I
AE
S In
t
er
na
t
io
na
l J
o
urna
l o
f
Art
if
icia
l In
t
ellig
ence
(
I
J
-
AI
)
Vo
l.
1
5
,
No
.
5
,
Octo
b
er
2
0
2
6
,
p
p
.
4
8
1
9
~
4
8
2
9
I
SS
N:
2
2
5
2
-
8
9
3
8
,
DOI
: 1
0
.
1
1
5
9
1
/ijai.v
15
.i
5
.
p
p
4
8
1
9
-
4
8
2
9
4819
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
a
i
.
ia
esco
r
e.
co
m
Ev
a
lua
ting YO
L
O
v
8
-
po
se
v
a
ria
nt
s for so
ccer
key
po
int
de
tectio
n
a
nd ho
mo
g
ra
phy
ma
pping
Durro
t
un
Na
s
hi
hin
1
,
Yo
ha
n
es Yo
ha
nie F
rideli
n P
a
nd
um
a
n
2
,
Dwi
Ca
hy
o
K
a
rt
i
k
o
3
,
A
nn
a
No
o
rdia
4
,
K
ho
irul Is
la
m
1
1
D
e
p
a
r
t
me
n
t
o
f
I
n
f
o
r
mat
i
c
s
E
n
g
i
n
e
e
r
i
n
g
,
S
t
a
t
e
U
n
i
v
e
r
si
t
y
o
f
S
u
r
a
b
a
y
a
,
S
u
r
a
b
a
y
a
,
I
n
d
o
n
e
s
i
a
2
D
e
p
a
r
t
me
n
t
o
f
M
u
l
t
i
me
d
i
a
E
n
g
i
n
e
e
r
i
n
g
,
T
h
e
U
n
i
v
e
r
si
t
y
o
f
O
s
a
k
a
,
S
u
i
t
a
,
Ja
p
a
n
3
D
e
p
a
r
t
me
n
t
o
f
P
h
y
si
c
a
l
E
d
u
c
a
t
i
o
n
,
S
t
a
t
e
U
n
i
v
e
r
si
t
y
o
f
S
u
r
a
b
a
y
a
,
S
u
r
a
b
a
y
a
,
I
n
d
o
n
e
si
a
4
D
e
p
a
r
t
me
n
t
o
f
S
p
o
r
t
s S
c
i
e
n
c
e
,
S
t
a
t
e
U
n
i
v
e
r
si
t
y
o
f
S
u
r
a
b
a
y
a
,
S
u
r
a
b
a
y
a
,
I
n
d
o
n
e
si
a
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Dec
1
,
2
0
2
5
R
ev
is
ed
Au
g
4
,
2
0
2
6
Acc
ep
ted
Au
g
1
2
,
2
0
2
6
S
o
c
c
e
r
a
n
a
ly
ti
c
s
in
te
n
se
ly
re
li
e
s
o
n
a
c
c
u
ra
te
p
it
c
h
k
e
y
p
o
i
n
t
d
e
t
e
c
ti
o
n
a
n
d
h
o
m
o
g
ra
p
h
y
m
a
p
p
i
n
g
,
y
e
t
so
c
c
e
r
b
ro
a
d
c
a
st
d
a
ta
p
o
se
s
c
h
a
ll
e
n
g
e
s
c
a
u
se
d
b
y
c
a
m
e
ra
a
n
g
le
v
a
riatio
n
a
n
d
p
a
rti
a
l
p
it
c
h
c
o
v
e
ra
g
e
.
T
h
is
stu
d
y
s
y
ste
m
a
ti
c
a
ll
y
e
v
a
lu
a
tes
y
o
u
o
n
l
y
lo
o
k
o
n
c
e
v
e
rsio
n
8
(
YO
LOv
8
)
-
p
o
se
v
a
rian
ts
(
n
a
n
o
,
sm
a
ll
,
m
e
d
iu
m
,
larg
e
,
a
n
d
x
-
larg
e
)
fo
r
k
e
y
p
o
in
t
d
e
tec
ti
o
n
a
n
d
h
o
m
o
g
ra
p
h
y
m
a
p
p
in
g
.
Th
e
se
m
o
d
e
ls
we
re
re
t
ra
in
e
d
o
n
b
ro
a
d
c
a
st
a
n
d
sc
o
u
t
in
g
d
a
tas
e
ts,
wh
il
e
h
o
m
o
g
ra
p
h
y
m
a
p
p
in
g
wa
s
c
o
m
p
u
te
d
u
sin
g
d
irec
t
li
n
e
a
r
tran
sfo
rm
(DLT
)
a
n
d
ra
n
d
o
m
sa
m
p
le
c
o
n
se
n
su
s
(
RAN
S
AC).
T
h
re
e
m
e
tri
c
s
we
re
u
se
d
fo
r
e
v
a
lu
a
ti
o
n
:
d
e
tec
ti
o
n
a
c
c
u
ra
c
y
,
c
o
m
p
u
tatio
n
a
l
e
fficie
n
c
y
,
a
n
d
h
o
m
o
g
ra
p
h
y
re
li
a
b
il
it
y
.
Re
su
lt
s
sh
o
we
d
th
a
t
th
e
m
e
d
i
u
m
v
a
rian
t
a
c
h
iev
e
d
th
e
b
e
st
b
a
lan
c
e
,
with
th
e
lo
we
st
m
e
a
n
p
ix
e
l
e
rro
r
(3
5
.
6
7
p
x
),
c
o
m
p
e
ti
ti
v
e
in
fe
re
n
c
e
sp
e
e
d
4
9
.
6
fra
m
e
s
p
e
r
se
c
o
n
d
(
FPS
),
a
n
d
m
o
d
e
ra
te
m
o
d
e
l
siz
e
(5
2
.
2
M
B).
F
u
rt
h
e
rm
o
re
,
it
a
lso
re
a
c
h
e
d
th
e
h
ig
h
e
st
h
o
m
o
g
ra
p
h
y
su
c
c
e
ss
ra
te
(9
6
.
2
%
)
,
su
r
p
a
ss
in
g
a
ll
o
t
h
e
r
v
a
rian
ts.
Th
e
se
fi
n
d
in
g
s
su
g
g
e
st
th
a
t
th
e
m
e
d
iu
m
v
a
rian
t
is
th
e
m
o
st
e
ffe
c
ti
v
e
b
a
lan
c
e
fo
r
re
a
l
-
ti
m
e
so
c
c
e
r
a
n
a
ly
ti
c
s,
p
ro
v
id
i
n
g
r
o
b
u
st
p
e
rfo
rm
a
n
c
e
in
a
c
c
u
ra
c
y
,
e
fficie
n
c
y
,
a
n
d
re
li
a
b
i
li
ty
u
n
d
e
r
d
iv
e
rse
b
r
o
a
d
c
a
st
d
a
ta.
F
u
t
u
re
st
u
d
ies
wo
u
l
d
fo
c
u
s
o
n
e
n
larg
i
n
g
t
h
e
d
a
tas
e
t
with
a
d
d
it
io
n
a
l
k
e
y
p
o
in
ts,
i
n
c
o
r
p
o
ra
ti
n
g
tem
p
o
ra
l
sm
o
o
th
i
n
g
,
a
n
d
i
n
teg
ra
ti
n
g
a
p
lay
e
r
trac
k
in
g
sy
ste
m
.
F
o
r
re
p
r
o
d
u
c
i
b
il
i
ty
,
t
h
e
d
a
tas
e
t
a
n
d
c
o
d
e
a
re
o
p
e
n
l
y
a
v
a
il
a
b
le o
n
G
it
Hu
b
.
K
ey
w
o
r
d
s
:
B
r
o
ad
ca
s
t d
ata
C
o
m
p
u
tatio
n
al
ef
f
icien
c
y
Dete
ctio
n
ac
cu
r
ac
y
Ho
m
o
g
r
a
p
h
y
m
ap
p
in
g
Ho
m
o
g
r
a
p
h
y
r
eliab
ilit
y
Key
p
o
in
t d
etec
tio
n
YOL
Ov
8
-
p
o
s
e
v
ar
ia
n
ts
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
:
Du
r
r
o
tu
n
Nash
ih
in
Dep
ar
tm
en
t o
f
I
n
f
o
r
m
atics E
n
g
in
ee
r
in
g
,
State
Un
iv
er
s
ity
o
f
Su
r
ab
ay
a
B
u
ild
in
g
A1
0
,
Ketin
tan
g
W
iy
ata
Stre
et,
Su
r
ab
ay
a
-
6
0
2
3
1
,
I
n
d
o
n
esia
E
m
ail: d
u
r
r
o
t
u
n
n
ash
ih
i
n
@
u
n
esa.
ac
.
id
1.
I
NT
RO
D
UCT
I
O
N
Desp
ite
th
e
r
ap
id
d
ev
elo
p
m
e
n
t
o
f
s
o
cc
er
an
aly
tics
,
o
p
en
-
s
o
u
r
ce
p
latf
o
r
m
s
an
d
ac
ce
s
s
ib
le
s
o
cc
er
s
co
u
tin
g
d
ata
r
em
ain
lim
ited
.
T
h
is
s
itu
atio
n
ca
n
b
e
ad
d
r
ess
ed
b
y
u
tili
zin
g
b
r
o
ad
ca
s
t
d
ata
f
o
r
lo
w
-
co
s
t
s
o
cc
er
an
aly
tics
.
Ma
jo
r
t
o
u
r
n
am
en
ts
r
u
n
b
y
Féd
ér
atio
n
I
n
te
r
n
atio
n
ale
d
e
Fo
o
tb
all
Ass
o
ciatio
n
(
FIFA
)
,
lik
e
E
n
g
lis
h
Pre
m
ier
L
ea
g
u
e
a
n
d
W
o
r
ld
C
u
p
,
ev
e
n
h
av
e
b
ee
n
g
en
er
atin
g
th
o
u
s
an
d
s
o
f
h
o
u
r
s
o
f
h
ig
h
p
lau
s
ib
le
d
ata
f
o
r
s
ca
lab
le
s
o
cc
er
an
aly
tics
[
1
]
.
T
h
is
co
n
d
itio
n
c
o
n
f
ir
m
s
th
at
th
e
b
r
o
ad
ca
s
t
d
ata
p
o
ten
tially
o
f
f
er
s
s
u
b
s
tan
tial
r
eso
u
r
ce
s
f
o
r
lar
g
e
-
s
ca
le
s
o
cc
er
an
aly
tics
.
Ho
wev
er
,
in
co
m
p
ar
is
o
n
to
s
co
u
tin
g
d
ata,
b
r
o
ad
ca
s
t
d
ata
u
s
u
ally
co
n
tain
s
s
o
m
e
ch
allen
g
es,
s
u
c
h
as
v
ar
y
i
n
g
ca
m
e
r
a
an
g
les,
z
o
o
m
lev
els,
a
n
d
o
cc
lu
d
ed
p
itc
h
r
eg
io
n
s
[
2
]
,
[
3
]
.
T
h
er
ef
o
r
e,
s
u
ch
ch
an
g
e
in
ca
m
er
a
p
er
s
p
ec
tiv
es
p
o
s
e
s
a
m
u
ch
m
o
r
e
ch
allen
g
in
g
task
to
en
s
u
r
e
ac
cu
r
ate
an
d
s
tab
le
s
o
cc
er
an
aly
tics
.
On
e
o
f
th
e
p
iv
o
tal
task
s
u
s
ed
to
s
u
p
p
o
r
t
s
o
cc
er
an
aly
tics
th
r
o
u
g
h
b
r
o
ad
ca
s
t
d
ata
is
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
1
5
,
No
.
5
,
Octo
b
er
2
0
2
6
:
4
8
1
9
-
4
8
2
9
4820
ac
cu
r
ate
d
etec
tio
n
o
f
f
iel
d
lan
d
m
ar
k
/k
e
y
p
o
in
t
[
4
]
.
T
h
is
p
r
o
c
ess
is
cr
u
cial
to
s
tan
d
a
r
d
ize
a
to
p
-
d
o
wn
v
iew
o
f
th
e
p
itch
,
a
p
r
o
ce
s
s
k
n
o
wn
as
h
o
m
o
g
r
a
p
h
y
m
ap
p
in
g
.
T
h
e
alig
n
m
en
t
b
etwe
en
k
ey
p
o
in
t
d
etec
tio
n
a
n
d
h
o
m
o
g
r
ap
h
y
m
ap
p
in
g
is
co
m
m
o
n
ly
a
p
r
er
eq
u
is
ite
f
o
r
d
o
w
n
s
tr
ea
m
task
s
,
s
u
ch
as
p
lay
er
tr
ac
k
in
g
an
d
tactica
l
an
aly
s
is
[
5
]
.
I
n
th
is
way
,
b
o
th
b
ec
o
m
e
two
co
r
e
co
m
p
o
n
e
n
ts
to
en
ab
le
tr
an
s
itio
n
f
r
o
m
b
r
o
ad
ca
s
t
d
ata
in
to
ac
cu
r
ate
an
d
s
ca
lab
le
s
o
cc
er
a
n
aly
tics
.
R
ec
en
t
ar
ch
itectu
r
es
p
r
o
v
i
d
e
u
s
ef
u
l
in
f
o
r
m
atio
n
to
war
d
s
ad
d
r
ess
in
g
th
ese
is
s
u
es.
Sp
o
r
tlig
h
t,
th
e
f
ir
s
t
win
n
er
o
f
th
e
2
0
2
3
So
cc
er
Net
ca
m
er
a
ca
lib
r
atio
n
c
h
allen
g
e
,
u
s
ed
h
ig
h
-
r
eso
lu
tio
n
n
etwo
r
k
v
er
s
io
n
2
-
wid
th
4
8
(
HR
NetV2
-
w4
8
)
to
d
etec
t
p
itc
h
lan
d
m
ar
k
s
v
ia
h
ea
tm
ap
s
.
T
h
ey
ac
h
iev
ed
ac
c
u
r
ate
h
o
m
o
g
r
a
p
h
y
m
a
p
p
in
g
f
r
o
m
v
id
eo
f
r
am
es
to
th
e
to
p
-
d
o
wn
p
itch
v
iew
[
6
]
.
Desp
ite
its
h
ig
h
-
q
u
ality
r
esu
lts
,
HR
Net
is
co
m
p
u
tatio
n
ally
co
s
tly
an
d
m
ig
h
t
n
o
t
b
e
a
p
p
li
ca
b
le
to
r
ea
l
-
tim
e
an
al
y
tics
.
Ho
wev
er
,
th
e
s
ec
o
n
d
win
n
er
,
Sp
iid
eo
,
u
s
ed
y
o
u
o
n
ly
l
o
o
k
o
n
ce
X
-
p
o
s
e
(
Y
OL
OX
-
p
o
s
e
)
as
p
lay
e
r
p
el
v
is
k
ey
p
o
i
n
ts
,
wh
ich
lea
d
s
to
p
r
o
v
in
g
th
at
e
v
en
lig
h
tweig
h
t
an
d
s
p
ee
d
-
o
r
ien
te
d
p
o
s
e
d
etec
to
r
s
ca
n
b
e
ef
f
ec
t
iv
e
[
7
]
.
B
ased
o
n
s
u
ch
in
s
ig
h
t
s
,
th
e
YOL
O
-
p
o
s
e
f
am
ily
m
ay
b
ec
o
m
e
an
ex
c
ellen
t
alter
n
ativ
e
to
d
etec
tin
g
s
o
cc
er
p
itch
lan
d
m
ar
k
s
.
YOL
O
-
p
o
s
e
ca
n
b
e
r
etr
ain
ed
to
d
etec
t
f
ield
lan
d
m
ar
k
s
,
alth
o
u
g
h
o
r
i
g
in
ally
it
wa
s
u
s
ed
to
d
etec
t
h
u
m
a
n
k
ey
p
o
i
n
t.
I
t
p
r
o
v
id
es
f
iv
e
v
ar
ian
ts
(
n
a
n
o
,
s
m
all,
m
ed
iu
m
,
lar
g
e
,
an
d
ex
t
r
a
-
lar
g
e
)
t
h
at
g
iv
e
f
lex
ib
ilit
y
in
ter
m
s
o
f
s
p
ee
d
a
n
d
th
e
co
m
p
lex
ity
o
f
th
e
m
o
d
el
[
8
]
.
As
an
ex
a
m
p
le,
YOL
Ov
8
n
-
p
o
s
e
(
n
an
o
)
to
o
k
1
3
1
.
8
m
s
to
r
u
n
o
n
C
PU
-
ONN
X
with
th
e
m
o
d
el
s
ize
o
f
3
.
3
MB
co
m
p
ar
ed
with
1
,
6
0
7
m
s
p
er
f
r
am
e
an
d
6
9
.
4
MB
in
YOL
Ov
8
x
-
p
o
s
e
(
ex
tr
a
-
lar
g
e
)
[
9
]
.
T
h
e
b
en
ch
m
ar
k
s
o
u
tlin
e
a
n
o
ticea
b
le
s
p
ee
d
ac
cu
r
ac
y
tr
ad
e
-
o
f
f
,
p
r
o
p
o
s
in
g
th
e
n
ee
d
to
f
o
c
u
s
o
n
th
e
c
h
o
ice
o
f
a
v
ar
ia
n
t th
at
will a
d
o
p
t th
e
tim
e
-
s
en
s
itiv
e
s
p
o
r
ts
v
is
u
al
an
aly
tics
.
R
e
g
a
r
d
l
es
s
o
f
f
le
x
i
b
i
li
t
y
f
o
r
ch
o
o
s
i
n
g
t
h
e
Y
O
L
O
-
p
o
s
e
m
o
d
e
l
s
,
it
is
n
o
t
s
p
ec
i
f
i
e
d
y
e
t
w
h
ic
h
v
a
r
i
a
n
t
p
r
o
v
i
d
e
s
t
h
e
m
o
s
t
e
f
f
e
ct
i
v
e
p
e
r
f
o
r
m
a
n
c
e
f
o
r
d
e
t
e
c
t
i
n
g
s
o
c
ce
r
p
i
t
c
h
k
e
y
p
o
i
n
t
s
i
n
d
i
v
e
r
s
e
c
a
m
e
r
a
a
n
g
l
e
s
a
n
d
p
a
r
t
i
a
l
o
cc
l
u
s
i
o
n
s
[
1
0
]
.
T
h
i
s
s
t
a
t
e
m
e
n
t
c
a
n
b
e
s
u
p
p
o
r
t
e
d
b
y
t
h
e
r
e
ce
n
t
s
t
u
d
i
es
d
e
m
o
n
s
t
r
a
ti
n
g
t
h
a
t
t
h
e
ac
t
u
al
e
f
f
e
c
t
i
v
e
n
es
s
o
f
e
ac
h
v
a
r
i
a
n
t
r
el
a
t
e
d
t
o
k
e
y
p
o
i
n
t
d
et
e
c
ti
o
n
o
r
s
o
c
c
e
r
a
n
a
l
y
t
i
cs
is
s
t
il
l
u
n
c
l
ea
r
[
9
]
,
[
1
1
]
.
M
o
r
e
o
v
e
r
,
t
h
e
a
b
i
l
i
t
y
o
f
a
r
e
l
i
a
b
l
e
a
l
g
o
r
i
th
m
t
o
m
a
p
k
e
y
p
o
i
n
t
d
e
t
e
c
t
e
d
w
i
t
h
Y
O
L
O
-
p
o
s
e
t
o
h
o
m
o
g
r
a
p
h
y
m
a
p
p
i
n
g
h
a
s
n
o
t
b
e
e
n
w
i
d
e
l
y
e
x
p
l
o
r
e
d
.
I
n
t
h
is
c
a
s
e
,
s
t
a
b
le
a
n
d
r
e
l
i
a
b
l
e
h
o
m
o
g
r
a
p
h
y
m
a
p
p
i
n
g
i
s
n
e
e
d
e
d
t
o
o
v
e
r
c
o
m
e
a
s
i
t
u
at
i
o
n
w
h
e
r
e
Y
OL
O
-
p
o
s
e
m
o
d
e
l
m
is
p
r
e
d
i
c
t
s
k
e
y
p
o
i
n
ts
,
r
es
u
lt
i
n
g
i
n
o
u
t
l
i
e
r
s
.
T
h
is
is
b
ec
a
u
s
e
i
n
a
cc
u
r
a
t
e
h
o
m
o
g
r
a
p
h
y
m
a
p
p
i
n
g
w
i
l
l
d
i
r
e
ct
l
y
i
m
p
a
ct
d
o
w
n
s
t
r
e
a
m
p
r
o
c
es
s
e
s
,
s
u
c
h
as
a
u
t
o
m
at
e
d
p
l
a
y
e
r
t
r
a
c
k
i
n
g
o
r
t
a
c
ti
c
a
l
a
n
a
l
y
s
is
t
r
a
c
k
i
n
g
[
1
2
]
.
T
h
e
s
e
is
s
u
es
h
i
g
h
l
i
g
h
t
a
c
l
e
a
r
g
a
p
i
n
u
n
d
e
r
s
t
a
n
d
i
n
g
t
h
e
r
e
a
l
p
e
r
f
o
r
m
a
n
c
e
o
f
Y
O
L
O
-
p
o
s
e
v
a
r
i
a
n
ts
f
o
r
k
e
y
p
o
i
n
t
d
e
t
e
c
t
i
o
n
a
n
d
w
h
a
t
i
s
t
h
e
a
p
p
r
o
p
r
i
a
t
e
w
a
y
t
o
u
t
i
l
i
z
e
t
h
e
s
e
d
et
e
c
t
e
d
k
ey
p
o
i
n
t
s
i
n
t
o
r
e
l
ia
b
l
e
h
o
m
o
g
r
a
p
h
y
m
a
p
p
i
n
g
.
T
h
e
r
e
f
o
r
e
,
a
d
e
e
p
e
r
i
n
v
e
s
ti
g
a
t
i
o
n
i
n
t
o
th
e
s
e
as
p
e
c
ts
i
s
i
n
e
v
i
t
a
b
l
y
n
e
e
d
e
d
.
Acc
o
r
d
in
g
to
th
e
m
en
tio
n
ed
is
s
u
es,
th
is
p
ap
er
aim
s
to
s
y
s
tem
atica
lly
ev
alu
ate
th
e
Y
OL
O
-
p
o
s
e
v
ar
ian
t
f
o
r
p
itc
h
k
ey
p
o
in
t
d
et
ec
tio
n
an
d
its
h
o
m
o
g
r
ap
h
y
m
ap
p
in
g
.
T
o
o
u
r
k
n
o
wled
g
e,
th
is
p
ap
er
in
tr
o
d
u
ce
s
th
e
f
ir
s
t
r
ep
r
o
d
u
cib
le
b
aselin
e
o
f
YOL
O
-
p
o
s
e
m
o
d
el
v
ar
a
in
ts
u
s
ed
in
s
o
cc
er
h
o
m
o
g
r
a
p
h
y
m
ap
p
in
g
.
Ou
r
co
n
tr
ib
u
tio
n
in
d
etail
ca
n
b
e
s
u
m
m
ed
u
p
as f
o
llo
ws:
i)
R
o
b
u
s
t
k
ey
p
o
in
t
d
etec
tio
n
ac
cu
r
ac
y
:
d
eter
m
in
e
wh
ich
YO
L
O
-
p
o
s
e
v
ar
ia
n
t
p
e
r
f
o
r
m
s
b
e
s
t
u
n
d
er
n
o
is
y
b
r
o
ad
ca
s
t d
ata
with
o
cc
lu
s
io
n
s
an
d
s
co
u
tin
g
d
ata.
ii)
C
o
m
p
u
tatio
n
al
ef
f
icien
cy
:
ex
am
in
e
th
e
laten
c
y
,
in
f
er
en
ce
s
p
ee
d
an
d
m
o
d
el
s
ize
to
lear
n
th
e
tr
ad
e
-
o
f
f
s
am
o
n
g
r
ea
l
-
tim
e
f
ea
s
ib
ilit
y
an
d
d
etec
tio
n
ac
c
u
r
ac
y
o
f
s
o
cc
er
an
aly
tics
ap
p
licatio
n
.
iii)
Ho
m
o
g
r
a
p
h
y
r
eliab
ilit
y
:
ass
ess
th
e
r
o
b
u
s
tn
ess
o
f
h
o
m
o
g
r
ap
h
y
m
a
p
p
in
g
s
u
n
d
e
r
v
a
r
y
in
g
v
i
d
eo
co
n
d
itio
n
s
to
s
u
p
p
o
r
t
d
o
wn
s
tr
ea
m
tactica
l a
n
aly
s
is
task
s
.
I
n
o
r
d
er
to
lead
th
e
r
ea
d
er
s
th
r
o
u
g
h
th
ese
co
n
tr
i
b
u
tio
n
s
,
th
e
p
ap
er
is
o
r
g
an
ized
in
t
h
e
f
o
llo
win
g
way
s
:
s
ec
tio
n
2
p
r
o
v
i
d
es
r
elate
d
wo
r
k
s
,
s
ec
tio
n
3
d
escr
ib
es
th
e
m
eth
o
d
o
lo
g
y
,
s
ec
tio
n
4
i
n
clu
d
es
r
esu
lts
an
d
d
is
cu
s
s
io
n
,
an
d
f
in
ally
,
s
ec
tio
n
5
g
iv
es th
e
co
n
clu
s
io
n
to
th
e
s
tu
d
y
.
2.
RE
L
AT
E
D
WO
RK
S
T
h
e
r
ec
en
t
in
n
o
v
atio
n
s
in
p
o
s
e
esti
m
atio
n
h
av
e
d
em
o
n
s
tr
a
ted
p
o
s
itiv
e
o
u
tco
m
es
in
f
o
o
t
b
all
f
ield
lo
ca
lizatio
n
[
6
]
,
[
7
]
.
T
h
e
p
er
f
o
r
m
an
ce
o
f
HR
Net
-
b
ased
m
o
d
els
in
k
ey
p
o
in
t
d
etec
tio
n
was
o
b
s
er
v
ed
th
r
o
u
g
h
th
e
So
cc
er
Net
ca
m
er
a
ca
lib
r
atio
n
ch
allen
g
e
.
Ho
wev
e
r
,
it
is
n
o
t
id
ea
l
f
o
r
lo
w
-
en
d
c
o
m
p
u
ter
s
o
r
r
ea
l
-
tim
e
co
n
d
itio
n
s
d
u
e
to
its
h
ig
h
co
m
p
u
tatio
n
al
co
s
t
[
6
]
.
C
o
n
v
er
s
ely
,
YOL
O
-
p
o
s
e
p
r
esen
ts
lig
h
ter
f
o
r
m
s
th
at
o
f
f
er
tr
ad
eo
f
f
s
b
etwe
en
ac
c
u
r
ac
y
a
n
d
s
p
ee
d
,
wh
ic
h
is
a
g
o
o
d
s
o
l
u
tio
n
f
o
r
lim
ited
c
o
m
p
u
ter
r
e
s
o
u
r
ce
s
[
7
]
,
[
8
]
.
As
th
e
u
s
e
o
f
YOL
O
f
r
am
ewo
r
k
s
h
as
b
ee
n
g
r
o
win
g
,
o
th
er
r
esear
ch
er
s
h
av
e
also
b
eg
u
n
to
co
m
p
ar
e
v
ar
i
o
u
s
YOL
O
g
en
er
atio
n
s
(
e.
g
.
,
Y
OL
Ov
5
,
YOL
Ov
7
,
an
d
YO
L
Ov
8
)
to
lear
n
m
o
r
e
ab
o
u
t
tr
ad
e
-
o
f
f
s
b
etwe
en
in
f
er
en
ce
s
p
ee
d
v
er
s
u
s
its
a
cc
u
r
ac
y
.
N
o
n
eth
eless
,
th
e
m
ajo
r
ity
o
f
th
ese
r
esear
ch
es
a
r
e
aim
ed
at
o
b
ject
d
etec
tio
n
an
d
n
o
t
at
k
ey
p
o
in
t
o
r
lan
d
m
ar
k
esti
m
atio
n
[
1
3
]
,
[
1
4
]
.
C
o
n
s
eq
u
e
n
tly
,
co
m
p
ar
is
o
n
s
in
p
er
f
o
r
m
an
ce
b
etwe
en
d
if
f
er
e
n
t
v
ar
ian
ts
o
f
a
p
ar
ticu
lar
YOL
O
-
p
o
s
e
ar
e
n
o
t
well
s
tu
d
ied
,
esp
ec
ially
wh
en
d
etec
tin
g
s
o
cc
er
f
ield
k
ey
p
o
in
ts
in
b
o
th
b
r
o
ad
c
ast an
d
s
co
u
tin
g
s
ce
n
ar
i
o
s
.
Ho
m
o
g
r
a
p
h
y
m
a
p
p
in
g
is
a
cr
u
cial
s
tep
to
ac
h
iev
e
s
o
cc
er
f
ield
lo
ca
lizatio
n
,
co
n
s
is
tin
g
o
f
esti
m
atin
g
a
g
eo
m
etr
ic
tr
an
s
f
o
r
m
atio
n
u
s
in
g
d
etec
ted
p
itch
f
ea
tu
r
es
in
t
o
a
s
tan
d
a
r
d
ized
f
ield
tem
p
late.
E
ar
lier
a
p
p
r
o
ac
h
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
E
va
lu
a
tin
g
YOLOv8
-
p
o
s
e
va
r
ia
n
ts
fo
r
s
o
cc
er k
ey
p
o
in
t d
etec
t
io
n
a
n
d
h
o
m
o
g
r
a
p
h
y
…
(
Du
r
r
o
tu
n
N
a
s
h
ih
in
)
4821
was
b
ased
o
n
class
ical
co
m
p
u
ter
v
is
io
n
alg
o
r
ith
m
s
lik
e
H
o
u
g
h
-
b
ased
lin
e
d
etec
tio
n
,
el
lip
s
e
f
itti
n
g
ce
n
ter
cir
cle,
an
d
tem
p
late
m
atch
in
g
p
en
alty
b
o
x
[
1
5
]
.
Alth
o
u
g
h
i
t
wo
r
k
ed
well
in
a
c
o
n
tr
o
lled
en
v
ir
o
n
m
en
t,
lik
e
s
co
u
tin
g
d
ata,
it
is
g
e
n
er
ally
w
o
r
s
e
wh
en
u
s
ed
i
n
a
b
r
o
a
d
ca
s
t e
n
v
ir
o
n
m
en
t.
T
h
er
ef
o
r
e,
it
is
n
o
t
a
p
r
o
p
e
r
o
p
tio
n
f
o
r
a
lar
g
e
-
s
ca
le
au
to
m
ated
an
aly
s
is
[
1
6
]
.
W
ith
th
e
em
e
r
g
en
ce
o
f
d
ee
p
lear
n
in
g
,
th
e
g
en
r
e
c
h
an
g
e
d
to
co
n
v
o
l
u
tio
n
al
n
eu
r
al
n
etwo
r
k
s
(
C
NN
s
)
with
th
e
ab
ilit
y
to
l
ea
r
n
d
is
cr
im
in
ativ
e
v
is
u
al
r
ep
r
esen
tatio
n
s
.
E
ar
ly
s
tu
d
ies
wer
e
b
ased
o
n
s
em
an
t
ic
s
eg
m
en
tatio
n
n
etwo
r
k
s
,
in
c
lu
d
in
g
f
u
lly
co
n
v
o
lu
tio
n
al
n
et
wo
r
k
s
(
FC
Ns)
an
d
Dee
p
L
ab
to
s
eg
m
en
t
f
ield
lin
es
o
r
ar
ea
s
o
f
in
ter
est
[
1
7
]
,
[
1
8
]
.
Seg
m
en
tatio
n
-
b
ased
p
i
p
e
lin
es,
th
o
u
g
h
m
o
r
e
r
o
b
u
s
t,
we
r
e
f
r
eq
u
e
n
tly
c
o
s
tly
in
ter
m
s
o
f
a
n
n
o
tatio
n
s
an
d
h
a
d
a
d
if
f
icu
lt
tim
e
co
m
p
e
tin
g
with
r
ea
l
-
tim
e
p
ip
elin
es.
Acc
o
r
d
in
g
to
t
h
is
cu
r
r
en
t
s
itu
atio
n
,
k
ey
p
o
in
t
-
b
as
ed
m
eth
o
d
s
is
n
ec
ess
ar
y
to
b
e
ex
p
lo
r
ed
as
th
ey
p
r
o
v
id
e
a
h
ig
h
ly
p
r
o
m
is
in
g
,
li
g
h
tweig
h
t
,
an
d
f
lex
ib
le
alter
n
a
tiv
e
f
o
r
h
o
m
o
g
r
ap
h
y
m
a
p
p
in
g
.
3.
M
E
T
H
O
D
Fig
u
r
e
1
p
r
esen
ts
th
e
s
y
s
tem
o
v
er
v
iew
o
f
th
is
s
tu
d
y
with
two
in
p
u
t
d
atasets
(
b
r
o
ad
ca
s
t
an
d
s
co
u
tin
g
)
,
f
iv
e
YOL
Ov
8
-
p
o
s
e
m
o
d
el
v
ar
ian
ts
f
o
r
k
e
y
p
o
i
n
t
d
etec
tio
n
,
an
d
th
r
ee
m
ai
n
ev
a
lu
atio
n
s
.
Alth
o
u
g
h
b
r
o
ad
ca
s
t
d
ata
is
o
u
r
m
ain
c
o
n
ce
r
n
o
f
th
is
s
tu
d
y
,
s
co
u
tin
g
d
ata
is
s
till
in
clu
d
ed
s
in
ce
we
wan
t
to
em
p
h
asize
th
at
o
u
r
ap
p
r
o
ac
h
ca
n
wo
r
k
well
ac
r
o
s
s
d
if
f
er
en
t
ca
m
er
a
p
er
s
p
ec
tiv
es
a
n
d
f
ield
co
n
d
itio
n
s
.
T
h
e
n
,
YOL
Ov
8
-
p
o
s
e
is
ch
o
s
en
b
ec
a
u
s
e
it
h
as
b
ee
n
wid
ely
u
s
ed
a
n
d
well
-
v
alid
ated
p
o
s
e
m
o
d
el
am
o
n
g
r
esear
ch
er
s
in
d
if
f
er
en
t
ar
ea
[
1
9
]
,
[
2
0
]
.
S
u
b
s
ec
tio
n
3
.
1
ex
p
lai
n
s
th
e
d
ataset,
p
r
ep
r
o
ce
s
s
in
g
an
d
tr
ain
in
g
s
tr
ateg
y
,
wh
ile
s
u
b
s
ec
tio
n
s
3
.
2
-
3
.
4
d
escr
ib
e
th
e
th
r
ee
m
ain
e
v
alu
atio
n
s
o
f
th
is
s
tu
d
y
:
ac
cu
r
ac
y
e
v
alu
ati
o
n
,
c
o
m
p
u
tatio
n
al
ef
f
icien
cy
,
a
n
d
h
o
m
o
g
r
ap
h
y
m
ap
p
in
g
r
eliab
ilit
y
.
T
h
e
d
etails
o
f
th
e
ev
al
u
atio
n
m
etr
ics ca
n
b
e
s
ee
n
in
T
ab
le
1
.
Fig
u
r
e
1
.
Sy
s
tem
o
v
er
v
iew
o
f
s
o
cc
er
f
ield
k
ey
p
o
in
t
d
etec
tio
n
an
d
h
o
m
o
g
r
ap
h
y
m
a
p
p
in
g
T
ab
le
1
.
Su
m
m
a
r
y
o
f
ev
alu
ati
o
n
m
etr
ics
Ev
a
l
u
a
t
i
o
n
M
e
t
r
i
c
s
A
c
c
u
r
a
c
y
e
v
a
l
u
a
t
i
o
n
Lo
ss
f
u
n
c
t
i
o
n
s
,
me
a
n
a
v
e
r
a
g
e
p
r
e
c
i
si
o
n
(
mA
P
)
a
n
d
mea
n
p
i
x
e
l
e
r
r
o
r
(
M
P
E)
C
o
m
p
u
t
a
t
i
o
n
a
l
e
f
f
i
c
i
e
n
c
y
La
t
e
n
c
y
,
i
n
f
e
r
e
n
c
e
s
p
e
e
d
,
a
n
d
m
o
d
e
l
s
i
z
e
H
o
mo
g
r
a
p
h
y
ma
p
p
i
n
g
r
e
l
i
a
b
i
l
i
t
y
R
a
t
i
o
o
f
su
c
c
e
ss ra
t
e
(
%)
3
.
1
.
Da
t
a
s
et
,
prepro
ce
s
s
ing
,
a
nd
t
ra
ini
ng
s
t
ra
t
eg
y
T
h
i
s
s
t
u
d
y
u
s
e
s
t
h
e
S
o
cc
e
r
f
ie
l
d
l
o
c
a
l
iz
a
t
i
o
n
d
a
ta
s
et
f
r
o
m
R
o
b
o
f
l
o
w
[
2
1
]
,
c
o
n
t
a
i
n
i
n
g
5
1
4
m
a
n
u
a
l
l
y
a
n
n
o
t
a
t
e
d
k
e
y
p
o
i
n
t
i
m
a
g
e
s
f
r
o
m
s
c
o
u
t
i
n
g
a
n
d
b
r
o
a
d
c
a
s
t
d
a
t
a.
F
o
r
f
u
r
t
h
e
r
r
e
p
r
o
d
u
c
i
b
i
l
i
t
y
,
d
a
t
a
s
et
a
n
d
c
o
d
e
c
a
n
b
e
a
c
c
e
s
s
e
d
in
G
i
t
H
u
b
a
t
h
t
t
p
s
:
/
/
g
i
t
h
u
b
.
c
o
m
/
d
u
r
r
o
t
u
n
n
a
s
h
i
h
i
n
/
y
o
l
o
v
8
p
o
s
e
-
s
o
c
c
e
r
-
k
e
y
p
o
i
n
t
-
h
o
m
o
g
r
a
p
h
y
-
m
a
p
p
i
n
g
.
T
o
m
a
k
e
i
t
c
l
e
a
r
,
t
h
e
m
e
a
n
i
n
g
o
f
k
e
y
p
o
i
n
t
i
n
t
h
i
s
r
e
s
e
a
r
c
h
is
a
g
i
v
e
n
p
o
i
n
t
o
n
t
h
e
s
o
c
c
e
r
f
i
e
l
d
,
l
i
k
e
c
o
r
n
e
r
s
o
r
m
a
r
k
e
r
s
o
f
p
e
n
a
l
t
y
a
r
e
a
s
,
a
n
n
o
t
a
t
e
d
w
i
t
h
p
i
x
el
c
o
o
r
d
i
n
a
t
e
s
.
Fo
r
t
h
e
p
r
e
p
r
o
c
e
s
s
i
n
g
s
t
e
p
,
t
h
e
f
l
i
p
_
i
d
x
p
a
r
a
m
e
t
e
r
i
n
t
h
e
d
a
t
a
.
y
a
m
l
w
as
s
l
i
g
h
t
l
y
c
h
an
g
e
d
t
o
e
n
s
u
r
e
t
h
a
t
t
h
e
l
e
f
t
-
r
i
g
h
t
k
e
y
p
o
i
n
t
s
a
r
e
p
r
o
p
e
r
l
y
m
a
t
ch
e
d
i
n
a
u
g
m
e
n
t
a
t
i
o
n
p
r
o
c
e
s
s
.
F
u
r
t
h
e
r
m
o
r
e
,
t
h
e
s
i
ze
o
f
a
l
l
i
n
p
u
t
i
m
a
g
e
s
w
as
s
t
a
n
d
a
r
d
i
z
e
d
t
o
a
c
e
r
t
a
i
n
r
es
o
l
u
ti
o
n
(
6
4
0
×
6
4
0
p
x
)
t
o
s
t
a
n
d
a
r
d
i
z
e
t
h
e
d
a
t
as
e
ts
a
n
d
o
p
t
i
m
i
z
e
c
o
m
p
u
t
i
n
g
e
f
f
i
c
i
e
n
c
y
.
I
n
t
h
e
t
r
a
i
n
i
n
g
s
t
r
a
t
e
g
y
,
t
h
e
h
y
p
e
r
p
a
r
a
m
e
t
e
r
s
o
f
al
l
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
1
5
,
No
.
5
,
Octo
b
er
2
0
2
6
:
4
8
1
9
-
4
8
2
9
4822
m
o
d
e
l
s
w
e
r
e
c
o
n
f
i
g
u
r
e
d
e
q
u
a
l
ly
.
T
h
e
p
u
r
p
o
s
e
o
f
t
h
is
s
e
tt
i
n
g
is
t
o
e
n
s
u
r
e
t
h
e
p
e
r
f
o
r
m
a
n
c
e
’
s
d
i
f
f
e
r
e
n
c
e
wa
s
b
a
s
ed
o
n
m
o
d
e
l
c
a
p
a
ci
t
y
,
a
n
d
n
o
t
o
n
o
p
t
i
m
i
z
at
i
o
n
b
i
as
.
T
h
e
n
,
t
h
e
s
t
o
c
h
a
s
ti
c
g
r
a
d
i
e
n
t
d
es
c
e
n
t
(
S
GD
)
o
p
t
i
m
iz
e
r
wi
t
h
i
n
i
t
i
al
r
a
t
e
0
.
0
1
was
c
h
o
s
e
n
d
u
e
t
o
i
ts
s
t
a
b
il
i
t
y
a
n
d
e
f
f
i
c
i
e
n
c
y
,
e
s
p
e
c
ia
l
l
y
f
o
r
la
r
g
e
-
s
c
al
e
t
r
ai
n
i
n
g
[
2
2
]
.
T
o
a
d
d
r
e
s
s
t
h
e
s
t
o
c
h
a
s
t
i
c
e
f
f
e
c
ts
,
t
h
e
d
a
t
a
a
u
g
m
e
n
t
a
t
i
o
n
w
a
s
a
s
l
i
g
h
t
a
s
p
o
s
s
i
b
l
e
,
t
u
r
n
i
n
g
m
o
s
a
i
c
a
u
g
m
e
n
t
a
t
i
o
n
o
f
f
,
a
n
d
u
s
i
n
g
s
i
m
p
le
g
e
o
m
e
t
r
i
c
t
r
a
n
s
f
o
r
m
a
t
i
o
n
s
.
3
.
2
.
Acc
ura
cy
ev
a
lua
t
i
o
n
T
h
e
ac
cu
r
ac
y
o
f
th
e
k
ey
p
o
in
t
d
etec
tio
n
m
o
d
els
was
ev
alu
ated
u
s
in
g
th
r
ee
m
et
r
ics:
lo
s
s
f
u
n
ctio
n
s
,
m
AP
an
d
MPE
.
Du
r
in
g
tr
ain
i
n
g
,
lo
s
s
f
u
n
ctio
n
s
(
tr
ain
in
g
an
d
v
alid
atio
n
l
o
s
s
es)
wer
e
u
s
ed
to
m
o
n
ito
r
m
o
d
el
p
er
f
o
r
m
an
ce
.
T
h
e
tr
ai
n
in
g
lo
s
s
m
ea
s
u
r
es
h
o
w
well
th
e
m
o
d
els
f
it
b
etwe
en
p
r
ed
icted
an
d
ac
tu
al
k
ey
p
o
i
n
ts
,
wh
ile
th
e
v
alid
atio
n
lo
s
s
r
ef
lects
th
e
m
o
d
el’
s
ca
p
ab
ilit
y
to
g
en
er
alize
th
e
p
r
e
v
io
u
s
u
n
s
ee
n
d
ata
[
2
3
]
.
I
n
ad
d
itio
n
,
m
AP
at
two
s
tan
d
ar
d
t
h
r
e
s
h
o
l
d
s
(
m
AP
@
0
.
5
an
d
m
AP@
0
.
5
:0
.
9
5
)
was
im
p
lem
e
n
ted
to
m
ea
s
u
r
e
t
h
e
ac
cu
r
ac
y
o
f
k
ey
p
o
in
t
lo
ca
lizat
io
n
u
n
d
er
d
if
f
er
e
n
t
lev
els
o
f
s
tr
ictn
ess
.
As
s
h
o
wn
in
(
1
)
,
th
e
m
AP
is
co
m
p
u
te
d
as
th
e
m
ea
n
o
f
av
er
ag
e
p
r
ec
i
s
io
n
(
AP)
v
alu
es
ac
r
o
s
s
m
u
lti
p
le
th
r
esh
o
ld
s
τ
[
2
4
]
.
T
o
co
m
p
lem
en
t
m
AP,
th
is
s
tu
d
y
ad
o
p
ts
MPE
f
r
o
m
p
r
io
r
wo
r
k
o
n
o
b
ject
p
o
s
e
esti
m
atio
n
to
q
u
an
tify
th
e
ac
c
u
r
ac
y
o
f
k
ey
p
o
i
n
t
d
etec
tio
n
in
o
u
r
s
o
cc
er
p
itch
k
e
y
p
o
in
t
d
etec
tio
n
task
[
2
5
]
.
T
h
is
m
etr
ic
as
in
(
2
)
d
ir
ec
tly
m
ea
s
u
r
es
t
h
e
av
er
a
g
e
E
u
clid
ea
n
d
is
tan
ce
in
p
ix
els b
etwe
en
p
r
e
d
icted
(
̂
,
̂
)
an
d
g
r
o
u
n
d
-
tr
u
th
(
,
)
k
ey
p
o
in
ts
.
=
1
|
|
∑
(
)
(
1
)
=
1
∑
√
(
−
̂
)
2
+
(
−
̂
)
2
=
1
(
2
)
3
.
3
.
Co
m
pu
t
a
t
io
na
l
e
f
f
iciency
T
o
m
ea
s
u
r
e
th
e
co
m
p
u
tatio
n
a
l
ef
f
icien
cy
,
th
is
s
tu
d
y
u
s
ed
t
h
r
ee
m
et
r
ics:
laten
cy
,
in
f
er
e
n
ce
s
p
ee
d
,
an
d
m
o
d
el
s
ize.
T
o
g
eth
e
r
,
th
e
co
m
b
in
atio
n
o
f
th
ese
m
et
r
ics
ca
n
s
u
p
p
o
r
t
th
e
p
r
ac
tical
im
p
l
em
en
tatio
n
o
f
r
ea
l
-
tim
e
s
o
cc
er
an
al
y
tics
.
First
m
etr
ic
is
laten
cy
d
ef
in
ed
as
th
e
av
er
a
g
e
tim
e
t
o
p
r
o
ce
s
s
a
s
in
g
le
f
r
a
m
e.
I
t
is
m
ath
em
atica
lly
s
tated
as
in
(
3
)
,
wh
er
e
is
th
e
s
u
m
o
f
in
f
e
r
en
ce
d
u
r
ati
o
n
o
v
e
r
test
f
r
a
m
es.
As
s
ee
n
in
(
3
)
,
th
e
s
ec
o
n
d
m
etr
ic
is
f
r
a
m
es
p
er
s
ec
o
n
d
(
FPS
)
,
d
escr
ib
in
g
h
o
w
m
an
y
to
tal
ex
ec
u
ted
f
r
am
es
th
at
m
o
d
els
ca
n
p
r
o
ce
s
s
p
er
s
ec
o
n
d
.
T
h
i
s
s
tu
d
y
co
n
s
id
er
ed
th
ese
m
etr
ics
b
ec
au
s
e
th
ey
ar
e
wid
el
y
u
s
ed
in
f
o
r
AI
ap
p
licatio
n
s
,
s
u
ch
as
YOL
O
-
f
all,
a
lig
h
tweig
h
t
ed
g
e
m
o
d
el
in
wh
ich
s
p
ee
d
an
d
ac
c
u
r
ac
y
ar
e
ess
en
tial
[
2
6
]
.
T
h
e
m
o
d
el
is
th
e
last
m
et
r
ic,
wh
er
e
lar
g
er
m
o
d
el
s
ize
u
s
u
ally
r
eq
u
ir
es
g
r
ea
ter
co
m
p
u
tatio
n
al
an
d
m
em
o
r
y
r
eso
u
r
ce
s
.
=
x
100
,
=
x
100
(
3
)
3
.
4
.
H
o
mo
g
ra
ph
y
m
a
pp
ing
re
lia
bil
it
y
T
h
e
h
o
m
o
g
r
ap
h
y
m
a
p
p
in
g
was
ca
lcu
lated
with
d
ir
ec
t
lin
ea
r
tr
an
s
f
o
r
m
(
DL
T
)
wh
ic
h
m
atch
es
id
en
tifie
d
f
ield
k
e
y
p
o
i
n
ts
(
s
o
u
r
ce
p
o
i
n
ts
)
with
th
eir
an
n
o
t
ated
r
ef
er
en
ce
p
o
in
ts
(
d
esti
n
atio
n
p
o
in
ts
)
o
n
th
e
s
o
cc
er
p
itch
tem
p
late
[
2
7
]
.
T
o
m
ak
e
it
r
o
b
u
s
t
ag
ain
s
t
n
o
is
e,
we
o
n
ly
u
s
ed
d
etec
ted
k
ey
p
o
in
ts
with
a
co
n
f
id
en
ce
lev
el
m
o
r
e
th
an
0
.
8
.
Fu
r
th
er
m
o
r
e,
th
e
h
o
m
o
g
r
ap
h
y
m
ap
p
in
g
is
co
n
s
id
er
ed
s
u
cc
ess
f
u
l
o
n
ly
if
m
in
im
al
d
etec
ted
k
ey
p
o
in
ts
ar
e
f
o
u
r
with
less
g
eo
m
etr
ic
alig
n
m
en
t.
B
esid
es,
we
also
u
s
ed
th
e
r
an
d
o
m
s
am
p
le
co
n
s
en
s
u
s
(
R
ANSA
C
)
alg
o
r
ith
m
,
wh
ich
r
ep
ea
ted
ly
ch
o
o
s
es
r
an
d
o
m
s
u
b
s
ets
o
f
k
ey
p
o
in
t
co
r
r
esp
o
n
d
en
ce
s
to
h
y
p
o
th
esize
a
h
o
m
o
g
r
ap
h
y
m
atr
ix
an
d
th
en
ev
alu
ates
h
o
w
m
an
y
co
r
r
esp
o
n
d
en
ce
s
ag
r
ee
with
th
is
h
y
p
o
th
esis
[
2
8
]
.
Key
p
o
in
ts
with
r
ep
r
o
jectio
n
er
r
o
r
less
th
an
th
e
th
r
esh
o
ld
ar
e
co
n
s
id
er
ed
i
n
lier
s
,
an
d
th
e
r
est
ar
e
o
u
tlier
s
.
T
o
m
ea
s
u
r
e
t
h
e
h
o
m
o
g
r
a
p
h
y
r
eliab
ilit
y
,
ea
c
h
YOL
Ov
8
-
p
o
s
e
v
a
r
ian
t
was
ev
alu
ated
o
n
ea
ch
im
ag
e
o
n
t
h
e
test
s
et.
Fo
r
ea
ch
im
ag
e,
a
b
in
ar
y
o
u
tco
m
e
was
o
b
tain
ed
,
s
u
cc
ess
f
u
l
m
ap
p
in
g
as
1
an
d
u
n
s
u
cc
ess
f
u
l
m
ap
p
in
g
as
0
.
As
s
ee
n
in
(
4
)
,
th
e
f
o
r
m
u
la
is
m
ad
e
to
ca
lcu
late
th
e
r
a
tio
o
f
im
ag
es
with
s
u
cc
ess
f
u
l
m
ap
p
in
g
o
v
e
r
all
im
ag
es.
Dif
f
er
en
t
with
m
AP
o
r
MPE
,
wh
ic
h
m
ea
s
u
r
e
ac
cu
r
ac
y
,
th
is
m
et
r
ic
d
ir
ec
tly
ca
p
tu
r
es
th
e
u
s
ab
ilit
y
o
f
th
e
s
y
s
tem
in
a
d
o
wn
s
tr
ea
m
task
,
i.e
.
tactica
l
an
aly
s
is
a
n
d
v
is
u
aliza
tio
n
o
f
p
lay
er
tr
ajec
to
r
ies.
=
1
∑
_
(
)
=
1
(
4
)
4.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
N
4
.
1
.
E
x
perim
ent
a
l
s
et
up
T
h
is
s
ec
tio
n
o
u
tlin
es
th
e
ex
p
er
im
en
tal
s
etu
p
u
s
ed
to
ev
alu
at
e
th
e
p
r
o
p
o
s
e
d
ap
p
r
o
ac
h
.
I
t
co
v
er
s
th
r
ee
k
ey
asp
ec
ts
:
d
a
t
a
s
e
t
co
m
p
o
s
iti
o
n
,
h
y
p
er
p
ar
am
eter
s
ettin
g
s
,
a
n
d
h
ar
d
war
e
s
p
ec
if
icatio
n
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
E
va
lu
a
tin
g
YOLOv8
-
p
o
s
e
va
r
ia
n
ts
fo
r
s
o
cc
er k
ey
p
o
in
t d
etec
t
io
n
a
n
d
h
o
m
o
g
r
a
p
h
y
…
(
Du
r
r
o
tu
n
N
a
s
h
ih
in
)
4823
4
.
1
.
1
.
Da
t
a
s
et
c
o
m
po
s
it
io
n
T
h
e
d
ataset
c
o
m
p
o
s
itio
n
is
s
u
m
m
ar
ized
i
n
T
a
b
le
2
,
wh
ile
th
e
3
2
a
n
n
o
tated
k
e
y
p
o
in
ts
a
r
e
illu
s
tr
ated
in
Fig
u
r
e
2
.
B
esid
es,
Fig
u
r
e
3
s
u
m
m
ar
izes
th
e
d
is
tr
ib
u
tio
n
o
f
k
ey
p
o
in
t
v
is
ib
ilit
y
in
th
e
So
cc
er
Fiel
d
L
o
ca
lizatio
n
d
ataset,
with
th
e
o
v
er
all
co
u
n
ts
f
o
r
th
e
th
r
e
e
v
is
ib
ilit
y
ca
teg
o
r
ies
(
in
v
is
i
b
le,
o
cc
lu
d
e
d
,
a
n
d
v
is
ib
le)
p
r
esen
ted
in
Fig
u
r
e
3
(
a)
,
a
n
d
th
e
d
is
tr
ib
u
tio
n
o
f
v
is
ib
le
k
e
y
p
o
in
ts
s
h
o
wn
i
n
Fig
u
r
e
3
(
b
)
.
T
h
e
h
ig
h
est
n
u
m
b
er
s
o
f
a
n
n
o
tate
d
k
e
y
p
o
i
n
ts
ar
e
o
b
s
er
v
ed
at
r
ef
e
r
en
ce
p
o
in
ts
1
5
,
1
4
,
a
n
d
1
3
,
in
d
icat
in
g
th
at
t
h
e
d
ataset
m
ajo
r
ly
c
o
v
er
s
th
e
m
id
d
le
ar
ea
o
f
t
h
e
p
itch
.
T
h
is
d
is
tr
ib
u
tio
n
p
r
o
v
id
es
a
n
im
p
o
r
tan
t
b
a
s
is
to
m
ea
s
u
r
e
th
e
p
er
f
o
r
m
an
ce
o
f
h
o
m
o
g
r
ap
h
y
m
ap
p
in
g
in
v
ar
i
o
u
s
v
is
ib
ilit
y
co
n
d
itio
n
s
.
T
ab
le
2
.
Data
s
et
co
m
p
o
s
itio
n
S
u
b
s
e
t
N
u
mb
e
r
o
f
i
m
a
g
e
s
D
e
scri
p
t
i
o
n
Tr
a
i
n
i
n
g
4
0
2
2
5
4
s
c
o
u
t
i
n
g
i
m
a
g
e
s,
1
4
8
b
r
o
a
d
c
a
st
i
n
g
i
ma
g
e
s
V
a
l
i
d
a
t
i
o
n
60
3
1
sc
o
u
t
i
n
g
i
ma
g
e
s
,
2
9
b
r
o
a
d
c
a
st
i
n
g
i
mag
e
s
Te
st
i
n
g
52
3
2
sc
o
u
t
i
n
g
i
ma
g
e
s
,
2
0
b
r
o
a
d
c
a
st
i
n
g
i
mag
e
s
To
t
a
l
5
1
4
-
Fig
u
r
e
2
.
Key
p
o
in
t a
n
n
o
tatio
n
o
f
a
s
o
cc
er
p
itch
with
3
2
r
ef
e
r
en
ce
k
ey
p
o
in
ts
(
a)
(
b
)
Fig
u
r
e
3
.
Dis
tr
ib
u
tio
n
o
f
k
ey
p
o
in
t v
is
ib
ilit
y
in
th
e
So
cc
er
Fi
eld
L
o
ca
lizatio
n
d
ataset
: (
a)
th
e
o
v
er
all
co
u
n
ts
f
o
r
ea
ch
v
is
ib
ilit
y
class
an
d
(
b
)
th
e
f
r
eq
u
e
n
cy
o
f
ea
ch
r
ef
er
e
n
ce
p
o
in
t w
h
en
la
b
eled
as v
is
ib
le
4
.
1
.
2
.
H
y
perpa
ra
m
et
er
s
et
t
i
ng
s
T
ab
le
3
d
is
p
lay
s
th
e
h
y
p
er
p
a
r
am
eter
s
ettin
g
s
in
th
is
s
tu
d
y
.
T
h
is
co
n
f
ig
u
r
atio
n
co
m
p
r
o
m
is
ed
b
o
t
h
co
m
p
u
tatio
n
al
ef
f
icien
cy
an
d
r
ep
r
o
d
u
cib
ilit
y
.
S
o
,
lig
h
tweig
h
t
an
d
h
ea
v
y
weig
h
t
YOL
Ov
8
-
p
o
s
e
v
ar
ia
n
ts
co
u
ld
b
e
tr
ain
ed
u
s
in
g
th
e
s
am
e
ex
p
er
im
en
tal
co
n
d
itio
n
s
.
T
ab
le
3
.
Hy
p
er
p
ar
a
m
eter
s
ettin
g
s
f
o
r
YOL
Ov
8
-
p
o
s
e
m
o
d
el
t
r
ain
in
g
P
a
r
a
me
t
e
r
V
a
l
u
e
N
o
t
e
s
Ep
o
c
h
s
5
0
0
Lo
n
g
sc
h
e
d
u
l
e
t
o
e
n
s
u
r
e
c
o
n
v
e
r
g
e
n
c
e
I
mag
e
si
z
e
6
4
0
×
6
4
0
F
i
x
e
d
i
n
p
u
t
r
e
s
o
l
u
t
i
o
n
B
a
t
c
h
si
z
e
32
B
a
l
a
n
c
e
d
f
o
r
G
P
U
me
mo
r
y
e
f
f
i
c
i
e
n
c
y
O
p
t
i
mi
z
e
r
S
G
D
S
t
a
b
l
e
f
o
r
k
e
y
p
o
i
n
t
d
e
t
e
c
t
i
o
n
t
a
s
k
s
I
n
i
t
i
a
l
l
e
a
r
n
i
n
g
r
a
t
e
(
l
r
0
)
0
.
0
1
D
e
f
a
u
l
t
w
a
r
m
-
u
p
r
a
t
e
M
o
s
a
i
c
a
u
g
me
n
t
a
t
i
o
n
0
.
0
D
i
sab
l
e
d
t
o
p
r
e
ser
v
e
p
i
t
c
h
g
e
o
m
e
t
r
y
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
1
5
,
No
.
5
,
Octo
b
er
2
0
2
6
:
4
8
1
9
-
4
8
2
9
4824
4
.
1
.
3
.
H
a
rdwa
re
s
pecif
ica
t
io
n
T
h
e
ex
p
e
r
im
en
tal
en
v
ir
o
n
m
e
n
t
is
o
p
er
atio
n
alize
d
o
n
Go
o
g
le
C
o
lab
Pro
.
G
o
o
g
le
C
o
la
b
Pro
is
a
clo
u
d
-
b
ased
p
latf
o
r
m
th
at
p
r
o
v
id
es
v
er
s
atile
co
m
p
u
tin
g
r
es
o
u
r
ce
s
f
o
r
d
ee
p
lear
n
in
g
ex
p
e
r
im
en
ts
.
Deta
ils
o
f
th
e
h
ar
d
war
e
s
p
ec
if
icatio
n
ca
n
b
e
s
ee
n
in
T
a
b
le
4
.
T
ab
le
4
.
Har
d
war
e
s
p
ec
if
icatio
n
o
f
th
e
ex
p
er
im
en
tal
en
v
ir
o
n
m
en
t
C
o
m
p
o
n
e
n
t
S
p
e
c
i
f
i
c
a
t
i
o
n
G
P
U
N
V
I
D
I
A
A
1
0
0
(
4
0
G
B
V
R
A
M
)
C
P
U
I
n
t
e
l
X
e
o
n
(
1
2
v
C
P
U
s,
~
2
.
2
0
G
H
z
)
R
A
M
8
3
.
5
G
B
D
i
sk
s
t
o
r
a
g
e
1
1
2
.
6
G
B
D
e
v
i
c
e
G
P
U
(
0
)
W
o
r
k
e
r
s
12
4
.
2
.
Acc
ura
cy
perf
o
rma
nce
a
cr
o
s
s
m
o
dels
Fig
u
r
e
4
d
ep
icts
th
e
lo
s
s
f
u
n
c
tio
n
ac
r
o
s
s
th
e
YOL
Ov
8
v
ar
ia
n
ts
o
v
er
5
0
0
ep
o
ch
s
.
Fig
u
r
e
4
(
a)
s
h
o
ws
th
e
tr
ain
in
g
lo
s
s
,
wh
ile
Fig
u
r
e
4
(
b
)
s
h
o
ws
th
e
v
alid
atio
n
l
o
s
s
.
Fig
u
r
e
5
d
e
p
icts
th
e
ac
c
u
r
ac
y
m
etr
ics
ac
r
o
s
s
th
e
YOL
Ov
8
v
ar
ian
ts
o
v
er
5
0
0
ep
o
ch
s
.
Fig
u
r
e
5
(
a)
s
h
o
ws
th
e
m
AP@
0
.
5
,
wh
ile
Fig
u
r
e
5
(
b
)
s
h
o
ws
th
e
m
AP@
0
.
5
:0
.
9
5
.
I
n
g
en
e
r
al,
th
e
m
ed
iu
m
m
o
d
el
s
h
o
u
ld
b
e
co
n
s
id
er
ed
as
t
h
e
b
est
m
o
d
el
d
u
e
to
its
lo
west v
alu
e
in
v
alid
atio
n
lo
s
s
an
d
its
h
ig
h
est
v
alu
e
in
m
AP
@
0
.
5
:0
.
9
5
a
f
ter
5
0
0
ep
o
ch
s
.
I
n
th
e
lo
s
s
f
u
n
ctio
n
cu
r
v
es,
th
e
n
an
o
m
o
d
el
alwa
y
s
h
as
a
h
ig
h
er
a
m
o
u
n
t
o
f
r
esid
u
al.
T
h
e
x
-
l
ar
g
e
m
o
d
el,
o
n
th
e
o
th
e
r
h
an
d
,
h
av
e
m
u
ch
f
aster
r
ed
u
ctio
n
s
in
v
alid
atio
n
lo
s
s
an
d
s
tab
ilizes
ea
r
lier
lik
e
th
e
m
ed
iu
m
o
n
e.
H
o
wev
er
,
a
f
ter
ar
o
u
n
d
4
5
0
e
p
o
ch
s
,
th
e
m
ed
iu
m
m
o
d
el
clea
r
ly
p
e
r
f
o
r
m
s
th
e
lo
west
th
an
t
h
e
o
t
h
er
s
.
I
n
th
e
m
AP@
0
.
5
a
n
d
m
AP@
0
.
5
:0
.
9
5
,
th
e
n
an
o
is
th
e
o
n
l
y
m
o
d
el
with
t
h
e
s
lo
west
co
n
v
er
g
en
ce
,
wh
ile
th
e
x
-
lar
g
e
ca
n
s
till
b
e
d
e
f
e
ated
b
y
th
e
m
e
d
iu
m
m
o
d
el,
esp
ec
ially
af
ter
4
5
0
e
p
o
ch
s
in
m
AP@
0
.
5
:0
.
9
5
.
T
o
s
u
p
p
o
r
t
th
at
th
e
m
e
d
iu
m
s
h
o
u
ld
b
e
ch
o
s
en
as
th
e
b
est
m
o
d
el,
T
ab
le
5
s
h
o
ws
th
e
MPE
o
f
all
m
o
d
el
v
ar
ia
n
ts
.
I
t
is
clea
r
th
at
th
e
m
ed
i
u
m
m
o
d
el
h
as
th
e
lo
west
MPE
(
3
5
.
6
7
p
x
)
th
a
n
th
e
o
th
er
s
.
I
t
also
p
r
o
o
f
s
th
at
th
is
m
o
d
el
ca
n
b
e
u
s
e
d
f
o
r
m
o
r
e
p
r
ec
is
e
k
ey
p
o
i
n
t d
etec
tio
n
.
(
a)
(
b
)
Fig
u
r
e
4
.
T
r
ain
in
g
a
n
d
v
alid
atio
n
lo
s
s
ac
r
o
s
s
YOL
Ov
8
v
ar
ia
n
ts
(
n
an
o
,
s
m
all,
m
e
d
iu
m
,
la
r
g
e,
an
d
x
-
lar
g
e)
o
v
er
5
0
0
e
p
o
ch
s
: (
a)
t
r
ain
in
g
l
o
s
s
an
d
(
b
)
v
alid
atio
n
l
o
s
s
(
a
)
(
b
)
Fig
u
r
e
5
.
Acc
u
r
ac
y
m
etr
ics ac
r
o
s
s
YOL
Ov
8
v
ar
ian
ts
(
n
a
n
o
,
s
m
all,
m
ed
iu
m
,
an
d
lar
g
e,
x
-
la
r
g
e)
o
v
er
5
0
0
ep
o
ch
s
: (
a)
m
AP@
0
.
5
an
d
(
b
)
m
AP@
0
.
5
:0
.
9
5
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
E
va
lu
a
tin
g
YOLOv8
-
p
o
s
e
va
r
ia
n
ts
fo
r
s
o
cc
er k
ey
p
o
in
t d
etec
t
io
n
a
n
d
h
o
m
o
g
r
a
p
h
y
…
(
Du
r
r
o
tu
n
N
a
s
h
ih
in
)
4825
T
ab
le
5
.
MPE
ac
r
o
s
s
YOL
Ov
8
-
p
o
s
e
m
o
d
el
v
ar
ia
n
ts
M
o
d
e
l
v
a
r
i
a
n
t
s
N
a
n
o
S
mal
l
M
e
d
i
u
m
La
r
g
e
X
-
l
a
r
ge
M
P
E
(
p
x
)
7
6
.
3
1
6
0
.
0
0
3
5
.
6
7
5
5
.
5
9
6
7
.
3
5
4
.
3
.
T
ra
de
-
o
f
f
s
in
co
m
pu
t
a
t
i
o
na
l e
f
f
iciency
T
a
b
l
e
6
s
h
o
w
s
t
h
e
c
o
m
p
u
t
a
t
i
o
n
e
f
f
i
c
i
e
n
c
y
v
a
l
u
e
s
a
c
r
o
s
s
t
h
e
f
i
v
e
Y
O
L
O
v
8
-
p
o
s
e
v
a
r
i
a
n
ts
.
T
h
e
f
i
n
d
i
n
g
s
d
e
m
o
n
s
t
r
a
t
e
t
h
at
t
h
e
n
a
n
o
m
o
d
e
l
p
r
e
s
e
n
ts
t
h
e
l
o
w
e
s
t
l
at
e
n
c
y
(
1
6
.
1
m
s
)
,
t
h
e
h
i
g
h
e
s
t
i
n
f
e
r
e
n
c
e
s
p
e
e
d
(
6
2
.
1
F
PS
)
a
n
d
t
h
e
s
m
a
l
l
es
t m
o
d
e
l
s
i
ze
(
7
.
3
5
M
B
)
.
H
o
w
e
v
e
r
,
t
h
i
s
m
o
d
e
l
d
o
e
s
n
o
t
n
e
ce
s
a
r
il
l
y
t
h
e
m
o
s
t
p
r
a
c
t
i
c
al
s
o
l
u
t
i
o
n
f
o
r
r
e
a
l
-
t
i
m
e
s
o
c
c
e
r
a
n
a
l
y
t
i
cs
a
s
a
c
c
u
r
a
c
y
p
e
r
f
o
r
m
a
n
c
e
i
s
a
ls
o
m
a
t
t
e
r
s
,
n
o
t
c
o
m
p
u
t
a
t
i
o
n
a
l
e
f
f
i
c
i
e
n
c
y
o
n
l
y
.
C
o
m
b
i
n
i
n
g
a
c
c
u
r
a
c
y
p
e
r
f
o
r
m
an
c
e
i
n
s
u
b
s
ec
t
i
o
n
4
.
1
a
n
d
c
o
m
p
u
t
a
t
i
o
n
a
l
e
f
f
i
ci
e
n
c
y
i
n
T
a
b
l
e
6
,
t
h
e
m
e
d
i
u
m
m
o
d
e
l
a
c
h
i
e
v
e
s
t
h
e
m
o
s
t
b
a
l
a
n
ce
d
s
o
lu
t
i
o
n
w
i
t
h
c
o
m
p
e
t
i
t
i
v
e
l
at
e
n
c
y
(
1
9
.
7
m
s
)
,
i
n
f
e
r
e
n
c
e
s
p
e
e
d
(
5
0
.
8
F
P
S
)
a
n
d
m
o
d
e
l
s
i
z
e
(
5
2
.
2
MB
)
,
w
h
i
le
m
a
i
n
ta
i
n
i
n
g
r
o
b
u
s
t
d
e
t
e
ct
i
o
n
a
c
c
u
r
a
c
y
.
T
h
e
p
r
i
o
r
r
e
s
e
a
r
c
h
h
a
s
i
d
e
n
t
i
f
i
e
d
a
s
i
m
il
a
r
t
r
a
d
e
-
o
f
f
b
e
t
w
e
e
n
s
p
e
e
d
a
n
d
ac
c
u
r
a
c
y
i
n
e
a
r
l
i
e
r
w
o
r
k
s
a
b
o
u
t
o
b
j
e
c
t
d
et
e
c
ti
o
n
w
it
h
d
e
e
p
l
ea
r
n
i
n
g
m
o
d
e
l
s
[
29
]
.
N
e
v
e
r
t
h
e
l
es
s
,
t
h
e
o
b
s
e
r
v
e
d
p
er
f
o
r
m
a
n
c
e
s
t
il
l
h
as
a
l
i
m
i
ta
t
i
o
n
,
w
h
e
r
e
t
h
e
li
m
i
te
d
d
a
t
as
e
t
an
d
f
i
x
e
d
r
e
s
o
l
u
t
i
o
n
(
6
4
0
×
6
4
0
)
m
a
y
f
a
v
o
r
m
o
d
e
l
s
o
f
m
e
d
i
u
m
c
a
p
a
c
i
t
y
s
i
n
c
e
l
a
r
g
er
m
o
d
e
l
s
m
a
y
n
o
t
f
u
l
l
y
e
x
p
l
o
i
t
t
h
e
i
r
p
o
t
e
n
t
i
al
.
T
ab
le
6
.
C
o
m
p
u
tatio
n
al
ef
f
icie
n
cy
m
ea
s
u
r
e
d
in
laten
cy
,
in
f
er
en
ce
s
p
ee
d
an
d
m
o
d
el
s
ize
M
o
d
e
l
v
a
r
i
a
n
t
s
N
a
n
o
S
mal
l
M
e
d
i
u
m
La
r
g
e
X
-
l
a
r
g
e
La
t
e
n
c
y
(
ms)
1
6
.
1
1
6
.
7
1
9
.
7
2
2
.
8
2
3
.
8
I
n
f
e
r
e
n
c
e
sp
e
e
d
(FPS)
6
2
.
1
5
9
.
8
5
0
.
8
4
3
.
9
4
2
.
0
X
-
M
o
d
e
l
si
z
e
(
M
B
)
7
.
3
5
2
3
.
6
5
2
.
2
8
6
.
3
1
3
3
.
7
4
.
4
.
H
o
mo
g
ra
ph
y
m
a
pp
ing
re
lia
bil
it
y
f
ind
ing
s
Fig
u
r
e
6
r
ep
o
r
ts
th
e
h
o
m
o
g
r
a
p
h
y
s
u
cc
ess
r
ate
o
f
d
if
f
er
en
t
YOL
Ov
8
-
p
o
s
e
m
o
d
el
v
ar
ian
t
s
b
ased
o
n
two
d
ataset
ty
p
es
(
b
r
o
ad
ca
s
t
an
d
s
co
u
tin
g
)
.
T
h
e
b
r
o
a
d
ca
s
t
d
ata
(
2
0
im
ag
es)
y
ield
s
n
an
o
8
0
%
(
1
6
/2
0
)
,
s
m
all
8
5
%
(
1
7
/
2
0
)
,
m
ed
iu
m
9
0
%
(
1
8
/2
0
)
,
lar
g
e
8
0
%
(
1
6
/2
0
)
,
a
n
d
x
-
lar
g
e
8
0
%
(
1
6
/2
0
)
.
T
h
e
n
,
th
e
s
co
u
tin
g
d
ata
(
3
2
im
ag
es)
h
as
a
m
o
r
e
s
tab
le
r
esu
lt,
wh
er
e
n
an
o
1
0
0
%
(
3
2
/
3
2
)
,
s
m
all
1
0
0
%
(
3
2
/3
2
)
,
m
ed
iu
m
1
0
0
%
(
3
2
/3
2
)
,
lar
g
e
1
0
0
%
(
3
2
/
3
2
)
,
an
d
x
-
lar
g
e
9
6
.
9
%
(
3
1
/
3
2
)
.
Acr
o
s
s
th
e
f
u
ll
d
ataset
(
5
2
im
a
g
es),
th
e
o
v
er
all
s
u
cc
ess
r
ates
ar
e
n
a
n
o
9
2
.
3
%,
s
m
all
9
4
.
2
%,
m
ed
iu
m
9
6
.
2
%,
lar
g
e
9
2
.
3
%,
an
d
x
-
lar
g
e
9
0
.
4
%.
T
w
o
co
n
s
is
ten
t
tr
en
d
s
em
er
g
e.
First,
s
co
u
tin
g
im
ag
es
ar
e
alm
o
s
t
al
way
s
s
o
lv
ab
l
e,
d
u
e
to
th
e
f
ix
ed
v
an
tag
e
p
o
in
t
an
d
f
u
ll
-
f
ield
co
v
er
ag
e.
Seco
n
d
,
b
r
o
ad
ca
s
t
im
ag
es
r
ep
r
esen
t
th
e
p
r
im
ar
y
c
h
allen
g
e:
s
u
cc
ess
r
ates
m
ajo
r
l
y
p
latea
u
n
ea
r
8
0
%
ev
en
f
o
r
th
e
lar
g
est m
o
d
els,
w
h
ile
th
e
m
ed
iu
m
ac
h
iev
es 9
0
% a
n
d
th
u
s
p
r
o
v
i
d
es th
e
b
est ch
o
ice
f
o
r
r
eliab
ilit
y
.
Fig
u
r
e
6
.
Su
cc
ess
co
u
n
ts
o
f
h
o
m
o
g
r
ap
h
y
m
a
p
p
in
g
ac
r
o
s
s
YOL
Ov
8
-
p
o
s
e
v
ar
ian
ts
Fig
u
r
e
7
illu
s
tr
ates
h
o
w
s
tr
o
n
g
esti
m
atio
n
ca
n
h
elp
to
o
v
er
co
m
e
th
e
o
u
tlier
c
o
r
r
esp
o
n
d
e
n
ce
s
u
s
in
g
R
ANSA
C
alg
o
r
ith
m
.
T
h
e
k
ey
p
o
in
t
lab
eled
as
#
2
7
(
h
ig
h
lig
h
ted
in
y
ello
w)
ac
ts
as
a
g
eo
m
etr
ic
o
u
tlier
o
f
th
e
p
itch
tem
p
late.
Yet,
R
AN
SA
C
co
r
r
ec
tly
ex
clu
d
es
it
d
u
r
in
g
m
o
d
el
f
itti
n
g
,
allo
win
g
th
e
h
o
m
o
g
r
a
p
h
y
to
b
e
ac
cu
r
ately
r
ec
o
v
er
e
d
.
I
n
th
is
s
tu
d
y
,
t
h
e
R
ANSAC
r
ep
r
o
jectio
n
th
r
esh
o
ld
was
s
et
to
3
p
x
,
wh
ich
is
th
e
d
e
f
au
lt
v
alu
e
f
r
o
m
O
p
en
C
V’
s
lib
r
ar
y
[
3
0
]
.
T
h
is
is
eq
u
iv
alen
t
to
ab
o
u
t
0
.
4
7
%
at
6
4
0
×
6
4
0
r
eso
l
u
tio
n
.
R
ANSAC
i
s
th
e
p
ar
ad
ig
m
o
f
ch
o
ice
in
r
o
b
u
s
t
g
eo
m
etr
ic
f
itti
n
g
i
n
th
e
f
ac
e
o
f
o
u
tlier
co
n
tam
in
atio
n
[
3
1
]
,
an
d
its
ef
f
ec
tiv
en
ess
h
er
e
h
ig
h
lig
h
ts
its
n
ec
ess
ity
f
o
r
s
tab
le
h
o
m
o
g
r
a
p
h
y
m
ap
p
in
g
in
t
h
e
p
r
esen
ce
o
f
n
o
is
y
d
etec
tio
n
s
.
Similar
ap
p
licatio
n
s
h
av
e
b
ee
n
ap
p
lied
t
o
s
p
o
r
ts
an
aly
tics
wh
er
eb
y
s
tr
o
n
g
o
u
tlier
m
an
a
g
em
en
t
is
ess
en
tial
to
r
eg
is
ter
in
g
th
e
co
r
r
ec
t p
itch
[
1
6
]
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
1
5
,
No
.
5
,
Octo
b
er
2
0
2
6
:
4
8
1
9
-
4
8
2
9
4826
Fig
u
r
e
7
.
H
o
m
o
g
r
a
p
h
y
m
ap
p
i
n
g
with
R
ANSAC
: k
ey
p
o
in
t #
2
7
(
y
ello
w)
ac
ts
as a
n
o
u
tlier
b
u
t is co
r
r
ec
tly
ex
clu
d
ed
,
en
ab
lin
g
ac
cu
r
ate
h
o
m
o
g
r
a
p
h
y
r
ec
o
v
er
y
d
esp
ite
n
o
is
y
d
etec
tio
n
s
4
.
5
.
Dis
cus
s
io
n
Acc
o
r
d
in
g
to
th
e
f
in
d
in
g
s
in
s
u
b
s
ec
tio
n
s
4
.
2
an
d
4
.
3
,
t
h
e
m
e
d
iu
m
m
o
d
el
o
f
f
er
s
th
e
b
est
co
m
p
r
o
m
is
e
as
it
o
b
tain
s
th
e
h
ig
h
est
k
e
y
p
o
in
t
d
etec
tio
n
ac
cu
r
ac
y
,
wh
ile
m
ain
tain
in
g
c
o
m
p
u
tatio
n
a
l
ef
f
icien
cy
.
T
h
is
in
d
icate
s
th
at
its
m
o
d
el
co
m
p
lex
ity
is
alr
ea
d
y
s
u
f
f
icien
t
to
c
ap
tu
r
e
th
e
ch
a
r
ac
ter
is
tic
o
f
o
u
r
d
ataset
with
f
i
x
ed
in
p
u
t
r
eso
l
u
tio
n
(
6
4
0
×
6
4
0
)
.
Fo
r
d
ee
p
e
r
u
n
d
er
s
tan
d
in
g
wh
y
th
is
o
cc
u
r
s
,
we
ca
n
t
ak
e
a
l
o
o
k
at
th
e
ch
ar
ac
ter
is
tic
o
f
th
e
n
an
o
as
th
e
s
m
allest
m
o
d
el
an
d
th
e
x
-
lar
g
e
as
th
e
lar
g
est
m
o
d
el.
T
h
e
s
m
aller
m
o
d
els,
lik
e
th
e
n
an
o
,
h
as
lo
wer
r
esid
u
als
in
th
e
lo
s
s
f
u
n
ctio
n
s
d
u
e
to
its
s
m
aller
r
ep
r
esen
tatio
n
al
ab
ilit
y
.
I
n
th
e
co
n
tr
ar
y
,
th
e
lar
g
er
m
o
d
els,
lik
e
x
-
lar
g
e
,
ca
n
n
o
t
b
e
th
e
b
est
as
it
ca
n
n
o
t
m
ax
im
ize
its
ab
ilit
y
to
ca
p
tu
r
e
th
e
lim
ited
d
ataset
with
f
ix
ed
in
p
u
t
r
eso
lu
tio
n
(
6
4
0
×
6
4
0
)
.
Ov
er
f
itti
n
g
ca
n
ten
d
to
b
e
c
o
m
m
o
n
wh
en
th
e
m
o
d
el
ca
p
ac
ity
is
lar
g
er
th
an
th
e
d
ata
u
s
ed
i
n
task
s
.
T
h
e
is
s
u
e
is
also
s
u
p
p
o
r
ted
b
y
o
t
h
er
s
tu
d
ies
d
escr
ib
in
g
th
at
a
m
o
d
el
ca
p
ac
ity
an
d
d
ataset
s
ize
ar
e
im
p
o
r
tan
t
f
ac
to
r
f
o
r
k
e
y
p
o
i
n
t d
etec
tio
n
[
3
2
]
,
[
3
3
]
.
T
ak
in
g
in
to
ac
c
o
u
n
t
th
e
f
in
d
in
g
s
in
s
u
b
s
ec
tio
n
4
.
4
,
th
is
p
r
o
p
o
s
e
d
s
y
s
tem
s
till
h
as
a
l
im
itati
o
n
d
esp
ite
th
e
r
o
b
u
s
tn
ess
o
f
t
h
e
h
o
m
o
g
r
ap
h
y
m
ap
p
i
n
g
.
Fig
u
r
e
8
e
x
a
m
in
es
th
e
two
f
ailu
r
e
ca
s
es
o
f
th
e
YOL
Ov
8
-
p
o
s
e
m
ed
iu
m
m
o
d
el,
wh
ich
o
th
er
wis
e
ac
h
iev
ed
th
e
h
ig
h
est
o
v
er
all
s
u
cc
ess
r
ate
(
9
6
.
2
%,
5
0
/5
2
im
a
g
es).
B
o
t
h
f
ailu
r
es
o
cc
u
r
in
b
r
o
a
d
ca
s
t
im
ag
es
an
d
s
h
ar
e
th
e
s
am
e
ca
u
s
e:
f
ewe
r
th
an
f
o
u
r
k
e
y
p
o
in
ts
wer
e
d
etec
ted
.
Sin
ce
a
m
in
im
u
m
o
f
f
o
u
r
n
o
n
-
c
o
llin
ea
r
co
r
r
esp
o
n
d
en
ce
s
is
m
a
th
em
atica
lly
r
eq
u
ir
e
d
to
co
m
p
u
te
a
p
r
o
jectiv
e
tr
an
s
f
o
r
m
atio
n
,
R
ANSAC
co
u
ld
n
o
t
p
r
o
d
u
ce
a
v
alid
h
o
m
o
g
r
ap
h
y
.
T
h
is
h
ig
h
lig
h
ts
wh
y
th
e
m
ed
i
u
m
m
o
d
el
co
u
ld
n
o
t
ac
h
iev
e
1
0
0
%
r
elia
b
ilit
y
u
n
d
er
b
r
o
a
d
ca
s
t
d
ata,
d
e
s
p
ite
o
u
tp
er
f
o
r
m
i
n
g
all
o
th
e
r
m
o
d
els.
I
n
c
o
n
tr
ast,
s
co
u
tin
g
im
a
g
es
co
n
s
is
ten
tly
p
r
o
v
id
e
en
o
u
g
h
well
-
d
is
tr
ib
u
ted
k
e
y
p
o
i
n
ts
,
r
esu
ltin
g
in
p
er
f
ec
t
h
o
m
o
g
r
ap
h
y
m
ap
p
in
g
.
Similar
co
n
s
tr
ain
ts
o
n
k
ey
p
o
in
t
d
en
s
ity
an
d
s
p
atial
s
p
r
ea
d
h
a
v
e
b
ee
n
o
b
s
er
v
ed
in
o
th
e
r
p
itch
r
eg
is
tr
atio
n
p
ip
elin
es,
wh
er
e
in
s
u
f
f
icien
t f
ea
tu
r
es lea
d
to
m
a
p
p
in
g
c
o
llap
s
e
[
1
6
]
.
Fig
u
r
e
8
.
Failu
r
e
ca
s
es o
f
YO
L
Ov
8
-
p
o
s
e
m
e
d
iu
m
in
b
r
o
ad
c
ast d
ata
d
u
e
to
f
ewe
r
th
an
f
o
u
r
d
etec
ted
k
ey
p
o
in
ts
I
n
s
u
m
m
a
r
y
,
th
r
ee
m
ain
lim
i
tatio
n
s
ca
n
b
e
id
e
n
tifie
d
b
ased
o
n
o
u
r
f
in
d
in
g
s
,
alo
n
g
wi
th
s
ev
er
al
p
o
ten
tial
s
o
lu
tio
n
s
.
First,
th
e
m
ed
iu
m
p
e
r
f
o
r
m
s
as
th
e
b
est
m
o
d
el
in
o
u
r
lim
ited
d
ataset
with
f
ix
ed
in
p
u
t
s
ize
(
6
4
0
×
6
4
0
)
.
I
n
p
r
ac
tice,
th
e
v
ar
iety
o
f
ca
m
e
r
a
an
g
les
ca
n
b
e
m
o
r
e
d
i
v
er
s
e,
s
o
en
lar
g
in
g
t
h
is
d
ataset
s
ize
with
d
iv
er
s
e
in
p
u
t
r
eso
lu
tio
n
is
n
ec
ess
ar
y
to
g
et
m
o
r
e
r
o
b
u
s
t
m
o
d
el.
T
h
en
,
th
e
ap
p
licab
ilit
y
o
f
th
is
ap
p
r
o
ac
h
m
ay
also
b
e
f
u
r
th
e
r
ev
alu
ated
wit
h
m
u
lti
-
ca
m
er
a
s
etu
p
s
o
r
in
o
th
er
s
p
o
r
ts
d
o
m
ain
s
,
lik
e
b
ask
etb
all.
Seco
n
d
,
k
ey
p
o
i
n
t
d
en
s
ity
in
th
e
v
is
ib
le
ar
ea
is
an
o
th
e
r
b
o
ttlen
ec
k
i
n
b
r
o
ad
ca
s
t
f
r
am
es.
E
v
en
s
u
ch
esti
m
ato
r
s
,
lik
e
DL
T
+
R
ANSA
C
,
ca
n
n
o
t
f
in
d
a
s
tab
le
h
o
m
o
g
r
a
p
h
y
with
o
u
t
s
u
f
f
icien
t
d
etec
tio
n
.
T
o
s
o
lv
e
th
is
,
th
e
d
ataset
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
E
va
lu
a
tin
g
YOLOv8
-
p
o
s
e
va
r
ia
n
ts
fo
r
s
o
cc
er k
ey
p
o
in
t d
etec
t
io
n
a
n
d
h
o
m
o
g
r
a
p
h
y
…
(
Du
r
r
o
tu
n
N
a
s
h
ih
in
)
4827
s
h
o
u
ld
b
e
im
p
r
o
v
e
d
b
y
ad
d
i
n
g
m
o
r
e
k
ey
p
o
in
ts
to
en
h
a
n
ce
h
o
m
o
g
r
a
p
h
y
s
tab
ilit
y
.
Ad
d
itio
n
ally
,
tem
p
o
r
al
s
m
o
o
th
in
g
b
etwe
en
f
r
am
es
b
ased
o
n
th
e
a
d
jace
n
t
f
r
am
e
in
f
o
r
m
atio
n
is
also
n
ec
ess
ar
y
to
r
ec
o
n
s
tr
u
ct
lo
s
t
k
ey
p
o
i
n
ts
[
34
]
.
T
h
ir
d
,
ac
cu
r
ate
k
ey
p
o
in
t
d
etec
tio
n
wit
h
s
tab
le
h
o
m
o
g
r
ap
h
y
m
a
p
p
in
g
ca
n
n
o
t
s
tan
d
in
d
ep
en
d
en
tly
with
o
u
t
p
la
y
er
tr
ac
k
in
g
.
T
h
u
s
,
in
teg
r
ati
n
g
it
with
p
lay
er
tr
ac
k
in
g
is
m
an
d
ato
r
y
to
e
n
ab
le
co
m
p
r
eh
e
n
s
iv
e
tactica
l a
n
aly
s
is
.
5.
CO
NCLU
SI
O
N
T
h
is
p
ap
er
p
r
esen
ts
a
s
y
s
tem
atic
ev
alu
atio
n
o
f
d
if
f
er
e
n
t
YOL
Ov
8
-
p
o
s
e
m
o
d
el
v
ar
ia
n
ts
f
o
r
k
ey
p
o
i
n
t
d
etec
tio
n
an
d
h
o
m
o
g
r
a
p
h
y
m
ap
p
in
g
u
s
in
g
b
r
o
ad
ca
s
t
an
d
s
co
u
tin
g
d
atasets
.
T
h
e
r
esu
lts
d
em
o
n
s
tr
ate
th
at
th
e
m
ed
iu
m
YOL
Ov
8
-
p
o
s
e
v
ar
ian
t
p
r
o
v
id
es
th
e
b
est
tr
ad
e
-
o
f
f
b
ala
n
ce
b
etwe
en
d
e
tectio
n
ac
cu
r
ac
y
,
co
m
p
u
tatio
n
al
ef
f
icien
cy
,
an
d
h
o
m
o
g
r
ap
h
y
m
ap
p
in
g
r
eliab
i
lity
.
C
o
m
p
ar
ed
to
o
th
er
v
a
r
ian
ts
,
it
ac
h
iev
es
th
e
b
est
d
etec
tio
n
ac
c
u
r
ac
y
,
wh
ile
m
ain
tain
in
g
r
ea
l
-
tim
e
in
f
er
e
n
ce
s
p
ee
d
a
n
d
m
o
d
el
co
m
p
le
x
ity
.
Fu
r
th
e
r
m
o
r
e
,
th
e
m
ed
iu
m
v
ar
ian
t
c
o
m
b
in
e
d
wi
th
DL
T
an
d
R
ANSAC
en
ab
les
r
o
b
u
s
t
h
o
m
o
g
r
ap
h
y
m
a
p
p
i
n
g
,
n
o
t
o
n
ly
u
n
d
e
r
s
co
u
tin
g
d
ata
b
u
t
also
b
r
o
a
d
c
ast
d
ata,
g
e
n
er
atin
g
m
o
r
e
d
iv
er
s
e
ca
m
er
a
an
g
les.
T
h
u
s
,
th
i
s
m
o
d
el
o
f
f
e
r
s
th
e
m
o
s
t
p
r
o
m
is
in
g
an
d
p
r
ac
tical
s
o
lu
tio
n
f
o
r
r
ea
l
-
tim
e
s
o
cc
er
a
n
aly
tics
.
T
o
th
e
b
est
o
f
o
u
r
k
n
o
wled
g
e,
th
is
wo
r
k
p
r
o
v
id
es
t
h
e
f
ir
s
t
r
ep
r
o
d
u
cib
le
b
en
ch
m
ar
k
o
f
d
if
f
e
r
en
t
YOL
Ov
8
-
p
o
s
e
v
a
r
ian
ts
f
o
r
s
o
cc
er
h
o
m
o
g
r
a
p
h
y
m
ap
p
in
g
.
Fin
ally
,
f
u
tu
r
e
wo
r
k
ca
n
f
o
c
u
s
o
n
u
tili
zin
g
l
ar
g
er
d
atasets
with
ad
d
itio
n
al
k
ey
p
o
in
t
lab
els,
in
co
r
p
o
r
atin
g
tem
p
o
r
al
s
m
o
o
t
h
in
g
,
an
d
in
teg
r
atin
g
a
tr
ac
k
i
n
g
s
y
s
tem
to
p
r
o
d
u
ce
m
o
r
e
r
o
b
u
s
t
an
d
c
o
m
p
lete
r
ea
l
-
tim
e
p
lay
er
h
o
m
o
g
r
ap
h
y
m
ap
p
in
g
.
F
UNDING
I
NF
O
R
M
A
T
I
O
N
T
h
is
r
esear
ch
was
f
u
n
d
e
d
b
y
a
n
o
n
-
APB
N
g
r
an
t
f
r
o
m
State
Un
iv
er
s
ity
o
f
Su
r
a
b
ay
a
u
n
d
e
r
C
o
n
tr
ac
t
No
.
B
/4
0
0
0
6
/UN3
8
.
I
I
I
.
1
/LK
.
0
4
.
0
0
/2
0
2
5
.
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
Du
r
r
o
tu
n
Nash
ih
in
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
Yo
h
an
es Yo
h
an
ie
Frid
elin
Pan
d
u
m
an
✓
✓
✓
✓
✓
Dwi
C
ah
y
o
Kar
tik
o
✓
✓
✓
An
n
a
No
o
r
d
ia
✓
✓
✓
✓
✓
Kh
o
ir
u
l I
s
lam
✓
✓
✓
✓
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
T
h
e
au
th
o
r
s
d
ec
lar
e
n
o
co
n
f
lict o
f
in
ter
est.
DATA AV
AI
L
AB
I
L
I
T
Y
T
h
e
d
ataset
with
c
o
d
es
is
o
p
en
ly
av
ailab
le
an
d
ca
n
b
e
ac
ce
s
s
ed
th
r
o
u
g
h
th
e
Gith
u
b
r
e
p
o
s
ito
r
y
at
h
ttp
s
://g
ith
u
b
.
co
m
/d
u
r
r
o
t
u
n
n
a
s
h
ih
in
/y
o
lo
v
8
p
o
s
e
-
s
o
cc
er
-
k
ey
p
o
in
t
-
h
o
m
o
g
r
ap
h
y
-
m
ap
p
in
g
.
RE
F
E
R
E
NC
E
S
[
1
]
A
.
D
e
l
i
e
g
e
e
t
a
l
.
,
“
S
o
c
c
e
r
N
e
t
-
v
2
:
a
d
a
t
a
se
t
a
n
d
b
e
n
c
h
mar
k
s
f
o
r
h
o
l
i
s
t
i
c
u
n
d
e
r
s
t
a
n
d
i
n
g
o
f
b
r
o
a
d
c
a
st
s
o
c
c
e
r
v
i
d
e
o
s
,
”
i
n
2
0
2
1
I
EEE/
C
VF
C
o
n
f
e
re
n
c
e
o
n
C
o
m
p
u
t
e
r
V
i
si
o
n
a
n
d
Pa
t
t
e
rn
R
e
c
o
g
n
i
t
i
o
n
W
o
rks
h
o
p
s
(
C
VPR
W)
,
Ju
n
.
2
0
2
1
,
p
p
.
4
5
0
3
–
4
5
1
4
,
d
o
i
:
1
0
.
1
1
0
9
/
C
V
P
R
W
5
3
0
9
8
.
2
0
2
1
.
0
0
5
0
8
.
[
2
]
S
.
A
k
a
n
a
n
d
S
.
V
a
r
l
ı
,
“
U
se
o
f
d
e
e
p
l
e
a
r
n
i
n
g
i
n
s
o
c
c
e
r
v
i
d
e
o
a
n
a
l
y
si
s
:
s
u
r
v
e
y
,
”
Mu
l
t
i
m
e
d
i
a
S
y
s
t
e
m
s
,
v
o
l
.
2
9
,
n
o
.
3
,
p
p
.
8
9
7
–
9
1
5
,
Ju
n
.
2
0
2
3
,
d
o
i
:
1
0
.
1
0
0
7
/
s0
0
5
3
0
-
0
2
2
-
0
1
0
2
7
-
0.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
1
5
,
No
.
5
,
Octo
b
er
2
0
2
6
:
4
8
1
9
-
4
8
2
9
4828
[
3
]
F
.
M
a
g
e
r
a
,
T.
H
o
y
o
u
x
,
O
.
B
a
r
n
i
c
h
,
a
n
d
M
.
V
.
D
r
o
o
g
e
n
b
r
o
e
c
k
,
“
B
r
o
a
d
T
r
a
c
k
:
b
r
o
a
d
c
a
s
t
c
a
mera
t
r
a
c
k
i
n
g
f
o
r
so
c
c
e
r
,
”
i
n
2
0
2
5
I
EEE/
C
VF
W
i
n
t
e
r
C
o
n
f
e
re
n
c
e
o
n
A
p
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)
,
F
e
b
.
2
0
2
5
,
p
p
.
6
1
7
7
–
6
1
8
7
,
d
o
i
:
1
0
.
1
1
0
9
/
W
A
C
V
6
1
0
4
1
.
2
0
2
5
.
0
0
6
0
2
.
[
4
]
H.
-
Y
.
W
a
n
g
,
P
.
-
Y
.
C
h
o
u
,
Y
.
-
Y
.
K
a
o
,
a
n
d
C
.
-
H
.
L
i
n
,
“
G
l
o
F
A
N
e
t
:
g
l
o
b
a
l
f
e
a
t
u
r
e
a
u
g
me
n
t
e
d
n
e
t
w
o
r
k
s
f
o
r
d
a
t
a
-
e
f
f
i
c
i
e
n
t
so
c
c
e
r
f
i
e
l
d
k
e
y
p
o
i
n
t
d
e
t
e
c
t
i
o
n
,
”
i
n
2
0
2
5
I
EEE
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
n
s
u
m
e
r
E
l
e
c
t
r
o
n
i
c
s
(
I
C
C
E)
,
Ja
n
.
2
0
2
5
,
p
p
.
1
–
2
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
C
E6
3
6
4
7
.
2
0
2
5
.
1
0
9
3
0
1
6
2
.
[
5
]
M
.
M
a
n
a
f
i
f
a
r
d
,
“
A
r
e
v
i
e
w
o
n
c
a
m
e
r
a
c
a
l
i
b
r
a
t
i
o
n
i
n
so
c
c
e
r
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
.
6
,
p
p
.
1
8
4
2
7
–
1
8
4
5
8
,
J
u
l
.
2
0
2
3
,
d
o
i
:
1
0
.
1
0
0
7
/
s1
1
0
4
2
-
023
-
1
6
1
4
5
-
8.
[
6
]
N
.
S
.
F
a
l
a
l
e
e
v
a
n
d
R
.
C
h
e
n
,
“
E
n
h
a
n
c
i
n
g
s
o
c
c
e
r
c
a
m
e
r
a
c
a
l
i
b
r
a
t
i
o
n
t
h
r
o
u
g
h
k
e
y
p
o
i
n
t
e
x
p
l
o
i
t
a
t
i
o
n
,
”
i
n
Pr
o
c
e
e
d
i
n
g
s
o
f
t
h
e
7
t
h
A
C
M
I
n
t
e
r
n
a
t
i
o
n
a
l
W
o
rks
h
o
p
o
n
Mu
l
t
i
m
e
d
i
a
C
o
n
t
e
n
t
A
n
a
l
y
si
s
i
n
S
p
o
r
t
s
,
O
c
t
.
2
0
2
4
,
p
p
.
6
5
–
73
,
d
o
i
:
1
0
.
1
1
4
5
/
3
6
8
9
0
6
1
.
3
6
8
9
0
7
4
.
[
7
]
H
.
A
r
d
ö
e
t
a
l
.
,
“
S
p
i
i
d
e
o
S
o
c
c
e
r
N
e
t
S
y
n
Lo
c
:
si
n
g
l
e
f
r
a
me
w
o
r
l
d
c
o
o
r
d
i
n
a
t
e
a
t
h
l
e
t
e
d
e
t
e
c
t
i
o
n
a
n
d
l
o
c
a
l
i
z
a
t
i
o
n
w
i
t
h
s
y
n
t
h
e
t
i
c
d
a
t
a
,”
i
n
Pr
o
c
e
e
d
i
n
g
s
o
f
t
h
e
2
0
t
h
I
n
t
e
rn
a
t
i
o
n
a
l
J
o
i
n
t
C
o
n
f
e
re
n
c
e
o
n
C
o
m
p
u
t
e
r
Vi
si
o
n
,
I
m
a
g
i
n
g
a
n
d
C
o
m
p
u
t
e
r
G
ra
p
h
i
c
s
T
h
e
o
ry
a
n
d
Ap
p
l
i
c
a
t
i
o
n
s
,
2
0
2
5
,
p
p
.
2
7
8
–
285
,
d
o
i
:
1
0
.
5
2
2
0
/
0
0
1
3
1
0
8
2
0
0
0
0
3
9
1
2
.
[
8
]
U
l
t
r
a
l
y
t
i
c
s,
“
E
x
p
l
o
r
e
u
l
t
r
a
l
y
t
i
c
s
Y
O
LO
v
8
,
”
d
o
c
s
.
u
l
t
r
a
l
y
t
i
c
s
.
c
o
m.
A
c
c
e
ss
e
d
:
A
u
g
.
1
5
,
2
0
2
5
.
[
O
n
l
i
n
e
]
.
A
v
a
i
l
a
b
l
e
:
h
t
t
p
s
:
/
/
d
o
c
s.
u
l
t
r
a
l
y
t
i
c
s
.
c
o
m/
m
o
d
e
l
s/
y
o
l
o
v
8
/
[
9
]
U
l
t
r
a
l
y
t
i
c
s,
“
P
o
s
e
e
st
i
ma
t
i
o
n
w
i
t
h
u
l
t
r
a
l
y
t
i
c
s
Y
O
LO
,
”
d
o
c
s
.
u
l
t
r
a
l
y
t
i
c
s
.
c
o
m.
A
c
c
e
sse
d
:
A
u
g
.
1
5
,
2
0
2
5
.
[
O
n
l
i
n
e
]
.
A
v
a
i
l
a
b
l
e
:
h
t
t
p
s
:
/
/
d
o
c
s.
u
l
t
r
a
l
y
t
i
c
s
.
c
o
m/
t
a
s
k
s/
p
o
se
/
[
1
0
]
P
.
S
k
a
l
sk
i
,
“
C
a
mera
c
a
l
i
b
r
a
t
i
o
n
i
n
sp
o
r
t
s
w
i
t
h
k
e
y
p
o
i
n
t
s,”
b
l
o
g
.
r
o
b
o
f
l
o
w
.
c
o
m
,
2
0
2
4
.
A
c
c
e
sse
d
:
A
u
g
.
1
5
,
2
0
2
5
.
[
O
n
l
i
n
e
]
.
A
v
a
i
l
a
b
l
e
:
h
t
t
p
s:
/
/
b
l
o
g
.
r
o
b
o
f
l
o
w
.
c
o
m
/
c
a
mer
a
-
c
a
l
i
b
r
a
t
i
o
n
-
s
p
o
r
t
s
-
c
o
mp
u
t
e
r
-
v
i
si
o
n
/
[
1
1
]
P
.
S
.
S
.
M
a
r
k
a
p
p
a
,
C
.
O
’
L
e
a
r
y
,
a
n
d
C
.
L
y
n
c
h
,
“
A
r
e
v
i
e
w
o
f
Y
O
LO
mo
d
e
l
s
f
o
r
so
c
c
e
r
-
b
a
s
e
d
o
b
j
e
c
t
d
e
t
e
c
t
i
o
n
,
”
i
n
2
0
2
4
S
i
x
t
h
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
r
e
n
c
e
o
n
I
n
t
e
l
l
i
g
e
n
t
C
o
m
p
u
t
i
n
g
i
n
D
a
t
a
S
c
i
e
n
c
e
s
(
I
C
D
S
)
,
O
c
t
.
2
0
2
4
,
p
p
.
1
–
7
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
D
S
6
2
0
8
9
.
2
0
2
4
.
1
0
7
5
6
4
4
3
.
[
1
2
]
G
.
N
o
u
si
a
s,
K
.
K
.
D
e
l
i
b
a
s
i
s,
a
n
d
I
.
G
.
M
a
g
l
o
g
i
a
n
n
i
s,
“
I
n
t
e
l
l
i
g
e
n
t
sam
p
l
i
n
g
c
o
n
se
n
su
s
f
o
r
h
o
m
o
g
r
a
p
h
y
e
st
i
m
a
t
i
o
n
i
n
f
o
o
t
b
a
l
l
v
i
d
e
o
s
u
si
n
g
f
e
a
t
u
r
e
l
e
ss
u
n
p
a
i
r
e
d
p
o
i
n
t
s
,
”
I
EEE
Ac
c
e
ss
,
v
o
l
.
1
3
,
p
p
.
1
8
7
8
4
3
–
1
8
7
8
5
7
,
2
0
2
5
,
d
o
i
:
1
0
.
1
1
0
9
/
A
C
C
ES
S
.
2
0
2
5
.
3
6
2
7
5
3
8
.
[
1
3
]
K
.
J
a
n
n
e
t
,
M
.
S
.
S
u
n
a
r
,
M
.
M
.
I
sl
a
m
M
o
l
l
a
,
a
n
d
M
.
A
.
A
s’
A
r
i
,
“
A
d
e
e
p
l
e
a
r
n
i
n
g
a
p
p
r
o
a
c
h
t
o
b
a
d
mi
n
t
o
n
p
l
a
y
e
r
f
o
o
t
w
o
r
k
d
e
t
e
c
t
i
o
n
b
a
s
e
d
o
n
Y
O
LO
mo
d
e
l
s
:
a
c
o
mp
a
r
a
t
i
v
e
st
u
d
y
,
”
i
n
2
0
2
4
I
EE
E
8
t
h
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
r
e
n
c
e
o
n
S
i
g
n
a
l
a
n
d
I
m
a
g
e
Pro
c
e
ss
i
n
g
Ap
p
l
i
c
a
t
i
o
n
s (I
C
S
I
PA)
,
S
e
p
.
2
0
2
4
,
p
p
.
1
–
6
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
S
I
P
A
6
2
0
6
1
.
2
0
2
4
.
1
0
6
8
6
5
3
7
.
[
1
4
]
M
.
A
t
i
k
,
Z.
D
u
r
a
n
,
a
n
d
R
.
Ö
z
g
ü
n
l
ü
k
,
“
C
o
m
p
a
r
i
s
o
n
o
f
Y
O
LO
v
e
r
si
o
n
s
f
o
r
o
b
j
e
c
t
d
e
t
e
c
t
i
o
n
f
r
o
m
a
e
r
i
a
l
i
ma
g
e
s
,”
I
n
t
e
r
n
a
t
i
o
n
a
l
J
o
u
rn
a
l
o
f
En
v
i
r
o
n
m
e
n
t
a
n
d
G
e
o
i
n
f
o
r
m
a
t
i
c
s
,
v
o
l
.
9
,
n
o
.
2
,
p
p
.
8
7
–
9
3
,
Ju
n
.
2
0
2
2
,
d
o
i
:
1
0
.
3
0
8
9
7
/
i
j
e
g
e
o
.
1
0
1
0
7
4
1
.
[
1
5
]
D
.
B
e
r
j
ó
n
,
C
.
C
u
e
v
a
s,
a
n
d
N
.
G
a
r
c
í
a
,
“
S
o
c
c
e
r
l
i
n
e
mark
se
g
me
n
t
a
t
i
o
n
a
n
d
c
l
a
ss
i
f
i
c
a
t
i
o
n
w
i
t
h
st
o
c
h
a
s
t
i
c
w
a
t
e
r
sh
e
d
t
r
a
n
sf
o
r
m,”
S
i
g
n
a
l
Pr
o
c
e
ssi
n
g
:
I
m
a
g
e
C
o
m
m
u
n
i
c
a
t
i
o
n
,
v
o
l
.
1
1
8
,
O
c
t
.
2
0
2
3
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
i
ma
g
e
.
2
0
2
3
.
1
1
7
0
1
4
.
[
1
6
]
M
.
G
.
-
P
é
r
e
z
a
n
d
A
.
A
g
u
d
o
,
“
P
n
LC
a
l
i
b
:
sp
o
r
t
s
f
i
e
l
d
r
e
g
i
st
r
a
t
i
o
n
v
i
a
p
o
i
n
t
s
a
n
d
l
i
n
e
s
o
p
t
i
mi
z
a
t
i
o
n
,”
C
o
m
p
u
t
e
r
V
i
si
o
n
a
n
d
I
m
a
g
e
U
n
d
e
rst
a
n
d
i
n
g
,
v
o
l
.
2
6
7
,
A
p
r
.
2
0
2
6
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
c
v
i
u
.
2
0
2
6
.
1
0
4
7
1
2
.
[
1
7
]
S
.
A
.
G
h
a
se
mza
d
e
h
,
G
.
V
.
Za
n
d
y
c
k
e
,
M
.
I
st
a
sse,
N
.
S
a
y
e
z
,
A
.
M
o
sh
t
a
g
h
p
o
u
r
,
a
n
d
C
.
D
.
V
l
e
e
sc
h
o
u
w
e
r
,
“
D
e
e
p
S
p
o
r
t
L
a
b
:
a
u
n
i
f
i
e
d
f
r
a
mew
o
r
k
f
o
r
b
a
l
l
d
e
t
e
c
t
i
o
n
,
p
l
a
y
e
r
i
n
s
t
a
n
c
e
s
e
g
m
e
n
t
a
t
i
o
n
a
n
d
p
o
s
e
e
s
t
i
mat
i
o
n
i
n
t
e
a
m
s
p
o
r
t
s
s
c
e
n
e
s
,
”
i
n
Pr
o
c
e
e
d
i
n
g
s
o
f
t
h
e
Bri
t
i
sh
M
a
c
h
i
n
e
V
i
si
o
n
C
o
n
f
e
re
n
c
e
2
0
2
1
,
2
0
2
1
,
d
o
i
:
1
0
.
5
2
4
4
/
C
.
3
5
.
2
1
0
.
[
1
8
]
B
.
M
.
A
.
,
S
.
V
e
n
k
a
t
a
r
a
ma
n
,
M
.
S
.
K
h
a
n
,
N
i
d
h
i
,
S
.
P
a
d
m
a
n
a
b
h
u
n
i
,
a
n
d
S
.
N
a
t
a
r
a
j
a
n
,
“
O
p
t
i
mi
z
i
n
g
se
ma
n
t
i
c
se
g
me
n
t
a
t
i
o
n
f
o
r
e
n
h
a
n
c
e
d
f
o
o
t
b
a
l
l
a
n
a
l
y
t
i
c
s:
a
p
i
x
e
l
-
l
e
v
e
l
a
p
p
r
o
a
c
h
,
”
Pr
o
c
e
d
i
a
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
,
v
o
l
.
2
3
5
,
p
p
.
2
6
6
2
–
2
6
7
3
,
2
0
2
4
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
p
r
o
c
s.
2
0
2
4
.
0
4
.
2
5
1
.
[
1
9
]
B
.
R
e
n
,
X
.
Zh
e
n
g
,
T.
G
u
a
n
,
a
n
d
J
.
W
a
n
g
,
“
V
i
b
r
a
t
o
r
r
a
c
k
p
o
se
e
st
i
ma
t
i
o
n
f
o
r
m
o
n
i
t
o
r
i
n
g
t
h
e
v
i
b
r
a
t
i
o
n
q
u
a
l
i
t
y
o
f
c
o
n
c
r
e
t
e
u
si
n
g
i
mp
r
o
v
e
d
Y
O
LO
v
8
-
p
o
s
e
a
n
d
v
a
n
i
s
h
i
n
g
p
o
i
n
t
s
,
”
B
u
i
l
d
i
n
g
s
,
v
o
l
.
1
4
,
n
o
.
1
0
,
O
c
t
.
2
0
2
4
,
d
o
i
:
1
0
.
3
3
9
0
/
b
u
i
l
d
i
n
g
s
1
4
1
0
3
1
7
4
.
[
2
0
]
C
.
D
o
n
g
a
n
d
G
.
D
u
,
“
A
n
e
n
h
a
n
c
e
d
r
e
a
l
-
t
i
me
h
u
ma
n
p
o
se
e
s
t
i
m
a
t
i
o
n
m
e
t
h
o
d
b
a
s
e
d
o
n
mo
d
i
f
i
e
d
Y
O
LO
v
8
f
r
a
me
w
o
r
k
,
”
S
c
i
e
n
t
i
f
i
c
Re
p
o
r
t
s
,
v
o
l
.
1
4
,
n
o
.
1
,
A
p
r
.
2
0
2
4
,
d
o
i
:
1
0
.
1
0
3
8
/
s4
1
5
9
8
-
024
-
5
8
1
4
6
-
z.
[
2
1
]
A
d
z
u
k
i
,
“
S
o
c
c
e
r
-
f
i
e
l
d
-
l
o
c
a
l
i
z
a
t
i
o
n
c
o
mp
u
t
e
r
v
i
si
o
n
m
o
d
e
l
,
”
u
n
i
v
e
r
se
.
r
o
b
o
f
l
o
w
.
c
o
m
.
A
c
c
e
sse
d
:
A
u
g
.
1
5
,
2
0
2
5
.
[
O
n
l
i
n
e
]
.
A
v
a
i
l
a
b
l
e
:
h
t
t
p
s:
/
/
u
n
i
v
e
r
s
e
.
r
o
b
o
f
l
o
w
.
c
o
m/
a
d
z
u
k
i
/
s
o
c
c
e
r
-
f
i
e
l
d
-
l
o
c
a
l
i
z
a
t
i
o
n
[
2
2
]
Y
.
T
i
a
n
,
Y
.
Zh
a
n
g
,
a
n
d
H
.
Zh
a
n
g
,
“
R
e
c
e
n
t
a
d
v
a
n
c
e
s
i
n
s
t
o
c
h
a
st
i
c
g
r
a
d
i
e
n
t
d
e
s
c
e
n
t
i
n
d
e
e
p
l
e
a
r
n
i
n
g
,
”
Ma
t
h
e
m
a
t
i
c
s
,
v
o
l
.
1
1
,
n
o
.
3
,
Jan
.
2
0
2
3
,
d
o
i
:
1
0
.
3
3
9
0
/
ma
t
h
1
1
0
3
0
6
8
2
.
[
2
3
]
Y
.
B
a
i
,
e
t
a
l
.
,
“
H
o
w
i
m
p
o
r
t
a
n
t
i
s
t
h
e
t
r
a
i
n
-
v
a
l
i
d
a
t
i
o
n
s
p
l
i
t
i
n
me
t
a
-
l
e
a
r
n
i
n
g
?
”
i
n
Pr
o
c
e
e
d
i
n
g
s
o
f
t
h
e
3
8
t
h
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
M
a
c
h
i
n
e
L
e
a
rn
i
n
g
(
I
C
ML)
-
PM
L
R
1
3
9
,
2
0
2
1
,
p
p
.
5
4
3
–
5
5
3
.
[
2
4
]
Z.
J
i
a
n
g
,
H
.
Ji
,
C
.
-
Y
.
Y
a
n
g
,
a
n
d
J
.
-
N
.
H
w
a
n
g
,
“
2
D
h
u
ma
n
p
o
se
e
s
t
i
m
a
t
i
o
n
c
a
l
i
b
r
a
t
i
o
n
a
n
d
k
e
y
p
o
i
n
t
v
i
s
i
b
i
l
i
t
y
c
l
a
ss
i
f
i
c
a
t
i
o
n
,
”
i
n
I
C
AS
S
P
2
0
2
4
-
2
0
2
4
I
E
EE
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
r
e
n
c
e
o
n
A
c
o
u
s
t
i
c
s,
S
p
e
e
c
h
a
n
d
S
i
g
n
a
l
Pr
o
c
e
ss
i
n
g
(
I
C
AS
S
P)
,
A
p
r
.
2
0
2
4
,
p
p
.
6
0
9
5
–
6
0
9
9
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
A
S
S
P
4
8
4
8
5
.
2
0
2
4
.
1
0
4
4
8
4
7
4
.
[
2
5
]
M
.
J
i
o
n
o
,
H
.
-
I
.
Li
n
,
W
.
-
H
.
C
h
e
n
,
a
n
d
S
.
N
a
n
d
a
,
“
6
-
D
o
F
p
o
s
e
e
s
t
i
m
a
t
i
o
n
u
si
n
g
a
n
e
n
d
-
to
-
e
n
d
p
i
p
e
l
i
n
e
o
f
n
e
u
r
a
l
n
e
t
w
o
r
k
w
i
t
h
BP
-
n
-
P
,
”
I
EEE
S
e
n
s
o
rs
J
o
u
r
n
a
l
,
v
o
l
.
2
5
,
n
o
.
2
,
p
p
.
3
8
3
7
–
3
8
4
7
,
Ja
n
.
2
0
2
5
,
d
o
i
:
1
0
.
1
1
0
9
/
JS
EN
.
2
0
2
4
.
3
5
0
5
8
3
2
.
[
2
6
]
X
.
Z
h
a
n
g
,
J
.
B
a
i
,
G
.
Q
i
a
o
,
X
.
X
i
a
o
,
L.
M
e
n
g
,
a
n
d
S
.
H
u
,
“
Y
O
LO
-
f
a
l
l
:
a
Y
O
LO
-
b
a
se
d
f
a
l
l
d
e
t
e
c
t
i
o
n
mo
d
e
l
w
i
t
h
h
i
g
h
p
r
e
c
i
si
o
n
,
sh
r
u
n
k
si
z
e
,
a
n
d
l
o
w
l
a
t
e
n
c
y
,
”
T
h
e
C
o
m
p
u
t
e
r
J
o
u
r
n
a
l
,
v
o
l
.
6
8
,
n
o
.
7
,
p
p
.
8
0
4
–
8
1
2
,
J
u
l
.
2
0
2
5
,
d
o
i
:
1
0
.
1
0
9
3
/
c
o
mj
n
l
/
b
x
a
f
0
0
5
.
[
2
7
]
O
p
e
n
C
V
,
“
B
a
si
c
c
o
n
c
e
p
t
s
o
f
t
h
e
h
o
mo
g
r
a
p
h
y
e
x
p
l
a
i
n
e
d
w
i
t
h
c
o
d
e
,
”
d
o
c
s.
o
p
e
n
c
v
.
o
r
g
.
A
c
c
e
sse
d
:
A
u
g
.
1
5
,
2
0
2
5
.
[
O
n
l
i
n
e
]
.
A
v
a
i
l
a
b
l
e
:
h
t
t
p
s:
/
/
d
o
c
s
.
o
p
e
n
c
v
.
o
r
g
/
4
.
1
3
.
0
/
d
9
/
d
a
b
/
t
u
t
o
r
i
a
l
_
h
o
mo
g
r
a
p
h
y
.
h
t
ml
[
2
8
]
W
.
H
e
e
t
a
l
.
,
“
A
r
e
a
l
-
t
i
m
e
a
n
d
h
i
g
h
p
r
e
c
i
s
i
o
n
h
a
r
d
w
a
r
e
i
m
p
l
e
me
n
t
a
t
i
o
n
o
f
R
A
N
S
A
C
a
l
g
o
r
i
t
h
m
f
o
r
v
i
s
u
a
l
S
LA
M
a
c
h
i
e
v
i
n
g
mi
sma
t
c
h
e
d
f
e
a
t
u
r
e
p
o
i
n
t
p
a
i
r
e
l
i
mi
n
a
t
i
o
n
,
”
I
EEE
T
ra
n
s
a
c
t
i
o
n
s
o
n
C
i
rc
u
i
t
s
a
n
d
S
y
st
e
m
s
I
:
R
e
g
u
l
a
r
P
a
p
e
rs
,
v
o
l
.
7
1
,
n
o
.
1
1
,
p
p
.
5
1
0
2
–
5
1
1
4
,
N
o
v
.
2
0
2
4
,
d
o
i
:
1
0
.
1
1
0
9
/
T
C
S
I
.
2
0
2
4
.
3
4
2
2
0
8
2
.
[
2
9
]
M
.
Zh
a
n
g
a
n
d
Z.
Z
h
o
u
,
“
S
p
e
e
d
-
a
c
c
u
r
a
c
y
t
r
a
d
e
-
o
f
f
i
n
l
i
g
h
t
w
e
i
g
h
t
-
b
a
se
d
h
a
n
d
p
o
se
e
st
i
ma
t
i
o
n
,
”
T
h
e
J
o
u
r
n
a
l
o
f
S
u
p
e
rc
o
m
p
u
t
i
n
g
,
v
o
l
.
8
1
,
n
o
.
1
2
,
A
u
g
.
2
0
2
5
,
d
o
i
:
1
0
.
1
0
0
7
/
s
1
1
2
2
7
-
0
2
5
-
0
7
6
9
9
-
5.
[
3
0
]
O
p
e
n
C
V
,
“
C
a
mera
c
a
l
i
b
r
a
t
i
o
n
a
n
d
3
D
r
e
c
o
n
s
t
r
u
c
t
i
o
n
,
”
d
o
c
s.
o
p
e
n
c
v
.
o
r
g
.
A
c
c
e
sse
d
:
A
u
g
.
1
5
,
2
0
2
5
.
[
O
n
l
i
n
e
]
.
A
v
a
i
l
a
b
l
e
:
h
t
t
p
s
:
/
/
d
o
c
s.
o
p
e
n
c
v
.
o
r
g
/
4
.
x
/
d
9
/
d
0
c
/
g
r
o
u
p
_
_
c
a
l
i
b
3
d
.
h
t
ml
[
3
1
]
F
.
K
l
u
g
e
r
a
n
d
B
.
R
o
se
n
h
a
h
n
,
“
P
A
R
S
A
C
:
a
c
c
e
l
e
r
a
t
i
n
g
r
o
b
u
st
m
u
l
t
i
-
m
o
d
e
l
f
i
t
t
i
n
g
w
i
t
h
p
a
r
a
l
l
e
l
s
a
mp
l
e
c
o
n
se
n
su
s
,”
Pro
c
e
e
d
i
n
g
s
o
f
t
h
e
AA
AI
C
o
n
f
e
re
n
c
e
o
n
Art
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
,
v
o
l
.
3
8
,
n
o
.
3
,
p
p
.
2
8
0
4
–
2
8
1
2
,
M
a
r
.
2
0
2
4
,
d
o
i
:
1
0
.
1
6
0
9
/
a
a
a
i
.
v
3
8
i
3
.
2
8
0
6
0
.
[
3
2
]
T.
Z
h
a
n
g
,
Q
.
Li
,
J.
W
e
n
,
a
n
d
C
.
L
.
P
.
C
h
e
n
,
“
En
h
a
n
c
e
me
n
t
a
n
d
o
p
t
i
mi
s
a
t
i
o
n
o
f
h
u
m
a
n
p
o
se
e
s
t
i
mat
i
o
n
w
i
t
h
m
u
l
t
i
-
sc
a
l
e
sp
a
t
i
a
l
a
t
t
e
n
t
i
o
n
a
n
d
a
d
v
e
r
s
a
r
i
a
l
d
a
t
a
a
u
g
m
e
n
t
a
t
i
o
n
,
”
I
n
f
o
rm
a
t
i
o
n
F
u
si
o
n
,
v
o
l
.
1
1
1
,
N
o
v
.
2
0
2
4
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
i
n
f
f
u
s.
2
0
2
4
.
1
0
2
5
2
2
.
[
3
3
]
R
.
B
.
N
e
u
p
a
n
e
,
K
.
L
i
,
a
n
d
T.
F
.
B
o
k
a
,
“
A
s
u
r
v
e
y
o
n
d
e
e
p
3
D
h
u
ma
n
p
o
se
e
s
t
i
m
a
t
i
o
n
,
”
A
rt
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
Re
v
i
e
w
,
v
o
l
.
5
8
,
n
o
.
1
,
N
o
v
.
2
0
2
4
,
d
o
i
:
1
0
.
1
0
0
7
/
s
1
0
4
6
2
-
0
2
4
-
1
1
0
1
9
-
3.
[
3
4
]
Y
.
Zh
a
o
a
n
d
P
.
K
r
ä
h
e
n
b
ü
h
l
,
“
R
e
a
l
-
t
i
me
o
n
l
i
n
e
v
i
d
e
o
d
e
t
e
c
t
i
o
n
w
i
t
h
t
e
m
p
o
r
a
l
sm
o
o
t
h
i
n
g
t
r
a
n
sf
o
r
mers
,
”
i
n
C
o
m
p
u
t
e
r
Vi
s
i
o
n
–
EC
C
V
2
0
2
2
,
2
0
2
2
,
p
p
.
4
8
5
–
5
0
2
,
d
o
i
:
1
0
.
1
0
0
7
/
9
7
8
-
3
-
0
3
1
-
1
9
8
3
0
-
4
_
2
8
.
Evaluation Warning : The document was created with Spire.PDF for Python.