I
nte
rna
t
io
na
l J
o
urna
l o
f
Rec
o
nfig
ura
ble a
nd
E
m
be
dd
e
d Sy
s
t
e
m
s
(
I
J
R
E
S)
Vo
l.
14
,
No
.
2
,
J
u
ly
20
25
,
p
p
.
339
~
3
5
2
I
SS
N:
2089
-
4864
,
DOI
:
1
0
.
1
1
5
9
1
/i
j
r
es
.
v
1
4
.
i
2
.
pp
3
3
9
-
352
339
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
r
es.ia
esco
r
e.
co
m
Sy
ste
m
a
t
ic revie
w
o
f
a
lightw
eig
ht
conv
o
lutiona
l ne
ura
l net
w
o
rk
a
rchitec
t
ures o
n
e
dg
e devices
M
uh
a
mm
a
d Ab
ba
s
Abu
T
a
l
ib
1
,
Sa
m
s
u
l Set
u
m
i
n
1
,
Siti
J
u
lia
na
Abu B
a
k
a
r
1
,
Adi
I
zha
r
Che
Ani
1
,
Deni
s
E
k
a
Ca
hy
a
ni
2
1
C
e
n
t
r
e
f
o
r
El
e
c
t
r
i
c
a
l
En
g
i
n
e
e
r
i
n
g
S
t
u
d
i
e
s,
U
n
i
v
e
r
si
t
i
T
e
k
n
o
l
o
g
i
M
A
R
A
C
a
w
a
n
g
a
n
P
u
l
a
u
P
i
n
a
n
g
,
P
e
r
mat
a
n
g
P
a
u
h
,
M
a
l
a
y
si
a
2
D
e
p
a
r
t
me
n
t
o
f
M
a
t
h
e
mat
i
c
s
,
F
a
c
u
l
t
y
o
f
M
a
t
h
e
mat
i
c
s
a
n
d
N
a
t
u
r
a
l
S
c
i
e
n
c
e
,
U
n
i
v
e
r
si
t
a
s
N
e
g
e
r
i
M
a
l
a
n
g
,
M
a
l
a
n
g
,
I
n
d
o
n
e
si
a
Art
icle
I
nfo
AB
ST
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
No
v
1
5
,
2
0
2
4
R
ev
i
s
ed
Ma
y
8
,
2
0
2
5
A
cc
ep
ted
J
u
n
1
0
,
2
0
2
5
A
li
g
h
t
w
e
i
g
h
t
c
o
n
v
o
lu
t
io
n
a
l
n
e
u
r
a
l
n
e
tw
o
rk
(CNN
)
h
a
s
b
e
c
o
m
e
o
n
e
o
f
th
e
m
a
jo
r
stu
d
ies
in
m
a
c
h
in
e
lea
rn
in
g
f
ield
t
o
o
p
ti
m
ize
it
s
p
o
t
e
n
ti
a
l
f
o
r
e
m
p
lo
y
in
g
it
o
n
th
e
re
so
u
rc
e
-
c
o
n
s
train
e
d
d
e
v
ice
s.
Ho
w
e
v
e
r,
a
b
e
n
c
h
m
a
r
k
f
o
r
f
a
ir
c
o
m
p
a
riso
n
is
sti
ll
m
issin
g
a
n
d
t
h
u
s,
th
is
p
a
p
e
r
a
im
s
to
id
e
n
ti
fy
th
e
re
c
e
n
t
s
tu
d
ies
re
g
a
rd
in
g
th
e
li
g
h
tw
e
ig
h
t
CNN
a
rc
h
it
e
c
tu
re
s
in
c
lu
d
i
n
g
th
e
ty
p
e
s
o
f
CNN
,
it
s
a
p
p
li
c
a
ti
o
n
s,
e
d
g
e
d
e
v
ice
s
u
sa
g
e
,
e
v
a
lu
a
ti
o
n
t
y
p
e
s
a
n
d
m
a
tri
c
e
s,
a
n
d
p
e
rf
o
r
m
a
n
c
e
c
o
m
p
a
riso
n
.
T
h
e
p
re
f
e
rre
d
re
p
o
rti
n
g
it
e
m
s
f
o
r
s
y
ste
m
a
ti
c
re
v
ie
ws
a
n
d
m
e
ta
-
a
n
a
l
y
sis
(P
RIS
M
A
)
f
ra
m
e
w
o
rk
w
a
s
u
se
d
a
s
th
e
m
a
in
a
p
p
ro
a
c
h
t
o
c
o
l
lec
t
a
n
d
i
n
terp
re
t
t
h
e
li
tera
tu
re
.
In
th
e
p
r
o
c
e
ss
,
3
7
p
a
p
e
rs
w
e
re
id
e
n
ti
f
ied
a
s
m
e
e
ti
n
g
th
e
c
rit
e
ria
f
o
r
li
g
h
tw
e
ig
h
t
CNN
s
a
im
e
d
a
t
im
a
g
e
c
la
ss
i
f
ica
ti
o
n
o
r
re
g
re
ss
io
n
tas
k
s.
Of
th
e
se
,
o
n
ly
2
0
st
u
d
ie
s
e
x
p
lo
re
d
th
e
u
se
o
f
th
e
se
m
o
d
e
ls
o
n
e
d
g
e
d
e
v
ice
s.
T
o
c
o
n
c
lu
d
e
,
M
o
b
il
e
Ne
t
a
p
p
e
a
re
d
a
s
th
e
m
o
st
u
se
d
a
rc
h
it
e
c
tu
re
,
wh
il
e
t
h
e
ty
p
e
s
o
f
CNN
f
o
c
u
se
d
o
n
im
a
g
e
c
las
si
f
ica
ti
o
n
f
o
r
th
e
g
e
n
e
ra
l
-
p
u
rp
o
se
a
p
p
li
c
a
ti
o
n
.
F
o
ll
o
w
in
g
th
a
t,
t
h
e
NV
IDIA
Je
tso
n
Na
n
o
w
a
s
th
e
m
o
st
u
ti
li
z
e
d
e
d
g
e
d
e
v
ice
in
re
c
e
n
t
re
se
a
rc
h
.
A
d
d
it
io
n
a
ll
y
,
p
e
rf
o
rm
a
n
c
e
e
v
a
l
u
a
ti
o
n
c
o
m
m
o
n
ly
in
c
lu
d
e
d
m
e
a
su
re
s
li
k
e
a
c
c
u
ra
c
y
a
n
d
ti
m
e
,
a
lo
n
g
w
it
h
m
e
tri
c
s
su
c
h
a
s
re
c
a
ll
,
p
re
c
isio
n
,
F
1
-
S
c
o
re
,
a
n
d
o
t
h
e
r
sim
il
a
r
in
d
ica
to
rs.
F
in
a
ll
y
,
th
e
a
v
e
ra
g
e
a
c
c
u
ra
c
y
f
o
r
p
e
rf
o
r
m
a
n
c
e
c
o
m
p
a
riso
n
c
a
n
se
rv
e
a
s
th
re
sh
o
l
d
v
a
lu
e
f
o
r
f
u
tu
re
re
se
a
rc
h
in
th
i
s
sc
o
p
e
o
f
stu
d
y
.
K
ey
w
o
r
d
s
:
E
d
g
e
d
ev
ice
I
m
ag
e
cla
s
s
i
f
icatio
n
L
i
g
h
t
w
ei
g
h
t
co
n
v
o
l
u
tio
n
a
l
n
eu
r
al
n
et
w
o
r
k
R
eso
u
r
ce
-
co
n
s
tr
ain
ed
S
y
s
te
m
a
tic
liter
at
u
r
e
r
ev
ie
w
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
:
Sa
m
s
u
l
Setu
m
i
n
C
en
tr
e
f
o
r
E
lectr
ical
E
n
g
i
n
ee
r
in
g
St
u
d
ies,
Un
i
v
er
s
i
ti T
ek
n
o
l
o
g
i M
A
R
A
C
a
w
a
n
g
a
n
P
u
la
u
P
in
an
g
P
er
m
ata
n
g
P
au
h
,
1
3
5
0
0
P
u
lau
P
in
an
g
,
Ma
la
y
s
ia
E
m
ail: sa
m
s
u
ls
@
u
it
m
.
ed
u
.
m
y
1.
I
NT
RO
D
UCT
I
O
N
R
esear
ch
er
s
th
e
s
e
d
a
y
s
ar
e
p
r
i
m
ar
il
y
f
o
c
u
s
ed
o
n
t
h
e
ad
v
a
n
ce
m
e
n
t
o
f
ar
tific
ial
i
n
telli
g
e
n
ce
(
A
I
)
tech
n
o
lo
g
y
in
o
r
d
er
to
en
h
a
n
c
e
s
o
ciet
y
's
q
u
alit
y
o
f
li
f
e
an
d
f
ac
ilit
ate
th
e
in
d
u
s
tr
ial
r
e
v
o
lu
t
io
n
.
A
lt
h
o
u
g
h
t
h
i
s
d
is
cip
lin
e
w
a
s
in
tr
o
d
u
ce
d
b
ac
k
in
t
h
e
1
9
5
0
s
,
it
h
as
g
o
n
e
th
r
o
u
g
h
r
ap
id
d
ev
elo
p
m
e
n
t
i
n
th
e
p
as
t
d
ec
ad
es,
w
h
ic
h
h
a
s
co
v
er
ed
b
o
th
i
n
s
id
e
an
d
o
u
ts
id
e
o
f
th
e
co
m
p
u
te
r
s
cien
ce
f
ield
[
1
]
,
[
2
]
.
I
t
ca
n
b
e
s
ee
n
t
h
at
m
an
y
tech
n
o
lo
g
ies
a
n
d
n
o
n
-
tec
h
n
o
lo
g
y
-
b
ased
j
o
u
r
n
als
h
a
v
e
p
u
b
lis
h
ed
ar
ticle
s
r
elate
d
to
AI
[
1
]
,
[
2
]
.
A
I
h
a
s
p
r
o
g
r
ess
ed
f
r
o
m
s
i
m
p
le
r
u
le
-
b
ased
s
y
s
te
m
s
to
m
o
r
e
co
m
p
licated
alg
o
r
ith
m
s
t
h
at
ca
n
m
ak
e
a
u
to
n
o
m
o
u
s
d
ec
is
io
n
s
a
n
d
s
o
lv
e
p
r
o
b
lem
s
.
T
h
e
p
r
im
ar
y
id
ea
u
n
d
er
l
y
i
n
g
A
I
is
to
d
ev
elo
p
s
y
s
te
m
s
ca
p
ab
le
o
f
d
o
in
g
ac
tiv
itie
s
t
h
at
w
o
u
ld
n
o
r
m
all
y
n
ee
d
h
u
m
an
in
tellect,
s
u
c
h
as
v
is
u
al
p
er
ce
p
tio
n
,
s
p
ee
ch
r
ec
o
g
n
itio
n
,
d
ec
is
io
n
-
m
ak
in
g
,
a
n
d
lan
g
u
a
g
e
tr
a
n
s
lat
io
n
[
1
]
,
[
2
]
.
Fig
u
r
e
1
s
h
o
w
s
th
e
in
ter
-
r
elatio
n
o
f
d
ata
s
cie
n
ce
to
ar
tif
icial
n
e
u
r
al
n
et
w
o
r
k
t
h
r
o
u
g
h
A
I
,
m
ac
h
i
n
e
lear
n
in
g
(
M
L
)
,
an
d
d
ee
p
lear
n
in
g
(
D
L
)
[
1
]
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
9
-
4864
I
n
t J
R
ec
o
n
f
i
g
u
r
ab
le
&
E
m
b
ed
d
ed
Sy
s
t
,
Vo
l.
14
,
No
.
2
,
J
u
l
y
20
25
:
339
-
3
5
2
340
Fu
r
t
h
er
m
o
r
e,
ML
is
o
n
e
o
f
t
h
e
m
o
s
t
co
m
m
o
n
s
u
b
f
ield
s
i
n
A
I
,
w
h
er
e
it
tak
e
s
a
d
if
f
er
e
n
t
ap
p
r
o
ac
h
f
r
o
m
a
c
lass
ical
p
r
o
g
r
a
m
m
i
n
g
m
et
h
o
d
.
So
,
in
s
tead
o
f
u
s
in
g
an
al
g
o
r
ith
m
f
o
r
a
s
p
ec
i
f
ic
p
r
o
b
lem
o
r
f
u
n
ct
io
n
,
ML
u
s
e
a
ce
r
tain
d
ataset
f
o
r
its
alg
o
r
ith
m
to
lear
n
,
p
r
ed
ict,
a
n
d
d
ec
id
e
th
e
o
u
tco
m
e
[
1
]
,
[
2
]
.
Prim
ar
i
l
y
,
M
L
is
t
y
p
icall
y
ca
teg
o
r
ized
in
to
f
o
u
r
m
aj
o
r
ty
p
es
s
u
c
h
as
s
u
p
er
v
i
s
ed
lear
n
in
g
,
w
h
ic
h
in
v
o
l
v
es
tr
ain
i
n
g
m
o
d
els
o
n
lab
elled
d
ata,
u
n
s
u
p
er
v
is
ed
le
ar
n
in
g
,
w
h
ic
h
i
n
v
o
l
v
es
s
ea
r
ch
i
n
g
f
o
r
p
atter
n
s
i
n
u
n
lab
eled
d
ata,
s
e
m
i s
u
p
er
v
i
s
ed
lear
n
in
g
,
w
h
ic
h
i
n
cl
u
d
es
b
o
t
h
s
u
p
er
v
i
s
ed
an
d
u
n
s
u
p
er
v
is
ed
lear
n
in
g
,
a
n
d
r
ein
f
o
r
ce
m
e
n
t
lear
n
in
g
,
w
h
ic
h
teac
h
es
m
o
d
els
to
m
a
k
e
d
ec
i
s
io
n
s
b
ased
o
n
tr
ial
a
n
d
er
r
o
r
[
1
]
,
[
2
]
.
I
n
ad
d
itio
n
,
th
e
ap
p
licatio
n
o
f
M
L
is
co
m
m
o
n
l
y
d
iv
id
ed
in
to
o
b
j
e
ct
class
if
icatio
n
o
r
r
eg
r
ess
io
n
(
i.e
.
,
p
r
ed
ictio
n
)
.
So
m
e
t
y
p
ical
ex
a
m
p
les
o
f
alg
o
r
ith
m
s
i
n
ML
in
cl
u
d
e
ar
tif
icia
l
n
eu
r
al
n
et
w
o
r
k
,
d
ec
is
io
n
tr
ee
s
,
lin
ea
r
r
eg
r
ess
io
n
,
an
d
s
u
p
p
o
r
t
v
ec
to
r
m
ac
h
in
e
[
1
]
,
[
2
]
.
Mo
r
eo
v
er
,
co
n
v
o
lu
tio
n
al
n
e
u
r
al
n
et
w
o
r
k
(
C
NN)
is
o
n
e
o
f
ML
’
s
ar
ti
f
ic
ial
n
eu
r
al
n
et
w
o
r
k
alg
o
r
ith
m
s
th
at
i
s
s
p
ec
ial
ized
f
o
r
i
m
ag
e
-
b
ased
tas
k
s
[
1
]
,
[
3
]
-
[
6
]
.
I
n
o
th
e
r
w
o
r
d
s
,
C
N
N
is
f
u
n
d
a
m
e
n
tal
in
m
a
n
y
co
m
p
u
ter
v
is
io
n
task
s
s
u
c
h
as
i
m
a
g
e
d
etec
tio
n
,
r
ec
o
g
n
itio
n
,
clas
s
i
f
icat
io
n
,
r
eg
r
ess
io
n
,
an
d
s
eg
m
e
n
tat
io
n
.
C
NN
is
m
ad
e
u
p
o
f
t
h
r
ee
m
ain
la
y
er
s
,
in
cl
u
d
in
g
co
n
v
o
l
u
tio
n
a
l,
p
o
o
lin
g
,
an
d
f
u
ll
y
co
n
n
ec
ted
la
y
er
s
[
1
]
,
[
3
]
-
[
6
]
.
Fig
u
r
e
2
d
ep
icts
th
e
b
asic
C
NN
ar
ch
it
ec
tu
r
e
an
d
its
tr
ain
i
n
g
p
r
o
ce
s
s
[
1
]
,
[
3
]
-
[
6
]
.
First,
co
n
v
o
lu
t
io
n
al
la
y
er
s
ap
p
l
y
f
ilter
s
to
in
co
m
in
g
d
ata,
ca
p
tu
r
in
g
s
p
atial
h
ier
ar
ch
ies
a
n
d
lo
ca
l
p
atter
n
s
t
h
at
ar
e
n
ec
e
s
s
ar
y
f
o
r
ap
p
licatio
n
s
s
u
c
h
as
i
m
a
g
e
id
en
ti
f
icatio
n
[
1
]
,
[
3
]
-
[
6
]
.
Seco
n
d
,
p
o
o
lin
g
la
y
er
s
lo
w
er
s
t
h
e
d
i
m
en
s
i
o
n
alit
y
o
f
t
h
e
d
ata,
in
cr
ea
s
i
n
g
co
m
p
u
ti
n
g
e
f
f
icie
n
c
y
a
n
d
r
esil
ie
n
ce
[
1
]
,
[
3
]
-
[
6
]
.
T
h
ir
d
,
f
u
ll
y
co
n
n
ec
ted
la
y
er
s
m
a
k
e
h
i
g
h
-
le
v
el
d
ec
is
io
n
s
b
ased
o
n
th
e
ex
tr
ac
ted
ch
ar
ac
ter
is
tics
[
1
]
,
[
3
]
-
[
6
]
.
Si
m
p
l
y
p
u
t,
th
e
f
ir
s
t
t
w
o
m
ain
la
y
er
s
p
er
f
o
r
m
f
ea
t
u
r
e
ex
tr
ac
tio
n
f
r
o
m
t
h
e
i
n
p
u
t d
ata
a
n
d
th
e
th
ir
d
m
ai
n
la
y
e
r
m
ap
s
t
h
e
e
x
tr
ac
ted
f
ea
t
u
r
es
t
o
d
e
cid
e
o
r
p
r
ed
ict
th
e
o
u
tp
u
t d
ata
[
1
]
,
[
3
]
-
[
6
]
.
Nev
er
th
e
less
,
t
h
e
d
ev
elo
p
m
e
n
t
o
f
C
NN
’
s
ap
p
licatio
n
s
u
s
u
a
ll
y
i
n
v
o
l
v
es
w
it
h
b
ig
d
ata
w
h
ich
r
elies
h
ea
v
i
l
y
o
n
clo
u
d
i
n
f
r
astr
u
ctu
r
e
an
d
r
eso
u
r
ce
s
f
o
r
h
i
g
h
co
m
p
u
tat
io
n
co
m
p
le
x
it
y
,
m
e
m
o
r
y
an
d
lo
ad
p
o
w
e
r
co
n
s
u
m
p
tio
n
[
4
]
-
[
8
]
.
I
n
r
ec
e
n
t
y
ea
r
s
,
t
h
e
w
id
e
s
p
r
ea
d
u
s
e
o
f
clo
u
d
co
m
p
u
t
in
g
i
n
m
a
n
y
f
ield
s
o
f
C
NN
’
s
ap
p
licatio
n
s
h
as
r
aised
s
o
m
e
co
n
ce
r
n
r
eg
ar
d
in
g
s
tr
ict
laten
c
y
r
eq
u
ir
em
e
n
ts
,
s
tr
ai
n
ed
n
et
w
o
r
k
ca
p
ac
it
y
,
as
w
el
l
as
p
r
iv
ac
y
a
n
d
s
ec
u
r
it
y
is
s
u
e
s
[
4
]
-
[
8
]
.
Ultim
a
tel
y
,
i
n
o
r
d
er
to
o
v
er
co
m
e
t
h
e
s
e
p
r
o
b
lem
s
a
n
d
o
p
ti
m
ize
C
NN
’
s
ap
p
licatio
n
s
,
an
i
n
cr
ea
s
i
n
g
d
em
an
d
f
o
r
d
ep
lo
y
i
n
g
DL
m
o
d
e
ls
d
ir
ec
tl
y
o
n
to
ed
g
e
d
ev
ice
s
to
en
ab
le
r
ea
l
-
ti
m
e
in
f
er
en
ce
an
d
d
ec
is
io
n
-
m
a
k
i
n
g
h
as
b
ee
n
in
tr
o
d
u
ce
d
.
E
d
g
e
c
o
m
p
u
ti
n
g
i
n
clu
d
e
s
p
r
o
ce
s
s
in
g
d
ata
at
o
r
n
ea
r
th
e
s
o
u
r
ce
o
f
d
ata
cr
ea
tio
n
w
h
ic
h
is
ca
lled
ed
g
e
d
ev
ice
s
,
s
u
ch
as
I
o
T
d
ev
ices,
s
m
ar
tp
h
o
n
e
s
,
o
r
s
en
s
o
r
s
,
r
ath
er
th
an
u
s
in
g
a
ce
n
tr
alize
d
clo
u
d
in
f
r
astru
ct
u
r
e.
Fig
u
r
e
1
.
T
h
e
in
ter
-
r
elatio
n
o
f
d
ata
s
cien
ce
to
ar
tif
icia
l n
e
u
r
al
n
et
w
o
r
k
t
h
r
o
u
g
h
A
I
,
ML
,
a
n
d
DL
[
1
]
,
[
2
]
Fig
u
r
e
2
.
T
h
e
b
asic CNN a
r
ch
itectu
r
e
an
d
it
s
tr
ain
i
n
g
p
r
o
ce
s
s
[
1
]
,
[
2
]
F
i
g
.
1
.
Th
e
b
a
si
c
C
N
N
a
r
c
h
i
t
e
c
t
u
r
e
a
n
d
i
t
s
t
r
a
i
n
i
n
g
p
r
o
c
e
ss
[
1
]
,
[
2
]
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
R
ec
o
n
f
i
g
u
r
ab
le
&
E
m
b
ed
d
ed
Sy
s
t
I
SS
N:
2089
-
4864
S
ystema
tic
r
ev
iew
o
f a
lig
h
tw
e
ig
h
t c
o
n
v
o
lu
tio
n
a
l
n
eu
r
a
l n
et
w
o
r
k
…
(
Mu
h
a
mma
d
A
b
b
a
s
A
b
u
Ta
lib
)
341
Ma
n
y
r
esear
ch
er
s
h
a
v
e
e
m
p
lo
y
ed
ed
g
e
co
m
p
u
ti
n
g
to
g
ain
v
ar
io
u
s
b
en
e
f
its
,
in
cl
u
d
i
n
g
s
h
o
r
ter
laten
c
y
,
lo
w
er
b
an
d
w
id
th
u
s
e,
an
d
g
r
ea
ter
p
r
iv
ac
y
an
d
s
ec
u
r
it
y
[
4]
-
[
8
]
.
Fo
r
th
at,
ed
g
e
d
ev
i
ce
s
eq
u
ip
p
ed
w
i
th
s
tr
o
n
g
ce
n
tr
al
p
r
o
ce
s
s
in
g
u
n
it
s
(
C
P
Us)
an
d
s
p
ec
ialized
h
ar
d
w
ar
e
ac
ce
ler
ato
r
s
lik
e
g
r
ap
h
ic
al
p
r
o
ce
s
s
in
g
u
n
it
s
(
GP
Us),
ten
s
ile
p
r
o
ce
s
s
in
g
u
n
i
ts
(
T
P
Us),
an
d
n
eu
r
al
p
r
o
ce
s
s
i
n
g
u
n
it
s
(
NP
Us)
ca
n
ex
ec
u
te
c
o
m
p
lica
ted
A
I
an
d
ML
m
o
d
els
lo
ca
ll
y
.
Ho
w
ev
er
,
o
p
tim
izi
n
g
C
NN
ar
ch
itect
u
r
e’
s
ef
f
icien
c
y
f
o
r
ed
g
e
d
ev
ices
d
ep
lo
y
m
e
n
t
p
o
s
es
a
cr
itical
ch
alle
n
g
e
s
i
n
ce
t
h
eir
a
p
p
licatio
n
s
v
ar
y
f
r
o
m
ea
c
h
o
th
er
an
d
d
u
e
to
th
e
l
i
m
ited
co
m
p
u
tatio
n
a
l r
eso
u
r
ce
s
an
d
p
o
w
er
co
n
s
tr
ai
n
t
s
o
f
th
e
ed
g
e
d
ev
ices.
T
o
d
ate,
m
o
s
t
s
t
u
d
y
in
t
h
is
s
u
b
j
ec
t
h
av
e
u
s
ed
d
if
f
er
e
n
t
m
et
h
o
d
o
f
o
p
tim
izatio
n
to
p
r
o
d
u
ce
th
eir
lig
h
t
w
ei
g
h
t
C
NN
f
o
r
ed
g
e
d
ev
ice
d
ep
lo
y
m
e
n
t
an
d
t
h
e
b
en
ch
m
ar
k
f
o
r
a
f
air
co
m
p
ar
is
o
n
is
s
till
m
is
s
i
n
g
[
8
]
-
[
1
2
]
.
No
w
,
t
h
i
s
s
y
s
te
m
s
l
iter
atu
r
e
r
ev
ie
w
(
S
L
R
)
ai
m
s
to
co
llec
t,
an
al
y
ze
,
a
n
d
in
ter
p
r
et
th
e
cu
r
r
en
t
o
r
r
ec
en
tl
y
p
u
b
lis
h
ed
ar
ticles
o
n
th
e
li
g
h
t
w
ei
g
h
t
C
NN
ar
ch
ite
ctu
r
es
f
o
r
ed
g
e
d
ev
ice
s
a
n
d
c
ateg
o
r
ized
th
e
m
i
n
ter
m
s
o
f
t
h
e
s
p
ec
if
ic
ar
ch
ite
ctu
r
e
o
r
b
ased
-
m
o
d
el,
C
NN’
s
t
y
p
es
(
i.e
.
,
class
i
f
icatio
n
o
r
r
eg
r
ess
io
n
)
,
an
d
ap
p
licatio
n
s
(
i.e
.
,
th
e
s
u
b
j
ec
t
o
r
p
u
r
p
o
s
e
o
f
ea
ch
C
N
N’
s
ar
c
h
itect
u
r
es)
u
s
ed
f
o
r
t
h
eir
r
esea
r
ch
.
Nex
t,
b
ased
o
n
th
e
t
y
p
es
o
f
ed
g
e
d
ev
ice,
th
e
ev
alu
a
tio
n
cr
iter
ia
(
e.
g
.
,
ti
m
e
an
d
ac
cu
r
ac
y
)
,
ev
alu
a
tio
n
m
atr
i
ce
s
(
e.
g
.
,
ac
cu
r
ac
y
,
p
r
ec
is
io
n
,
F1
-
Sco
r
e
,
an
d
r
o
o
t m
ea
n
s
q
u
ar
ed
er
r
o
r
(
R
MSE
)
)
,
an
d
p
er
f
o
r
m
a
n
ce
co
m
p
ar
is
o
n
b
et
w
ee
n
ea
ch
s
t
u
d
y
in
th
e
f
ir
s
t
q
u
esti
o
n
w
ill
b
e
r
ec
o
r
d
e
d
an
d
d
is
cu
s
s
ed
.
I
n
s
h
o
r
t,
th
e
r
esear
ch
q
u
es
tio
n
s
in
th
i
s
p
ap
er
ar
e
as
f
o
llo
w
s
a
n
d
ca
n
b
e
s
ee
n
as
il
lu
s
tr
ated
in
Fi
g
u
r
e
3
:
−
R
Q1
:
W
h
at
is
th
e
c
u
r
r
en
t
li
g
h
t
w
ei
g
h
t
C
N
N
ar
ch
itect
u
r
es
u
s
ed
o
n
th
e
li
m
ited
co
m
p
u
ta
tio
n
al
r
eso
u
r
ce
s
o
r
ed
g
e
d
ev
ices?
−
R
Q2
:
W
h
at
is
t
h
e
c
u
r
r
en
t
p
er
f
o
r
m
a
n
ce
o
f
t
h
e
li
g
h
t
w
ei
g
h
t
C
N
N
ar
ch
itec
tu
r
e
u
s
ed
o
n
th
e
li
m
ited
co
m
p
u
tatio
n
al
r
eso
u
r
c
es o
r
ed
g
e
d
ev
ices?
T
h
is
SLR
i
s
d
iv
id
ed
i
n
to
f
o
u
r
m
aj
o
r
s
ec
tio
n
s
.
I
n
t
h
e
f
ir
s
t
s
ec
tio
n
,
th
e
in
tr
o
d
u
ctio
n
i
s
r
ev
ie
w
ed
r
eg
ar
d
in
g
th
e
b
ac
k
g
r
o
u
n
d
k
n
o
w
led
g
e
o
f
A
I
,
ML
,
C
N
N,
a
n
d
ed
g
e
d
e
v
ice,
t
h
e
f
o
cu
s
ed
li
m
ita
tio
n
s
,
an
d
t
h
e
r
esear
ch
q
u
esti
o
n
s
o
f
th
i
s
liter
atu
r
e.
Fo
r
th
e
s
ec
o
n
d
s
ec
tio
n
,
m
o
s
t
o
f
th
e
r
elate
d
ar
ticles
o
n
th
e
lig
h
t
w
ei
g
h
t
C
NN
o
n
ed
g
e
d
ev
ice
s
w
ill
b
e
co
llected
an
d
m
ap
p
ed
b
y
u
s
in
g
t
h
e
p
r
ef
er
r
ed
r
ep
o
r
tin
g
it
e
m
s
f
o
r
s
y
s
te
m
atic
r
ev
ie
w
s
a
n
d
m
e
ta
-
a
n
al
y
s
i
s
(
P
R
I
SM
A
)
f
r
a
m
e
w
o
r
k
a
s
t
h
e
m
et
h
o
d
o
lo
g
y
p
ar
t
o
f
t
h
is
s
t
u
d
y
.
T
h
ir
d
l
y
,
all
th
e
r
esu
lt
s
w
ill
b
e
an
al
y
ze
d
an
d
d
is
cu
s
s
ed
in
o
r
d
er
to
an
s
w
er
th
e
r
esear
ch
q
u
e
s
tio
n
s
o
f
t
h
e
cu
r
r
en
t
li
g
h
t
w
ei
g
h
t
C
NN
ar
ch
i
tectu
r
e
s
u
s
ed
o
n
th
e
li
m
ited
co
m
p
u
tatio
n
al
r
eso
u
r
ce
s
o
r
ed
g
e
d
ev
ices
an
d
its
p
er
f
o
r
m
an
ce
s
,
as
t
h
e
th
ir
d
s
ec
tio
n
o
f
t
h
i
s
p
ap
er
.
Fin
all
y
,
t
h
is
S
L
R
w
ill
b
e
co
n
clu
d
ed
in
t
h
e
f
o
u
r
th
s
ec
tio
n
b
y
p
r
o
v
id
in
g
th
e
s
u
m
m
ar
ized
f
i
n
d
in
g
s
a
n
d
an
i
n
s
i
g
h
t f
o
r
f
u
t
u
r
e
r
ec
o
m
m
e
n
d
at
io
n
s
o
n
t
h
is
s
co
p
e
o
f
s
tu
d
y
.
Fig
u
r
e
3
.
Min
d
m
ap
ill
u
s
tr
atio
n
o
f
t
h
e
r
esear
ch
q
u
e
s
tio
n
s
2.
RE
S
E
ARCH
M
E
T
H
O
D
I
n
t
h
i
s
S
L
R
,
a
d
etailed
tec
h
n
iq
u
e
f
o
r
id
en
ti
f
y
i
n
g
r
elev
an
t
r
esear
ch
o
n
t
h
e
s
u
b
j
ec
t
r
eg
ar
d
in
g
lig
h
t
w
ei
g
h
t
C
N
N
ar
ch
itect
u
r
e
s
o
n
ed
g
e
d
ev
ice
s
w
as
co
n
d
u
cted
.
T
h
is
ap
p
r
o
ac
h
w
as
ad
o
p
ted
b
ased
o
n
th
e
P
R
I
SMA
f
r
a
m
e
w
o
r
k
[
1
3
]
,
an
d
th
e
m
o
d
if
ied
f
lo
w
c
h
ar
t
in
F
ig
u
r
e
4
s
h
o
w
s
t
h
e
p
r
ac
tical
v
i
e
w
o
f
ea
ch
s
tep
o
f
th
is
S
L
R
’
s
m
eth
o
d
o
lo
g
y
.
B
a
s
icall
y
,
t
h
er
e
w
er
e
t
h
r
ee
m
a
j
o
r
p
h
ases
in
v
o
lv
ed
in
co
m
p
letin
g
t
h
i
s
p
ap
er
.
I
n
itiall
y
,
t
h
e
id
en
ti
f
icat
io
n
p
h
a
s
e
d
eter
m
i
n
ed
th
e
r
ec
o
r
d
s
ac
q
u
ir
ed
f
r
o
m
t
h
e
s
ea
r
ch
s
tr
ateg
y
u
s
ed
i
n
an
y
k
i
n
d
o
f
ac
ad
e
m
ic
r
esear
ch
d
atab
ase.
S
ec
o
n
d
l
y
,
th
e
i
n
it
ial
p
ar
t
o
f
th
e
s
cr
ee
n
i
n
g
p
h
ase
i
n
v
o
l
v
es
e
x
ec
u
ti
n
g
t
h
e
s
e
lectio
n
cr
iter
ia
in
o
r
d
er
to
o
n
l
y
co
n
s
i
d
er
th
e
n
ec
e
s
s
ar
y
ca
te
g
o
r
ies
f
o
r
th
e
d
escr
ip
tiv
e
an
al
y
s
is
.
T
h
ir
d
l
y
,
t
h
e
q
u
alit
y
ass
es
s
m
en
t
is
al
s
o
in
clu
d
ed
as
th
e
s
ec
o
n
d
p
ar
t
o
f
th
e
s
cr
ee
n
in
g
p
h
ase
i
n
o
r
d
er
t
o
f
in
d
o
u
t
th
e
eli
g
ib
ilit
y
o
f
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
9
-
4864
I
n
t J
R
ec
o
n
f
i
g
u
r
ab
le
&
E
m
b
ed
d
ed
Sy
s
t
,
Vo
l.
14
,
No
.
2
,
J
u
l
y
20
25
:
339
-
3
5
2
342
ea
ch
ar
ticle
f
o
r
th
is
s
co
p
e
o
f
s
tu
d
y
.
Fi
n
all
y
,
t
h
e
in
cl
u
d
ed
p
h
ase
s
h
o
w
s
t
h
e
f
i
n
al
n
u
m
b
er
o
f
ar
ticles
th
at
w
il
l
b
e
u
s
ed
in
l
iter
atu
r
e
clas
s
i
f
icatio
n
in
o
r
d
er
to
s
atis
f
y
al
l th
e
r
ese
ar
ch
q
u
esti
o
n
s
s
tated
b
ef
o
r
eh
a
n
d
.
Fig
u
r
e
4
.
T
h
e
m
o
d
if
ied
P
R
I
S
MA
f
r
a
m
e
w
o
r
k
[
1
3
]
w
ith
al
l t
h
e
r
ec
o
r
d
s
f
o
r
th
is
S
L
R
’
s
m
et
h
o
d
o
lo
g
y
2
.
1
.
Sea
rc
h
s
t
ra
t
eg
y
First
o
f
al
l,
th
e
s
ea
r
ch
s
tr
ateg
y
w
a
s
s
p
ec
i
f
icall
y
ex
ec
u
ted
i
n
t
w
o
ac
ad
e
m
ic
d
atab
ase
in
d
e
x
er
s
,
s
u
c
h
a
s
Sco
p
u
s
,
an
d
w
eb
o
f
s
cie
n
c
e
(
W
o
S),
as
w
ell
a
s
t
w
o
p
u
b
lis
h
ed
ar
ticle
d
atab
ases
,
s
u
ch
a
s
I
E
E
E
an
d
Scien
ce
Dir
ec
t.
Mo
r
eo
v
er
,
t
h
e
s
ea
r
ch
o
n
l
y
i
n
cl
u
d
ed
j
o
u
r
n
al
ar
ticles,
r
ev
ie
w
p
ap
er
s
,
an
d
c
o
n
f
er
en
ce
p
ap
er
s
o
r
p
r
o
ce
ed
in
g
s
f
o
r
Sco
p
u
s
an
d
W
o
S,
w
h
ile
o
n
l
y
r
ec
o
r
d
s
f
o
r
jo
u
r
n
al
ar
ticles
f
r
o
m
I
E
E
E
an
d
Scien
ce
Dir
ec
t
w
er
e
ex
tr
ac
ted
.
Nex
t,
th
e
k
e
y
w
o
r
d
s
u
s
ed
f
o
r
all
t
h
e
d
atab
ase
s
ea
r
ch
es
w
er
e
“
li
g
h
t
w
eig
h
t
C
NN”
A
ND
“
ed
g
e
d
ev
ices”
in
t
h
e
s
ea
r
ch
f
ield
s
o
f
titl
e,
ab
s
tr
ac
t,
an
d
au
th
o
r
’
s
k
e
y
w
o
r
d
.
T
h
en
,
in
o
r
d
er
t
o
f
o
cu
s
m
o
r
e
o
n
th
e
r
ec
en
t
an
d
u
p
d
ated
p
ap
er
s
,
th
e
p
u
b
licatio
n
s
’
y
ea
r
s
i
n
t
h
e
d
at
ab
ases
w
er
e
li
m
ited
f
r
o
m
2
0
2
0
to
2
0
2
4
.
A
ls
o
,
t
h
e
s
ea
r
ch
f
o
c
u
s
ed
o
n
p
ap
er
s
p
u
b
l
is
h
ed
o
n
l
y
i
n
E
n
g
li
s
h
.
B
y
ap
p
l
y
in
g
t
h
ese
ter
m
s
,
t
h
e
s
ea
r
c
h
was
n
ar
r
o
w
ed
d
o
w
n
to
a
s
p
ec
if
ic
ar
ea
an
d
s
co
p
e
r
elate
d
to
th
is
s
tu
d
y
.
A
t
t
h
is
s
t
ag
e,
a
to
tal
r
ec
o
r
d
o
f
4
6
0
ar
ti
cles’
m
etad
ata
w
a
s
o
b
tain
ed
th
r
o
u
g
h
o
u
t th
e
s
ea
r
c
h
.
2
.
2
.
Select
io
n
cr
it
er
ia
Fo
r
th
e
s
elec
tio
n
cr
iter
ia,
all
th
e
r
ec
o
r
d
ed
m
e
tad
ata
w
as
c
o
m
b
i
n
ed
i
n
a
s
i
n
g
le
s
p
r
ea
d
s
h
ee
t
f
o
r
th
e
s
cr
ee
n
i
n
g
p
r
o
ce
s
s
.
T
h
e
m
aj
o
r
g
o
al
w
as
to
m
ap
t
h
e
av
a
ilab
le
liter
atu
r
e
o
n
t
h
e
u
s
e
o
f
lig
h
t
weig
h
t
C
NN
i
n
ed
g
e
d
ev
ices
ac
co
r
d
in
g
to
th
e
s
o
u
r
ce
titl
e,
j
o
u
r
n
al
p
u
b
lis
h
er
,
y
e
ar
o
f
p
u
b
licatio
n
s
,
r
esear
ch
f
i
eld
,
an
d
n
u
m
b
er
o
f
citatio
n
s
,
as
th
e
s
e
ca
teg
o
r
ies
w
il
l
b
e
u
s
ed
in
th
e
d
escr
ip
tiv
e
an
al
y
s
i
s
o
f
th
e
r
esu
lt
an
d
d
is
cu
s
s
io
n
s
ec
tio
n
.
A
ll
d
ata
f
o
r
o
th
er
ca
te
g
o
r
ies
w
a
s
ex
cl
u
d
ed
an
d
r
e
m
o
v
ed
.
F
o
r
th
e
n
e
x
t
s
tep
,
all
t
h
e
p
a
p
er
s
’
d
i
g
i
tal
o
b
j
ec
t
id
en
ti
f
ier
s
(
DOI
s
)
w
er
e
s
o
r
ted
o
u
t
in
o
r
d
er
to
r
em
o
v
e
d
u
p
licate
r
ec
o
r
d
s
ea
s
il
y
u
s
in
g
t
h
e
s
p
r
ea
d
s
h
ee
t
'
s
to
o
l.
L
ast
b
u
t
n
o
t
least,
r
e
v
ie
w
p
ap
er
s
an
d
co
n
f
er
en
ce
p
r
o
ce
ed
in
g
s
w
er
e
also
e
x
clu
d
ed
i
n
o
r
d
er
to
k
ee
p
th
e
r
ec
o
r
d
s
m
o
r
e
r
ele
v
an
t.
D
u
e
to
th
e
s
e
cr
iter
ia,
2
8
1
r
esear
ch
p
u
b
licatio
n
s
w
er
e
r
ej
ec
ted
d
u
r
in
g
th
e
in
itial
s
cr
ee
n
i
n
g
p
r
o
ce
s
s
,
an
d
o
n
l
y
1
7
9
r
ec
o
r
d
s
w
er
e
le
f
t f
o
r
f
u
r
t
h
er
ass
e
s
s
m
e
n
t.
2
.
3
.
Q
ua
lity
a
s
s
ess
m
ent
Fo
llo
w
i
n
g
th
e
in
i
tial
s
cr
ee
n
i
n
g
p
h
ase,
a
q
u
alit
y
ass
e
s
s
m
e
n
t
w
a
s
p
er
f
o
r
m
ed
o
n
ea
ch
r
e
s
ea
r
ch
p
ap
er
in
o
r
d
er
to
f
u
r
th
er
en
s
u
r
e
t
h
at
o
n
l
y
th
e
m
o
s
t
e
lig
ib
le
s
t
u
d
ies
w
er
e
i
n
cl
u
d
ed
in
t
h
is
SLR
f
o
r
a
cr
itical
r
ev
ie
w
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
R
ec
o
n
f
i
g
u
r
ab
le
&
E
m
b
ed
d
ed
Sy
s
t
I
SS
N:
2089
-
4864
S
ystema
tic
r
ev
iew
o
f a
lig
h
tw
e
ig
h
t c
o
n
v
o
lu
tio
n
a
l
n
eu
r
a
l n
et
w
o
r
k
…
(
Mu
h
a
mma
d
A
b
b
a
s
A
b
u
Ta
lib
)
343
E
ac
h
ar
ticle's
titl
e
an
d
ab
s
tr
ac
t
w
er
e
s
cr
u
ti
n
ized
th
o
r
o
u
g
h
l
y
to
en
s
u
r
e
th
e
ir
r
elev
an
c
y
a
n
d
co
n
tr
ib
u
tio
n
to
th
e
to
p
ic
u
n
d
er
r
ev
ie
w
.
B
y
e
x
ec
u
tin
g
t
h
is
p
r
o
ce
s
s
,
it
h
elp
s
to
p
u
r
if
y
t
h
e
s
e
lectio
n
,
e
n
s
u
r
i
n
g
o
n
l
y
p
er
tin
e
n
t
a
n
d
h
ig
h
-
q
u
al
it
y
ac
ad
e
m
ic
liter
at
u
r
e
is
i
n
cl
u
d
ed
in
t
h
e
r
e
v
ie
w
p
r
o
ce
s
s
.
As
a
r
es
u
lt,
a
to
tal
o
f
1
3
2
ar
ticles
w
er
e
r
e
m
o
v
ed
f
r
o
m
t
h
e
r
ec
o
r
d
s
,
w
i
th
th
r
ee
o
f
t
h
e
m
b
ei
n
g
a
r
ev
ie
w
p
ap
er
s
an
d
o
th
er
s
b
ein
g
ar
ticles
th
at
w
er
e
n
o
t
r
elate
d
to
i
m
a
g
e
cla
s
s
i
f
icatio
n
o
r
r
eg
r
ess
io
n
(
i.e
.
,
s
i
g
n
a
l
clas
s
if
ica
tio
n
,
s
o
u
n
d
clas
s
i
f
icatio
n
,
o
b
j
ec
t
d
etec
tio
n
,
s
eg
m
e
n
tatio
n
,
an
d
lo
ca
lizatio
n
)
.
A
t
th
i
s
s
tag
e,
th
er
e
w
er
e
o
n
l
y
3
7
r
esear
ch
p
a
p
er
s
lef
t
i
n
t
h
e
r
ec
o
r
d
f
o
r
t
h
e
f
i
n
al
p
r
o
ce
s
s
o
f
d
ata
ex
tr
ac
tio
n
.
2
.
4
.
Da
t
a
ex
t
ra
ct
i
o
n
Du
r
in
g
th
e
d
ata
ex
tr
ac
tio
n
p
h
ase,
3
7
p
u
b
licatio
n
s
w
er
e
ca
r
ef
u
ll
y
s
elec
ted
f
o
r
th
eir
r
elev
an
ce
an
d
ca
p
ac
it
y
to
ad
d
r
ess
th
e
r
esear
ch
q
u
esti
o
n
s
g
i
v
e
n
in
t
h
e
p
r
ec
ed
in
g
s
ec
tio
n
.
Fo
r
t
h
at,
b
y
u
n
d
er
s
ta
n
d
in
g
th
e
cu
r
r
en
t tr
e
n
d
o
f
li
g
h
t
w
eig
h
t
C
NN
o
n
ed
g
e
d
ev
ices,
all
t
h
e
d
a
ta
w
ill b
e
an
al
y
ze
d
as a
li
ter
atu
r
e
class
if
ica
tio
n
i
n
th
e
lat
ter
p
ar
t
o
f
t
h
e
r
e
s
u
l
t
a
n
d
d
is
cu
s
s
io
n
s
ec
t
io
n
.
W
it
h
t
h
at,
v
ar
io
u
s
li
g
h
t
w
ei
g
h
t
C
NN
ar
c
h
itect
u
r
es
t
h
at
h
a
v
e
b
ee
n
co
n
d
u
cted
in
p
r
ev
io
u
s
r
esear
ch
w
ith
r
eso
u
r
ce
-
co
n
s
tr
ain
ed
d
ev
ices
w
i
ll
b
e
h
ig
h
l
ig
h
ted
,
in
clu
d
in
g
th
eir
task
-
b
ased
ca
teg
o
r
ies,
ap
p
licatio
n
s
,
as
w
ell
as
t
h
e
t
y
p
es
o
f
ed
g
e
d
ev
ices
an
d
th
e
ir
s
p
ec
if
i
ca
tio
n
s
.
T
h
en
,
th
e
k
e
y
p
er
f
o
r
m
an
ce
o
f
lig
h
t
w
e
ig
h
t
C
N
N
ar
ch
ite
ct
u
r
es
o
n
t
h
e
s
e
lo
w
-
r
eso
u
r
ce
d
ev
ice
s
w
ill
b
e
ex
a
m
in
ed
i
n
ter
m
s
o
f
ev
al
u
atio
n
t
y
p
es,
m
atr
ices,
an
d
p
er
f
o
r
m
an
ce
co
m
p
ar
i
s
o
n
s
as r
ep
o
r
ted
b
y
v
ar
io
u
s
r
esear
c
h
er
s
.
3.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
N
I
n
t
h
is
s
ec
tio
n
,
all
t
h
e
r
e
s
u
l
t
s
o
b
tain
ed
a
f
ter
co
n
d
u
cti
n
g
t
h
e
ap
p
r
o
ac
h
d
is
c
u
s
s
ed
in
th
e
p
r
ev
io
u
s
s
ec
tio
n
w
i
ll
b
e
o
b
s
er
v
ed
an
d
a
n
al
y
ze
d
.
T
h
is
s
ec
tio
n
i
s
d
iv
id
ed
in
to
t
w
o
s
u
b
s
ec
tio
n
s
.
T
h
e
f
ir
s
t
s
u
b
s
ec
tio
n
w
ill
f
o
cu
s
o
n
t
h
e
d
escr
ip
ti
v
e
an
a
l
y
s
is
to
s
ee
th
e
g
e
n
er
al
tr
en
d
o
f
th
e
r
esear
c
h
,
an
d
t
h
e
s
ec
o
n
d
s
u
b
s
ec
tio
n
,
liter
at
u
r
e
class
i
f
icatio
n
s
,
w
ill
f
u
r
th
er
d
is
cu
s
s
th
e
co
n
te
n
t i
n
o
r
d
er
to
f
u
l
f
il t
h
e
r
esear
ch
q
u
e
s
tio
n
s
in
t
h
is
p
ap
er
.
3
.
1
.
Descript
iv
e
a
na
ly
s
is
Fro
m
t
h
e
m
et
h
o
d
o
lo
g
y
co
n
d
u
cted
,
th
e
o
b
tain
ed
l
iter
atu
r
e
f
o
r
th
is
s
y
s
te
m
atic
liter
at
u
r
e
r
ev
ie
w
h
as
a
to
tal
o
f
3
7
p
ap
er
s
th
at
ar
e
s
p
e
cif
icall
y
r
elate
d
to
t
h
e
r
esear
c
h
o
f
l
ig
h
t
w
ei
g
h
t
C
NN
ar
ch
ite
ctu
r
es
i
m
p
le
m
en
ted
o
r
w
er
e
d
esig
n
ed
f
o
r
r
eso
u
r
ce
-
co
n
s
tr
ain
ed
ed
g
e
d
e
v
ices.
B
ased
o
n
T
ab
le
1
,
all
th
e
p
ap
er
s
w
er
e
cla
s
s
i
f
ied
ac
co
r
d
in
g
to
th
e
y
ea
r
o
f
p
u
b
li
ca
tio
n
,
j
o
u
r
n
al
p
u
b
lis
h
er
s
,
an
d
n
u
m
b
er
o
f
citatio
n
s
.
T
h
en
,
t
h
e
n
u
m
b
er
o
f
r
elate
d
p
ap
er
s
p
u
b
lis
h
ed
in
t
h
e
f
o
ll
o
w
i
n
g
y
ea
r
,
2
0
2
1
–
2
0
2
4
,
is
d
ep
icted
in
Fig
u
r
e
5
,
p
u
b
lis
h
er
class
i
f
icat
io
n
i
n
Fig
u
r
e
6
,
an
d
th
e
n
u
m
b
er
o
f
ci
tatio
n
s
f
r
o
m
ea
c
h
p
ap
er
in
Fig
u
r
e
7
.
T
ab
le
1
.
R
esear
ch
d
atab
ase
d
e
s
cr
ip
tiv
e
an
a
l
y
s
is
R
e
f
.
N
u
m
b
e
r
Y
e
a
r
P
u
b
l
i
s
h
e
r
C
i
t
e
d
R
e
f
.
N
u
m
b
e
r
Y
e
a
r
P
u
b
l
i
s
h
e
r
C
i
t
e
d
[
1
4
]
2
0
2
4
El
se
v
i
e
r
1
[
3
3
]
2
0
2
3
El
se
v
i
e
r
2
[
1
5
]
2
0
2
4
El
se
v
i
e
r
0
[
3
4
]
2
0
2
3
El
se
v
i
e
r
9
[
1
6
]
2
0
2
4
W
i
l
e
y
0
[
3
5
]
2
0
2
2
I
EEE
40
[
1
7
]
2
0
2
4
El
se
v
i
e
r
1
[
3
6
]
2
0
2
2
S
p
r
i
n
g
e
r
11
[
1
8
]
2
0
2
4
El
se
v
i
e
r
4
[
3
7
]
2
0
2
2
M
D
P
I
0
[
1
9
]
2
0
2
4
M
D
P
I
0
[
3
8
]
2
0
2
2
K
I
P
S
6
[
2
0
]
2
0
2
4
M
D
P
I
3
[
3
9
]
2
0
2
2
W
i
l
e
y
3
[
2
1
]
2
0
2
3
C
S
I
R
-
N
I
S
c
P
R
2
[
4
0
]
2
0
2
2
S
p
r
i
n
g
e
r
2
[
2
2
]
2
0
2
3
C
S
I
R
-
N
I
S
c
P
R
1
[
4
1
]
2
0
2
2
El
se
v
i
e
r
22
[
2
3
]
2
0
2
3
S
p
r
i
n
g
e
r
3
[
4
2
]
2
0
2
2
S
p
r
i
n
g
e
r
1
[
2
4
]
2
0
2
3
W
i
l
e
y
1
[
4
3
]
2
0
2
1
M
D
P
I
3
[
2
5
]
2
0
2
3
I
EEE
0
[
4
4
]
2
0
2
1
I
EEE
12
[
2
6
]
2
0
2
3
I
EEE
7
[
4
5
]
2
0
2
1
W
i
l
e
y
16
[
2
7
]
2
0
2
3
El
se
v
i
e
r
3
[
4
6
]
2
0
2
1
K
S
S
2
[
2
8
]
2
0
2
3
El
se
v
i
e
r
4
[
4
7
]
2
0
2
1
I
EEE
33
[
2
9
]
2
0
2
3
El
se
v
i
e
r
16
[
4
8
]
2
0
2
1
El
se
v
i
e
r
41
[
3
0
]
2
0
2
3
C
S
I
R
-
N
I
S
c
P
R
1
[
4
9
]
2
0
2
1
M
D
P
I
5
[
3
1
]
2
0
2
3
C
S
I
R
-
N
I
S
c
P
R
5
[
5
0
]
2
0
2
1
M
D
P
I
18
[
3
2
]
2
0
2
3
I
EEE
9
I
n
Fig
u
r
e
5
,
th
e
p
ie
ch
ar
t
s
h
o
w
s
t
h
at
i
n
th
e
y
ea
r
2
0
2
1
,
th
e
n
u
m
b
er
o
f
p
u
b
lis
h
ed
p
ap
er
s
w
as
8
,
w
h
ic
h
w
a
s
also
th
e
s
a
m
e
n
u
m
b
er
p
r
o
d
u
ce
d
in
2
0
2
2
.
Ho
w
ev
er
,
in
2
0
2
3
,
th
e
n
u
m
b
er
w
as
al
m
o
s
t
t
w
ice
th
e
p
r
ev
io
u
s
p
u
b
lis
h
ed
p
ap
er
,
w
h
ic
h
w
a
s
1
4
.
Mo
r
eo
v
er
,
b
y
t
h
e
m
id
-
y
ea
r
o
f
2
0
2
4
,
th
e
alr
ea
d
y
-
p
u
b
lis
h
ed
ar
ticles
w
er
e
7
.
Hen
ce
,
it
ca
n
b
e
s
ee
n
an
d
p
r
ed
icted
th
at
b
y
t
h
e
en
d
o
f
2
0
2
4
,
th
e
n
u
m
b
er
o
f
p
u
b
lis
h
ed
p
ap
er
s
w
ill
b
e
t
w
ice
as
lar
g
e.
Nex
t,
Fi
g
u
r
e
6
s
h
o
w
s
th
e
n
u
m
b
er
o
f
r
esear
ch
ar
ticles
b
y
p
u
b
lis
h
er
s
.
A
cc
o
r
d
in
g
to
th
i
s
p
ie
ch
ar
t,
th
e
h
ig
h
e
s
t
n
u
m
b
er
o
f
p
ap
er
s
w
e
r
e
p
u
b
lis
h
ed
b
y
E
ls
ev
ier
,
w
h
ich
is
1
1
p
ap
er
s
an
d
2
9
.
7
%
o
f
th
e
to
tal
p
ap
e
r
s
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
9
-
4864
I
n
t J
R
ec
o
n
f
i
g
u
r
ab
le
&
E
m
b
ed
d
ed
Sy
s
t
,
Vo
l.
14
,
No
.
2
,
J
u
l
y
20
25
:
339
-
3
5
2
344
r
ev
ie
w
ed
in
th
i
s
ar
ticle.
Fo
llo
w
i
n
g
t
h
at,
I
E
E
E
an
d
MD
PI
p
u
b
lis
h
ed
s
ix
p
ap
er
s
,
w
ith
1
6
.
2
%
f
r
o
m
ea
c
h
p
u
b
lis
h
er
.
O
th
er
t
h
a
n
t
h
at,
4
p
ap
er
s
an
d
1
0
.
8
%
w
er
e
p
u
b
lis
h
ed
b
y
ea
c
h
o
f
W
ile
y
,
Sp
r
i
n
g
er
,
a
n
d
C
SIR
-
NI
ScP
R
,
lea
v
i
n
g
1
p
ap
er
an
d
2
.
7
% o
f
t
h
e
to
tal
p
ap
er
s
b
ein
g
p
u
b
li
s
h
ed
b
y
KI
P
S a
n
d
KSS,
r
esp
ec
tiv
e
l
y
.
Fig
u
r
e
5
.
Nu
m
b
er
o
f
p
ap
er
s
p
u
b
lis
h
ed
f
r
o
m
ea
c
h
y
ea
r
w
ith
in
2
0
2
1
u
n
til 2
0
2
4
Fig
u
r
e
6
.
Nu
m
b
er
o
f
p
ap
er
s
p
u
b
lis
h
ed
b
y
ea
ch
p
u
b
lis
h
er
Af
ter
t
h
at,
Fi
g
u
r
e
7
an
al
y
ze
d
th
e
n
u
m
b
er
o
f
ci
tatio
n
s
f
r
o
m
ea
ch
p
ap
er
in
t
h
is
S
L
R
.
T
h
e
h
o
r
izo
n
tal
ax
is
d
en
o
tes
th
e
r
e
f
er
en
ce
n
u
m
b
er
o
f
ea
c
h
ar
ticle,
a
n
d
t
h
e
v
er
tical
a
x
i
s
s
h
o
w
s
th
e
n
u
m
b
er
o
f
its
c
itatio
n
.
B
ased
o
n
th
e
b
ar
g
r
ap
h
,
t
h
e
h
i
g
h
e
s
t
n
u
m
b
er
o
f
citatio
n
s
is
4
1
,
f
o
llo
w
ed
b
y
4
0
,
3
3
,
an
d
2
2
.
On
l
y
4
p
ap
er
s
f
r
o
m
th
e
to
tal
ar
ticle
h
a
v
e
t
h
e
n
u
m
b
er
o
f
citatio
n
s
ab
o
v
e
2
0
;
o
th
er
th
a
n
th
at,
m
o
s
t
p
ap
er
s
h
av
e
t
h
e
n
u
m
b
er
o
f
citatio
n
s
b
elo
w
2
0
,
w
h
ich
r
a
n
g
e
f
r
o
m
0
to
1
8
.
I
n
s
u
m
m
ar
y
,
b
y
o
b
s
er
v
in
g
a
n
d
an
a
l
y
zi
n
g
t
h
ese
s
i
m
p
le
liter
at
u
r
e
class
i
f
icatio
n
s
,
it
s
u
g
g
est
s
t
h
a
t
th
e
r
esear
c
h
f
o
c
u
s
ed
o
n
t
h
i
s
f
ield
i
s
s
t
ill
c
u
r
r
en
tl
y
in
th
e
b
eg
in
n
i
n
g
p
h
ase.
T
h
er
ef
o
r
e,
f
u
r
th
er
r
esear
ch
is
n
ee
d
ed
in
o
r
d
er
to
c
o
n
tr
ib
u
te
m
o
r
e
n
o
v
elt
y
a
n
d
a
s
tate
-
of
-
t
h
e
-
ar
t
ap
p
r
o
ac
h
to
th
e
s
t
u
d
y
o
f
lig
h
t
w
ei
g
h
t C
NN
o
n
ed
g
e
d
ev
ices.
Fig
u
r
e
7
.
Nu
m
b
er
o
f
citat
io
n
s
f
r
o
m
ea
ch
p
ap
er
3
.
2
.
L
it
er
a
t
ure
cla
s
s
if
ica
t
io
n
s
Fo
r
th
e
liter
atu
r
e
cla
s
s
i
f
icatio
n
,
th
e
r
e
v
ie
w
ed
ar
ticles
w
er
e
class
i
f
ied
ac
co
r
d
in
g
l
y
b
y
r
e
f
er
r
in
g
to
t
h
e
lig
h
t
w
ei
g
h
t
C
NN
ar
ch
itectu
r
e
s
f
o
r
th
e
p
u
r
p
o
s
e
o
f
ed
g
e
d
ev
ice
i
m
p
le
m
e
n
tat
io
n
.
T
ab
le
2
s
u
m
m
ar
ize
all
t
h
o
s
e
lig
h
t
w
ei
g
h
t
C
NN
ar
ch
itect
u
r
e
s
o
r
b
ased
m
o
d
els,
th
e
t
y
p
es
o
f
th
e
C
NN,
an
d
its
ap
p
lica
tio
n
s
.
Me
an
w
h
ile,
T
ab
le
3
f
o
cu
s
ed
o
n
th
o
s
e
lig
h
t
w
ei
g
h
t
C
NN
ar
ch
i
tectu
r
es
t
h
at
h
av
e
b
ee
n
ex
p
er
i
m
e
n
ted
o
n
ed
g
e
d
ev
ices,
w
h
ic
h
in
cl
u
d
es
it
s
p
er
f
o
r
m
a
n
ce
e
v
al
u
atio
n
in
ter
m
s
o
f
t
h
eir
e
v
al
u
a
tio
n
t
y
p
e
s
,
m
atr
ices,
a
n
d
p
er
f
o
r
m
an
ce
co
m
p
ar
is
o
n
in
ter
m
s
o
f
a
v
er
ag
e
ac
cu
r
ac
y
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
R
ec
o
n
f
i
g
u
r
ab
le
&
E
m
b
ed
d
ed
Sy
s
t
I
SS
N:
2089
-
4864
S
ystema
tic
r
ev
iew
o
f a
lig
h
tw
e
ig
h
t c
o
n
v
o
lu
tio
n
a
l
n
eu
r
a
l n
et
w
o
r
k
…
(
Mu
h
a
mma
d
A
b
b
a
s
A
b
u
Ta
lib
)
345
T
ab
le
2
.
R
esear
ch
d
atab
ase
liter
atu
r
e
class
i
f
icatio
n
s
R
e
f
.
A
r
c
h
i
t
e
c
t
u
r
e
s/
mo
d
e
l
s
Ty
p
e
s o
f
C
N
N
A
p
p
l
i
c
a
t
i
o
n
s
[
1
4
]
L
i
t
e
-
M
D
C
C
l
a
ssi
f
i
c
a
t
i
o
n
t
y
p
e
s o
f
p
i
g
e
o
n
p
e
a
's
d
i
se
a
se
s
P
l
a
n
t
d
i
se
a
se
d
e
t
e
c
t
i
o
n
f
o
r
p
i
g
e
o
n
p
e
a
[
1
5
]
VGG
-
16
C
l
a
ssi
f
i
c
a
t
i
o
n
o
f
c
r
a
c
k
e
d
/
n
o
n
-
c
r
a
c
k
e
d
s
u
r
f
a
c
e
s
A
u
t
o
mat
e
d
c
r
a
c
k
d
e
t
e
c
t
i
o
n
i
n
b
u
i
l
d
i
n
g
i
n
s
p
e
c
t
i
o
n
a
n
d
m
a
i
n
t
e
n
a
n
c
e
[
1
6
]
S
D
L
M
C
l
a
ssi
f
i
c
a
t
i
o
n
o
f
c
a
t
a
r
a
c
t
/
n
o
n
-
c
a
t
a
r
a
c
t
e
y
e
Ca
t
a
r
a
c
t
e
y
e
d
e
t
e
c
t
i
o
n
[
1
7
]
S
h
u
f
f
l
e
N
e
t
V
2
_
Y
O
L
O
v
5
s
C
l
a
ssi
f
i
c
a
t
i
o
n
t
y
p
e
s o
f
c
a
n
o
l
a
k
e
r
n
e
l
g
r
a
d
e
s
R
e
a
l
-
t
i
me
c
a
n
o
l
a
d
a
mag
e
d
e
t
e
c
t
i
o
n
[
1
8
]
O
n
D
e
v
-
L
C
T
C
l
a
ssi
f
i
c
a
t
i
o
n
G
e
n
e
r
a
l
p
u
r
p
o
se
f
o
r
i
mag
e
c
l
a
ssi
f
i
c
a
t
i
o
n
[
1
9
]
Z
e
r
o
-
F
V
e
i
n
N
e
t
C
l
a
ssi
f
i
c
a
t
i
o
n
t
y
p
e
s o
f
f
i
n
g
e
r
s v
e
i
n
F
i
n
g
e
r
v
e
i
n
r
e
c
o
g
n
i
t
i
o
n
[
2
0
]
Y
O
L
O
v
5
C
l
a
ssi
f
i
c
y
t
h
e
ma
t
u
r
i
t
y
o
f
b
l
u
e
b
e
r
r
y
f
r
u
i
t
B
l
u
e
b
e
r
r
y
f
r
u
i
t
m
a
t
u
r
i
t
y
d
e
t
e
c
t
i
on
[
2
1
]
M
o
b
i
l
e
N
e
t
V
2
,
C
o
n
d
e
n
se
N
e
t
V
2
,
S
h
u
f
f
l
e
N
e
t
V
2
C
l
a
ssi
f
i
c
a
t
i
o
n
o
f
f
a
b
r
i
c
,
su
r
f
a
c
e
,
a
n
d
c
a
st
i
n
g
d
e
f
e
c
t
P
r
o
d
u
c
t
d
e
f
e
c
t
d
e
t
e
c
t
i
o
n
i
n
M
F
G
i
n
d
u
st
r
i
e
s
[
2
2
]
M
o
b
i
l
e
N
e
t
V
2
C
l
a
ssi
f
i
c
a
t
i
o
n
t
y
p
e
s o
f
f
a
b
r
i
c
d
e
f
e
c
t
F
a
b
r
i
c
d
e
f
e
c
t
d
e
t
e
c
t
i
o
n
i
n
t
e
x
t
i
l
e
M
F
G
[
2
3
]
C
C
N
N
e
t
C
l
a
ssi
f
i
c
a
t
i
o
n
t
y
p
e
s o
f
t
r
a
d
i
t
i
o
n
a
l
C
h
i
n
e
se
me
d
i
c
i
n
e
(
T
C
M
)
T
r
a
d
i
t
i
o
n
a
l
C
h
i
n
e
se
me
d
i
c
i
n
e
i
m
a
g
e
c
l
a
ssi
f
i
c
a
t
i
o
n
[
2
4
]
M
o
b
i
l
e
N
e
t
C
l
a
ssi
f
i
c
a
t
i
o
n
G
e
n
e
r
a
l
p
u
r
p
o
se
f
o
r
i
mag
e
c
l
a
ssi
f
i
c
a
t
i
o
n
[
2
5
]
V
G
G
1
6
_
B
N
,
R
e
sN
e
t
-
50
,
R
e
g
N
e
t
-
X
C
l
a
ssi
f
i
c
a
t
i
o
n
G
e
n
e
r
a
l
p
u
r
p
o
se
f
o
r
i
mag
e
c
l
a
ssi
f
i
c
a
t
i
o
n
[
2
6
]
CH
-
C
N
N
C
l
a
ssi
f
i
c
a
t
i
o
n
G
e
n
e
r
a
l
p
u
r
p
o
se
f
o
r
i
mag
e
c
l
a
ssi
f
i
c
a
t
i
o
n
[
2
7
]
M
o
b
i
l
e
V
i
TF
a
c
e
C
l
a
ssi
f
i
c
a
t
i
o
n
o
f
d
i
f
f
e
r
e
n
t
b
r
e
e
d
s o
f
sh
e
e
p
S
h
e
e
p
f
a
c
e
r
e
c
o
g
n
i
t
i
o
n
[
2
8
]
V
G
G
N
e
t
-
16
,
R
e
sN
e
t
-
5
0
/
5
6
/
1
1
0
,
G
o
o
g
L
e
N
e
t
,
D
e
n
se
N
e
t
-
40
C
l
a
ssi
f
i
c
a
t
i
o
n
G
e
n
e
r
a
l
p
u
r
p
o
se
f
o
r
i
mag
e
c
l
a
ssi
f
i
c
a
t
i
o
n
[
2
9
]
L
i
t
e
C
N
N
Cl
a
ssi
f
i
c
a
t
i
o
n
o
f
d
i
f
f
e
r
e
n
t
t
y
p
e
s o
f
p
l
a
n
t
d
i
se
a
se
s
P
l
a
n
t
d
i
se
a
se
i
d
e
n
t
i
f
i
c
a
t
i
o
n
[
3
0
]
S
h
u
f
f
l
e
N
e
t
v
2
C
l
a
ssi
f
i
c
a
t
i
o
n
o
f
d
e
f
e
c
t
i
v
e
/
n
o
n
d
e
f
e
c
t
i
v
e
c
a
s
t
i
n
g
C
a
st
i
n
g
d
e
f
e
c
t
d
e
t
e
c
t
i
o
n
[
3
1
]
C
o
n
d
e
n
se
N
e
t
V
2
C
l
a
ssi
f
i
c
a
t
i
o
n
t
y
p
e
s o
f
su
r
f
a
c
e
d
e
f
e
c
t
S
u
r
f
a
c
e
d
e
f
e
c
t
d
e
t
e
c
t
i
o
n
i
n
i
n
d
u
s
t
r
i
a
l
i
n
t
e
l
l
i
g
e
n
t
p
r
o
d
u
c
t
i
o
n
[
3
2
]
C
N
N
C
l
a
ssi
f
i
c
a
t
i
o
n
t
y
p
e
s o
f
f
a
c
i
a
l
e
mo
t
i
o
n
F
a
c
i
a
l
e
mo
t
i
o
n
r
e
c
o
g
n
i
t
i
o
n
f
o
r
V
I
P
[
3
3
]
EB
N
A
S
C
l
a
ssi
f
i
c
a
t
i
o
n
G
e
n
e
r
a
l
p
u
r
p
o
se
f
o
r
i
mag
e
c
l
a
ssi
f
i
c
a
t
i
o
n
[
3
4
]
Y
O
L
O
v
5
s
-
B
i
F
P
N
C
l
a
ssi
f
i
c
a
t
i
o
n
o
f
p
i
g
b
o
d
y
r
e
g
i
o
n
o
f
i
n
t
e
r
e
st
(
R
o
I
)
,
r
e
g
r
e
ssi
o
n
o
f
p
i
g
b
o
d
y
t
e
mp
e
r
a
t
u
r
e
P
i
g
b
o
d
y
t
e
mp
e
r
a
t
u
r
e
a
u
t
o
ma
t
i
c
d
e
t
e
c
t
i
o
n
f
o
r
e
a
r
l
y
d
i
se
a
se
w
a
r
n
i
n
g
[
3
5
]
Ed
g
e
F
i
r
e
S
m
o
k
e
C
l
a
ssi
f
y
t
h
e
o
c
c
u
r
a
n
c
e
o
f
f
o
r
e
st
f
i
r
e
s
F
i
r
e
-
smo
k
e
d
e
t
e
c
t
i
o
n
o
f
f
o
r
e
st
f
i
r
e
s
[
3
6
]
T
r
i
p
l
e
N
e
t
C
l
a
ssi
f
i
c
a
t
i
o
n
G
e
n
e
r
a
l
p
u
r
p
o
se
f
o
r
i
mag
e
c
l
a
ssi
f
i
c
a
t
i
o
n
[
3
7
]
C
o
n
d
e
n
se
N
e
X
t
V
2
C
l
a
ssi
f
i
c
a
t
i
o
n
G
e
n
e
r
a
l
p
u
r
p
o
se
f
o
r
i
mag
e
c
l
a
ssi
f
i
c
a
t
i
o
n
[
3
8
]
I
n
c
e
p
t
i
o
n
V
3
,
M
o
b
i
l
e
N
e
t
,
V
G
G
1
6
C
l
a
ssi
f
i
c
a
t
i
o
n
t
y
p
e
s o
f
f
a
c
e
w
i
t
h
mask
/
w
i
t
h
o
u
t
m
a
s
k
F
a
c
e
mask
c
l
a
ss
i
f
i
c
a
t
i
o
n
[
3
9
]
H
F
EN
e
t
C
l
a
s
si
f
i
c
a
t
i
o
n
o
f
d
e
f
e
c
t
i
v
e
/
n
o
n
d
e
f
e
c
t
i
v
e
c
e
r
a
mi
c
t
i
l
e
su
r
f
a
c
e
C
e
r
a
mi
c
t
i
l
e
su
r
f
a
c
e
d
e
f
e
c
t
d
e
t
e
c
t
i
o
n
[
4
0
]
S
h
u
f
f
l
e
N
e
t
C
l
a
ssi
f
i
c
a
t
i
o
n
o
f
3
D
o
b
j
e
c
t
i
m
a
g
e
s
3
D
o
b
j
e
c
t
r
e
c
o
g
n
i
t
i
o
n
f
o
r
3
D
s
c
a
n
i
n
g
t
e
c
h
n
o
l
o
g
y
[
4
1
]
M
o
b
i
l
e
N
e
t
v
2
R
e
g
r
e
ssi
o
n
o
f
c
r
o
w
d
d
e
c
si
t
y
e
st
i
mat
i
o
n
Est
i
m
a
t
i
n
g
c
r
o
w
d
d
e
n
si
t
y
f
o
r
p
u
b
l
i
c
se
c
u
r
i
t
y
man
a
g
e
me
n
t
[
4
2
]
R
D
P
N
e
t
C
l
a
ssi
f
i
c
a
t
i
o
n
G
e
n
e
r
a
l
p
u
r
p
o
se
f
o
r
i
mag
e
c
l
a
ssi
f
i
c
a
t
i
o
n
[
4
3
]
En
se
mb
l
e
B
i
n
a
r
i
z
e
d
D
r
o
N
e
t
(
EB
D
N
)
C
l
a
ssi
f
i
c
a
t
i
o
n
t
a
sk
f
o
r
c
o
l
l
i
si
o
n
-
a
v
o
i
d
a
n
c
e
,
r
e
g
r
e
ssi
o
n
t
a
sk
f
o
r
p
r
e
d
i
c
t
i
o
n
o
f
d
e
si
r
e
d
st
e
e
r
i
n
g
a
n
g
l
e
A
u
t
o
n
o
mo
u
s
d
r
i
v
i
n
g
f
o
r
u
n
ma
n
n
e
d
a
u
t
o
n
o
mo
u
s v
e
h
i
c
l
e
s (U
A
V
)
[
4
4
]
BC
-
N
e
t
C
l
a
ssi
f
i
c
a
t
i
o
n
G
e
n
e
r
a
l
p
u
r
p
o
se
f
o
r
i
mag
e
c
l
a
ssi
f
i
c
a
t
i
o
n
,
sp
e
e
c
h
r
e
c
o
g
n
i
t
i
o
n
o
f
k
e
y
w
o
r
d
sp
o
t
t
i
n
g
,
f
a
c
i
a
l
e
x
p
r
e
ssi
o
n
r
e
c
o
g
n
i
t
i
o
n
[
4
5
]
M
o
b
i
l
e
N
e
t
-
v2
C
l
a
ssi
f
i
c
a
t
i
o
n
t
y
p
e
s o
f
so
l
i
d
w
a
st
e
W
a
st
e
c
l
a
ssi
f
i
c
a
t
i
o
n
f
o
r
so
l
i
d
w
a
st
e
man
a
g
e
me
n
t
[
4
6
]
D
e
n
se
N
e
t
C
l
a
ssi
f
i
c
a
t
i
o
n
G
e
n
e
r
a
l
p
u
r
p
o
se
f
o
r
i
mag
e
c
l
a
ssi
f
i
c
a
t
i
o
n
[
4
7
]
M
o
b
i
l
e
N
e
t
V
3
C
l
a
ssi
f
i
c
a
t
i
o
n
t
y
p
e
s o
f
i
c
i
n
g
g
r
a
d
e
s
I
c
i
n
g
mo
n
i
t
o
r
i
n
g
o
f
t
r
a
n
smiss
i
o
n
l
i
n
e
s
[
4
8
]
S
p
a
r
k
N
e
t
C
l
a
ssi
f
i
c
a
t
i
o
n
G
e
n
e
r
a
l
p
u
r
p
o
se
f
o
r
i
mag
e
c
l
a
ssi
f
i
c
a
t
i
o
n
[
4
9
]
M
o
b
i
l
e
N
e
t
V
2
&
S
q
u
e
e
z
e
N
e
t
C
l
a
ssi
f
i
c
a
t
i
o
n
t
y
p
e
s o
f
w
a
st
e
R
e
v
e
r
s
e
v
e
n
d
i
n
g
m
a
c
h
i
n
e
f
o
r
t
y
p
e
s o
f
w
a
st
e
r
e
c
y
c
l
e
s
[
5
0
]
A
S
I
R
-
N
e
t
C
l
a
ssi
f
i
c
a
t
i
o
n
t
y
p
e
s o
f
d
i
f
f
e
r
e
n
t
g
r
o
u
n
d
v
e
h
i
c
l
e
t
a
r
g
e
t
A
u
t
o
mat
i
c
t
a
r
g
e
t
r
e
c
o
g
n
i
t
i
o
n
(
A
T
R
)
i
n
sy
n
t
h
e
t
i
c
a
p
e
r
t
u
r
e
r
a
d
a
r
(
S
A
R
)
i
mag
e
s fo
r
mi
l
i
t
a
r
y
su
r
v
e
i
l
l
a
n
c
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
9
-
4864
I
n
t J
R
ec
o
n
f
i
g
u
r
ab
le
&
E
m
b
ed
d
ed
Sy
s
t
,
Vo
l.
14
,
No
.
2
,
J
u
l
y
20
25
:
339
-
3
5
2
346
T
ab
le
3
.
R
esear
ch
d
atab
ase
ed
g
e
d
ev
ices
’
p
er
f
o
r
m
a
n
ce
clas
s
if
icatio
n
s
R
e
f
.
Ed
g
e
d
e
v
i
c
e
t
y
p
e
s
Ev
a
l
u
a
t
i
o
n
t
y
p
e
s
Ev
a
l
u
a
t
i
o
n
ma
t
r
i
c
e
s
A
v
e
.
a
c
c
u
r
a
c
y
(
%)
[
1
5
]
R
a
s
p
b
e
r
r
y
P
i
3
B
+
-
A
c
c
u
r
a
c
y
-
M
o
d
e
l
si
z
e
-
R
o
b
u
st
n
e
ss
-
A
c
c
u
r
a
c
y
-
R
e
c
a
l
l
-
P
r
e
c
i
si
o
n
-
F1
-
S
c
o
r
e
9
5
.
3
0
[
1
6
]
A
n
d
r
o
i
d
S
m
a
r
t
p
h
o
n
e
-
A
c
c
u
r
a
c
y
-
M
o
d
e
l
si
z
e
-
T
i
me
-
A
c
c
u
r
a
c
y
-
I
n
f
e
r
e
n
c
e
t
i
me
-
P
a
r
a
me
t
e
r
s
9
5
.
6
3
[
1
7
]
N
V
I
D
I
A
Je
t
so
n
N
a
n
o
-
S
p
e
e
d
-
S
e
n
si
t
i
v
i
t
y
-
P
r
e
c
i
si
o
n
-
R
e
c
a
l
l
-
F1
-
S
c
o
r
e
-
I
n
f
e
r
e
n
c
e
s
p
e
e
d
-
[
2
1
]
N
V
I
D
I
A
Je
t
so
n
N
a
n
o
-
A
c
c
u
r
a
c
y
-
S
e
n
si
t
i
v
i
t
y
-
S
p
e
c
i
f
i
c
i
t
y
-
A
c
c
u
r
a
c
y
-
R
e
c
a
l
l
-
P
r
e
c
i
si
o
n
-
F1
-
S
c
o
r
e
9
7
.
0
0
[
2
2
]
N
V
I
D
I
A
Je
t
so
n
N
a
n
o
-
A
c
c
u
r
a
c
y
-
S
e
n
si
t
i
v
i
t
y
-
A
c
c
u
r
a
c
y
-
R
e
c
a
l
l
-
P
r
e
c
i
si
o
n
-
F1
-
S
c
o
r
e
9
6
.
5
2
[
2
5
]
-
N
V
I
D
I
A
A
G
X
X
a
v
i
e
r
-
N
V
I
D
I
A
Je
t
so
n
N
a
n
o
-
A
c
c
u
r
a
c
y
-
C
o
mp
u
t
a
t
i
o
n
s
c
o
mp
l
e
x
i
t
y
-
T
i
me
-
M
o
d
e
l
s
i
z
e
-
T
o
p
-
1
a
c
c
u
r
a
c
y
-
M
A
C
s
-
L
a
t
e
n
c
y
-
P
a
r
a
me
t
e
r
s
7
5
.
5
7
[
2
7
]
N
V
I
D
I
A
Je
t
so
n
N
a
n
o
-
A
c
c
u
r
a
c
y
-
A
c
c
u
r
a
c
y
-
P
r
e
c
i
si
o
n
-
R
e
c
a
l
l
9
7
.
1
3
[
2
9
]
Z
Y
N
Q
Z
7
-
L
i
t
e
7
0
2
0
F
P
G
A
-
A
c
c
u
r
a
c
y
-
S
p
e
e
d
-
T
i
me
-
A
c
c
u
r
a
c
y
-
I
n
f
e
r
e
n
c
e
s
p
e
e
d
-
L
a
t
e
n
c
y
9
5
.
7
1
[
3
0
]
N
V
I
D
I
A
Je
t
so
n
N
a
n
o
-
A
c
c
u
r
a
c
y
-
S
e
n
si
t
i
v
i
t
y
-
P
r
e
c
i
si
o
n
-
R
e
c
a
l
l
-
F1
-
S
c
o
r
e
-
A
c
c
u
r
a
c
y
9
9
.
5
8
[
3
1
]
N
V
I
D
I
A
Je
t
so
n
X
a
v
i
e
r
N
x
-
A
c
c
u
r
a
c
y
-
S
e
n
si
t
i
v
i
t
y
-
A
c
c
u
r
a
c
y
-
R
e
c
a
l
l
-
P
r
e
c
i
si
o
n
-
F1
-
S
c
o
r
e
9
1
.
4
0
[
3
5
]
N
V
I
D
I
A
Je
t
so
n
N
a
n
o
-
A
c
c
u
r
a
c
y
-
S
e
n
si
t
i
v
i
t
y
-
T
i
me
-
A
c
c
u
r
a
c
y
-
R
e
c
a
l
l
-
P
r
e
c
i
si
o
n
-
F1
-
S
c
o
r
e
-
H
a
mm
i
n
g
l
o
ss
9
8
.
9
7
[
3
6
]
R
a
s
p
b
e
r
r
y
P
i
4
-
T
i
me
-
C
o
mp
u
t
a
t
i
o
n
s
c
o
mp
l
e
x
i
t
y
-
L
a
t
e
n
c
y
-
F
L
O
P
S
-
[
3
7
]
N
X
P
B
l
u
e
B
o
x
2
.
0
-
M
o
d
e
l
S
i
z
e
-
A
c
c
u
r
a
c
y
-
T
i
me
-
C
o
mp
u
t
a
t
i
o
n
s
c
o
mp
l
e
x
i
t
y
-
F
L
O
P
S
-
P
a
r
a
me
t
e
r
s
-
T
o
p
-
1
a
c
c
u
r
a
c
y
-
I
n
f
e
r
e
n
c
e
t
i
me
8
4
.
5
5
[
3
8
]
R
a
s
p
b
e
r
r
y
P
i
4
-
A
c
c
u
r
a
c
y
-
S
p
e
e
d
-
L
o
ss
-
A
c
c
u
r
a
c
y
-
P
r
e
c
i
si
o
n
-
R
e
c
a
l
l
-
F1
-
S
c
o
r
e
9
5
.
5
1
[
3
7
]
R
a
s
p
b
e
r
r
y
P
i
4
-
S
p
e
e
d
-
C
o
mp
u
t
a
t
i
o
n
s
c
o
mp
l
e
x
i
t
y
-
T
i
me
-
M
e
a
n
a
b
so
l
u
t
e
e
r
r
o
r
(
M
A
E)
-
R
M
S
E
-
I
n
f
e
r
e
n
c
e
sp
e
e
d
a
n
d
t
i
me
-
F
L
O
P
S
-
[
4
3
]
X
i
l
i
n
x
Z
y
n
q
7
Z
1
0
0
F
P
G
A
-
A
c
c
u
r
a
c
y
-
P
r
e
c
i
si
o
n
-
S
p
e
e
d
-
R
M
S
E
-
F1
-
S
c
o
r
e
-
F
P
S
9
5
.
6
0
[
4
4
]
N
U
C
L
EO
-
F
7
6
7
ZI
w
i
t
h
S
T
M
3
2
H
7
4
3
-
T
i
me
-
M
S
E
-
L
a
t
e
n
c
y
-
[
4
7
]
H
u
a
w
e
i
A
t
l
a
s 2
0
0
D
K
-
A
c
c
u
r
a
c
y
-
P
r
e
c
i
si
o
n
-
T
i
me
-
A
c
c
u
r
a
c
y
-
T
i
me
-
F
P
S
7
4
.
5
0
[
4
8
]
I
n
t
e
l
A
r
r
i
a
1
0
G
X
1
1
5
0
F
P
G
A
-
T
i
me
-
S
p
e
e
d
-
I
n
f
e
r
e
n
c
e
t
i
me
-
S
p
e
e
d
u
p
-
[
4
9
]
-
N
V
I
D
I
A
Je
t
so
n
N
a
n
o
-
N
V
I
D
I
A
Je
t
so
n
T
X
1
-
A
c
c
u
r
a
c
y
-
T
i
me
-
A
c
c
u
r
a
c
y
9
5
.
0
0
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
R
ec
o
n
f
i
g
u
r
ab
le
&
E
m
b
ed
d
ed
Sy
s
t
I
SS
N:
2089
-
4864
S
ystema
tic
r
ev
iew
o
f a
lig
h
tw
e
ig
h
t c
o
n
v
o
lu
tio
n
a
l
n
eu
r
a
l n
et
w
o
r
k
…
(
Mu
h
a
mma
d
A
b
b
a
s
A
b
u
Ta
lib
)
347
3
.
2
.
1
.
L
ig
htw
eig
ht
c
o
nv
o
lutio
na
l neura
l net
w
o
rk
a
rc
hite
ct
ures
Firstl
y
,
Fi
g
u
r
e
8
s
h
o
w
s
th
e
n
u
m
b
er
o
f
ar
ch
itect
u
r
es
u
s
ed
f
r
o
m
all
t
h
e
r
esear
ch
i
n
T
ab
le
2
.
B
ased
o
n
th
e
ar
ch
itect
u
r
es
o
r
b
ased
m
o
d
el,
it
ca
n
b
e
s
ee
n
th
at
th
er
e
w
er
e
s
ev
er
al
s
i
m
i
lar
n
et
w
o
r
k
s
t
h
at
w
er
e
b
ein
g
u
s
ed
as
th
eir
ap
p
r
o
ac
h
es,
s
u
c
h
as
Mo
b
ileNet
w
it
h
9
(
1
9
.
1
%),
Sh
u
f
f
leNe
t
w
it
h
4
(
8
.
5
%),
VG
G
-
1
6
w
it
h
4
(
8
.
5
%),
C
o
n
d
en
s
eNe
t
w
it
h
3
(
6
.
4
%),
r
esp
ec
tiv
el
y
in
ter
m
s
o
f
n
u
m
b
er
o
f
th
eir
u
s
a
g
e.
Me
an
w
h
i
le,
th
e
Ot
h
er
s
w
it
h
2
1
(
4
4
.
7
%),
r
e
p
r
esen
t
th
e
n
u
m
b
e
r
o
f
d
if
f
er
en
t
m
o
d
el
s
w
it
h
o
n
l
y
o
n
e
u
s
ag
e.
All
t
h
ese
ar
e
s
o
m
e
o
f
t
h
e
s
tate
-
of
-
th
e
-
ar
t
ap
p
r
o
ac
h
es
th
at
ar
e
cu
r
r
en
tl
y
b
ein
g
u
s
ed
b
y
r
esear
c
h
er
s
i
n
th
i
s
f
ield
.
W
h
ile
s
o
m
e
o
f
th
e
r
esear
ch
er
s
u
s
e
t
h
e
m
as
b
en
c
h
m
ar
k
s
,
th
e
r
e
ar
e
also
s
ev
er
al
o
th
er
s
t
h
at
m
o
d
if
y
t
h
ese
o
r
ig
i
n
al
n
et
w
o
r
k
s
w
it
h
v
ar
io
u
s
v
er
s
io
n
s
to
i
m
p
r
o
v
e
th
eir
p
er
f
o
r
m
a
n
ce
s
.
A
s
id
e
f
r
o
m
th
at,
t
h
er
e
w
er
e
a
f
e
w
w
i
th
h
y
b
r
id
m
o
d
els
o
r
m
u
lt
i
m
o
d
al
w
h
ic
h
co
m
b
in
ed
t
w
o
o
r
m
o
r
e
n
et
w
o
r
k
s
to
g
et
h
er
b
y
u
s
in
g
n
o
v
el
ap
p
r
o
ac
h
es.
Fo
r
ex
am
p
le,
r
esear
ch
in
[
2
7
]
co
m
b
i
n
ed
a
lig
h
t
w
ei
g
h
t
C
NN
ar
ch
itect
u
r
e’
s
Mo
b
ileNet
m
o
d
el
w
it
h
a
Vis
io
n
T
r
an
s
f
o
r
m
er
ar
ch
itect
u
r
e.
A
ll
i
n
all,
Mo
b
ileNet
in
clu
d
i
n
g
its
v
ar
io
u
s
v
er
s
io
n
is
t
h
e
m
o
s
t
u
s
e
d
lig
h
t
w
e
ig
h
t
C
NN
ar
ch
itec
tu
r
e
f
o
r
ed
g
e
d
ev
ices
i
m
p
le
m
en
ta
tio
n
.
Fig
u
r
e
8
.
Sev
er
al
d
if
f
er
en
t c
o
m
m
o
n
li
g
h
t
w
ei
g
h
t
C
NN
ar
ch
it
ec
tu
r
es
3
.
2
.
2
.
T
y
pes
o
f
co
nv
o
lutio
na
l
neura
l net
w
o
rk
Nex
t,
as
f
o
r
th
e
t
y
p
es
o
f
l
ig
h
t
w
ei
g
h
t
C
NN
ar
ch
itect
u
r
es
i
n
T
a
b
le
2
,
m
o
s
t
o
f
th
e
r
esear
c
h
u
s
ed
f
o
r
class
i
f
icatio
n
w
it
h
a
f
e
w
o
f
th
e
m
u
s
ed
f
o
r
r
eg
r
ess
io
n
,
a
n
d
s
o
m
e
co
m
b
i
n
ed
b
o
th
th
e
class
i
f
icatio
n
an
d
r
eg
r
ess
io
n
.
B
ased
o
n
Fig
u
r
e
9
,
it
s
u
g
g
e
s
ts
t
h
at
th
e
clas
s
i
f
ica
tio
n
t
y
p
e
w
a
s
th
e
m
ai
n
ap
p
r
o
ac
h
th
at
w
as
b
ein
g
r
esear
ch
ed
w
ith
3
4
(
9
1
.
9
%)
an
d
th
e
r
e
g
r
ess
io
n
t
y
p
e
w
i
th
1
(
2
.
7
%)
w
as
m
u
ch
m
o
r
e
co
m
p
lica
ted
to
b
e
r
esea
r
ch
ed
o
n
.
Sin
ce
t
h
e
r
eg
r
e
s
s
io
n
t
y
p
es
o
f
C
NN
r
eq
u
ir
e
s
co
n
tin
u
o
u
s
d
ata
f
o
r
p
r
ed
ictio
n
,
its
i
m
p
le
m
e
n
tatio
n
f
o
r
r
eso
u
r
ce
-
co
n
s
tr
ain
ed
d
ev
i
ce
s
m
a
y
r
eq
u
ir
e
a
h
ig
h
er
co
m
p
u
tat
io
n
al
u
s
ag
e
co
m
p
ar
ed
to
th
e
class
i
f
icatio
n
t
y
p
es
a
n
d
th
u
s
,
t
h
e
r
es
u
lt
i
n
d
i
ca
ted
th
at
o
n
l
y
m
i
n
o
r
r
esear
ch
h
a
s
b
ee
n
d
o
n
e
f
o
r
r
eg
r
ess
io
n
s
ta
s
k
.
Ho
w
e
v
er
,
th
er
e
w
er
e
s
till
a
f
e
w
s
t
u
d
ies
th
at
ap
p
lied
th
e
r
eg
r
ess
io
n
lig
h
t
w
ei
g
h
t
C
NN,
an
d
s
o
m
e
als
o
u
s
ed
it
in
h
y
b
r
id
m
o
d
el
s
w
it
h
class
i
f
icat
io
n
an
d
r
eg
r
ess
io
n
w
it
h
2
(
5
.
4
%).
I
n
s
h
o
r
t,
m
o
s
t
o
f
th
e
r
esear
ch
in
lig
h
t
w
e
ig
h
t
C
NN
w
a
s
lean
i
n
g
to
w
ar
d
s
th
e
clas
s
i
f
icatio
n
t
y
p
es o
f
C
NN.
Fig
u
r
e
9
.
T
h
e
co
m
m
o
n
t
y
p
e
s
o
f
C
N
N
3
.
2
.
3
.
M
o
dels
’
a
pp
lica
t
io
ns
Fro
m
th
e
a
s
p
ec
t
o
f
ap
p
licatio
n
s
,
T
ab
le
2
d
ep
icted
th
at
th
is
f
ield
o
f
r
esear
c
h
is
v
er
y
co
m
p
r
eh
en
s
iv
e.
T
h
e
s
tu
d
ies
w
er
e
co
n
d
u
cted
f
r
o
m
v
ar
io
u
s
f
ield
in
cl
u
d
i
n
g
i
n
d
u
s
tr
ial
m
a
n
u
f
ac
tu
r
i
n
g
(
MF
G)
w
it
h
8
(
2
1
.
6
%),
p
u
b
lic
s
u
r
v
eilla
n
ce
a
n
d
s
af
e
t
y
w
it
h
8
(
2
1
.
6
%),
h
ea
lth
w
it
h
4
(
1
0
.
8
%),
w
a
s
te
m
a
n
a
g
e
m
en
t
w
it
h
2
(
5
.
4
%),
ag
r
icu
l
tu
r
e
(
i.e
.
,
an
i
m
a
l
an
d
p
l
an
t)
d
is
ea
s
e
d
etec
tio
n
w
i
th
2
(
5
.
4
%),
an
d
m
ilit
ar
y
w
it
h
1
(
2
.
7
%)
as
ill
u
s
tr
ate
i
n
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
9
-
4864
I
n
t J
R
ec
o
n
f
i
g
u
r
ab
le
&
E
m
b
ed
d
ed
Sy
s
t
,
Vo
l.
14
,
No
.
2
,
J
u
l
y
20
25
:
339
-
3
5
2
348
Fig
u
r
e
1
0
.
Asi
d
e
f
r
o
m
t
h
at,
t
h
e
h
ig
h
es
t
n
u
m
b
er
o
f
ap
p
licat
io
n
s
w
er
e
f
o
r
g
e
n
er
al
-
p
u
r
p
o
s
e
(
GP
)
u
s
es
w
i
th
1
2
(
3
2
.
4
%)
w
h
ic
h
s
h
o
w
s
th
e
f
lex
ib
ilit
y
a
n
d
r
eliab
ilit
y
o
f
th
e
n
et
w
o
r
k
to
b
e
u
s
ed
w
i
th
m
a
n
y
d
i
f
f
er
e
n
t
ap
p
licatio
n
s
.
Fig
u
r
e
1
0
.
Var
io
u
s
ap
p
licatio
n
s
o
f
li
g
h
t
w
eig
h
t
C
NN
u
s
e
f
o
r
d
if
f
er
e
n
t secto
r
o
f
i
n
d
u
s
tr
y
3
.
2
.
4
.
T
y
pes
o
f
edg
e
dev
ices
On
t
h
e
o
t
h
er
h
a
n
d
,
b
ased
o
n
T
ab
le
3
,
2
0
s
tu
d
y
w
er
e
f
i
lte
r
ed
o
u
t
f
o
r
t
h
e
p
er
f
o
r
m
a
n
ce
an
al
y
s
is
o
f
lig
h
t
w
ei
g
h
t
C
N
N
ar
ch
itect
u
r
e
d
ep
lo
y
ed
in
ed
g
e
d
e
v
ice.
A
s
i
d
e
f
r
o
m
th
e
s
e
2
0
r
esear
ch
,
th
e
o
th
er
1
7
r
esear
ch
ea
r
lier
d
id
n
o
t
co
n
tin
u
e
u
s
i
n
g
ed
g
e
d
ev
ices
w
h
ile
o
n
l
y
ev
a
lu
ati
n
g
t
h
eir
li
g
h
t
w
ei
g
h
t
C
N
N
w
i
th
s
o
m
e
o
f
t
h
e
b
en
ch
m
ar
k
li
g
h
t
w
ei
g
h
t
m
o
d
els.
Fig
u
r
e
1
1
s
u
m
m
ar
izes
t
h
e
t
y
p
e
s
an
d
n
u
m
b
er
s
o
f
ed
g
e
d
ev
ices
b
ased
o
n
th
e
r
esear
ch
i
n
T
ab
le
3
.
Firstl
y
,
i
t
ca
n
b
e
s
ee
n
th
at
ed
g
e
d
e
v
ic
es
ca
n
b
e
ca
teg
o
r
ized
i
n
to
t
wo
t
y
p
es
w
h
ic
h
ar
e
on
-
t
h
e
-
s
h
el
f
d
ev
ices
i
n
cl
u
d
in
g
s
ev
er
al
s
y
s
te
m
-
on
-
a
-
ch
ip
(
So
C
)
(
e.
g
.
,
NVI
DI
A
J
etso
n
s
er
ie
s
an
d
R
asp
b
er
r
y
P
i
s
er
ies),
m
icr
o
co
n
tr
o
ller
(
MCU
)
w
it
h
it
s
d
ev
elo
p
m
e
n
t
b
o
ar
d
s
u
ch
a
s
t
h
e
NU
C
L
E
O
-
F7
6
7
Z
I
w
it
h
ST
M3
2
H7
4
3
,
an
d
An
d
r
o
id
S
m
ar
tp
h
o
n
e.
Me
an
w
h
ile,
th
e
o
f
f
-
t
he
-
s
h
e
lf
d
ev
ices
in
cl
u
d
e
th
o
s
e
b
ein
g
i
m
p
l
e
m
en
ted
as
FP
GAs
(
e.
g
.
,
Xilin
x
Z
y
n
q
7
Z
1
0
0
,
Z
Y
NQ
Z
7
-
L
ite
7
0
2
0
,
an
d
I
n
tel
Ar
r
ia
1
0
GX1
1
5
0
)
.
Hen
ce
,
m
o
s
t
o
f
th
e
r
ec
en
t
s
t
u
d
y
s
h
o
w
s
th
at
NVI
DI
A
J
etso
n
N
an
o
as
t
h
e
m
o
s
t
u
s
ed
ed
g
e
d
ev
ice
w
it
h
7
(
3
1
.
8
%),
f
o
llo
w
ed
b
y
R
asp
b
e
r
r
y
P
i
4
w
it
h
3
(
1
3
.
6
%),
an
d
FP
GAs
al
s
o
w
ith
3
(
1
3
.
6
%)
w
h
i
le
o
th
er
s
(
So
C
)
w
it
h
7
(
3
1
.
8
%)
r
ep
r
esen
t
t
h
e
o
th
er
t
y
p
es
o
f
So
C
ed
g
e
d
ev
ice
s
th
a
t
w
er
e
u
s
ed
o
n
l
y
o
n
ce
.
Fig
u
r
e
1
1
.
So
m
e
o
f
th
e
g
en
er
a
l e
d
g
e
d
ev
ices
u
s
ed
to
e
m
b
ed
a
lig
h
t
w
ei
g
h
t
C
NN
ar
ch
itect
u
r
e
3
.
2
.
5
.
E
v
a
lua
t
io
n
t
y
pes
a
nd
m
a
t
rice
s
L
ast
b
u
t
n
o
t
least,
f
o
r
th
e
ev
al
u
atio
n
t
y
p
e
s
an
d
m
atr
ices,
m
o
s
t
o
f
th
ese
s
t
u
d
ies
f
o
cu
s
ed
o
n
ac
cu
r
ac
y
,
ti
m
e,
s
p
ee
d
,
s
en
s
it
iv
it
y
,
m
o
d
el
s
ize,
co
m
p
u
tat
io
n
co
m
p
lex
it
y
,
an
d
o
th
er
s
.
Mo
r
e
o
v
er
,
s
ev
er
al
m
a
tr
ices,
s
u
c
h
as
ac
cu
r
ac
y
,
r
ec
all,
p
r
ec
is
io
n
,
F1
-
Sco
r
e
,
in
f
er
en
ce
ti
m
e
an
d
s
p
e
ed
,
laten
c
y
,
p
ar
a
m
eter
s
,
f
lo
ati
n
g
p
o
in
t
o
p
er
atio
n
s
(
FL
OP
s
)
,
an
d
o
t
h
er
s
,
w
er
e
a
l
w
a
y
s
b
ein
g
u
s
ed
i
n
o
r
d
er
to
en
s
u
r
e
t
h
at
t
h
e
li
g
h
t
w
ei
g
h
t
C
N
N
f
o
r
ed
g
e
d
e
v
ices
p
er
f
o
r
m
a
n
ce
w
as
o
p
ti
m
ized
.
So
m
e
o
f
th
e
ev
a
lu
atio
n
m
atr
i
ce
s
r
ep
r
esen
t
ea
ch
o
f
th
e
ev
al
u
atio
n
t
y
p
e
s
as
th
e
m
atr
ices
ar
e
t
h
e
s
p
ec
if
ic
ass
ess
m
en
t
o
f
ea
ch
o
f
th
e
e
v
al
u
atio
n
t
y
p
es.
Fo
r
ex
a
m
p
le,
ac
cu
r
ac
y
a
n
d
to
p
-
1
ac
cu
r
ac
y
ar
e
th
e
e
v
alu
at
io
n
m
atr
ices
f
o
r
ev
alu
a
tio
n
t
y
p
e
o
f
ac
cu
r
ac
y
,
i
n
f
er
e
n
ce
ti
m
e
an
d
laten
c
y
ar
e
t
h
e
ev
alu
a
tio
n
m
atr
ices
f
o
r
e
v
alu
a
tio
n
t
y
p
e
o
f
ti
m
e,
a
n
d
F
L
OP
s
an
d
m
u
ltip
l
y
-
ac
cu
m
u
late
o
p
er
atio
n
s
(
M
AC
s
)
ar
e
th
e
ev
a
lu
atio
n
m
atr
ices
f
o
r
ev
alu
a
tio
n
t
y
p
e
o
f
co
m
p
u
tati
o
n
co
m
p
lex
i
t
y
.
T
ab
le
3
d
esc
r
ib
es
s
o
m
e
o
f
t
h
e
Evaluation Warning : The document was created with Spire.PDF for Python.