I
nte
rna
t
io
na
l J
o
urna
l o
f
E
lect
rica
l a
nd
Co
m
p
ute
r
E
ng
in
ee
ring
(
I
J
E
CE
)
Vo
l.
9
,
No
.
4
,
A
u
g
u
s
t
201
9
,
p
p
.
3
3
1
4
~
3
3
2
2
I
SS
N:
2088
-
8708
,
DOI
: 1
0
.
1
1
5
9
1
/
i
j
ec
e
.
v
9
i
4
.
p
p
3
3
1
4
-
3322
3314
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ia
e
s
co
r
e
.
co
m/
jo
u
r
n
a
ls
/in
d
ex
.
p
h
p
/
I
JE
C
E
O
f
fline
sig
na
tur
e
v
erification us
ing
DAG
-
CN
N
J
a
v
ier
O
.
P
inzó
n
-
Are
na
s
,
Ro
bin
s
o
n J
i
m
é
nez
-
M
o
re
no
,
Césa
r
G
.
P
a
chó
n
-
Su
esc
ún
F
a
c
u
lt
y
o
f
En
g
in
e
e
rin
g
,
Nu
e
v
a
G
r
a
n
a
d
a
M
il
it
a
ry
Un
iv
e
rsit
y
,
Co
lo
m
b
ia
Art
icle
I
nfo
AB
ST
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Au
g
6
,
2
0
1
8
R
ev
i
s
ed
Mar
29
,
2
0
1
9
A
cc
ep
ted
A
p
r
8
,
2
0
1
9
T
h
is
p
a
p
e
r
p
re
se
n
ts
th
e
im
p
lem
e
n
tatio
n
o
f
a
D
AG
-
CNN
w
h
i
c
h
a
im
s
to
c
las
si
fy
a
n
d
v
e
ri
fy
th
e
a
u
th
e
n
ti
c
it
y
o
f
th
e
o
ff
li
n
e
sig
n
a
tu
re
s
o
f
3
u
se
rs,
u
sin
g
th
e
w
rit
e
r
-
in
d
e
p
e
n
d
e
n
t
m
e
th
o
d
.
I
n
o
rd
e
r
to
d
e
v
e
lo
p
t
h
is
w
o
rk
,
2
d
a
tab
a
se
s
(train
in
g
/
v
a
li
d
a
ti
o
n
a
n
d
tes
ti
n
g
)
we
re
b
u
il
t
m
a
n
u
a
ll
y
,
i.
e
.
th
e
m
a
n
u
a
l
c
o
ll
e
c
ti
o
n
o
f
th
e
sig
n
a
tu
re
s
o
f
th
e
3
u
se
rs
a
s
w
e
ll
a
s
f
o
rg
e
d
sig
n
a
tu
re
s
m
a
d
e
b
y
p
e
o
p
le
n
o
t
b
e
lo
n
g
in
g
to
th
e
b
a
se
a
n
d
a
lt
e
re
d
b
y
th
e
sa
m
e
u
se
rs
w
e
r
e
d
o
n
e
,
a
n
d
sig
n
a
tu
re
s o
f
a
n
o
th
e
r
1
1
5
p
e
o
p
le w
e
r
e
u
se
d
to
c
re
a
te t
h
e
c
a
t
e
g
o
ry
o
f
n
o
n
-
m
e
m
b
e
rs.
On
c
e
th
e
n
e
tw
o
rk
is
tr
a
in
e
d
,
it
s
v
a
li
d
a
ti
o
n
a
n
d
su
b
se
q
u
e
n
t
tes
ti
n
g
is
p
e
rf
o
r
m
e
d
,
o
b
tain
in
g
o
v
e
ra
ll
a
c
c
u
ra
c
ies
o
f
9
9
.
4
%
a
n
d
9
9
.
3
%
,
re
sp
e
c
ti
v
e
l
y
,
sh
o
w
in
g
th
e
f
e
a
tu
re
s
lea
rn
e
d
b
y
t
h
e
n
e
tw
o
rk
a
n
d
v
e
ri
fy
in
g
th
e
a
b
il
it
y
o
f
th
is
c
o
n
f
ig
u
ra
ti
o
n
o
f
n
e
u
ra
l
n
e
tw
o
rk
to
b
e
u
se
d
i
n
a
p
p
li
c
a
ti
o
n
s
f
o
r
id
e
n
ti
f
ica
ti
o
n
a
n
d
v
e
rif
ica
ti
o
n
o
f
o
f
f
li
n
e
sig
n
a
tu
re
s.
K
ey
w
o
r
d
s
:
B
ilin
ea
r
ar
ch
itect
u
r
e
DAG
co
n
v
o
l
u
tio
n
al
n
e
u
r
al
n
et
w
o
r
k
Han
d
w
r
iti
n
g
r
ec
o
g
n
itio
n
L
a
y
er
ac
tiv
atio
n
Of
f
li
n
e
s
i
g
n
a
tu
r
e
v
er
i
f
icat
io
n
Co
p
y
rig
h
t
©
2
0
1
9
In
stit
u
te o
f
A
d
v
a
n
c
e
d
E
n
g
i
n
e
e
rin
g
a
n
d
S
c
ien
c
e
.
Al
l
rig
h
ts re
se
rv
e
d
.
C
o
r
r
e
s
p
o
nd
ing
A
uth
o
r
:
R
o
b
in
s
o
n
J
i
m
én
ez
Mo
r
en
o
,
Me
ch
atr
o
n
ic
s
E
n
g
in
ee
r
i
n
g
P
r
o
g
r
a
m
,
Fac
u
lt
y
o
f
E
n
g
i
n
ee
r
i
n
g
,
Nu
e
v
a
Gr
an
ad
a
Militar
y
Un
i
v
er
s
it
y
,
C
ar
r
er
a
1
1
#
1
0
1
-
8
0
,
B
o
g
o
tá
D
.
C
.
,
C
o
lo
m
b
ia.
E
m
ail:
r
o
b
in
s
o
n
.
j
i
m
en
ez
@
u
n
i
m
ilit
ar
.
ed
u
.
co
1.
I
NT
RO
D
UCT
I
O
N
Sig
n
at
u
r
es,
ap
ar
t
f
r
o
m
b
io
m
etr
ic
id
en
tif
ica
tio
n
s
s
u
c
h
as
f
i
n
g
er
p
r
in
t
s
,
ar
e
p
er
s
o
n
au
th
en
ticatio
n
s
y
s
te
m
s
t
h
at
ar
e
cu
r
r
en
tl
y
ac
c
ep
ted
in
m
o
s
t
s
o
cial
co
n
te
x
t
s
as
a
m
ea
n
s
o
f
id
en
tify
i
n
g
an
d
v
alid
ati
n
g
id
e
n
tit
y
.
Ho
w
e
v
er
,
th
e
f
o
r
g
er
ie
s
o
f
s
i
g
n
at
u
r
es
is
a
p
r
o
b
le
m
i
m
m
er
s
ed
in
t
h
i
s
s
y
s
te
m
th
at
r
e
p
r
esen
ts
a
b
ig
p
r
o
b
le
m
w
h
ic
h
h
as
b
ee
n
tr
ied
to
s
o
lv
e
o
r
m
i
tig
a
te
i
n
d
i
f
f
er
en
t
w
a
y
s
,
u
s
in
g
d
i
f
f
er
en
t
m
eth
o
d
s
o
f
v
er
if
ica
tio
n
o
f
th
e
au
th
e
n
tic
it
y
o
f
a
s
i
g
n
at
u
r
e.
I
n
th
e
f
ir
s
t
p
lace
,
it
m
u
s
t
b
e
ta
k
e
n
i
n
to
ac
co
u
n
t
th
at
t
h
er
e
ar
e
2
w
a
y
s
o
f
o
b
tain
i
n
g
d
ata
f
r
o
m
a
s
ig
n
at
u
r
e,
w
h
er
e
th
e
f
ir
s
t
o
n
e
i
s
o
f
f
lin
e,
s
i
g
n
atu
r
es
th
at
ar
e
o
b
tain
ed
f
r
o
m
d
o
cu
m
e
n
t
s
alr
ea
d
y
s
ig
n
ed
,
w
h
ic
h
ar
e
s
ca
n
n
ed
in
o
r
d
er
to
ac
q
u
ir
e
a
d
ig
italizatio
n
o
f
th
e
s
a
m
e
[
1
]
,
eith
er
b
y
p
ass
i
n
g
t
h
e
i
m
a
g
e
i
n
g
r
a
y
s
ca
le
o
r
b
y
b
in
ar
izat
io
n
.
W
h
ile
t
h
e
s
ec
o
n
d
i
s
o
n
lin
e,
wh
er
e
d
ata
is
ac
q
u
ir
ed
d
u
r
in
g
t
h
e
p
r
o
ce
s
s
o
f
m
ak
i
n
g
th
e
s
i
g
n
at
u
r
e
[
2
,
3
]
,
tak
in
g
in
f
o
r
m
at
io
n
s
u
c
h
as
th
e
s
tar
t
a
n
d
en
d
p
o
s
itio
n
,
n
u
m
b
er
o
f
t
i
m
e
s
th
e
p
e
n
is
li
f
ted
,
s
p
ee
d
,
p
r
ess
u
r
e,
etc.
C
o
m
p
ar
i
n
g
t
h
e
u
s
e
o
f
o
n
li
n
e
o
r
o
f
f
li
n
e
s
i
g
n
at
u
r
es,
in
g
e
n
er
a
l,
t
h
e
u
s
e
o
f
o
n
li
n
e
m
et
h
o
d
s
h
as
a
g
r
ea
ter
d
eg
r
ee
o
f
ac
cu
r
a
c
y
,
h
o
w
e
v
er
,
th
eir
p
r
o
b
le
m
is
th
at
t
h
e
d
ata
ca
n
o
n
l
y
b
e
o
b
tain
ed
th
r
o
u
g
h
s
p
ec
ial
d
ev
ices
a
n
d
at
t
h
e
ti
m
e
o
f
s
ig
n
in
g
,
s
o
it
ca
n
b
e
a
p
r
o
b
le
m
w
h
en
th
e
y
ar
e
s
ig
n
at
u
r
e
s
alr
ea
d
y
m
ad
e
in
d
o
cu
m
en
ts
o
r
b
an
k
c
h
ec
k
s
,
f
o
r
th
is
r
ea
s
o
n
,
t
h
e
u
s
e
o
f
m
e
th
o
d
s
w
it
h
o
f
f
li
n
e
s
i
g
n
at
u
r
es
th
at
v
alid
ate
t
h
e
m
,
is
n
ec
es
s
ar
y
.
On
e
o
f
t
h
e
m
o
s
t
i
m
p
o
r
tan
t
ca
u
s
e
s
f
o
r
th
e
m
et
h
o
d
s
o
f
v
er
if
i
ca
tio
n
o
f
o
f
f
li
n
e
s
ig
n
at
u
r
es
d
o
es
n
o
t
h
a
v
e
th
e
d
e
g
r
ee
o
f
ac
c
u
r
ac
y
o
f
o
n
li
n
e,
is
t
h
e
lo
s
s
o
f
d
y
n
a
m
ic
i
n
f
o
r
m
atio
n
,
w
h
ic
h
m
a
k
es
t
h
e
p
r
o
b
le
m
m
o
r
e
d
if
f
ic
u
lt
[
4
]
.
On
th
e
o
th
er
h
a
n
d
,
w
it
h
i
n
o
f
f
li
n
e
s
y
s
te
m
s
,
t
h
er
e
ar
e
tw
o
m
e
th
o
d
s
,
w
r
iter
-
d
ep
en
d
en
t
(
W
D)
,
w
h
ic
h
is
to
p
er
f
o
r
m
a
v
er
if
icatio
n
s
y
s
te
m
f
o
r
ea
ch
u
s
er
in
d
i
v
id
u
all
y
,
an
d
w
r
iter
-
i
n
d
ep
en
d
en
t
(
W
I
)
,
w
h
er
e
a
s
in
g
le
s
y
s
te
m
co
v
er
s
th
e
v
er
i
f
icatio
n
o
f
all
r
eg
is
ter
ed
u
s
er
s
[
5
]
,
n
ev
er
t
h
el
ess
,
a
W
I
s
y
s
te
m
r
ep
r
ese
n
ts
a
g
r
ea
ter
ch
alle
n
g
e,
s
in
ce
it
i
s
p
o
s
s
ib
le
n
o
t
to
f
o
c
u
s
th
e
s
y
s
te
m
o
n
r
ec
o
g
n
izi
n
g
th
e
s
p
ec
if
ic
d
etail
s
o
f
a
u
s
er
,
b
u
t
to
g
e
n
er
alize
ch
ar
ac
ter
is
tic
s
[
6
]
,
as
w
ell
as e
r
r
o
n
eo
u
s
l
y
to
s
a
y
t
h
at
th
e
s
i
g
n
atu
r
e
o
f
o
n
e
u
s
er
b
elo
n
g
s
to
an
o
th
er
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8708
Offlin
e
s
ig
n
a
tu
r
e
ve
r
ifica
tio
n
u
s
in
g
DA
G
-
CNN
(
Ja
vier O
.
P
i
n
z
ó
n
-
A
r
en
a
s
)
3315
Du
e
to
th
e
ab
o
v
e,
s
e
v
er
al
tec
h
n
iq
u
es
h
a
v
e
b
ee
n
i
m
p
le
m
e
n
t
ed
f
o
r
th
e
o
f
f
li
n
e
s
i
g
n
a
tu
r
e
v
er
if
icatio
n
,
s
u
c
h
as
th
e
o
n
e
p
r
esen
ted
in
[
7
]
,
in
w
h
ich
b
o
th
t
h
e
W
I
a
n
d
W
D
m
et
h
o
d
s
ar
e
u
s
ed
to
g
en
er
ate
a
h
y
b
r
id
class
i
f
ier
,
r
ea
ch
in
g
ac
cu
r
ac
ie
s
f
r
o
m
8
6
.
0
4
%
u
p
to
9
4
.
6
2
%.
L
i
k
e
w
is
e,
t
h
an
k
s
to
th
e
h
ig
h
p
er
f
o
r
m
a
n
ce
o
f
n
eu
r
al
n
et
w
o
r
k
s
in
t
h
e
clas
s
i
f
icatio
n
o
f
u
s
er
s
u
s
i
n
g
h
a
n
d
w
r
itin
g
s
e
m
p
lo
y
in
g
th
e
W
I
m
et
h
o
d
,
as
ca
n
b
e
s
ee
n
in
[
8
]
a
n
d
[
9
]
w
ith
a
cc
u
r
ac
ies
o
f
u
p
to
9
9
%,
h
av
e
b
eg
u
n
to
b
e
u
s
ed
to
i
m
p
le
m
en
t
s
y
s
te
m
s
f
o
r
v
er
if
y
i
n
g
o
f
f
li
n
e
s
ig
n
at
u
r
es.
An
e
x
a
m
p
le
o
f
th
i
s
is
d
escr
ib
ed
i
n
[
1
0
]
,
w
h
er
e
it i
s
b
u
il
t a
n
eu
r
al
n
et
w
o
r
k
ar
ch
i
t
ec
tu
r
e
w
h
ich
,
af
ter
b
ein
g
tr
ai
n
ed
w
ith
a
d
atab
ase
o
f
1
1
1
u
s
er
s
,
o
b
tain
ed
a
9
6
.
3
% in
it
s
o
v
er
all
ac
cu
r
ac
y
.
W
ith
th
e
r
i
s
e
o
f
t
h
e
Dee
p
L
e
ar
n
in
g
tec
h
n
iq
u
es
[
1
1
]
,
m
o
r
e
r
o
b
u
s
t
n
et
w
o
r
k
s
,
s
u
ch
a
s
C
o
n
v
o
lu
tio
n
al
Neu
r
al
Net
w
o
r
k
s
(
C
N
N)
[
1
2
]
,
h
av
e
b
ee
n
s
tar
ted
to
b
e
u
s
ed
in
r
ec
o
g
n
itio
n
o
f
p
atter
n
s
in
i
m
a
g
es.
T
h
o
s
e
n
et
w
o
r
k
s
ar
e
ab
le
t
o
e
x
tr
a
ct
h
i
g
h
er
lev
e
l
f
ea
t
u
r
es,
all
o
w
i
n
g
its
u
s
e
i
n
co
m
p
lex
d
ev
elo
p
m
e
n
t
s
r
elate
d
to
letter
s
an
d
w
o
r
d
s
,
s
u
c
h
a
s
in
t
h
e
r
ec
o
g
n
itio
n
o
f
lic
en
s
e
p
late
s
o
f
ca
r
s
[
1
3
]
an
d
w
o
r
d
s
p
o
tti
n
g
ap
p
licatio
n
s
[
1
4
,
1
5
]
,
am
o
n
g
o
th
er
s
.
T
h
e
C
NN
h
as
a
ls
o
d
em
o
n
s
tr
ated
to
h
a
v
e
b
etter
p
er
f
o
r
m
a
n
ce
t
h
an
t
h
e
co
m
m
o
n
ar
tif
ic
ial
n
e
u
r
al
n
et
w
o
r
k
s
,
s
u
r
p
as
s
i
n
g
t
h
e
ac
cu
r
ac
y
in
m
o
r
e
th
a
n
1
0
%
[
1
6
]
.
T
h
e
ad
v
an
ta
g
e
th
at
t
h
e
C
NNs
h
a
v
e
is
t
h
at
th
e
y
d
o
es
n
o
t
o
n
l
y
f
o
cu
s
t
h
eir
lear
n
i
n
g
o
n
f
in
er
d
etail
s
b
u
t
o
n
g
e
n
er
a
l
ap
p
ea
r
an
ce
o
f
t
h
e
s
ig
n
at
u
r
es
[
1
7
]
.
T
h
is
ca
p
ab
ilit
y
h
a
s
g
iv
e
n
t
h
e
p
o
s
s
ib
ili
t
y
o
f
ap
p
ly
i
n
g
t
h
e
m
in
w
o
r
k
s
li
k
e
t
h
e
o
n
e
d
ev
e
lo
p
ed
in
[1
8
]
,
w
h
er
e
a
Dee
p
C
NN
i
s
u
s
ed
in
o
r
d
er
to
d
is
cr
im
i
n
ate
wh
ich
s
i
g
n
at
u
r
es
b
elo
n
g
ed
to
w
h
ich
u
s
er
u
s
i
n
g
th
e
W
I
m
et
h
o
d
,
an
d
th
e
n
m
o
v
e
o
n
to
an
o
th
er
p
h
ase,
w
h
er
e
W
D
i
s
u
s
ed
to
class
i
f
y
th
e
s
i
g
n
at
u
r
e
as g
e
n
u
in
e
o
r
n
o
t.
Oth
er
s
i
m
ilar
w
o
r
k
is
d
escr
ib
ed
in
[
1
9
]
,
w
h
er
e
a
h
y
b
r
id
W
I
-
W
D
m
o
d
el
is
i
m
p
le
m
en
ted
,
u
s
in
g
a
2
-
ch
a
n
n
el
C
NN
w
it
h
a
ch
an
n
el
f
o
r
a
q
u
er
y
s
ig
n
at
u
r
e
an
d
o
th
er
f
o
r
a
r
ef
er
en
ce
s
i
g
n
a
tu
r
e,
ad
d
itio
n
a
ll
y
,
ad
d
in
g
a
n
W
D
s
tag
e
w
i
th
a
s
u
p
p
o
r
t v
ec
to
r
m
a
ch
in
e
(
S
VM
)
tr
ain
ed
w
it
h
th
e
C
NN
f
ea
tu
r
e
s
o
f
th
e
la
s
t d
o
w
n
s
a
m
p
li
n
g
la
y
er
.
I
n
[
2
0
]
,
in
co
n
tr
ast
to
t
h
e
a
f
o
r
em
e
n
tio
n
ed
w
o
r
k
s
,
t
h
e
a
u
t
h
o
r
s
p
r
o
p
o
s
e
3
s
tag
es,
t
h
at
w
er
e
ca
lled
“f
ea
t
u
r
e
lear
n
i
n
g
”,
w
h
er
e
a
p
r
e
-
tr
ain
ed
r
e
s
id
u
al
C
N
N
w
a
s
u
s
ed
to
lear
n
th
e
ch
ar
ac
ter
is
tic
s
o
f
th
e
s
i
g
n
atu
r
e
s
,
th
en
,
2
ad
d
itio
n
al
s
tag
e
s
“
ac
ti
v
e
lear
n
i
n
g
”
an
d
“
f
in
al
v
er
i
f
ic
atio
n
”
ar
e
i
m
p
le
m
en
ted
w
i
th
S
MV
,
th
e
f
ir
s
t o
n
e
to
f
i
n
all
y
s
ep
ar
ate
t
h
e
g
en
u
i
n
e
s
ig
n
at
u
r
e
f
r
o
m
t
h
e
f
o
r
g
er
ies,
an
d
t
h
e
o
th
er
to
f
i
n
all
y
v
er
i
f
y
th
e
au
th
e
n
ticit
y
.
Ho
w
e
v
er
,
th
o
s
e
d
ev
elo
p
m
e
n
t
n
ee
d
s
at
least
2
s
ta
g
es
to
f
in
al
l
y
v
er
i
f
y
if
t
h
e
s
i
g
n
atu
r
e
i
s
g
e
n
u
i
n
e
o
r
n
o
t,
h
e
n
ce
,
th
e
id
ea
o
f
o
u
r
w
o
r
k
is
to
co
m
b
in
e
t
h
e
t
w
o
s
ta
g
e
i
n
o
n
l
y
o
n
e
n
et
w
o
r
k
ca
p
ab
le
o
f
id
en
tify
a
n
u
s
er
s
i
g
n
atu
r
e
an
d
v
er
if
y
it
s
au
t
h
en
ticit
y
a
n
d
also
d
is
ca
r
d
th
e
s
ig
n
at
u
r
es
n
o
b
elo
n
g
in
g
to
th
e
u
s
er
s
r
e
g
is
ter
ed
.
A
s
i
m
ilar
ap
p
r
o
ac
h
to
o
u
r
w
o
r
k
i
s
p
r
esen
ted
in
[
2
1
]
,
w
h
er
e
th
e
W
I
an
d
W
D
s
tag
es
ar
e
co
m
b
in
ed
in
t
o
o
n
ly
o
n
e
Sia
m
e
s
e
C
NN,
n
ev
er
t
h
ele
s
s
,
i
t
n
ee
d
to
co
m
p
ar
e
t
h
e
s
i
m
ilar
it
y
b
et
w
ee
n
a
p
air
o
f
s
ig
n
at
u
r
es,
i.e
.
a
s
a
m
p
le
o
f
a
g
en
u
i
n
e
s
ig
n
at
u
r
e
is
n
ee
d
ed
.
I
n
o
u
r
w
o
r
k
,
n
o
s
a
m
p
le
to
co
m
p
ar
e
is
n
ee
d
ed
in
o
r
d
er
t
o
v
er
if
y
th
e
s
ig
n
at
u
r
e
en
ter
ed
to
th
e
n
et
w
o
r
k
.
I
n
o
r
d
er
t
o
ad
v
an
ce
in
th
e
d
ev
elo
p
m
e
n
t
o
f
th
e
tec
h
n
iq
u
es
u
s
ed
f
o
r
o
f
f
li
n
e
s
i
g
n
at
u
r
e
v
er
if
icatio
n
,
th
is
w
o
r
k
p
r
esen
ts
th
e
u
s
e
o
f
a
DA
G
-
C
NN,
a
n
et
w
o
r
k
th
a
t
co
n
s
is
ts
o
f
a
C
NN
co
n
f
i
g
u
r
ed
w
it
h
a
s
tr
u
ct
u
r
e
t
y
p
e
Dir
ec
ted
A
c
y
clic
Gr
ap
h
(
D
AG)
Net
w
o
r
k
[
22
]
,
w
h
ic
h
h
as
as
ad
v
a
n
ta
g
e
o
v
er
a
C
NN
co
n
f
ig
u
r
ed
in
a
li
n
ea
r
w
a
y
th
e
u
s
e
o
f
m
u
ltip
le
p
ath
s
o
r
b
r
an
ch
es
in
o
r
d
er
to
allo
w
t
h
e
n
et
w
o
r
k
to
in
cr
ea
s
e
its
d
ep
th
(
n
u
m
b
er
o
f
co
n
v
o
lu
tio
n
la
y
er
s
u
s
ed
)
w
it
h
o
u
t
h
a
v
i
n
g
to
m
a
k
e
it
lo
n
g
er
.
A
l
s
o
,
ea
ch
p
ath
ca
n
b
e
co
n
f
ig
u
r
ed
i
n
d
if
f
er
en
t
w
a
y
s
,
p
r
o
v
id
i
n
g
th
e
n
et
w
o
r
k
w
it
h
t
h
e
p
o
s
s
ib
ilit
y
to
lear
n
d
if
f
er
en
t
f
ea
t
u
r
es
d
ir
ec
tl
y
f
r
o
m
th
e
o
r
ig
i
n
al
in
p
u
t.
W
ith
t
h
is
n
et
w
o
r
k
,
w
e
s
ee
k
t
o
v
er
if
y
th
e
au
t
h
e
n
ticit
y
o
f
d
if
f
er
en
t
s
i
g
n
atu
r
e
s
u
s
i
n
g
t
h
e
W
I
m
eth
o
d
f
o
r
a
n
ap
p
licatio
n
w
h
er
e
a
r
eg
i
s
tr
at
io
n
o
f
3
u
s
er
s
i
s
d
o
n
e,
w
h
i
ch
ar
e
ac
ce
p
ted
b
y
m
ea
n
s
o
f
t
h
eir
s
i
g
n
at
u
r
e,
ad
d
itio
n
all
y
,
t
h
e
n
et
w
o
r
k
m
u
s
t
b
e
ab
le
to
k
n
o
w
i
f
t
h
e
s
ig
n
atu
r
e
th
a
t
is
m
ad
e
b
elo
n
g
s
o
r
n
o
t
to
t
h
e
p
eo
p
le
in
r
eg
i
s
ter
ed
.
T
h
is
w
o
r
k
is
d
iv
id
ed
in
to
5
s
ec
tio
n
s
,
w
h
er
e
s
ec
tio
n
2
d
escr
ib
es
th
e
d
atab
ase
b
u
ilt
f
o
r
th
e
i
m
p
le
m
en
ta
tio
n
.
Sectio
n
3
p
r
esen
ts
t
h
e
p
r
o
p
o
s
ed
a
r
ch
itectu
r
e
to
g
eth
er
w
it
h
its
r
esp
e
ctiv
e
tr
ain
i
n
g
an
d
v
alid
atio
n
.
Sectio
n
4
p
r
esen
ts
th
e
r
es
u
lts
o
b
tain
ed
th
r
o
u
g
h
a
n
e
w
d
atab
ase
an
d
t
h
e
a
n
al
y
s
i
s
o
f
th
e
b
e
h
av
io
r
o
f
th
e
n
et
w
o
r
k
.
Fi
n
all
y
,
s
ec
tio
n
5
g
iv
e
s
th
e
co
n
clu
s
io
n
s
r
ea
ch
ed
.
2.
M
E
T
H
O
DO
L
O
G
Y
2
.
1
.
Da
t
a
ba
s
e
B
ec
au
s
e
t
h
is
w
o
r
k
is
f
o
cu
s
ed
o
n
th
e
v
er
i
f
icatio
n
o
f
a
g
r
o
u
p
o
f
3
s
p
ec
if
ic
u
s
er
s
,
w
e
b
u
il
t
o
u
r
o
w
n
d
atab
ase.
T
h
is
d
atab
ase
co
n
s
is
ts
o
f
3
3
9
g
e
n
u
i
n
e
s
ig
n
at
u
r
es
an
d
2
6
0
f
o
r
g
er
ies
m
ad
e
m
an
u
all
y
b
y
u
s
er
s
,
ad
d
itio
n
all
y
,
a
s
it
is
r
eq
u
ir
ed
t
h
at
i
f
a
n
o
th
er
p
er
s
o
n
s
i
g
n
s
w
it
h
t
h
eir
o
w
n
s
i
g
n
atu
r
e
is
n
o
t r
e
co
g
n
ized
w
i
th
i
n
t
h
e
u
s
er
s
,
a
n
ad
d
itio
n
al
ca
te
g
o
r
y
i
s
ad
d
ed
w
ith
1
s
i
g
n
atu
r
e
o
f
1
1
5
d
if
f
er
en
t
u
s
er
s
,
w
h
ic
h
w
er
e
o
b
tain
ed
f
r
o
m
t
h
e
UT
Sig
Data
s
et
[
23
].
2
.
1
.
1
.
Da
t
a
ba
s
e
c
o
llect
i
o
n
T
h
e
co
llectio
n
o
f
th
e
g
e
n
u
in
e
s
ig
n
at
u
r
es
is
d
o
n
e
w
it
h
ea
c
h
o
f
th
e
u
s
er
s
,
w
r
iti
n
g
i
t
o
n
a
wh
ite
s
h
ee
t,
w
it
h
o
u
t r
estrictio
n
o
f
s
ize
a
n
d
d
ir
ec
tio
n
,
i.e
.
as t
h
e
p
er
s
o
n
d
o
es n
o
t a
l
w
a
y
s
s
ig
n
i
n
t
h
e
s
a
m
e
w
a
y
,
t
h
e
s
ig
n
at
u
r
e
ca
n
b
e
m
ad
e
lar
g
er
o
r
s
m
a
ller
,
j
u
s
t
as
it
m
a
y
h
av
e
v
a
r
iatio
n
s
in
it
s
in
c
lin
a
tio
n.
I
n
ad
d
itio
n
to
th
is
,
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
n
t J
E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
9
,
No
.
4
,
A
u
g
u
s
t
201
9
:
3
3
1
4
-
332
2
3316
th
e
co
llectio
n
o
f
all
th
e
s
i
g
n
at
u
r
es
i
s
d
o
n
e
in
a
s
i
n
g
le
d
a
y
,
i
n
o
th
er
w
o
r
d
s
,
p
eo
p
le
s
tar
t
w
i
th
t
h
e
r
ested
h
an
d
,
b
u
t
w
h
ile
t
h
e
y
ad
v
an
ce
i
n
t
h
e
w
r
iti
n
g
,
th
e
y
ar
e
g
o
in
g
to
f
e
el
an
n
o
y
a
n
ce
o
r
f
ati
g
u
e
w
h
e
n
w
r
iti
n
g
,
d
o
in
g
th
a
t
th
e
y
h
av
e
m
aj
o
r
s
v
ar
iatio
n
s
i
n
th
e
elab
o
r
atio
n
o
f
th
e
s
i
g
n
a
tu
r
e,
th
is
to
h
a
v
e
m
aj
o
r
v
ar
iatio
n
in
th
e
g
en
u
i
n
e
s
ig
n
at
u
r
es.
I
n
to
tal,
3
3
9
g
e
n
u
i
n
e
s
ig
n
at
u
r
es
ar
e
o
b
tai
n
ed
.
T
h
e
f
o
r
g
er
ies
co
n
s
is
t
o
f
2
t
y
p
es,
s
o
m
e
elab
o
r
ated
b
y
th
e
s
a
m
e
u
s
er
,
s
o
th
at
h
e
m
a
k
es
s
lig
h
t
v
ar
iatio
n
s
w
h
er
e
h
e/
s
h
e
is
ab
le
to
id
en
tify
th
at
i
s
n
o
t
h
is
/
h
er
s
ig
n
at
u
r
e,
an
d
m
ad
e
b
y
th
ir
d
p
ar
ties
(
p
e
o
p
le
n
o
t
b
elo
n
g
in
g
to
th
e
d
ata
b
ase)
,
co
llectin
g
a
to
tal
o
f
2
6
0
f
o
r
g
ed
s
ig
n
at
u
r
es.
A
s
a
m
p
le
o
f
t
h
e
d
atab
ase
ca
n
b
e
s
ee
n
in
F
ig
u
r
e
1
.
Fig
u
r
e
1
.
Sa
m
p
les o
f
ea
c
h
u
s
e
r
2
.
1
.
2
.
P
re
pro
ce
s
s
ing
T
w
o
p
h
a
s
es
o
f
p
r
ep
r
o
ce
s
s
in
g
w
er
e
ca
r
r
ied
o
u
t:
th
e
f
ir
s
t
p
h
as
e
w
a
s
d
o
n
e
m
a
n
u
a
ll
y
,
w
h
er
e
e
ac
h
o
f
th
e
s
ig
n
at
u
r
es
w
as
d
i
g
italized
w
it
h
a
s
ca
n
n
er
o
f
3
0
0
d
p
i
r
eso
lu
tio
n
,
s
av
ed
in
J
P
G
f
o
r
m
at,
f
o
ll
o
w
ed
b
y
th
i
s
,
ea
c
h
o
f
th
e
s
i
g
n
at
u
r
es
w
as
cu
t
t
o
a
p
r
e
d
ef
in
ed
s
ize,
tak
i
n
g
in
to
ac
co
u
n
t
th
e
lar
g
e
s
t
s
i
g
n
atu
r
e,
w
h
ich
w
a
s
570
×
9
2
0
p
x
,
w
ith
o
u
t
s
ca
li
n
g
,
i.e
.
w
it
h
its
o
r
ig
in
a
l
d
ig
itizati
o
n
s
ize,
an
d
f
in
al
l
y
,
t
h
e
r
esid
u
es
n
o
t
b
elo
n
g
i
n
g
to
th
e
s
i
g
n
at
u
r
es,
s
u
ch
as
s
tai
n
s
o
r
d
ir
t
,
w
er
e
e
li
m
in
ated
.
I
n
th
e
s
ec
o
n
d
p
h
ase,
th
e
i
m
p
r
o
v
e
m
en
t
o
f
t
h
e
i
m
a
g
e
is
p
er
f
o
r
m
ed
,
w
h
ic
h
co
n
s
i
s
ts
o
f
p
ass
in
g
it
to
g
r
a
y
s
ca
le
an
d
m
a
k
in
g
a
co
n
tr
ast
ad
j
u
s
t
m
e
n
t
to
h
ig
h
li
g
h
t
th
e
d
etails
o
f
th
e
s
i
g
n
at
u
r
es u
s
i
n
g
(
1
)
.
,
=
+
(
ℎ
ℎ
−
)
(
′
,
ℎ
ℎ
−
)
,
ℎ
:
′
,
=
{
0
(
,
−
(
)
∗
)
−
≤
0
ℎ
(
,
−
(
)
∗
)
−
(
1
)
W
h
er
e
,
is
th
e
o
u
tp
u
t
p
i
x
el
at
th
e
p
o
s
itio
n
(
i,j)
o
f
t
h
e
p
r
o
ce
s
s
ed
im
a
g
e
an
d
,
th
e
in
p
u
t
p
i
x
el,
an
d
ℎ
ℎ
ar
e
th
e
u
p
p
er
an
d
lo
w
er
li
m
it
s
o
f
s
ca
li
n
g
o
f
th
e
p
r
o
ce
s
s
ed
i
m
ag
e,
an
d
ℎ
ℎ
ar
e
th
e
u
p
p
er
an
d
lo
w
er
li
m
it
s
o
f
s
ca
lin
g
o
f
th
e
i
n
p
u
t
i
m
a
g
e,
a
n
d
i
s
g
a
m
m
a
,
v
ar
iab
le
t
h
at
m
o
d
if
ies
th
e
m
ap
p
i
n
g
cu
r
v
e
b
et
w
ee
n
in
p
u
t a
n
d
o
u
tp
u
t.
Fo
r
th
i
s
ca
s
e,
=
0
,
ℎ
ℎ
=
1
,
=
0
.
5059
,
ℎ
ℎ
=
1
an
d
=
1
.
2
.
1
.
3
.
Da
t
a
a
ug
m
e
nta
t
io
n
T
o
in
cr
ea
s
e
th
e
d
atab
ase
w
it
h
o
u
t
n
ee
d
i
n
g
to
r
eq
u
ir
e
m
o
r
e
s
i
g
n
at
u
r
es
o
f
th
e
u
s
er
s
a
n
d
in
t
u
r
n
to
m
ak
e
ch
an
g
es
i
n
t
h
e
p
o
s
it
io
n
s
o
f
th
e
s
i
g
n
at
u
r
e
s
,
it
i
s
d
ec
id
ed
t
o
c
ar
r
y
o
u
t
a
d
ata
a
u
g
m
e
n
tatio
n
.
I
n
o
r
d
er
to
d
o
th
i
s
au
g
m
e
n
tatio
n
,
a
lo
ca
tio
n
alg
o
r
ith
m
is
ap
p
lied
u
s
i
n
g
m
o
r
p
h
o
lo
g
ical
o
p
er
atio
n
s
[
2
4
]
an
d
a
co
n
d
itio
n
al
o
f
ar
ea
,
i.e
.
th
e
im
a
g
e
is
b
i
n
ar
ized
w
it
h
a
f
i
x
ed
th
r
es
h
o
ld
,
w
h
er
e
p
ix
el
s
w
it
h
v
al
u
es
g
r
ea
ter
th
an
2
5
2
b
ec
o
m
e
1
,
o
th
er
w
i
s
e,
b
ec
o
m
e
0
,
to
th
en
ap
p
l
y
i
n
g
a
clo
s
e
o
p
er
atio
n
(
d
ilatio
n
f
o
llo
w
ed
b
y
er
o
s
io
n
)
f
o
llo
w
ed
b
y
an
o
p
en
o
p
er
atio
n
(
er
o
s
io
n
f
o
llo
w
ed
b
y
d
ila
tio
n
)
,
in
o
r
d
er
to
eli
m
i
n
ated
t
h
e
n
o
is
e
t
h
at
was
n
o
t
er
ased
i
n
t
h
e
f
ir
s
t
p
h
a
s
e
o
f
th
e
p
r
ep
r
o
ce
s
s
in
g
.
W
ith
t
h
is
,
a
co
n
d
itio
n
al
is
ap
p
lied
o
v
er
th
e
r
em
ai
n
in
g
a
r
ea
s
,
w
h
er
e
i
f
th
e
y
h
av
e
less
th
a
n
5
0
0
p
x
,
th
e
y
ar
e
er
ased
,
in
o
r
d
er
to
av
o
id
tak
i
n
g
in
to
ac
co
u
n
t r
e
s
id
u
al
s
tai
n
s
,
s
in
ce
t
h
e
r
es
u
lt
in
g
ar
ea
s
ar
e
u
s
ed
to
g
en
er
ate
t
h
e
b
o
u
n
d
in
g
b
o
x
t
h
at
w
i
ll
co
n
tain
t
h
e
s
i
g
n
at
u
r
e.
T
o
th
e
b
o
u
n
d
in
g
b
o
x
,
2
0
p
x
ar
e
ad
d
ed
t
o
ea
ch
s
id
e,
in
o
r
d
er
to
co
v
er
p
o
s
s
ib
le
p
ar
ts
o
f
th
e
tr
a
ce
s
th
at
m
ig
h
t
b
e
o
u
t
f
r
o
m
it,
an
d
th
en
,
t
h
e
b
o
x
is
lo
ca
ted
o
n
th
e
o
r
ig
in
al
i
m
a
g
e.
T
h
is
p
r
o
ce
s
s
ca
n
b
e
s
ee
n
in
Fi
g
u
r
e
2
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8708
Offlin
e
s
ig
n
a
tu
r
e
ve
r
ifica
tio
n
u
s
in
g
DA
G
-
CNN
(
Ja
vier O
.
P
i
n
z
ó
n
-
A
r
en
a
s
)
3317
Fig
u
r
e
2
.
P
r
o
ce
s
s
to
lo
ca
te
th
e
r
eg
io
n
o
f
i
n
ter
est o
f
t
h
e
s
i
g
n
at
u
r
e
in
t
h
e
i
m
ag
e
I
n
th
is
w
a
y
,
th
e
r
eg
io
n
o
f
in
t
er
est
o
f
ea
ch
s
ig
n
at
u
r
e
is
f
o
u
n
d
,
f
r
o
m
w
h
ic
h
th
e
v
al
u
e
s
o
f
lo
ca
tio
n
o
f
th
e
v
er
tices
o
f
th
e
b
o
u
n
d
i
n
g
b
o
x
ar
e
o
b
tain
ed
,
co
m
p
ar
i
n
g
wh
at
v
er
tex
is
m
o
r
e
s
ep
ar
ated
f
r
o
m
th
e
ed
g
e
o
f
t
h
e
i
m
a
g
e.
T
ak
in
g
i
n
to
ac
co
u
n
t
t
h
is
s
ep
ar
atio
n
,
a
s
i
g
n
at
u
r
e
s
h
i
f
t
is
p
er
f
o
r
m
ed
,
w
h
o
s
e
v
alu
e
is
r
an
d
o
m
b
et
w
ee
n
th
e
p
o
in
t
o
f
t
h
e
v
er
tex
a
n
d
h
al
f
th
e
d
i
s
ta
n
ce
f
r
o
m
th
e
f
ar
th
e
s
t
ed
g
e,
b
o
th
in
th
e
h
ei
g
h
t
an
d
in
th
e
w
id
t
h
o
f
t
h
e
i
m
a
g
e.
T
h
is
p
r
o
ce
s
s
is
p
er
f
o
r
m
ed
at
leas
t
4
ti
m
e
s
i
n
ea
c
h
c
ateg
o
r
y
,
r
es
u
lti
n
g
in
a
to
tal
o
f
1
3
5
6
g
en
u
i
n
e
an
d
1
3
8
1
f
o
r
g
er
ies,
ad
d
in
g
4
6
1
s
ig
n
a
tu
r
es
n
o
t
b
elo
n
g
i
n
g
to
t
h
e
3
u
s
er
s
.
Fro
m
t
h
at
d
atab
as
e,
5
0
ar
e
r
an
d
o
m
l
y
ex
tr
ac
ted
f
r
o
m
ea
c
h
ca
te
g
o
r
y
to
ca
r
r
y
o
u
t
t
h
e
v
alid
atio
n
o
f
t
h
e
n
et
w
o
r
k
,
o
b
tain
i
n
g
a
d
ataset
o
f
2
,
8
4
8
s
ig
n
at
u
r
es
f
o
r
tr
ain
in
g
an
d
3
5
0
f
o
r
v
alid
atio
n
.
2
.
2
.
DAG
-
CNN
2
.
2
.
1
.
Arc
hite
ct
ure
T
o
p
er
f
o
r
m
th
e
v
er
if
icat
io
n
o
f
ea
ch
s
i
g
n
at
u
r
e,
a
DA
G
-
C
N
N
w
it
h
its
o
w
n
b
ili
n
ea
r
ar
ch
i
tectu
r
e
is
u
s
ed
,
i
n
s
u
c
h
a
w
a
y
th
at
a
v
er
y
d
ee
p
C
N
N
ca
n
b
e
ac
h
iev
ed
b
u
t
d
iv
id
ed
i
n
to
t
w
o
s
ec
tio
n
s
f
o
cu
s
ed
o
n
lear
n
in
g
d
if
f
er
e
n
t
ch
ar
ac
ter
is
tics
.
A
s
s
h
o
w
n
i
n
T
ab
le
1
,
ea
ch
d
iv
is
io
n
co
n
s
is
t
s
o
f
a
d
if
f
er
e
n
t
co
n
f
ig
u
r
atio
n
,
w
h
er
e
p
ath
1
,
b
y
th
e
s
izes
o
f
its
f
il
ter
s
,
is
f
o
cu
s
ed
o
n
lear
n
i
n
g
d
etails
o
f
th
e
s
i
g
n
at
u
r
es,
f
o
r
th
is
r
ea
s
o
n
,
it
also
co
n
s
is
ts
o
f
a
lar
g
e
n
u
m
b
er
o
f
f
ilter
s
in
all
la
y
er
s
to
b
e
ab
le
t
o
lear
n
en
o
u
g
h
f
ea
t
u
r
es.
O
n
t
h
e
o
th
er
h
an
d
,
p
ath
2
h
as
f
e
w
er
f
ilter
s
w
i
th
lar
g
e
r
s
izes
in
ea
c
h
la
y
er
,
w
h
ic
h
allo
w
s
t
h
i
s
d
iv
i
s
io
n
to
le
ar
n
m
o
r
e
g
e
n
er
al
ch
ar
ac
ter
is
tic
s
o
f
t
h
e
s
ig
n
at
u
r
e
s
,
s
u
c
h
a
s
its
d
i
m
e
n
s
io
n
s
a
n
d
s
ec
tio
n
s
b
y
lar
g
er
d
etail
s
.
Du
e
to
th
e
o
r
ig
i
n
al
s
ize
o
f
th
e
i
m
a
g
es
o
f
th
e
d
atase
t,
th
e
co
m
p
u
tatio
n
al
co
s
t
w
o
u
ld
b
e
v
er
y
lar
g
e,
f
o
r
th
is
r
ea
s
o
n
,
w
h
en
e
n
ter
i
n
g
t
h
e
n
et
w
o
r
k
,
it
w
as
r
ed
u
ce
d
to
a
s
ize
w
h
er
e
th
e
c
h
ar
ac
ter
is
t
ics
o
f
t
h
e
s
i
g
n
atu
r
e
w
er
e
n
o
t
lo
s
t,
b
ein
g
r
e
s
ized
to
3
0
%
o
f
its
s
ize,
r
e
m
ai
n
i
n
g
w
it
h
d
i
m
e
n
s
io
n
s
o
f
1
7
1
×
2
7
6
p
x
at
t
h
e
e
n
tr
a
n
ce
o
f
th
e
n
et
w
o
r
k
.
I
n
g
e
n
er
al,
th
e
ar
ch
itect
u
r
e
m
u
s
t
b
e
ab
le
to
class
if
y
7
d
if
f
er
en
t
ca
te
g
o
r
ies,
o
f
w
h
ic
h
ar
e:
a
ca
teg
o
r
y
o
f
g
en
u
i
n
e
s
i
g
n
atu
r
e
s
an
d
o
n
e
o
f
f
o
r
g
er
ies
f
o
r
ea
ch
u
s
er
,
h
a
v
i
n
g
6
ca
teg
o
r
ies
in
to
tal,
p
lu
s
an
ad
d
itio
n
al
o
n
e
o
f
s
ig
n
at
u
r
es
n
o
t
b
elo
n
g
i
n
g
to
t
h
e
d
atab
ase.
T
ab
le
1
.
DA
G
-
C
NN
A
r
c
h
itect
u
r
e
p
r
o
p
o
s
ed
.
I
n
th
e
co
n
f
ig
u
r
a
tio
n
o
f
t
h
e
la
y
er
s
,
F i
s
th
e
a
m
o
u
n
t o
f
f
i
lter
s
,
S r
ep
r
esen
ts
t
h
e
Strid
e
an
d
P
th
e
ze
r
o
p
ad
d
in
g
.
I
n
th
e
la
y
er
s
,
B
N
an
d
FC
ar
e
th
e
B
atch
No
r
m
aliza
tio
n
an
d
Fu
ll
y
C
o
n
n
ec
ted
la
y
er
s
,
r
e
s
p
ec
tiv
el
y
I
n
p
u
t
1
7
1
x
2
7
6
Pa
t
h
1
Pa
t
h
2
C
o
n
v
o
l
u
t
i
o
n
(
R
e
L
U
+
B
N
)
1
6
x
1
6
F
=
3
2
S
=
3
/
P
=
2
C
o
n
v
o
l
u
t
i
o
n
(
R
e
L
U
+
B
N
)
3
0
x
3
0
F
=
1
6
S
=
1
/
P
=
0
C
o
n
v
o
l
u
t
i
o
n
(
R
e
L
U
+
B
N
)
5
x
5
F
=
6
4
S
=
1
/
P
=
2
C
o
n
v
o
l
u
t
i
o
n
(
R
e
L
U
+
B
N
)
2
0
x
2
0
F
=
1
6
S
=
1
/
P
=
0
M
a
x
P
o
o
l
i
n
g
2
x
3
S
=
2
P
=
1
x
0
M
a
x
P
o
o
l
i
n
g
3
x
2
S
=
2
P
=
0
C
o
n
v
o
l
u
t
i
o
n
(
R
e
L
U
+
B
N
)
4
x
4
F
=
9
6
S
=
1
/
P
=
1
C
o
n
v
o
l
u
t
i
o
n
(
R
e
L
U
+
B
N
)
1
0
x
1
0
F
=
3
2
S
=
1
/
P
=
2
C
o
n
v
o
l
u
t
i
o
n
(
R
e
L
U
+
B
N
)
4
x
4
F
=
1
2
8
S
=
1
/
P
=
1
M
a
x
P
o
o
l
i
n
g
2
x
3
S
=
2
P
=
0
M
a
x
P
o
o
l
i
n
g
2
x
2
S
=
2
P
=
0
C
o
n
v
o
l
u
t
i
o
n
(
R
e
L
U
+
B
N
)
9
x
9
F
=
3
2
S
=
1
/
P
=
2
C
o
n
v
o
l
u
t
i
o
n
(
R
e
L
U
+
B
N
)
4
x
4
F
=
2
5
6
S
=
1
/
P
=
0
M
a
x
P
o
o
l
i
n
g
2
x
2
S
=
2
P
=
0
C
o
n
v
o
l
u
t
i
o
n
(
R
e
L
U
+
B
N
)
4
x
4
F
=
6
4
S
=
1
/
P
=
2
M
a
x
P
o
o
l
i
n
g
2
x
2
S
=
2
P
=
0
M
a
x
P
o
o
l
i
n
g
2
x
3
S
=
2
P
=
0
F
C
(
R
e
L
U
+
D
r
o
p
o
u
t
)
1
0
2
4
F
C
(
R
e
L
U
+
D
r
o
p
o
u
t
)
2
5
6
F
C
(
R
e
L
U
+
D
r
o
p
o
u
t
)
2
0
4
8
F
C
(
R
e
L
U
+
D
r
o
p
o
u
t
)
2
0
4
8
FC
7
S
o
f
t
M
a
x
C
l
a
ssi
f
i
c
a
t
i
o
n
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
n
t J
E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
9
,
No
.
4
,
A
u
g
u
s
t
201
9
:
3
3
1
4
-
332
2
3318
2
.
2
.
2
.
T
ra
ini
ng
p
ro
ce
s
s
Fo
r
th
e
ex
ec
u
tio
n
o
f
th
e
tr
ain
i
n
g
,
b
atch
e
s
o
f
3
2
s
ig
n
at
u
r
es
wer
e
u
s
ed
p
er
iter
atio
n
,
in
o
th
er
w
o
r
d
s
,
to
co
m
p
lete
a
n
ep
o
ch
,
8
9
iter
atio
n
s
w
er
e
n
ee
d
ed
,
s
i
n
ce
if
a
lar
g
er
a
m
o
u
n
t
o
f
d
ata
w
e
r
e
u
s
ed
p
er
b
atch
,
th
e
co
m
p
u
tatio
n
al
co
s
t
w
o
u
l
d
b
e
v
er
y
h
i
g
h
.
Du
e
to
t
h
is
,
w
e
ch
o
s
e
to
u
s
e
a
lear
n
i
n
g
r
ate
o
f
1
0
-
4
f
o
r
t
h
e
n
et
w
o
r
k
to
b
e
ab
le
to
lear
n
th
e
ch
ar
ac
ter
is
tics
m
o
r
e
g
en
er
all
y
an
d
p
r
ev
e
n
t
t
h
e
n
et
w
o
r
k
f
r
o
m
f
a
lli
n
g
i
n
to
a
s
tate
o
f
o
v
er
f
itti
n
g
,
in
ad
d
itio
n
,
th
e
o
r
d
er
o
f
th
e
s
ig
n
at
u
r
es
ch
an
g
es
r
a
n
d
o
m
l
y
in
ea
c
h
ep
o
ch
,
in
o
r
d
er
n
o
t
to
b
ias
t
h
e
n
et
w
o
r
k
to
lear
n
d
ata
s
o
r
ted
in
a
s
p
ec
if
ic
o
r
d
er
.
W
ith
t
h
e
s
e
p
ar
a
m
eter
s
,
t
h
e
D
AG
-
C
NN
w
a
s
tr
ai
n
ed
f
o
r
6
0
e
p
o
ch
s
,
s
h
o
w
in
g
t
h
e
b
eh
av
io
r
o
f
Fi
g
u
r
e
3
,
o
b
tain
in
g
an
ac
cu
r
ac
y
o
f
1
0
0
%
w
i
th
t
h
e
t
r
ain
in
g
d
ataset
an
d
9
9
.
4
3
%
w
it
h
th
e
v
alid
atio
n
,
w
h
ich
i
s
a
v
er
y
r
eliab
le
ac
cu
r
ac
y
i
n
ter
m
s
o
f
id
en
ti
f
icatio
n
o
f
f
o
r
g
er
ies,
i.e
.
o
f
th
e
3
5
0
v
alid
atio
n
s
ig
n
at
u
r
es,
it
o
n
l
y
h
ad
m
i
s
clas
s
i
f
icatio
n
i
n
2
s
ig
n
a
tu
r
es,
o
n
e
g
i
v
en
a
s
g
e
n
u
i
n
e
w
h
e
n
it
w
as
f
o
r
g
er
y
f
o
r
u
s
er
1
,
an
d
o
n
e
r
ec
o
g
n
ized
as
f
o
r
g
er
y
w
h
en
i
t
w
a
s
g
e
n
u
in
e
f
o
r
u
s
er
3
,
as
s
h
o
w
n
in
t
h
e
co
n
f
u
s
io
n
m
atr
i
x
o
f
Fi
g
u
r
e
4
.
L
ik
e
w
i
s
e,
all
th
e
s
i
g
n
atu
r
e
s
o
f
t
h
e
1
1
5
u
s
er
s
n
o
t
b
elo
n
g
i
n
g
to
th
e
v
er
if
iab
le
o
n
es
w
er
e
co
r
r
ec
tly
cla
s
s
i
f
ied
.
(
a)
(
b
)
Fig
u
r
e
3
.
T
r
ain
in
g
an
d
v
a
lid
ati
o
n
b
eh
av
io
r
o
f
t
h
e
D
A
G
-
C
NN
,
a)
A
cc
u
r
ac
y
,
b
)
L
o
s
s
Fig
u
r
e
4
.
C
o
n
f
u
s
io
n
Ma
tr
i
x
o
f
th
e
last
ep
o
ch
tr
ai
n
ed
o
f
th
e
n
et
w
o
r
k
u
s
in
g
t
h
e
Valid
atio
n
D
ataset
F a
r
e
th
e
Fo
r
g
er
y
a
n
d
G
ar
e
th
e
Gen
u
i
n
e
Si
g
n
at
u
r
es
3.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
NS
T
o
test
th
e
n
et
w
o
r
k
,
an
ad
d
iti
o
n
al
d
atab
ase
is
b
u
il
t,
ca
lled
T
est
d
ataset,
w
h
ich
co
n
s
i
s
ts
o
f
2
3
g
en
u
i
n
e
s
ig
n
at
u
r
es
an
d
2
3
f
o
r
g
er
ies
o
f
ea
ch
u
s
er
to
w
h
ic
h
a
d
ata
au
g
m
en
tatio
n
w
a
s
p
er
f
o
r
m
ed
w
ith
c
h
an
g
e
s
o
f
r
an
d
o
m
p
o
s
itio
n
,
i.e
.
in
t
h
e
s
a
m
e
w
a
y
to
th
e
a
u
g
m
e
n
tati
o
n
d
o
n
e
in
s
ec
tio
n
2
,
o
b
tain
in
g
a
to
tal
o
f
3
4
5
s
ig
n
at
u
r
es
p
er
ca
te
g
o
r
y
(
2
4
1
5
in
to
tal
o
f
t
h
e
d
atab
ase)
.
Fo
r
t
h
e
ca
te
g
o
r
y
o
f
o
t
h
er
s
,
t
h
e
1
1
5
u
s
er
s
o
f
th
e
UT
Si
g
Data
s
et
w
er
e
u
s
ed
,
b
u
t c
h
o
o
s
i
n
g
3
d
if
f
er
en
t sa
m
p
le
s
to
th
o
s
e
alr
ea
d
y
u
s
ed
in
t
h
e
p
r
ev
io
u
s
d
atab
ases
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8708
Offlin
e
s
ig
n
a
tu
r
e
ve
r
ifica
tio
n
u
s
in
g
DA
G
-
CNN
(
Ja
vier O
.
P
i
n
z
ó
n
-
A
r
en
a
s
)
3319
T
h
is
d
atab
ase
is
e
n
ter
ed
to
t
h
e
tr
ai
n
ed
n
et
w
o
r
k
to
p
er
f
o
r
m
t
h
e
test
,
in
w
h
ich
9
9
.
3
%
o
f
o
v
er
all
ac
cu
r
ac
y
i
s
o
b
tain
ed
,
as
s
h
o
wn
i
n
Fi
g
u
r
e
5
,
w
h
er
e
it
ca
n
b
e
s
ee
n
t
h
at
th
e
g
e
n
u
in
e
s
i
g
n
a
tu
r
es
o
f
al
l
u
s
er
s
w
er
e
r
ec
o
g
n
ized
w
it
h
p
er
ce
n
tag
e
s
g
r
ea
ter
th
a
n
9
8
%.
Ho
w
e
v
er
,
s
o
m
e
f
o
r
g
er
ies
m
an
a
g
ed
to
p
ass
a
s
g
en
u
i
n
e,
b
u
t
th
eir
co
r
r
ec
t
clas
s
i
f
icatio
n
w
as
h
i
g
h
er
th
a
n
9
8
.
5
%
i
n
all
ca
s
e
s
,
d
e
m
o
n
s
tr
ati
n
g
th
e
ef
f
ec
t
iv
e
n
es
s
o
f
t
h
e
n
et
w
o
r
k
in
v
er
i
f
y
in
g
t
h
e
u
s
er
s
r
eg
i
s
ter
ed
,
ev
en
h
a
v
i
n
g
f
o
r
g
er
ies
w
i
t
h
g
r
ea
t
s
i
m
i
lar
it
y
(
t
h
o
s
e
elab
o
r
ated
b
y
th
e
s
a
m
e
u
s
er
w
it
h
s
li
g
h
t c
h
a
n
g
es).
Fig
u
r
e
5
.
C
o
n
f
u
s
io
n
Ma
tr
i
x
u
s
in
g
t
h
e
T
est Da
taset
T
o
u
n
d
er
s
tan
d
t
h
e
in
ter
n
al
o
p
er
atio
n
o
f
t
h
e
n
et
w
o
r
k
,
th
e
s
t
r
o
n
g
est
ac
ti
v
atio
n
s
o
f
ea
c
h
c
o
n
v
o
l
u
tio
n
ar
e
ex
tr
ac
ted
af
ter
th
e
ap
p
li
ca
tio
n
o
f
R
eL
U
(
R
ec
ti
f
ied
L
i
n
ea
r
Un
i
t,
w
h
o
s
e
f
u
n
c
tio
n
is
to
eli
m
in
ate
t
h
e
n
eg
at
iv
e
v
al
u
es
o
f
t
h
e
o
u
tp
u
t
v
o
lu
m
e)
,
in
s
u
c
h
a
w
a
y
t
h
at
it
ca
n
b
e
o
b
s
er
v
ed
h
o
w
th
e
n
et
w
o
r
k
b
eh
a
v
es
w
i
th
th
e
i
m
a
g
e
th
r
o
u
g
h
it.
I
n
Fig
u
r
e
6
,
it
ca
n
s
ee
th
e
d
if
f
er
e
n
t
ac
tiv
atio
n
s
o
f
th
e
t
w
o
d
iv
i
s
io
n
s
o
f
t
h
e
n
et
w
o
r
k
.
P
ath
1
s
h
o
w
s
ac
ti
v
atio
n
s
m
ai
n
l
y
f
o
c
u
s
ed
o
n
t
h
e
in
ter
n
al
d
etails
o
f
t
h
e
s
i
g
n
at
u
r
e,
co
n
to
u
r
an
d
s
h
ap
e
in
th
e
la
s
t
t
w
o
co
n
v
o
l
u
tio
n
s
,
w
h
ile
p
ath
2
f
o
c
u
s
es
lar
g
el
y
o
n
t
h
e
g
en
e
r
al
f
o
r
m
o
f
th
e
s
i
g
n
atu
r
e,
e
m
p
ty
s
ec
t
io
n
s
w
it
h
i
n
th
e
co
n
to
u
r
a
n
d
th
e
b
o
u
n
d
ar
ies o
f
p
o
s
s
ib
le
s
i
g
n
a
tu
r
e
lo
ca
tio
n
(
co
n
v
o
l
u
tio
n
2
)
.
I
n
a
m
o
r
e
d
etailed
w
a
y
to
v
i
s
u
alize
th
e
ac
ti
v
atio
n
s
o
v
er
t
h
e
s
ig
n
atu
r
e,
h
ea
t
m
ap
s
o
v
er
lap
p
ed
o
n
th
e
o
r
i
g
in
al
i
m
ag
e
ar
e
u
s
ed
.
I
n
th
o
s
e
m
ap
s
,
th
e
s
tr
o
n
g
e
s
t
ac
ti
v
at
io
n
s
ar
e
r
ep
r
ese
n
ted
b
y
r
ed
co
lo
r
an
d
t
h
e
w
ea
k
es
t
o
n
es
b
y
d
ar
k
b
lu
e.
Fo
r
th
is
,
t
h
e
b
est
s
e
v
en
ac
tiv
a
tio
n
s
o
b
tain
ed
i
n
t
h
e
f
ir
s
t
la
y
er
o
f
ea
c
h
p
ath
ar
e
s
h
o
w
n
i
n
Fig
u
r
e
7
.
W
ith
in
th
e
p
ath
1
,
th
e
f
ilte
r
s
w
er
e
ab
le
to
lear
n
f
in
er
d
etails
o
f
th
e
s
i
g
n
at
u
r
e,
n
o
t
o
n
l
y
r
elate
d
to
th
e
s
h
ap
e
b
u
t
o
f
d
i
f
f
er
e
n
t
k
e
y
p
ar
ts
,
s
u
ch
a
s
t
h
e
f
ir
s
t
let
ter
o
f
t
h
e
s
ig
n
at
u
r
e,
as
ca
n
b
e
s
ee
n
i
n
t
h
e
last
ac
ti
v
atio
n
o
f
th
e
b
o
tto
m
,
an
d
s
p
ec
if
ic
tr
ac
es
th
at
ca
n
d
if
f
er
en
tiate
a
f
o
r
g
e
r
y
f
r
o
m
a
g
e
n
u
in
e,
s
u
c
h
as
t
h
e
th
ir
d
ac
tiv
atio
n
at
th
e
to
p
an
d
th
e
b
o
tto
m
.
On
th
e
o
th
er
h
a
n
d
,
in
p
ath
2
,
as
ex
p
ec
ted
,
g
en
er
alize
th
e
“
a
p
p
ea
r
an
ce
”
o
f
th
e
s
ig
n
at
u
r
e,
ac
ti
v
ati
n
g
in
d
ep
en
d
en
tl
y
th
e
lo
w
er
,
ce
n
ter
a
n
d
to
p
p
ar
t
o
f
th
e
s
i
g
n
a
tu
r
e,
al
s
o
it
ac
ti
v
ate
s
e
m
p
t
y
r
eg
io
n
s
a
s
i
n
th
e
s
ec
o
n
d
co
n
v
o
l
u
tio
n
la
y
er
,
as
s
h
o
w
n
i
n
Fig
u
r
e
6
.
T
h
e
ab
o
v
e
s
h
o
w
s
th
at
t
h
e
p
r
o
p
o
s
ed
ar
ch
itect
u
r
e
o
b
tain
ed
th
e
b
eh
a
v
io
r
th
at
w
as d
es
ir
ed
,
d
escr
ib
e
d
in
s
ec
tio
n
3
.
Fig
u
r
e
6
.
A
cti
v
atio
n
s
o
f
ea
c
h
co
n
v
o
lu
tio
n
.
A
t t
h
e
to
p
,
th
e
p
ath
1
;
at
th
e
b
o
tto
m
,
t
h
e
p
ath
2
.
T
h
e
co
n
v
o
lu
tio
n
s
ar
e
s
o
r
ter
f
r
o
m
le
f
t (
co
n
v
o
lu
tio
n
1
)
to
r
ig
h
t (
co
n
v
o
lu
t
io
n
5
)
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
n
t J
E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
9
,
No
.
4
,
A
u
g
u
s
t
201
9
:
3
3
1
4
-
332
2
3320
(
a)
(
b
)
Fig
u
r
e
7
.
A
cti
v
atio
n
s
o
f
t
h
e
f
ir
s
t c
o
n
v
o
l
u
tio
n
la
y
er
u
s
i
n
g
o
v
e
r
lap
p
ed
h
ea
t m
ap
s
,
a)
P
ath
1
,
b
)
P
ath
2
4.
CO
NCLU
SI
O
NS
T
h
is
p
ap
er
p
r
esen
te
d
t
h
e
ap
p
li
ca
tio
n
o
f
a
D
A
G
-
C
N
N
f
o
cu
s
e
d
o
n
t
h
e
o
f
f
lin
e
s
i
g
n
at
u
r
e
v
er
i
f
icatio
n
,
as
an
ad
v
an
ce
i
n
th
e
u
s
e
o
f
Dee
p
L
ea
r
n
i
n
g
p
atter
n
r
ec
o
g
n
itio
n
tech
n
iq
u
es
i
n
th
i
s
t
y
p
e
o
f
tas
k
,
g
iv
i
n
g
a
n
et
w
o
r
k
ca
p
ab
le
o
f
class
i
f
y
i
n
g
a
n
d
au
t
h
en
t
icate
at
t
h
e
s
a
m
e
ti
m
e
,
s
h
o
w
i
n
g
th
e
p
er
f
o
r
m
an
ce
o
f
t
h
i
s
t
y
p
e
o
f
n
e
t
w
o
r
k
.
Usi
n
g
a
D
AG
-
C
NN
al
lo
w
ed
i
t
to
lear
n
d
if
f
er
en
t
c
h
ar
ac
ter
is
tics
o
f
t
h
e
s
i
g
n
at
u
r
es,
as
s
h
o
w
n
i
n
Fi
g
u
r
e
6
an
d
Fig
u
r
e
7
,
w
h
er
e,
d
ep
en
d
i
n
g
o
n
t
h
e
t
y
p
e
o
f
co
n
f
i
g
u
r
atio
n
th
at
is
s
et
i
n
ea
c
h
b
r
an
c
h
,
th
e
n
et
w
o
r
k
f
o
c
u
s
ed
o
n
ce
r
tain
s
ec
tio
n
s
an
d
d
etai
ls
o
f
th
e
s
i
g
n
a
tu
r
e
i
n
o
r
d
er
to
b
e
a
b
le
to
d
if
f
er
en
tiate,
f
ir
s
t,
w
h
et
h
er
it
b
elo
n
g
ed
o
r
n
o
t
to
r
e
g
is
ter
ed
u
s
er
s
,
an
d
t
h
en
v
er
i
f
y
t
h
e
g
e
n
u
in
e
n
es
s
o
f
it.
T
h
e
r
es
u
lts
o
b
tain
ed
a
n
d
th
e
an
al
y
s
is
o
f
th
e
b
eh
av
io
r
o
f
th
e
tr
ai
n
ed
n
et
w
o
r
k
,
w
h
er
e
an
ac
cu
r
ac
y
g
r
ea
ter
th
an
9
9
%
w
as
o
b
tain
ed
,
allo
w
u
s
to
v
er
i
f
y
t
h
e
ef
f
icien
c
y
o
f
t
h
e
C
NN,
i
n
th
i
s
ca
s
e
o
f
a
DAG
-
C
NN,
f
o
r
th
e
r
ec
o
g
n
it
io
n
o
f
t
h
e
g
e
n
u
i
n
en
es
s
o
f
s
i
g
n
at
u
r
es
o
f
f
li
n
e.
On
t
h
e
o
th
er
h
an
d
,
to
ac
h
iev
e
ev
en
a
h
i
g
h
er
p
er
ce
n
tag
e,
it
ca
n
b
e
ad
d
ed
o
n
lin
e
d
ata
co
llectio
n
m
et
h
o
d
s
an
d
a
n
al
y
s
i
s
o
f
th
e
s
e
,
m
a
k
in
g
r
ea
l
-
ti
m
e
v
er
i
f
icatio
n
m
u
c
h
m
o
r
e
r
eliab
le,
i.e
.
at
t
h
e
ti
m
e
o
f
s
i
g
n
i
n
g
,
h
o
w
ev
er
,
it
r
eq
u
ir
es
th
e
u
s
e
o
f
s
p
ec
ialized
d
ev
ices
to
p
er
f
o
r
m
s
aid
d
ata
ac
q
u
is
itio
n
,
w
h
ic
h
ca
n
b
e
ap
p
lied
in
f
u
tu
r
e
w
o
r
k
.
ACK
NO
WL
E
D
G
E
M
E
NT
S
T
h
e
r
esear
ch
f
o
r
th
is
p
ap
er
w
as
s
u
p
p
o
r
ted
b
y
Dav
i
n
ci
R
e
s
e
ar
ch
Gr
o
u
p
o
f
Nu
ev
a
Gr
an
ad
a
Militar
y
Un
i
v
er
s
it
y
.
RE
F
E
R
E
NC
E
S
[1
]
D.
I
m
p
e
d
o
v
o
a
n
d
G
.
P
irl
o
,
"
A
u
to
m
a
ti
c
si
g
n
a
tu
re
v
e
ri
f
ic
a
ti
o
n
:
T
h
e
sta
te
o
f
th
e
a
rt,
"
EE
E
T
ra
n
sa
c
ti
o
n
s
o
n
S
y
ste
ms
,
M
a
n
,
a
n
d
Cy
b
e
rn
e
ti
c
s,
Pa
rt
C
(
Ap
p
li
c
a
ti
o
n
s
a
n
d
Rev
iews
)
,
v
o
l.
3
8
,
n
o
5
,
p
p
.
6
0
9
-
6
3
5
,
2
0
0
8
.
DO
I:
1
0
.
1
1
0
9
/T
S
M
CC.
2
0
0
8
.
9
2
3
8
6
6
[2
]
A
.
Kh
o
l
m
a
to
v
a
n
d
B.
Ya
n
ik
o
g
lu
,
"
Id
e
n
ti
ty
a
u
th
e
n
ti
c
a
ti
o
n
u
sin
g
i
m
p
ro
v
e
d
o
n
li
n
e
sig
n
a
tu
re
v
e
ri
f
ic
a
ti
o
n
m
e
th
o
d
,
"
Pa
tt
e
rn
re
c
o
g
n
it
i
o
n
letter
s
,
v
o
l.
2
6
,
n
o
1
5
,
p
p
.
2
4
0
0
-
2
4
0
8
,
2
0
0
5
.
D
OI:
1
0
.
1
0
1
6
/j
.
p
a
trec
.
2
0
0
5
.
0
4
.
0
1
7
[3
]
J.
Bro
m
le
y
,
I.
G
u
y
o
n
,
Y.
L
e
Cu
n
,
E.
S
ä
c
k
in
g
e
r
a
n
d
R.
S
h
a
h
,
"
S
ig
n
a
tu
re
v
e
rif
ic
a
ti
o
n
u
sin
g
a
"
sia
m
e
se
"
ti
m
e
d
e
la
y
n
e
u
ra
l
n
e
tw
o
rk
,
"
Ad
v
a
n
c
e
s in
Ne
u
ra
l
I
n
fo
rm
a
ti
o
n
Pr
o
c
e
ss
in
g
S
y
ste
ms
,
p
p
.
7
37
-
7
4
4
,
1
9
9
4
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8708
Offlin
e
s
ig
n
a
tu
r
e
ve
r
ifica
tio
n
u
s
in
g
DA
G
-
CNN
(
Ja
vier O
.
P
i
n
z
ó
n
-
A
r
en
a
s
)
3321
[4
]
L
.
G
.
Ha
f
e
m
a
n
n
,
R.
S
a
b
o
u
rin
a
n
d
L
.
S
.
Oliv
e
ira,
"
L
e
a
rn
in
g
f
e
a
tu
re
s
f
o
r
o
ff
li
n
e
h
a
n
d
w
rit
ten
sig
n
a
tu
re
v
e
rif
ic
a
ti
o
n
u
sin
g
d
e
e
p
c
o
n
v
o
l
u
ti
o
n
a
l
n
e
u
ra
l
n
e
tw
o
rk
s,"
Pa
tt
e
rn
R
e
c
o
g
n
it
i
o
n
,
v
o
l.
7
0
,
p
p
.
1
6
3
-
1
7
6
,
2
0
1
7
.
DO
I:
1
0
.
1
0
1
6
/
j.
p
a
tco
g
.
2
0
1
7
.
0
5
.
0
1
2
[5
]
S.
N.
S
rih
a
ri,
A
.
X
u
a
n
d
M
.
K.
Ka
lera
,
"
Lea
rn
in
g
stra
teg
i
e
s
a
n
d
c
las
sif
ica
ti
o
n
m
e
th
o
d
s
f
o
r
o
ff
-
li
n
e
sig
n
a
tu
re
v
e
ri
f
ica
ti
o
n
,
"
in
Fr
o
n
ti
e
rs
i
n
h
a
n
d
writi
n
g
re
c
o
g
n
i
ti
o
n
,
2
0
0
4
.
IW
FHR
-
9
2
0
0
4
.
N
in
t
h
i
n
ter
n
a
ti
o
n
a
l
wo
rk
sh
o
p
o
n
,
p
p
.
1
6
1
-
166
,
2
0
0
4
.
DO
I:
1
0
.
1
1
0
9
/IW
F
HR.
2
0
0
4
.
6
1
[6
]
A
.
S
o
lei
m
a
n
i,
B.
N.
A
ra
a
b
i
a
n
d
K.
F
o
u
la
d
i,
"
De
e
p
m
u
lt
it
a
sk
m
e
tri
c
lea
rn
in
g
f
o
r
o
ff
li
n
e
sig
n
a
tu
re
v
e
ri
f
ica
ti
o
n
,
"
Pa
tt
e
rn
Rec
o
g
n
it
io
n
L
e
tt
e
rs
,
v
o
l
.
8
0
,
p
p
.
8
4
-
9
0
,
2
0
1
6
.
DO
I:
1
0
.
1
0
1
6
/j
.
p
a
trec
.
2
0
1
6
.
0
5
.
0
2
3
[7
]
G
.
S
.
Esk
a
n
d
e
r,
R.
S
a
b
o
u
ri
n
a
n
d
E
.
G
ra
n
g
e
r,
"
H
y
b
rid
w
rit
e
r
-
in
d
e
p
e
n
d
e
n
t
–
w
rit
e
r
-
d
e
p
e
n
d
e
n
t
o
f
f
li
n
e
sig
n
a
tu
re
v
e
ri
f
ica
ti
o
n
sy
ste
m
,
"
IET
b
io
me
trics
,
v
o
l.
2
,
n
o
4
,
p
p
.
1
6
9
-
1
8
1
,
2
0
1
3
.
DO
I:
1
0
.
1
0
4
9
/
iet
-
b
m
t.
2
0
1
3
.
0
0
2
[8
]
H.
C.
F
u
,
H.
Y.
C
h
a
n
g
,
Y.
Y.
Xu
,
a
n
d
H.
T
.
P
a
o
,
"
Us
e
r
a
d
a
p
ti
v
e
h
a
n
d
w
rit
in
g
re
c
o
g
n
it
i
o
n
b
y
s
e
l
f
-
g
ro
w
in
g
p
ro
b
a
b
il
isti
c
d
e
c
isio
n
-
b
a
se
d
n
e
u
ra
l
n
e
tw
o
rk
s,"
IEE
E
T
ra
n
sa
c
ti
o
n
s
o
n
n
e
u
r
a
l
n
e
two
rk
s
,
v
o
l.
1
1
,
n
o
6
,
p
p
.
1
3
7
3
-
1
3
8
4
,
2
0
0
0
.
DO
I:
1
0
.
1
1
0
9
/
7
2
.
8
8
3
4
5
1
[9
]
L
.
X
in
g
a
n
d
Y.
Qia
o
,
"
De
e
p
w
r
it
e
r:
A
m
u
lt
i
-
stre
a
m
d
e
e
p
CNN
f
o
r
tex
t
-
in
d
e
p
e
n
d
e
n
t
w
rit
e
r
id
e
n
ti
f
ica
ti
o
n
,
"
i
n
Fro
n
ti
e
rs
in
Ha
n
d
writ
in
g
Rec
o
g
n
it
i
o
n
(
ICFHR),
2
0
1
6
1
5
th
I
n
t
e
rn
a
ti
o
n
a
l
C
o
n
fer
e
n
c
e
o
n
,
2
0
1
6
,
p
p
.
5
8
4
-
5
8
9
.
DO
I:
1
0
.
1
1
0
9
/
ICF
HR.2
0
1
6
.
0
1
1
2
[1
0
]
A
.
M
c
C
a
b
e
,
J.
T
r
e
v
a
th
a
n
a
n
d
W
.
Re
a
d
,
"
Ne
u
ra
l
n
e
t
w
o
rk
-
b
a
se
d
h
a
n
d
w
rit
ten
sig
n
a
tu
re
v
e
rif
ic
a
ti
o
n
,
"
J
o
u
r
n
a
l
o
f
c
o
mp
u
ter
s
,
v
o
l.
3
,
p
p
.
9
-
2
2
,
2
0
0
8
.
DO
I:
1
0
.
4
3
0
4
/j
c
p
.
3
.
8
.
9
-
22
[1
1
]
Y.
L
e
Cu
n
,
Y.
Be
n
g
io
a
n
d
G
.
Hin
to
n
,
"
De
e
p
lea
rn
in
g
,"
Na
tu
re
,
v
o
l.
5
2
1
,
n
o
7
5
5
3
,
p
p
.
4
3
6
-
4
4
4
,
2
0
1
5
.
DO
I:
1
0
.
1
0
3
8
/
n
a
tu
re
1
4
5
3
9
[1
2
]
M
.
D.
Zeiler
a
n
d
R.
F
e
rg
u
s,
"
V
isu
a
li
z
in
g
a
n
d
U
n
d
e
rsta
n
d
in
g
Co
n
v
o
lu
ti
o
n
a
l
Ne
tw
o
rk
s,"
in
Eu
ro
p
e
a
n
c
o
n
fer
e
n
c
e
o
n
c
o
mp
u
ter
v
isio
n
,
S
p
ri
n
g
e
r,
Ch
a
m
,
p
p
.
8
1
8
-
8
3
3
,
2
0
1
4
.
DO
I:
1
0
.
1
0
0
7
/9
7
8
-
3
-
3
1
9
-
1
0
5
9
0
-
1
_
5
3
[1
3
]
P
.
M
a
rz
u
k
i,
e
t
a
l.
,
"
A
d
e
si
g
n
o
f
li
c
e
n
se
p
late
re
c
o
g
n
it
io
n
sy
ste
m
u
s
in
g
c
o
n
v
o
lu
ti
o
n
a
l
n
e
u
ra
l
n
e
tw
o
rk
"
,
In
ter
n
a
ti
o
n
a
l
J
o
u
rn
a
l
o
f
El
e
c
trica
l
a
n
d
Co
mp
u
ter
E
n
g
i
n
e
e
rin
g
(
IJ
ECE
)
,
v
o
l.
9
,
n
o
3
,
p
p
.
2
1
9
6
~
2
2
0
4
,
2
0
1
9
.
DO
I:
1
0
.
1
1
5
9
1
/i
jec
e
.
v
9
i3
.
p
p
2
1
9
6
-
2
2
0
4
[1
4
]
S
.
S
u
d
h
o
l
t,
a
n
d
G
.
F
i
n
k
,
"
P
HO
CNe
t:
A
d
e
e
p
c
o
n
v
o
l
u
ti
o
n
a
l
n
e
u
ra
l
n
e
tw
o
rk
f
o
r
w
o
rd
sp
o
tt
i
n
g
in
h
a
n
d
w
rit
ten
d
o
c
u
m
e
n
ts,
"
i
n
Fro
n
ti
e
rs
in
Ha
n
d
writi
n
g
Rec
o
g
n
it
io
n
(
ICFHR),
2
0
1
6
1
5
th
In
ter
n
a
ti
o
n
a
l
Co
n
fer
e
n
c
e
o
n
,
IEE
E,
p
p
.
2
7
7
-
282
,
2
0
1
6
.
DO
I:
1
0
.
1
1
0
9
/ICF
HR.2
0
1
6
.
0
0
6
0
.
[1
5
]
F
.
W
o
lf
,
De
n
se
ly
Co
n
n
e
c
ted
Co
n
v
o
lu
ti
o
n
a
l
Ne
tw
o
rk
s
fo
r
W
o
rd
S
p
o
tt
in
g
i
n
H
a
n
d
wri
tt
e
n
D
o
c
u
me
n
ts
,
M
a
ste
r
T
h
e
sis,
Co
m
p
u
ter S
c
ien
c
e
De
p
a
rtm
e
n
t,
Tec
h
n
ica
l
Un
iv
e
rsity
o
f
Do
rt
m
u
n
d
,
Do
rtm
u
n
d
,
G
e
rm
a
n
y
,
2
0
1
8
.
[1
6
]
U.
S
h
u
k
la
a
n
d
V.
Bh
a
teja
,
"
Co
m
p
a
riso
n
b
e
tw
e
e
n
CNN
a
n
d
AN
N
in
o
f
f
li
n
e
sig
n
a
tu
re
v
e
ri
f
ica
t
io
n
,"
i
n
S
e
c
o
n
d
In
ter
n
a
t
io
n
a
l
C
o
n
fer
e
n
c
e
o
n
C
o
mp
u
ti
n
g
,
Co
mm
u
n
ica
t
io
n
a
n
d
Co
n
t
ro
l
T
e
c
h
n
o
l
o
g
y
,
p
p
.
1
3
6
-
1
4
0
,
2
0
1
8
.
[1
7
]
L
.
G
.
Ha
fe
m
a
n
n
,
R.
S
a
b
o
u
r
in
a
n
d
L
.
S
.
Oliv
e
ira,
"
A
n
a
l
y
z
in
g
fe
a
tu
re
s lea
rn
e
d
f
o
r
o
ff
li
n
e
sig
n
a
tu
re
v
e
rif
ica
ti
o
n
u
sin
g
De
e
p
CNN
s
,"
i
n
2
0
1
6
2
3
r
d
In
ter
n
a
ti
o
n
a
l
Co
n
fer
e
n
c
e
o
n
Pa
tt
e
rn
Rec
o
g
n
it
io
n
(
ICPR
)
,
IEE
E,
p
p
.
2
9
8
9
-
2
9
9
4
,
2
0
1
6
.
DO
I:
1
0
.
1
1
0
9
/
IC
P
R.
2
0
1
6
.
7
9
0
0
0
9
2
[1
8
]
L
.
G
.
Ha
f
e
m
a
n
n
,
R.
S
a
b
o
u
rin
a
n
d
L
.
S
.
Oliv
e
ira,
"
W
rit
e
r
-
in
d
e
p
e
n
d
e
n
t
f
e
a
tu
re
lea
rn
in
g
f
o
r
o
ff
li
n
e
sig
n
a
tu
re
v
e
ri
f
ica
ti
o
n
u
si
n
g
d
e
e
p
c
o
n
v
o
l
u
ti
o
n
a
l
n
e
u
ra
l
n
e
tw
o
rk
s,"
Ne
u
ra
l
n
e
two
rk
s
(
IJ
CNN),
2
0
1
6
I
n
ter
n
a
ti
o
n
a
l
J
o
in
t
Co
n
fer
e
n
c
e
o
n
,
IEE
E,
p
p
.
2
5
7
6
-
2
5
8
3
,
2
0
1
6
.
DO
I:
1
0
.
1
1
0
9
/IJCNN
.
2
0
1
6
.
7
7
2
7
5
2
1
[1
9
]
M
.
B.
Yilm
a
z
a
n
d
K.
Öz
tü
rk
,
"
H
y
b
rid
Us
e
r
-
In
d
e
p
e
n
d
e
n
t
a
n
d
Us
e
r
-
De
p
e
n
d
e
n
t
Off
li
n
e
S
ig
n
a
tu
re
V
e
ri
f
ica
ti
o
n
w
it
h
a
Tw
o
-
Ch
a
n
n
e
l
CNN
,
"
i
n
2
0
1
8
I
EE
E/
CVF
Co
n
fer
e
n
c
e
o
n
Co
m
p
u
ter
Vi
sio
n
a
n
d
Pa
t
ter
n
Rec
o
g
n
it
io
n
W
o
rk
sh
o
p
s
(
CVP
RW
)
,
IEE
E,
p
p
.
6
3
9
-
6
3
9
8
,
2
0
1
8
.
DO
I:
1
0
.
1
1
0
9
/C
VPRW
.
2
0
1
8
.
0
0
0
9
4
[2
0
]
T
.
Yo
u
n
e
sia
n
,
S
.
M
a
so
u
d
n
ia,
R
.
Ho
ss
e
in
i
a
n
d
B.
N.
A
ra
a
b
i,
"
A
c
ti
v
e
T
ra
n
s
f
e
r
L
e
a
rn
in
g
f
o
r
P
e
rsia
n
Off
li
n
e
S
ig
n
a
tu
re
V
e
rif
ica
ti
o
n
,
"
a
rXiv
p
re
p
rin
t
a
rXiv:1
9
0
3
.
0
6
2
5
5
,
2
0
1
9
.
[2
1
]
Z.
J.
X
in
g
,
F
.
Yi
n
,
Y.
C.
W
u
a
n
d
C.
L
.
L
iu
,
"
O
ff
li
n
e
sig
n
a
tu
re
v
e
ri
f
ica
ti
o
n
u
s
in
g
c
o
n
v
o
lu
ti
o
n
S
iam
e
se
n
e
t
w
o
rk
,
"
i
n
Nin
th
In
ter
n
a
ti
o
n
a
l
C
o
n
fer
e
n
c
e
o
n
Gr
a
p
h
ic
a
n
d
Ima
g
e
Pro
c
e
ss
in
g
(
ICGIP
2
0
1
7
)
,
I
n
tern
a
ti
o
n
a
l
S
o
c
iety
f
o
r
Op
ti
c
s
a
n
d
P
h
o
t
o
n
ics
,
p
p
.
1
0
6
1
5
1
I
,
2
0
1
8
.
DO
I:
1
0
.
1
1
1
7
/1
2
.
2
3
0
3
3
8
0
[2
2
]
K.
T
h
u
las
iram
a
n
a
n
d
M
.
N.
S
w
a
m
y
,
Gr
a
p
h
s: t
h
e
o
ry
a
n
d
a
lg
o
rit
h
ms
,
Jo
h
n
W
il
e
y
&
S
o
n
s,
2
0
1
1
.
[2
3
]
A
.
S
o
lei
m
a
n
i,
K.
F
o
u
lad
i
a
n
d
B.
N.
A
r
a
a
b
i,
"
U
T
S
ig
:
A
P
e
rsia
n
o
ff
li
n
e
sig
n
a
tu
re
d
a
tas
e
t,
"
IET
Bi
o
me
trics
,
v
o
l.
6
,
n
o
1
,
p
p
.
1
-
8
,
2
0
1
6
.
DO
I:
1
0
.
1
0
4
9
/i
e
t
-
b
m
t.
2
0
1
5
.
0
0
5
8
[2
4
]
M
.
S
o
n
k
a
,
V
.
Hla
v
a
c
a
n
d
R.
B
o
y
le,
Ima
g
e
p
ro
c
e
ss
in
g
,
a
n
a
lys
is,
a
n
d
ma
c
h
i
n
e
v
isio
n
,
Ce
n
g
a
g
e
L
e
a
rn
i
n
g
,
2
0
1
4
.
B
I
O
G
RAP
H
I
E
S
O
F
AUTH
O
RS
J
a
v
ier
O
r
la
n
d
o
Pi
n
z
ó
n
A
r
e
n
a
s
w
a
s
b
o
rn
in
S
o
c
o
rr
o
-
S
a
n
tan
d
e
r,
Co
lo
m
b
ia,
in
1
9
9
0
.
He
re
c
e
i
v
e
d
h
is
d
e
g
re
e
in
M
e
c
h
a
tro
n
ics
En
g
i
n
e
e
rin
g
(Cu
m
L
a
u
d
e
)
a
n
d
S
p
e
c
ializa
ti
o
n
in
En
g
in
e
e
rin
g
P
r
o
jec
t
M
a
n
a
g
e
m
e
n
t
a
t
th
e
Nu
e
v
a
Gra
n
a
d
a
M
il
it
a
r
y
Un
iv
e
rsit
y
-
UMN
G
-
in
2
0
1
3
a
n
d
2
0
1
6
,
re
sp
e
c
ti
v
e
l
y
.
He
h
a
s
e
x
p
e
rien
c
e
in
th
e
a
re
a
s
o
f
a
u
to
m
a
ti
o
n
,
e
lec
tro
n
ic
c
o
n
tro
l
a
n
d
m
a
c
h
in
e
lea
rn
in
g
.
Cu
rre
n
tl
y
,
h
e
is
stu
d
y
in
g
f
o
r
a
M
a
ste
r’s
d
e
g
r
e
e
in
M
e
c
h
a
tro
n
ics
En
g
in
e
e
rin
g
a
n
d
w
o
rk
in
g
a
s
Re
se
a
rc
h
As
sist
a
n
t
a
t
t
h
e
UMNG
w
it
h
e
m
p
h
a
sis
o
n
R
o
b
o
ti
c
s
a
n
d
M
a
c
h
in
e
L
e
a
rn
in
g
.
E
-
m
a
il
:
u
3
9
0
0
2
3
1
@u
n
i
m
il
it
a
r.
e
d
u
.
c
o
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
n
t J
E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
9
,
No
.
4
,
A
u
g
u
s
t
201
9
:
3
3
1
4
-
332
2
3322
Ro
b
i
n
s
o
n
J
i
m
é
n
e
z
M
o
r
e
n
o
w
a
s
b
o
r
n
i
n
B
o
g
o
tá,
C
o
lo
m
b
ia,
i
n
1
9
7
8
.
He
re
c
e
iv
e
d
th
e
En
g
i
n
e
e
r
d
e
g
re
e
in
El
e
c
tro
n
ics
a
t
t
h
e
F
ra
n
c
isc
o
Jo
sé
d
e
Ca
ld
a
s District
Un
iv
e
rsit
y
-
UD
-
in
2
0
0
2
.
M
.
S
c
.
i
n
In
d
u
strial
A
u
to
m
a
ti
o
n
f
ro
m
th
e
Un
i
v
e
rsid
a
d
Na
c
io
n
a
l
d
e
Co
l
o
m
b
ia
-
2
0
1
2
a
n
d
P
h
D
i
n
En
g
in
e
e
rin
g
a
t
th
e
F
ra
n
c
isc
o
Jo
s
é
d
e
Ca
ld
a
s
District
Un
iv
e
rsity
-
2
0
1
8
.
He
is
c
u
rre
n
tl
y
w
o
rk
in
g
a
s
a
P
ro
f
e
ss
o
r
in
th
e
M
e
c
h
a
tro
n
ics
En
g
in
e
e
rin
g
P
r
o
g
ra
m
a
t
th
e
Nu
e
v
a
G
ra
n
a
d
a
M
il
it
a
r
y
Un
iv
e
rsit
y
-
UMN
G
.
He
h
a
s
e
x
p
e
rien
c
e
in
th
e
a
re
a
s
o
f
In
stru
m
e
n
tatio
n
a
n
d
El
e
c
tro
n
ic
Co
n
tro
l
,
a
c
ti
n
g
m
a
in
ly
in
:
Ro
b
o
ti
c
s,
c
o
n
tro
l,
p
a
tt
e
rn
re
c
o
g
n
it
i
o
n
a
n
d
im
a
g
e
p
ro
c
e
ss
in
g
.
E
-
m
a
il
:
ro
b
in
so
n
.
ji
m
e
n
e
z
@u
n
im
il
it
a
r.
e
d
u
.
c
o
Cé
sa
r
G
io
v
a
n
y
Pa
c
h
ó
n
S
u
e
s
c
ú
n
w
a
s
b
o
rn
in
B
o
g
o
tá,
Co
l
o
m
b
ia,
in
1
9
9
6
.
He
re
c
e
iv
e
d
h
is
d
e
g
re
e
in
M
e
c
h
a
tro
n
ics
En
g
in
e
e
r
in
g
f
ro
m
th
e
P
il
o
t
Un
iv
e
rsity
o
f
Co
lo
m
b
ia
in
2
0
1
8
.
Cu
rre
n
tl
y
,
h
e
is
stu
d
y
in
g
h
is
M
a
ste
r’s
d
e
g
re
e
in
M
e
c
h
a
tro
n
ics
En
g
in
e
e
ri
n
g
a
n
d
w
o
rk
in
g
a
s
Re
se
a
rc
h
A
s
sista
n
t
a
t
th
e
Nu
e
v
a
G
ra
n
a
d
a
M
il
it
a
ry
Un
iv
e
r
sit
y
w
it
h
e
m
p
h
a
s
is
o
n
Ro
b
o
ti
c
s
a
n
d
M
a
c
h
i
n
e
L
e
a
rn
in
g
.
E
-
m
a
il
:
u
3
9
0
0
2
5
9
@
u
n
i
m
il
it
a
r.
e
d
u
.
c
o
Evaluation Warning : The document was created with Spire.PDF for Python.