I
A
E
S
I
n
t
e
r
n
at
io
n
al
Jou
r
n
al
of
A
r
t
if
ic
ia
l
I
n
t
e
ll
ig
e
n
c
e
(
I
J
-
AI
)
V
ol
.
10
, N
o.
4
,
D
e
c
e
m
be
r
202
1
, pp.
872
~
878
I
S
S
N
:
2252
-
8938
,
D
O
I
:
10.11591/
ij
a
i.
v
10
.i
4
.pp
872
-
878
872
Jou
r
n
al
h
om
e
page
:
ht
tp
:
//
ij
ai
.
ia
e
s
c
or
e
.c
om
E
f
f
e
c
t
o
f
f
i
l
t
e
r
si
z
e
s on
i
m
age
c
l
ass
i
f
i
c
at
i
on
i
n
C
N
N
:
a
c
ase
st
u
d
y
on
C
FIR
10 an
d
F
ash
i
o
n
-
M
N
IS
T
d
at
ase
t
s
O
w
ai
s
M
u
j
t
ab
a K
h
an
d
ay,
S
am
ad
D
ad
van
d
ip
ou
r
, M
oh
d
A
a
q
ib
L
on
e
Institut
e of Informati
on Science, Un
iversity
of Miskol
c, Miskol
c,
Hungary
A
r
t
ic
le
I
n
f
o
A
B
S
T
R
A
C
T
A
r
ti
c
le
h
is
to
r
y
:
R
e
c
e
iv
e
d
N
ov 23, 2020
R
e
vi
s
e
d A
ug 27, 2021
A
c
c
e
pt
e
d S
e
p 10, 2021
Convolution
neural
network
s
(CNN
or
ConvNet
),
a
deep
neural
network
class
inspired
by
biological
processe
s,
are
immensely
used
for
image
classi
fication
or
visual
imagery.
These
networks
need
various
parameters
or
attrib
utes
like
number
of
filters,
filter
size,
number
of
input
channels,
padding
str
i
de
and
dilation,
for
doing
the
required
task.
In
this
paper,
we
focused
on
the
hyperparameter,
i.e.,
filter
size.
Filter
sizes
come
in
various
sizes
li
ke
3×
3,
5×
5,
and
7×
7.
We
varied
the
filter
sizes
and
recorded
their
effects
on
the
models'
accuracy.
The
models'
architecture
is
kept
intact
and
only
the
filter
sizes
are
varied.
This
gives
a
better
understanding
of
the
effect
of
filt
er
sizes
on
image
classification.
CIFAR10
and
FashionMNIST
datasets
are
used
for
this
study.
Experimental
results
showed
the
ac
curacy
is
inversely
proportional
to
the
filter
size.
The
accuracy
using
3×
3
filters
on
CIFAR10
and
Fashion
-
MNIST is 73.04% and 93
.68%, respec
tively.
K
e
y
w
o
r
d
s
:
C
I
F
A
R
10
C
onvolut
io
na
l
ne
ur
a
l
ne
twor
k
D
e
e
p l
e
a
r
ni
ng
F
a
s
hi
on
-
M
N
I
S
T
F
il
te
r
s
iz
e
This is an
open
acce
ss artic
le unde
r
the
CC BY
-
SA
license
.
C
or
r
e
s
pon
di
n
g A
u
th
or
:
O
w
a
is
M
uj
ta
ba
K
ha
nda
y
I
ns
ti
tu
te
of
I
nf
or
m
a
ti
on
S
c
ie
nc
e
U
ni
ve
r
s
it
y of
M
is
kol
c
E
gyt
e
m
va
r
os
, 3525,
M
is
kol
c
, H
unga
r
y
E
m
a
il
:
a
it
ow
a
is
@
uni
-
m
is
kol
c
.hu
1.
I
N
T
R
O
D
U
C
T
I
O
N
A
r
ti
f
ic
ia
l
i
nt
e
ll
ig
e
nc
e
(
A
I
)
ha
s
m
in
im
iz
e
d
th
e
ga
p
be
twe
e
n
h
um
a
n
a
nd
m
a
c
hi
ne
c
a
pa
bi
li
ti
e
s
.
T
he
r
e
s
e
a
r
c
he
r
s
a
nd
e
nt
hus
ia
s
ts
a
r
e
doi
ng
e
nor
m
ous
w
or
k
on
num
e
r
ous
a
s
p
e
c
ts
of
th
e
f
ie
ld
to
m
a
k
e
e
xt
r
a
or
di
na
r
y
th
in
gs
ha
ppe
n
in
m
a
ny
dom
a
in
s
.
T
h
e
dom
a
in
of
c
om
put
e
r
vi
s
io
n
is
one
s
uc
h
a
r
e
a
.
V
a
r
io
us
r
e
c
ogni
ti
on
or
c
la
s
s
if
ic
a
ti
on
a
lg
or
it
hm
s
w
e
r
e
de
v
e
lo
pe
d
li
ke
s
uppor
t
ve
c
to
r
m
a
c
hi
ne
s
(
S
V
M
)
,
n
e
ur
a
l
ne
twor
ks
(
N
N
)
,
m
ul
ti
-
le
ve
l
pe
r
c
e
pt
r
ons
(
M
L
P
)
, a
nd ma
ny mor
e
. A
l
ot
of
m
a
c
hi
ne
l
e
a
r
ni
ng a
lg
or
it
hm
s
ha
ve
be
e
n pr
opos
e
d t
o a
c
hi
v
e
th
e
ta
s
k
of
c
ha
r
a
c
te
r
r
e
c
ogni
ti
on
[
1]
-
[
3]
.
T
he
s
e
a
dv
a
nc
e
m
e
nt
s
he
lp
e
d
m
a
c
hi
ne
s
to
pe
r
c
e
iv
e
im
a
ge
a
nd
vi
de
o
r
e
c
ogni
ti
on,
na
tu
r
a
l
la
ngua
ge
pr
oc
e
s
s
in
g,
a
nd
m
uc
h
m
or
e
.
D
e
e
p
le
a
r
ni
ng
is
e
nor
m
ous
ly
pe
r
f
e
c
te
d
w
it
h
ti
m
e
,
pr
im
a
r
il
y
ove
r
c
onvolut
io
na
l
ne
ur
a
l
ne
twor
ks
(
C
N
N
)
,
a
nd
gi
v
e
s
m
uc
h
b
e
tt
e
r
r
e
s
ul
ts
th
a
n
th
e
ot
he
r
s
[
4]
-
[
7]
.
C
N
N
is
now
a
w
e
ll
-
e
s
ta
bl
is
he
d
m
a
c
hi
ne
le
a
r
ni
ng
to
ol
us
e
d
f
o
r
im
a
ge
c
la
s
s
if
ic
a
ti
on
pr
obl
e
m
s
,
e
s
pe
c
ia
ll
y
in
m
e
di
c
a
l
s
c
ie
nc
e
s
a
nd
pr
a
c
ti
c
a
l
li
f
e
li
ke
de
te
c
ti
ng
r
oa
d
s
ig
ns
,
r
e
c
ogni
z
in
g
hum
a
n
a
c
ti
vi
ty
,
a
nd
f
a
c
ia
l
e
xpr
e
s
s
io
n
r
e
c
ogni
ti
on [
8]
, [
9]
.
C
N
N
is
a
de
e
p
le
a
r
ni
ng
a
lg
or
it
hm
th
a
t
ta
ke
s
a
n
im
a
ge
a
s
in
put
a
nd
r
e
c
ogni
z
e
s
it
.
T
he
a
r
c
hi
te
c
tu
r
e
of
C
N
N
is
b
a
s
e
d
on
hum
a
n
ne
ur
ons
.
T
h
e
s
e
a
r
e
f
e
e
d
-
f
or
w
a
r
d
ne
ur
a
l
ne
twor
ks
th
a
t
c
a
n
c
a
pt
ur
e
th
e
te
m
por
a
l
a
nd
s
pa
ti
a
l
de
pe
nde
nc
ie
s
by
a
ppl
yi
ng
r
e
le
va
nt
f
il
te
r
s
.
T
he
ne
twor
k
c
a
n
e
xt
r
a
c
t
f
e
a
tu
r
e
s
w
it
hout
m
a
nua
l
ha
ndl
in
g
[
10]
.
C
N
N
is
us
e
d
f
or
a
w
id
e
r
r
a
nge
of
im
a
ge
r
e
c
ogni
ti
on
ta
s
k
li
ke
m
e
di
c
a
l
im
a
ge
r
e
c
ogni
ti
on
[
11]
,
[
12]
,
x
r
a
ys
r
e
c
ogni
ti
on
[
13]
,
[
14]
,
ha
ndw
r
it
te
n
c
ha
r
a
c
t
e
r
r
e
c
ogni
ti
o
n[
15]
-
[
17
]
,
a
nd
of
f
li
ne
c
ha
r
a
c
te
r
r
e
c
ogni
ti
on
[
18]
,
[
19
]
.
A
C
N
N
ha
s
th
e
in
put
la
ye
r
,
c
onvolut
io
na
l
la
ye
r
,
a
m
a
x
-
pool
in
g
la
ye
r
,
a
nd
a
f
ul
ly
c
onne
c
te
d
la
ye
r
.
I
n
th
e
e
nd,
w
e
h
a
ve
th
e
S
of
tM
a
x
a
ppl
ie
d
to
c
la
s
s
if
y
th
e
im
a
ge
to
it
s
r
e
s
pe
c
ti
ve
c
la
s
s
[
20]
.
T
h
e
c
onvolut
io
na
l
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
E
ff
e
c
t
of
f
il
te
r
s
iz
e
s
on i
m
ag
e
c
la
s
s
if
ic
at
io
n i
n C
N
N
…
(
O
w
ai
s
M
uj
ta
ba K
handay
)
873
la
ye
r
is
a
2D
la
y
e
r
,
a
nd
th
e
f
il
te
r
s
a
r
e
u
s
e
d
in
th
is
la
ye
r
.
C
N
N
in
put
la
ye
r
ha
s
th
r
e
e
di
m
e
n
s
io
ns
.
T
he
in
put
im
a
ge
is
a
th
r
e
e
-
di
m
e
ns
io
n
a
l
im
a
ge
c
om
pr
is
in
g
he
ig
ht
,
w
id
th
,
a
nd
de
pt
h.
T
h
e
de
pt
h
is
t
a
ke
n
a
s
on
e
f
or
th
e
gr
e
ys
c
a
le
i
m
a
ge
a
nd t
hr
e
e
f
or
t
he
R
G
B
i
m
a
ge
. T
he
f
il
te
r
s
a
r
e
t
h
e
f
e
a
tu
r
e
m
a
tr
ix
e
s
of
va
r
io
us
s
iz
e
s
,
e
.g.
3
×
3
,
5
×
5
, a
nd
7
×
7
. T
he
f
il
te
r
s
a
r
e
r
e
pr
e
s
e
nt
e
d by the
ve
c
to
r
of
w
e
ig
ht
s
, w
hi
c
h t
r
ie
s
t
o f
in
d
w
he
th
e
r
t
he
f
e
a
tu
r
e
s
a
r
e
a
va
il
a
bl
e
in
th
e
in
put
im
a
ge
.
T
he
ve
c
to
r
of
th
e
w
e
ig
ht
s
r
e
pr
e
s
e
nt
s
a
f
e
a
t
ur
e
s
uc
h
a
s
a
c
ur
ve
,
e
dg
e
,
a
nd
s
ha
pe
.
W
e
h
a
ve
c
ons
id
e
r
e
d
th
r
e
e
ty
pe
s
of
f
il
te
r
s
:
3
×
3
,
5
×
5
,
a
nd
7
×
7
,
a
nd
c
he
c
ke
d
th
e
im
pa
c
t
of
th
e
s
e
f
il
te
r
s
on t
he
i
m
a
ge
c
la
s
s
if
ic
a
ti
on.
I
n
th
e
c
onvolut
io
na
l
la
ye
r
,
th
e
out
put
is
c
a
lc
ul
a
te
d
a
s
a
dot
pr
o
duc
t
of
th
e
w
e
ig
ht
s
a
n
d
th
e
in
put
a
nd
th
e
n
a
dds
s
om
e
bi
a
s
.
T
he
f
il
te
r
s
s
li
de
ove
r
th
e
in
put
im
a
ge
m
a
tr
ix
a
nd
pr
oduc
e
th
e
c
onvolut
io
na
l
ou
tp
ut
.
I
f
th
e
in
put
is
h
a
vi
ng
di
m
e
ns
io
n
s
of
×
×
1
a
nd
th
e
is
th
e
num
be
r
of
f
il
te
r
s
a
ppl
ie
d,
th
e
n
th
e
out
put
of
th
e
c
onvolut
io
na
l
la
ye
r
i
s
×
×
.
T
he
s
tr
id
e
is
a
ppl
ie
d
to
de
te
r
m
in
e
th
e
num
be
r
of
pi
xe
ls
s
ki
ppe
d
f
o
r
th
e
ne
xt
c
onvolut
io
n.
I
f
s
tr
id
e
is
one
,
th
e
n
th
e
f
il
te
r
m
ove
s
one
pi
xe
l
a
nd
c
a
lc
ul
a
te
s
th
e
c
onvolut
io
na
l
out
put
.
I
f
s
tr
id
e
is
a
c
ons
id
e
r
a
bl
e
v
a
lu
e
,
th
e
n
th
e
c
onvolut
io
na
l
out
p
ut
c
om
pl
e
xi
ty
de
c
r
e
a
s
e
s
,
but
it
a
ls
o
a
f
f
e
c
ts
th
e
a
c
c
ur
a
c
y.
T
he
ge
ne
r
a
l
r
ul
e
s
ugge
s
te
d
i
s
to
ta
k
e
th
e
s
tr
id
e
'
s
v
a
l
ue
le
s
s
th
a
n
twi
c
e
th
e
f
il
te
r
s
iz
e
[
21
]
-
[
23]
.
A
pool
in
g l
a
ye
r
i
s
us
e
d f
or
t
he
dow
ns
a
m
pl
in
g a
nd i
s
a
ppl
ie
d
a
lo
ng t
he
s
pa
ti
a
l
di
m
e
ns
io
ns
. I
f
t
he
s
tr
id
e
i
s
2, t
he
n
th
e
r
e
s
ul
ti
ng
vol
um
e
is
−
2
×
−
2
×
.
T
h
e
f
ul
ly
c
onn
e
c
te
d
l
a
ye
r
c
om
put
e
s
th
e
va
li
d
c
la
s
s
s
c
or
e
.
I
n
th
is
la
ye
r
,
e
ve
r
y
ne
ur
on
is
c
onne
c
te
d
to
e
ve
r
y
ot
he
r
ne
ur
on
in
th
e
pr
e
vi
ous
la
ye
r
,
a
nd
th
e
out
put
s
iz
e
w
il
l
be
1
×
1
×
,
w
he
r
e
M
is
th
e
num
be
r
of
out
put
c
la
s
s
e
s
,
e
.g.,
f
or
th
e
h
a
ndw
r
it
te
n
di
gi
t
c
la
s
s
if
ic
a
ti
on
M
is
10
f
or
th
e
c
if
a
r
10 da
ta
s
e
t
M
i
s
10 a
nd f
or
c
if
a
r
100 da
ta
s
e
t
M
i
s
100.
T
he
hype
r
pa
r
a
m
e
te
r
c
a
ll
e
d f
il
te
r
s
iz
e
s
i
s
t
a
ke
n i
nt
o c
ons
id
e
r
a
ti
o
n i
n t
hi
s
s
tu
dy. A
f
il
te
r
i
s
a
2d
s
qua
r
e
m
a
tr
i
x
th
a
t
is
a
ppl
ie
d
to
e
ve
r
y
c
onvolut
io
na
l
la
ye
r
.
T
he
s
e
m
a
tr
ix
e
s
a
r
e
of
va
r
io
us
s
iz
e
s
li
ke
3
×
3
,
5
×
5
,
a
nd
7
×
7
.
W
e
va
r
ie
d
th
e
s
e
f
il
te
r
s
iz
e
s
ke
e
pi
ng
a
ll
th
e
ot
he
r
hype
r
pa
r
a
m
e
te
r
s
th
e
s
a
m
e
to
s
e
e
th
e
ir
e
f
f
e
c
t
on
th
e
a
c
c
ur
a
c
y a
nd t
o s
e
e
w
hi
c
h one
s
pe
r
f
or
m
t
he
be
s
t
a
m
ong whic
h
c
ir
c
um
s
ta
nc
e
s
.
2.
D
A
T
A
S
E
T
S
W
e
pe
r
f
or
m
e
d
th
e
e
xpe
r
im
e
nt
s
w
it
h
th
e
f
ol
lo
w
in
g
two
da
ta
s
e
ts
i
)
C
I
F
A
R
10
[
24]
a
nd
ii
)
F
a
s
hi
onM
N
I
S
T
[
25]
da
ta
s
e
t.
T
he
s
e
two
d
a
ta
s
e
ts
a
r
e
s
m
a
ll
a
n
d
pr
e
c
is
e
da
ta
s
e
ts
ha
vi
ng
lo
w
c
om
put
a
ti
ona
l
c
os
ts
.
T
he
r
e
s
ul
ts
dr
a
w
n
f
r
om
th
e
s
e
da
ta
s
e
ts
c
a
n
be
a
ppl
ie
d
t
o
m
os
t
of
th
e
da
ta
s
e
ts
.
S
o,
w
e
us
e
th
e
s
e
two
da
ta
s
e
ts
to
a
voi
d
th
e
c
om
put
a
ti
on
c
o
s
t
of
m
or
e
e
xt
e
ns
iv
e
da
ta
s
e
ts
.
C
I
F
A
R
10
is
th
e
da
ta
s
e
t
of
th
e
60000
im
a
ge
s
of
th
e
s
e
c
a
te
gor
ie
s
pl
a
ne
,
c
a
r
,
bi
r
d,
c
a
t,
d
e
e
r
,
dog,
f
r
og,
hor
s
e
,
a
nd
s
hi
p.
T
hi
s
da
ta
s
e
t
ha
s
10
out
put
c
la
s
s
e
s
.
T
he
s
e
c
ond
one
i
s
th
e
F
a
s
hi
onM
N
I
S
T
da
t
a
s
e
t
th
a
t
c
ont
a
in
s
10
di
f
f
e
r
e
nt
c
a
te
gor
ie
s
,
s
o
th
e
out
put
c
la
s
s
e
s
a
r
e
10.
A
ll
th
e
da
t
a
s
e
t
s
ha
ve
60000
im
a
ge
s
,
out
of
w
hi
c
h
w
e
us
e
40000
f
or
th
e
tr
a
in
in
g
a
nd
10000
f
or
th
e
va
li
da
ti
on,
a
nd
th
e
r
e
s
t
of
1000
f
or
te
s
ti
ng
th
e
a
c
c
ur
a
c
y
of
t
he
m
ode
l.
T
he
da
ta
s
e
t
is
s
pl
it
in
to
a
va
li
da
ti
on
da
ta
s
e
t
to
tr
a
in
our
m
ode
l
e
f
f
ic
ie
nt
ly
.
T
he
va
li
da
ti
on
da
ta
s
e
t
e
ns
ur
e
s
th
e
m
ode
l
doe
s
not
c
a
us
e
ove
r
f
it
ti
ng
or
unde
r
f
it
ti
ng on tr
a
in
in
g da
ta
a
nd pe
r
f
or
m
s
be
s
t
on t
e
s
t
da
ta
.
3.
A
R
C
H
I
T
E
C
T
U
R
E
F
or
bot
h
th
e
da
ta
s
e
ts
,
C
I
F
A
R
10
a
s
w
e
ll
a
s
M
N
I
S
T
da
ta
s
e
t,
w
e
bui
ld
a
C
N
N
m
ode
l
w
it
h
th
ir
te
e
n
la
ye
r
s
.
T
h
e
f
ir
s
t
la
y
e
r
is
th
e
in
put
la
ye
r
,
tr
a
c
ke
d
by
two
C
on
v2D
la
ye
r
s
.
I
n
e
a
c
h
la
ye
r
,
32
f
il
te
r
s
a
r
e
u
s
e
d.
T
he
n
ba
tc
h
nor
m
a
li
z
a
ti
on
is
us
e
d
f
or
s
ta
nda
r
di
z
in
g
th
e
in
put
s
f
ol
lo
w
e
d
by
th
e
M
a
xP
ool
in
g2D
la
ye
r
,
a
(
2,
2
)
pool
i
s
a
ppl
ie
d. T
he
dr
op
out
l
a
ye
r
i
s
us
e
d f
or
pr
o
te
c
ti
ng t
he
m
o
de
l
f
r
om
ove
r
f
it
ti
ng, a
nd t
he
n
th
e
F
la
tt
e
n l
a
ye
r
is
us
e
d t
o f
la
tt
e
n t
he
i
npu
ts
. T
w
o D
r
opout l
a
ye
r
s
a
r
e
us
e
d i
n
t
he
w
hol
e
m
ode
l,
a
nd t
he
n
a
t
th
e
s
e
nd, one
de
ns
e
la
ye
r
is
u
s
e
d. T
w
o
ty
pe
s
of
a
c
ti
va
ti
on
f
unc
ti
ons
a
r
e
u
s
e
d
.
O
n
e
i
s
R
e
L
U
, a
nd
a
t
th
e
e
nd,
S
of
tM
a
x
i
s
us
e
d
a
s
a
n
a
c
ti
va
ti
on
f
unc
ti
on.
A
da
m
opt
im
iz
e
r
is
e
m
pl
oye
d
f
or
opt
im
iz
in
g
th
e
m
ode
l
w
it
h
th
e
le
a
r
ni
ng
r
a
te
of
0.001,
a
nd
th
e
c
a
te
gor
ic
a
l
c
r
os
s
-
e
nt
r
opy
f
unc
ti
on
is
us
e
d
to
c
a
lc
ul
a
te
th
e
e
r
r
or
.
T
he
s
a
m
e
m
ode
l
is
us
e
d
w
it
h
a
ll
th
r
e
e
di
f
f
e
r
e
nt
f
il
te
r
s
iz
e
s
,
a
nd
th
e
e
f
f
e
c
t
on
th
e
a
c
c
ur
a
c
y
is
m
e
a
s
ur
e
d.
T
h
e
a
r
c
hi
te
c
tu
r
e
of
th
e
m
ode
ls
i
s
ke
pt
unc
ha
nge
d t
o t
r
a
c
k t
he
e
f
f
e
c
t
of
f
il
te
r
s
iz
e
s
on t
he
a
c
c
ur
a
c
y a
nd
lo
s
s
of
t
he
m
ode
l.
4.
E
X
P
E
R
I
M
E
N
T
A
L
R
E
S
U
L
T
S
F
or
th
e
e
xpe
r
im
e
nt
,
w
e
us
e
d
a
s
im
pl
e
c
onvolut
io
n
ne
ur
a
l
ne
tw
or
k
w
it
h
a
to
ta
l
o
f
13
la
ye
r
s
a
nd
ju
s
t
two
C
onv2D
la
ye
r
s
.
T
he
m
or
e
la
ye
r
s
w
e
us
e
in
C
N
N
,
th
e
m
or
e
a
c
c
ur
a
c
y
w
e
ge
t,
but
it
is
not
a
lwa
ys
tr
ue
th
e
m
ode
l
c
oul
d
r
un
in
to
ove
r
f
it
ti
ng.
A
ls
o,
th
e
c
om
put
a
ti
ona
l
c
os
t
i
nc
r
e
a
s
e
s
e
f
f
e
c
ti
ve
ly
w
it
h
e
a
c
h
l
a
ye
r
.
W
e
ha
v
e
to
f
in
d a
ba
la
nc
e
be
twe
e
n t
he
numb
e
r
of
l
a
ye
r
s
w
e
us
e
.
T
he
a
r
c
hi
te
c
tu
r
e
i
s
gi
ve
n i
n
F
ig
ur
e
1
.
I
n
th
e
f
ir
s
t
e
xpe
r
im
e
nt
,
bo
th
th
e
c
onvolut
io
na
l
la
ye
r
s
a
r
e
tr
a
in
e
d
w
it
h
a
3
×
3
f
il
te
r
,
a
nd
in
e
a
c
h
la
y
e
r
,
32
f
il
te
r
s
a
r
e
us
e
d.
I
n
th
e
s
e
c
ond
e
xp
e
r
im
e
nt
,
w
e
h
a
ve
us
e
d
th
e
s
a
m
e
a
r
c
hi
te
c
tu
r
e
th
e
num
be
r
of
th
e
f
il
te
r
s
is
ke
pt
unc
h
a
nge
d,
onl
y
th
e
f
il
te
r
s
i
z
e
s
a
r
e
c
ha
nge
d
to
5
×
5
,
a
nd
in
th
e
th
ir
d
e
xpe
r
im
e
nt
,
th
e
f
il
te
r
s
iz
e
w
a
s
a
lt
e
r
e
d t
o 7
×
7
,
a
nd t
he
r
e
s
t
of
t
he
ne
twor
k i
s
t
he
s
a
m
e
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
,
V
ol
.
10
, N
o.
4
,
D
e
c
e
m
be
r
202
1
:
872
-
878
874
F
ig
ur
e
1
.
A
r
c
hi
te
c
tu
r
e
of
C
N
N
T
a
bl
e
1
.
A
c
c
ur
a
c
y
of
th
e
C
I
F
A
R
10
da
ta
s
e
t
us
in
g
di
f
f
e
r
e
nt
f
il
te
r
s
iz
e
s
s
how
s
th
e
a
c
c
ur
a
c
y
of
th
e
tr
a
in
in
g,
va
li
da
ti
on,
a
nd
te
s
t
da
ta
s
e
t
f
or
C
I
F
A
R
10
u
s
in
g
di
f
f
e
r
e
nt
f
il
te
r
s
iz
e
s
,
a
nd
T
a
bl
e
2
s
how
s
th
e
lo
s
s
on
th
e
C
I
F
A
R
10
da
ta
s
e
t
us
in
g
di
f
f
e
r
e
nt
f
i
lt
e
r
s
iz
e
s
.
T
he
a
c
c
ur
a
c
y
on
th
e
C
I
F
A
R
10
da
ta
s
e
t
us
in
g
a
f
il
te
r
s
iz
e
of
3
×
3
is
hi
ghe
s
t,
i.
e
.,
94.26%
on
th
e
tr
a
in
in
g
da
ta
s
e
t,
72.75%
va
l
i
da
ti
on
da
ta
s
e
t,
a
nd
63.5%
on
th
e
te
s
t
da
ta
s
e
t,
a
nd t
he
l
ow
e
s
t
a
c
c
ur
a
c
y i
s
w
he
n
7
×
7
f
il
te
r
s
a
r
e
us
e
d. A
ll
t
he
m
ode
l
c
onf
ig
ur
a
ti
ons
a
r
e
t
r
a
in
e
d on 50
e
poc
hs
,
a
nd
th
e
num
be
r
of
f
il
te
r
s
us
e
d
is
32.
T
he
a
c
c
ur
a
c
y
a
nd
th
e
lo
s
s
dur
in
g
tr
a
in
in
g
a
nd
va
li
da
ti
on
a
r
e
s
how
n i
n F
ig
ur
e
s
2 a
nd 3, r
e
s
pe
c
ti
ve
ly
.
T
a
bl
e
1
.
A
c
c
ur
a
c
y of
t
he
C
I
F
A
R
10 da
ta
s
e
t
us
in
g
di
f
f
e
r
e
nt
f
il
te
r
s
iz
e
s
F
i
l
t
e
r
S
i
z
e
T
r
a
i
ni
ng D
a
t
a
V
a
l
i
da
t
i
on D
a
t
a
T
e
s
t
D
a
t
a
3
×
3
0.942625
0.7275
0.7304
5
×
5
0.923275
0.7261
0.7297
7
×
7
0.87725
0.7067
0.635
T
a
bl
e
2
.
L
os
s
o
f
th
e
C
I
F
A
R
10 da
t
a
s
e
t
di
f
f
e
r
e
nt
f
il
te
r
s
iz
e
s
F
i
l
t
e
r
S
i
z
e
T
r
a
i
ni
ng D
a
t
a
V
a
l
i
da
t
i
on D
a
t
a
T
e
s
t
D
a
t
a
3
×
3
0.1659
1.02
1.01
5
×
5
0.2209
0.97
0.97
7
×
7
0.3377
1.01
2.97
I
t
is
c
le
a
r
f
r
om
F
ig
ur
e
2
th
e
a
c
c
ur
a
c
y
is
in
c
r
e
a
s
in
g
a
s
th
e
e
poc
hs
in
c
r
e
a
s
e
,
but
th
e
r
a
te
of
in
c
r
e
a
s
e
de
c
r
e
a
s
e
s
a
nd
r
e
a
c
he
s
a
poi
nt
a
bove
w
hi
c
h
th
e
a
c
c
ur
a
c
y
doe
s
n'
t
c
ha
nge
.
T
he
a
c
c
ur
a
c
y
c
ur
ve
of
th
e
3
×
3
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
E
ff
e
c
t
of
f
il
te
r
s
iz
e
s
on i
m
ag
e
c
la
s
s
if
ic
at
io
n i
n C
N
N
…
(
O
w
ai
s
M
uj
ta
ba K
handay
)
875
f
il
te
r
is
a
bove
th
e
7
×
7
a
nd
5
×
5
f
il
te
r
.
T
he
va
li
da
ti
on
a
c
c
ur
a
c
y
lo
oks
th
e
s
a
m
e
a
s
tr
a
in
in
g
a
c
c
ur
a
c
y,
but
th
e
r
e
a
r
e
m
a
ny
s
pi
ke
s
,
but
le
s
s
e
r
f
il
te
r
s
iz
e
c
ur
ve
s
a
r
e
hi
ghe
r
t
ha
n
th
e
bi
gge
r
f
il
te
r
s
iz
e
s
.
T
a
bl
e
1
s
how
s
th
a
t
a
c
c
ur
a
c
y
is
th
e
hi
ghe
s
t
w
it
h
a
3
×
3
f
il
te
r
on
th
e
te
s
t
da
ta
.
S
o,
it
is
c
onc
lu
de
d
th
a
t
th
e
s
m
a
ll
e
r
f
il
te
r
s
iz
e
s
te
nd
to
gi
ve
be
tt
e
r
a
c
c
ur
a
c
y t
ha
n us
in
g l
a
r
ge
r
f
il
te
r
s
iz
e
s
.
F
ig
ur
e
2
. T
r
a
in
in
g a
c
c
ur
a
c
y a
nd va
li
da
ti
on a
c
c
ur
a
c
y us
in
g di
f
f
e
r
e
nt
f
il
te
r
s
iz
e
s
on t
he
C
I
F
A
R
10 da
ta
s
e
t
F
ig
ur
e
3
s
how
s
th
a
t
th
e
tr
a
in
in
g
lo
s
s
is
gi
ve
n
m
in
im
um
by
th
e
3
×
3
f
il
te
r
a
nd
th
e
hi
ghe
s
t
by
th
e
7
×
7
f
il
te
r
.
W
it
h
th
e
va
li
da
ti
on
lo
s
s
,
it
lo
ok
s
a
bi
t
di
f
f
e
r
e
nt
,
but
not
to
o
m
uc
h.
T
h
e
r
e
a
r
e
m
a
ny
s
pi
ke
s
,
but
if
w
e
s
e
e
th
e
ge
ne
r
a
li
ty
,
th
e
lo
s
s
de
c
r
e
a
s
e
s
a
s
w
e
us
e
th
e
s
m
a
ll
e
r
f
il
te
r
s
iz
e
a
nd
in
c
r
e
a
s
e
s
if
w
e
us
e
th
e
la
r
ge
r
f
il
te
r
s
iz
e
.
T
he
s
e
c
ond e
xpe
r
im
e
nt
i
s
done
on t
he
F
a
s
hi
onM
N
I
S
T
da
ta
s
e
t
w
it
h t
he
s
a
m
e
ne
twor
k c
onf
ig
ur
a
ti
on
a
s
done
on
th
e
C
I
F
A
R
10
da
ta
s
e
t.
T
hi
r
ty
-
two
f
il
te
r
s
w
it
h
th
r
e
e
di
f
f
e
r
e
nt
f
il
te
r
s
iz
e
s
a
r
e
us
e
d
in
bot
h
th
e
C
onv2D la
ye
r
s
, a
nd t
he
t
r
a
in
in
g, va
li
da
ti
on, a
nd t
e
s
t
a
c
c
ur
a
c
y a
nd l
os
s
a
r
e
m
e
a
s
ur
e
d.
F
ig
ur
e
3
. T
r
a
in
in
g l
os
s
a
nd
va
li
da
ti
on l
os
s
us
in
g di
f
f
e
r
e
nt
f
il
te
r
s
iz
e
s
on t
he
C
I
F
A
R
10 da
ta
s
e
t
T
a
bl
e
3
s
how
s
,
t
he
a
c
c
ur
a
c
y
of
th
e
tr
a
in
in
g
da
ta
s
e
t,
va
li
da
ti
on
da
ta
s
e
t,
a
nd
te
s
t
da
ta
s
e
t
is
m
a
xi
m
um
,
i.
e
., 92.68%
, 92.35, a
nd
92.68%
,
r
e
s
pe
c
ti
ve
ly
, w
he
n
3
×
3
f
il
te
r
s
a
r
e
us
e
d a
nd l
ow
e
s
t,
i
.e
., 91.1%
w
he
n
7
×
7
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
,
V
ol
.
10
, N
o.
4
,
D
e
c
e
m
be
r
202
1
:
872
-
878
876
f
il
te
r
s
a
r
e
us
e
d.
T
a
bl
e
4
s
how
s
t
he
l
os
s
of
t
he
t
r
a
in
in
g
da
ta
s
e
t,
v
a
li
da
ti
on da
ta
s
e
t
,
a
nd t
e
s
t
da
ta
s
e
t
is
m
a
xi
m
um
f
or
3
×
3
f
il
te
r
s
w
hi
le
in
th
e
lo
w
e
s
t
a
7
×
7
f
il
te
r
s
a
r
e
us
e
d.
A
c
c
ur
a
c
y
gr
a
phs
f
or
F
a
s
hi
onM
N
I
S
T
da
ts
e
t
is
pl
ot
te
d
in
F
ig
u
r
e
4
.
T
he
5
×
5
a
nd
3
×
3
c
ur
ve
s
a
r
e
a
lm
os
t
th
e
s
a
m
e
on
th
e
tr
a
in
in
g
da
ta
,
a
nd
th
e
ir
pe
r
f
or
m
a
nc
e
s
a
r
e
th
e
s
a
m
e
,
but
th
e
r
e
s
ul
t
is
s
e
e
n
in
th
e
va
li
d
a
ti
on
a
nd
te
s
t
da
ta
s
e
ts
w
he
r
e
th
e
3
×
3
f
il
te
r
out
pe
r
f
or
m
s
t
he
ot
he
r
f
il
te
r
s
.
T
a
bl
e
3
.
A
c
c
ur
a
c
y of
F
a
s
hi
onM
N
I
S
T
d
a
ta
s
e
t
u
s
in
g
di
f
f
e
r
e
nt
f
il
te
r
s
iz
e
s
F
i
l
t
e
r
s
i
z
e
T
r
a
i
ni
ng da
t
a
V
a
l
i
da
t
i
on da
t
a
T
e
s
t
da
t
a
3
×
3
0.929
0.9235
0.9268
5
×
5
0.926
0.9196
0.9264
7
×
7
0.918
0.910
0.911
T
a
bl
e
4
.
L
os
s
of
F
a
s
hi
on
M
N
I
S
T
da
ta
s
e
t
us
in
g
di
f
f
e
r
e
nt
f
il
te
r
s
iz
e
s
F
i
l
t
e
r
s
i
z
e
T
r
a
i
ni
ng da
t
a
V
a
l
i
da
t
i
on da
t
a
T
e
s
t
da
t
a
3
×
3
0.185
0.217
0.200
5
×
5
0.189
0.228
0.202
7
×
7
0.216
0.256
0.243
T
he
lo
s
s
c
ur
ve
f
or
F
a
s
hi
onM
N
I
S
T
da
ta
s
e
t
u
s
in
g
di
f
f
e
r
e
nt
f
il
te
r
s
iz
e
s
is
pl
ot
te
d
in
F
ig
ur
e
5
.
I
n
bot
h,
th
e
gr
a
phs
lo
s
s
i
s
m
in
im
um
w
he
n t
he
3
×
3
f
il
te
r
i
s
us
e
d a
nd ma
xi
m
u
m
w
he
n
7
×
7
is
us
e
d.
F
ig
ur
e
4
. T
r
a
in
in
g a
c
c
ur
a
c
y a
nd va
li
da
ti
on a
c
c
ur
a
c
y us
in
g di
f
f
e
r
e
nt
f
il
te
r
s
iz
e
s
on t
he
F
a
s
hi
onM
N
I
S
T
d
a
ta
s
e
t
F
ig
ur
e
5
. T
r
a
in
in
g l
os
s
a
nd va
li
da
ti
on l
os
s
us
in
g di
f
f
e
r
e
nt
f
il
te
r
s
iz
e
s
on t
he
F
a
s
hi
onM
N
I
S
T
da
t
a
s
e
t
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
E
ff
e
c
t
of
f
il
te
r
s
iz
e
s
on i
m
ag
e
c
la
s
s
if
ic
at
io
n i
n C
N
N
…
(
O
w
ai
s
M
uj
ta
ba K
handay
)
877
5.
C
O
N
C
L
U
S
I
O
N
I
n
th
is
pa
p
e
r
,
di
f
f
e
r
e
nt
C
N
N
s
tr
uc
tu
r
e
s
a
r
e
us
e
d
f
or
im
a
ge
c
la
s
s
if
ic
a
ti
on.
T
he
e
xp
e
r
im
e
nt
s
ha
ve
be
e
n
a
ppl
ie
d
on
C
F
I
R
10
a
nd
F
a
s
hi
on
M
N
I
S
T
da
t
a
s
e
t
s
,
a
nd
w
e
u
s
e
d
th
r
e
e
di
f
f
e
r
e
nt
f
il
te
r
s
s
iz
e
,
a
nd
th
e
num
be
r
of
f
il
te
r
s
is
ke
pt
c
ons
ta
nt
32.
F
r
om
th
e
e
xpe
r
im
e
nt
r
e
s
ul
ts
,
f
il
t
e
r
s
iz
e
ha
s
a
m
a
jo
r
e
f
f
e
c
t
on
c
la
s
s
if
ic
a
ti
on
a
c
c
ur
a
c
y,
th
e
a
c
c
ur
a
c
y
of
tr
a
in
in
g,
a
nd
v
a
li
da
ti
on.
A
c
c
ur
a
c
y
w
it
h
f
il
te
r
s
iz
e
3×
3
is
hi
ghe
r
th
a
n
5
×
5
a
nd
7
×
7,
but
th
e
tr
a
in
in
g
ti
m
e
is
hi
gh
w
he
n
a
s
m
a
ll
e
r
f
il
te
r
s
iz
e
is
us
e
d
.
B
e
s
id
e
s
,
w
he
n
f
il
te
r
s
iz
e
3
×
3
pi
x
e
ls
,
t
e
s
ti
ng
a
c
c
ur
a
c
y
is
b
e
tt
e
r
th
a
n
f
il
te
r
s
iz
e
5
×
5
pi
x
e
ls
a
nd
7
×
7
pi
xe
ls
. T
h
e
a
c
c
ur
a
c
y s
how
n
by
us
in
g a
3
×
3
f
i
lt
e
r
s
iz
e
in
bot
h
th
e
da
ta
s
e
ts
is
92.68%
in
F
a
s
hi
onM
N
I
S
T
a
nd
73.04%
in
th
e
C
I
F
A
R
10
da
ta
s
e
t.
T
he
onl
y
tr
a
de
of
us
in
g
le
s
s
f
il
te
r
s
iz
e
i
s
t
h
e
c
om
put
a
ti
ona
l
c
os
t
a
s
s
o
c
ia
te
d w
it
h t
he
m
o
de
l.
R
E
F
E
R
E
N
C
E
S
[1]
A.
Pal
and
D.
Singh,
“Handwritten
English
character
recog
nition
using
neural
network,”
Internati
onal
Journal
of
Computer
Scienc
e and Com
munication
, vol. 1, no. 2, 2010, pp. 141
-
144.
[2]
L.
Heutte,
T.
Paquet,
J.
V.
Moreau,
Y.
Lecourtier,
and
C.
Olivier,
“A
structural/statistical
feature
based
vector
fo
r
handwritten
character
recognition,”
Pattern
Recognition
Letters
,
vol.
19,
no.
7,
pp.
629
-
641,
1998
,
doi: 10.1016/S0167
-
8655(98)00039
-
7.
[3]
N
.
Kato,
M.
Suzuki,
S.
Omachi,
H.
Aso,
and
Y.
Nemoto,
“A
han
dwritten
character
recognition
system
using
directional
element
feature
and
asymmetric
Mahalanobis
distance,”
IEEE
Transactions
on
Pattern
Analysis
and
Machine Intell
igence
, vol. 21, no. 3, pp. 258
-
262, 1999, doi: 10.1109/34.754617.
[4]
T.
Chan,
K.
Jia,
S.
Gao,
J.
Lu,
Z.
Zeng,
and
Y.
Ma,
“PCANet:
A
simple
deep
learning
baseline
for
image
classifi
cation?,”
IEEE
Transactions
on
Image
Processing
,
v
ol.
24,
no.
12,
pp.
5017
-
5032,
2015,
doi: 10.1109/TIP.2015
.2475625.
[5]
W.
S.
Ahmed
and
A.
A.
A.
Karim,
“The
impact
of
filter
size
and
n
umber
of
filters
on
classification
accuracy
in
CNN,”
in
2020
International
Conferen
ce
on
Computer
Science
and
Software
Engineering
(CSASE)
,
2020,
pp. 88
-
93, doi: 10.1109/CSASE48920.
2020.9142089.
[6]
O.
M.
Khandy
and
S.
Dadvandipour,
“Analysis
of
machine
learning
algorithms
for
character
recognition:
a
case
study
on
handwritten
digit
recognition,”
Indonesian
Journal
Electric
al
Engineering
Computer
Science
(IJEECS)
,
vol. 21, no. 1, p. 574
-
581, 2021, doi: 10.11591/ijeecs.v21.i1.pp574
-
581.
[7]
O.
M.
Khanday
and
S.
Dadvandipour,
“Convolutional
neural
net
works
and
impact
of
filter
sizes
on
image
classifi
cation,”
Multidis
zciplinár
is Tudo
mányok
, vol. 10, no. 1, pp. 55
-
60, 2020, doi: 10.35925/j.multi.
2020.1.7
[8]
Y.
Li,
J.
Zeng,
S.
Shan,
and
X.
Chen,
“Occlusion
awar
e
facial
expr
ession
recognition
using
CNN
with
attention
mechanism,”
IEEE
Transactions
on
Image
Processing
,
vo
l.
28,
no.
5,
pp.
2439
-
2450,
2018,
doi: 10.1109/TIP.2018.2886767
[9]
S.
Xie
and
H.
Hu,
“Facial
expression
recognition
with
FRR‐CNN,”
El
ectronics
Letters
,
vol.
53,
no.
4,
pp.
235
-
237,
2017, doi: 10.1049/el.2016.4328.
[10]
H.
-
H. Zhao and H.
Liu, “Multi
ple
classifiers fusion
and CNN feature
extraction for handwritt
en digits
recogni
tion,”
Granular Comp
uting
, vol. 5, pp. 411
-
418, 2020, doi: 10.1007/s41066
-
019
-
00158
-
6.
[11]
Q.
Li,
W.
Cai,
X.
Wang,
Y.
Zhou,
D.
D.
Feng,
and
M.
Chen,
“Me
dical
image
classification
with
convolutional
neural
network,”
in
2014
13th
International
Conference
on
Cont
rol
Automation
Robotics
and
Vision
(ICARCV)
,
2014,
pp. 844
-
848, doi: 10.1109/ICARCV.2014.7064414.
[12]
S.
S.
Yadav
and
S.
M.
Jadhav,
“Deep
convolutional
neural
network
b
ased
medical
image
classification
for
disease
diagnosis,”
Journal
of Bi
g Data
, vol. 6, no. 1
13, 2019, doi: 10.1186/s40537
-
019
-
0276
-
2.
[13]
T.
Rahmat,
A.
Ismail,
and
S.
Aliman,
“Chest
x
-
ray
image
classificati
on
using
faster
R
-
CNN,”
Malaysian
Journal
of Computing
(MJoC),
vol. 4, no. 1, 2019, doi: 10.24191/mjoc.v4i1.6095.
[14]
A. A. Reshi
et al
., “An
efficie
nt CNN model for
COVID
-
19 disease de
tection based on X
-
ray image class
ification,
”
Complexit
y
, vol. 2021, 2021, doi: 10.1155/2021/6621607.
[15]
D.
S.
Maitra,
U.
Bhattacharya,
and
S.
K.
Parui,
“CNN
based
c
ommon
approach
to
handwritten
characte
r
recognition
of
mul
tiple
scripts,”
in
2015
13th
International
Confere
nce
on
Document
Analysis
and
Recognitio
n
(ICDAR)
, 2015, pp. 1021
-
1025, doi: 10.1109/ICDAR.2015.7333916.
[16]
M.
M.
Rahman,
M.
A.
H.
Akhand,
S.
Islam,
P.
Chandra
Shill,
and
M.
M.
Hafizur
Rahman,
“Bangla
handwritt
en
character
recogniti
on
using
convolut
ional
neural
network,”
Internati
onal
Journal
Image
Graphics
Signal
Processing
, vol. 7, no. 8, pp. 42
-
49, 2015, doi: 10.5815/ijigsp.2015.08.05.
[17]
L.
Chen,
S.
Wang,
W.
Fan,
J.
Sun,
and
S.
Naoi,
“Beyond
human
recognition:
A
CNN
-
based
framework
for
handwritten
character
recognition,”
2015
3rd
IAPR
Asian
Confere
nce
on
Pattern
Recognition
(ACPR,
)
2015
,
pp. 695
-
699, doi: 10.1109/ACPR.2015.7486592.
[18]
Z.
Zhong,
L.
Jin,
and
Z.
Xie,
“High
performance
offline
hand
written
Chinese
cha
racter
recognition
usin
g
GoogLeNet
and
directional
feature
maps,”
in
2015
13th
Internatio
nal
Conference
on
Document
Analysis
and
Recognition (ICDAR)
,
pp. 846
-
850, 2015. doi: 10.1109/ICDAR.2015.7333881.
[19]
Z.
Li,
N.
Teng,
M.
Jin,
and
H.
Lu,
“Building
efficient
CNN
architect
ure
for
offline
handwritten
Chinese
character
recognition
,”
Computer
Scienc
e
, 2018.
[20]
M. Hussain, J. J. Bird, an
d
D. R. Faria,
“A study on CNN
transfer le
ar
ning for image c
lassification,” in
Advances
in
Intelligen
t Systems a
nd Comput
in
, pp. 191
-
202, 2018.
[21]
J.
Hannink
et
al
.,
“Mobile
stride
length
estimation
with
deep
conv
olutional
neural
networks,”
IEEE
Journal
of
Biomedical and Health Informatics,
vol. 22, no. 2, pp. 354
-
362, 2018, doi: 10.1109/JBHI.2017.2679486.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
,
V
ol
.
10
, N
o.
4
,
D
e
c
e
m
be
r
202
1
:
872
-
878
878
[22]
S.
Albawi,
T.
A.
Mohammed,
and
S
.
Al
-
Zawi,
“Understanding
of
a
convolutional
neural
network,”
Internati
onal
Conferenc
e on Eng
ineering
and Tech
nology (
ICET),
2017,
pp. 1
-
6, doi: 10.1109/ICEngTechnol.2017.8308186.
[23]
S.
Lai,
L.
Jin,
and
W.
Yang,
“Toward
high
-
performance
online
HCCR:
a
CNN
app
roach
with
DropDistortion,
path
signature
and
spatial
stochastic
max
-
pooling,”
Pattern
Recogni
tion
Letters
,
vol.
89,
pp.
60
-
66,
2017,
doi: 10.1016/j.patrec.2017.02.011.
[24]
“CIFAR
-
10
and
CIFAR
-
100
datasets,”
Toronto.edu.
[Online].
Available:
https://www.cs.toronto.edu/~
kriz/cifar.html . (Accesse
d: Jan, 5, 2021).
[25]
Zalando
Research,
“Fashion
MNIST,”
[Online].
Available:
https://ww
w.kaggle.com/zalando
-
research/fashionm
nist
(Accessed: Jan,
5, 2021).
B
I
O
G
R
A
P
H
I
E
S
O
F
A
U
T
H
O
R
S
Owais
Mujtaba
Khand
ay
received
his
B.Sc.
(I.T)
from
the
Uni
versity
of
Kashmir
(S.P
College
Srinagar
)
and
M.Sc.
(Computer
Scienc
e)
from
the
Univer
sity
of
Pondiche
rry.
Currentl
y,
he
is
a
Ph.D
.
Student
at
the
University
of
Miskolc,
Hungar
y,
under
the
Stip
endium
Hungaricum Scholarshi
p program. aitowais@
uni
-
miskolc.hu +
36704202865
Samad
Dadvandipour
,
Associate
Professor,
Institute
of
Information
Sciences,
University
o
f
Miskolc, 3515 Hungar
y
dr.samad@
uni
-
miskolc.hu
Evaluation Warning : The document was created with Spire.PDF for Python.