I
nte
rna
t
io
na
l J
o
urna
l o
f
E
lect
rica
l a
nd
Co
m
pu
t
er
E
ng
ineering
(
I
J
E
CE
)
Vo
l.
15
,
No
.
4
,
A
u
g
u
s
t
20
25
,
p
p
.
3
7
2
7
~
3
7
3
6
I
SS
N:
2088
-
8
7
0
8
,
DOI
: 1
0
.
1
1
5
9
1
/ijece.
v
15
i
4
.
pp
3
7
2
7
-
3
7
3
6
3727
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
ec
e.
ia
esco
r
e.
co
m
Enha
ncing
a
no
ma
ly
det
ection
perf
o
rma
nce using
Re
sNet50
and
BiL
STM
network
s o
n benchma
rk
d
a
tas
ets
Dipa
k
Ra
mo
liy
a
1
,
Am
i
t
G
a
n
a
t
ra
2
1
D
e
p
a
r
t
me
n
t
o
f
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
a
n
d
E
n
g
i
n
e
e
r
i
n
g
,
D
e
v
a
n
g
P
a
t
e
l
I
n
st
i
t
u
t
e
o
f
A
d
v
a
n
c
e
T
e
c
h
n
o
l
o
g
y
a
n
d
R
e
s
e
a
r
c
h
(
D
EPS
TA
R
)
,
F
a
c
u
l
t
y
o
f
T
e
c
h
n
o
l
o
g
y
a
n
d
E
n
g
i
n
e
e
r
i
n
g
,
C
h
a
r
o
t
a
r
U
n
i
v
e
r
si
t
y
o
f
S
c
i
e
n
c
e
a
n
d
T
e
c
h
n
o
l
o
g
y
(
C
H
A
R
U
S
A
T)
,
A
n
a
n
d
,
I
n
d
i
a
2
D
e
p
a
r
t
me
n
t
o
f
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
a
n
d
E
n
g
i
n
e
e
r
i
n
g
,
F
a
c
u
l
t
y
o
f
E
n
g
i
n
e
e
r
i
n
g
a
n
d
T
e
c
h
n
o
l
o
g
y
,
P
a
r
u
l
U
n
i
v
e
r
si
t
y
(
P
U
)
,
V
a
d
o
d
a
r
a
,
I
n
d
i
a
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Au
g
3
,
2
0
2
4
R
ev
is
ed
Ma
r
2
4
,
2
0
2
5
Acc
ep
ted
Ma
y
2
3
,
2
0
2
5
De
tec
ti
o
n
o
f
a
b
n
o
rm
a
l
a
c
ti
v
it
y
fro
m
larg
e
v
id
e
o
se
q
u
e
n
c
e
s
is
o
n
e
o
f
t
h
e
b
ig
g
e
st
c
h
a
ll
e
n
g
e
s
b
e
c
a
u
se
o
f
a
m
b
ig
u
it
y
in
d
iffere
n
t
a
c
ti
v
it
ies
.
O
v
e
r
th
e
las
t
m
a
n
y
y
e
a
rs,
se
v
e
ra
l
c
a
m
e
ra
s
h
a
v
e
b
e
e
n
p
lac
e
d
t
o
c
o
v
e
r
t
h
e
p
u
b
li
c
a
n
d
p
riv
a
te
se
c
to
rs
t
o
m
o
n
it
o
r
a
b
n
o
rm
a
l
h
u
m
a
n
a
c
ti
v
it
y
a
n
d
su
r
v
e
i
ll
a
n
c
e
.
In
re
c
e
n
t
y
e
a
rs,
d
e
e
p
lea
rn
in
g
a
n
d
c
o
m
p
u
ter
v
isi
o
n
h
a
v
e
si
g
n
ifi
c
a
n
tl
y
imp
a
c
ted
th
is
k
in
d
o
f
su
r
v
e
il
lan
c
e
.
In
telli
g
e
n
t
sy
ste
m
s
t
h
a
t
c
a
n
a
u
t
o
m
a
ti
c
a
ll
y
i
d
e
n
ti
f
y
u
n
u
s
u
a
l
e
v
e
n
ts
in
v
id
e
o
stre
a
m
s
a
re
c
u
rre
n
t
ly
in
h
ig
h
d
e
m
a
n
d
.
A
d
e
e
p
lea
rn
in
g
-
b
a
se
d
c
o
m
b
i
n
a
ti
o
n
a
l
m
o
d
e
l
h
a
s
b
e
e
n
p
r
o
p
o
se
d
t
o
d
e
tec
t
a
b
n
o
rm
a
l
a
c
ti
v
it
y
fro
m
in
p
u
t
v
id
e
o
stre
a
m
s.
Th
e
p
r
o
p
o
se
d
st
u
d
y
u
se
s
a
c
o
m
b
in
a
ti
o
n
o
f
c
o
n
v
o
lu
ti
o
n
a
n
d
se
q
u
e
n
ti
a
l
m
o
d
e
ls.
A
Re
sN
e
t5
0
n
e
two
r
k
wit
h
a
re
sid
u
a
l
c
o
n
n
e
c
ti
o
n
wa
s
u
se
d
fo
r
i
n
it
ial
fe
a
tu
re
e
x
trac
ti
o
n
.
T
h
e
p
r
o
p
o
se
d
b
i
d
irec
ti
o
n
a
l
lo
n
g
sh
o
rt
-
term
m
e
m
o
ry
(
BiL
S
TM
)
n
e
tw
o
rk
h
a
s
imp
r
o
v
e
d
t
h
e
e
x
trac
ted
Re
sN
e
t5
0
fe
a
tu
re
s.
S
imu
lati
o
n
o
f
th
e
p
r
o
p
o
se
d
m
o
d
e
l
wa
s
e
x
p
e
ri
m
e
n
ted
o
n
two
b
e
n
c
h
m
a
rk
d
a
tas
e
ts
fo
r
a
n
o
m
a
ly
d
e
tec
ti
o
n
UCF
C
rime
a
n
d
S
h
a
n
g
h
a
iT
e
c
h
.
S
imu
lati
o
n
o
f
p
r
o
p
o
se
d
a
rc
h
it
e
c
tu
re
h
a
s
a
c
h
iev
e
d
9
7
.
5
5
%
a
n
d
9
1
.
9
4
%
re
m
a
rk
a
b
le
a
c
c
u
ra
c
y
fo
r
UCF
Crime
a
n
d
S
h
a
n
g
h
a
iT
e
c
h
d
a
tas
e
ts
re
sp
e
c
ti
v
e
ly
.
K
ey
w
o
r
d
s
:
An
o
m
aly
d
etec
tio
n
B
iLST
M
C
o
m
p
u
ter
v
is
io
n
T
r
an
s
f
er
lear
n
i
n
g
Vid
eo
s
u
r
v
eillan
ce
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
:
Dip
ak
R
am
o
liy
a
Dep
ar
tm
en
t o
f
C
o
m
p
u
ter
Scie
n
ce
an
d
E
n
g
in
ee
r
in
g
,
De
v
an
g
Patel
I
n
s
titu
te
o
f
Ad
v
an
ce
T
ec
h
n
o
lo
g
y
an
d
R
esear
ch
(
DE
PS
T
AR
)
,
Facu
lt
y
o
f
T
ec
h
n
o
l
o
g
y
a
n
d
E
n
g
in
ee
r
in
g
,
C
h
ar
o
tar
U
n
iv
er
s
ity
o
f
Scien
ce
an
d
T
ec
h
n
o
lo
g
y
(
C
HARUS
AT
)
C
h
an
g
a,
An
an
d
-
3
8
8
4
2
1
,
I
n
d
i
a
E
m
ail: d
ip
ak
r
am
o
liy
a.
ce
@
ch
ar
u
s
at.
ac
.
in
1.
I
NT
RO
D
UCT
I
O
N
I
n
s
u
r
v
eillan
ce
a
p
p
licatio
n
s
,
a
n
o
m
alo
u
s
ac
tiv
ity
d
etec
tio
n
p
lay
s
a
cr
u
cial
r
o
le.
I
t
is
p
o
s
s
ib
le
to
u
s
e
au
to
m
atic
v
i
d
eo
ca
p
tu
r
e
to
r
e
co
r
d
an
o
m
al
o
u
s
h
u
m
an
b
e
h
av
io
r
with
o
u
t
th
e
n
ee
d
f
o
r
s
y
s
tem
in
ter
v
en
tio
n
[
1
]
.
Du
e
to
its
co
m
p
lex
ity
,
th
e
co
u
n
tles
s
p
o
ten
tial
f
o
r
ab
er
r
an
t
ca
s
es,
an
d
th
e
s
ca
r
city
o
f
a
v
ailab
le
n
o
r
m
al
o
n
es,
it
r
em
ain
s
o
n
e
o
f
t
h
e
m
o
s
t
ch
allen
g
in
g
a
n
d
well
-
k
n
o
wn
r
esear
ch
ar
ea
s
[
2
]
.
Su
r
v
eillan
ce
c
am
er
as
ar
e
alm
o
s
t
ev
er
y
wh
er
e
t
h
ese
d
ay
s
,
f
r
o
m
s
ec
u
r
ity
-
s
en
s
itiv
e
lo
ca
tio
n
s
—
s
u
ch
as
b
o
r
d
e
r
s
an
d
m
ilit
ar
y
i
n
s
tallatio
n
s
u
s
ed
to
m
o
n
ito
r
ter
r
o
r
is
t
ac
tiv
it
y
—
to
p
r
iv
ate
r
esid
en
c
es
m
ea
n
t
t
o
d
eter
b
u
r
g
lar
ies.
Du
e
to
wea
r
in
ess
an
d
m
o
n
o
to
n
y
,
m
an
u
al
wo
r
k
e
r
s
wer
e
u
n
ab
le
t
o
m
ain
tain
d
ilig
en
t
m
o
n
ito
r
in
g
o
v
er
an
e
x
ten
d
ed
le
n
g
th
o
f
tim
e
d
u
e
to
th
e
lar
g
e
v
o
lu
m
e
o
f
d
ata
i
n
v
o
lv
e
d
.
T
r
ad
itio
n
ally
,
th
e
m
o
n
ito
r
in
g
f
u
n
ctio
n
is
h
an
d
led
b
y
h
u
m
a
n
o
p
er
ato
r
s
.
T
h
ese
o
p
er
ato
r
s
h
av
e
to
s
im
u
ltan
eo
u
s
ly
v
is
u
ally
ev
alu
ate
m
an
y
ca
m
er
a
f
ee
d
s
.
T
ec
h
n
o
lo
g
ical
ad
v
an
ce
m
e
n
ts
an
d
f
allin
g
co
s
ts
h
av
e
led
to
an
ac
ce
ler
atio
n
o
f
th
e
in
s
tallatio
n
o
f
s
u
r
v
eillan
ce
ca
m
er
as
in
b
o
th
p
u
b
lic
an
d
p
r
iv
at
e
s
p
ac
es
[
3
]
.
T
h
e
m
ain
p
r
o
b
le
m
with
th
is
a
p
p
r
o
ac
h
is
t
h
at
th
e
d
etec
tio
n
r
ate
o
f
an
o
m
ali
es
d
ec
r
ea
s
es
r
ap
id
ly
wh
en
u
s
ed
in
c
r
o
wd
ed
en
v
ir
o
n
m
en
ts
wh
en
o
cc
lu
s
io
n
s
an
d
cl
u
tter
ar
is
e.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
15
,
No
.
4
,
Au
g
u
s
t
20
25
:
3
7
2
7
-
3736
3728
An
o
m
aly
d
etec
tio
n
,
a
cr
itical
task
in
v
ar
io
u
s
d
o
m
ain
s
s
u
ch
as
f
in
an
ce
,
cy
b
er
s
ec
u
r
ity
,
in
d
u
s
tr
ial
m
o
n
ito
r
in
g
an
d
h
ea
lth
ca
r
e
in
v
o
lv
es
r
ec
o
g
n
izi
n
g
p
atter
n
s
in
d
ata
th
at
d
ev
iate
m
ea
n
in
g
f
u
lly
f
r
o
m
t
h
e
n
o
r
m
.
T
h
ese
an
o
m
alies
ca
n
in
d
icate
s
ig
n
if
ican
t
b
u
t
r
ar
e
e
v
en
ts
lik
e
f
r
au
d
,
s
y
s
tem
f
ailu
r
es,
o
r
u
n
u
s
u
al
m
ed
ica
l
co
n
d
itio
n
s
,
n
ec
ess
itatin
g
tim
ely
an
d
ac
cu
r
ate
d
etec
tio
n
to
m
itig
ate
p
o
ten
tial
r
is
k
s
an
d
i
m
p
ac
ts
.
T
r
ad
itio
n
al
m
eth
o
d
s
f
o
r
a
n
o
m
aly
d
etec
tio
n
o
f
ten
s
tr
u
g
g
le
with
th
e
co
m
p
lex
ity
an
d
h
ig
h
d
im
e
n
s
io
n
ali
ty
o
f
m
o
d
er
n
d
ata,
m
ak
in
g
a
d
v
an
ce
d
tech
n
iq
u
es e
s
s
en
tial f
o
r
ef
f
ec
tiv
e
an
aly
s
is
[
4
]
.
L
o
n
g
s
h
o
r
t
-
ter
m
m
em
o
r
y
(
L
STM
)
n
etwo
r
k
s
,
a
f
o
r
m
o
f
r
ec
u
r
r
en
t
n
e
u
r
al
n
etwo
r
k
(
R
NN)
,
h
av
e
d
em
o
n
s
tr
ated
e
x
tr
ao
r
d
in
ar
y
s
u
cc
ess
in
s
eq
u
en
tial
d
ata
p
r
o
c
ess
in
g
d
u
e
to
th
eir
ca
p
ac
ity
t
o
ca
p
tu
r
e
l
o
n
g
-
te
r
m
d
ep
en
d
e
n
cies.
Ho
wev
er
,
s
tan
d
ar
d
L
STM
m
o
d
els
p
r
o
c
ess
d
ata
in
a
s
in
g
le
tem
p
o
r
al
d
ir
ec
tio
n
,
p
o
ten
tially
m
is
s
in
g
co
n
tex
tu
al
in
f
o
r
m
ati
o
n
th
at
c
o
u
ld
e
n
h
an
ce
an
o
m
aly
d
etec
tio
n
ac
c
u
r
ac
y
[
4
]
.
B
id
ir
ec
tio
n
al
L
STM
(
B
iLST
M)
n
etwo
r
k
s
o
v
e
r
co
m
e
th
is
p
r
o
b
lem
b
y
p
r
o
ce
s
s
in
g
d
ata
in
b
o
th
f
o
r
war
d
a
n
d
b
a
ck
war
d
d
i
r
ec
tio
n
s
,
th
u
s
lev
er
ag
in
g
th
e
co
m
p
lete
t
em
p
o
r
a
l c
o
n
tex
t
[
5
]
.
An
o
m
aly
d
etec
tio
n
u
s
in
g
d
ee
p
lear
n
in
g
p
r
esen
ts
s
ev
er
al
s
ig
n
if
ican
t
ch
all
en
g
es,
d
esp
ite
its
p
o
ten
tial
to
o
u
tp
e
r
f
o
r
m
tr
ad
itio
n
al
m
et
h
o
d
s
.
On
e
o
f
t
h
e
p
r
i
m
ar
y
c
h
allen
g
es
is
th
e
r
eq
u
i
r
em
en
t
f
o
r
lar
g
e
am
o
u
n
ts
o
f
lab
elled
tr
ain
in
g
d
ata
,
wh
ich
i
s
o
f
ten
s
ca
r
ce
in
r
ea
l
-
wo
r
l
d
s
ce
n
ar
io
s
[
6
]
.
A
n
o
m
alies,
b
y
n
atu
r
e,
ar
e
r
a
r
e
an
d
d
iv
er
s
e,
m
ak
in
g
it
d
if
f
icu
lt
to
g
ath
er
a
r
e
p
r
esen
tativ
e
an
d
s
u
f
f
icien
tly
lar
g
e
d
ataset
f
o
r
tr
ai
n
in
g
d
ee
p
lear
n
in
g
m
o
d
els.
Ad
d
itio
n
ally
,
d
ee
p
lea
r
n
in
g
m
o
d
els,
esp
ec
ially
th
o
s
e
with
co
m
p
lex
ar
ch
itectu
r
es,
a
r
e
co
m
p
u
tatio
n
ally
in
ten
s
iv
e
an
d
r
eq
u
ir
e
s
u
b
s
tan
t
ial
r
eso
u
r
ce
s
f
o
r
b
o
th
tr
ain
in
g
an
d
in
f
er
en
ce
[
4
]
,
[
5
]
.
T
h
is
ca
n
b
e
a
b
ar
r
ier
to
d
ep
lo
y
m
e
n
t
in
en
v
ir
o
n
m
e
n
ts
wi
th
lim
ited
co
m
p
u
tatio
n
al
c
ap
ab
ilit
ies.
Mo
r
eo
v
er
,
d
ee
p
l
ea
r
n
in
g
m
o
d
els
ca
n
s
o
m
etim
es
s
tr
u
g
g
le
with
in
ter
p
r
etab
ilit
y
,
m
a
k
in
g
it
h
ar
d
to
u
n
d
er
s
tan
d
wh
y
a
p
a
r
ticu
lar
in
s
tan
ce
was
class
if
ied
as
an
an
o
m
aly
,
wh
i
ch
is
cr
u
cial
in
m
an
y
ap
p
licatio
n
s
wh
er
e
tr
an
s
p
a
r
en
c
y
an
d
ex
p
lain
ab
ilit
y
a
r
e
im
p
o
r
tan
t.
Fu
r
th
er
m
o
r
e,
th
es
e
m
o
d
els
a
r
e
s
en
s
itiv
e
to
h
y
p
er
p
a
r
am
eter
t
u
n
in
g
an
d
ar
ch
itectu
r
e
d
esig
n
,
r
eq
u
ir
in
g
ex
ten
s
iv
e
ex
p
er
im
en
tatio
n
an
d
e
x
p
er
tis
e
to
ac
h
iev
e
o
p
tim
al
p
er
f
o
r
m
a
n
ce
[
4
]
.
I
n
th
is
ar
ticle,
th
e
au
th
o
r
s
h
av
e
p
r
o
p
o
s
ed
an
ad
v
an
ce
d
ar
c
h
itectu
r
e
f
o
r
an
o
m
aly
d
etec
tio
n
.
T
h
e
m
ajo
r
co
n
tr
ib
u
tio
n
s
o
f
th
e
ar
ticle
ar
e
i)
p
r
o
p
o
s
ed
a
d
v
an
ce
d
co
n
v
o
l
u
tio
n
an
d
B
I
L
STM
-
b
ased
s
eq
u
en
tial
ar
ch
itectu
r
e
f
o
r
an
o
m
aly
d
etec
tio
n
,
ii)
s
i
m
u
latio
n
o
f
th
e
p
r
o
p
o
s
ed
s
tu
d
y
ex
p
lo
r
e
t
h
e
ef
f
ec
tiv
en
ess
o
f
th
e
s
tu
d
y
b
y
co
m
p
ar
in
g
th
e
SOTA
d
ee
p
lear
n
in
g
m
o
d
el
f
o
r
a
n
o
m
aly
d
etec
tio
n
.
T
h
e
p
r
o
p
o
s
ed
s
t
u
d
y
co
n
tr
ib
u
tes
to
im
p
r
o
v
in
g
v
id
e
o
s
u
r
v
eillan
ce
i
n
s
en
s
itiv
e
s
ec
to
r
s
.
T
h
e
r
est
o
f
th
e
a
r
ticle
is
o
r
g
an
ized
as
s
ec
tio
n
2
d
em
o
n
s
tr
ates
liter
atu
r
e
an
aly
s
is
an
d
co
m
p
ar
ativ
e
an
aly
s
is
o
f
r
ec
en
t
wo
r
k
o
n
an
o
m
aly
d
etec
tio
n
u
s
in
g
d
ee
p
le
ar
n
in
g
m
et
h
o
d
s
.
Sectio
n
3
p
r
e
s
en
ts
th
e
p
r
o
p
o
s
ed
B
iLST
M
ar
ch
itectu
r
e
with
r
esid
u
al
co
n
n
ec
tio
n
s
.
Sectio
n
4
illu
s
tr
ates
th
e
s
im
u
latio
n
o
f
th
e
p
r
o
p
o
s
ed
ar
ch
itectu
r
e
an
d
s
tatis
tica
l a
n
a
ly
s
is
with
SOTA
d
ee
p
lear
n
in
g
m
o
d
els o
v
er
d
if
f
er
en
t
b
en
ch
m
ar
k
d
atasets
.
2.
L
I
T
E
R
AT
U
RE
SU
RVE
Y
T
h
er
e
ar
e
two
ty
p
es
o
f
m
eth
o
d
s
f
o
r
id
e
n
tify
in
g
u
n
u
s
u
al
a
ctiv
ity
:
lo
w
-
lev
el
an
o
m
aly
d
e
tectio
n
an
d
h
ig
h
-
lev
el
a
n
o
m
aly
d
etec
tio
n
.
L
o
w
-
lev
el
an
o
m
aly
d
etec
tio
n
tech
n
iq
u
es
lo
ca
te
lo
ca
l
s
p
at
io
tem
p
o
r
al
r
e
g
io
n
s
th
at
lik
ely
h
av
e
a
b
er
r
an
t
l
o
w
-
lev
el
f
ea
tu
r
e
p
atter
n
s
b
ef
o
r
e
h
i
g
h
-
lev
el
an
al
y
s
is
,
in
clu
d
in
g
ac
tiv
ity
class
if
icatio
n
an
d
o
b
ject
tr
ac
k
in
g
,
is
ca
r
r
ied
o
u
t.
[
7
]
in
ter
-
ac
tiv
ity
c
o
n
tex
t
f
ea
tu
r
e
-
b
ased
ar
ch
ite
ctu
r
e
p
r
o
p
o
s
ed
b
y
Z
hu
et
a
l.
[
7
]
.
T
h
e
p
r
o
p
o
s
e
d
s
tu
d
y
u
s
es
a
g
r
ee
d
y
s
ea
r
c
h
alg
o
r
ith
m
f
o
r
p
o
in
t
-
b
ased
co
n
tex
tu
al
an
o
m
al
y
d
etec
tio
n
.
VI
R
AT
g
r
o
u
n
d
d
ata
s
et
u
s
ed
f
o
r
th
e
s
im
u
latio
n
o
f
t
h
e
p
r
o
p
o
s
ed
s
tu
d
y
.
Ng
u
y
en
an
d
Me
u
n
ier
[
8
]
p
r
o
p
o
s
ed
a
U
-
Net
-
b
ased
ar
ch
itectu
r
e
f
o
r
an
o
m
al
y
d
etec
tio
n
f
r
o
m
v
id
eo
s
eq
u
en
ce
s
.
T
h
e
p
r
o
p
o
s
ed
s
tu
d
y
u
s
ed
c
o
n
v
o
lu
tio
n
ar
c
h
itectu
r
e
in
clu
d
i
n
g
4
s
tr
ea
m
s
o
f
co
n
v
o
lu
tio
n
s
o
f
f
ilte
r
s
izes
1
×1
,
3
×3
,
5
×5
an
d
7
×
7
f
o
r
f
ea
tu
r
e
e
x
tr
ac
tio
n
.
Sim
u
latio
n
o
f
th
e
p
r
o
p
o
s
ed
s
tu
d
y
u
s
ed
b
en
ch
m
a
r
k
d
atasets
lik
e
Av
en
u
e
a
n
d
UC
SD
Ped
2
.
T
h
e
p
r
o
p
o
s
ed
s
tu
d
y
was
ab
le
to
ac
h
iev
e
0
.
8
6
9
an
d
0
.
9
6
2
ac
cu
r
ac
y
r
esp
ec
tiv
ely
.
A
f
aster
R
-
C
NN
-
b
ased
Dee
p
An
o
m
aly
m
o
d
el
was
p
r
o
p
o
s
e
d
b
y
C
h
r
is
tian
s
en
et
a
l
.
[
9
]
.
to
id
en
tify
u
n
u
s
u
al
ac
tiv
ity
in
f
ilm
s
.
T
h
e
s
u
g
g
ested
ar
ch
itectu
r
e
ca
n
id
en
tify
p
eo
p
le
u
p
to
9
0
m
eter
s
awa
y
.
T
o
id
e
n
tify
an
o
m
alies,
p
r
e
-
tr
ai
n
co
n
v
o
lu
t
io
n
m
o
d
els
wer
e
em
p
lo
y
ed
.
T
h
e
s
u
g
g
ested
s
tu
d
y
m
ak
es
u
s
e
o
f
a
p
r
o
p
o
s
ed
ap
p
r
o
ac
h
th
at
o
f
f
er
s
b
etwe
en
3
0
0
a
n
d
2
,
0
0
0
r
eg
i
o
n
s
p
e
r
im
ag
e.
Su
b
s
eq
u
en
tly
,
a
ca
teg
o
r
izatio
n
n
etwo
r
k
d
esig
n
ates
a
lab
el
f
o
r
ev
er
y
r
eg
io
n
an
d
r
o
u
tes
it
v
ia
it.
Dee
p
An
o
m
aly
is
a
h
elp
f
u
l
tech
n
iq
u
e
f
o
r
r
eg
io
n
s
u
g
g
esti
o
n
wh
en
th
e
r
e
ar
e
f
ew
o
r
n
o
ar
ea
s
p
er
i
m
ag
e.
Sab
o
k
r
o
u
et
a
l.
[
1
0
]
p
r
o
p
o
s
ed
an
a
u
to
en
co
d
er
(
AE
)
b
as
ed
m
o
d
el
f
o
r
a
n
o
m
aly
d
etec
tio
n
.
T
h
e
p
r
o
p
o
s
ed
ar
c
h
itectu
r
e
f
i
n
d
s
er
r
o
r
s
in
ab
n
o
r
m
al
p
ar
ts
in
t
h
e
r
ec
o
n
s
tr
u
ctio
n
p
h
ase
o
f
s
p
ar
s
e
AE
.
Simu
latio
n
o
f
p
r
o
p
o
s
ed
ar
ch
itectu
r
e
u
s
es
UC
SD
Ped
2
an
d
UM
N
b
en
ch
m
ar
k
d
atasets
an
d
p
atch
s
ize
with
3
0
×
30
x
1
0
Pro
p
o
s
ed
s
tu
d
y
m
a
r
k
s
p
at
ch
es
as
ab
n
o
r
m
al
if
m
o
r
e
th
a
n
4
0
%
p
ix
els
ar
e
d
etec
ted
.
Simu
latio
n
f
i
n
d
s
9
9
.
6
AUC with
an
8
2
% d
etec
tio
n
r
ate.
Kh
an
et
a
l.
[
1
1
]
p
r
o
p
o
s
ed
co
n
v
o
lu
tio
n
-
b
ased
ar
ch
itectu
r
e
f
o
r
m
o
n
ito
r
i
n
g
o
f
ir
r
eg
u
lar
h
u
m
an
ac
tio
n
s
,
an
d
tr
af
f
ic
s
u
r
v
eillan
ce
.
Au
th
o
r
s
h
av
e
also
co
n
tr
ib
u
ted
to
s
u
r
v
eillan
ce
d
atasets
s
u
ch
as
t
h
e
v
eh
icle
ac
ci
d
en
t
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
-
8
7
0
8
E
n
h
a
n
cin
g
a
n
o
ma
ly
d
etec
tio
n
p
erfo
r
ma
n
ce
u
s
in
g
R
esN
et5
0
a
n
d
…
(
Dip
a
k
R
a
m
o
liya
)
3729
im
ag
e
d
ataset
(
VAI
D)
.
T
h
e
d
ataset
co
n
tain
s
o
f
1
3
6
0
ac
cid
en
tal
im
ag
es
o
f
v
eh
icles.
T
h
e
au
t
h
o
r
s
h
a
v
e
ex
ten
d
ed
th
e
s
im
u
latio
n
o
f
th
e
p
r
o
p
o
s
ed
ar
ch
itectu
r
e
with
d
if
f
er
en
t ty
p
es o
f
p
o
o
lin
g
,
ac
tiv
at
io
n
,
an
d
o
p
tim
izer
m
eth
o
d
s
.
T
h
e
p
r
o
p
o
s
ed
s
tu
d
y
f
in
d
s
8
2
%
ac
c
u
r
ac
y
u
s
in
g
m
o
d
if
ie
d
co
n
v
o
lu
tio
n
al
n
e
u
r
al
n
etwo
r
k
(
C
NN)
m
o
d
els.
L
e
an
d
Kim
[
1
2
]
p
r
o
p
o
s
ed
a
v
id
eo
an
o
m
aly
d
etec
tio
n
s
y
s
te
m
em
p
lo
y
in
g
a
n
atten
tio
n
-
b
as
ed
r
esid
u
al
au
to
en
co
d
er
ar
ch
itectu
r
e
.
T
h
e
s
u
g
g
ested
n
etwo
r
k
u
s
es
b
o
t
h
s
p
atial
an
d
tem
p
o
r
al
in
f
o
r
m
atio
n
in
a
s
in
g
le,
in
teg
r
ated
n
etwo
r
k
an
d
is
b
a
s
ed
o
n
u
n
s
u
p
e
r
v
is
ed
lear
n
in
g
.
Simu
latio
n
o
f
th
e
p
r
o
p
o
s
ed
s
tu
d
y
u
s
es
th
r
ee
d
if
f
er
en
t
b
en
ch
m
ar
k
d
atasets
UC
SD
Ped
2
,
C
UHK,
an
d
Sh
an
g
h
aitec
h
.
Simu
latio
n
o
f
th
e
p
r
o
p
o
s
ed
s
tu
d
y
f
in
d
s
AUC
v
alu
es
as
7
3
.
6
%,
8
6
.
7
%
an
d
9
7
.
4
%
f
o
r
Sh
an
g
h
aiT
ec
h
,
C
UHK
Av
en
u
e,
an
d
U
C
SD
Ped
2
d
ataset
s
r
esp
ec
tiv
ely
.
Av
o
la
et
a
l.
[
1
3
]
p
r
o
p
o
s
ed
an
o
m
aly
d
etec
tio
n
b
ased
o
n
g
en
er
ativ
e
ad
v
er
s
ar
ial
n
etwo
r
k
(
GAN)
.
T
h
e
p
r
o
p
o
s
ed
s
tu
d
y
u
s
es
GAN
ar
ch
itectu
r
e
f
o
r
ae
r
ial
v
id
eo
s
u
r
v
ei
llan
ce
at
lo
w
altitu
d
es.
Pro
p
o
s
ed
s
tu
d
y
f
o
cu
s
es
o
n
lo
w
-
altitu
d
e
s
eq
u
en
c
es
in
co
m
p
lex
s
ce
n
ar
io
s
wh
er
ein
a
ti
n
y
o
b
ject
o
r
g
ad
g
et
m
ay
b
e
ca
u
s
e
f
o
r
co
n
ce
r
n
o
r
d
an
g
er
.
Simu
latio
n
o
f
th
e
s
tu
d
y
u
s
es
b
en
ch
m
ar
k
ch
an
g
e
d
etec
tio
n
(
UM
C
D)
an
d
u
n
m
an
n
ed
ae
r
ial
v
eh
icle
(
UAV)
I
m
o
s
aick
in
g
d
ataset.
Simu
latio
n
o
f
GAN
ar
ch
itectu
r
e
f
in
d
s
9
5
.
7
%
s
tr
u
ctu
r
al
s
im
il
ar
ity
in
d
e
x
(
SS
I
M)
.
T
ab
le
1
p
r
esen
ts
a
co
m
p
ar
ativ
e
ex
am
in
atio
n
o
f
v
ar
io
u
s
d
ee
p
lear
n
in
g
m
eth
o
d
o
l
o
g
ies
f
o
r
an
o
m
aly
id
en
tific
atio
n
[
1
4
]
–
[
1
8
]
.
T
ab
le
1
.
C
o
m
p
a
r
ativ
e
an
aly
s
is
with
o
th
er
d
ee
p
lear
n
i
n
g
ap
p
r
o
ac
h
es f
o
r
a
n
o
m
aly
d
etec
tio
n
[
1
4
]
–
[
1
8
]
A
u
t
h
o
r
s
Y
e
a
r
M
o
d
e
l
D
a
t
a
s
e
t
R
e
s
u
l
t
Za
h
e
e
r
e
t
a
l
.
[
1
4
]
2
0
2
0
O
n
e
-
C
l
a
ss
C
l
a
ss
i
f
i
e
r
U
C
S
D
P
e
d
2
9
8
.
1
%
Ta
n
g
e
t
a
l
.
[
1
5
]
2
0
2
0
U
-
N
e
t
A
v
e
n
u
e
8
5
.
1
%
Zh
o
u
e
t
a
l
.
[
1
6
]
2
0
1
9
LSTM
A
v
e
n
u
e
8
6
.
1
%
W
a
n
e
t
a
l
.
[
1
7
]
2
0
2
1
LSTM
U
C
F
C
r
i
me
9
5
.
1
2
%
D
o
sh
i
a
n
d
Y
i
l
ma
z
[
1
8
]
2
0
2
3
Y
O
LO
-
v4
A
v
e
n
u
e
8
5
.
7
8
%
3.
M
E
T
H
O
DO
L
O
G
Y
T
h
e
p
r
o
p
o
s
ed
s
tu
d
y
u
s
es
a
co
m
b
in
atio
n
al
a
p
p
r
o
ac
h
f
o
r
an
o
m
aly
d
etec
tio
n
.
T
h
e
p
r
o
p
o
s
ed
s
tu
d
y
u
s
ed
R
esNet
ar
ch
itectu
r
e
f
o
r
f
ea
tu
r
e
ex
tr
ac
tio
n
f
o
llo
wed
b
y
B
iLST
M
[
4
]
to
en
h
an
ce
f
ea
tu
r
es
b
ased
o
n
f
ee
d
b
ac
k
lear
n
in
g
.
T
h
e
ar
c
h
itectu
r
e
o
f
th
e
p
r
o
p
o
s
ed
s
tu
d
y
is
m
ain
l
y
d
iv
id
e
d
in
to
t
h
r
ee
m
ajo
r
c
ateg
o
r
ies
as
f
ea
tu
r
e
ex
tr
ac
tio
n
with
R
esNet5
0
,
f
ea
tu
r
e
en
h
an
ce
m
e
n
t
with
B
iLST
M,
an
d
class
if
icatio
n
with
m
u
lti
-
lay
er
p
er
ce
p
tr
o
n
(
ML
P)
[
1
9
]
.
Fig
u
r
e
1
d
em
o
n
s
tr
ates
th
e
co
n
ce
p
tu
al
a
r
ch
itectu
r
e
o
f
th
e
p
r
o
p
o
s
ed
ar
ch
itect
u
r
e.
Fiv
e
lay
er
s
o
f
co
n
v
o
l
u
tio
n
lay
er
s
in
ter
co
n
n
e
cted
with
m
ax
p
o
o
lin
g
f
o
r
f
e
atu
r
e
ex
ac
tio
n
s
.
A
s
tack
o
f
c
o
n
v
o
lu
ti
o
n
lay
er
s
is
in
ter
co
n
n
ec
ted
with
a
r
esid
u
al
lay
er
to
r
ed
u
ce
th
e
is
s
u
e
o
f
v
an
is
h
in
g
g
r
a
d
ien
ts
in
d
ee
p
n
e
u
r
al
n
etwo
r
k
s
.
T
h
e
s
ec
o
n
d
p
h
ase
o
f
a
r
ch
itectu
r
e
co
n
s
is
ts
o
f
B
iLST
M.
L
STM
ar
ch
itectu
r
e
also
h
el
p
s
to
p
r
ev
e
n
t
v
an
is
h
in
g
is
s
u
es
f
o
r
f
ea
tu
r
e
lear
n
in
g
p
r
o
p
o
s
ed
s
tu
d
y
h
as
u
s
ed
f
o
u
r
lay
e
r
s
o
f
B
iLST
M.
W
h
er
e
a
s
er
ies
o
f
L
STM
m
o
d
els
in
ter
co
n
n
ec
ted
in
co
u
n
ter
o
r
d
e
r
with
ea
ch
o
t
h
er
,
g
en
er
ally
ca
lled
a
s
f
o
r
war
d
a
n
d
b
ac
k
war
d
p
ass
o
f
th
e
L
STM
n
etwo
r
k
[
2
0
]
.
Fig
u
r
e
1
.
R
esid
u
al
-
b
ased
B
iL
STM
p
r
o
p
o
s
ed
a
r
ch
itectu
r
e
f
o
r
an
o
m
aly
d
etec
tio
n
[
2
0
]
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
15
,
No
.
4
,
Au
g
u
s
t
20
25
:
3
7
2
7
-
3736
3730
T
h
e
p
r
o
p
o
s
ed
ar
c
h
it
ec
tu
r
e
u
s
e
s
f
iv
e
lay
er
s
o
f
co
n
v
o
lu
ti
o
n
w
ith
s
k
ip
co
n
n
ec
tio
n
with
o
u
t
p
u
t
s
izes
o
f
(
1
1
2
×
1
1
2
,
5
6
×
5
6
,
2
8
×
2
8
,
1
4
×
14.
7
×
7
)
f
o
ll
o
wed
b
y
a
p
o
llin
g
lay
er
.
I
n
ter
m
ed
iate
lay
er
s
o
f
r
esid
u
al
co
n
n
ec
tio
n
s
ar
e
s
ep
ar
ate
d
with
n
o
r
m
alize
d
la
y
er
s
with
th
e
s
am
e
k
er
n
el
s
ize.
T
h
e
r
esid
u
al
m
o
d
u
le
aim
s
to
in
clu
d
e
th
e
n
etwo
r
k
-
e
x
tr
ac
te
d
f
ea
tu
r
es.
Fu
r
th
e
r
m
o
r
e
,
th
e
v
an
is
h
in
g
g
r
ad
ien
t
p
r
o
b
lem
is
m
itig
ated
b
y
th
e
u
s
ag
e
o
f
s
h
o
r
t
co
n
n
ec
tio
n
s
.
W
ith
R
e
s
Net
s
er
v
in
g
as
th
e
f
o
u
n
d
atio
n
,
Fig
u
r
e
2
illu
s
tr
ates
th
e
f
u
n
d
am
en
tal
wo
r
k
in
g
s
o
f
c
o
n
v
o
lu
tio
n
lear
n
in
g
.
E
q
u
atio
n
(
1
)
illu
s
tr
ates
th
e
n
o
r
m
alizin
g
p
r
o
ce
d
u
r
e
u
s
ed
to
n
o
r
m
alize
f
ea
tu
r
es
d
u
r
in
g
R
eL
U'
s
ac
t
iv
atio
n
.
T
h
r
o
u
g
h
th
e
u
s
e
o
f
s
k
ip
co
n
n
ec
tio
n
s
,
wh
ich
co
m
b
in
e
th
e
o
u
t
p
u
ts
o
f
s
tack
ed
lay
er
s
with
th
o
s
e
f
r
o
m
p
r
ev
io
u
s
lay
er
s
,
s
ig
n
if
ican
tly
d
ee
p
er
n
etwo
r
k
s
m
ay
n
o
w
b
e
tr
ain
ed
th
an
in
th
e
p
ast
[
2
1
]
.
(
)
=
{
,
>
0
0
,
ℎ
(
1
)
Fig
u
r
e
2
.
C
o
n
ce
p
tu
al
d
esig
n
o
f
r
esid
u
al
co
n
n
ec
tio
n
u
s
ed
in
t
h
e
p
r
o
p
o
s
ed
s
tu
d
y
[
2
1
]
T
h
e
R
esNet
m
o
d
u
le
is
co
m
p
o
s
ed
o
f
t
h
e
b
asic
b
lo
ck
an
d
th
e
id
en
tity
m
ap
p
in
g
.
Ass
u
m
e
th
a
t
th
e
in
p
u
t
v
ec
to
r
o
f
R
esNet
in
th
e
−
ℎ
m
o
d
u
l
e
is
an
d
th
e
o
u
tp
u
t
v
e
cto
r
o
f
−
ℎ
m
o
d
u
le
is
+
1
.
th
e
o
u
tp
u
t
v
ec
to
r
o
f
th
e
R
esNet
m
o
d
u
le
is
f
o
r
m
u
lated
as
(
2
)
a
n
d
(
3
)
[
2
2
]
.
(
,
{
}
)
=
(
(
2
)
(
1
)
)
(
2
)
+
1
=
(
(
,
{
}
)
)
+
(
3
)
3
.
1
.
B
iL
ST
M
L
o
n
g
s
h
o
r
t
-
ter
m
m
e
m
o
r
y
,
o
r
L
STM
,
is
f
r
eq
u
en
tly
th
o
u
g
h
t
o
f
as
a
d
ev
elo
p
m
en
t
o
f
r
ec
u
r
r
en
t
n
eu
r
al
n
etwo
r
k
s
(
R
NN)
.
T
h
e
ca
p
ac
ity
o
f
R
NN
to
f
u
n
ctio
n
as
“
s
h
o
r
t
-
ter
m
m
em
o
r
y
”
m
a
d
e
it
p
o
s
s
ib
le
to
ap
p
ly
p
r
io
r
k
n
o
wled
g
e
—
b
u
t o
n
ly
u
p
to
a
ce
r
tain
p
o
in
t
—
f
o
r
th
e
cu
r
r
en
t
t
ask
.
L
STM
ar
ch
itectu
r
e,
wh
ich
is
an
ex
ten
s
io
n
o
f
R
NN,
o
f
f
er
s
“
lo
n
g
-
ter
m
m
em
o
r
y
,
”
o
r
th
e
ab
ilit
y
to
s
to
r
e
all
o
f
th
e
p
r
io
r
d
ata
f
o
r
a
g
iv
en
n
eu
r
al
n
o
d
e,
r
ath
er
th
an
ju
s
t a
s
p
ec
if
ic
p
o
in
t i
n
ti
m
e
[
2
3
]
.
L
STM
ar
ch
itectu
r
e
co
n
s
is
ts
o
f
th
r
ee
g
ates
I
n
p
u
t,
o
u
t
p
u
t,
an
d
f
o
r
g
et.
T
h
e
n
etwo
r
k
'
s
p
r
esen
t
lo
n
g
-
ter
m
m
em
o
r
y
,
wh
ic
h
h
o
u
s
es
th
e
h
is
to
r
y
o
f
d
ata,
is
r
ef
er
r
ed
to
as
th
e
“
ce
ll
s
ta
te
.
”
Sh
o
r
t
-
ter
m
m
em
o
r
y
ca
n
b
e
th
o
u
g
h
t
o
f
as
th
e
o
u
tp
u
t
at
a
p
r
ev
io
u
s
p
o
in
t
in
tim
e,
wh
ich
is
r
ef
er
r
ed
t
o
as
th
e
p
r
e
v
io
u
s
Hid
d
en
State.
T
h
e
in
p
u
t
v
alu
e
at
t
h
e
cu
r
r
en
t
tim
e
s
tep
is
co
n
tain
ed
in
th
e
i
n
p
u
t
d
ata.
B
iLST
M
ad
o
p
ts
co
n
s
id
er
atio
n
o
f
b
o
th
t
h
e
p
ast an
d
f
u
t
u
r
e
p
r
o
p
er
ties
o
f
t
h
e
n
etwo
r
k
.
T
h
e
B
iLST
M
h
id
d
en
lay
er
co
n
s
is
ts
o
f
two
co
m
p
o
n
en
ts
: th
e
f
o
r
war
d
an
d
b
ac
k
war
d
L
STM
ce
ll
s
t
ates.
T
o
p
a
r
ticip
ate
in
th
e
f
o
r
war
d
a
n
d
r
ev
er
s
e
ca
lc
u
latio
n
s
,
n
etwo
r
k
tr
af
f
ic
in
f
o
r
m
atio
n
f
ir
s
t
en
ter
s
th
e
h
id
d
en
lay
er
th
r
o
u
g
h
th
e
in
p
u
t
lay
er
,
an
d
t
h
e
d
ata
co
m
p
u
ted
b
y
th
e
h
id
d
en
lay
e
r
is
s
u
b
s
eq
u
en
tly
co
m
m
u
n
icate
d
to
th
e
o
u
tp
u
t
lay
er
;
u
ltima
tely
,
th
e
o
u
tp
u
t
lay
e
r
co
m
b
i
n
e
s
th
e
f
o
r
war
d
an
d
r
ev
er
s
e
L
STM
o
u
tp
u
ts
b
ased
o
n
a
p
r
e
d
eter
m
in
ed
weig
h
t
t
o
p
r
o
d
u
ce
th
e
d
esire
d
o
u
tp
u
t.
Fig
u
r
e
3
s
h
o
ws
th
e
s
tr
u
ctu
r
e
o
f
t
h
e
B
iLST
M
n
etwo
r
k
.
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
-
8
7
0
8
E
n
h
a
n
cin
g
a
n
o
ma
ly
d
etec
tio
n
p
erfo
r
ma
n
ce
u
s
in
g
R
esN
et5
0
a
n
d
…
(
Dip
a
k
R
a
m
o
liya
)
3731
Fig
u
r
e
3
.
Fu
n
ctio
n
al
ar
c
h
itectu
r
e
o
f
B
iLST
M
u
s
ed
f
o
r
an
o
m
aly
d
etec
tio
n
[
2
3
]
Bi
-
d
ir
ec
tio
n
L
STM
n
etwo
r
k
allo
ws
th
e
m
o
d
el
to
lear
n
in
t
er
-
d
ep
e
n
d
en
t
f
ea
tu
r
es
f
r
o
m
in
p
u
t
v
id
e
o
.
I
n
itially
,
f
ea
tu
r
e
m
atr
ices
ar
e
p
r
o
ce
s
s
ed
with
a
f
o
r
wa
r
d
p
ass
o
f
L
STM
an
d
co
n
s
eq
u
en
tly
ca
lcu
lated
in
b
ac
k
war
d
p
ass
also
.
T
h
e
p
r
o
p
o
s
ed
s
tu
d
y
h
as u
s
ed
d
if
f
er
e
n
t
i
n
ter
co
n
n
ec
te
d
lay
e
r
s
o
f
B
iLST
M
.
Ho
wev
er
,
g
o
o
d
ac
cu
r
ac
y
is
f
o
u
n
d
i
n
th
e
f
o
u
r
l
ay
er
s
ar
ch
itectu
r
e
o
f
B
iLST
M.
T
h
e
h
id
d
e
n
s
tate
o
f
th
e
p
r
o
p
o
s
ed
ar
ch
itectu
r
e
h
t
ca
n
b
e
ca
lcu
lated
as
(
1
)
.
W
h
ar
e
C
t
,
as lea
r
n
in
g
o
f
ce
ll st
ate
f
o
r
m
u
lated
as
(
2
)
[
2
4
]
.
ℎ
=
×
ℎ
(
)
(
1
)
=
∙
−
1
+
∙
̃
(
2
)
Fo
r
war
d
an
d
b
ac
k
wa
r
d
p
ass
es
co
n
s
is
t
o
f
th
r
ee
g
ates
as
th
e
v
an
illa
ar
ch
itectu
r
e
o
f
L
STM
.
T
h
e
in
p
u
t
g
ate
o
f
th
e
p
r
o
p
o
s
ed
ar
ch
itectu
r
e
h
elp
s
to
id
en
tify
r
elev
an
t
in
f
o
r
m
atio
n
f
r
o
m
an
o
m
al
y
f
r
am
es
(
v
id
eo
)
.
T
h
e
o
u
tp
u
t
o
f
in
p
u
t
g
ate
is
s
et
as
e
n
ab
lin
g
ab
n
o
r
m
al
f
ea
tu
r
es
f
o
r
th
e
r
est
o
f
th
e
v
id
e
o
d
ata
its
s
et
to
0
as
d
is
ab
led
.
Similar
ly
,
o
u
tp
u
t
g
ate
e
n
ab
les
an
d
d
is
ab
les
b
ased
o
n
f
ea
t
u
r
e
in
f
o
r
m
atio
n
,
wh
eth
er
it
s
h
o
u
ld
b
e
f
o
r
war
d
ed
to
th
e
n
ex
t la
y
er
o
r
n
o
t.
Fo
r
m
u
lat
io
n
o
f
i
n
p
u
t a
n
d
o
u
tp
u
t
g
ate
as
(
3
)
an
d
(
4
)
[
2
5
]
.
=
(
⋅
[
ℎ
−
1
,
]
+
)
(
3
)
=
(
⋅
[
ℎ
−
1
,
]
+
)
(
4
)
Ou
tp
u
t
an
d
in
p
u
t
g
ates
ar
e
in
t
er
co
n
n
ec
ted
in
f
o
r
war
d
an
d
b
ac
k
war
d
p
r
o
p
a
g
atio
n
a
n
o
m
aly
f
ea
tu
r
es.
Ho
wev
er
,
th
e
b
eh
a
v
io
r
o
f
in
p
u
t
an
d
o
u
t
p
u
t
g
ate
co
n
tr
o
l
b
y
m
em
o
r
y
ce
ll,
wh
ich
is
u
p
d
ated
b
y
f
o
r
g
et
g
ate
(
f
o
r
m
u
lated
as
eq
u
atio
n
5
)
o
f
t
h
e
p
r
o
p
o
s
ed
ar
c
h
itectu
r
e.
T
h
e
o
u
tp
u
t
o
f
th
e
B
iLST
M
lay
er
is
th
e
co
n
ca
ten
atio
n
o
f
th
e
f
o
r
war
d
a
n
d
b
ac
k
war
d
h
id
d
en
s
tates
at
ea
ch
tim
e
s
tep
as
eq
u
atio
n
6
.
W
h
er
e
is
th
e
f
o
r
war
d
h
i
d
d
en
s
tate
at
tim
e
s
tep
t a
n
d
ℎ
(
)
is
th
e
b
ac
k
war
d
h
id
d
en
s
tate
at
tim
e
s
tep
t.
=
(
⋅
[
ℎ
−
1
,
]
+
)
(
5
)
ℎ
=
[
ℎ
(
)
,
ℎ
(
)
]
(
6
)
T
h
e
p
r
o
p
o
s
ed
ar
c
h
itectu
r
e
im
p
r
o
v
es
th
e
r
ec
o
g
n
itio
n
o
f
a
b
n
o
r
m
al
ac
tiv
ity
f
r
o
m
v
id
eo
s
eq
u
en
ce
s
[
2
5
]
.
T
h
e
co
m
b
in
atio
n
al
s
tu
d
y
o
f
co
n
v
o
l
u
tio
n
a
n
d
r
ec
u
r
r
en
t
n
e
two
r
k
s
h
elp
s
t
o
e
x
tr
ac
t
ad
v
a
n
ce
d
f
ea
tu
r
es
f
r
o
m
co
m
p
lex
ac
tiv
ity
v
id
e
o
s
.
T
h
e
d
ee
p
er
ar
ch
itectu
r
e
o
f
R
esNet
5
0
h
elp
s
to
r
ed
u
ce
v
a
n
is
h
in
g
i
s
s
u
es a
n
d
g
en
er
ate
a
f
ea
tu
r
e
m
atr
i
x
wh
ich
is
f
u
r
th
e
r
an
aly
ze
d
with
B
i
-
d
ir
ec
tio
n
L
STM
ar
ch
itectu
r
e
f
o
r
im
p
r
o
v
i
s
es
in
th
e
d
etec
tio
n
o
f
ab
n
o
r
m
ality
in
co
m
m
o
n
ar
e
as.
4.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
N
I
n
th
is
s
ec
tio
n
,
it
i
s
ex
p
l
ain
ed
th
e
r
esu
lts
o
f
r
esear
ch
an
d
at
th
e
s
am
e
tim
e
i
s
g
iv
en
th
e
co
m
p
r
eh
en
s
iv
e
d
is
cu
s
s
io
n
.
R
esu
lts
ca
n
b
e
p
r
esen
ted
in
f
ig
u
r
es,
g
r
ap
h
s
,
tab
les
an
d
o
t
h
er
s
th
at
m
a
k
e
th
e
r
ea
d
er
u
n
d
er
s
tan
d
ea
s
ily
[
1
4
]
,
[
1
5
]
.
Simu
latio
n
o
f
p
r
o
p
o
s
ed
co
m
b
in
atio
n
al
ar
ch
itectu
r
e
i
m
p
lem
en
ted
o
n
two
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
15
,
No
.
4
,
Au
g
u
s
t
20
25
:
3
7
2
7
-
3736
3732
d
iv
er
s
e
an
o
m
al
y
d
etec
tio
n
d
a
tasets
as
Un
iv
er
s
ity
o
f
C
en
tr
al
Flo
r
id
a
(
UC
F)
C
r
im
e
[
2
6
]
an
d
Sh
an
g
h
aiT
ec
h
C
am
p
u
s
d
ataset
[
2
7
]
.
Pro
p
er
ti
es o
f
b
en
ch
m
ar
k
a
n
o
m
aly
d
ete
ctio
n
d
atasets
ar
e
d
em
o
n
s
tr
ate
d
in
T
ab
le
2
.
T
ab
le
2
.
Data
s
et
in
f
o
r
m
atio
n
D
a
t
a
s
e
t
C
l
a
s
ses
A
v
g
.
F
r
a
m
e
s
Le
n
g
t
h
U
C
F
C
r
i
me
14
7
2
4
7
1
2
8
h
r
S
h
a
n
g
h
a
i
T
e
c
h
13
2
4
4
1
5
1
3
0
h
r
T
h
e
p
r
o
p
o
s
ed
a
p
p
r
o
ac
h
is
s
im
u
lated
o
n
a
PC
with
3
2
GB
R
AM
,
GeFo
r
ce
R
T
X
3
0
8
0
,
an
d
C
o
r
e
i7
p
r
o
ce
s
s
o
r
.
T
h
e
m
o
d
el
was
ex
ec
u
ted
u
s
in
g
Py
th
o
n
'
s
Ker
as
p
ac
k
ag
e
an
d
T
en
s
o
r
Flo
w.
E
q
u
atio
n
s
(
7
),
(
8
),
an
d
(
9
)
ar
e
u
s
ed
to
b
u
ild
ev
alu
atio
n
m
atr
ices
s
u
ch
as
p
r
ec
i
s
io
n
,
r
ec
all
,
F1
-
Sco
r
e,
a
n
d
ac
cu
r
ac
y
[
2
8
]
.
T
h
e
s
im
u
latio
n
o
f
th
e
p
r
o
p
o
s
ed
s
tu
d
y
h
as
ac
h
iev
ed
r
em
a
r
k
ab
le
p
er
f
o
r
m
a
n
ce
as
9
7
.
5
5
%
ac
cu
r
ac
y
f
o
r
th
e
UC
F
C
r
im
e
d
ataset
an
d
9
1
.
9
4
%
ac
cu
r
ac
y
f
o
r
th
e
Sh
an
g
h
aiT
ec
h
d
ataset.
T
h
e
ac
cu
r
ac
y
an
d
lo
s
s
g
r
ap
h
o
f
th
e
p
r
o
p
o
s
ed
R
esNet5
0
with
B
iLS
T
M
ar
ch
itectu
r
e
is
illu
s
tr
ated
in
Fig
u
r
e
4
.
=
+
(
7
)
=
+
(
8
)
=
+
+
+
+
(
9
)
Fig
u
r
e
4
.
Acc
u
r
ac
y
a
n
d
lo
s
s
g
r
ap
h
s
o
f
th
e
p
r
o
p
o
s
ed
m
o
d
el
w
ith
b
o
th
b
en
ch
m
a
r
k
d
atasets
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
-
8
7
0
8
E
n
h
a
n
cin
g
a
n
o
ma
ly
d
etec
tio
n
p
erfo
r
ma
n
ce
u
s
in
g
R
esN
et5
0
a
n
d
…
(
Dip
a
k
R
a
m
o
liya
)
3733
I
n
itially
,
ex
p
er
im
en
ts
wer
e
co
n
d
u
cte
d
o
n
th
e
co
r
e
ar
c
h
itectu
r
e
o
f
co
n
v
o
lu
tio
n
a
n
d
r
ec
u
r
r
en
t
n
etwo
r
k
s
to
id
e
n
tify
th
e
p
o
t
en
tial
o
f
an
o
m
aly
d
etec
tio
n
f
o
r
b
o
th
d
atasets
,
as
d
em
o
n
s
tr
ated
in
T
ab
le
3
.
I
m
p
lem
en
tatio
n
o
f
an
o
m
aly
d
etec
tio
n
u
s
in
g
co
r
e
an
d
s
tate
-
of
-
t
he
-
ar
t
m
o
d
els
f
in
d
s
g
o
o
d
ac
cu
r
ac
y
,
b
u
t
n
o
t
p
r
o
m
is
in
g
to
m
ak
e
th
e
m
o
d
el
m
o
r
e
g
en
er
alize
d
an
d
r
o
b
u
s
t.
T
h
e
s
im
u
latio
n
o
f
th
e
s
tu
d
y
u
s
es
co
r
e
(
4
-
lay
er
)
r
ec
u
r
r
en
t
n
etwo
r
k
m
o
d
els
s
u
ch
as
r
ec
u
r
r
en
t
n
e
u
r
al
n
etwo
r
k
(
R
NN)
,
l
o
n
g
s
h
o
r
t
-
ter
m
m
e
m
o
r
y
(
L
STM
)
,
an
d
g
ated
r
ec
u
r
r
en
t
u
n
it
(
GR
U)
[
2
9
]
.
All
th
is
r
ec
u
r
r
e
n
t
a
d
d
itio
n
o
f
th
e
B
iLST
M
n
etwo
r
k
im
p
r
o
v
es
th
e
r
ec
o
g
n
itio
n
r
ate
o
f
ab
n
o
r
m
al
ac
ti
v
ities
in
b
o
th
d
atasets
.
T
ab
le
3
.
R
elativ
e
an
aly
s
is
o
f
i
n
d
iv
id
u
al
d
ee
p
lear
n
in
g
m
eth
o
d
s
f
o
r
an
o
m
aly
d
etec
tio
n
M
o
d
e
l
D
a
t
a
s
e
t
A
c
c
u
r
a
c
y
C
N
N
U
C
F
9
0
.
6
1
S
h
a
n
g
h
a
i
T
e
c
h
8
6
.
0
6
R
N
N
U
C
F
8
5
.
8
2
S
h
a
n
g
h
a
i
T
e
c
h
7
4
.
3
7
LSTM
U
C
F
9
4
.
5
1
S
h
a
n
g
h
a
i
T
e
c
h
8
9
.
7
6
G
R
U
U
C
F
9
4
.
7
8
S
h
a
n
g
h
a
i
T
e
c
h
8
8
.
4
5
B
i
L
S
T
M
UCF
9
5
.
2
2
S
h
a
n
g
h
a
i
T
e
c
h
9
0
.
0
1
Simu
latio
n
o
f
th
e
p
r
o
p
o
s
ed
s
tu
d
y
also
co
m
p
ar
e
d
with
d
i
f
f
er
en
t
c
o
m
b
in
atio
n
s
o
v
er
c
o
n
v
o
lu
tio
n
m
o
d
el
s
u
ch
as
v
an
illa
C
NN,
Den
s
eNe
t,
R
e
s
Net5
0
/3
4
,
an
d
VGG1
6
[
3
0
]
,
[
3
1
]
.
T
a
b
le
4
d
em
o
n
s
tr
ates
a
co
m
p
ar
ativ
e
an
aly
s
is
o
f
t
h
e
S
OT
A
co
n
v
o
l
u
tio
n
m
o
d
el
with
th
e
p
r
o
p
o
s
ed
Dee
p
B
iLST
M
m
o
d
el.
Simu
latio
n
o
f
th
e
p
r
o
p
o
s
ed
s
tu
d
y
also
ex
t
en
d
s
to
an
aly
ze
th
e
im
p
ac
t o
f
d
if
f
er
en
t h
y
p
er
p
ar
am
eter
s
s
u
ch
as b
atch
s
ize,
s
p
lit
r
atio
,
an
d
lea
r
n
in
g
r
ate.
Fig
u
r
e
5
illu
s
tr
ates
t
h
e
im
p
a
ct
o
f
t
h
e
lear
n
i
n
g
r
ate
a
n
d
tr
ain
-
test
s
p
lit
r
atio
f
o
r
b
o
th
d
atasets
.
T
h
e
s
im
u
latio
n
was
also
an
aly
ze
d
with
d
if
f
e
r
en
t
n
u
m
b
er
s
o
f
ep
o
c
h
s
an
d
lear
n
in
g
lay
er
s
in
th
e
p
r
o
p
o
s
ed
B
iLST
M
ar
ch
itectu
r
e.
T
ab
le
4
.
C
o
m
p
a
r
ativ
e
an
aly
s
is
o
f
d
if
f
er
en
t c
o
m
b
in
atio
n
al
ar
c
h
itectu
r
es f
o
r
a
n
o
m
aly
d
etec
tio
n
M
o
d
e
l
P
a
r
a
ms
A
c
c
u
r
a
c
y
P
r
e
c
i
s
i
o
n
R
e
c
a
l
l
C
N
N
+
B
i
LST
M
1
4
9
8
9
9
2
.
6
7
0
.
9
2
0
.
9
2
V
G
G
1
6
+
B
i
LST
M
1
5
2
0
8
9
3
.
0
9
0
.
9
3
0
.
9
2
D
e
n
seN
e
t
+
B
i
LS
TM
1
7
5
9
4
9
5
.
9
1
0
.
9
5
0
.
9
6
R
e
sN
e
t
3
4
+
B
i
LS
TM
1
6
2
4
8
9
4
.
2
8
0
.
9
4
0
.
9
4
R
e
sN
e
t
5
0
+
B
i
L
S
T
M
1
8
4
6
4
9
7
.
5
5
0
.
9
7
0
.
9
7
Fig
u
r
e
5
.
C
o
m
p
a
r
ativ
e
an
aly
s
i
s
o
f
p
r
o
p
o
s
ed
ar
c
h
itectu
r
e
wit
h
d
if
f
er
e
n
t le
ar
n
in
g
p
ar
a
m
eter
s
5.
DIS
CU
SS
I
O
N
T
h
e
p
r
o
p
o
s
ed
s
tu
d
y
an
al
y
ze
s
th
e
d
etec
tio
n
o
f
ab
n
o
r
m
al
ac
tiv
ity
f
r
o
m
b
e
n
ch
m
ar
k
d
atasets
lik
e
UC
F
cr
im
e
an
d
Sh
a
n
g
h
aiT
ec
h
u
s
in
g
en
h
a
n
ce
d
d
ee
p
lear
n
i
n
g
m
o
d
els
s
u
ch
as
R
esNet5
0
with
B
iLST
M.
Simu
latio
n
75,00%
80,00%
85,00%
90,00%
95,00%
100,00%
0.2
0.3
0.2
0.3
0.2
0.3
0.01
0.001
0.0001
Acc
u
ra
cy
Learn
in
g ra
te wi
th
d
if
f
e
re
n
t
s
p
lit
-
ra
tio
C
o
mp
a
r
a
t
i
v
e
a
n
a
l
y
si
s w
i
t
h
d
i
f
f
e
r
n
t
l
e
a
r
n
i
n
g
r
a
t
e
a
n
d
sp
l
i
t
r
a
t
i
o
U
CF
S
h
an
g
h
ai
T
e
ch
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
15
,
No
.
4
,
Au
g
u
s
t
20
25
:
3
7
2
7
-
3736
3734
o
f
p
r
o
p
o
s
ed
a
r
ch
itectu
r
e
f
i
n
d
s
r
em
ar
k
ab
le
p
e
r
f
o
r
m
an
ce
w
ith
9
3
.
8
1
an
d
8
7
.
9
1
%
ac
cu
r
a
cy
f
o
r
b
en
ch
m
ar
k
d
atasets
.
T
h
e
s
im
u
latio
n
s
ec
t
io
n
also
an
aly
ze
s
th
e
co
m
p
a
r
is
o
n
with
d
if
f
er
en
t
SOTA
m
o
d
els
f
o
r
an
o
m
aly
d
etec
tio
n
an
d
id
en
tifie
s
th
e
i
m
p
ac
t
o
f
s
tan
d
alo
n
e
co
n
v
o
lu
tio
n
an
d
r
ec
u
r
r
en
t
n
etwo
r
k
s
.
Simu
latio
n
o
f
th
e
p
r
o
p
o
s
ed
s
tu
d
y
im
p
r
o
v
es
r
ec
o
g
n
itio
n
ac
c
u
r
ac
y
f
o
r
a
b
n
o
r
m
al
ac
tiv
ity
b
y
c
o
m
b
in
i
n
g
co
n
v
o
l
u
tio
n
a
n
d
r
ec
u
r
r
en
t
ar
ch
itectu
r
e.
T
h
e
p
er
f
o
r
m
an
c
e
o
f
th
e
p
r
o
p
o
s
ed
a
r
ch
itectu
r
e
f
o
r
u
n
s
ee
n
p
r
ed
ictio
n
d
at
a
is
illu
s
tr
ated
in
Fig
u
r
e
6
as a
co
n
f
u
s
io
n
m
atr
ix
f
o
r
UC
F c
r
im
e
d
atasets
.
Fig
u
r
e
6
.
C
o
n
f
u
s
io
n
m
atr
i
x
o
f
p
r
o
p
o
s
ed
ar
ch
itectu
r
e
o
n
u
n
s
e
en
UC
F d
ataset
6.
CO
NCLU
SI
O
N
Vid
eo
s
u
r
v
eillan
ce
b
ec
o
m
e
e
s
s
en
tial
to
p
r
ev
e
n
t
u
n
ex
p
ec
ted
ac
cid
e
n
ts
in
p
u
b
lic
p
lace
s
.
Au
to
m
atic
d
etec
tio
n
o
f
ab
n
o
r
m
al
ac
tiv
ity
h
elp
s
to
r
ed
u
ce
cr
im
e
an
d
im
p
r
o
v
e
cr
it
ical
ca
r
e
i
n
th
e
m
ed
ical
s
ec
to
r
.
T
h
e
p
r
o
p
o
s
ed
s
tu
d
y
h
elp
s
to
im
p
r
o
v
e
th
e
d
etec
tio
n
o
f
a
b
n
o
r
m
al
ac
tiv
ity
f
r
o
m
v
id
eo
s
tr
ea
m
s
.
T
h
e
p
r
o
p
o
s
ed
s
tu
d
y
u
s
es
co
n
v
o
l
u
tio
n
an
d
s
eq
u
en
ti
al
ar
ch
itectu
r
e
to
en
h
a
n
ce
th
e
ac
cu
r
ac
y
o
f
an
o
m
aly
d
etec
tio
n
.
Simu
latio
n
o
f
th
e
p
r
o
p
o
s
ed
h
as
p
r
o
v
en
im
p
ac
t
o
f
th
e
p
r
o
p
o
s
ed
ar
ch
itectu
r
e
o
v
er
two
b
en
c
h
m
ar
k
d
atasets
.
Simu
latio
n
h
as
also
ex
p
lo
r
ed
th
e
c
o
m
p
ass
io
n
o
f
t
h
e
p
r
o
p
o
s
ed
a
r
ch
itectu
r
e
with
s
tate
-
of
-
th
e
-
ar
t
d
ee
p
lear
n
in
g
m
o
d
els
f
o
r
an
o
m
aly
d
etec
tio
n
.
T
h
e
p
r
o
p
o
s
ed
co
m
b
in
atio
n
al
ar
ch
itect
u
r
e
ca
n
ac
h
iev
e
r
em
ar
k
ab
le
p
e
r
f
o
r
m
an
c
e
with
9
7
.
5
5
%
an
d
9
1
.
9
4
%
ac
c
u
r
ac
y
f
o
r
UC
F
an
d
Sh
an
g
h
aiT
ec
h
d
atasets
r
esp
ec
tiv
ely
.
Simu
latio
n
h
as
also
ex
p
lo
r
e
d
th
e
p
er
f
o
r
m
an
ce
o
f
d
if
f
er
en
t
c
o
m
b
in
atio
n
s
o
f
s
eq
u
en
tial
m
o
d
els
with
co
n
v
o
lu
tio
n
m
o
d
els,
b
u
t
B
iLST
M
f
in
d
s
p
r
o
m
is
in
g
r
esu
lts
co
m
p
ar
e
d
t
o
o
th
e
r
s
eq
u
en
tial
m
o
d
els
lik
e
GR
U
an
d
L
STM
.
T
h
e
p
r
o
p
o
s
ed
s
tu
d
y
ca
n
b
e
ex
ten
d
ed
to
d
etec
t
ab
n
o
r
m
al
a
ctiv
ity
f
r
o
m
liv
e
v
id
e
o
f
ee
d
s
.
An
ex
ten
s
io
n
o
f
th
is
wo
r
k
ca
n
b
e
d
esig
n
ed
with
r
o
b
u
s
t a
n
d
ad
v
a
n
ce
d
ar
c
h
itectu
r
e
f
o
r
an
o
m
aly
d
etec
tio
n.
ACK
NO
WL
E
DG
E
M
E
NT
S
W
e
wo
u
ld
lik
e
to
th
an
k
th
e
“
C
h
ar
o
tar
Un
iv
er
s
ity
o
f
Scien
ce
an
d
T
ec
h
n
o
l
o
g
y
–
C
HARU
SAT
”
f
o
r
ca
r
r
y
in
g
o
u
t th
is
r
esear
ch
an
d
u
tili
zin
g
r
eso
u
r
ce
s
.
F
UNDING
I
NF
O
R
M
A
T
I
O
N
Au
th
o
r
s
s
tate
n
o
f
u
n
d
in
g
in
v
o
lv
ed
.
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
-
8
7
0
8
E
n
h
a
n
cin
g
a
n
o
ma
ly
d
etec
tio
n
p
erfo
r
ma
n
ce
u
s
in
g
R
esN
et5
0
a
n
d
…
(
Dip
a
k
R
a
m
o
liya
)
3735
AUTHO
R
CO
NT
RI
B
UT
I
O
NS
ST
A
T
E
M
E
N
T
T
h
is
jo
u
r
n
al
u
s
es
th
e
C
o
n
tr
ib
u
to
r
R
o
les
T
ax
o
n
o
m
y
(
C
R
ed
iT)
to
r
ec
o
g
n
ize
in
d
iv
id
u
al
au
th
o
r
co
n
tr
ib
u
tio
n
s
,
r
ed
u
ce
au
th
o
r
s
h
ip
d
is
p
u
tes,
an
d
f
ac
ilit
ate
co
llab
o
r
atio
n
.
Na
m
e
o
f
Aut
ho
r
C
M
So
Va
Fo
I
R
D
O
E
Vi
Su
P
Fu
Dip
ak
R
am
o
liy
a
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
Am
it Ga
n
atr
a
✓
✓
✓
✓
✓
C
:
C
o
n
c
e
p
t
u
a
l
i
z
a
t
i
o
n
M
:
M
e
t
h
o
d
o
l
o
g
y
So
:
So
f
t
w
a
r
e
Va
:
Va
l
i
d
a
t
i
o
n
Fo
:
Fo
r
mal
a
n
a
l
y
s
i
s
I
:
I
n
v
e
s
t
i
g
a
t
i
o
n
R
:
R
e
so
u
r
c
e
s
D
:
D
a
t
a
C
u
r
a
t
i
o
n
O
:
W
r
i
t
i
n
g
-
O
r
i
g
i
n
a
l
D
r
a
f
t
E
:
W
r
i
t
i
n
g
-
R
e
v
i
e
w
&
E
d
i
t
i
n
g
Vi
:
Vi
su
a
l
i
z
a
t
i
o
n
Su
:
Su
p
e
r
v
i
s
i
o
n
P
:
P
r
o
j
e
c
t
a
d
mi
n
i
st
r
a
t
i
o
n
Fu
:
Fu
n
d
i
n
g
a
c
q
u
i
si
t
i
o
n
CO
NF
L
I
C
T
O
F
I
N
T
E
R
E
S
T
ST
A
T
E
M
E
NT
Au
th
o
r
s
s
tate
n
o
co
n
f
lict
o
f
in
t
er
est.
E
T
H
I
CAL AP
P
RO
V
AL
T
h
e
co
n
d
u
cted
r
esear
ch
is
n
o
t
r
elate
d
to
eith
er
h
u
m
an
o
r
a
n
i
m
al
u
s
e.
DATA AV
AI
L
AB
I
L
I
T
Y
Der
iv
ed
d
ata
s
u
p
p
o
r
tin
g
th
e
f
i
n
d
in
g
s
o
f
th
is
s
tu
d
y
ar
e
av
aila
b
le
f
r
o
m
th
e
co
r
r
esp
o
n
d
i
n
g
a
u
th
o
r
,
DR
,
o
n
r
eq
u
est.
RE
F
E
R
E
NC
E
S
[
1
]
Y
.
T
a
n
,
C
.
H
u
,
K
.
Zh
a
n
g
,
K
.
Z
h
e
n
g
,
E
.
A
.
D
a
v
i
s,
a
n
d
J
.
S
.
P
a
r
k
,
“
LST
M
-
b
a
se
d
a
n
o
m
a
l
y
d
e
t
e
c
t
i
o
n
f
o
r
n
o
n
-
l
i
n
e
a
r
d
y
n
a
m
i
c
a
l
sy
st
e
m,
”
I
EEE
A
c
c
e
ss
,
v
o
l
.
8
,
p
p
.
1
0
3
3
0
1
–
1
0
3
3
0
8
,
2
0
2
0
,
d
o
i
:
1
0
.
1
1
0
9
/
a
c
c
e
ss.2
0
2
0
.
2
9
9
9
0
6
5
.
[
2
]
S
.
G
a
r
g
a
n
d
S
.
B
a
t
r
a
,
“
A
n
o
v
e
l
e
n
se
mb
l
e
d
t
e
c
h
n
i
q
u
e
f
o
r
a
n
o
ma
l
y
d
e
t
e
c
t
i
o
n
,
”
I
n
t
e
r
n
a
t
i
o
n
a
l
J
o
u
rn
a
l
o
f
C
o
m
m
u
n
i
c
a
t
i
o
n
S
y
st
e
m
s
,
v
o
l
.
3
0
,
n
o
.
1
1
,
J
a
n
.
2
0
1
7
,
d
o
i
:
1
0
.
1
0
0
2
/
d
a
c
.
3
2
4
8
.
[
3
]
M
.
S
a
i
d
E
l
sa
y
e
d
,
N
.
-
A
.
Le
-
K
h
a
c
,
S
.
D
e
v
,
a
n
d
A
.
D
.
Ju
r
c
u
t
,
“
N
e
t
w
o
r
k
a
n
o
m
a
l
y
d
e
t
e
c
t
i
o
n
u
si
n
g
LS
TM
b
a
se
d
a
u
t
o
e
n
c
o
d
e
r
,
”
i
n
Pro
c
e
e
d
i
n
g
s
o
f
t
h
e
1
6
t
h
A
C
M
S
y
m
p
o
s
i
u
m
o
n
Q
o
S
a
n
d
S
e
c
u
r
i
t
y
f
o
r
W
i
rel
e
ss
a
n
d
Mo
b
i
l
e
N
e
t
w
o
r
k
s
,
N
o
v
.
2
0
2
0
.
d
o
i
:
1
0
.
1
1
4
5
/
3
4
1
6
0
1
3
.
3
4
2
6
4
5
7
.
[
4
]
H
.
W
e
i
,
K
.
Li
,
H
.
Li
,
Y
.
L
y
u
,
a
n
d
X
.
H
u
,
“
D
e
t
e
c
t
i
n
g
v
i
d
e
o
a
n
o
ma
l
y
w
i
t
h
a
st
a
c
k
e
d
c
o
n
v
o
l
u
t
i
o
n
a
l
LST
M
f
r
a
mew
o
r
k
,
”
i
n
C
o
m
p
u
t
e
r
Vi
s
i
o
n
S
y
st
e
m
s
,
S
p
r
i
n
g
e
r
I
n
t
e
r
n
a
t
i
o
n
a
l
P
u
b
l
i
sh
i
n
g
,
2
0
1
9
,
p
p
.
3
3
0
–
3
4
2
.
d
o
i
:
1
0
.
1
0
0
7
/
9
7
8
-
3
-
0
3
0
-
3
4
9
9
5
-
0
_
3
0
.
[
5
]
D
.
M
a
n
e
,
“
R
e
a
l
-
t
i
me
a
n
o
m
a
l
y
d
e
t
e
c
t
i
o
n
i
n
v
i
d
e
o
s
u
r
v
e
i
l
l
a
n
c
e
:
A
ma
t
h
e
ma
t
i
c
a
l
m
o
d
e
l
i
n
g
a
n
d
n
o
n
l
i
n
e
a
r
a
n
a
l
y
si
s
p
e
r
s
p
e
c
t
i
v
e
w
i
t
h
M
o
b
i
l
e
N
e
t
a
n
d
B
i
-
LST
M
,
”
C
o
m
m
u
n
i
c
a
t
i
o
n
s
o
n
A
p
p
l
i
e
d
N
o
n
l
i
n
e
a
r
An
a
l
y
si
s
,
v
o
l
.
3
1
,
n
o
.
2
s,
p
p
.
3
0
6
–
3
1
9
,
Ju
n
.
2
0
2
4
,
d
o
i
:
1
0
.
5
2
7
8
3
/
c
a
n
a
.
v
3
1
.
6
5
1
.
[
6
]
E.
D
u
ma
n
a
n
d
O
.
A
.
Er
d
e
m,
“
A
n
o
ma
l
y
d
e
t
e
c
t
i
o
n
i
n
v
i
d
e
o
s
u
s
i
n
g
o
p
t
i
c
a
l
f
l
o
w
a
n
d
c
o
n
v
o
l
u
t
i
o
n
a
l
a
u
t
o
e
n
c
o
d
e
r
,
”
I
EEE
A
c
c
e
ss
,
v
o
l
.
7
,
p
p
.
1
8
3
9
1
4
–
1
8
3
9
2
3
,
2
0
1
9
,
d
o
i
:
1
0
.
1
1
0
9
/
a
c
c
e
ss.
2
0
1
9
.
2
9
6
0
6
5
4
.
[
7
]
Y
.
Zh
u
,
N
.
M
.
N
a
y
a
k
,
a
n
d
A
.
K
.
R
o
y
-
C
h
o
w
d
h
u
r
y
,
“
C
o
n
t
e
x
t
-
a
w
a
r
e
a
c
t
i
v
i
t
y
r
e
c
o
g
n
i
t
i
o
n
a
n
d
a
n
o
m
a
l
y
d
e
t
e
c
t
i
o
n
i
n
v
i
d
e
o
,
”
I
E
EE
J
o
u
rn
a
l
o
f
S
e
l
e
c
t
e
d
T
o
p
i
c
s
i
n
S
i
g
n
a
l
P
ro
c
e
ssi
n
g
,
v
o
l
.
7
,
n
o
.
1
,
p
p
.
9
1
–
1
0
1
,
F
e
b
.
2
0
1
3
,
d
o
i
:
1
0
.
1
1
0
9
/
j
st
sp
.
2
0
1
2
.
2
2
3
4
7
2
2
.
[
8
]
T.
N
.
N
g
u
y
e
n
a
n
d
J.
M
e
u
n
i
e
r
,
“
A
n
o
m
a
l
y
d
e
t
e
c
t
i
o
n
i
n
v
i
d
e
o
se
q
u
e
n
c
e
w
i
t
h
a
p
p
e
a
r
a
n
c
e
-
m
o
t
i
o
n
c
o
r
r
e
s
p
o
n
d
e
n
c
e
,
”
i
n
2
0
1
9
I
EEE/
C
VF
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
C
o
m
p
u
t
e
r
Vi
s
i
o
n
(
I
C
C
V)
,
O
c
t
.
2
0
1
9
,
p
p
.
1
2
7
3
–
1
2
8
3
.
d
o
i
:
1
0
.
1
1
0
9
/
i
c
c
v
.
2
0
1
9
.
0
0
1
3
6
.
[
9
]
P
.
C
h
r
i
st
i
a
n
se
n
,
L.
N
i
e
l
s
e
n
,
K
.
S
t
e
e
n
,
R
.
J
ø
r
g
e
n
s
e
n
,
a
n
d
H
.
K
a
r
s
t
o
f
t
,
“
D
e
e
p
A
n
o
ma
l
y
:
C
o
m
b
i
n
i
n
g
b
a
c
k
g
r
o
u
n
d
s
u
b
t
r
a
c
t
i
o
n
a
n
d
d
e
e
p
l
e
a
r
n
i
n
g
f
o
r
d
e
t
e
c
t
i
n
g
o
b
st
a
c
l
e
s
a
n
d
a
n
o
ma
l
i
e
s
i
n
a
n
a
g
r
i
c
u
l
t
u
r
a
l
f
i
e
l
d
,
”
S
e
n
s
o
rs
,
v
o
l
.
1
6
,
n
o
.
1
1
,
p
.
1
9
0
4
,
N
o
v
.
2
0
1
6
,
d
o
i
:
1
0
.
3
3
9
0
/
s
1
6
1
1
1
9
0
4
.
[
1
0
]
M
.
S
a
b
o
k
r
o
u
,
M
.
F
a
t
h
y
,
a
n
d
M
.
H
o
se
i
n
i
,
“
V
i
d
e
o
a
n
o
ma
l
y
d
e
t
e
c
t
i
o
n
a
n
d
l
o
c
a
l
i
s
a
t
i
o
n
b
a
se
d
o
n
t
h
e
s
p
a
r
s
i
t
y
a
n
d
r
e
c
o
n
s
t
r
u
c
t
i
o
n
e
r
r
o
r
o
f
a
u
t
o
‐
e
n
c
o
d
e
r
,
”
E
l
e
c
t
r
o
n
i
c
s
L
e
t
t
e
rs
,
v
o
l
.
5
2
,
n
o
.
1
3
,
p
p
.
1
1
2
2
–
1
1
2
4
,
J
u
n
.
2
0
1
6
,
d
o
i
:
1
0
.
1
0
4
9
/
e
l
.
2
0
1
6
.
0
4
4
0
.
[
1
1
]
S
.
W
.
K
h
a
n
e
t
a
l
.
,
“
A
n
o
ma
l
y
d
e
t
e
c
t
i
o
n
i
n
t
r
a
f
f
i
c
s
u
r
v
e
i
l
l
a
n
c
e
v
i
d
e
o
s
u
si
n
g
d
e
e
p
l
e
a
r
n
i
n
g
,
”
S
e
n
s
o
rs
,
v
o
l
.
2
2
,
n
o
.
1
7
,
p
.
6
5
6
3
,
A
u
g
.
2
0
2
2
,
d
o
i
:
1
0
.
3
3
9
0
/
s2
2
1
7
6
5
6
3
.
[
1
2
]
V.
-
T.
Le
a
n
d
Y
.
-
G
.
K
i
m
,
“
A
t
t
e
n
t
i
o
n
-
b
a
se
d
r
e
si
d
u
a
l
a
u
t
o
e
n
c
o
d
e
r
f
o
r
v
i
d
e
o
a
n
o
m
a
l
y
d
e
t
e
c
t
i
o
n
,
”
A
p
p
l
i
e
d
I
n
t
e
l
l
i
g
e
n
c
e
,
v
o
l
.
5
3
,
n
o
.
3
,
p
p
.
3
2
4
0
–
3
2
5
4
,
M
a
y
2
0
2
2
,
d
o
i
:
1
0
.
1
0
0
7
/
s
1
0
4
8
9
-
0
2
2
-
0
3
6
1
3
-
1.
[
1
3
]
D
.
A
v
o
l
a
e
t
a
l
.
,
“
A
n
o
v
e
l
G
A
N
-
b
a
se
d
a
n
o
ma
l
y
d
e
t
e
c
t
i
o
n
a
n
d
l
o
c
a
l
i
z
a
t
i
o
n
met
h
o
d
f
o
r
a
e
r
i
a
l
v
i
d
e
o
s
u
r
v
e
i
l
l
a
n
c
e
a
t
l
o
w
a
l
t
i
t
u
d
e
,
”
Re
m
o
t
e
S
e
n
s
i
n
g
,
v
o
l
.
1
4
,
n
o
.
1
6
,
p
.
4
1
1
0
,
A
u
g
.
2
0
2
2
,
d
o
i
:
1
0
.
3
3
9
0
/
r
s1
4
1
6
4
1
1
0
.
[
1
4
]
M
.
Za
i
g
h
a
m
Za
h
e
e
r
,
J.
-
H
.
Le
e
,
M
.
A
st
r
i
d
,
a
n
d
S
.
-
I
.
Le
e
,
“
O
l
d
i
s
g
o
l
d
:
R
e
d
e
f
i
n
i
n
g
t
h
e
a
d
v
e
r
sari
a
l
l
y
l
e
a
r
n
e
d
o
n
e
-
c
l
a
ss
c
l
a
ss
i
f
i
e
r
t
r
a
i
n
i
n
g
p
a
r
a
d
i
g
m,
”
i
n
2
0
2
0
I
E
EE/
C
VF
C
o
n
f
e
r
e
n
c
e
o
n
C
o
m
p
u
t
e
r
Vi
s
i
o
n
a
n
d
Pa
t
t
e
rn
Re
c
o
g
n
i
t
i
o
n
(
C
VP
R)
,
Ju
n
.
2
0
2
0
,
p
p
.
1
4
1
7
1
–
1
4
1
8
1
.
d
o
i
:
1
0
.
1
1
0
9
/
c
v
p
r
4
2
6
0
0
.
2
0
2
0
.
0
1
4
1
9
.
[
1
5
]
Y
.
T
a
n
g
,
L
.
Zh
a
o
,
S
.
Zh
a
n
g
,
C
.
G
o
n
g
,
G
.
Li
,
a
n
d
J.
Y
a
n
g
,
“
I
n
t
e
g
r
a
t
i
n
g
p
r
e
d
i
c
t
i
o
n
a
n
d
r
e
c
o
n
s
t
r
u
c
t
i
o
n
f
o
r
a
n
o
mal
y
d
e
t
e
c
t
i
o
n
,
”
Pa
t
t
e
r
n
Re
c
o
g
n
i
t
i
o
n
L
e
t
t
e
rs
,
v
o
l
.
1
2
9
,
p
p
.
1
2
3
–
1
3
0
,
Ja
n
.
2
0
2
0
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
p
a
t
r
e
c
.
2
0
1
9
.
1
1
.
0
2
4
.
[
1
6
]
J.
T.
Z
h
o
u
,
J.
D
u
,
H
.
Z
h
u
,
X
.
P
e
n
g
,
Y
.
L
i
u
,
a
n
d
R
.
S
.
M
.
G
o
h
,
“
A
n
o
ma
l
y
N
e
t
:
A
n
a
n
o
mal
y
d
e
t
e
c
t
i
o
n
n
e
t
w
o
r
k
f
o
r
v
i
d
e
o
su
r
v
e
i
l
l
a
n
c
e
,
”
I
EE
E
T
r
a
n
s
a
c
t
i
o
n
s
o
n
I
n
f
o
rm
a
t
i
o
n
F
o
re
n
s
i
c
s
a
n
d
S
e
c
u
ri
t
y
,
v
o
l
.
1
4
,
n
o
.
1
0
,
p
p
.
2
5
3
7
–
2
5
5
0
,
O
c
t
.
2
0
1
9
,
d
o
i
:
1
0
.
1
1
0
9
/
t
i
f
s.
2
0
1
9
.
2
9
0
0
9
0
7
.
[
1
7
]
B
.
W
a
n
,
W
.
J
i
a
n
g
,
Y
.
F
a
n
g
,
Z.
L
u
o
,
a
n
d
G
.
D
i
n
g
,
“
A
n
o
ma
l
y
d
e
t
e
c
t
i
o
n
i
n
v
i
d
e
o
se
q
u
e
n
c
e
s
:
A
b
e
n
c
h
mark
a
n
d
c
o
m
p
u
t
a
t
i
o
n
a
l
mo
d
e
l
,
”
I
ET
I
m
a
g
e
Pro
c
e
ss
i
n
g
,
v
o
l
.
1
5
,
n
o
.
1
4
,
p
p
.
3
4
5
4
–
3
4
6
5
,
M
a
y
2
0
2
1
,
d
o
i
:
1
0
.
1
0
4
9
/
i
p
r
2
.
1
2
2
5
8
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
15
,
No
.
4
,
Au
g
u
s
t
20
25
:
3
7
2
7
-
3736
3736
[
1
8
]
K
.
D
o
s
h
i
a
n
d
Y
.
Y
i
l
ma
z
,
“
T
o
w
a
r
d
s
i
n
t
e
r
p
r
e
t
a
b
l
e
v
i
d
e
o
a
n
o
ma
l
y
d
e
t
e
c
t
i
o
n
,
”
i
n
2
0
2
3
I
EEE
/
C
VF
W
i
n
t
e
r
C
o
n
f
e
re
n
c
e
o
n
Ap
p
l
i
c
a
t
i
o
n
s
o
f
C
o
m
p
u
t
e
r
Vi
s
i
o
n
(
WA
C
V)
,
J
a
n
.
2
0
2
3
,
p
p
.
2
6
5
4
–
2
6
6
3
.
d
o
i
:
1
0
.
1
1
0
9
/
w
a
c
v
5
6
6
8
8
.
2
0
2
3
.
0
0
2
6
8
.
[
1
9
]
D
.
R
.
K
o
t
h
a
d
i
y
a
,
C
.
M
.
B
h
a
t
t
,
T.
S
a
b
a
,
A
.
R
e
h
m
a
n
,
a
n
d
S
.
A
.
B
a
h
a
j
,
“
S
I
G
N
F
O
R
M
E
R
:
D
e
e
p
V
i
s
i
o
n
t
r
a
n
sf
o
r
m
e
r
f
o
r
si
g
n
l
a
n
g
u
a
g
e
r
e
c
o
g
n
i
t
i
o
n
,
”
I
EEE
Ac
c
e
ss
,
v
o
l
.
1
1
,
p
p
.
4
7
3
0
–
4
7
3
9
,
2
0
2
3
,
d
o
i
:
1
0
.
1
1
0
9
/
a
c
c
e
ss.
2
0
2
2
.
3
2
3
1
1
3
0
.
[
2
0
]
D
.
K
o
t
h
a
d
i
y
a
,
C
.
B
h
a
t
t
,
K
.
S
a
p
a
r
i
y
a
,
K
.
P
a
t
e
l
,
A
.
-
B
.
G
i
l
-
G
o
n
z
á
l
e
z
,
a
n
d
J
.
M
.
C
o
r
c
h
a
d
o
,
“
D
e
e
p
s
i
g
n
:
S
i
g
n
l
a
n
g
u
a
g
e
d
e
t
e
c
t
i
o
n
a
n
d
r
e
c
o
g
n
i
t
i
o
n
u
s
i
n
g
d
e
e
p
l
e
a
r
n
i
n
g
,
”
El
e
c
t
r
o
n
i
c
s
,
v
o
l
.
1
1
,
n
o
.
1
1
,
p
.
1
7
8
0
,
J
u
n
.
2
0
2
2
,
d
o
i
:
1
0
.
3
3
9
0
/
e
l
e
c
t
r
o
n
i
c
s
1
1
1
1
1
7
8
0
.
[
2
1
]
P
.
R
a
m
a
c
h
a
n
d
r
a
n
,
B
.
Z
o
p
h
,
a
n
d
Q
.
V
.
Le
,
“
S
e
a
r
c
h
i
n
g
f
o
r
a
c
t
i
v
a
t
i
o
n
f
u
n
c
t
i
o
n
s,
”
a
rX
i
v
:
1
7
1
0
.
0
5
9
4
1
,
O
c
t
.
2
0
1
7
,
[
O
n
l
i
n
e
]
.
A
v
a
i
l
a
b
l
e
:
h
t
t
p
:
/
/
a
r
x
i
v
.
o
r
g
/
a
b
s
/
1
7
1
0
.
0
5
9
4
1
[
2
2
]
L.
X
u
,
X
.
D
i
n
g
,
D
.
Zh
a
o
,
A
.
X
.
Li
u
,
a
n
d
Z
.
Z
h
a
n
g
,
“
A
t
h
r
e
e
-
d
i
me
n
s
i
o
n
a
l
R
e
sN
e
t
a
n
d
t
r
a
n
sf
o
r
mer
-
b
a
se
d
a
p
p
r
o
a
c
h
t
o
a
n
o
m
a
l
y
d
e
t
e
c
t
i
o
n
i
n
mu
l
t
i
v
a
r
i
a
t
e
t
e
mp
o
r
a
l
–
s
p
a
t
i
a
l
d
a
t
a
,
”
E
n
t
r
o
p
y
,
v
o
l
.
2
5
,
n
o
.
2
,
p
.
1
8
0
,
Ja
n
.
2
0
2
3
,
d
o
i
:
1
0
.
3
3
9
0
/
e
2
5
0
2
0
1
8
0
.
[
2
3
]
K
.
Y
o
u
s
a
f
a
n
d
T.
N
a
w
a
z
,
“
A
n
a
t
t
e
n
t
i
o
n
m
e
c
h
a
n
i
sm
-
b
a
s
e
d
C
N
N
-
B
i
LST
M
c
l
a
ssi
f
i
c
a
t
i
o
n
mo
d
e
l
f
o
r
d
e
t
e
c
t
i
o
n
o
f
i
n
a
p
p
r
o
p
r
i
a
t
e
c
o
n
t
e
n
t
i
n
c
a
r
t
o
o
n
v
i
d
e
o
s
,
”
Mu
l
t
i
m
e
d
i
a
T
o
o
l
s
a
n
d
Ap
p
l
i
c
a
t
i
o
n
s
,
v
o
l
.
8
3
,
n
o
.
1
1
,
p
p
.
3
1
3
1
7
–
3
1
3
4
0
,
S
e
p
.
2
0
2
3
,
d
o
i
:
1
0
.
1
0
0
7
/
s
1
1
0
4
2
-
0
2
3
-
1
6
7
2
7
-
6.
[
2
4
]
T.
C
h
e
n
,
R
.
X
u
,
Y
.
H
e
,
a
n
d
X
.
W
a
n
g
,
“
I
mp
r
o
v
i
n
g
s
e
n
t
i
m
e
n
t
a
n
a
l
y
s
i
s
v
i
a
se
n
t
e
n
c
e
t
y
p
e
c
l
a
ssi
f
i
c
a
t
i
o
n
u
si
n
g
B
i
LSTM
-
C
R
F
a
n
d
C
N
N
,
”
Ex
p
e
r
t
S
y
s
t
e
m
s
w
i
t
h
A
p
p
l
i
c
a
t
i
o
n
s
,
v
o
l
.
7
2
,
p
p
.
2
2
1
–
2
3
0
,
A
p
r
.
2
0
1
7
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
e
sw
a
.
2
0
1
6
.
1
0
.
0
6
5
.
[
2
5
]
S
.
C
h
e
n
,
X
.
Zh
o
n
g
,
L.
L
i
,
W
.
L
i
u
,
C
.
G
u
,
a
n
d
L.
Z
h
o
n
g
,
“
A
d
a
p
t
i
v
e
l
y
c
o
n
v
e
r
t
i
n
g
a
u
x
i
l
i
a
r
y
a
t
t
r
i
b
u
t
e
s
a
n
d
t
e
x
t
u
a
l
e
mb
e
d
d
i
n
g
f
o
r
v
i
d
e
o
c
a
p
t
i
o
n
i
n
g
b
a
s
e
d
o
n
B
i
LS
TM
,
”
N
e
u
r
a
l
Pro
c
e
ss
i
n
g
L
e
t
t
e
rs
,
v
o
l
.
5
2
,
n
o
.
3
,
p
p
.
2
3
5
3
–
2
3
6
9
,
S
e
p
.
2
0
2
0
,
d
o
i
:
1
0
.
1
0
0
7
/
s
1
1
0
6
3
-
0
2
0
-
1
0
3
5
2
-
2.
[
2
6
]
W
.
S
u
l
t
a
n
i
,
C
.
C
h
e
n
,
a
n
d
M
.
S
h
a
h
,
“
R
e
a
l
-
w
o
r
l
d
a
n
o
ma
l
y
d
e
t
e
c
t
i
o
n
i
n
su
r
v
e
i
l
l
a
n
c
e
v
i
d
e
o
s,
”
i
n
2
0
1
8
I
EEE
/
C
V
F
C
o
n
f
e
re
n
c
e
o
n
C
o
m
p
u
t
e
r
Vi
s
i
o
n
a
n
d
P
a
t
t
e
r
n
Re
c
o
g
n
i
t
i
o
n
,
J
u
n
.
2
0
1
8
.
d
o
i
:
1
0
.
1
1
0
9
/
c
v
p
r
.
2
0
1
8
.
0
0
6
7
8
.
[
2
7
]
W
.
L
i
u
,
W
.
L
u
o
,
D
.
Li
a
n
,
a
n
d
S
.
G
a
o
,
“
F
u
t
u
r
e
f
r
a
me
p
r
e
d
i
c
t
i
o
n
f
o
r
a
n
o
mal
y
d
e
t
e
c
t
i
o
n
-
A
n
e
w
b
a
se
l
i
n
e
,
”
i
n
2
0
1
8
I
E
EE/
C
VF
C
o
n
f
e
re
n
c
e
o
n
C
o
m
p
u
t
e
r
Vi
s
i
o
n
a
n
d
Pa
t
t
e
r
n
Re
c
o
g
n
i
t
i
o
n
,
J
u
n
.
2
0
1
8
,
p
p
.
6
5
3
6
–
6
5
4
5
.
d
o
i
:
1
0
.
1
1
0
9
/
c
v
p
r
.
2
0
1
8
.
0
0
6
8
4
.
[
2
8
]
D
.
R
a
m
o
l
i
y
a
a
n
d
A
.
G
a
n
a
t
r
a
,
“
I
n
si
g
h
t
s
o
f
d
e
e
p
l
e
a
r
n
i
n
g
-
b
a
se
d
v
i
d
e
o
a
n
o
m
a
l
y
d
e
t
e
c
t
i
o
n
a
p
p
r
o
a
c
h
e
s
,
”
i
n
I
n
t
e
l
l
i
g
e
n
t
C
o
m
m
u
n
i
c
a
t
i
o
n
T
e
c
h
n
o
l
o
g
i
e
s
a
n
d
V
i
rt
u
a
l
Mo
b
i
l
e
N
e
t
w
o
rks
,
S
p
r
i
n
g
e
r
N
a
t
u
r
e
S
i
n
g
a
p
o
r
e
,
2
0
2
3
,
p
p
.
6
6
3
–
6
7
6
.
d
o
i
:
1
0
.
1
0
0
7
/
9
7
8
-
9
8
1
-
99
-
1
7
6
7
-
9
_
4
8
.
[
2
9
]
A
.
S
h
e
w
a
l
k
a
r
,
D
.
N
y
a
v
a
n
a
n
d
i
,
a
n
d
S
.
A
.
Lu
d
w
i
g
,
“
P
e
r
f
o
r
m
a
n
c
e
e
v
a
l
u
a
t
i
o
n
o
f
d
e
e
p
n
e
u
r
a
l
n
e
t
w
o
r
k
s a
p
p
l
i
e
d
t
o
s
p
e
e
c
h
r
e
c
o
g
n
i
t
i
o
n
:
R
N
N
,
LSTM
a
n
d
G
R
U
,
”
J
o
u
r
n
a
l
o
f
Art
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
a
n
d
S
o
f
t
C
o
m
p
u
t
i
n
g
Re
s
e
a
r
c
h
,
v
o
l
.
9
,
n
o
.
4
,
p
p
.
2
3
5
–
2
4
5
,
A
u
g
.
2
0
1
9
,
d
o
i
:
1
0
.
2
4
7
8
/
j
a
i
scr
-
2
0
1
9
-
0
0
0
6
.
[
3
0
]
C
.
Z
h
a
n
g
e
t
a
l
.
,
“
R
e
sN
e
t
o
r
D
e
n
s
e
N
e
t
?
I
n
t
r
o
d
u
c
i
n
g
d
e
n
se
s
h
o
r
t
c
u
t
s
t
o
R
e
s
N
e
t
,
”
i
n
2
0
2
1
I
EE
E
Wi
n
t
e
r
C
o
n
f
e
re
n
c
e
o
n
Ap
p
l
i
c
a
t
i
o
n
s
o
f
C
o
m
p
u
t
e
r V
i
si
o
n
(
W
AC
V)
,
Ja
n
.
2
0
2
1
.
d
o
i
:
1
0
.
1
1
0
9
/
w
a
c
v
4
8
6
3
0
.
2
0
2
1
.
0
0
3
5
9
.
[
3
1
]
D
.
Th
e
c
k
e
d
a
t
h
a
n
d
R
.
R
.
S
e
d
a
m
k
a
r
,
“
D
e
t
e
c
t
i
n
g
a
f
f
e
c
t
s
t
a
t
e
s u
si
n
g
V
G
G
1
6
,
R
e
sN
e
t
5
0
a
n
d
S
E
-
R
e
sN
e
t
5
0
n
e
t
w
o
r
k
s
,
”
S
N
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
,
v
o
l
.
1
,
n
o
.
2
,
M
a
r
.
2
0
2
0
,
d
o
i
:
1
0
.
1
0
0
7
/
s
4
2
9
7
9
-
0
2
0
-
0
1
1
4
-
9.
B
I
O
G
RAP
H
I
E
S O
F
AUTH
O
RS
Dip
a
k
Ra
m
o
li
y
a
re
c
e
iv
e
d
th
e
B
.
En
g
g
.
a
n
d
M
.
E
n
g
g
.
d
e
g
re
e
s
in
c
o
m
p
u
ter
e
n
g
in
e
e
rin
g
fr
o
m
G
u
jara
t
Tec
h
n
o
lo
g
ica
l
Un
iv
e
rsit
y
,
G
u
jara
t,
in
2
0
1
3
a
n
d
2
0
1
5
re
sp
e
c
ti
v
e
ly
.
Cu
rre
n
tl
y
,
h
e
is
a
n
a
ss
istan
t
p
ro
fe
ss
o
r
a
t
t
h
e
De
p
a
rtme
n
t
o
f
Co
m
p
u
ter
S
c
ien
c
e
a
n
d
En
g
i
n
e
e
rin
g
,
Ch
a
ro
tar
Un
iv
e
rsit
y
o
f
S
c
ien
c
e
a
n
d
Tec
h
n
o
lo
g
y
,
C
h
a
n
g
a
,
G
u
jara
t.
His
re
se
a
rc
h
in
tere
sts
in
c
lu
d
e
m
a
c
h
in
e
lea
rn
i
n
g
,
d
e
e
p
lea
rn
i
n
g
,
c
lo
u
d
c
o
m
p
u
ti
n
g
a
n
d
h
i
g
h
-
p
e
rfo
rm
a
n
c
e
c
o
m
p
u
ti
n
g
.
He
h
a
s
e
x
p
e
rti
se
i
n
d
a
tab
a
se
sy
ste
m
s
a
n
d
d
a
tab
a
s
e
d
e
sig
n
s.
He
h
a
s
wo
r
k
e
d
o
n
d
a
tab
a
se
d
e
sig
n
s
o
f
v
a
rio
u
s
ERP
s
a
n
d
a
p
p
li
c
a
ti
o
n
s.
He
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
d
ip
a
k
ra
m
o
l
iy
a
.
c
e
@c
h
a
ru
sa
t.
a
c
.
in
.
Am
it
G
a
n
a
tr
a
re
c
e
iv
e
d
a
P
h
.
D.
d
e
g
re
e
in
c
o
m
p
u
ter
e
n
g
i
n
e
e
rin
g
fro
m
Ka
d
i
S
a
rv
a
Vish
wa
Vid
y
a
lay
a
(KSV),
G
a
n
d
h
in
a
g
a
r,
G
u
jara
t,
In
d
ia.
He
h
a
s
b
e
e
n
se
rv
in
g
a
s
th
e
P
r
o
v
o
st
o
f
P
a
ru
l
Un
i
v
e
rsity
,
Va
d
o
d
a
ra
sin
c
e
Au
g
u
st
2
0
2
2
.
Be
fo
re
t
h
a
t,
h
e
se
rv
e
d
a
s
th
e
De
a
n
,
F
a
c
u
lt
y
o
f
Tec
h
n
o
l
o
g
y
-
CHA
RUSAT,
G
u
jara
t.
His
a
re
a
s
o
f
in
tere
st
in
c
l
u
d
e
a
rti
ficia
l
in
telli
g
e
n
c
e
,
m
a
c
h
in
e
lea
rn
in
g
a
n
d
d
e
e
p
lea
rn
i
n
g
.
He
is
a
n
a
c
ti
v
e
m
e
m
b
e
r
o
f
IE
EE
(S
r
.
M
e
m
b
e
r),
ACM
a
n
d
CS
I
p
ro
fe
ss
io
n
a
l
so
c
iety
c
h
a
p
te
rs
.
He
h
a
s
su
p
e
rv
ise
d
(
g
u
i
d
e
d
)
m
o
re
th
a
n
1
5
0
+
i
n
d
u
stry
p
ro
jec
ts
a
t
th
e
g
ra
d
u
a
te
lev
e
l
a
n
d
1
0
0
+
d
isse
rtatio
n
s
a
t
th
e
g
ra
d
u
a
te
lev
e
l.
1
5
P
h
.
D
.
S
c
h
o
lars
h
a
v
e
c
o
m
p
lete
d
t
h
e
ir
P
h
.
D.
u
n
d
e
r
h
is
su
p
e
r
v
isio
n
a
n
d
c
u
rre
n
t
ly
,
h
e
is
su
p
e
rv
isi
n
g
(g
u
i
d
in
g
)
0
8
re
se
a
rc
h
sc
h
o
lars
a
t
th
e
P
h
.
D.
lev
e
l
a
n
d
0
2
sc
h
o
lars
a
t
th
e
P
o
std
o
c
l
e
v
e
l.
Ema
il
:
g
a
n
a
traa
m
it
p
@g
m
a
il
.
c
o
m
.
Evaluation Warning : The document was created with Spire.PDF for Python.