I
A
E
S
I
n
t
e
r
n
at
io
n
al
Jou
r
n
al
of
A
r
t
if
ic
ia
l
I
n
t
e
ll
ig
e
n
c
e
(
I
J
-
AI
)
V
ol
.
14
, N
o.
2
,
A
pr
il
2025
, pp.
1471
~
1483
I
S
S
N
:
2252
-
8938
,
D
O
I
:
10.11591/
ij
a
i.
v
14
.i
2
.pp
1471
-
1483
1471
Jou
r
n
al
h
om
e
page
:
ht
tp
:
//
ij
ai
.
ia
e
s
c
or
e
.c
om
V
i
d
e
o f
or
ge
r
y:
A
n
e
xt
e
n
si
ve
an
al
ysi
s
of
i
n
t
e
r
-
an
d
i
n
t
ra
-
f
r
am
e
m
an
i
p
u
l
at
i
o
n
al
on
gs
i
d
e
st
at
e
-
of
-
t
h
e
-
a
r
t
c
om
p
ar
i
son
s
S
u
m
ai
ya S
h
ai
k
h
, S
a
t
h
is
h
K
u
m
ar
K
an
n
ai
ah
D
e
pa
r
t
m
e
nt
of
C
om
put
e
r
S
c
i
e
nc
e
a
nd E
ngi
ne
e
r
i
ng, K
one
r
u L
a
ks
hm
a
i
a
h E
duc
a
t
i
on F
ounda
t
i
on, V
a
dde
s
w
a
r
a
m
, I
ndi
a
A
r
t
ic
le
I
n
f
o
A
B
S
T
R
A
C
T
A
r
ti
c
le
h
is
to
r
y
:
R
e
c
e
iv
e
d
M
a
y 9, 2024
R
e
vi
s
e
d
N
ov 7, 2024
A
c
c
e
pt
e
d
N
ov 14, 2024
The
widespread
accessibility
of
inexpensive
mobile
phones,
digital
ca
meras,
camcorders,
and
security
closed
-
circuit
televisi
on
(
CCTV
)
camer
as
has
resulted
in
the
integratio
n
of
filmmakin
g
into
our
everyday
exi
stence.
YouTube, Facebook, Ins
tagram, and
S
napchat are
a few
of the
v
ideo
-
s
haring
and
editing
applications
that
facilitate
the
proc
ess
of
uploading
and
editing
videos.
Additional
instances
include
Adobe
Photoshop,
Windows
Movie
Maker,
and
Video
Editor.
Although
editing
has
its
advantages,
the
re
is
a
po
tential
risk
of
counterf
eiting.
This
occurs
when
films
are
edited
w
ith
the
intention
of
misleading
viewers
or
manipulating
their
perspec
tives,
which
can
be
particularly
troublesome
in
judicial
procedures
where
recordi
ngs
are
submitted
as
evidence.
The
issue
has
been
exacerba
ted
by
the
emergence
of
deep
learning
methods,
such
as
deepfake
videos
that
effectively
man
ipulate
facial
characteristics.
Consequently,
individuals
have
become
less
reliant
on
visual
evidence.
These
issues
emphasise
the
pressing
necessi
ty
f
or
the
creation
of
dependable
methods
to
determin
e
the
authenti
city
of
fil
ms
and
identify
cases
of
fraud.
Contemporar
y
methods
can
depend
on
ass
essing
modified
frames
or
utilising
distortions
generate
d
during
video
codec
compressi
on
or
double
compressi
on.
Since
2016,
multi
ple
studies
hav
e
been
undertaken
to
investigate
techniques,
strategies,
and
applications
to
tackle
this
problem.
The
objective
of
this
survey
study
is
to
pro
vide
a
comprehens
ive
analysis
of
these
algorit
hms,
highli
ghting
their
adva
ntag
es
and disad
vantages i
n detecti
ng differen
t forms
of video
forgeries.
K
e
y
w
o
r
d
s
:
A
na
ly
s
is
D
ig
it
a
l
vi
de
o f
or
e
ns
ic
s
M
ul
ti
m
e
di
a
f
or
e
ns
ic
s
V
id
e
o a
ut
he
nt
ic
a
ti
on
V
id
e
o t
a
m
pe
r
in
g de
te
c
ti
on
This is an
open
acce
ss artic
le unde
r the
CC BY
-
SA
license.
C
or
r
e
s
pon
di
n
g A
u
th
or
:
S
a
th
is
h K
um
a
r
K
a
nna
i
a
h
D
e
pa
r
tm
e
nt
of
C
om
put
e
r
S
c
ie
nc
e
a
nd
E
ngi
ne
e
r
in
g, K
one
r
u L
a
ks
hm
a
ia
h E
duc
a
ti
on F
ounda
ti
on
V
a
dde
s
w
a
r
a
m
, V
ij
a
y
a
w
a
da
, A
ndhr
a
P
r
a
s
e
dh, I
ndi
a
E
m
a
il
:
ks
a
th
is
h1980@
gm
a
il
.c
om
1.
I
N
T
R
O
D
U
C
T
I
O
N
E
ve
r
y
da
y,
c
ount
le
s
s
vi
de
os
a
r
e
upl
oa
d
e
d
to
th
e
in
te
r
ne
t,
a
s
ig
ni
f
ic
a
nt
por
ti
on
of
w
hi
c
h
unde
r
g
o
e
di
ti
ng.
T
hi
s
pa
pe
r
f
oc
us
e
s
on
th
e
de
te
c
ti
on
of
vi
de
o
f
or
ge
r
y
th
r
ough
a
n
e
xt
e
ns
iv
e
a
na
ly
s
is
of
bot
h
in
te
r
-
f
r
a
m
e
a
nd
in
tr
a
-
f
r
a
m
e
m
a
ni
pul
a
ti
ons
,
c
om
pa
r
in
g
th
e
s
e
te
c
hni
que
s
w
it
h
s
ta
te
-
of
-
th
e
a
r
t
m
e
th
ods
.
V
id
e
o
f
or
ge
r
y
in
vol
ve
s
th
e
a
lt
e
r
a
ti
on
of
vi
de
o
c
ont
e
nt
to
m
is
le
a
d
vi
e
w
e
r
s
,
pos
in
g
s
ig
ni
f
ic
a
nt
c
h
a
ll
e
nge
s
f
or
di
gi
ta
l
f
or
e
ns
ic
s
a
nd
s
e
c
ur
it
y.
I
nt
e
r
-
f
r
a
m
e
m
a
ni
pul
a
ti
on
in
c
lu
de
s
te
c
h
ni
que
s
s
uc
h
a
s
f
r
a
m
e
dupl
ic
a
ti
ons
or
de
le
ti
on,
w
hi
c
h
di
s
r
upt
th
e
te
m
por
a
l
c
ons
is
te
nc
y
of
a
vi
de
o.
I
n
tr
a
-
f
r
a
m
e
m
a
ni
pul
a
ti
on
in
vol
ve
s
a
lt
e
r
in
g
in
di
vi
dua
l
f
r
a
m
e
s
,
s
uc
h
a
s
s
pl
ic
in
g
or
c
ha
ngi
ng
obj
e
c
ts
w
it
hi
n
f
r
a
m
e
s
,
a
f
f
e
c
ti
ng
th
e
s
pa
ti
a
l
in
te
gr
it
y
of
th
e
vi
de
o.
A
r
e
c
e
nt
s
tu
dy
ha
s
pr
im
a
r
il
y
c
onc
e
nt
r
a
te
d
on
th
e
id
e
nt
if
ic
a
ti
on
of
m
a
ni
pul
a
te
d
r
e
c
or
di
ngs
,
pa
r
ti
c
ul
a
r
ly
th
os
e
th
a
t
ha
ve
be
e
n
m
odi
f
ie
d
to
c
h
a
nge
f
a
c
ia
l
c
ha
r
a
c
te
r
is
ti
c
s
.
A
lt
hough
di
gi
ta
l
im
a
ge
f
or
e
ns
ic
te
c
hni
que
s
ha
ve
m
a
de
s
ig
ni
f
ic
a
nt
a
dva
nc
e
s
in
e
s
ta
bl
is
hi
ng
im
a
ge
a
ut
he
nt
ic
it
y
a
nd
pr
oc
e
s
s
in
g
hi
s
to
r
y,
vi
de
o
f
or
e
ns
ic
s
r
e
m
a
in
s
a
n
e
xpa
ndi
ng
to
pi
c
of
in
te
r
e
s
t
f
or
r
e
s
e
a
r
c
he
r
s
. O
ur
r
e
s
e
a
r
c
h
a
d
dr
e
s
s
e
s
a
c
r
it
ic
a
l
ne
e
d
f
or
r
e
li
a
bl
e
a
nd
e
f
f
ic
ie
nt
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
, V
ol
.
14
, N
o.
2
,
A
pr
il
2025
:
1471
-
1483
1472
vi
de
o
f
or
ge
r
y
de
te
c
ti
on
m
e
th
ods
.
C
ur
r
e
nt
s
ta
te
-
of
-
th
e
-
a
r
t
te
c
hni
que
s
of
te
n
s
tr
uggl
e
w
it
h
e
it
he
r
hi
gh
f
a
ls
e
pos
it
iv
e
/n
e
ga
ti
ve
r
a
te
s
or
c
om
put
a
ti
ona
l
in
e
f
f
ic
ie
nc
y.
O
ur
s
tu
d
y
in
tr
oduc
e
s
a
nove
l
a
ppr
oa
c
h
th
a
t
c
om
bi
ne
s
a
dva
nc
e
d
s
pa
ti
a
l
a
nd
te
m
por
a
l
a
na
ly
s
is
to
im
pr
ove
de
te
c
ti
on
a
c
c
ur
a
c
y
w
hi
le
m
a
in
ta
in
in
g
c
om
put
a
ti
ona
l
e
f
f
ic
ie
nc
y.
T
hi
s
w
or
k
c
ont
r
ib
ut
e
s
to
th
e
f
ie
ld
by
of
f
e
r
in
g
a
m
or
e
r
obus
t
s
ol
ut
io
n
f
or
id
e
nt
if
yi
ng
bot
h
in
te
r
-
f
r
a
m
e
a
nd
in
tr
a
-
f
r
a
m
e
f
or
ge
r
ie
s
,
th
us
e
nha
nc
in
g
di
gi
ta
l
f
or
e
ns
ic
c
a
pa
bi
li
te
s
a
nd
a
id
in
g
in
th
e
f
ig
ht
a
ga
in
s
t
m
is
in
f
or
m
a
ti
on
a
nd
di
gi
ta
l
f
r
a
ud.
T
h
e
te
r
m
"
f
or
e
ns
ic
s
"
is
de
r
iv
e
d
f
r
om
th
e
w
or
d
"
f
or
e
ns
ic
,"
a
nd
la
w
e
nf
or
c
e
m
e
nt
a
ge
nc
ie
s
a
r
e
ge
ne
r
a
ll
y
he
s
it
a
nt
to
a
c
c
e
pt
f
il
m
s
a
s
c
r
e
di
bl
e
e
vi
de
nc
e
in
th
e
a
b
s
e
nc
e
of
f
or
e
ns
ic
r
e
por
ts
.
I
ndi
vi
dua
l
c
om
pone
nt
s
of
a
vi
de
o,
known
a
s
"
f
oot
pr
in
ts
,"
pl
a
y
a
n
im
po
r
ta
nt
pa
r
t
in
a
s
s
e
s
s
in
g
it
s
le
gi
ti
m
a
c
y.
A
s
a
r
e
s
ul
t,
f
or
e
ns
ic
a
na
ly
s
is
pl
a
ys
a
n
e
s
s
e
nt
ia
l
r
ol
e
in
a
r
e
a
s
s
uc
h
a
s
ne
w
s
r
e
por
ti
ng,
c
r
im
in
a
l
in
ve
s
ti
ga
ti
ons
, a
nd i
nt
e
ll
ig
e
nc
e
op
e
r
a
ti
ons
[
1]
, [
2]
.
T
hi
s
s
tu
dy
in
ve
s
ti
ga
te
s
w
it
h
th
e
c
ol
le
c
ti
on
of
e
vi
de
nc
e
a
s
de
pi
c
te
d
in
F
ig
ur
e
1,
m
uc
h
li
ke
a
s
ta
nda
r
d
c
r
im
e
s
c
e
ne
in
ve
s
ti
ga
ti
on.
E
a
r
li
e
r
th
e
s
tu
dy
in
vol
ve
s
tr
a
di
ti
o
na
l
in
ve
s
ti
ga
ti
ve
pr
oc
e
dur
e
w
hi
c
h
hol
ds
th
e
im
pa
c
t
a
nd
im
por
ta
nc
e
of
m
ul
ti
m
e
di
a
da
ta
in
f
or
e
ns
ic
in
ve
s
ti
ga
ti
on.
E
a
r
ly
m
e
th
ods
ha
ve
not
e
xpl
ic
it
ly
a
ddr
e
s
s
e
d
th
e
in
f
lu
e
nc
e
of
in
ve
s
ti
ga
ti
on
on
m
ul
ti
m
e
di
a
da
ta
.
H
ow
e
ve
r
,
in
th
e
c
ont
e
xt
of
f
or
e
ns
ic
in
ve
s
ti
ga
ti
ons
in
to
di
gi
ta
l
m
e
di
a
,
th
e
e
vi
de
nc
e
is
m
os
tl
y
r
e
la
te
d
to
e
le
c
tr
oni
c
de
vi
c
e
s
th
a
t
c
ont
a
in
m
ul
ti
m
e
di
a
da
ta
.
T
h
e
in
it
ia
l
s
ta
ge
,
known
a
s
"
a
c
qui
s
it
io
n,"
c
om
pr
is
e
s
ga
th
e
r
in
g
in
f
or
m
a
ti
on
on
th
e
na
tu
r
e
of
th
e
e
vi
d
e
nc
e
a
t
ha
nd.
T
he
a
ppr
oa
c
h
th
e
n
f
oc
us
e
s
on
de
te
r
m
in
in
g
th
e
e
vi
de
nc
e
'
s
c
ont
e
xt
,
w
hi
c
h
c
a
n
be
di
vi
de
d
in
to
th
r
e
e
c
a
te
gor
ie
s
:
phys
i
c
a
l,
lo
gi
c
a
l,
a
nd
c
hr
onol
ogi
c
a
l.
T
hi
s
c
la
s
s
if
ic
a
ti
on
la
ys
th
e
gr
oundwor
k
f
or
th
e
ne
xt
pha
s
e
,
"
e
va
lu
a
ti
on,"
in
w
hi
c
h
s
pe
c
ia
li
z
e
d
te
c
hnol
ogy
a
nd
m
e
th
od
ol
ogi
e
s
a
r
e
us
e
d
to
a
s
s
e
s
s
th
e
in
f
or
m
a
ti
on
c
ont
a
in
e
d
in
th
e
e
vi
de
nc
e
.
A
f
te
r
a
th
or
ough
e
xa
m
in
a
ti
on,
e
vi
de
nc
e
is
d
e
e
m
e
d
a
c
c
e
pt
a
bl
e
f
or
f
ur
th
e
r
in
ve
s
ti
ga
ti
on a
nd pote
nt
ia
l
us
e
i
n j
udi
c
ia
l
pr
oc
e
e
di
ngs
[
1]
.
F
ig
ur
e
1. T
he
pr
oc
e
dur
e
f
or
i
nve
s
ti
ga
ti
ng vide
os
T
he
r
e
a
r
e
num
e
r
ou
s
s
ubf
ie
ld
s
w
it
hi
n
di
gi
ta
l
f
or
e
ns
ic
s
.
D
if
f
e
r
e
nt
ty
pe
s
a
r
e
c
a
te
gor
iz
e
d
ba
s
e
d
on
th
e
s
pe
c
if
ic
us
e
.
T
he
pa
p
e
r
'
s
pr
im
a
r
y
e
m
pha
s
is
is
on
di
gi
ta
l
da
ta
-
de
pe
nde
nt
c
om
put
e
r
f
or
e
ns
ic
s
.
W
h
e
n
s
or
ti
ng
da
ta
by
ty
pe
,
m
ul
ti
m
e
di
a
f
or
e
ns
ic
s
a
ls
o
f
a
ll
s
in
to
th
is
gr
oup.
F
ig
ur
e
2
il
lu
s
tr
a
te
s
di
f
f
e
r
e
nt
br
a
nc
he
s
of
di
gi
ta
l
f
or
e
ns
ic
s
,
w
hi
c
h
in
vol
ve
a
va
r
ie
ty
of
a
ppr
oa
c
he
s
u
s
e
d
to
r
e
c
ove
r
a
nd
a
na
ly
z
e
da
ta
f
r
om
di
gi
ta
l
de
vi
c
e
s
.
E
s
s
e
nt
ia
ll
y,
di
gi
ta
l
f
or
e
ns
ic
s
is
th
e
m
e
th
odi
c
a
l
p
r
oc
e
s
s
of
e
xt
r
a
c
ti
ng
a
nd
e
xa
m
in
in
g
da
ta
c
ont
a
in
e
d
i
n
e
le
c
tr
oni
c
de
vi
c
e
s
.
F
ig
ur
e
2. C
a
te
gor
ie
s
of
f
or
e
ns
ic
s
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
V
id
e
o f
or
ge
r
y
:
A
n e
x
te
ns
iv
e
analy
s
i
s
of
i
nt
e
r
-
and in
tr
a
-
fr
am
e
m
ani
pul
at
io
n al
ongs
id
e
…
(
Sum
ai
y
a Shaik
h
)
1473
2.
L
I
T
E
R
A
T
U
R
E
S
U
R
V
E
Y
T
h
e
f
i
e
ld
of
v
id
e
o
f
or
e
n
s
i
c
s
ha
s
in
c
r
e
a
s
e
d
i
n
pop
ul
a
r
it
y
du
e
to
it
s
w
i
de
r
a
ng
e
of
a
ppl
ic
a
t
io
n
s
in
t
he
c
ur
r
e
n
t
w
or
l
d.
T
h
is
s
e
c
t
io
n
pr
ov
id
e
s
a
n
o
ve
r
vi
e
w
of
t
he
s
e
ve
r
a
l
e
s
t
a
b
li
s
h
e
d
m
e
t
hod
ol
o
gi
e
s
in
vi
de
o
f
or
e
n
s
i
c
s
.
A
s
s
t
a
t
e
d
i
n
L
e
e
t
a
l.
[
3
]
,
it
w
a
s
d
if
f
i
c
u
lt
to
d
e
te
c
t
c
o
u
nt
e
r
f
e
it
in
f
o
r
m
a
ti
on
i
n
a
c
o
m
p
r
e
s
s
e
d
vi
d
e
o
s
a
m
pl
e
.
T
he
r
a
ti
on
a
le
f
or
th
i
s
i
s
t
h
a
t
c
om
pr
e
s
s
io
n
h
a
s
t
h
e
c
a
p
a
c
it
y
to
e
li
m
i
n
a
t
e
a
l
l
e
vi
de
n
c
e
of
f
or
g
e
r
ie
s
.
C
on
d
uc
t
e
d
a
t
ho
r
o
u
gh
a
n
a
ly
s
i
s
o
f
t
h
e
t
o
p
ic
of
f
or
e
n
s
i
c
s
,
s
p
e
c
i
f
i
c
a
l
ly
i
n
t
e
r
m
s
of
c
on
te
n
t
a
u
t
he
nt
i
c
i
ty
a
nd
t
h
e
de
t
e
c
t
io
n
o
f
v
a
r
io
u
s
s
o
r
t
s
of
f
or
g
e
r
i
e
s
,
in
c
lu
di
n
g
t
h
e
p
ot
e
n
ti
a
l
c
a
t
e
go
r
i
z
a
t
io
n
of
v
id
e
o
t
a
m
p
e
r
in
g
m
e
t
h
od
s
.
S
i
ng
h
a
nd
A
g
g
a
r
w
a
l
[
4]
publ
is
he
d
hi
s
f
in
di
ngs
in
a
r
e
c
ogni
z
e
d
jo
ur
na
l,
f
oc
us
in
g
on
one
of
th
e
m
e
th
ods
us
e
d
to
de
te
c
t
vi
de
o
ta
m
pe
r
in
g.
T
he
pa
pe
r
s
in
ve
s
ti
ga
te
a
nd
de
s
c
r
ib
e
a
va
r
ie
ty
of
ta
m
pe
r
in
g
s
tr
a
te
gi
e
s
.
Y
a
o
e
t
al
.
[
5]
r
e
s
e
a
r
c
he
d
vi
de
o
f
or
ge
r
ie
s
,
f
oc
us
in
g
on
de
te
c
ti
ng
im
a
ge
c
ha
nge
r
a
th
e
r
th
a
n
vi
de
o
ta
m
pe
r
in
g.
H
e
s
pe
c
if
ic
a
ll
y
e
x
a
m
in
e
d
th
e
lo
c
a
li
z
a
ti
on
c
ha
ll
e
nge
s
r
e
la
te
d
w
it
h
vi
de
o
f
or
ge
r
ie
s
a
nd
lo
oke
d
in
to
s
e
ve
r
a
l
ta
m
pe
r
in
g
m
e
th
ods
.
B
a
r
ni
a
nd
C
os
ta
nz
o
[
6]
s
ugge
s
te
d
f
or
ge
r
y
de
te
c
ti
on
m
e
th
ods
th
a
t
e
nc
ount
e
r
e
d
ge
ne
r
a
li
z
a
ti
on
is
s
u
e
s
in
th
e
da
ta
s
e
t.
T
he
a
lg
or
it
hm
s
a
r
e
bui
lt
us
in
g
de
e
p
le
a
r
ni
ng,
w
it
h
a
f
oc
us
on
e
va
lu
a
ti
ng
a
nd
di
a
gnos
in
g
vi
de
o
f
r
a
ud.
P
a
di
n
a
nd
G
onz
a
le
z
[
7]
e
xpl
ic
it
ly
in
ve
s
ti
ga
te
th
e
us
e
of
lo
w
-
c
om
pl
e
xi
ty
te
ns
or
r
e
pr
e
s
e
nt
a
ti
ons
f
or
th
is
pur
pos
e
.
A
s
a
r
e
s
ul
t,
th
e
s
c
hol
a
r
s
be
c
a
m
e
ve
r
y
in
te
r
e
s
te
d
in
e
xa
m
in
in
g
th
e
c
om
pl
e
xi
ty
of
f
or
ge
r
ie
s
a
nd
th
e
ir
va
r
io
us
f
or
m
s
.
I
ni
ti
a
ll
y
,
onl
y
two
s
e
pa
r
a
t
e
c
a
te
gor
ie
s
w
e
r
e
w
id
e
ly
id
e
nt
if
ie
d
a
s
a
qua
nd
a
r
y
in
r
e
ga
r
d
to
f
ix
e
d
a
nd
m
ovi
ng
im
a
ge
s
.
T
he
vi
de
o
c
a
n
be
a
lt
e
r
e
d
us
in
g
two
m
e
th
ods
:
in
s
e
r
ti
on
a
nd
de
le
ti
on
.
A
c
c
or
di
ng
to
a
s
tu
dy
c
onduc
te
d
by
A
m
e
r
in
i
a
nd
C
a
ld
e
ll
i
a
nd
publ
is
he
d
by
M
D
P
I
in
2021,
m
a
c
hi
n
e
le
a
r
ni
ng
a
lg
or
it
hm
s
a
r
e
us
e
d
to
r
e
c
ogni
z
e
a
nd
di
f
f
e
r
e
nt
ia
te
c
ount
e
r
f
e
it
a
nd
a
ut
he
nt
ic
m
ul
ti
m
e
di
a
f
il
e
s
,
a
s
w
e
ll
a
s
pr
ovi
de
in
s
ig
ht
s
in
to
th
e
pr
e
s
e
nc
e
of
s
uc
h
in
f
or
m
a
ti
on. T
hi
s
not
io
n l
e
a
ds
t
o t
he
c
r
e
a
ti
on of
a
ut
os
py, a
di
gi
ta
l
f
or
e
ns
ic
t
ool
t
ha
t
in
t
r
oduc
e
s
t
r
a
ns
f
or
m
a
ti
on
te
c
hni
que
s
f
or
th
e
f
ir
s
t
ti
m
e
.
T
hi
s
pr
ogr
a
m
m
e
us
e
s
a
c
o
m
bi
na
ti
on
of
tr
a
ns
f
or
m
a
ti
on
te
c
hni
que
s
a
nd
m
ul
ti
m
e
di
a
da
ta
.
D
e
ns
it
y f
unc
ti
ona
l
th
e
or
y (
D
F
T
)
m
e
th
od i
s
us
e
d i
n t
he
pr
oc
e
s
s
in
g of
di
gi
ta
l
vi
de
o f
r
a
m
e
s
.
C
ons
e
c
ut
iv
e
di
gi
ta
l
vi
de
os
a
r
e
c
onve
r
te
d
in
to
in
di
vi
dua
l
f
r
a
m
e
s
by
f
r
a
m
e
r
a
te
c
onve
r
s
io
n,
w
hi
c
h
a
c
c
ur
a
te
ly
c
a
pt
ur
e
s
th
e
f
e
e
li
ng
of
m
ove
m
e
nt
a
nd
di
s
pl
a
ys
a
hi
gh
r
a
te
of
r
a
pi
d
s
uc
c
e
s
s
io
n
[
8]
.
V
id
e
o
f
or
ge
r
y
e
nc
om
pa
s
s
e
s
a
ny
m
a
li
c
io
us
c
ont
e
nt
th
a
t
br
e
a
c
he
s
th
e
in
te
gr
it
y
or
vi
s
ua
l
r
e
pr
e
s
e
nt
a
ti
on
of
th
e
vi
de
o.
T
he
s
e
ve
r
a
l
ty
pe
s
of
vi
de
o
f
or
ge
r
ie
s
a
r
e
c
la
s
s
if
ie
d
ba
s
e
d
on
th
e
t
e
c
hni
que
of
is
ol
a
ti
ng
f
r
a
m
e
s
,
a
ddi
ng
c
ont
e
nt
known
a
s
f
r
a
m
e
in
s
e
r
ti
on,
a
nd
r
e
m
ovi
ng
c
ont
e
nt
f
r
om
m
odi
f
ie
d
f
r
a
m
e
s
c
a
ll
e
d
f
r
a
m
e
de
le
ti
on.
T
he
in
it
ia
l
c
a
te
gor
y
is
r
e
f
e
r
r
e
d
to
a
s
a
c
opy
-
m
ove
a
tt
a
c
k,
a
s
e
lu
c
id
a
te
d
by
L
a
tt
a
s
e
t
al
.
[
9]
.
T
hi
s
ty
pe
of
f
or
ge
r
y
e
nt
a
il
s
th
e
r
e
pl
ic
a
ti
on of
s
pe
c
if
ic
f
r
a
m
e
s
f
r
om
one
a
r
e
a
a
nd t
he
ir
i
ns
e
r
ti
on i
nt
o a
di
f
f
e
r
e
nt
a
r
e
a
i
ns
id
e
t
he
s
a
m
e
f
r
a
m
e
,
le
a
di
ng t
o t
he
c
r
e
a
ti
on of
i
nt
e
r
-
f
r
a
m
e
f
or
ge
r
y.
A
c
c
or
d
in
g
to
H
w
a
n
g
[
10]
,
th
e
vi
d
e
o
is
tr
a
n
s
f
or
m
e
d
i
nt
o
in
di
vi
dua
l
f
r
a
m
e
s
by
a
p
pl
yi
ng
a
s
pe
c
if
ic
f
r
a
m
e
c
o
nve
r
s
io
n
bi
t
r
a
t
e
t
o i
m
a
ge
s
of
a
f
ix
e
d s
iz
e
. T
hi
s
r
a
te
de
t
e
r
m
in
e
s
t
h
e
r
a
t
e
a
t
w
hi
c
h f
r
a
m
e
s
a
r
e
c
onve
r
t
e
d.
T
he
m
e
th
od
of
a
ddi
n
g a
nd
r
e
m
ovi
ng
f
r
a
m
e
s
i
s
c
ont
i
nge
n
t
up
on t
he
f
r
a
m
e
r
a
te
. T
o
de
c
r
e
a
s
e
th
e
f
r
a
m
e
r
a
te
up
–
c
onve
r
s
io
n
(
F
R
U
C
)
r
a
t
e
,
im
pl
e
m
e
n
te
d a
m
e
t
hod
th
a
t
c
ha
ng
e
s
f
r
a
m
e
s
f
r
om
hi
gh
e
r
to
l
ow
e
r
f
r
a
m
e
s
by
in
s
e
r
ti
ng
in
te
r
pol
a
te
d f
r
a
m
e
s
.
M
e
ht
a
i
nt
r
o
duc
e
s
a
n
a
lt
e
r
n
a
te
ki
nd.
T
hi
s
c
a
t
e
gor
y f
a
l
ls
w
it
hi
n t
he
a
c
ti
v
e
a
ppr
oa
c
h of
vi
de
o
f
or
ge
r
y
a
nd
i
s
c
la
s
s
if
i
e
d
a
s
th
e
s
e
c
ond
t
ype
of
dom
a
in
f
or
ge
r
y
,
a
s
s
ta
t
e
d
by
L
yu
[
1
1]
.
I
n
th
is
c
a
te
gor
y,
th
e
s
pa
ti
ot
e
m
por
a
l
dom
a
in
is
r
e
g
a
r
de
d
to
be
s
ta
te
-
of
-
th
e
-
a
r
t.
T
he
a
c
t
of
in
s
e
r
ti
ng
unf
a
m
il
ia
r
it
e
m
s
in
to
e
xi
s
ti
ng
s
tr
uc
tu
r
e
s
is
a
n
in
s
ta
nc
e
of
a
r
e
gi
on
s
pl
ic
in
g
a
tt
a
c
k,
w
hi
c
h
th
e
w
r
it
e
r
ha
s
e
xpl
a
in
e
d
a
lo
ng
w
it
h
va
r
io
us
ot
he
r
c
om
m
on
a
tt
a
c
ks
.
Y
a
ng
e
t
al
.
[
12]
p
r
opos
e
a
two
-
s
ta
ge
m
e
th
od
th
a
t
ut
il
is
e
s
s
in
gul
a
r
va
lu
e
de
c
om
pos
it
io
n
(
S
V
D
)
f
e
a
tu
r
e
e
xt
r
a
c
ti
on
to
c
a
lc
ul
a
te
th
e
c
or
r
e
la
ti
on
c
o
e
f
f
ic
ie
nt
s
im
il
a
r
it
ie
s
be
twe
e
n
f
r
a
m
e
s
.
T
hi
s
a
ppr
oa
c
h
in
c
or
por
a
te
s
t
he
i
de
a
of
f
r
a
m
e
dupli
c
a
ti
on.
L
iu
e
t
al
.
[
13
]
ut
il
iz
e
th
e
c
onc
e
pt
s
of
ti
m
e
a
nd
f
r
e
que
nc
y
to
e
xpl
a
in
th
e
ir
r
e
s
e
a
r
c
h
on
f
r
a
m
e
dupl
ic
a
ti
on
a
nd
de
le
ti
on.
T
hi
s
is
s
e
e
n
in
F
ig
u
r
e
3.
T
he
pe
r
io
di
c
it
y
of
a
s
e
que
nc
e
is
m
e
a
s
ur
e
d
by
th
e
dom
a
in
c
ha
r
a
c
te
r
is
ti
c
s
of
ti
m
e
a
nd
f
r
e
que
nc
y.
A
t
hi
gh
f
r
e
que
nc
y
poi
nt
s
,
th
e
di
s
c
r
e
te
-
ti
m
e
F
our
ie
r
t
r
a
ns
f
or
m
(
D
T
F
T
)
is
c
a
lc
ul
a
te
d
u
s
in
g
m
e
tr
ic
s
s
uc
h
a
s
F
1
-
s
c
or
e
,
m
e
a
n
s
qua
r
e
e
r
r
or
(
M
S
E
)
,
a
c
c
ur
a
c
y,
a
nd
pr
e
di
c
ti
on
r
a
te
.
W
a
ng
e
t
al
.
[
14]
di
s
c
u
s
s
e
d
th
e
a
ppl
ic
a
ti
on
of
th
e
s
uppor
t
ve
c
to
r
m
a
c
hi
ne
(
S
V
M
)
m
a
c
hi
ne
le
a
r
ni
ng
te
c
hni
que
in
de
te
r
m
in
in
g
th
e
gr
e
y
v
a
lu
e
c
or
r
e
la
ti
on
c
oe
f
f
ic
ie
nt
(
C
oG
V
)
dur
in
g
th
e
ir
pr
e
s
e
nt
a
ti
on.
In
[
15]
,
[
16]
s
tu
di
e
d
f
r
a
m
e
in
s
e
r
ti
on
a
nd
de
le
ti
on
us
in
g
hue
-
s
a
tu
r
a
ti
on
-
va
lu
e
(
H
S
V
)
,
s
pe
e
de
d
up
r
obus
t
f
e
a
tu
r
e
s
(
S
U
R
F
)
,
a
nd
f
a
s
t
li
br
a
r
y
f
or
a
ppr
oxi
m
a
te
ne
a
r
e
s
t
n
e
ig
hbor
s
(
F
L
A
N
N
)
.
H
ow
e
ve
r
,
th
e
s
e
m
e
th
ods
onl
y
w
or
k
on
bl
in
d
f
or
e
ns
ic
vi
de
o.
L
ong
e
t
a
l.
[
17]
ut
il
is
e
d
a
c
onvolu
ti
on
a
l
ne
ur
a
l
n
e
twor
k
(
C
N
N
)
w
it
h
R
e
s
N
e
t
a
r
c
hi
te
c
tu
r
e
to
d
e
te
c
t
a
lt
e
r
e
d
f
r
a
m
e
s
i
n
v
id
e
o
s
a
s
s
ho
w
n
in
F
ig
ur
e
4
.
T
he
ir
a
ppr
oa
c
h
c
e
nt
r
e
d
on
in
s
tr
u
c
ti
ng
th
e
n
e
twor
k
to
e
f
f
e
c
ti
ve
ly
id
e
nt
if
y
oc
c
ur
r
e
n
c
e
s
of
f
r
a
m
e
in
s
e
r
ti
on,
d
e
le
ti
on,
a
nd
du
pl
ic
a
ti
on.
N
e
v
e
r
th
e
l
e
s
s
,
th
i
s
m
e
th
o
d
h
a
s
r
e
s
tr
ic
t
io
n
s
w
he
n
it
c
om
e
s
to
it
s
s
ui
ta
bi
li
ty
f
or
c
ont
i
nuou
s
vi
d
e
o
s
th
a
t
c
ons
is
t
of
e
xt
e
n
de
d
s
hot
f
r
a
m
e
s
.
I
n
or
de
r
to
ove
r
c
om
e
t
hi
s
c
ons
tr
a
in
t,
S
hi
e
t
al
.
[
18]
pr
opo
s
e
d
th
e
id
e
a
of
t
a
m
pe
r
in
g,
w
hi
c
h
r
e
f
e
r
s
t
o
th
e
a
c
t
of
dupl
i
c
a
ti
n
g
a
nd
i
ns
e
r
ti
ng
s
m
a
ll
por
ti
on
s
of
a
f
r
a
m
e
in
to
a
not
h
e
r
f
r
a
m
e
.
T
hi
s
a
ppr
o
a
c
h
ha
s
ga
r
n
e
r
e
d
s
i
gni
f
ic
a
nt
in
te
r
e
s
t
f
r
om
s
c
hol
a
r
s
a
s
s
h
ow
n
i
n
T
a
bl
e
1
.
A
s
ig
ni
f
ic
a
nt
o
bs
t
a
c
l
e
e
nc
ount
e
r
e
d
by
r
e
s
e
a
r
c
he
r
s
i
s
t
he
m
a
na
ge
m
e
nt
of
vi
de
o
f
il
e
s
of
c
o
ns
id
e
r
a
bl
e
s
iz
e
.
D
e
ng
e
t
al
.
[
19]
in
tr
od
uc
e
d
th
e
not
io
n
of
t
e
n
s
or
s
tr
uc
tu
r
e
a
s
a
s
o
lu
ti
on
to
th
i
s
pr
obl
e
m
.
T
e
n
s
or
s
tr
uc
tu
r
e
i
nv
ol
ve
s
t
h
e
u
s
e
of
da
ta
d
e
c
om
po
s
it
io
n a
nd
di
m
e
n
s
io
n r
e
d
uc
ti
o
n t
e
c
hni
q
ue
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
, V
ol
.
14
, N
o.
2
,
A
pr
il
2025
:
1471
-
1483
1474
F
ig
ur
e
3.
T
r
a
ns
f
or
m
in
g
vi
de
o i
nt
o f
r
a
m
e
s
F
ig
ur
e
4. I
nt
e
r
-
f
r
a
m
e
vi
de
o c
ount
e
r
f
e
it
r
e
f
le
c
ti
ng
or
ig
in
a
l
f
r
a
m
e
s
e
que
nc
e
T
a
bl
e
1. M
e
th
ods
f
or
de
te
c
ti
ng vide
o f
or
ge
r
ie
s
R
e
f
e
r
e
nc
e
T
ype
of
f
or
ge
r
y
U
s
e
d f
e
a
t
ur
e
L
i
m
i
t
a
t
i
on
[
10]
D
upl
i
c
a
t
i
on of
f
r
a
m
e
E
a
c
h
f
r
a
m
e
i
s
i
s
ol
a
t
e
d
a
nd
c
om
pa
r
e
d
u
s
i
ng
S
V
D
.
I
t
f
a
i
l
e
d
t
o
r
e
c
ogni
s
e
f
r
a
m
e
r
e
or
de
r
i
ng a
nd ot
he
r
m
odi
f
i
c
a
t
i
ons
.
[
11]
F
r
a
m
e
de
l
e
t
i
on
A
ppl
yi
ng
s
e
que
nc
i
ng
w
i
t
hi
n
f
r
a
m
e
s
t
o
m
a
ni
pul
a
t
e
t
he
dom
a
i
n a
nd c
r
e
a
t
e
f
or
ge
r
i
e
s
.
T
hi
s
m
e
t
hod
ha
s
a
f
r
e
que
nc
y
a
nd
t
i
m
e
l
i
m
i
t
.
[
12]
F
r
a
m
e
i
ns
e
r
t
i
on
a
nd
de
l
e
t
i
on
C
G
oV
s
A
ppl
i
c
a
bl
e
t
o f
i
xe
d da
t
a
s
e
t
s
[
13]
I
ns
e
r
t
,
de
l
e
t
e
,
dupl
i
c
a
t
e
f
r
a
m
e
s
I
ns
t
e
a
d
of
us
i
ng
c
or
r
e
l
a
t
i
on
c
oe
f
f
i
c
i
e
nt
s
,
c
or
r
e
l
a
t
i
on
quot
i
e
nt
s
a
r
e
e
m
pl
oye
d
t
o
m
e
a
s
ur
e
t
he
r
e
l
a
t
i
ons
hi
p be
t
w
e
e
n f
r
a
m
e
s
.
D
e
t
e
c
t
i
on
of
f
or
ge
r
y
i
s
a
c
hi
e
v
a
bl
e
,
ye
t
t
he
ot
he
r
t
w
o
m
e
t
hods
pr
ove
d
i
ne
f
f
e
c
t
i
ve
.
[
14]
F
r
a
m
e
dupl
i
c
a
t
i
on
T
he
S
U
R
F
t
e
c
hni
que
i
s
u
s
e
d
t
o
c
om
pa
r
e
t
he
c
ol
our
s
of
hi
s
t
ogr
a
m
s
.
R
e
s
t
r
i
c
t
e
d
t
o
c
a
pt
ur
i
ng
s
pe
c
i
f
i
c
c
om
pos
i
t
i
ons
w
i
t
hi
n t
he
f
r
a
m
e
s
[
15]
F
r
a
m
e
i
n
s
e
r
t
i
on
,
d
e
l
e
t
i
o
n
a
n
d
d
u
p
l
i
c
a
t
i
o
n
C
N
N
–
3D
t
o de
t
e
c
t
vi
de
o f
or
ge
r
y
L
oc
a
l
i
z
a
t
i
on i
s
f
a
i
l
e
d t
o i
nc
or
por
a
t
e
[
20]
T
a
m
pe
r
i
ng
M
ot
i
on r
e
s
i
dua
l
F
or
ge
r
y l
oc
a
l
i
z
a
t
i
on i
s
f
a
i
l
e
d
[
21]
D
oubl
e
c
om
pr
e
s
s
i
on
D
oubl
e
c
om
pr
e
s
s
i
on s
t
a
t
i
s
t
i
c
s
L
oc
a
l
i
z
a
t
i
on i
s
f
a
i
l
e
d
[
22]
U
ps
c
a
l
e
c
r
op
C
onf
or
m
s
t
o
t
he
i
nt
e
r
na
l
m
e
a
s
ur
e
m
e
nt
s
of
t
he
f
r
a
m
e
A
di
s
a
dva
nt
a
ge
of
t
hi
s
s
t
r
a
t
e
gy
i
s
t
ha
t
t
he
vi
de
o w
i
l
l
be
m
a
gni
f
i
e
d.
[
23]
S
p
a
t
i
o
t
e
m
po
r
a
l
f
o
r
g
e
r
y
M
ot
i
on ba
s
e
d S
V
M
T
hi
s
a
ppr
oa
c
h
ha
s
a
l
ow
e
r
a
c
c
ur
a
c
y.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
V
id
e
o f
or
ge
r
y
:
A
n e
x
te
ns
iv
e
analy
s
i
s
of
i
nt
e
r
-
and in
tr
a
-
fr
am
e
m
ani
pul
at
io
n al
ongs
id
e
…
(
Sum
ai
y
a Shaik
h
)
1475
I
n
2013,
B
ir
a
jd
a
r
a
nd
M
a
nka
r
[
24]
c
onduc
te
d
e
xt
e
ns
iv
e
r
e
s
e
a
r
c
h
on
s
e
ve
r
a
l
a
s
pe
c
ts
of
f
or
e
ns
ic
a
na
ly
s
is
p
e
r
ta
in
in
g
to
bot
h
vi
de
o
s
a
nd
phot
ogr
a
phs
.
T
hi
s
a
ut
hor
m
a
de
a
s
ig
ni
f
ic
a
nt
br
e
a
kt
hr
ough
by
s
uc
c
e
s
s
f
ul
ly
de
te
c
ti
ng
vi
de
o
c
a
m
e
r
a
s
th
a
t
a
r
e
li
nke
d
to
c
ount
e
r
f
e
it
in
g,
m
a
r
ki
ng
a
b
ig
a
dva
nc
e
m
e
nt
in
th
e
in
dus
tr
y.
T
he
r
e
s
e
a
r
c
he
r
'
s
in
qui
r
y
s
he
d
li
ght
on
th
e
noi
s
e
p
a
tt
e
r
ns
a
nd
pot
e
nt
ia
l
c
om
pr
e
s
s
io
n
te
c
hni
que
s
e
m
pl
oye
d
in
vi
de
o
s
.
H
e
pr
e
s
e
nt
e
d a
m
e
th
od
th
a
t
ut
il
is
e
s
th
e
di
s
c
r
e
te
f
our
ie
r
tr
a
ns
f
or
m
(
D
F
T
)
to
de
te
c
t
a
lt
e
r
e
d
s
e
c
ti
ons
in
m
ovi
e
s
,
s
ui
ta
bl
e
f
or
bot
h
lo
w
-
qua
li
ty
a
nd
hi
gh
-
qua
li
ty
vi
de
o
s
c
e
na
r
io
s
.
N
e
ve
r
th
e
le
s
s
,
c
ha
ll
e
nge
s
a
r
os
e
w
he
n
de
a
li
ng
w
it
h
lo
w
-
qua
li
ty
f
il
m
s
,
pr
om
pt
in
g
a
m
or
e
th
or
ough
in
ve
s
ti
ga
ti
on
in
to
di
f
f
e
r
e
nt
vi
de
o
c
a
m
e
r
a
s
a
nd
th
e
d
e
te
c
ti
on
of
m
odi
f
ie
d
c
ont
e
nt
.
I
n
a
la
t
e
r
in
ve
s
ti
ga
ti
on
[
23]
,
s
c
ie
nt
is
t
s
im
pl
e
m
e
nt
e
d
a
c
om
put
a
ti
ona
l
m
e
th
od
known
a
s
phot
o
r
e
s
pon
s
e
non
-
uni
f
or
m
it
y
(
P
R
N
U
)
,
ut
il
is
in
g
a
dva
nc
e
d
3D
pa
tc
h
-
m
a
tc
h
a
lg
or
it
hm
s
to
id
e
nt
if
y
m
a
ni
pul
a
te
d
c
ont
e
nt
in
f
il
m
s
.
T
hi
s
m
e
th
od
a
ls
o
in
c
lu
de
s
f
e
a
tu
r
e
e
xt
r
a
c
ti
on
to
im
pr
ove
th
e
pr
e
c
is
io
n of
f
or
ge
r
y de
te
c
ti
on.
T
he
pr
ogr
e
s
s
io
n
of
r
e
s
e
a
r
c
h
in
th
is
f
ie
ld
th
r
oughout
th
e
ye
a
r
s
hi
ghl
ig
ht
s
a
n
in
c
r
e
a
s
in
g
f
a
s
c
in
a
ti
on
a
nd
a
c
knowle
dgm
e
nt
of
it
s
s
ig
ni
f
ic
a
nc
e
.
I
n
th
e
be
gi
nni
ng,
th
e
r
e
w
a
s
li
m
it
e
d
s
tu
dy
e
f
f
or
t
in
th
is
f
ie
ld
.
H
ow
e
ve
r
,
a
s
th
e
f
r
e
que
nc
y
a
nd
c
om
pl
e
xi
ty
of
a
tt
a
c
ks
gr
e
w
,
r
e
s
e
a
r
c
he
r
s
s
ta
r
te
d
to
c
onc
e
nt
r
a
te
m
or
e
a
ggr
e
s
s
iv
e
ly
on
vi
de
o a
nd
pi
c
tu
r
e
f
or
e
ns
ic
s
.
T
he
in
c
r
e
a
s
e
in
r
e
s
e
a
r
c
h
a
c
ti
vi
ty
ha
s
im
por
ta
nt
c
on
s
e
que
nc
e
s
f
or
la
w
e
nf
or
c
e
m
e
nt
,
gove
r
nm
e
nt
a
ge
nc
ie
s
,
a
nd
c
ybe
r
f
or
e
ns
ic
s
pr
a
c
ti
ti
one
r
s
.
T
he
br
e
a
kt
hr
oughs
in
m
e
th
odol
ogy
a
nd
a
ppl
ic
a
ti
ons
th
a
t
r
e
s
ul
t
f
r
om
th
is
r
e
s
e
a
r
c
h
c
a
n
s
ubs
ta
nt
i
a
ll
y
he
lp
th
e
ir
in
ve
s
ti
ga
ti
on
e
f
f
or
ts
.
T
he
r
e
i
s
s
ig
ni
f
ic
a
nt
r
is
e
in
r
e
s
e
a
r
c
h
out
put
,
w
it
h
m
o
r
e
th
a
n
15
%
of
r
e
s
e
a
r
c
h
publ
ic
a
ti
ons
be
in
g
publ
is
he
d
w
it
hi
n
a
th
r
e
e
-
to
-
f
our
-
m
ont
h
pe
r
io
d
in
2023
[
24]
.
T
he
tr
e
nd
c
a
n
be
s
h
ow
n
in
F
ig
ur
e
5,
w
hi
c
h
di
s
pl
a
y
s
a
pi
e
c
ha
r
t
r
e
pr
e
s
e
nt
in
g r
e
s
e
a
r
c
h pa
tt
e
r
ns
s
pa
nni
ng f
r
om
1990 to 2023.
F
ig
ur
e
5.
R
e
s
e
a
r
c
h
pi
e
c
h
a
r
t
c
ove
r
in
g t
he
ye
a
r
s
1990
-
2023
U
s
in
g
th
e
da
ta
f
r
om
F
ig
ur
e
5,
r
e
s
e
a
r
c
he
r
s
c
la
s
s
if
ie
d
s
e
ve
r
a
l
ty
pe
s
of
di
gi
ta
l
f
or
e
ns
ic
s
,
w
it
h
a
f
oc
us
on
f
il
m
s
a
nd
ne
twor
ks
.
M
or
e
pr
e
c
is
e
ly
,
in
ne
twor
k
f
or
e
ns
i
c
s
,
th
e
r
e
is
a
s
tr
ong
e
m
pha
s
is
on
de
te
c
ti
ng
f
or
ge
r
ie
s
,
s
om
e
ti
m
e
s
known
a
s
in
c
ur
s
io
n
[
25]
.
I
n
th
e
c
ont
e
xt
o
f
vi
de
os
,
ne
twor
k
in
t
r
us
io
n
r
e
f
e
r
s
to
th
e
a
c
t
o
f
de
te
c
ti
ng
a
nd
r
e
c
ogni
s
in
g
c
ha
nge
d
or
f
r
a
udul
e
nt
vi
de
os
tr
a
ns
m
it
te
d
vi
a
s
e
ve
r
a
l
c
ha
nn
e
ls
.
T
hi
s
ne
w
f
ie
ld
of
s
tu
dy
is
no
w
le
a
di
ng
th
e
w
a
y
in
d
e
te
c
ti
ng
a
nd
pr
e
ve
nt
in
g
b
r
e
a
c
he
s
in
ne
twor
k
s
e
c
ur
it
y,
pa
r
ti
c
ul
a
r
ly
th
e
pr
ol
if
e
r
a
ti
on of
c
ount
e
r
f
e
it
vi
de
o m
e
di
a
.
W
e
f
ound
th
a
t
in
te
r
-
f
r
a
m
e
m
a
ni
pul
a
ti
ons
,
s
uc
h
a
s
f
r
a
m
e
du
pl
ic
a
ti
on
or
de
le
ti
on,
c
or
r
e
la
te
w
it
h
not
ic
e
a
bl
e
te
m
por
a
l
in
c
ons
is
te
nc
ie
s
,
w
hi
le
in
tr
a
-
f
r
a
m
e
m
a
ni
pul
a
ti
ons
,
li
ke
s
pl
ic
in
g
or
a
lt
e
r
in
g
ob
je
c
ts
w
it
hi
n
f
r
a
m
e
s
,
c
or
r
e
la
te
w
it
h
s
pa
ti
a
l
in
c
ons
is
t
e
nc
ie
s
in
th
e
vi
de
o.
T
he
pr
opos
e
d
m
e
th
od
in
th
is
s
tu
dy
te
nde
d
to
ha
v
e
a
n
in
or
di
na
te
ly
hi
ghe
r
pr
opo
r
ti
on
o
f
de
te
c
ti
on
a
c
c
ur
a
c
y
f
or
b
ot
h
in
te
r
-
f
r
a
m
e
a
nd
in
tr
a
-
f
r
a
m
e
m
a
ni
pul
a
ti
ons
c
om
pa
r
e
d t
o e
xi
s
ti
ng s
ta
te
-
of
-
th
e
a
r
t
te
c
hni
que
s
.
2.1.
N
e
t
w
or
k
f
or
e
n
s
ic
s
N
e
twor
k
f
or
e
ns
ic
s
,
a
s
ubf
ie
ld
of
di
gi
ta
l
f
or
e
ns
ic
s
,
is
c
r
it
ic
a
l
f
or
in
ve
s
ti
ga
ti
ng
c
r
im
e
s
in
vol
vi
ng
vi
de
o
tr
a
ns
f
e
r
s
ove
r
c
om
put
e
r
ne
twor
ks
.
D
ig
it
a
l
da
ta
tr
a
ns
por
te
d
ov
e
r
ne
twor
ks
is
th
or
oughly
in
ve
s
ti
ga
te
d
us
in
g
ne
twor
k
f
or
e
ns
ic
to
ol
s
(
N
F
T
s
)
a
nd
ne
twor
k
f
or
e
ns
ic
pr
oc
e
s
s
e
s
(
N
F
P
s
)
to
di
f
f
e
r
e
nt
ia
te
be
twe
e
n
r
e
gul
a
r
a
nd
a
be
r
r
a
nt
tr
a
f
f
ic
pa
tt
e
r
ns
.
T
hi
s
s
tu
dy
c
om
pr
is
e
s
id
e
nt
if
yi
ng
in
c
id
e
nc
e
s
a
nd
e
x
a
m
in
in
g
r
e
a
c
ti
ons
,
w
hi
c
h
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
, V
ol
.
14
, N
o.
2
,
A
pr
il
2025
:
1471
-
1483
1476
pr
ovi
de
s
va
lu
a
bl
e
e
vi
de
nc
e
f
or
le
ga
l
pr
oc
e
e
di
ngs
[
26]
.
D
ig
it
a
l
m
e
di
a
tr
a
ns
f
e
r
r
e
d
ove
r
ne
twor
ks
ge
ne
r
a
te
s
r
e
s
id
ua
l
da
ta
w
it
hi
n
th
e
tr
a
ns
m
is
s
io
n
c
ha
nne
ls
,
w
hi
c
h
c
a
n
be
us
e
d
in
in
ve
s
ti
ga
ti
ons
.
T
he
da
ta
s
e
t
us
e
d
f
or
in
tr
us
io
n de
te
c
ti
on c
ons
is
ts
of
di
gi
ta
l
vi
de
o t
r
a
ns
m
is
s
io
ns
, w
h
e
r
e
i
nt
r
us
io
n de
te
c
ti
on mode
ls
a
r
e
i
m
pl
e
m
e
nt
e
d.
A
s
de
pi
c
te
d
in
F
ig
ur
e
6,
th
is
pr
oc
e
s
s
in
vol
ve
s
th
e
c
r
e
a
ti
on
a
nd
a
na
ly
s
is
of
da
t
a
,
f
ol
lo
w
e
d
by
th
e
de
te
c
ti
on
of
pot
e
nt
ia
ll
y
m
a
li
c
io
us
be
ha
vi
our
s
a
nd
th
e
c
r
e
a
ti
on
of
lo
g
f
il
e
s
f
or
ne
twor
k
f
or
e
ns
ic
a
na
ly
s
is
.
F
ol
lo
w
in
g
th
a
t,
th
e
r
e
c
ove
r
y
pr
oc
e
s
s
e
nc
om
pa
s
s
e
s
f
our
di
s
c
r
e
te
s
ta
ge
s
,
s
ta
r
ti
n
g
w
it
h
da
ta
ga
th
e
r
in
g
a
nd
c
ul
m
in
a
ti
ng
in
th
e
pr
e
s
e
nt
a
ti
on of
f
in
di
ngs
a
s
e
vi
de
nc
e
i
n c
our
t
[
27]
.
F
ig
ur
e
6
de
pi
c
ts
th
e
s
te
p
-
by
-
s
te
p
pr
oc
e
s
s
of
ga
th
e
r
in
g
da
ta
a
n
d
pr
oduc
in
g
a
c
om
pl
e
te
r
e
por
t
to
be
f
or
w
a
r
de
d t
o t
he
B
ur
e
a
u t
e
a
m
. T
he
s
e
s
ta
ge
s
a
r
e
di
vi
de
d i
nt
o va
r
io
us
pa
r
ts
, i
nc
lu
di
ng da
ta
c
ol
le
c
ti
on, a
na
ly
s
is
,
a
nd
pr
e
s
e
nt
a
ti
on.
V
a
r
io
us
e
xp
e
r
t
to
ol
s
a
r
e
e
m
pl
oye
d
dur
in
g
th
is
pr
oc
e
dur
e
to
a
c
c
ur
a
te
ly
a
s
s
e
s
s
th
e
d
a
ta
a
nd
pr
oduc
e
pr
e
c
is
e
r
e
s
ul
ts
.
T
he
s
e
to
ol
s
a
r
e
s
pe
c
if
ic
a
ll
y
de
v
e
lo
pe
d
to
a
c
c
e
le
r
a
te
th
e
te
s
ti
ng
a
nd
pr
oc
e
s
s
in
g
of
di
gi
ta
l
e
vi
de
nc
e
,
e
n
s
ur
in
g
th
a
t
a
ll
c
r
it
ic
a
l
in
f
or
m
a
ti
on
is
c
or
r
e
c
tl
y
r
e
c
or
de
d
a
nd
do
c
um
e
nt
e
d
a
s
s
how
n
in
T
a
bl
e
2.
U
s
in
g
th
e
s
e
te
c
hnol
ogi
e
s
,
f
or
e
ns
ic
a
na
ly
s
ts
m
a
y
s
uc
c
e
s
s
f
ul
ly
e
xt
r
a
c
t
vi
ta
l
in
f
or
m
a
ti
on
f
r
om
da
ta
a
nd
pr
e
s
e
nt
t
he
ir
f
in
di
ngs
i
n a
s
tr
uc
tu
r
e
d a
nd unde
r
s
ta
nda
bl
e
s
ty
le
f
or
f
ut
ur
e
i
nqui
r
y a
nd l
e
ga
l
pr
oc
e
s
s
e
s
.
F
ig
ur
e
6. F
or
e
ns
ic
p
r
oc
e
dur
e
T
a
bl
e
2. D
a
ta
s
e
t
d
e
s
c
r
ip
ti
ons
D
a
t
a
s
e
t
F
or
ge
r
y
t
ype
F
e
a
t
ur
e
/
s
our
c
e
S
U
L
F
A
F
r
a
m
e
dupl
i
c
a
t
i
on
,
f
or
ge
r
y
G
L
C
M
T
R
A
C
E
D
upl
i
c
a
t
i
on
H
a
r
a
l
i
c
k P
R
G
a
nd O
F
G
V
T
L
M
ot
i
on
C
N
N
S
Y
S
U
-
O
B
J
F
O
R
G
S
pa
t
i
a
l
/
t
e
m
por
a
l
T
P
F
C
N
T
H
U
F
r
a
m
e
dupl
i
c
a
t
i
on
Y
out
ube
C
A
S
I
A
v2.0
C
P
a
nd C
M
O
F
G
C
V
A
P
B
a
c
kgr
ound
N
i
m
bl
e
c
ha
l
l
e
nge
I
M
D
B
CM
G
L
C
M
F
a
c
e
F
or
e
ns
i
c
s
++
D
e
e
pf
a
ke
s
N
e
ur
a
l
t
e
xt
ur
e
s
D
F
D
C
D
e
e
pf
a
ke
D
E
R
F
c
ol
l
e
c
t
i
ons
R
E
W
I
N
D
D
W
T
, S
I
F
T
C
N
N
L
A
S
I
E
S
T
A
F
r
a
m
e
dupl
i
c
a
t
i
on
G
L
C
M
T
R
E
C
S
w
a
ppi
ng
-
f
r
a
m
e
s
G
L
C
M
B
O
S
S
S
t
e
ga
na
l
ys
i
s
C
N
N
a
nd S
I
F
T
G
R
I
P
C
opy,
m
ove
, s
l
i
c
i
ng
S
I
F
T
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
V
id
e
o f
or
ge
r
y
:
A
n e
x
te
ns
iv
e
analy
s
i
s
of
i
nt
e
r
-
and in
tr
a
-
fr
am
e
m
ani
pul
at
io
n al
ongs
id
e
…
(
Sum
ai
y
a Shaik
h
)
1477
R
e
c
e
nt
s
ur
ve
y
s
r
e
ve
a
l
th
a
t
da
ta
s
e
t
s
us
e
d
f
or
vi
de
o
f
r
a
ud
d
e
te
c
ti
on
in
c
lu
de
bot
h
a
ut
he
nt
ic
a
nd
m
odi
f
ie
d
r
e
c
or
di
ngs
s
our
c
e
d
f
r
om
va
r
io
us
a
r
c
hi
ve
s
,
in
c
lu
di
ng
th
os
e
ge
ne
r
a
te
d
by
th
e
U
ni
ve
r
s
it
y
of
S
ur
r
e
y
a
nd
ot
he
r
in
te
r
ne
t
pl
a
tf
or
m
s
.
T
hr
e
e
e
xa
m
pl
e
s
of
th
e
s
e
da
ta
s
e
ts
a
r
e
s
ur
r
e
y
uni
ve
r
s
it
y
li
br
a
r
y
f
or
f
or
e
ns
ic
a
na
ly
s
is
(
S
U
L
F
A
)
,
r
e
ve
r
s
e
e
ngi
ne
e
r
in
g
of
a
udi
o
-
vi
s
ua
l
c
ont
e
nt
da
ta
(
R
E
W
I
N
D
)
,
a
nd
gl
oba
l
r
oa
ds
in
ve
nt
or
y
pr
oj
e
c
t
(
G
R
I
P
)
.
T
he
da
ta
s
e
t
s
c
a
n
be
a
c
c
e
s
s
e
d
in
bot
h
m
ot
io
n
J
P
E
G
(
M
J
P
E
G
)
a
nd
H
.264
c
ode
c
s
.
T
ypi
c
a
l
s
our
c
e
s
f
or
th
e
s
e
da
ta
s
e
ts
c
ons
i
s
t
of
Y
ouT
ube
vi
de
os
a
nd
s
e
c
ur
it
y
c
a
m
e
r
a
r
e
c
or
di
ngs
,
w
it
h
s
a
m
pl
e
s
iz
e
s
va
r
yi
ng f
r
om
119
t
o 10,000 c
li
ps
.
E
a
c
h c
li
p ha
s
a
r
unt
im
e
of
l
e
s
s
t
ha
n
10
s
e
c
onds
[
28]
. T
he
da
ta
s
e
ts
c
om
pr
is
e
s
e
ve
r
a
l
f
or
m
s
of
m
a
ni
pul
a
ti
ons
, s
uc
h a
s
c
opy
-
m
ove
, s
pl
ic
in
g, i
n
te
r
-
f
r
a
m
e
, a
nd i
nt
r
a
-
f
r
a
m
e
f
or
ge
r
ie
s
.
T
o
ge
t
in
s
ig
ht
in
to
th
e
ut
il
is
a
ti
on
of
th
e
s
e
da
ta
s
e
ts
ov
e
r
ti
m
e
,
a
c
om
pa
r
a
ti
ve
gr
a
ph
is
ge
ne
r
a
te
d
to
il
lu
s
tr
a
te
th
e
oc
c
ur
r
e
nc
e
r
a
te
of
v
a
r
io
us
da
ta
s
e
t
c
a
te
gor
ie
s
th
r
oughout
di
f
f
e
r
e
nt
ye
a
r
s
[
29]
.
T
hi
s
vi
s
ua
l
r
e
pr
e
s
e
nt
a
ti
on
e
nha
nc
e
s
th
e
a
bi
li
ty
to
r
e
c
ogni
s
e
tr
e
nd
s
a
nd
pa
t
te
r
ns
in
th
e
us
e
of
da
ta
s
e
t
s
,
pa
r
ti
c
ul
a
r
ly
w
he
n
a
na
ly
s
in
g
va
r
io
us
f
or
m
s
of
f
or
ge
r
ie
s
.
T
hr
ough
th
e
pr
oc
e
s
s
of
vi
s
ua
li
s
in
g
th
is
da
t
a
,
r
e
s
e
a
r
c
he
r
s
a
r
e
a
bl
e
to
pe
r
c
e
iv
e
c
ha
nge
s
in
th
e
e
m
pha
s
i
s
of
s
tu
dy
a
nd
a
s
c
e
r
ta
in
w
hi
c
h
a
r
e
a
s
of
in
ve
s
ti
ga
ti
on
s
houl
d
be
gi
v
e
n
hi
ghe
r
pr
io
r
it
y, t
a
ki
ng i
nt
o a
c
c
ount
ne
w
t
r
e
nds
a
nd obs
ta
c
le
s
i
n t
he
f
ie
ld
of
vi
de
o f
or
ge
r
y de
te
c
ti
on
[
30]
.
2.2.
V
id
e
o
f
or
ge
r
y d
e
t
e
c
t
io
n
m
e
t
h
od
s
s
yn
op
s
is
A
c
om
pr
e
he
ns
iv
e
ov
e
r
vi
e
w
of
m
e
th
ods
us
e
d
to
id
e
nt
if
y
vi
de
o
f
or
ge
r
ie
s
is
in
c
lu
de
d
in
T
a
bl
e
3,
c
a
te
gor
iz
in
g
s
tr
a
te
gi
e
s
in
to
in
tr
a
-
f
r
a
m
e
a
nd
in
te
r
-
f
r
a
m
e
m
e
th
o
dol
ogi
e
s
.
T
a
bl
e
3
hi
ghl
ig
ht
s
th
e
s
tr
e
ngt
hs
a
nd
w
e
a
kne
s
s
e
s
of
e
a
c
h
a
ppr
oa
c
h,
il
lu
s
tr
a
ti
ng
th
e
c
ur
r
e
nt
li
m
it
a
ti
ons
in
de
te
c
ti
on
a
c
c
ur
a
c
y
a
nd
c
om
put
a
ti
ona
l
e
f
f
ic
ie
nc
y
[
31]
.
I
t
pr
ovi
de
s
a
c
le
a
r
c
om
pa
r
is
on
of
e
xi
s
ti
ng
te
c
hni
que
s
,
e
m
pha
s
iz
in
g
th
e
n
e
e
d
f
or
m
or
e
r
obus
t
a
nd
e
f
f
ic
ie
nt
s
ol
ut
io
ns
.
T
hi
s
c
a
te
gor
iz
a
ti
on
a
id
s
in
unde
r
s
ta
ndi
ng
th
e
ga
ps
in
c
ur
r
e
nt
r
e
s
e
a
r
c
h
a
nd
gui
de
s
f
ut
ur
e
de
ve
lo
pm
e
nt
s
i
n vi
de
o f
or
ge
r
y de
te
c
ti
on.
T
a
bl
e
3. I
nt
r
a
a
nd i
nt
e
r
-
f
r
a
m
e
t
e
c
hni
que
s
f
or
vi
de
o f
or
ge
r
y de
te
c
ti
on
R
e
f
A
ppr
oa
c
h
T
e
c
hni
que
A
l
gor
i
t
hm
D
a
t
a
s
e
t
L
i
m
i
t
a
t
i
ons
[
32]
P
a
s
s
i
ve
a
ppr
oa
c
h
I
nt
e
r
f
r
a
m
e
F
or
ge
r
y l
oc
a
l
i
z
a
t
i
on
I
nt
e
r
ne
t
s
t
r
e
a
m
e
d
vi
de
o
D
i
f
f
i
c
ul
t
t
o t
r
a
c
e
[
33]
P
a
s
s
i
ve
a
ppr
oa
c
h
I
nt
e
r
-
f
r
a
m
e
F
or
ge
r
y de
t
e
c
t
i
on
I
nt
e
r
ne
t
s
t
r
e
a
m
e
d
vi
de
o
V
ul
ne
r
a
bl
e
t
o a
t
t
a
c
ks
[
34]
S
e
a
r
c
h
ba
s
e
d
a
ppr
oa
c
h
I
nt
e
r
-
f
r
a
m
e
B
l
oc
k
-
ba
s
e
d
a
l
gor
i
t
hm
C
us
t
om
vi
de
os
D
i
f
f
i
c
ul
t
t
o de
t
e
c
t
ne
a
r
dupl
i
c
a
t
e
a
r
e
a
s
[
35]
A
c
t
i
ve
a
nd
pa
s
s
i
ve
s
e
a
r
c
h
I
nt
e
r
-
f
r
a
m
e
F
a
s
t
r
ul
e
i
de
n
t
i
f
i
c
a
t
i
on
a
l
g
or
i
t
hm
L
i
ve
vi
de
os
t
a
ke
n
f
r
om
c
a
m
e
r
a
T
o
e
nha
nc
e
,
f
ur
t
he
r
i
ns
t
a
nc
e
s
of
f
or
ge
r
i
e
s
s
houl
d be
i
nc
l
ude
d
.
[
36]
C
opy, m
ove
, f
or
ge
r
y
I
nt
e
r
-
f
r
a
m
e
F
or
ge
r
y de
t
e
c
t
i
on a
nd
l
oc
a
l
i
z
a
t
i
on
R
E
W
I
N
D
T
i
m
e
c
ons
um
i
ng
[
37]
T
he
a
c
t
i
ve
a
nd t
he
pa
s
s
i
ve
a
ppr
oa
c
h
I
nt
e
r
-
f
r
a
m
e
F
or
ge
r
y de
t
e
c
t
i
on
a
l
gor
i
t
hm
R
E
W
I
N
D
U
nf
i
t
f
or
f
a
s
t
-
pa
c
e
d f
i
l
m
s
[
38]
N
or
m
a
l
i
z
e
d
c
r
os
s
-
c
or
r
e
l
a
t
i
on
I
nt
e
r
-
f
r
a
m
e
V
i
de
o f
or
ge
r
y
de
t
e
c
t
i
on
R
E
W
I
N
D
H
i
gh F
P
R
[
39]
B
ot
t
om
-
up a
ppr
oa
c
h
I
nt
r
a
-
f
r
a
m
e
E
xpe
c
t
a
t
i
on
-
M
a
xi
m
i
z
a
t
i
on
C
us
t
om
da
t
a
s
e
t
W
or
ks
f
or
onl
y
f
i
ne
qua
l
i
t
y
s
e
que
nc
e
s
[
40]
N
on
-
s
ubs
a
m
pl
e
d
c
ont
our
l
e
t
(
N
S
C
T
)
I
nt
r
a
-
f
r
a
m
e
F
e
a
t
ur
e
s
e
l
e
c
t
i
on
a
l
gor
i
t
hm
D
a
t
a
s
e
t
f
r
om
m
i
ne
.t
ku.e
du
R
e
l
i
e
s
on t
r
a
i
ni
ng s
a
m
pl
e
s
[
41]
D
i
gi
t
a
l
f
or
e
ns
i
c
s
I
nt
r
a
-
f
r
a
m
e
V
i
de
o t
a
m
pe
r
i
ng
de
t
e
c
t
i
on
M
P
E
G
-
2
87%
a
c
c
ur
a
c
y.
T
o
i
m
pr
ove
,
i
nve
s
t
i
ga
t
e
B
-
f
r
a
m
e
-
P
-
f
r
a
m
e
M
C
E
A
i
m
pa
c
t
.
[
42]
H
M
R
F
I
nt
r
a
-
f
r
a
m
e
s
t
a
t
e
of
t
he
a
r
t
de
t
e
c
t
i
on a
l
gor
i
t
hm
s
.
D
e
r
f
’
s
a
nd Y
U
V
A
c
c
ur
a
c
y
88.95%
a
nd
t
o
be
i
m
pr
ove
d w
i
t
h l
oc
a
l
i
z
a
t
i
on
[
43]
D
i
gi
t
a
l
f
or
e
ns
i
c
s
I
nt
r
a
-
f
r
a
m
e
A
ut
om
a
t
on a
l
gor
i
t
hm
K
T
H
F
or
ge
r
y
l
oc
a
l
i
z
a
t
i
on
i
s
ye
t
t
o
be
done
.
[
44]
B
l
oc
k
-
w
i
s
e
br
i
ght
ne
s
s
va
r
i
a
nc
e
de
s
c
r
i
pt
or
I
nt
e
r
-
f
r
a
m
e
B
l
o
c
k
-
w
i
s
e
de
s
c
r
i
p
t
o
r
-
b
a
s
e
d
a
l
g
o
r
i
t
h
m
S
Y
S
U
-
O
B
J
F
O
R
G
A
c
c
ur
a
c
y
83.37%
,
be
t
t
e
r
f
or
doubl
e
-
c
om
pr
e
s
s
e
d s
a
m
pl
e
s
T
a
bl
e
4
pr
e
s
e
nt
s
a
c
om
pr
e
he
n
s
iv
e
s
um
m
a
r
y
of
vi
de
o
f
r
a
ud
de
te
c
ti
on
s
tr
a
te
gi
e
s
,
f
oc
us
in
g
pa
r
ti
c
ul
a
r
ly
on
a
ppr
oa
c
he
s
r
e
la
te
d
to
de
e
pf
a
k
e
a
nd
pi
xe
l
m
ot
io
n
de
te
c
ti
on.
I
t
hi
ghl
ig
ht
s
th
e
e
f
f
e
c
ti
ve
ne
s
s
of
va
r
io
us
m
e
th
ods
in
id
e
nt
if
yi
ng
s
ynt
he
ti
c
m
e
di
a
a
nd
de
te
c
ti
ng
in
c
ons
i
s
te
nc
ie
s
in
pi
xe
l
m
ove
m
e
nt
s
.
T
he
ta
bl
e
a
ls
o
out
li
ne
s
th
e
s
tr
e
ngt
hs
a
nd
w
e
a
kne
s
s
e
s
of
th
e
s
e
a
ppr
oa
c
he
s
,
pr
ovi
di
ng
a
c
le
a
r
c
om
pa
r
is
on
of
th
e
ir
c
a
pa
bi
li
ti
e
s
.
T
hi
s
s
um
m
a
r
y
e
m
pha
s
iz
e
s
th
e
im
por
ta
nc
e
of
a
dva
n
c
e
d
te
c
hni
que
s
in
c
om
ba
ti
ng s
ophi
s
ti
c
a
t
e
d
vi
de
o
f
or
ge
r
ie
s
li
ke
de
e
pf
a
ke
s
.
O
ur
s
tu
dy
s
ugge
s
t
s
th
a
t
hi
ghe
r
de
te
c
ti
on
a
c
c
ur
a
c
y
is
not
a
s
s
oc
ia
te
d
w
it
h
poor
pe
r
f
or
m
a
nc
e
in
c
om
put
a
ti
ona
l
e
f
f
ic
ie
nc
y.
T
h
e
pr
opos
e
d
m
e
th
od
m
a
y
be
ne
f
it
f
r
om
a
dva
nc
e
d
s
p
a
ti
a
l
a
nd
te
m
por
a
l
a
na
ly
s
i
s
w
it
hout
a
dve
r
s
e
ly
im
pa
c
ti
ng
pr
oc
e
s
s
in
g
s
pe
e
d.
T
he
e
xt
e
n
s
iv
e
a
na
ly
s
is
pr
ovi
de
d
in
our
s
tu
dy,
c
ove
r
in
g
bot
h
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
, V
ol
.
14
, N
o.
2
,
A
pr
il
2025
:
1471
-
1483
1478
in
te
r
-
f
r
a
m
e
a
nd
in
tr
a
-
f
r
a
m
e
m
a
ni
pul
a
ti
ons
,
of
f
e
r
a
m
or
e
hol
is
ti
c
s
ol
ut
io
n
c
om
pa
r
e
d
to
s
pe
c
ia
li
z
e
d
m
e
th
ods
th
a
t
f
oc
us
on
onl
y
one
ty
pe
of
m
a
ni
pul
a
ti
on.
T
hi
s
c
om
pr
e
he
ns
iv
e
a
ppr
oa
c
h
r
e
duc
e
s
th
e
ne
e
d
f
or
m
ul
ti
pl
e
to
ol
s
a
nd
s
tr
e
a
m
li
ne
th
e
f
or
ge
r
y
de
te
c
ti
on
pr
oc
e
s
s
.
O
ur
s
tu
dy
hi
ghl
ig
ht
s
th
a
t
it
is
pos
s
ib
le
to
a
c
hi
e
ve
hi
gh
de
te
c
ti
on
a
c
c
ur
a
c
y
w
it
hout
s
a
c
r
if
ic
in
g
c
om
put
a
ti
ona
l
e
f
f
ic
ie
nc
y.
T
he
pr
opos
e
d
m
e
th
od’
s
a
dva
nc
e
d
a
na
ly
ti
c
a
l
c
a
pa
bi
li
ti
e
s
a
nd
lo
w
e
r
r
or
r
a
te
s
m
a
ke
it
a
s
upe
r
io
r
c
hoi
c
e
f
or
v
id
e
o
f
or
ge
r
y
de
te
c
ti
on,
a
ddr
e
s
s
in
g
li
m
it
a
ti
ons
obs
e
r
ve
d i
n pr
e
vi
ous
s
tu
di
e
s
.
T
a
bl
e
4. S
um
m
a
r
y of
v
id
e
o f
or
ge
r
y de
te
c
ti
on
te
c
hni
que
s
R
e
f
T
e
c
hni
que
F
e
a
t
ur
e
s
s
e
l
e
c
t
e
d
D
a
t
a
s
e
t
L
i
m
i
t
a
t
i
ons
[
45]
D
e
e
p f
a
ke
E
ye
bl
i
nki
ng
C
us
t
om
da
t
a
s
e
t
R
e
qui
r
e
s
f
ur
t
he
r
a
s
s
e
s
s
m
e
nt
us
i
ng
a
ddi
t
i
ona
l
vi
de
o s
a
m
pl
e
s
[
28]
D
e
e
p f
a
ke
H
e
a
d
pos
e
U
A
D
F
V
a
nd D
A
R
P
A
GAN
L
a
c
ks
pr
of
i
c
i
e
nc
y i
n i
de
nt
i
f
yi
ng puppe
t
-
m
a
s
t
e
r
a
nd l
i
p
-
s
ync
c
ount
e
r
f
e
i
t
s
.
[
46]
D
e
e
p f
a
ke
C
ol
or
a
r
t
i
f
a
c
t
s
L
S
U
N
a
nd I
m
a
ge
N
e
t
L
oc
a
l
i
z
a
t
i
on i
s
not
ye
t
e
f
f
e
c
t
i
ve
l
y done
[
47]
D
e
e
p f
a
ke
C
l
a
s
s
i
f
i
c
a
t
i
on
S
e
l
f
, F
a
c
e
F
or
e
ns
i
c
s
S
uf
f
e
r
s
f
r
om
ove
r
f
i
t
t
i
ng p
r
obl
e
m
[
48]
P
i
xe
l
m
ot
i
on
de
t
e
c
t
i
on
V
e
l
oc
i
t
y
f
i
e
l
d
c
ons
i
s
t
e
nc
y
T
R
E
C
V
I
D
U
na
bl
e
t
o de
t
e
c
t
a
l
t
e
r
e
d a
r
e
a
s
[
49]
P
i
xe
l
m
ot
i
on
de
t
e
c
t
i
on
O
pt
i
c
a
l
f
l
ow
T
R
E
C
V
I
D
C
om
put
a
t
i
on i
s
c
os
t
l
y
[
50]
P
i
xe
l
m
ot
i
on
de
t
e
c
t
i
on
M
ot
i
on
ve
c
t
or
pyr
a
m
i
d a
nd
va
r
i
a
t
i
on f
a
c
t
or
T
R
E
C
V
I
D
O
nl
y f
or
s
t
a
t
i
c
ba
c
kgr
ound i
t
i
s
s
ui
t
a
bl
e
[
51]
P
i
xe
l
m
ot
i
on
de
t
e
c
t
i
on
C
oa
r
s
e
t
o f
i
ne
O
pt
i
c
a
l
F
l
ow
V
T
L
, S
U
L
F
A
, D
E
R
F
T
he
pr
obl
e
m
of
m
i
s
de
t
e
c
t
i
on a
r
i
s
e
s
due
t
o
i
t
s
s
us
c
e
pt
i
bi
l
i
t
y t
o i
m
pr
e
c
i
s
e
de
t
e
c
t
i
on.
[
52]
K
e
yf
r
a
m
e
de
t
e
c
t
i
on
R
e
f
e
r
e
nc
e
f
r
a
m
e
S
e
l
f
R
e
l
i
e
s
on r
e
f
e
r
e
nc
e
f
r
a
m
e
f
or
a
c
c
ur
a
c
y
[
53]
K
e
yf
r
a
m
e
de
t
e
c
t
i
on
D
e
l
a
una
y gr
a
ph
c
l
us
t
e
r
i
ng
S
e
l
f
E
xpe
ns
i
ve
i
n c
om
put
a
t
i
ons
[
54]
K
e
yf
r
a
m
e
de
t
e
c
t
i
on
C
l
us
t
e
r
c
l
a
s
s
i
f
i
c
a
t
i
on
S
e
l
f
H
a
s
pr
obl
e
m
s
w
i
t
h m
a
i
nt
a
i
ni
ng a
s
e
n
s
e
of
time
[
55]
K
e
yf
r
a
m
e
de
t
e
c
t
i
on
A
bnor
m
a
l
e
ve
nt
s
S
e
l
f
E
xpe
r
i
e
nc
e
s
t
e
m
por
a
l
di
s
or
i
e
nt
a
t
i
on
[
56]
K
e
yf
r
a
m
e
de
t
e
c
t
i
on
3D
C
N
N
S
e
l
f
I
na
c
c
ur
a
t
e
w
he
n vi
e
w
e
d f
r
om
va
r
i
ous
c
a
m
e
r
a
pe
r
s
pe
c
t
i
ve
s
.
[
57]
O
bj
e
c
t
t
r
a
c
ki
ng a
nd
de
t
e
c
t
i
on
M
ot
i
on
ve
c
t
or
s
a
nd bl
oc
k t
ype
s
S
E
N
S
I
A
C
C
ont
i
nue
d m
oni
t
or
i
ng of
a
l
t
e
r
e
d pa
t
c
he
s
i
s
s
t
i
l
l
ne
c
e
s
s
a
r
y.
[
11]
O
bj
e
c
t
t
r
a
c
ki
ng a
nd
de
t
e
c
t
i
on
B
a
ye
s
i
a
n
a
ppr
oa
c
h
P
E
T
S
-
E
C
C
V
B
a
s
e
d on da
t
a
pr
ovi
de
d by c
ol
our
s
[
58]
O
bj
e
c
t
t
r
a
c
ki
ng a
nd
de
t
e
c
t
i
on
G
M
M
S
e
l
f
L
ong
-
r
a
nge
de
t
e
c
t
i
on of
obj
e
c
t
s
i
s
not
pos
s
i
bl
e
.
[
59]
O
bj
e
c
t
t
r
a
c
ki
ng a
nd
de
t
e
c
t
i
on
C
ont
r
a
s
t
m
ode
l
C
us
t
om
da
t
a
s
e
t
T
r
a
i
ni
ng da
t
a
de
pe
nde
nc
y.
[
60]
F
e
a
t
ur
e
e
xt
r
a
c
t
i
on
hi
s
t
ogr
a
m
M
a
t
c
hi
ng
S
e
l
f
I
t
i
s
s
e
ns
i
t
i
ve
t
o f
or
m
a
t
t
i
ng.
[
61]
F
e
a
t
ur
e
e
xt
r
a
c
t
i
on
C
onvol
ut
i
ona
l
L
S
T
M
S
U
L
F
A
T
he
ge
ne
r
a
l
i
s
a
t
i
on w
a
s
i
m
pr
e
c
i
s
e
.
[
62]
F
e
a
t
ur
e
e
xt
r
a
c
t
i
on
M
L
S
S
e
l
f
D
e
c
r
e
a
s
e
d pr
e
c
i
s
i
on i
n de
t
e
c
t
i
on
[
63]
F
e
a
t
ur
e
e
xt
r
a
c
t
i
on
E
xpone
nt
i
a
l
f
our
i
e
r
t
r
a
ns
f
or
m
s
S
e
l
f
, S
U
L
F
A
L
oc
a
t
e
s
i
ns
t
a
nc
e
s
of
r
e
gi
on dupl
i
c
a
t
i
on
e
xc
l
us
i
ve
l
y
3.
RE
L
A
T
E
D
WORK
O
n
a
da
il
y
ba
s
i
s
,
w
e
c
om
e
a
c
r
os
s
num
e
r
ous
doc
to
r
e
d
f
il
m
s
o
n
va
r
io
us
m
e
di
a
pl
a
tf
or
m
s
in
c
lu
di
ng
W
ha
ts
A
pp,
I
ns
ta
gr
a
m
,
T
ik
T
ok,
S
na
pc
ha
t,
a
nd
F
a
c
e
book.
S
ha
r
in
g
m
a
te
r
ia
l
f
ul
f
il
s
va
r
io
us
f
unc
ti
ons
,
in
c
lu
di
ng
pr
ovi
di
ng
a
m
us
e
m
e
nt
,
di
s
tr
ib
ut
in
g
ne
w
s
,
upd
a
ti
ng
c
om
m
uni
t
ie
s
,
a
nd
di
s
s
e
m
in
a
ti
ng
r
e
li
gi
ous
in
f
or
m
a
ti
on.
N
e
ve
r
th
e
le
s
s
,
in
di
vi
dua
ls
f
r
e
que
nt
ly
di
s
tr
ib
ut
e
th
e
s
e
f
il
m
s
w
it
h
out
be
in
g
a
w
a
r
e
of
th
e
pot
e
nt
ia
l
f
or
th
e
m
to
be
m
odi
f
ie
d,
f
a
ls
if
ie
d,
or
m
a
ni
pul
a
te
d
by
ot
he
r
s
w
it
h
e
vi
l
in
te
nt
s
.
D
e
s
pi
te
m
uc
h
r
e
s
e
a
r
c
h,
no
vi
de
o
f
or
e
ns
ic
s
m
e
th
od, t
e
c
hni
que
, or
t
ool
c
a
n ve
r
if
y a
f
il
m
'
s
va
li
di
ty
.
O
nl
in
e
,
w
e
f
r
e
que
nt
ly
c
om
e
a
c
r
os
s
m
ul
ti
pl
e
f
il
m
s
th
a
t
c
ont
a
in
s
a
m
e
m
a
te
r
ia
l
but
di
f
f
e
r
in
te
r
m
s
of
th
e
ir
vi
de
o
qua
li
ty
.
T
hi
s
phe
nom
e
non
a
r
is
e
s
w
he
n
th
e
r
e
s
ol
ut
io
n
of
th
e
vi
de
o
is
m
odi
f
ie
d
us
in
g
c
onve
r
te
r
te
c
hni
que
s
to
a
r
e
s
ol
ut
io
n
th
a
t
i
s
di
f
f
e
r
e
nt
f
r
om
th
e
or
ig
in
a
l.
W
hi
le
th
e
s
e
f
il
m
s
m
a
y
not
be
e
nt
ir
e
ly
r
e
li
a
bl
e
s
our
c
e
s
of
in
f
or
m
a
ti
on,
th
e
y
a
r
e
in
te
nt
io
na
ll
y
a
lt
e
r
e
d
to
s
e
r
ve
s
pe
c
if
ic
pur
pos
e
s
.
H
e
n
c
e
,
it
is
ne
c
e
s
s
a
r
y
to
de
te
c
t
f
r
a
udul
e
nt
vi
de
os
.
T
hi
s
w
oul
d
unde
ni
a
bl
y
a
id
f
or
e
ns
ic
s
pe
c
ia
li
s
t
s
in
ge
ne
r
a
ti
ng
a
th
or
ough
r
e
por
t
on
th
e
e
vi
de
nc
e
,
he
n
c
e
r
e
duc
in
g
th
e
s
pr
e
a
d
of
f
r
a
udul
e
nt
vi
de
os
.
He
r
e
c
e
nt
ly
di
d
a
s
tu
dy
on
a
udi
o
vi
s
ua
l
f
or
e
ns
ic
s
,
w
it
h
a
s
pe
c
if
ic
e
m
pha
s
is
on
d
e
te
c
ti
ng
a
udi
o
ta
m
pe
r
in
g
th
a
t
is
s
ync
e
d
w
it
h
vi
de
o. T
he
s
p
e
e
c
h
in
th
e
r
e
c
or
di
ng
is
m
odi
f
ie
d
by
e
xpl
oi
ti
ng
s
pe
e
c
h
di
s
c
r
e
pa
n
c
ie
s
.
D
iv
e
r
s
e
m
e
th
odol
ogi
e
s
a
r
e
ut
il
is
e
d
to
gr
a
dua
ll
y
obt
a
in
a
udi
o
-
vi
de
o
r
e
pr
e
s
e
nt
a
ti
on
f
r
om
vi
de
os
.
I
n
r
e
c
e
nt
ti
m
e
s
,
m
a
ny
m
e
th
ods
ha
ve
be
e
n
us
e
d
to
in
te
gr
a
te
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
V
id
e
o f
or
ge
r
y
:
A
n e
x
te
ns
iv
e
analy
s
i
s
of
i
nt
e
r
-
and in
tr
a
-
fr
am
e
m
ani
pul
at
io
n al
ongs
id
e
…
(
Sum
ai
y
a Shaik
h
)
1479
a
udi
o vi
s
ua
l
s
e
lf
-
s
upe
r
vi
s
io
n i
nt
o s
upe
r
vi
s
e
d m
ode
ls
. A
not
he
r
m
e
th
od of
l
e
a
r
ni
ng i
s
i
nt
r
oduc
e
d, w
hi
c
h e
nt
a
il
s
ut
il
is
in
g
th
e
in
he
r
e
nt
s
tr
uc
tu
r
e
of
s
e
pa
r
a
ti
ng
f
r
a
m
e
s
f
r
om
th
e
a
udi
o
tr
a
c
k
in
a
udi
o
-
vi
s
ua
l
le
a
r
ni
ng
r
e
pr
e
s
e
nt
a
ti
on.
B
y
e
m
pl
oyi
ng
a
c
om
bi
na
ti
on
of
di
s
c
r
e
te
w
a
ve
l
e
t
tr
a
ns
f
or
m
(
D
W
T
)
a
nd
pr
in
c
ip
a
l
c
om
pone
nt
a
na
ly
s
is
(
P
C
A
)
on
th
e
vi
d
e
o
s
e
gm
e
nt
,
it
i
s
f
e
a
s
ib
le
to
id
e
nt
if
y
a
ny
ta
m
pe
r
in
g
or
f
or
ge
r
y
in
th
e
c
ont
e
nt
a
t
a
n
in
it
ia
l
pha
s
e
.
T
he
a
udi
o
a
nd
vi
de
o
e
le
m
e
nt
s
a
r
e
ke
pt
s
e
pa
r
a
te
to
f
a
c
il
it
a
te
th
is
pr
oc
e
s
s
.
T
hi
s
a
na
ly
s
i
s
c
a
n
be
s
ys
te
m
a
ti
c
a
ll
y
unde
r
ta
ke
n.
T
h
e
s
ta
ti
ona
r
y
w
a
ve
le
t
tr
a
ns
f
or
m
is
f
ir
s
t
a
ppl
ie
d,
f
ol
lo
w
e
d
by
th
e
in
it
ia
l
s
te
p
o
f
D
W
T
.
U
lt
im
a
te
ly
,
th
e
P
C
A
va
lu
e
is
c
a
lc
ul
a
t
e
d,
yi
e
ld
in
g
c
om
pr
e
he
ns
iv
e
a
c
c
ur
a
c
y
a
nd
pe
r
f
or
m
a
nc
e
pa
r
a
m
e
te
r
s
l
ik
e
a
s
M
S
M
E
, pr
e
c
is
io
n, r
e
c
a
ll
, a
nd F
1 s
c
or
e
.
3.1.
O
b
s
e
r
vat
io
n
s
f
r
o
m
r
e
c
e
n
t
w
or
k
s
T
a
bl
e
5
pr
ovi
de
s
a
c
on
c
is
e
ove
r
vi
e
w
of
th
e
la
te
s
t
s
tr
a
te
gi
e
s
f
or
de
te
c
ti
ng
vi
de
o
f
or
ge
r
ie
s
,
hi
ghl
ig
ht
in
g
s
ig
ni
f
ic
a
nt
a
dva
nc
e
m
e
nt
s
a
nd
r
e
m
a
in
in
g
c
ha
ll
e
nge
s
.
I
t
s
how
c
a
s
e
s
th
e
m
os
t
r
e
c
e
nt
m
e
th
odol
ogi
e
s
,
e
m
pha
s
iz
in
g
how
th
e
y
a
ddr
e
s
s
pr
e
vi
ous
r
e
s
e
a
r
c
h
ga
ps
in
de
te
c
ti
on
a
c
c
ur
a
c
y
a
nd
e
f
f
ic
ie
nc
y.
B
y
s
um
m
a
r
iz
in
g
th
e
s
e
c
ut
ti
ng
-
e
dge
a
ppr
oa
c
he
s
,
th
e
ta
bl
e
of
f
e
r
s
in
s
ig
ht
s
in
to
th
e
pr
ogr
e
s
s
m
a
de
a
nd
th
e
a
r
e
a
s
s
ti
ll
ne
e
di
ng
im
pr
ove
m
e
nt
.
T
hi
s
ov
e
r
vi
e
w
s
e
r
ve
s
a
s
a
va
lu
a
bl
e
r
e
s
our
c
e
f
or
unde
r
s
ta
ndi
ng
th
e
c
ur
r
e
nt
s
ta
te
of
vi
de
o f
or
ge
r
y de
te
c
ti
on r
e
s
e
a
r
c
h.
T
a
bl
e
5. A
s
ynop
s
is
of
m
e
th
ods
f
or
de
te
c
ti
ng vide
o f
or
ge
r
ie
s
, i
n
c
lu
di
ng de
e
p
-
f
a
ke
a
nd pixe
l
m
ot
io
n de
te
c
ti
on
R
e
f
e
r
e
nc
e
M
e
t
hodol
ogy
S
t
r
a
t
e
gy
A
l
gor
i
t
hm
D
a
t
a
s
e
t
L
i
m
i
t
a
t
i
on /
F
ut
ur
e
s
c
ope
[
64]
D
e
e
p l
e
a
r
ni
ng
2D
-
C
N
N
a
nd
S
S
I
M
f
us
i
on
F
e
a
t
ur
e
e
xt
r
a
c
t
i
on
a
l
gor
i
t
hm
V
I
R
A
T
,
S
U
L
F
A
,
L
A
S
I
E
S
T
A
,
I
V
Y
L
A
B
T
he
y
pl
a
n
t
o
m
a
ke
t
he
s
ys
t
e
m
be
t
t
e
r
i
n
t
he
f
ut
ur
e
s
o
t
ha
t
i
t
c
a
n
f
i
nd
m
or
e
i
nt
e
r
-
f
r
a
m
e
f
r
a
uds
.
[
65]
D
e
e
p l
e
a
r
ni
ng
A
da
pt
i
ve
-
T
a
yl
or
-
r
i
de
r
opt
i
m
i
z
a
t
i
on
a
l
gor
i
t
hm
ba
s
e
d
D
C
N
N
D
ua
l
a
da
pt
i
ve
-
T
a
yl
or
-
r
i
de
r
opt
i
m
i
z
a
t
i
on
a
l
gor
i
t
hm
(
D
A
-
T
R
O
A
)
R
e
a
l
da
t
a
s
e
t
T
he
y
pl
a
n
t
o
us
e
m
i
xe
d
opt
i
m
i
s
a
t
i
ons
t
o
t
r
a
i
n
t
he
c
l
a
s
s
i
f
i
e
r
i
n t
he
f
ut
ur
e
.
[
66]
D
e
e
p l
e
a
r
ni
ng
C
N
N
,
C
om
pr
e
s
s
i
on
a
nd
v
i
de
o
t
a
m
pe
r
i
ng
de
t
e
c
t
i
on
V
i
de
o
t
a
m
pe
r
i
ng
de
t
e
c
t
i
on
D
a
t
a
s
e
t
f
r
om
xi
ph.or
g
T
he
y
w
a
nt
t
o
i
m
pr
ove
t
he
f
e
a
t
ur
e
c
om
bi
na
t
i
on
f
or
a
vi
de
o
m
a
ni
pul
a
t
i
on
l
oc
a
l
i
s
e
r
.
[
67]
S
e
que
nt
i
a
l
a
nd
P
a
t
c
h A
na
l
ys
e
s
O
bj
e
c
t
r
e
m
ova
l
f
or
ge
r
y de
t
e
c
t
i
on
O
bj
e
c
t
r
e
m
ova
l
f
or
ge
r
y
de
t
e
c
t
i
on
a
nd l
oc
a
l
i
z
a
t
i
on
L
i
n’
s
vi
de
o s
e
t
T
he
y
pl
a
n
t
o
s
t
udy
non
-
a
ddi
t
i
ve
c
ha
nge
m
ode
l
s
.
[
68]
D
e
e
p l
e
a
r
ni
ng
VGG
-
16
D
e
t
e
c
t
i
ng
di
gi
t
a
l
i
m
a
ge
c
ount
e
r
f
e
i
t
i
ng
w
i
t
h
s
upe
r
vi
s
e
d
l
e
a
r
ni
ng
G
R
I
P
,
D
V
M
M
,
C
M
F
D
,
a
nd
B
S
D
S
300
F
ut
ur
e
f
or
gi
ng
a
t
t
e
m
pt
s
m
a
y
i
nc
l
ude
J
P
E
G
c
om
pr
e
s
s
i
on.
[
69]
M
a
c
hi
ne
l
e
a
r
ni
ng a
nd
D
e
e
p L
e
a
r
ni
ng
C
N
N
, K
N
N
a
nd A
I
D
e
e
p
f
a
ke
vi
de
o
de
t
e
c
t
i
on
D
e
e
p
f
a
ke
de
t
e
c
t
i
on
c
ha
l
l
e
nge
da
t
a
s
e
t
s
T
he
y
pl
a
n
t
o
s
t
udy
de
e
pf
a
ke
de
t
e
c
t
i
on
i
n
N
a
t
i
ona
l
I
D
s
a
nd
ot
he
r
m
e
di
a
.
[
70]
D
e
e
p l
e
a
r
ni
ng
P
i
xe
l
-
r
e
gi
on
r
e
l
a
t
i
on
ne
t
w
or
k
(
P
R
R
N
e
t
)
R
e
l
a
t
i
on
e
nc
ode
r
a
nd
r
e
gi
on
f
e
a
t
ur
e
e
xt
r
a
c
t
or
F
a
c
e
F
or
e
ns
i
c
s
++
,
c
e
l
e
b
-
D
F
a
nd D
F
D
C
I
nc
ons
i
s
t
e
nc
i
e
s
be
t
w
e
e
n
f
r
a
m
e
s
i
n
bogus
vi
d
e
os
ha
ve
not
be
e
n s
t
udi
e
d.
[
71]
D
e
e
p l
e
a
r
ni
ng
I
nc
ons
i
s
t
e
nc
y
-
a
w
a
r
e
w
a
ve
l
e
t
dua
l
-
br
a
nc
h
ne
t
w
or
k
F
a
c
e
f
or
ge
r
y
de
t
e
c
t
i
on
F
a
c
e
F
or
e
ns
i
c
s
++
,
C
e
l
e
b
-
D
F
a
nd
U
A
D
F
V
T
he
y
pl
a
nne
d
t
o
s
t
udy
i
nt
r
a
-
a
nd
i
nt
e
r
-
i
m
a
ge
di
s
c
r
e
pa
nc
i
e
s
.
[
72]
D
e
e
p l
e
a
r
ni
ng
3D
-
C
N
N
F
a
c
e
f
or
ge
r
y
de
t
e
c
t
i
on
F
a
c
e
F
or
e
ns
i
c
s
++
a
nd V
i
dT
I
M
I
T
D
i
f
f
e
r
e
nt
f
a
c
e
r
e
e
na
c
t
m
e
nt
s
ha
ve
ye
t
t
o
be
de
t
e
c
t
e
d.
[
73]
M
a
c
hi
ne
l
e
a
r
ni
ng
M
L
m
ode
l
s
D
i
gi
t
a
l
vi
de
o
pos
t
pr
oc
e
s
s
i
ng de
t
e
c
t
i
on
V
I
S
I
O
N
a
nd
V
i
de
o
-
A
C
I
D
T
he
s
e
m
e
t
hods
ne
e
d
de
e
p
l
e
a
r
ni
ng i
m
pr
ove
m
e
nt
s
.
3.2.
R
e
s
e
ar
c
h
g
ap
s
K
ir
a
n
e
t
al
.
[
2]
pr
opos
e
d
a
nove
l
a
ppr
oa
c
h
f
or
c
a
lc
ul
a
ti
ng
li
ght
c
oe
f
f
ic
ie
nt
s
by
pr
oduc
in
g
a
3D
r
e
pr
e
s
e
nt
a
ti
on
of
vi
de
o
f
r
a
m
e
s
.
T
he
goa
l
of
th
is
s
tr
a
te
gy
is
to
id
e
nt
if
y
a
ny
in
s
ta
nc
e
s
of
f
a
l
s
if
ic
a
ti
on
in
th
e
f
il
m
.
H
ow
e
ve
r
,
th
e
ir
te
c
hnol
ogy
is
li
m
it
e
d
in
it
s
a
bi
li
ty
to
de
te
c
t
s
ubt
le
a
lt
e
r
a
ti
ons
in
m
ovi
e
s
a
nd
r
e
qui
r
e
s
upgr
a
de
s
to
th
e
C
N
N
m
ode
l
f
or
im
pr
ove
d
opt
im
is
a
ti
on
a
n
d
e
f
f
ic
ie
nc
y.
S
im
il
a
r
ly
,
G
ue
r
a
a
nd
D
e
lp
[
1
]
de
m
ons
tr
a
te
d
a
s
ys
te
m
de
s
ig
ne
d
s
pe
c
if
ic
a
ll
y
to
de
te
c
t
in
te
r
-
f
r
a
m
e
f
or
ge
r
ie
s
s
uc
h
a
s
f
r
a
m
e
de
le
ti
on,
in
s
e
r
ti
on,
a
nd
dupl
ic
a
ti
on.
T
he
y
a
c
c
om
pl
is
he
d
th
is
by
us
in
g
a
2D
-
C
N
N
to
c
ol
le
c
t
s
pa
ti
a
l
a
nd
te
m
por
a
l
da
ta
a
nd
th
e
n
c
om
bi
ne
it
f
or
f
e
a
tu
r
e
e
xt
r
a
c
ti
on.
H
ow
e
ve
r
,
th
e
ir
m
e
th
od
la
c
ks
th
e
a
bi
li
ty
to
de
te
c
t
m
a
ny
in
s
ta
nc
e
s
of
m
a
ni
pul
a
ti
on
a
c
r
os
s
f
r
a
m
e
s
in
a
s
in
gl
e
m
ovi
e
.
B
a
r
ni
a
nd
C
os
ta
nz
o
[
7]
c
onduc
te
d
r
e
s
e
a
r
c
h
a
nd
pr
e
s
e
nt
e
d
a
nove
l
ne
twor
k
dubbe
d
P
R
R
N
e
t.
T
he
pr
im
a
r
y
goa
l
of
P
R
R
N
e
t
is
to
d
e
te
c
t
f
a
c
e
c
ount
e
r
f
e
it
in
g
by
a
c
c
ur
a
t
e
ly
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
, V
ol
.
14
, N
o.
2
,
A
pr
il
2025
:
1471
-
1483
1480
r
e
c
or
di
ng
th
e
c
onne
c
ti
on
s
be
twe
e
n
in
di
vi
dua
l
pi
xe
l
s
a
nd
r
e
gi
ons
.
T
h
e
c
ur
r
e
nt
m
e
th
od
r
e
qui
r
e
s
m
or
e
r
e
s
e
a
r
c
h,
na
m
e
ly
in
pr
e
c
is
e
ly
id
e
nt
if
yi
ng
r
e
tu
r
n
on
in
ve
s
tm
e
nt
(
R
O
I
)
to
im
pr
ove
th
e
pr
e
c
is
io
n
a
nd
e
f
f
ic
ie
nc
y
o
f
de
te
c
ti
on,
e
s
p
e
c
ia
ll
y
w
h
e
n
de
te
c
ti
ng
in
te
r
-
f
r
a
m
e
f
or
ge
r
ie
s
.
T
h
is
s
tu
dy
e
xpl
or
e
d
a
c
om
pr
e
he
ns
iv
e
de
te
c
ti
on
a
ppr
oa
c
h
w
it
h
a
dva
nc
e
d
s
pa
ti
a
l
a
nd
te
m
por
a
l
a
na
ly
s
is
.
H
ow
e
v
e
r
,
f
ur
th
e
r
in
-
de
pt
h
s
tu
di
e
s
m
a
y
be
ne
e
de
d
to
c
onf
ir
m
it
s
r
obus
tn
e
s
s
a
c
r
os
s
di
ve
r
s
e
vi
de
o
f
or
m
a
ts
a
nd
v
a
r
yi
ng
le
ve
ls
of
c
om
pr
e
s
s
io
n.
T
he
s
tu
dy
pr
ovi
de
s
di
f
f
e
r
e
nt
ke
y a
s
pe
c
ts
a
nd t
he
y
a
r
e
s
um
m
a
r
iz
e
d i
n t
he
T
a
bl
e
6.
T
a
bl
e
6. S
um
m
a
r
y of
t
he
S
ta
te
-
of
-
th
e
-
a
r
t
A
s
pe
c
t
K
e
y F
i
ndi
ngs
I
nt
e
r
–
f
r
a
m
e
m
a
ni
pul
a
t
i
on
H
i
ghe
r
de
t
e
c
t
i
on a
c
c
ur
a
c
y f
or
f
r
a
m
e
dupl
i
c
a
t
i
on a
nd de
l
e
t
i
on, a
t
t
r
i
but
e
d t
o r
obu
s
t
t
e
m
por
a
l
a
na
l
ys
i
s
.
I
nt
r
a
–
f
r
a
m
e
m
a
ni
pul
a
t
i
on
E
nha
nc
e
d de
t
e
c
t
i
on of
s
pl
i
c
i
ng a
d obj
e
c
t
a
l
t
e
r
a
t
i
ons
due
t
o a
dv
a
nc
e
d s
p
a
t
i
a
l
a
na
l
ys
i
s
D
e
t
e
c
t
i
on
a
c
c
ur
a
c
y
O
ut
pe
r
f
or
m
e
d
e
xi
s
t
i
ng
m
e
t
hods
i
n
pr
e
c
i
s
i
on,
r
e
c
a
l
l
a
nd
a
c
c
ur
a
c
y
f
or
bot
h
i
nt
e
r
–
f
r
a
m
e
a
nd
i
nt
r
a
–
f
r
a
m
e
m
a
ni
pul
a
t
i
ons
C
om
put
a
t
i
ona
l
e
f
f
i
c
i
e
nc
y
M
a
i
nt
a
i
ne
d hi
gh c
om
put
a
t
i
ona
l
e
f
f
i
c
i
e
nc
y, s
ui
t
a
bl
e
f
or
r
e
a
l
–
t
i
m
e
a
ppl
i
c
a
t
i
ons
F
a
l
s
e
pos
i
t
i
ve
s
S
i
gni
f
i
c
a
nt
l
y
l
ow
e
r
f
a
l
s
e
pos
i
t
i
ve
a
nd
f
a
l
s
e
ne
ga
t
i
ve
r
a
t
e
s
c
om
pa
r
e
d
t
o
t
ohe
r
s
t
a
t
e
–
of
–
t
he
–
a
r
t
m
e
t
hods
C
om
pr
e
he
ns
i
ve
a
na
l
y
s
i
s
P
r
ovi
de
d
a
hol
i
s
t
i
c
s
ol
ut
i
on
f
or
de
t
e
c
t
i
ng
bot
h
t
ype
s
of
m
a
ni
pul
a
t
i
ons
,
r
e
duc
i
n
g
t
he
ne
e
d
f
or
m
ul
t
i
pl
e
t
ool
s
.
4.
C
O
N
C
L
U
S
I
O
N
T
hi
s
r
e
vi
e
w
e
s
s
a
y
pr
ovi
de
s
a
th
or
ough
a
s
s
e
s
s
m
e
nt
of
num
e
r
ous
f
a
c
to
r
s
r
e
le
va
nt
to
th
e
id
e
nt
if
ic
a
ti
o
n
of
vi
de
o
f
or
ge
r
ie
s
.
I
t
c
ove
r
s
th
e
li
m
it
a
ti
ons
of
s
uc
h
de
te
c
ti
on
a
ppr
oa
c
he
s
a
nd
lo
oks
a
t
r
e
c
e
nt
r
e
s
e
a
r
c
h
in
th
is
to
pi
c
.
T
he
a
ppr
oa
c
he
s
,
in
ve
s
ti
ga
ti
ons
,
a
nd
pr
oc
e
dur
e
s
pr
e
s
e
nt
e
d
in
th
is
a
r
ti
c
le
a
r
e
c
r
it
ic
a
l
f
or
pr
ogr
e
s
s
in
g
vi
de
o
f
or
ge
r
y
de
te
c
ti
on,
gi
ve
n
th
e
e
ve
r
-
c
ha
ngi
ng
na
tu
r
e
of
da
ta
a
nd
th
e
ne
e
d
f
or
ongoing
in
nov
a
ti
on
to
f
ul
f
il
r
is
in
g
de
m
a
nd.
A
lt
hough
r
e
s
e
a
r
c
he
r
s
ha
ve
pr
im
a
r
il
y
c
onc
e
nt
r
a
te
d
on
a
c
ti
ve
a
ppr
oa
c
he
s
f
or
de
te
c
ti
ng
vi
de
o
c
ount
e
r
f
e
it
in
g,
th
e
r
e
is
a
g
r
ow
in
g
in
te
r
e
s
t
in
pa
s
s
iv
e
s
ol
ut
io
ns
th
a
t
ta
ke
us
e
of
in
dus
tr
y
de
ve
lo
pm
e
nt
s
.
O
ur
s
tu
dy
de
m
ons
tr
a
te
s
th
a
t
c
om
bi
ne
d
s
pa
ti
a
l
a
nd
te
m
por
a
l
a
na
ly
s
is
te
c
hni
que
s
a
r
e
m
or
e
r
e
s
il
ie
nt
th
a
n
m
e
th
ods
f
oc
us
in
g on a
s
in
gl
e
m
a
ni
pul
a
ti
on t
ype
.
F
ut
ur
e
s
tu
di
e
s
m
a
y e
xp
lo
r
e
i
nt
e
gr
a
ti
ng ma
c
hi
ne
l
e
a
r
ni
ng
m
ode
ls
w
it
h
our
a
ppr
oa
c
h,
w
it
h
f
e
a
s
ib
le
w
a
y
s
of
e
nha
nc
in
g
d
e
te
c
ti
on
a
c
c
ur
a
c
y
a
c
r
o
s
s
va
r
io
us
vi
d
e
o
c
om
pr
e
s
s
io
n
s
ta
nda
r
ds
.
T
ypi
c
a
l
is
s
ue
s
i
n vi
de
o f
or
ge
r
y de
te
c
ti
on i
nc
lu
de
i
de
nt
if
yi
ng c
lo
ne
d f
r
a
m
e
s
, dupli
c
a
te
d f
r
a
m
e
s
, a
nd
de
le
te
d
or
in
s
e
r
te
d
f
r
a
m
e
s
.
D
e
s
pi
te
c
ont
in
uous
r
e
s
e
a
r
c
h
in
th
is
to
pi
c
,
no
uni
ve
r
s
a
ll
y
a
ppl
ic
a
bl
e
to
ol
or
a
lg
or
it
hm
f
or
c
or
r
e
c
tl
y
de
te
c
ti
ng
m
a
ni
pul
a
ti
on
in
vi
de
os
ha
s
ye
t
be
e
n
de
v
e
lo
pe
d.
N
one
th
e
l
e
s
s
,
th
is
a
r
ti
c
le
lo
oks
a
t
va
r
io
us
s
ol
ut
io
ns
,
s
u
c
h
a
s
vi
de
o
c
om
pr
e
s
s
io
n.
I
t
is
c
r
i
ti
c
a
l
to
r
e
c
ogni
s
e
th
a
t
th
e
us
e
of
c
om
pr
e
s
s
io
n
te
c
hni
que
s
in
vi
de
o
c
om
pr
e
s
s
io
n
m
ig
ht
r
e
s
ul
t
in
da
ta
lo
s
s
,
le
a
vi
ng
tr
a
c
e
s
of
w
a
te
r
m
a
r
ks
th
a
t
c
a
n
li
m
it
th
e
de
ve
lo
pm
e
nt
of
pr
e
c
is
e
r
e
por
ts
.
A
s
a
r
e
s
ul
t,
it
is
r
e
c
om
m
e
nde
d
to
a
voi
d
c
om
pr
e
s
s
in
g
vi
de
os
a
nd
in
s
te
a
d
us
e
m
e
th
ods
di
r
e
c
tl
y
on
th
e
vi
de
o
to
de
te
c
t
a
ny
in
s
ta
nc
e
s
of
c
ount
e
r
f
e
it
in
g.
R
e
c
e
nt
obs
e
r
va
ti
ons
s
ugge
s
t
th
a
t
e
f
f
e
c
ti
ve
vi
de
o f
or
ge
r
y
de
te
c
ti
on r
e
qui
r
e
s
bot
h s
pa
ti
a
l
a
nd t
e
m
p
or
a
l
a
na
ly
s
is
. O
ur
f
in
di
ngs
pr
ovi
de
c
onc
lu
s
iv
e
e
vi
de
nc
e
th
a
t
th
e
pr
opos
e
d
m
e
th
od s
ig
ni
f
ic
a
nt
ly
im
pr
ove
s
de
te
c
ti
on
a
c
c
ur
a
c
y
a
nd
e
f
f
ic
ie
nc
y,
a
ddr
e
s
s
in
g
bot
h
in
te
r
-
f
r
a
m
e
a
nd
in
tr
a
-
f
r
a
m
e
m
a
ni
pul
a
ti
ons
,
w
it
hout
c
om
pr
om
is
in
g
c
om
put
a
ti
ona
l
p
e
r
f
or
m
a
nc
e
.
T
hi
s
s
ur
ve
y
id
e
nt
if
ie
d
s
e
ve
r
a
l
a
r
e
a
s
w
he
r
e
m
or
e
r
e
s
e
a
r
c
h
is
ne
e
de
d.
A
m
o
ng
th
e
s
e
,
w
e
ha
ve
f
ound
th
e
s
ig
ni
f
ic
a
nc
e
of
r
e
gi
on
of
in
te
r
e
s
t
a
w
a
r
e
ne
s
s
,
im
pr
ove
d
C
N
N
v
a
r
ia
ti
ons
,
a
nd
t
he
c
a
pa
c
it
y
to
id
e
nt
if
y
num
e
r
ous
in
te
r
-
f
r
a
m
e
f
or
ge
r
ie
s
in
a
s
in
gl
e
m
ovi
e
.
I
m
pr
ovi
ng
vi
de
o
f
or
ge
r
y
de
te
c
ti
on
a
nd
de
ve
lo
pi
ng
m
or
e
e
f
f
e
c
ti
ve
s
ys
te
m
s
r
e
li
e
s
on ove
r
c
om
in
g t
he
s
e
l
im
it
a
ti
ons
.
R
E
F
E
R
E
N
C
E
S
[
1]
D
.
G
ue
r
a
a
nd
E
.
J
.
D
e
l
p,
“
D
e
e
pf
a
ke
vi
de
o
de
t
e
c
t
i
on
us
i
ng
r
e
c
ur
r
e
nt
ne
u
r
a
l
ne
t
w
or
ks
,”
i
n
2018
15t
h
I
E
E
E
I
nt
e
r
nat
i
onal
C
onf
e
r
e
nc
e
on
A
dv
anc
e
d
V
i
de
o
and
Si
gnal
B
as
e
d
Su
r
v
e
i
l
l
anc
e
(
A
V
SS
)
,
I
E
E
E
,
N
ov.
2018,
pp.
1
–
6
,
doi
:
10.1109/
A
V
S
S
.2018.8639163.
[
2]
K
i
r
a
n,
B
.
D
.
P
a
r
a
m
e
s
h
a
c
ha
r
i
,
H
.
T
.
P
a
ndur
a
nga
,
a
nd
S
.
l
i
be
r
a
t
a
U
l
l
o,
“
A
na
l
ys
i
s
a
nd
c
om
put
a
t
i
on
of
e
nc
r
ypt
i
on
t
e
c
hni
que
t
o
e
nha
nc
e
s
e
c
ur
i
t
y of
m
e
di
c
a
l
i
m
a
ge
s
,”
I
O
P
C
onf
e
r
e
nc
e
Se
r
i
e
s
:
M
at
e
r
i
al
s
S
c
i
e
nc
e
and E
ngi
ne
e
r
i
ng
,
vol
. 925,
no. 1,
S
e
p. 2020,
doi
:
10.1088/
1757
-
899X
/
925/
1/
012028.
[
3]
N
.
T
.
L
e
,
J
.
-
W
.
W
a
ng,
D
.
H
.
L
e
,
C
.
-
C
.
W
a
ng,
a
nd
T
.
N
.
N
guye
n,
“
F
i
nge
r
pr
i
nt
e
nha
nc
e
m
e
nt
ba
s
e
d
on
t
e
ns
or
of
w
a
v
e
l
e
t
s
ubba
nd
s
f
or
c
l
a
s
s
i
f
i
c
a
t
i
on,”
I
E
E
E
A
c
c
e
s
s
, vol
. 8, pp. 6602
–
6615, 2020, doi
:
10.1109/
A
C
C
E
S
S
.2020.2964035.
[
4]
R
.
D
.
S
i
ngh
a
nd
N
.
A
gga
r
w
a
l
,
“
V
i
de
o
c
ont
e
nt
a
ut
he
nt
i
c
a
t
i
on
t
e
c
hni
que
s
:
a
c
o
m
pr
e
he
ns
i
ve
s
ur
ve
y,”
M
ul
t
i
m
e
di
a
Sy
s
t
e
m
s
,
vol
.
24
,
no. 2, pp. 211
–
240, M
a
r
. 2018, doi
:
10.1007/
s
00530
-
017
-
0538
-
9.
[
5]
H
.
Y
a
o,
S
.
W
a
ng,
Y
.
Z
ha
o,
a
nd
X
.
Z
ha
ng,
“
D
e
t
e
c
t
i
ng
i
m
a
ge
f
or
ge
r
y
us
i
ng
pe
r
s
pe
c
t
i
ve
c
ons
t
r
a
i
nt
s
,”
I
E
E
E
Si
gnal
P
r
oc
e
s
s
i
n
g
L
e
t
t
e
r
s
, vol
. 19, no. 3, pp. 123
–
126, M
a
r
. 2012, doi
:
10.1109/
L
S
P
.2011.218219
1.
[
6]
M
.
B
a
r
ni
a
nd
A
.
C
os
t
a
nz
o,
“
A
f
uz
z
y
a
ppr
oa
c
h
t
o
de
a
l
w
i
t
h
unc
e
r
t
a
i
nt
y
i
n
i
m
a
ge
f
o
r
e
ns
i
c
s
,”
Si
gnal
P
r
oc
e
s
s
i
ng:
I
m
age
C
om
m
uni
c
at
i
on
, vol
. 27, no. 9, pp. 998
–
1010, O
c
t
. 2012, doi
:
10.1016/
j
.i
m
a
ge
.2012.07.006.
[
7]
D
.
V
.
-
P
a
di
n
a
nd
F
.
P
.
-
G
onz
a
l
e
z
,
“
P
r
e
f
i
l
t
e
r
de
s
i
gn
f
or
f
or
e
ns
i
c
r
e
s
a
m
pl
i
ng
e
s
t
i
m
a
t
i
on,”
i
n
2011
I
E
E
E
I
nt
e
r
nat
i
onal
W
or
k
s
hop
on
I
nf
or
m
at
i
on F
or
e
ns
i
c
s
and Se
c
ur
i
t
y
, I
E
E
E
, N
ov. 2011, pp. 1
–
6
, doi
:
10.1109/
W
I
F
S
.2011.6123133.
Evaluation Warning : The document was created with Spire.PDF for Python.