p
r
o
ce
s
s
f
l
o
w
is
ill
u
s
t
r
ate
d
i
n
Fi
g
u
r
e
3
.
T
h
e
p
r
o
ce
s
s
b
eg
in
s
with
p
r
o
b
lem
f
o
r
m
u
latio
n
,
wh
er
e
t
h
e
g
o
al
is
to
f
in
d
th
e
o
p
tim
al
co
m
b
in
atio
n
b
etwe
en
th
e
n
u
m
b
er
o
f
tr
ain
i
n
g
ep
o
c
h
s
an
d
f
ea
tu
r
e
e
x
tr
ac
t
io
n
lay
er
s
u
s
ed
.
T
h
e
o
u
tp
u
t
i
n
d
icato
r
will
b
e
th
e
ac
cu
r
ac
y
ac
h
iev
e
d
o
n
t
h
e
v
alid
atio
n
d
ata
d
u
r
i
n
g
th
e
f
in
al
ep
o
ch
o
f
m
o
d
el
tr
ain
in
g
,
with
th
e
g
o
al
o
f
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
4
,
No
.
6
,
Dec
em
b
er
2
0
2
5
:
5
1
4
0
-
5
1
5
6
5144
m
ax
im
izin
g
th
is
v
alu
e
.
Nex
t,
co
n
tr
o
l
f
ac
t
o
r
s
ar
e
id
en
tifie
d
:
th
e
n
u
m
b
er
o
f
tr
ain
in
g
ep
o
c
h
s
,
wh
ich
af
f
ec
ts
th
e
r
is
k
o
f
u
n
d
er
f
itti
n
g
o
r
o
v
er
f
itti
n
g
,
a
n
d
th
e
n
u
m
b
er
o
f
f
ea
tu
r
e
ex
tr
ac
tio
n
la
y
er
s
,
wh
ic
h
im
p
a
cts
th
e
d
e
p
th
o
f
th
e
m
o
d
el’
s
ar
ch
itectu
r
e.
T
h
en
,
ex
p
er
im
en
ts
ar
e
co
n
d
u
cted
u
s
in
g
co
m
b
in
atio
n
s
o
f
th
ese
f
ac
to
r
s
b
ased
o
n
a
p
r
ed
ef
in
e
d
d
esig
n
.
Usi
n
g
th
e
L
9
o
r
t
h
o
g
o
n
al
ar
r
ay
,
n
in
e
c
o
m
b
in
atio
n
s
with
th
r
ee
lev
els
ar
e
id
en
tifie
d
:
t
h
e
n
u
m
b
er
o
f
e
p
o
ch
s
s
et
at
5
0
,
7
5
,
an
d
1
0
0
,
a
n
d
th
e
n
u
m
b
er
o
f
f
ea
tu
r
e
e
x
tr
ac
tio
n
lay
er
s
s
et
at
5
,
6
,
an
d
7
.
T
h
e
tr
ain
in
g
p
r
o
ce
s
s
is
ca
r
r
ied
o
u
t
u
s
in
g
th
e
s
am
e
eq
u
ip
m
en
t
to
m
in
im
ize
n
o
is
e.
T
h
e
ex
p
e
r
im
en
t
is
an
aly
ze
d
b
ased
o
n
th
e
s
ig
n
al
-
to
-
n
o
is
e
(
S/N)
r
atio
,
with
th
e
"lar
g
er
is
b
etter
"
cr
iter
io
n
u
s
ed
to
m
ax
i
m
ize
ac
cu
r
ac
y
.
T
h
is
ap
p
r
o
ac
h
h
elp
s
id
en
tify
a
r
o
b
u
s
t
co
m
b
in
atio
n
th
at
will
y
ield
th
e
h
ig
h
est
m
o
d
el
ac
cu
r
ac
y
.
T
h
e
r
esu
lts
o
f
th
is
an
aly
s
is
s
er
v
e
as th
e
b
asis
f
o
r
m
o
d
el
d
e
v
elo
p
m
e
n
t.
3.
4
.
M
o
del per
f
o
rma
nce
ev
a
lua
t
io
n ind
ica
t
o
rs
Sev
er
al
in
d
icato
r
s
ar
e
u
s
ed
to
an
aly
ze
th
e
p
er
f
o
r
m
an
ce
o
f
th
e
d
ev
el
o
p
ed
m
o
d
el.
I
f
th
e
m
o
d
el
is
f
o
u
n
d
lack
in
g
b
ased
o
n
th
e
in
d
icato
r
s
lis
ted
in
T
a
b
le
1
,
a
h
y
p
er
p
ar
am
eter
tu
n
i
n
g
p
r
o
ce
s
s
will
b
e
in
itiated
.
T
h
e
k
ey
in
d
icato
r
s
u
s
ed
to
ass
ess
wh
eth
er
a
m
o
d
el
is
co
n
s
id
er
e
d
o
p
tim
al
in
clu
d
e
ac
h
iev
i
n
g
th
e
h
ig
h
est
ac
cu
r
ac
y
,
p
r
ec
is
io
n
,
an
d
r
e
ca
ll,
with
th
e
lo
west
p
o
s
s
ib
le
lo
s
s
,
wh
ile
a
v
o
id
in
g
s
ig
n
s
o
f
o
v
er
f
itti
n
g
o
r
u
n
d
e
r
f
itti
n
g
d
u
r
i
n
g
tr
ain
in
g
a
n
d
v
alid
atio
n
.
Ad
d
itio
n
ally
,
th
e
m
o
d
el
m
u
s
t
e
x
h
ib
it
s
tr
o
n
g
p
e
r
f
o
r
m
an
ce
o
n
th
e
test
d
ata,
as
d
em
o
n
s
tr
ated
b
y
h
i
g
h
v
al
u
es
in
ac
cu
r
ac
y
,
p
r
ec
is
io
n
,
r
ec
all,
an
d
F1
-
s
co
r
e
.
C
o
m
p
u
tatio
n
t
im
e,
esp
ec
ially
f
o
r
test
in
g
d
ata,
is
also
a
cr
u
cial
f
ac
to
r
wh
en
d
ete
r
m
in
in
g
w
h
ich
m
o
d
el
to
ap
p
ly
to
r
ea
l
-
wo
r
ld
p
r
o
b
lem
s
,
as
f
aster
m
o
d
els m
ay
b
e
m
o
r
e
p
r
ac
tical
f
o
r
im
p
lem
en
tatio
n
.
Fig
u
r
e
3
.
T
a
g
u
ch
i
m
eth
o
d
d
p
r
o
ce
s
s
f
lo
w
T
ab
le
1
.
Mo
d
el
p
er
f
o
r
m
a
n
ce
e
v
alu
atio
n
in
d
icato
r
s
P
a
r
a
me
t
e
r
I
n
d
i
c
a
t
o
r
Tr
a
i
n
i
n
g
r
e
s
u
l
t
Th
e
a
v
e
r
a
g
e
v
a
l
u
e
s
o
f
l
o
ss,
a
c
c
u
r
a
c
y
,
p
r
e
c
i
si
o
n
,
a
n
d
r
e
c
a
l
l
o
n
t
h
e
t
r
a
i
n
i
n
g
a
n
d
v
a
l
i
d
a
t
i
o
n
d
a
t
a
,
a
s
w
e
l
l
a
s
t
h
e
t
r
a
i
n
i
n
g
t
i
me
.
Te
st
i
n
g
r
e
su
l
t
A
n
a
l
y
s
i
n
g
t
h
e
a
v
e
r
a
g
e
a
c
c
u
r
a
c
y
,
p
r
e
c
i
si
o
n
,
r
e
c
a
l
l
,
a
n
d
F
1
-
sc
o
r
e
f
o
r
b
o
t
h
c
l
a
s
ses,
a
l
o
n
g
w
i
t
h
t
h
e
t
e
st
i
n
g
t
i
me.
4.
E
XP
E
R
I
M
E
N
T
S AN
D
RE
S
UL
T
S
4
.
1
.
Da
t
a
s
et
T
h
is
s
tu
d
y
u
tili
ze
s
a
d
ataset
co
n
s
is
tin
g
o
f
p
h
o
to
s
co
llected
b
ased
o
n
v
ar
io
u
s
ty
p
es
o
f
c
o
f
f
ee
b
ea
n
s
f
r
o
m
ac
r
o
s
s
I
n
d
o
n
esia.
B
r
o
a
d
ly
s
p
ea
k
in
g
,
th
er
e
ar
e
f
o
u
r
m
ain
ty
p
es
o
f
co
f
f
ee
b
ea
n
s
:
Ar
ab
ica,
R
o
b
u
s
ta,
L
ib
er
ica,
an
d
E
x
ce
ls
a.
T
h
e
d
a
taset
in
clu
d
es
1
1
6
d
is
tin
ct
ty
p
es
o
f
co
f
f
ee
b
ea
n
s
ca
teg
o
r
ize
d
b
y
th
eir
g
r
o
win
g
r
eg
io
n
s
,
co
m
p
r
is
in
g
a
to
tal
o
f
1
4
,
5
2
5
im
ag
es.
Fig
u
r
e
4
p
r
es
en
ts
r
ep
r
esen
tativ
e
ex
am
p
les
o
f
th
r
ee
I
n
d
o
n
esian
co
f
f
ee
b
ea
n
ty
p
es
—
R
o
b
u
s
ta
Dam
p
it
in
Fig
u
r
e
4
(
a)
,
Ar
ab
ic
a
C
ih
u
r
ip
in
Fig
u
r
e
4
(
b
)
,
an
d
L
ib
er
ica
Su
m
ed
an
g
in
Fig
u
r
e
4
(
c)
—
s
elec
ted
f
r
o
m
a
to
tal
o
f
1
1
6
v
ar
ieties
in
clu
d
ed
in
th
e
d
ataset.
T
h
ese
ex
a
m
p
les
h
ig
h
lig
h
t
th
e
n
o
tab
le
m
o
r
p
h
o
l
o
g
ical
an
d
c
o
lo
r
v
ar
iatio
n
s
am
o
n
g
co
f
f
e
e
b
ea
n
s
o
r
ig
in
atin
g
f
r
o
m
d
if
f
er
en
t
r
eg
io
n
s
an
d
s
p
ec
ies,
illu
s
tr
atin
g
th
e
v
is
u
al
d
iv
er
s
ity
t
h
at
th
e
class
if
icati
o
n
m
o
d
el
m
u
s
t
r
ec
o
g
n
ize
.
T
h
e
d
etails
o
f
d
atasets
ar
e
p
r
esen
ted
in
T
ab
le
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
C
la
s
s
i
fica
tio
n
o
f sin
g
le
o
r
ig
in
I
n
d
o
n
esia
n
co
ffee
b
e
a
n
s
u
s
in
g
co
n
vo
lu
tio
n
a
l …
(
A
c
h
ma
d
P
r
a
ta
ma
R
ifa
i)
5145
(
a)
(
b
)
(
c)
Fig
u
r
e
4
.
E
x
am
p
les o
f
th
e
im
a
g
es f
r
o
m
th
r
ee
co
f
f
ee
v
ar
ieties
of
(
a)
R
o
b
u
s
ta
Dam
p
it,
(
b
)
Ar
ab
ica
C
ih
u
r
ip
,
an
d
(
c)
L
ib
er
ica
Su
m
e
d
an
g
T
ab
le
2
.
Deta
ils
o
f
th
e
d
atasets
Ty
p
e
o
f
c
o
f
f
e
e
b
e
a
n
s
N
u
mb
e
r
o
f
d
a
t
a
Ty
p
e
o
f
c
o
f
f
e
e
b
e
a
n
s
N
u
mb
e
r
o
f
d
a
t
a
Li
b
e
r
i
c
a
La
w
u
2
0
9
A
r
a
b
i
c
a
P
a
n
d
a
n
s
a
r
i
B
r
e
b
e
s
1
1
6
R
o
b
u
st
a
La
w
u
1
7
2
R
o
b
u
st
a
M
e
r
a
p
i
1
1
2
A
r
a
b
i
c
a
F
l
o
r
e
s
B
a
j
a
w
a
W
a
s
h
1
6
8
R
o
b
u
st
a
Te
ma
n
g
g
u
n
g
N
a
t
u
r
a
l
1
1
2
A
r
a
b
i
c
a
C
i
w
i
d
e
y
W
i
n
e
1
6
4
A
r
a
b
i
c
a
A
r
g
o
p
u
r
o
1
1
2
A
r
a
b
i
c
a
P
a
p
u
a
D
o
g
i
y
a
i
1
6
4
A
r
a
b
i
c
a
Te
m
a
n
g
g
u
n
g
H
o
n
e
y
1
1
1
A
r
a
b
i
c
a
G
u
n
u
n
g
S
u
mb
i
n
g
1
6
4
A
r
a
b
i
c
a
M
e
r
a
p
i
W
i
n
e
1
1
1
A
r
a
b
i
c
a
M
e
w
a
r
w
a
n
g
i
1
6
4
R
o
b
u
st
a
M
e
d
a
n
1
1
0
A
r
a
b
i
c
a
W
a
n
o
j
a
K
a
mo
j
a
n
g
1
6
4
R
o
b
u
st
a
F
l
o
r
e
s
1
1
0
Ex
c
e
l
s
a
G
u
n
u
n
g
S
i
n
d
o
r
o
1
6
4
A
r
a
b
i
c
a
La
w
u
1
0
9
A
r
a
b
i
c
a
P
u
n
t
a
n
g
J
a
w
a
B
a
r
a
t
1
6
3
A
r
a
b
i
c
a
A
c
e
h
G
a
y
o
H
o
n
e
y
1
0
8
Ex
c
e
l
s
a
I
j
e
n
R
a
u
n
g
1
6
2
R
o
b
u
st
a
D
a
m
p
i
t
1
0
8
A
r
a
b
i
c
a
To
r
a
j
a
R
a
n
t
e
k
a
r
u
a
1
6
0
A
r
a
b
i
c
a
C
i
w
i
d
e
y
L
u
w
a
k
1
0
8
A
r
a
b
i
c
a
I
j
e
n
C
a
r
b
o
n
i
c
M
a
s
c
e
r
a
s
i
N
a
t
u
r
a
l
1
5
6
A
r
a
b
i
c
a
G
o
w
a
M
a
l
i
n
o
S
p
e
c
i
a
l
i
t
y
1
0
8
A
r
a
b
i
c
a
To
r
a
j
a
N
a
p
o
1
5
6
A
r
a
b
i
k
S
u
b
a
n
g
N
a
t
u
r
a
l
1
0
8
R
o
b
u
st
a
M
a
n
d
a
i
l
i
n
g
S
u
m
a
t
r
a
1
5
4
A
r
a
b
i
c
a
R
a
n
c
a
b
a
l
i
H
o
n
e
y
1
0
6
A
r
a
b
i
c
a
En
r
e
k
a
n
g
L
a
t
i
m
o
j
o
n
g
1
5
4
A
r
a
b
i
c
a
B
e
n
g
k
u
l
u
1
0
6
A
r
a
b
i
c
a
G
n
.
P
a
t
u
h
a
1
5
4
A
r
a
b
i
c
a
To
r
a
j
a
F
u
l
l
w
a
sh
1
0
4
A
r
a
b
i
c
a
Li
n
t
o
n
g
N
i
h
u
t
a
1
5
2
R
o
b
u
st
a
La
mp
u
n
g
1
0
4
A
r
a
b
i
c
a
F
l
o
r
e
s
M
a
n
g
g
a
r
a
i
1
5
2
A
r
a
b
i
c
a
Te
m
a
n
g
g
u
n
g
F
u
l
l
w
a
sh
1
0
4
A
r
a
b
i
c
a
M
a
masa
1
5
2
A
r
a
b
i
c
a
K
l
a
t
e
n
Y
e
l
l
o
w
B
o
u
r
b
o
n
1
0
4
A
r
a
b
i
c
a
G
u
n
u
n
g
T
i
l
u
1
5
2
R
o
b
u
st
a
I
j
e
n
C
a
r
b
o
n
i
c
M
a
sce
r
a
si
N
a
t
u
r
a
l
1
0
4
A
r
a
b
i
c
a
C
i
h
u
r
i
p
1
5
2
A
r
a
b
i
c
a
Lu
w
a
k
A
c
e
h
G
a
y
o
1
0
4
A
r
a
b
i
c
a
B
a
l
i
K
i
n
t
a
ma
n
i
1
5
2
A
r
a
b
i
c
a
G
u
n
u
n
g
H
a
l
u
1
0
4
A
r
a
b
i
c
a
C
i
w
i
d
e
y
N
a
t
u
r
a
l
1
4
8
R
o
b
u
st
a
M
a
g
e
l
a
n
g
S
e
mar
M
e
sem
1
0
4
A
r
a
b
i
c
a
C
i
w
i
d
e
y
H
o
n
e
y
1
4
8
R
o
b
u
st
a
B
a
l
i
K
i
n
t
a
ma
n
i
1
0
4
R
o
b
u
st
a
G
u
n
u
n
g
U
n
g
a
r
a
n
1
4
8
A
r
a
b
i
c
a
B
a
l
i
P
r
e
mi
u
m
1
0
3
A
r
a
b
i
c
a
G
n
.
H
a
l
i
m
u
n
S
u
k
a
b
u
mi
1
4
8
R
o
b
u
st
a
S
u
mb
a
w
a
1
0
2
R
o
b
u
st
a
W
o
n
o
g
i
r
i
1
4
8
R
o
b
u
st
a
B
a
l
i
P
u
p
u
a
n
1
0
2
A
r
a
b
i
c
a
J
a
v
a
C
a
r
l
o
s
W
a
s
h
1
4
8
A
r
a
b
i
c
a
F
l
o
r
e
s J
u
r
i
a
M
a
n
g
g
a
r
a
i
1
0
2
R
o
b
u
st
a
G
u
n
u
n
g
K
e
l
i
r
1
4
6
R
o
b
u
st
a
T
o
r
a
j
a
K
a
l
o
si
1
0
1
Li
b
e
r
i
c
a
S
u
me
d
a
n
g
1
4
4
R
o
b
u
st
a
F
i
n
e
G
a
r
u
t
1
0
1
A
r
a
b
i
c
a
B
u
r
a
n
g
r
a
n
g
1
4
0
A
r
a
b
i
c
a
Te
m
a
n
g
g
u
n
g
A
n
a
e
r
o
b
1
0
1
A
r
a
b
i
c
a
I
j
e
n
H
o
n
e
y
1
4
0
R
o
b
u
st
a
S
i
d
i
k
a
l
a
n
g
S
u
ma
t
r
a
1
0
0
A
r
a
b
i
c
a
A
r
g
o
p
u
r
o
D
a
m
a
r
k
a
n
d
a
n
g
C
a
r
b
o
n
i
c
M
a
s
c
e
r
a
s
i
1
4
0
R
o
b
u
st
a
W
i
n
e
Te
ma
n
g
g
u
n
g
1
0
0
Li
b
e
r
i
c
a
G
u
n
u
n
g
M
u
r
i
a
1
3
8
A
r
a
b
i
c
a
D
a
mp
i
t
1
0
0
A
r
a
b
i
c
a
Ti
m
o
r
L
e
st
e
Er
m
e
r
a
1
3
6
R
o
b
u
st
a
P
a
p
u
a
A
m
b
a
i
d
i
r
u
1
0
0
R
o
b
u
st
a
A
c
e
h
La
m
n
o
1
3
6
R
o
b
u
st
a
S
o
l
o
k
P
a
d
a
n
g
1
0
0
A
r
a
b
i
c
a
S
e
mer
u
1
3
6
R
o
b
u
st
a
K
u
d
u
s
1
0
0
A
r
a
b
i
c
a
S
a
mo
si
r
1
3
6
A
r
a
b
i
c
a
G
u
n
u
n
g
H
a
l
u
F
u
l
l
w
a
s
h
1
0
0
A
r
a
b
i
c
a
G
u
n
u
n
g
S
i
n
d
o
r
o
1
3
2
A
r
a
b
i
c
a
J
a
v
a
I
j
e
n
N
a
t
u
r
a
l
1
0
0
A
r
a
b
i
c
a
En
e
r
e
k
a
n
g
1
3
2
R
o
b
u
st
a
La
n
a
n
g
D
a
mp
i
t
1
0
0
A
r
a
b
i
c
a
W
o
n
o
g
i
r
i
1
3
2
R
o
b
u
st
a
Ja
v
a
C
a
r
l
o
s
M
a
l
a
n
g
1
0
0
A
r
a
b
i
c
a
To
r
a
j
a
S
a
p
a
n
1
3
2
A
r
a
b
i
c
a
M
i
n
a
n
g
S
o
l
o
k
1
0
0
A
r
a
b
i
c
a
C
i
k
u
r
a
y
1
3
2
R
o
b
u
st
a
G
a
y
o
A
c
e
h
1
0
0
A
r
a
b
i
c
a
A
c
e
h
G
a
y
o
Ja
g
o
n
g
J
e
g
e
t
A
n
a
e
r
o
b
N
a
t
u
r
a
l
1
3
2
A
r
a
b
i
c
a
P
a
p
u
a
Y
a
h
o
k
i
mo
1
0
0
A
r
a
b
i
c
a
G
u
n
u
n
g
A
r
j
u
n
a
A
n
a
e
r
o
b
1
2
0
H
N
a
t
u
r
a
l
1
3
0
R
o
b
u
st
a
F
i
n
e
P
a
g
a
r
a
l
a
m
1
0
0
A
r
a
b
i
c
a
P
a
p
u
a
W
a
m
e
n
a
1
2
8
A
r
a
b
i
c
a
P
u
n
t
a
n
g
N
a
t
u
r
a
l
1
0
0
R
o
b
u
st
a
C
i
w
i
d
e
y
1
2
8
A
r
a
b
i
c
a
S
u
m
b
a
w
a
1
0
0
R
o
b
u
st
a
G
u
n
u
n
g
A
r
j
u
n
a
1
2
6
R
o
b
u
st
a
W
o
n
o
sa
l
a
m
96
A
r
a
b
i
c
a
G
a
y
o
F
u
l
l
w
a
s
h
1
2
4
A
r
a
b
i
c
a
M
a
n
g
l
a
y
a
n
g
80
A
r
a
b
i
c
a
M
a
n
d
h
e
l
i
n
g
1
2
4
Li
b
e
r
i
c
a
J
a
m
b
i
1
2
0
Ex
c
e
l
s
a
W
o
n
o
s
a
l
a
m
1
2
4
Ex
c
e
l
s
a
W
o
n
o
g
i
r
i
1
2
0
A
r
a
b
i
c
a
A
c
e
h
G
a
y
o
A
v
a
t
a
r
a
Th
e
r
mal
N
a
t
u
r
a
l
1
2
4
A
r
a
b
i
c
a
G
u
n
u
n
g
P
r
a
u
1
2
0
A
r
a
b
i
c
a
J
a
v
a
C
a
r
l
o
s N
a
t
u
r
a
l
1
2
4
A
r
a
b
i
c
a
B
a
l
i
B
u
l
e
l
e
n
g
A
n
a
e
r
o
b
1
2
0
R
o
b
u
st
a
K
e
r
i
n
c
i
1
2
4
R
o
b
u
st
a
P
a
p
u
a
M
i
mi
k
a
1
1
8
A
r
a
b
i
c
a
S
i
d
i
k
a
l
a
n
g
1
2
4
A
r
a
b
i
c
a
A
c
e
h
G
a
y
o
M
o
n
d
e
l
1
1
6
A
r
a
b
i
c
a
S
e
me
n
d
o
1
2
4
A
r
a
b
i
c
a
G
a
r
u
t
D
o
n
O
b
a
r
La
v
e
n
d
e
r
1
1
6
A
r
a
b
i
c
a
F
l
o
r
e
s
B
a
j
a
w
a
F
u
l
l
W
a
sh
1
2
0
R
o
b
u
st
a
B
e
n
g
k
u
l
u
1
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
4
,
No
.
6
,
Dec
em
b
er
2
0
2
5
:
5
1
4
0
-
5
1
5
6
5146
4
.
2
.
Da
t
a
prepro
ce
s
s
ing
4
.
2
.
1
.
S
h
uff
lin
g
a
nd
s
pli
t
t
ing
da
t
a
Sh
u
f
f
lin
g
is
a
s
tep
u
s
ed
to
r
an
d
o
m
ly
r
ea
r
r
a
n
g
e
th
e
d
ataset
b
ased
o
n
a
r
a
n
d
o
m
p
ar
am
eter
.
T
h
is
p
r
o
ce
s
s
is
d
o
n
e
s
im
u
ltan
eo
u
s
l
y
with
th
e
d
ata
s
p
litt
in
g
s
tep
.
B
y
d
o
in
g
s
o
,
th
e
d
ata
is
s
h
u
f
f
led
an
d
th
en
d
iv
id
e
d
in
to
tr
ain
in
g
,
v
alid
atio
n
,
an
d
t
est
s
ets.
T
h
e
p
u
r
p
o
s
e
o
f
th
is
d
iv
is
io
n
is
to
en
s
u
r
e
th
at
th
e
te
s
t
d
ata
h
as
n
o
t
b
ee
n
u
s
ed
in
th
e
t
r
ain
in
g
s
et,
an
d
v
ice
v
er
s
a.
Sh
u
f
f
lin
g
a
n
d
s
p
litt
i
n
g
a
r
e
e
x
p
ec
ted
to
h
elp
th
e
m
o
d
el
lear
n
th
e
d
ata
with
o
u
t
r
ely
i
n
g
o
n
a
s
p
ec
if
ic
o
r
d
er
,
b
u
t
in
s
tead
o
n
th
e
co
m
p
lex
p
atter
n
s
with
in
t
h
e
d
ata.
T
h
e
d
ata
is
s
p
lit
in
to
8
0
%
tr
ain
in
g
,
1
0
%
v
alid
atio
n
,
an
d
1
0
%
test
s
ets,
r
esu
lti
n
g
in
1
1
,
6
2
0
tr
ain
in
g
im
ag
e
s
,
1
,
4
5
2
v
alid
atio
n
im
ag
es,
an
d
1
,
4
5
3
test
im
ag
es.
T
h
e
lar
g
e
tr
ain
i
n
g
s
et
is
in
ten
d
ed
to
h
elp
th
e
m
o
d
el
b
etter
u
n
d
er
s
tan
d
th
e
v
ar
iab
les in
th
e
d
ata,
p
r
ev
en
ti
n
g
b
o
t
h
o
v
e
r
f
itti
n
g
a
n
d
u
n
d
er
f
itti
n
g
.
4
.
2
.
2
.
Resize
a
nd
re
s
ca
le
T
h
e
r
esizin
g
s
tep
in
v
o
lv
es setti
n
g
th
e
im
ag
e
d
im
en
s
io
n
s
to
1
5
0
×
1
5
0
p
ix
els.
T
h
is
en
s
u
r
es th
at
all
d
ata
h
as
a
u
n
if
o
r
m
s
ize.
Nex
t,
n
o
r
m
aliza
tio
n
is
ap
p
lied
to
co
n
v
er
t
p
ix
el
v
alu
es
i
n
to
a
m
atr
i
x
o
f
v
alu
es
r
an
g
in
g
b
etwe
en
ze
r
o
a
n
d
o
n
e,
u
s
in
g
t
h
e
p
ar
am
eter
1
/2
5
5
.
4
.
2
.
3
.
Da
t
a
a
ug
m
ent
a
t
io
n
D
a
t
a
a
u
g
m
e
n
t
a
t
i
o
n
is
a
p
p
l
i
e
d
t
o
i
n
c
r
e
a
s
e
t
h
e
v
a
r
i
e
t
y
o
f
d
a
t
a
s
et
s
.
T
o
e
n
h
a
n
c
e
t
h
e
r
o
b
u
s
t
n
es
s
a
n
d
g
e
n
e
r
a
l
i
z
a
ti
o
n
c
a
p
a
b
il
i
t
y
o
f
t
h
e
m
o
d
e
l
,
s
e
v
e
r
a
l
d
a
t
a
a
u
g
m
e
n
t
a
ti
o
n
t
e
c
h
n
i
q
u
es
as
s
h
o
w
n
i
n
Fi
g
u
r
e
5
w
e
r
e
a
p
p
l
i
ed
t
o
t
h
e
t
r
ai
n
i
n
g
d
a
t
as
e
t
.
T
h
e
s
e
in
c
l
u
d
e
h
e
i
g
h
t
s
h
i
f
t
i
n
F
i
g
u
r
e
5
(
a
)
,
s
h
e
a
r
i
n
F
i
g
u
r
e
5
(
b
)
,
z
o
o
m
i
n
F
i
g
u
r
e
5
(
c
)
,
an
d
h
o
r
izo
n
tal
f
lip
tr
a
n
s
f
o
r
m
atio
n
s
in
Fig
u
r
e
5
(
d
)
.
I
n
ea
ch
s
u
b
f
ig
u
r
e,
th
e
im
a
g
e
o
n
th
e
lef
t
r
e
p
r
esen
ts
th
e
o
r
ig
in
al
co
f
f
ee
b
ea
n
,
wh
ile
th
e
im
ag
e
o
n
th
e
r
i
g
h
t
s
h
o
ws
th
e
c
o
r
r
es
p
o
n
d
in
g
au
g
m
en
ted
v
er
s
io
n
,
d
em
o
n
s
tr
atin
g
h
o
w
th
ese
tech
n
iq
u
es
in
cr
ea
s
e
th
e
d
iv
er
s
ity
o
f
th
e
d
ataset.
T
h
ese
s
tep
s
g
en
er
ate
n
ew
im
ag
es
to
en
r
ich
th
e
tr
ain
in
g
d
ataset,
en
h
an
cin
g
th
e
m
o
d
el’
s
ab
ilit
y
to
g
en
er
alize
.
(
a)
(
b
)
(
c)
(
d
)
Fig
u
r
e
5.
Data
au
g
m
en
tatio
n
tech
n
iq
u
es
of
(
a)
h
eig
h
t sh
if
t,
(
b
)
s
h
ea
r
,
(
c
)
zo
o
m
,
an
d
(
d
)
h
o
r
izo
n
tal
f
lip
4
.
3
.
H
y
perpa
ra
m
e
t
er
s
et
t
ing
s
Sev
er
al
h
y
p
e
r
p
ar
a
m
eter
s
ap
p
li
ed
to
th
e
m
o
d
el
h
a
v
e
b
ee
n
p
r
e
d
eter
m
in
ed
,
as
s
h
o
wn
i
n
T
ab
l
e
3
.
T
h
ese
s
ettin
g
s
wer
e
s
tan
d
ar
d
ized
ac
r
o
s
s
all
m
o
d
els
b
ein
g
tr
ain
ed
to
en
s
u
r
e
f
air
tr
ea
tm
en
t
an
d
f
ac
i
litate
co
m
p
ar
is
o
n
s
.
T
h
e
I
n
ce
p
tio
n
V3
tr
an
s
f
er
le
ar
n
in
g
m
o
d
el
was
tr
ain
ed
f
o
r
5
0
ep
o
c
h
s
,
wh
ile
th
e
n
u
m
b
e
r
o
f
ep
o
c
h
s
f
o
r
th
e
b
u
ild
-
f
r
o
m
-
s
cr
atch
m
o
d
el
was
d
eter
m
in
e
d
t
h
r
o
u
g
h
T
ag
u
ch
i e
x
p
er
im
en
ts
.
T
h
e
m
o
d
el
d
e
v
el
o
p
m
en
t w
as c
ar
r
ied
o
u
t
u
s
in
g
T
en
s
o
r
Flo
w,
with
th
e
co
d
e
wr
itten
in
Py
th
o
n
.
T
h
e
tr
ain
in
g
p
r
o
ce
s
s
was
co
n
d
u
cted
u
s
in
g
J
u
p
y
ter
No
teb
o
o
k
o
n
a
co
m
p
u
ter
eq
u
ip
p
ed
with
an
I
n
tel
C
o
r
e
i9
-
1
3
9
0
0
KF
3
GHz
C
P
U,
3
2
G
B
o
f
R
A
M,
an
d
an
NVI
DI
A
GeFo
r
ce
R
T
X
4
0
8
0
GPU.
4
.
4
.
Arc
hite
ct
ure
o
f
t
ra
ns
f
er
lea
rning
wit
h
I
ncept
io
n V3
I
n
s
ce
n
ar
io
1
,
th
e
f
ea
tu
r
e
ex
tr
ac
tio
n
lay
er
s
ar
e
f
r
o
ze
n
,
u
tili
zin
g
th
e
p
r
etr
ain
ed
weig
h
ts
o
f
I
n
ce
p
tio
n
V3
.
T
h
is
ap
p
r
o
ac
h
f
o
cu
s
es
o
n
tr
ain
in
g
o
n
l
y
th
e
class
if
icatio
n
p
ar
t
o
f
th
e
m
o
d
el.
T
h
e
n
u
m
b
e
r
o
f
ep
o
ch
s
u
s
ed
in
th
is
p
h
ase
is
2
5
.
Af
ter
th
at,
f
in
e
-
tu
n
in
g
i
s
p
er
f
o
r
m
e
d
b
y
u
n
f
r
ee
zin
g
h
alf
o
f
th
e
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
C
la
s
s
i
fica
tio
n
o
f sin
g
le
o
r
ig
in
I
n
d
o
n
esia
n
co
ffee
b
e
a
n
s
u
s
in
g
co
n
vo
lu
tio
n
a
l …
(
A
c
h
ma
d
P
r
a
ta
ma
R
ifa
i)
5147
ex
tr
ac
tio
n
lay
er
s
,
s
p
ec
if
ically
1
5
5
o
u
t
o
f
th
e
to
tal
3
1
1
lay
e
r
s
.
T
h
is
f
in
e
-
tu
n
in
g
p
h
ase
also
r
u
n
s
f
o
r
2
5
e
p
o
ch
s
,
b
r
in
g
in
g
th
e
to
tal
tr
ain
in
g
ep
o
ch
s
f
o
r
th
is
s
ce
n
ar
io
to
5
0
.
I
n
s
ce
n
ar
io
2
,
all
t
h
e
f
ea
tu
r
e
e
x
tr
ac
tio
n
lay
e
r
s
ar
e
u
n
f
r
o
z
en
.
T
h
is
ap
p
r
o
ac
h
allo
ws
th
e
m
o
d
el
to
lear
n
th
e
d
ataset
f
r
o
m
th
e
v
e
r
y
f
i
r
s
t
lay
er
,
ad
d
r
ess
in
g
an
y
s
ig
n
if
ican
t
d
if
f
er
en
ce
s
b
etwe
e
n
th
e
d
ataset
u
s
ed
in
th
is
s
tu
d
y
an
d
th
e
o
n
e
u
s
ed
f
o
r
p
r
etr
ai
n
in
g
th
e
m
o
d
el.
T
h
e
d
if
f
er
en
ce
in
t
h
e
n
u
m
b
er
o
f
p
a
r
am
eter
s
ca
n
b
e
s
ee
n
in
T
a
b
le
4
,
an
d
th
e
m
o
d
el
ar
ch
itectu
r
e
o
f
tr
an
s
f
e
r
lear
n
i
n
g
is
illu
s
tr
ated
in
Fig
u
r
es
6
—
Fig
u
r
e
6
(
a)
f
o
r
s
ce
n
ar
io
1
with
all
lay
er
s
f
r
o
ze
n
,
Fig
u
r
e
6
(
b
)
f
o
r
s
ce
n
ar
i
o
1
with
h
alf
o
f
t
h
e
lay
er
s
u
n
f
r
o
ze
n
,
an
d
Fig
u
r
e
6
(
c)
f
o
r
s
ce
n
ar
io
2
.
T
ab
le
3
.
Hy
p
er
p
ar
a
m
eter
s
ettin
g
s
H
y
p
e
r
p
a
r
a
me
t
e
r
V
a
l
u
e
Lo
ss f
u
n
c
t
i
o
n
B
i
n
a
r
y
c
r
o
sse
n
t
r
o
p
y
B
a
t
c
h
si
z
e
2
5
6
O
p
t
i
mi
z
e
r
A
d
a
m
I
n
p
u
t
si
z
e
1
9
2
×
75
Le
a
r
n
i
n
g
r
a
t
e
0
.
0
1
Ev
a
l
u
a
t
i
o
n
me
t
r
i
c
A
c
c
u
r
a
c
y
,
p
r
e
c
i
s
i
o
n
,
r
e
c
a
l
l
T
ab
le
4
.
Nu
m
b
er
o
f
tr
an
s
f
er
lear
n
in
g
m
o
d
els p
ar
am
ete
r
s
S
c
e
n
a
r
i
o
S
t
a
g
e
s
P
a
r
a
me
t
e
r
t
y
p
e
Tr
a
i
n
a
b
l
e
N
o
n
-
t
r
a
i
n
a
b
l
e
#1
F
r
e
e
z
e
a
l
l
l
a
y
e
r
2
,
1
0
7
,
4
0
1
2
1
,
8
0
2
,
7
8
4
U
n
f
r
e
e
z
e
h
a
l
f
l
a
y
e
r
1
8
,
8
9
6
,
8
4
1
5
,
0
1
3
,
3
4
4
#2
U
n
f
r
e
e
z
e
a
l
l
l
a
y
e
r
2
3
,
8
6
7
,
5
5
3
3
4
,
4
3
2
(
a)
(
b
)
(
c)
Fig
u
r
e
6
.
Ar
c
h
itectu
r
e
o
f
tr
an
s
f
er
lear
n
in
g
m
o
d
el
of
(
a)
s
ce
n
ar
io
1
:
f
r
ee
ze
all
lay
er
,
(
b
)
s
ce
n
ar
io
1
: u
n
f
r
ee
ze
h
alf
la
y
er
,
an
d
(
c)
s
ce
n
a
r
io
2
:
u
n
f
r
ee
ze
h
alf
lay
er
4
.
5
.
Arc
hite
ct
ure
o
f
t
ra
ini
ng
f
ro
m
s
cr
a
t
ch
m
o
del
T
h
e
ar
c
h
itectu
r
e
d
e
v
elo
p
m
en
t
f
o
r
th
e
b
u
ild
-
f
r
o
m
-
s
cr
atch
m
o
d
el
was
ca
r
r
ied
o
u
t
b
as
ed
o
n
th
e
an
aly
s
is
u
s
in
g
th
e
T
a
g
u
ch
i
m
eth
o
d
,
r
esu
ltin
g
in
a
r
o
b
u
s
t
c
o
m
b
in
atio
n
o
f
th
e
n
u
m
b
er
o
f
f
ea
tu
r
e
ex
tr
ac
tio
n
lay
er
s
an
d
th
e
tr
ain
in
g
ep
o
ch
s
.
T
h
e
f
ea
tu
r
e
e
x
tr
ac
tio
n
lay
e
r
s
an
aly
ze
d
co
n
s
is
ted
o
f
5
,
6
,
a
n
d
7
co
n
v
o
l
u
tio
n
al
lay
er
s
.
T
h
e
f
ir
s
t
th
r
ee
la
y
er
s
a
r
e
id
en
tical,
f
o
llo
we
d
b
y
lay
er
s
4
,
5
,
6
,
an
d
7
,
ea
ch
co
n
tain
i
n
g
6
4
f
ilter
s
with
a
s
ize
o
f
3
×
3
,
ap
p
ly
i
n
g
th
e
R
eL
U
ac
tiv
atio
n
f
u
n
ctio
n
.
Pad
d
in
g
is
ap
p
lied
to
all
lay
er
s
ex
ce
p
t
th
e
th
ir
d
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
4
,
No
.
6
,
Dec
em
b
er
2
0
2
5
:
5
1
4
0
-
5
1
5
6
5148
co
n
v
o
l
u
tio
n
al
lay
er
.
E
ac
h
co
n
v
o
lu
tio
n
al
lay
e
r
is
f
o
llo
wed
b
y
a
2
×
2
p
o
o
lin
g
lay
er
.
Me
a
n
wh
ile,
th
e
n
u
m
b
er
o
f
ep
o
ch
s
test
ed
in
clu
d
es
5
0
,
7
5
,
an
d
1
0
0
ep
o
ch
s
.
Valid
atio
n
ac
cu
r
ac
y
is
u
s
ed
as
th
e
e
v
alu
atio
n
o
u
tp
u
t,
b
ased
o
n
th
e
S/N
r
atio
,
with
th
e
"lar
g
er
is
b
etter
"
ap
p
r
o
ac
h
.
Deta
ils
o
f
th
e
f
ac
to
r
co
m
b
in
atio
n
s
ar
e
s
h
o
wn
in
T
ab
le
5
,
an
d
th
e
a
n
aly
s
is
r
esu
lts
ar
e
p
r
esen
ted
in
Fig
u
r
e
7
.
T
h
e
T
ag
u
ch
i
an
aly
s
is
r
ev
ea
le
d
th
at
th
e
m
o
s
t
r
o
b
u
s
t
co
m
b
in
atio
n
f
o
r
t
h
e
m
o
d
el
c
o
n
s
is
ts
o
f
7
f
ea
tu
r
e
ex
tr
ac
tio
n
lay
er
s
an
d
1
0
0
tr
ain
in
g
ep
o
ch
s
.
T
h
is
co
m
b
in
atio
n
y
ield
s
th
e
h
ig
h
est
ac
cu
r
ac
y
co
m
p
ar
ed
t
o
o
th
er
test
ed
co
n
f
ig
u
r
atio
n
s
.
B
ased
o
n
th
is
r
esu
lt,
th
e
d
ev
elo
p
e
d
m
o
d
el
will
h
av
e
an
ar
ch
ite
ctu
r
e
co
n
s
is
tin
g
o
f
7
co
n
v
o
lu
tio
n
al
lay
e
r
s
.
T
h
e
f
ir
s
t
lay
er
will
h
av
e
1
6
f
ilter
s
,
th
e
s
ec
o
n
d
lay
er
will
h
a
v
e
3
2
f
ilter
s
,
an
d
th
e
r
em
ain
in
g
lay
er
s
will
h
av
e
6
4
f
ilter
s
,
all
with
a
f
ilter
s
ize
o
f
3
×
3
.
E
ac
h
c
o
n
v
o
lu
tio
n
al
lay
er
is
f
o
llo
wed
b
y
a
m
ax
-
p
o
o
lin
g
lay
er
with
a
p
o
o
l
s
ize
o
f
2
×
2
.
T
h
is
s
tr
u
ctu
r
e
is
th
en
co
n
n
ec
ted
to
a
g
lo
b
al
av
e
r
ag
e
p
o
o
lin
g
lay
er
,
wh
ich
co
n
v
er
ts
th
e
f
ea
t
u
r
e
m
atr
ix
in
to
a
v
ec
to
r
th
at
is
p
ass
ed
to
a
f
u
lly
co
n
n
ec
ted
l
ay
er
.
T
h
e
f
ir
s
t
f
u
lly
co
n
n
ec
ted
lay
e
r
h
as 5
1
2
n
eu
r
o
n
s
,
wh
ile
th
e
s
ec
o
n
d
f
u
lly
co
n
n
ec
ted
lay
er
h
as 1
1
6
n
eu
r
o
n
s
.
E
ac
h
lay
er
u
s
es th
e
R
eL
U
ac
tiv
atio
n
f
u
n
ctio
n
,
ex
c
ep
t
f
o
r
th
e
f
i
n
al
lay
er
,
wh
ich
u
tili
ze
s
So
f
tm
ax
f
o
r
m
u
lticlas
s
class
if
icatio
n
.
T
h
e
m
o
d
el
co
n
tain
s
a
to
tal
o
f
2
6
4
,
0
8
4
tr
ain
ab
le
p
ar
am
eter
s
a
n
d
will
b
e
tr
ain
e
d
f
o
r
1
0
0
e
p
o
ch
s
.
T
h
e
m
o
d
el
ar
ch
itectu
r
e
is
illu
s
tr
ated
in
Fig
u
r
e
8
.
T
ab
le
5
.
Desig
n
o
f
ex
p
er
im
en
t
r
esu
lts
f
o
r
tr
ain
in
g
f
r
o
m
s
cr
at
ch
m
o
d
el
N
u
mb
e
r
o
f
f
e
a
t
u
r
e
e
x
t
r
a
c
t
i
o
n
l
a
y
e
r
N
u
mb
e
r
o
f
e
p
o
c
h
V
a
l
i
d
a
t
i
o
n
a
c
c
u
r
a
c
y
(
%)
5
50
4
9
.
4
5
5
75
5
5
.
4
4
5
1
0
0
5
5
.
9
9
6
50
5
1
.
1
7
6
75
5
6
.
4
7
6
1
0
0
6
3
.
5
0
7
50
5
4
.
8
9
7
75
6
2
.
4
0
7
1
0
0
6
7
.
9
8
Fig
u
r
e
7
.
T
a
g
u
ch
i
a
n
aly
s
is
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
C
la
s
s
i
fica
tio
n
o
f sin
g
le
o
r
ig
in
I
n
d
o
n
esia
n
co
ffee
b
e
a
n
s
u
s
in
g
co
n
vo
lu
tio
n
a
l …
(
A
c
h
ma
d
P
r
a
ta
ma
R
ifa
i)
5149
Fig
u
r
e
8
.
Ar
c
h
itectu
r
e
o
f
tr
ain
in
g
f
r
o
m
s
cr
atch
m
o
d
el
4
.
6
.
T
ra
ini
ng
r
esu
lt
s
T
h
e
tr
ain
i
n
g
p
h
ase
was
ca
r
r
ie
d
o
u
t
f
o
r
5
0
ep
o
ch
s
o
n
t
h
e
I
n
ce
p
tio
n
V3
t
r
an
s
f
er
lear
n
in
g
m
o
d
el
in
b
o
th
s
ce
n
ar
i
o
s
.
Me
an
wh
ile,
th
e
b
u
ild
-
f
r
o
m
-
s
cr
atch
m
o
d
el
w
as
tr
ain
ed
f
o
r
1
0
0
e
p
o
ch
s
,
as
d
eter
m
in
ed
b
y
t
h
e
T
ag
u
ch
i
m
eth
o
d
an
aly
s
is
.
T
h
e
tr
ain
in
g
p
r
o
ce
s
s
f
o
r
th
e
I
n
c
ep
tio
n
V3
tr
an
s
f
er
lear
n
in
g
m
o
d
el
in
s
ce
n
ar
io
1
,
s
ce
n
ar
io
2
,
a
n
d
th
e
b
u
ild
-
f
r
o
m
-
s
cr
atch
m
o
d
el
is
s
h
o
wn
i
n
Fig
u
r
es
9
to
11
,
r
esp
ec
tiv
ely
.
Fig
u
r
e
9
.
T
r
ain
in
g
p
r
o
g
r
ess
o
f
I
n
c
ep
tio
n
V
3
m
o
d
el
u
s
in
g
s
ce
n
ar
io
1
Evaluation Warning : The document was created with Spire.PDF for Python.