In
d
o
n
e
sian
Jou
r
n
al of
Ele
c
tr
i
c
a
l
En
g
in
e
erin
g
a
n
d
C
om
pu
ter S
c
ien
ce
Vol.
14, No.
1, April 2019,
pp.
327~332
ISSN: 2502-
4752,
DOI
:
10.115
91/ijeecs.
v
14.
i
1
.
pp327-332
3
27
Jou
rn
a
l
h
o
me
pa
ge
:
ht
tp:
//i
a
e
score
.
com
/
j
o
u
r
na
l
s
/
i
n
d
e
x
.
p
hp/
i
j
eec
s
Eva
l
ua
tion of basic conv
ol
utio
nal neural network and bag of
features for leaf recogn
ition
Nu
r
u
l
F
a
t
i
ha
h
Sa
h
i
da
n
,
A
h
m
a
d
K
ha
i
r
i
J
u
ha
,
Za
i
d
a
h
I
br
a
h
i
m
Facu
lt
y
of
C
o
m
put
er
a
nd
M
athe
m
a
tical
S
c
i
en
c
e
s
,
U
n
i
versi
t
i
Tek
n
o
log
i
M
ARA, S
h
a
h Al
am
,
Selan
g
o
r, Ma
l
aysia
Art
i
cl
e In
fo
ABSTRACT
A
r
tic
le hist
o
r
y
:
R
e
ce
i
v
e
d
Jul 9,
2018
Re
vise
d S
e
p 23,
201
8
Ac
ce
p
t
ed
No
v
1
8
, 2
018
Th
is
p
aper
p
resen
t
s
t
h
e
ev
alu
a
tion
of
b
as
ic
C
onvol
utio
nal
Neu
r
a
l
Net
w
ork
(CNN
)
and
Bag
of
F
e
a
t
u
res
(BoF
)
f
o
r
L
eaf
R
e
c
og
ni
ti
on.
I
n
th
is
s
tudy,
the
p
e
rform
a
nc
e
o
f
b
a
s
ic
C
NN
a
nd
B
o
F
f
or
l
e
a
f
r
e
c
og
nition
u
sing
a
p
ub
lic
l
y
avai
lab
l
e
datas
e
t
call
e
d
Fo
li
o
da
taset
has
been
i
nvestiga
t
ed.
C
NN
h
a
s
p
r
o
ve
n
its
p
o
w
erf
u
l
f
eature
repr
esen
tat
i
on
pow
er
i
n
com
p
u
t
er
v
i
s
i
on.
T
h
e
s
am
e
go
es
with
B
oF
w
here
i
t
h
a
s
s
e
t
n
e
w
p
e
rf
o
r
m
a
nce
s
t
and
a
rds
on
popu
lar
imag
e
classificati
o
n
benchmarks
a
nd
h
as
achieved
scal
abilit
y
brea
kthr
oug
h
in
im
ag
e
ret
r
iev
a
l.
T
he
f
eatu
r
e
th
at
i
s
be
i
n
g
utili
zed
i
n
th
e
BoF
i
s
S
peed
ed-Up
Rob
u
s
t
F
eatu
r
e
(S
URF
)
t
extu
re
f
eatu
r
e.
T
h
e
e
xp
erim
ent
a
l
resul
t
s
i
n
d
icat
e
th
at BoF
achi
e
ves
better accuracy
com
p
a
red
to b
asic
C
NN
.
K
eyw
ord
s
:
Ba
g of
f
e
a
t
ures
CN
N
Deep
l
earn
in
g
Le
af
re
c
o
gn
iti
on
Co
pyri
gh
t © 2
019 In
stit
u
t
e
of Advanced
En
gi
neeri
n
g
an
d
Scien
ce.
All
rights
res
e
rv
ed.
Corres
pon
d
i
n
g
Au
th
or:
Za
id
a
h
Ib
r
ah
im,
F
a
cult
y
o
f
C
o
m
put
e
r
a
n
d
Ma
t
hem
a
t
i
ca
l
S
c
i
e
nce
s
,
Un
iv
ersit
i
T
ekno
l
o
g
i
M
ARA,
Sha
h
Alam
,
S
el
a
n
g
o
r,
M
alays
i
a.
Em
ail:
zaida
h
@
tm
sk.u
i
t
m.
edu.m
y
1.
I
N
TR
OD
U
C
TI
O
N
S
e
ve
ral
pa
rts
of
a
p
lan
t
c
a
n
b
e
u
s
ed
b
y
a
bo
ta
ni
st
i
n
or
de
r
t
o
r
ecog
n
i
ze
a
p
la
n
t
.
Th
is
i
nc
l
udes
fl
ow
er
s,
l
e
a
ve
s,
a
nd
roo
t
s.
H
ow
ever
,
leave
s
a
re
t
he
m
ost
w
i
de
l
y
u
se
d
as
i
t
is
m
ore
c
onve
nie
n
t
t
o
b
e
u
s
e
d
a
n
d
the
re
su
lt
s
are
grea
t
[1].
T
he
pur
pose
o
f
i
de
nt
i
f
y
i
ng
p
l
a
n
ts
i
s
t
o
c
a
t
eg
oriz
e
the
pla
n
ts
f
or
r
ec
ordi
n
g
pur
pose
s
.
Th
e
p
r
o
c
e
s
s
of
i
d
e
nt
i
f
yi
ng
a
p
l
a
nt
u
si
ng
l
e
a
v
e
s
i
s
a
n
ea
sy
t
a
s
k
f
o
r
b
ot
an
ist
s
a
s
t
h
ey
can
s
i
m
ply
rec
ogni
z
e
i
t
us
i
n
g
t
h
e
i
r
sense
s
[
2].
O
n
t
he
c
ontra
ry,
for
ma
chines
t
o
ach
iev
e
t
h
e
sa
me
r
ec
ogn
iti
o
n
r
esul
ts
r
equ
i
res
perform
ing
i
m
age
-
proce
ssi
ng
t
e
c
h
n
i
qu
e
s
t
o
extra
c
t
v
i
s
ua
l
inf
o
rm
ation
a
n
d
c
o
m
p
ar
e
t
h
e
m
t
o
e
x
i
s
ti
n
g
s
ets
of
data
[
3].
S
t
r
u
ct
ured lear
n
i
n
g
o
r be
t
t
e
r
k
n
o
w
n
a
s de
ep le
a
rn
in
g,
has be
e
n
r
ec
o
g
n
i
ze
d a
s
a n
ew are
a i
n
com
pu
t
e
r
vi
si
o
n
tha
t
has be
en r
ep
orted
to
p
ro
duce
exc
e
lle
n
t
re
s
ults [
4].
D
e
ep
l
ear
nin
g
,
a
cla
ss
of
m
a
c
hi
ne-lea
r
n
i
n
g
tec
h
n
i
que
s
us
ed
t
o
e
xt
ra
ct
c
h
a
ra
ct
e
r
i
s
ti
cs
o
f
d
a
t
a
,
a
n
d
C
N
N
(Ne
t
wo
rk
N
e
u
ra
l
Con
v
o
l
ut
ion
a
l
)
,
a
se
ri
es
o
f
arti
fi
c
i
al
n
eu
r
a
l
n
e
t
wo
rk
s
t
h
at
h
av
e
b
e
en
e
xp
an
d
e
d
in
t
o
sp
a
c
e
u
si
ng
s
ha
re
d
we
i
ght
s
,
h
av
e
b
e
e
n
f
o
und
to
b
e
su
it
ab
l
e
f
o
r
co
mp
ut
e
r
v
i
s
i
o
n
t
a
sk
s
[5
].
W
i
t
hi
n
th
e
p
a
st
f
e
w
y
e
ars,
d
e
e
p
l
ea
rn
i
n
g
a
l
go
ri
thms
p
a
r
t
i
c
ul
a
r
ly
c
on
vol
u
t
i
o
n
a
l
n
e
u
r
al
n
etworks
(
C
NN
)
ha
ve
p
roven
t
h
e
i
r
m
u
ch
pow
er
ful fe
a
t
u
r
e repr
esenta
t
i
on p
o
w
e
r
in c
o
m
puter
v
isio
n
[6].
C
o
n
v
o
lu
ti
o
n
al
N
eura
l
N
e
tw
ork
has
resu
l
t
ed
i
n
gr
ou
nd
br
eaki
n
g
de
c
i
sions
o
v
e
r
t
h
e
l
a
st
d
ec
ad
e
i
n
vari
ous
f
i
e
l
d
s
rela
t
e
d
t
o
p
a
t
te
r
n
r
e
c
og
n
i
t
i
on
;
from
ima
g
e
pr
oce
s
s
i
ng
t
o
v
o
i
c
e
r
e
c
o
gni
ti
o
n
[7
].
C
NN’s
ca
pab
i
lit
ies
h
a
ve
b
ec
ome
a
kn
ow
n
a
nd
u
s
e
d
i
n
v
o
ice
anal
ys
is
[
8]
,
im
age
class
i
fica
t
i
o
n
[
9],
sc
ene
class
i
fica
t
i
o
n
[
10],
ve
h
i
c
l
e
rec
o
g
n
i
t
i
on
[1
1],
frui
t
c
l
a
ssi
fic
a
tio
n
an
d
ripe
ness
gr
a
d
i
n
g
r
ecog
n
i
t
i
on
[1
2
]
,
and
f
ood
re
c
o
g
n
iti
on
[
13
].
Ba
g
of
F
ea
ture
s
(Bo
F
)
m
o
d
e
l
i
s
a
lso
w
i
de
ly
i
mplem
e
n
t
e
d
i
n
ima
g
e
p
r
o
ce
ssin
g
a
n
d
c
la
ss
ifica
tio
n
tas
k
s
[14]
.
It
i
s
a
m
a
c
h
ine l
e
arni
n
g
m
ode
l
a
n
d
an
i
n
t
e
g
ra
t
i
o
n
o
f
Ba
g
of Words
(
B
o
W)
w
hi
ch
m
ak
e
s
i
t
su
it
ab
l
e
f
o
r
i
m
a
g
e
cl
a
s
si
fi
c
a
t
i
o
n
[1
5],
f
o
r
ex
a
m
pl
e,
i
n
b
r
e
a
st
h
istop
a
th
o
l
og
y
i
m
a
g
e
s
r
e
c
og
n
i
t
i
on
a
nd
A
ra
b
i
c
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
250
2-
4
7
5
2
In
do
n
e
sia
n
J
Elec Eng
&
C
o
mp
S
ci,
Vo
l. 1
4
,
No
. 1
,
Ap
r
i
l
2
0
1
9
:
327
–
3
3
2
32
8
ha
ndwr
itte
n
w
o
rd
r
e
c
o
g
n
i
tio
n
[1
6].
In
t
his
pape
r,
a
c
ompa
riso
n
b
e
t
w
een
C
NN
an
d
B
o
F
i
s
b
e
i
ng
c
o
nduc
t
e
d
in
or
de
r
to
a
na
ly
z
e
the
acc
ur
ac
y
pe
r
f
or
m
a
nc
e
o
f
l
ea
f
r
ecogn
i
t
i
on
b
etw
e
e
n
t
he
t
w
o
m
odels.
Thi
s
p
ap
er
p
r
e
se
nt
s
e
v
alu
a
tion
of
b
a
s
i
c
C
onv
olu
tio
n
a
l
Neura
l
N
e
tw
or
k
an
d
Ba
g
o
f
F
ea
tu
r
e
s
f
o
r
Leaf
Rec
o
g
n
i
t
i
o
n.
I
n
th
is
s
t
u
d
y
,
c
o
mpa
r
is
on
i
s
m
a
de
t
o
de
t
e
r
m
ine
w
h
i
c
h
m
ode
l
i
s
t
he
m
ost
a
p
pr
op
r
i
a
t
e
t
o
re
cog
n
ize
the
leaf
from
F
o
li
o
da
tase
t.
Resea
r
ch
a
bo
u
t
l
e
a
f
re
c
og
ni
tio
n
has
bee
n
c
o
n
d
u
c
t
e
d
by
se
ver
a
l
r
e
sear
cher
s
usi
n
g
va
r
i
o
u
s
tec
h
n
i
que
s.
O
ne
t
echn
i
que
u
se
d
is
S
u
p
po
rt
V
ecto
r
Mach
in
e
(S
VM)
with
t
ex
tu
re
fe
at
ur
e
s
a
n
d
t
h
e
r
esult
a
c
h
ie
v
e
d
is
9
9%
a
cc
u
r
ac
y
[
2
]
.
W
it
h
da
ta
a
u
g
m
e
n
t
at
i
on,
t
he
a
c
c
ur
a
c
y
o
f
9
9.
04
%
usi
n
g
A
l
exN
e
t
a
n
d
99.
42
u
si
n
g
G
o
o
g
l
e
N
et
a
r
e
o
bt
a
i
ne
d
[
1
6]
.
Bes
i
de
s
t
h
a
t
,
sh
a
p
e
f
eatu
r
e
s
a
n
d
c
o
l
ou
r
hi
st
og
r
a
m
w
i
t
h
k
-
n
ear
est
ne
ig
hb
o
u
r
c
l
a
s
sif
i
e
r
s
ha
ve
b
e
e
n
a
p
plie
d
w
i
t
h
8
7.
2%
acc
ur
ac
y
[
3
]
.
S
i
n
c
e
t
he
r
es
ul
ts
f
r
o
m
usi
n
g
th
is
d
a
t
ase
t
h
as
b
e
e
n
ve
r
y
p
osi
tive
,
t
h
i
s
da
t
a
s
e
t
ha
s
bee
n
c
ho
s
en
t
o
be
e
xper
i
m
e
nte
d
i
n
t
h
i
s
r
ese
a
r
c
h.
2.
RESEARCH
M
E
T
HO
D
2.
1.
B
ag
of
F
eat
u
re
s
(
B
oF)
B
a
g-
of-
F
ea
t
u
r
e
s
(
B
oF
)
r
e
pr
ese
n
ts
t
he
i
ma
ges
b
y
i
n
s
ta
n
c
e
s
o
f
l
o
c
a
l
f
e
a
t
u
r
e
s
e
x
t
r
a
c
t
e
d
f
r
o
m
t
h
e
im
age.
T
h
i
s
fra
m
e
w
ork
can
b
e
perc
ei
ve
d
i
n
t
o
a
t
wo-le
v
e
l
fra
me
wo
rk.
The
firs
t
l
e
ve
l
is
a
s
s
oc
iate
d
t
o
t
he
p
i
x
e
l
in
te
ns
i
t
y of the
i
m
a
ge
a
nd
e
x
t
r
ac
tion o
f
l
oc
a
l
f
eat
ur
e
s
t
y
p
e.
F
or
t
he
s
ec
o
n
d
leve
l
,
i
t
c
o
n
s
i
s
ts
o
f
tw
o par
t
w
hi
c
h
are
e
n
c
odi
ng
a
n
d
poo
li
ng
[
16].
Th
e
e
n
c
o
d
i
ng
p
a
r
t
con
v
e
r
t
e
d
l
o
c
a
l
fea
t
ur
es
i
nt
o
c
o
de b
oo
ks,
i
n
w
hic
h
t
h
e
m
ost
r
e
pr
ese
n
t
a
tive
vi
sua
l
v
oc
ab
u
l
ar
y
pa
t
t
er
ns
a
r
e
c
o
d
e
d
a
s
v
i
s
u
al
o
r
c
o
de
w
or
ds
u
s
i
n
g
c
o
de
b
o
o
k
l
e
a
r
n
i
ng.
T
he
n,
a
h
i
s
t
o
g
ra
m
o
r
f
e
a
t
u
re
v
e
c
t
o
r
is
p
r
o
d
u
c
e
d
t
h
ro
u
g
h
a
n
ea
sy
f
r
e
q
u
en
c
y
a
na
lys
i
s
o
f
eac
h
c
odew
o
r
d
i
ns
id
e
t
h
a
t
im
age
in
t
he
p
oo
l
i
n
g
p
ar
t
[
1
5
]
.
F
i
g
u
r
e
1
.
show
s
the
gene
r
a
l
str
uc
t
u
r
e
o
f
Bo
F
fr
ame
w
or
k.
I
n
th
i
s
p
r
o
jec
t
,
S
p
ee
de
d
u
p
R
obus
t
F
e
a
t
ur
es
(
SU
RF
)
ha
s
be
en
u
s
e
d
i
n
B
o
F
be
ca
use
t
h
e
pe
rform
ance
of
t
h
i
s
f
e
a
t
ur
e
is
e
xce
l
len
t
a
nd
o
n
l
y
r
eq
u
i
r
e
l
ow
c
om
p
u
ta
t
i
o
n
a
l
c
o
s
t
[1
8]
.
I
t
i
s
a
n
i
m
a
ge
d
e
t
ec
to
r
a
nd
de
sc
r
i
ptor
t
ha
t
i
s
base
d
o
n
H
e
ssia
n
m
a
tr
i
x
m
ea
sur
e
s.
I
t use
s
a
2D
H
aa
r
w
a
vele
t
tr
an
sf
or
m
f
o
r
a descr
i
pt
o
r
t
ha
t
uses
o
nly
6
4
d
i
m
e
n
si
ons
l
e
a
d
i
n
g
t
o
q
uic
k
f
ea
tur
e
e
x
t
r
act
io
n
[1
8
].
F
o
r
BoF
t
r
ai
ning,
t
he
s
tr
on
g
e
st
f
e
a
t
u
r
e
s
fr
om
e
ac
h
cate
gor
y
ar
e
s
et
t
o
8
0
p
e
r
cen
t
.
B
ased
o
n
t
h
e
r
e
sult,
t
he
a
ve
r
a
ge
a
ccur
a
c
y
i
s
0.
85
w
h
ic
h
show
s
t
h
a
t
by
us
ing
t
h
is
m
et
ho
d,
t
he
a
cc
u
r
acy
i
s
m
o
r
e
t
h
a
n
8
0
pe
r
c
e
n
t.
T
he
c
l
u
ster
in
g
of
t
h
e
d
a
t
a
ha
s
be
e
n
c
omple
t
e
d
o
n
t
h
e
2
0t
h
i
t
e
r
a
t
i
on
i
n
w
h
i
c
h
i
t
i
s
ab
o
u
t
4.
3
9
sec
o
nds/
iter
a
t
i
on.
F
igur
e
2
sh
ow
s
t
h
e
V
i
s
u
al
W
or
d
gr
ap
h
f
o
r
BoF
b
ased
o
n
the
data
u
se
d i
n
th
i
s
pro
j
ec
t
.
F
i
gur
e
1.
G
ene
r
al
s
tr
uc
tur
e
o
f
B
o
F
fr
a
m
ew
ork
[
1
6]
F
i
gur
e
2.
V
isu
a
l
Wor
d
g
r
a
ph
base
d
on
t
h
e
BoF
in
th
is pro
jec
t
Lite
r
a
t
u
r
e
r
evi
e
w
tha
t
h
as
b
ee
n
do
ne
a
u
t
h
o
r
use
d
i
n
t
h
e
c
h
ap
ter
"
I
n
tr
od
uc
ti
on"
t
o
exp
l
a
i
n
the
di
ffe
r
e
nc
e
of
t
he
m
anusc
r
i
p
t
w
ith
o
the
r
p
a
p
e
r
s,
t
hat
it
i
s
in
no
v
a
tive
,
it
a
r
e
u
se
d
i
n
t
he
c
ha
pter
"
R
e
se
ar
c
h
M
e
t
hod
"
t
o
d
escri
b
e
t
h
e
st
ep
o
f
re
se
arc
h
a
n
d
u
sed
i
n
t
h
e
c
h
a
p
t
e
r
"R
esul
ts
a
nd
D
i
s
c
u
ss
io
n
"
t
o
su
pport
t
h
e
a
n
al
ysis
o
f
t
h
e
r
e
sults
[
2].
I
f
t
he
m
an
uscr
ip
t
wa
s
writt
e
n
r
ea
l
l
y
h
a
v
e
h
i
gh
or
i
g
ina
l
i
t
y
,
w
h
ich
pr
o
pose
d
a
n
e
w
m
e
tho
d
o
r
a
l
g
o
r
ithm
,
t
he
a
d
d
it
io
na
l
c
h
a
p
te
r
after
the
"Intr
o
d
u
c
tio
n"
c
hap
t
er
a
nd
b
e
f
or
e
the
"
R
e
s
e
a
r
c
h
Me
th
o
d
"
c
h
ap
t
e
r
ca
n
be
a
dded
to
e
x
p
l
ai
n
br
ie
fl
y
t
h
e
th
e
o
r
y
a
nd
/
or
t
he
p
r
o
p
o
se
d
m
e
tho
d
/
al
gor
ithm
[
4
]
.
2.
2.
Co
n
v
o
l
u
t
io
nal Neura
l
N
et
w
o
rk
(CN
Ns
)
Con
v
o
l
u
t
i
o
n
al
N
eur
a
l
N
e
tw
or
k
(
C
N
N
s
)
c
ons
ists
o
f
f
o
u
r
t
ypes
o
f
l
a
y
e
r
s
wh
i
c
h
a
r
e
co
nv
ol
u
tion
laye
r
s
,
po
ol
in
g
la
ye
r
,
R
e
c
t
i
f
i
ed
L
ine
a
r
un
it
(
R
e
L
u)
l
a
y
er
a
nd
f
u
l
l
y
c
on
necte
d
l
ayer
s.
C
o
n
v
o
l
ut
i
on
laye
r
s
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
ones
i
a
n
J
E
lec
En
g & Co
mp
S
c
i
IS
S
N
: 2502-
47
52
Eva
l
u
a
tio
n
of
bas
i
c
c
onv
olu
t
i
o
n
a
l
neu
r
a
l
ne
t
w
ork
an
d
bag
of
fe
a
t
u
r
e
s
fo
r lea
f
… (Nur
u
l
Fa
tih
ah
Sa
hi
d
an)
32
9
extra
c
t
t
he
i
n
p
u
t
o
f
an
i
ma
g
e
b
y
us
ing
con
v
o
l
u
tio
n
o
p
e
ra
t
i
on
an
d
pr
od
uce
a
fe
at
ur
e
ma
p
[1].
M
ul
tip
le
con
v
o
l
u
t
i
o
n
a
l
l
ayer
s
ca
n
be
a
pp
lied
for
d
i
ffere
nt
f
e
a
t
u
re
m
aps
a
s
w
e
l
l
.
T
his
me
thod
i
s
t
o
e
n
sure
c
o
m
plet
e
extra
c
tio
n
o
f
v
ar
i
o
us
f
ea
tur
e
s.
N
e
x
t
,
p
o
o
l
in
g
laye
r
low
e
r
t
h
e
s
i
z
e
o
f
t
h
e
f
e
a
t
u
r
e
m
a
p
s
.
T
h
i
s
p
r
o
c
e
s
s
m
a
k
e
s
t
h
e
in
put
r
o
bust
a
g
ai
ns
t
n
o
ise
a
n
d
dist
ort
i
o
n
[6
].
N
e
u
ral
ne
tw
orks
a
nd
CN
N
par
tic
ula
r
ly
r
e
l
y
o
n
t
he
t
hird
l
a
y
e
r
w
h
ic
h
is
t
he
a
cti
v
a
t
i
o
n
func
tio
n.
C
N
N
m
a
y
use
sp
ec
ific
f
u
n
c
t
i
o
n
s
suc
h
a
s
ReLU
s
functions
to
e
f
f
ici
e
ntly
implem
e
n
t
no
n
-
li
ner
trig
geri
n
g
.
A
l
l
ne
ga
tive
pi
xe
l
va
lu
es
i
n
t
h
e
fe
a
t
ure
m
a
p
a
r
e
re
place
d
with
z
ero
b
y
u
s
i
n
g
Re
L
u
l
a
y
e
r
[
3]
.
F
u
l
l
y
co
n
n
ec
te
d
la
ye
r
w
h
ic
h
is
t
he
l
as
t
l
a
ye
r,
to
tal
the
w
e
igh
t
a
g
e
o
f
p
re
v
i
ous la
y
er
o
f
fe
atures
to de
t
er
mine
t
h
e
out
p
u
t.
F
i
gure
3
s
h
ow
s
t
h
e
CN
N
arc
h
ite
c
t
ur
e
tha
t
e
xtra
c
t
s
fe
at
ures
by
u
s
i
n
g
c
on
vo
lut
i
on
t
ec
h
n
ique
o
n
t
h
e
in
put
i
m
a
ge
,
r
e
size
the
fe
at
u
r
e
ma
p
durin
g
po
o
lin
g
l
a
yer
and
c
l
a
s
s
i
f
i
e
s
i
t
i
n
t
h
e
f
u
l
l
y
c
o
n
n
e
c
t
e
d
l
a
y
e
r
.
T
h
e
firs
t
co
n
v
o
l
ut
i
on
l
a
yer
us
ua
l
l
y
e
x
t
r
acts
t
h
e
low
-
le
ve
l
fe
a
t
ures
s
uc
h
as
e
dge
s
w
h
ile
t
he
s
eco
nd
c
o
nv
o
l
u
t
i
o
n
layer
ex
tra
c
ts t
he
h
ig
h-
l
e
vel fe
ature
s
s
uc
h a
s
the
s
ha
pe.
F
i
gure
3.
S
truc
t
u
r
e
of
Basic
C
N
N
[19]
3.
RESULT
S
A
N
D
ANALY
S
IS
T
h
e
l
a
p
t
o
p
u
s
e
d
t
o
r
u
n
t
h
e
C
N
N
a
n
d
B
o
F
f
o
r
t
h
i
s
p
r
o
j
e
c
t
w
a
s
L
e
n
o
v
o
w
ith
W
i
ndow
s
10
as
t
h
e
opera
tin
g
sy
st
e
m
,
Intel
Core
i
7
pr
oc
ess
o
r,
a
nd
a
n
8
.
0
0
G
B
R
A
M
w
hi
le
t
h
e
s
oftw
ar
e
use
d
w
as
M
a
t
lab
201
8a
.
The
da
tas
e
t
us
ed
i
s
F
o
li
o
Le
af
D
ata
Set
[1
9].
Leave
s
p
i
c
tures
a
re
t
a
k
en
f
ro
m
pl
a
n
t
s
o
n
th
e
f
a
rm
o
f
th
e
Uni
v
ersi
ty
o
f
Ma
urit
ius
a
n
d
near
by
l
o
c
a
t
ions.
There
ar
e
32
ca
t
e
g
o
ries
o
f
p
l
a
n
t
a
n
d
for
e
ach
c
a
t
eg
ory
2
0
i
m
ag
e
s
o
f
l
eav
e
s
a
re
e
xp
e
r
ime
n
t
e
d
.
A
l
l
th
e
i
m
a
g
es
a
re
r
es
i
zed
i
n
t
o
224
b
y
2
24
pixe
l
s
t
o
e
n
sur
e
t
he
con
s
is
t
e
ncy
of
t
he
d
a
t
a
for
ea
ch
m
e
t
h
od.
F
i
g
ure
4
sh
ow
s
sa
m
p
le
ima
g
es
o
f
F
o
lio
L
ea
f
da
tase
t
fr
o
m
a
l
l
t
he
3
2
ca
t
e
g
o
ries.
Fi
g
u
re
4
.
Sa
mp
l
e
i
ma
g
e
s
o
f
F
o
l
i
o
L
e
a
f
d
a
t
a
se
t
Ex
pe
rime
n
t
s
w
e
re
c
ond
uc
te
d
by
c
h
a
ngi
ng
th
e
num
ber
o
f
l
a
y
ers,
t
h
e
va
l
u
e
s
o
f
the
pa
ram
e
ter
s
i
n
t
h
e
con
v
o
l
v
e
layer
,
poo
l
i
n
g
l
a
y
e
r
a
nd
the
lea
r
n
i
n
g
r
ate.
T
he
pur
pos
e
is
t
o
de
t
e
rm
ine
the
bes
t
c
om
b
i
na
ti
on
of
val
u
es
t
o
pro
d
u
ce
t
h
e
hi
g
h
es
t
a
c
c
u
rac
y
f
or
l
eaf
r
ec
ogn
i
t
i
o
n
fro
m
Fo
l
i
o
d
at
as
e
t
.
Th
e
resul
t
o
f
th
e
ex
p
e
rimen
t
s
was
recorded
i
n
Table
1.
B
y
refe
rr
i
ng
to
T
able
1
.
the
first
col
um
n
in
dic
a
t
es
t
he
numbe
r
o
f
s
t
a
c
k
s
of
l
a
y
er
s
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SSN: 2502-
4752
I
n
do
n
e
si
an
J
E
l
e
c
E
n
g
&
C
o
m
p
S
ci
, V
o
l
.
1
4
,
No. 1, April 2019 :
327 –
3
32
33
0
wher
e
a
stac
k
c
onsis
ts
o
f
o
n
e
co
nv
o
l
ve
l
a
y
e
r
,
one
m
ax-po
o
l
i
n
g
l
ayer
a
nd
o
n
e
Re
L
u
l
a
y
e
r
.
I
n
col
u
mn
Con
v
o
l
v
e
laye
r
,
t
he
f
irs
t
n
u
m
ber
in
t
he
s
q
u
ar
e
brac
ke
t
repre
s
en
t
s
t
h
e
s
i
z
e
o
f
t
h
e
c
o
n
v
o
l
v
e
f
i
l
t
e
r
w
h
i
l
e
t
h
e
seco
nd
n
um
ber
re
presen
ts
t
he
num
ber
of
c
o
n
v
o
l
ve
f
il
ters.
T
he
t
h
i
rd
l
ay
e
r
r
ep
re
sent
s
t
h
e
si
ze
o
f
max
-
poo
li
n
g
fi
l
t
e
r
and
t
he n
umbe
r of stri
d
e
.
The n
u
mbe
r
o
f
e
p
oc
h
a
nd l
ear
n
ing ra
te
i
s show
n i
n
c
o
l
u
m
n 4.
The
nu
m
b
er o
f
epoc
h
determ
i
n
es
t
he
numbe
r
of
r
epe
t
i
t
i
o
n
s
o
f
a
ll
th
e
t
r
ai
ni
ng
d
a
t
a
w
h
i
l
e
t
h
e
l
e
a
r
n
i
n
g
r
a
t
e
i
s
t
h
e
a
m
o
u
n
t
o
f
adj
u
s
t
m
e
n
t
t
ha
t
is
b
ei
n
g
m
a
d
e
to
t
he
w
e
i
g
h
t
s
dur
in
g
the
tr
ain
i
n
g
proc
es
s.
B
y
l
o
o
k
i
ng
a
t
T
a
b
le
1
.
the
be
st
ac
cura
cy
i
s
ac
hie
v
e
d
w
h
e
n
t
h
er
e
a
r
e
t
h
re
e
stac
ks
o
f
l
a
y
e
rs,
but
t
h
e
a
ccu
ra
cy
s
t
a
rts
t
o
d
ec
re
a
s
e
whe
n
t
h
e
numbe
r of s
tac
k
s
i
s
m
or
e tha
n
t
hree
.
T
a
b
l
e 1.
Expe
rime
nt
a
l
r
e
s
u
lts
on pa
ram
e
te
r t
u
n
i
ng
f
or
b
asic
C
N
N
No
o
f
S
t
ack
of
L
aye
r
s
C
onvolve
L
a
y
e
r
Pooli
ng
la
ye
r
a
nd
St
r
i
d
e
Epoc
h,
Lea
r
n
ing
Ra
te
Ac
c
u
r
a
c
y
(%)
T
o
t
al T
ime
/
s
1
[3
,
16]
3
10,
0
.
001
71.
92
6
m
i
n
32s
[5
,
20]
2
10,
0
.
0001
65.
62
8
m
i
n
23s
2
[3,
16],
[
3,
16]
3
10,
0
.
001
79.
81
4
m
i
n
27s
[3,
80],
[
3,
64]
2
10,
0
.
001
73.
82
15m
in
40s
3
[3
,
16]
,
[
3
,
16]
,
[
3,
3
2]
3
10,
0
.
001
76.
66
4
m
i
n
44s
[5
,
20]
,
[
3
,
20]
,
[
3,
16]
3
10,
0
.
001
82.
03
3
m
i
n
35s
F
i
gure
5.
s
h
o
w
s
t
r
a
i
n
i
n
g
pr
o
g
r
e
ss
for
c
o
n
v
o
lve
layer
[5,2
0],
[3
,
20],
[
3
,16]
,
po
oli
n
g
la
ye
r
an
d
s
t
ri
d
e
3,
e
poc
h
10
a
nd
lear
nin
g
r
at
e
0.00
1.
B
y
us
in
g
t
h
ese
par
a
m
e
ter
v
alues
the
a
ccura
cy
r
esu
lt
re
ac
he
d
a
t
82.0
3
w
i
t
h
e
l
a
p
s
e
d
t
i
m
e
o
f
3
m
i
n
u
t
e
s
a
n
d
3
5
s
e
c
o
n
d
s
w
h
i
c
h
i
s
f
a
s
t
e
r
c
ompa
re
t
o
t
h
e
ot
her
r
e
sul
t
s.
T
he
s
e
l
a
y
ers
an
d
para
me
ters
i
s muc
h
m
ore
accurate
com
par
e
to
o
t
he
rs.
Fi
g
u
re
5
.
T
r
a
i
ni
ng
P
r
o
g
r
e
s
s
Ta
b
l
e
2.
S
how
s
an
over
v
i
e
w
of
t
he
acc
urac
y
p
e
rfor
m
a
n
ce
of
b
a
s
ic
C
N
N
com
p
are
d
t
o
B
o
F
base
d
o
n
Foli
o
Leaf
D
a
t
aset.
B
y
l
o
o
k
i
n
g
a
t
T
a
b
l
e
2,
w
e
c
a
n
see
t
h
at
B
oF
i
s
b
e
t
ter
t
h
an
b
a
s
ic
C
N
N
bu
t
i
t
t
oo
k
a
l
o
n
g
er
ti
m
e
t
o
ac
h
i
e
v
e
th
i
s
r
e
s
ul
t.
Th
is
i
s
bec
a
us
e
e
x
tra
c
ti
n
g
o
f
t
h
e
S
U
R
F
fe
at
ures
i
s
lon
g
e
r
c
om
pare
t
o
the
time
t
o
extra
c
t
the
l
ow
-
l
e
v
e
l
a
nd m
i
ddle
le
ve
l
fea
t
ur
es by
the
basic
CN
N.
Tab
l
e
2. The
Perform
ance
ov
e
r
v
i
e
w for bas
i
c CNN and BoF
for
F
o
l
i
o
Leaf
Da
t
a
s
et
Mo
d
e
l
Ba
s
i
c
C
N
N
B
o
F
Valid
at
ion
Ac
c
u
rac
y
0.
82
0
.
8
5
E
l
apse
d T
i
me
(s)
177
276
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
ones
i
a
n
J
E
lec
En
g & Co
mp
S
c
i
IS
S
N
: 2502-
47
52
Eva
l
u
a
tio
n
of
bas
i
c
c
onv
olu
t
i
o
n
a
l
neu
r
a
l
ne
t
w
ork
an
d
bag
of
fe
a
t
u
r
e
s
fo
r lea
f
… (Nur
u
l
Fa
tih
ah
Sa
hi
d
an)
33
1
4.
CONCL
US
IO
N
T
h
i
s
p
a
p
e
r
h
a
s
p
r
e
s
e
n
t
e
d
t
h
e
e
v
a
l
u
a
t
i
o
n
o
f
b
a
s
i
c
C
N
N
a
n
d
B
o
F
f
o
r
leaf
r
e
c
ogn
i
t
i
o
n.
I
n
t
h
i
s
p
a
p
er
,
the
a
ccur
acy
p
erforma
n
ce
f
or
l
eaf
r
e
c
o
g
n
i
t
i
on
ba
se
d
on
F
o
li
o
l
e
a
f
d
a
t
a
s
e
t
i
s
co
mp
ared
b
et
we
en
b
a
s
i
c
C
NN
and
BoF
.
T
he
e
xper
i
m
e
nta
l
r
esu
l
t
s
s
h
o
w
t
h
at
b
a
s
ic
C
N
N
a
c
hie
v
e
s
a
l
o
w
e
r
acc
urac
y
r
a
te
c
om
p
a
red
t
o
B
oF
sinc
e
it
r
e
q
u
i
res
a
h
u
g
e
am
ou
nt
o
f
da
t
a
c
om
par
e
d
to
B
oF
.
The
r
ef
or
e,
i
f
t
h
e
numbe
r
of
d
ata
is
l
im
ited,
B
oF
stil
l
prov
i
d
es
a
g
o
o
d
r
e
s
u
lt
an
d
p
r
efe
r
a
b
le
c
o
m
par
e
d
to C
N
N
.
F
or
t
he
f
ut
ure
r
e
sea
r
ch,
we
p
la
n t
o
e
n
h
a
n
ce
t
h
e
C
N
N a
r
c
h
i
t
e
ctu
r
e a
n
d
in
c
r
ea
se
s t
h
e nu
mb
er o
f
d
a
t
a
set
s
to
ob
t
a
in
a m
ore
ac
curate and
f
aster
results.
ACKNOW
LEDG
E
MEN
T
S
The
au
t
hors
w
oul
d
l
i
k
e
to
t
ha
nk
F
acult
y
o
f
C
om
pu
t
e
r
a
nd
M
a
t
h
e
m
a
tic
al
S
cience
s,
U
n
i
v
ersi
tiTe
k
n
o
lo
gi
M
A
R
A
, S
ha
h A
l
am
,
S
e
l
a
n
g
o
r,
f
or
s
po
nsor
in
g
t
hi
s r
e
sear
ch.
REFE
RENCES
[1]
Z.
I
brahim,
et a
l
.,
"
Mu
lti-m
a
xpoo
li
ng
Con
v
o
l
u
tio
na
l
Neu
r
al
N
etw
o
rk
f
or
M
edici
n
al
H
erb
L
eaf
R
ec
ognition
,
”
Pr
oceedi
n
gs of
the 6t
h II
AE
Inter
n
ati
onal Con
f
erence on Int
e
lligent Sys
t
ems and Image Pr
ocess
i
ng
, Sh
i
m
ane,
Jap
a
n
, p
p.
32
7
-3
3
1
,Sep
t
e
mb
er 20
1
8
.
[2]
Z.
I
brahi
m
,
et
al
.
,
“
Leaf
R
ecognition
using
Text
ur
e
Features
f
or
H
e
r
bal
Plant
I
d
entif
i
catio
n,
”
Ind
onesi
an Jou
r
n
a
l
of
El
ectr
i
cal
En
gi
n
eer
ing
a
nd
Co
m
p
ut
er S
c
ien
c
e
,
V
o
l
. 9
, No
. 1
, p
p
. 15
2
-1
5
6
,
Jan
u
ar
y 20
1
8
.
[3]
Mu
nisa
mi.
T
,
et a
l
.,
“
P
lant
l
eaf
r
ecog
n
i
tio
n
u
s
in
g
s
h
ap
e
f
eatures
a
n
d
c
o
l
o
u
r
h
i
s
t
og
ra
m
with
k
-n
e
a
r
e
s
t
ne
ig
hb
ou
r
clas
sifiers,”
Pr
o
cedi
a
Co
mp
uter
S
c
ience (Els
evier) Jour
na
l
.
58,
p
p
.
7
40
-747
,
2
015.
[4]
L.
D
eng
and
D
.
Y
u
,
“
Deep
L
earn
i
ng
:
M
e
t
h
o
d
s
and
Ap
pli
cati
ons,”
Fo
u
n
d
a
t
io
n
s
an
d
Tre
n
ds
in
S
i
gn
al
Proc
e
s
sin
g
,
vo
l.
7
,
n
o
.
3
–
4
,
p
p
.
19
7–
387
,
2
013.
[5]
E. N
i
s
han
i
A
nd
B
. Cico,
“
C
o
m
puter Visio
n App
r
oach
es
B
as
ed
O
n
D
e
ep
L
earni
ng
An
d
N
eural Net
w
ork
s
,”
20
17 6th
M
e
d
i
ter
r
a
n
ea
n Co
nf
e
r
ence
O
n
E
m
b
e
dd
ed
Compu
t
i
ng (
M
eco)
,
Ba
r, M
onten
egr
o
,
11
-15
Ju
ne
2
01
7
.
[6]
M.
A
fiq,
et a
l
.
,
“
Deep
L
earni
ng
f
o
r
Roman
Han
d
written
C
h
aracter
R
ecog
n
ition,”
Ind
o
n
e
si
an Jo
urn
a
l
of
Electr
i
cal
En
gi
neeri
ng an
d
Co
mp
u
t
er S
c
ien
c
e
.
Vo
l.
1
2
,
no.
2
,
p
p
.
4
55
-460
, N
ov
emb
e
r
20
18
.
[7]
S
.
A
lb
awi,
e
t
al
.
,
"
U
nd
erstandi
ng
o
f
a
conv
ol
utio
nal
neu
r
al
n
et
w
o
rk,"
2
017
In
ter
n
a
t
io
n
a
l Con
f
eren
ce o
n
En
gi
neeri
ng an
d
T
ech
no
log
y
(
I
CET)
, An
ta
l
y
a
, p
p
. 1
-6,2
01
7
.
[8]
R.
D
urr
a
tu
n,
et al
.,“P
erf
o
rm
an
ce
E
v
al
uati
on
f
o
r
V
isio
n-Based
Vehi
cle
Clas
sifficat
io
n
Us
in
g
Con
v
o
l
ut
io
na
l
Ne
ura
l
Net
w
ork,”
Internat
io
nal Jo
ur
nal o
f
E
ngineer
in
g
&
T
ech
no
log
y
,
7
(3.
15)
86-9
0
,
2018
.
[9]
T.
G
uo
,
et a
l
.
,
"S
im
p
l
e
con
v
o
l
u
tio
na
l
n
e
ural
n
etw
o
rk
o
n
i
m
age
cl
assi
ficatio
n,
"
2
0
1
7
IE
EE
2n
d In
ter
natio
na
l
Conference on B
i
g
Da
t
a
Analysis
(
I
C
B
DA
)
,
B
e
ijing
,
p
p.
7
2
1
-72
4
,
201
7.
[10]
G.
C
heng,
et a
l
.,
“
S
c
en
e
cl
assif
i
catio
n
o
f
h
igh
reso
lutio
n
rem
o
t
e
s
en
si
ng
im
a
g
e
s
u
sing
c
o
n
vo
lu
tion
a
l
n
e
ura
l
net
w
ork
s
,
”
IEEE Int
e
rn
a
t
i
o
n
a
l
Co
nfer
ence in
G
e
osci
ence an
d Remo
te Sen
s
in
g Sym
p
o
s
ium
(
I
G
A
RSS
)
,
pp.
7
67
-770,
20
16
.
[11]
J.
S
oc
ho
r,
et
al
.
,
"
B
oxCars:
3D
B
oxes
a
s
CN
N
Input
f
o
r
Im
prove
d
F
i
n
e-Grai
ned
V
e
h
icl
e
R
ecognit
ion,"
201
6 IEEE
Con
f
eren
ce on
Com
p
u
t
er Vi
si
on an
d P
a
tt
ern
Reco
gn
ition
(
C
VP
R),
Las Vega
s,
NV
,
p
p
.
3
00
6-3
0
1
5
,
2
01
6.
[12]
Z.
I
brahim,
et al
.
,
"P
alm
Oil
Fres
h
F
r
uit
Bu
nch
Ri
penes
s
G
radi
ng
R
eco
gni
ti
on
Usi
n
g
C
on
vol
utio
nal
Neu
r
al
Net
w
ork,"
Jou
r
nal
o
f
T
e
leco
mmuni
cation
,
Elect
ro
n
i
c
&
Comp
u
t
er
E
n
g
i
neer
in
g
,
vol.
9
,
n
o
.
3-2,
p
p
. 1
09
-1
13
,
2
018
.
[13]
C.
C
hi
n
,
et
al.
,
"S
m
a
rtp
hon
e-bas
e
d
f
o
o
d
cat
eg
ory
and
nu
trition
q
u
ant
i
t
y
r
eco
gn
i
t
i
o
n
in
f
o
o
d
im
ag
e
wit
h
d
eep
learning
a
lgorith
m,"
20
1
6
In
te
rn
atio
na
l Con
f
e
r
e
n
c
e
on
Fuzzy
Th
e
o
ry
a
n
d
Its Ap
plic
a
t
io
n
s
(iFuzz
y
)
,
Ta
ic
hu
ng
,
pp.
1-1
,
2
016
.
[14]
R.A.
C
h
e
n
,
e
t
al.
,
"
En
han
ced
B
a
g
-of
-
F
eat
ures
m
od
el
f
or
i
m
a
g
e
c
l
assif
i
cation,"
2014 IEEE
W
o
r
ks
h
o
p
on A
d
van
ced
Res
e
arch
an
d
Te
ch
no
lo
gy i
n
In
dustr
y
Ap
p
l
icati
ons
(
W
ART
I
A)
, Ottawa,
ON
, pp
.
1
19
5-
11
98
,
2
0
1
4
.
[15]
S
.
H
.
Bhan
dari
,
"A bag-o
f
-
f
eatures
a
pp
roa
c
h
f
o
r m
a
lig
nancy
d
e
te
cti
o
n
in
b
reas
t histo
p
at
holog
y images
,"
20
15
IEE
E
Int
e
rna
t
i
o
n
a
l
Co
n
f
er
e
n
ce on
Image
Pro
cessing
(
I
CIP)
, Qu
ebec Cit
y
,
Q
C
,
p
p
.
4
9
3
2
-4
936
,
2
01
5.
[16]
M
.
O
.
As
say
ony
and
S
.
A
.
M
a
h
m
o
u
d
,
"
In
tegratio
n
of
G
ab
or
F
eature
s
wit
h
B
ag-o
f
-
Feat
ures
F
ramewo
rk
f
or
A
rab
i
c
Han
d
writt
en
W
ord
Reco
gni
ti
on,
"
20
17
9
t
h IEEE-GCC Con
f
eren
ce
an
d Exh
i
bi
ti
on
(
G
CCCE)
,
M
ana
ma
,
p
p
.
1-4,
20
17
.
[17]
P
.
P
aw
ara,
e
t
a
l
.,
“
D
a
t
a
Au
gm
entat
i
on
f
o
r
P
l
a
nt
C
l
a
ssif
icatio
n,
”
A
d
vanced Concept
s
for
In
t
e
lligent
Visi
on Systems:
18
th
In
t
e
rna
t
i
onal
Conferen
ce,
ACIVS
201
7
,
An
twerp
,
B
e
lgium
, P
ro
ceedin
gs
, S
e
pt
em
ber 1
8
-2
1, 2
01
7.
[18]
A.
K
.
K
h
an,
et
a
l
.,
“
E
v
a
lu
ati
on
of
S
IF
T
an
d
S
U
RF
u
si
ng
B
ag
o
f
W
ords
M
o
d
el
on
a
Very
L
arge
D
ataset
,”
Si
nd
h
Univ.
R
e
s
.
Jo
ur
.
(
S
ci
.
S
e
r.
)
,
V
o
l
.
4
5 (3),
p
p
.
492-49
5
,
2013.
[19]
J.W.
S
u
and
R
.
S
.
Y
ong
,
"
P
l
ant
Le
af
R
ecognition
Usi
ng
a
Convo
l
ut
iona
l
Neural
N
etwork,"
I
n
te
r
n
a
t
io
na
l J
o
u
r
na
l
of
Fu
zz
y
L
ogic a
nd Int
e
lli
gen
t
S
y
st
e
m
s
,
vo
l.
17,
no.
1
,
p
p.
26-3
4
,
M
a
rc
h
2
017
.
[20]
Folio
D
at
a Se
t.
R
eri
e
ved f
r
om h
tt
p
://archive.
i
cs.
u
ci
.edu/
m
l/dat
as
et
s/F
o
li
o
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
250
2-
4
7
5
2
In
do
n
e
sia
n
J
Elec Eng
&
C
o
mp
S
ci,
Vo
l. 1
4
,
No
. 1
,
Ap
r
i
l
2
0
1
9
:
327
–
3
3
2
33
2
BIOGRAPHI
E
S
OF AU
T
H
ORS
Nu
rul
F
a
ti
hah
Bin
t
i
S
a
hi
dan
is
a
M
a
s
ter’s
s
tud
e
nt
i
n
Com
p
u
t
er
S
ci
e
n
ce
(W
eb
T
echno
lo
g
y
)
at
Un
iv
ersi
tiT
ekn
o
l
ogi
M
ARA,
S
h
a
h
Al
am,
S
e
l
a
ngo
r,
M
al
aysia.
H
e
r
a
r
ea
of
i
nt
erests
i
s
im
ag
e
p
r
oce
s
s
i
n
g
,
dat
a
sci
ence an
d b
i
g
dat
a
, m
ulti
me
d
i
a
c
o
mp
ut
in
g
a
n
d we
b-b
a
se
d
t
echn
o
lo
gy
.
Ah
m
a
d
Khai
ri
B
in
J
uh
a
is
a
M
as
ter’s
s
t
u
d
e
nt
i
n
Co
mpu
t
er
S
c
i
ence
(
W
e
b
Tech
no
logy
)
at
Un
iv
ersi
tiT
ekn
o
l
ogi
M
A
R
A
,
S
h
a
h
A
l
am,
Se
langor,
Malaysia
.
Hi
s
ar
ea
o
f
i
nt
erests
i
s
im
ag
e
p
r
oce
s
s
i
n
g
,
dat
a
sci
ence an
d b
i
g
dat
a
, b
u
si
ness co
m
putin
g
and
web-b
a
sed
tec
hn
olo
g
y
.
Zai
d
ah
I
brah
im
i
s
an
A
sso
ciate
Pro
f
ess
o
r
at
t
h
e
F
acul
t
y
of
C
o
m
p
ut
er
a
nd
M
ath
e
m
a
tical
S
cien
ces,
Un
iv
ersi
tiT
ekn
o
l
ogi
M
ARA,
S
hah
Al
am,
S
e
l
a
ng
or,
M
a
lay
s
ia.
S
h
e
h
a
s
been
t
e
ach
i
ng
co
urses
rel
a
ted
t
o
Artif
i
cia
l
I
nt
ellig
e
nce
f
o
r
ov
er
t
en
y
e
a
rs.
S
h
e
is
acti
v
el
y
i
n
v
ol
v
e
d
in
r
es
earch
a
n
d
p
ub
li
ca
t
i
o
n
s
u
nder
Di
g
i
ta
l
Im
age,
A
u
d
i
o
a
nd
S
p
eech
T
ech
no
lo
gy
(DIA
ST)
res
ear
ch
i
nt
erest
g
r
ou
p
t
h
at
i
nclu
de
t
ex
t
and
ob
ject
recog
n
ition
Evaluation Warning : The document was created with Spire.PDF for Python.