T
E
L
K
O
M
NIKA
T
elec
o
mm
un
ica
t
io
n Co
m
pu
t
i
ng
E
lect
ro
nics
a
nd
Co
ntr
o
l
Vo
l.
24
,
No
.
1
,
Feb
r
u
ar
y
20
26
,
p
p
.
1
26
~1
41
I
SS
N:
1
6
9
3
-
6
9
3
0
,
DOI
: 1
0
.
1
2
9
2
8
/
T
E
L
KOM
NI
K
A
.
v
24
i
1
.
27
448
126
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//jo
u
r
n
a
l.u
a
d
.
a
c.
id
/in
d
ex
.
p
h
p
/TELK
OM
N
I
K
A
Perf
o
r
m
a
nce
enh
a
nce
m
e
nt
o
f
e
m
b
edded obj
ect
det
e
ction v
ia
neura
l hardw
a
re accelera
tion
Alw
in H
a
rt
o
no
L
i
m
a
ra
n,
Ag
un
g
Wica
k
s
o
no
,
P
a
t
a
h H
er
wa
nto
D
e
p
a
r
t
me
n
t
o
f
I
n
f
o
r
mat
i
c
s
En
g
i
n
e
e
r
i
n
g
,
F
a
c
u
l
t
y
o
f
I
n
f
o
r
mat
i
o
n
T
e
c
h
n
o
l
o
g
y
,
S
T
M
I
K
I
n
d
o
n
e
si
a
M
a
n
d
i
r
i
,
B
a
n
d
u
n
g
,
I
n
d
o
n
e
si
a
Art
icle
I
nfo
AB
ST
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Au
g
4
,
2025
R
ev
i
s
ed
Oct
24
2
0
2
5
A
cc
ep
ted
Dec
8
,
2
0
2
5
T
h
is
p
a
p
e
r
p
re
se
n
ts
th
e
f
irst
b
e
n
c
h
m
a
r
k
in
g
o
f
y
o
u
o
n
ly
lo
o
k
o
n
c
e
v
e
rsio
n
1
1
(YO
L
O1
1
)
o
n
t
h
e
Ro
c
k
c
h
i
p
RK3
5
6
6
n
e
u
ra
l
p
r
o
c
e
ss
in
g
u
n
it
(N
P
U)
w
it
h
in
th
e
Ora
n
g
e
P
i
3
B
p
latf
o
rm
.
P
e
rf
o
rm
a
n
c
e
wa
s
c
o
m
p
a
re
d
b
e
tw
e
e
n
th
e
q
u
a
d
-
c
o
re
A
RM
Co
rtex
-
A
5
5
CP
U
a
n
d
th
e
in
teg
ra
ted
NP
U
u
sin
g
th
e
COCO
2
0
1
7
d
a
tas
e
t,
e
v
a
lu
a
ti
n
g
late
n
c
y
,
e
n
e
rg
y
,
a
n
d
a
c
c
u
ra
c
y
.
NP
U
a
c
c
e
lera
ti
o
n
a
c
h
ie
v
e
d
>
8
0
%
late
n
c
y
re
d
u
c
ti
o
n
a
n
d
≈
9
4
%
lo
w
e
r
p
e
r
-
i
n
f
e
re
n
c
e
e
n
e
r
g
y
c
o
n
su
m
p
ti
o
n
,
w
it
h
sp
e
e
d
u
p
o
f
u
p
to
1
6
.
7
×
w
h
il
e
m
a
i
n
tain
i
n
g
a
c
c
u
ra
c
y
w
it
h
in
0
.
0
3
m
e
a
n
a
v
e
ra
g
e
p
re
c
isio
n
(
m
A
P
)
o
f
th
e
b
a
se
li
n
e
.
Av
e
ra
g
e
p
o
we
r
re
m
a
in
e
d
n
e
a
rl
y
c
o
n
sta
n
t
(3
.
6
0
W
c
e
n
tral
p
ro
c
e
ss
in
g
u
n
it
(
CP
U
)
v
s.
3
.
5
9
W
N
P
U),
in
d
ica
ti
n
g
th
a
t
t
h
e
e
ff
ici
e
n
c
y
g
a
in
s
ste
m
f
ro
m
re
d
u
c
e
d
i
n
f
e
re
n
c
e
ti
m
e
ra
th
e
r
th
a
n
lo
w
e
r
w
a
tt
a
g
e
.
L
i
m
it
a
ti
o
n
s
in
c
lu
d
e
d
u
n
sta
b
le
INT
8
q
u
a
n
ti
z
a
ti
o
n
d
u
e
t
o
u
n
su
p
p
o
r
ted
o
p
e
ra
to
rs
a
n
d
c
a
li
b
ra
ti
o
n
-
ra
n
g
e
m
is
m
a
t
c
h
,
a
s
w
e
ll
a
s
m
in
o
r
C
P
U
-
sid
e
o
v
e
rh
e
a
d
in
p
re
p
ro
c
e
ss
in
g
a
n
d
n
o
n
-
m
a
x
i
m
u
m
su
p
p
re
ss
io
n
.
T
h
e
f
in
d
in
g
s
c
o
n
f
irm
th
a
t
th
e
RK3
5
6
6
N
P
U
d
e
li
v
e
rs
su
b
sta
n
ti
a
l
e
ff
icie
n
c
y
g
a
in
s
w
it
h
o
u
t
a
c
c
u
ra
c
y
lo
ss
,
e
n
a
b
li
n
g
c
o
m
p
a
c
t
a
n
d
l
o
w
-
c
o
st
p
l
a
tf
o
r
m
s
to
su
sta
in
m
o
d
e
rn
o
b
jec
t
-
d
e
tec
ti
o
n
w
o
rk
lo
a
d
s.
T
h
is
d
e
m
o
n
stra
tes
th
a
t
a
ff
o
rd
a
b
le
NP
Us
c
a
n
p
ro
v
id
e
re
li
a
b
le,
re
a
l
-
ti
m
e
a
rti
f
icia
l
in
telli
g
e
n
c
e
(
AI
)
in
f
e
re
n
c
e
f
o
r
e
m
b
e
d
d
e
d
v
isio
n
,
i
n
tern
e
t
o
f
th
in
g
s (
Io
T
)
,
a
n
d
r
o
b
o
ti
c
s
a
p
p
li
c
a
ti
o
n
s.
K
ey
w
o
r
d
s
:
E
n
er
g
y
e
f
f
icie
n
c
y
L
o
w
-
laten
c
y
i
m
a
g
e
in
f
er
en
ce
Or
an
g
e
P
i 3
B
R
K3
5
6
6
NP
U
YOL
O1
1
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
:
A
l
w
in
Har
to
n
o
L
i
m
ar
an
Dep
ar
t
m
en
t o
f
I
n
f
o
r
m
atic
s
E
n
g
in
ee
r
i
n
g
,
Facu
lt
y
o
f
I
n
f
o
r
m
at
io
n
T
ec
h
n
o
lo
g
y
ST
MI
K
I
n
d
o
n
esia M
an
d
ir
i
B
elitu
n
g
Stre
et
No
.
7
,
Me
r
d
ek
a,
Su
m
u
r
B
an
d
u
n
g
,
B
an
d
u
n
g
,
W
est J
av
a,
I
n
d
o
n
esia
E
m
ail:
al
w
i
n
.
h
ar
to
n
o
@
p
r
o
to
n
m
ail.
co
m
1.
I
NT
RO
D
UCT
I
O
N
Ob
j
ec
t d
etec
tio
n
h
as b
ec
o
m
e
a
ce
n
tr
al
co
m
p
o
n
en
t o
f
m
o
d
er
n
co
m
p
u
ter
v
is
io
n
,
p
o
w
er
i
n
g
ap
p
licatio
n
s
in
s
u
r
v
eil
lan
ce
,
r
o
b
o
tics
,
au
t
o
n
o
m
o
u
s
v
e
h
icles,
a
n
d
th
e
i
n
ter
n
et
o
f
th
i
n
g
s
(
I
o
T
)
[
1
]
.
Am
o
n
g
t
h
e
av
ailab
le
ap
p
r
o
ac
h
es,
th
e
y
o
u
o
n
l
y
lo
o
k
o
n
ce
(
YO
L
O)
alg
o
r
it
h
m
[
2
]
is
w
id
el
y
u
s
ed
b
ec
au
s
e
it
b
ala
n
ce
s
ac
c
u
r
ac
y
w
i
t
h
co
m
p
u
tatio
n
al
ef
f
icie
n
c
y
.
U
n
lik
e
r
eg
io
n
p
r
o
p
o
s
a
l
-
b
ased
m
et
h
o
d
s
th
at
o
p
er
ate
in
m
u
l
tip
le
s
tag
es,
YO
L
O
p
r
o
ce
s
s
es
an
en
t
ir
e
i
m
ag
e
in
a
s
i
n
g
le
p
ass
[
3
]
,
[
4
]
,
s
i
m
u
lta
n
eo
u
s
l
y
p
r
ed
ictin
g
b
o
u
n
d
i
n
g
b
o
x
es,
co
n
f
id
en
ce
s
co
r
es,
an
d
class
lab
els.
T
h
is
d
esig
n
en
ab
le
s
lo
w
-
la
ten
c
y
i
n
f
e
r
en
ce
,
ev
en
o
n
r
eso
u
r
ce
-
li
m
ite
d
p
latf
o
r
m
s
.
YOL
O1
1
e
x
ten
d
s
th
is
li
n
e
o
f
d
ev
elo
p
m
en
t
w
it
h
s
e
v
er
al
i
m
p
r
o
v
e
m
e
n
ts
.
I
t
i
n
tr
o
d
u
ce
s
f
as
ter
in
f
er
en
c
e
o
p
tim
ized
f
o
r
lo
w
-
p
o
w
er
d
ev
i
ce
s
,
in
teg
r
ate
s
g
lo
b
al
s
ce
n
e
a
w
ar
en
ess
f
o
r
m
o
r
e
p
r
ec
is
e
b
o
u
n
d
in
g
b
o
x
p
r
ed
ictio
n
s
[
5
]
,
[
6
]
,
an
d
o
f
f
er
s
b
u
ilt
-
i
n
s
u
p
p
o
r
t
f
o
r
em
b
ed
d
ed
ac
ce
ler
atio
n
o
n
n
e
u
r
al
p
r
o
ce
s
s
in
g
u
n
its
(
NP
Us)
[
7
]
,
[
8
]
.
T
h
ese
f
ea
t
u
r
es
p
o
s
iti
o
n
YO
L
O1
1
as
a
s
tr
o
n
g
ca
n
d
id
ate
f
o
r
d
ep
lo
y
m
e
n
t
in
ed
g
e
e
n
v
ir
o
n
m
e
n
ts
w
h
er
e
late
n
c
y
a
n
d
en
er
g
y
e
f
f
icien
c
y
ar
e
cr
itica
l.
T
h
e
p
r
esen
t
s
t
u
d
y
r
estricte
d
its
s
co
p
e
to
lo
w
-
late
n
c
y
,
p
er
-
i
m
ag
e
in
f
er
e
n
ce
o
n
s
tatic
i
m
a
g
es
a
n
d
is
o
lated
co
m
p
u
te
-
b
ac
k
en
d
e
f
f
ec
ts
(
ce
n
tr
al
p
r
o
ce
s
s
in
g
u
n
it
(
C
P
U
)
v
s
.
R
K
3
5
6
6
N
P
U)
o
n
p
er
-
i
m
a
g
e
laten
c
y
a
n
d
en
er
g
y
at
m
atch
ed
ac
cu
r
ac
y
;
s
e
q
u
e
n
tial o
r
te
m
p
o
r
al
p
r
o
ce
s
s
in
g
w
a
s
ex
cl
u
d
ed
.
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KOM
NI
K
A
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l
C
o
n
tr
o
l
P
erfo
r
ma
n
ce
en
h
a
n
ce
men
t o
f
emb
ed
d
ed
o
b
ject
d
etec
tio
n
via
n
eu
r
a
l h
a
r
d
w
a
r
e
..
(
A
lw
in
Ha
r
t
o
n
o
Lima
r
a
n
)
127
P
r
ev
io
u
s
s
t
u
d
ies
h
a
v
e
s
u
r
v
e
y
e
d
YOL
O
v
er
s
io
n
s
an
d
d
e
m
o
n
s
tr
ated
th
e
p
o
ten
tial
o
f
ea
r
lier
m
o
d
el
s
o
n
af
f
o
r
d
ab
le
p
latf
o
r
m
s
s
u
c
h
a
s
t
h
e
R
asp
b
er
r
y
P
i
[
9
]
,
[
1
0
]
an
d
NVI
DI
A
J
etso
n
Nan
o
[
1
1
]
.
A
li
an
d
Z
h
a
n
g
[
1
2
]
p
r
o
v
id
ed
a
co
m
p
ar
ativ
e
r
ev
ie
w
o
f
YO
L
O
v
er
s
io
n
s
1
th
r
o
u
g
h
1
1
,
w
h
er
ea
s
Z
ag
i
to
v
et
a
l.
[
1
3
]
be
n
ch
m
ar
k
ed
YOL
O
v
5
–
v
8
o
n
e
m
b
ed
d
ed
b
o
ar
d
s
,
s
h
o
w
i
n
g
th
e
i
m
p
o
r
tan
c
e
o
f
h
ar
d
w
ar
e
ac
ce
ler
atio
n
a
n
d
q
u
an
t
izatio
n
f
o
r
s
u
s
tai
n
in
g
i
n
f
er
e
n
ce
p
er
f
o
r
m
an
ce
.
Ho
w
e
v
er
,
to
th
e
b
est
o
f
o
u
r
k
n
o
w
led
g
e,
YO
L
O1
1
h
as
n
o
t
y
et
b
ee
n
s
y
s
te
m
a
ticall
y
e
v
al
u
ated
o
n
th
e
R
o
ck
ch
ip
R
K3
5
6
6
NP
U
w
i
t
h
in
a
co
s
t
-
e
f
f
ec
t
iv
e
s
i
n
g
le
-
b
o
ar
d
co
m
p
u
ter
s
u
ch
as
th
e
Or
an
g
e
P
i 3
B
.
T
h
i
s
i
s
th
e
f
i
r
s
t
s
tu
d
y
t
o
b
e
n
ch
m
a
r
k
Y
OL
O
1
1
o
n
t
h
e
O
r
an
g
e
Pi
3
B
(
R
K
3
5
6
6
N
PU
)
a
n
d
t
o
c
o
m
p
a
r
e
C
PU
-
b
a
s
e
d
an
d
N
PU
-
b
a
s
e
d
e
x
e
cu
t
io
n
s
u
n
d
e
r
r
e
p
r
es
en
t
a
tiv
e
r
e
a
l
-
w
o
r
l
d
s
c
en
a
r
i
o
s
.
W
e
ev
al
u
a
t
e
d
e
n
d
-
to
-
en
d
l
at
en
cy
,
i
n
cl
u
d
i
n
g
p
r
e
p
r
o
c
e
s
s
in
g
,
in
f
e
r
e
n
c
e
,
an
d
p
o
s
t
p
r
o
c
es
s
i
n
g
,
a
l
o
n
g
w
it
h
p
e
r
-
in
f
e
r
e
n
c
e
en
e
r
g
y
an
d
m
o
d
e
l
a
cc
u
r
a
cy
.
N
PU
a
c
c
e
l
e
r
at
i
o
n
r
e
d
u
c
e
d
in
f
er
e
n
c
e
l
a
t
en
cy
b
y
8
2
–
9
4
%
a
n
d
p
e
r
-
in
f
e
r
en
c
e
en
e
r
g
y
b
y
8
0
–
9
4
%
,
w
h
i
le
a
c
cu
r
a
cy
r
em
ai
n
e
d
w
i
th
in
0
.
0
2
6
–
0
.
0
3
0
m
e
an
av
e
r
ag
e
p
r
e
c
is
i
o
n
(
m
A
P
)
o
f
t
h
e
Ul
t
r
a
ly
ti
cs
b
a
s
e
lin
e
[
1
4
]
.
T
h
e
s
e
f
in
d
i
n
g
s
c
o
n
f
i
r
m
th
a
t
c
o
m
p
a
c
t
,
af
f
o
r
d
ab
l
e
p
l
a
t
f
o
r
m
s
c
an
s
u
s
t
a
in
m
o
d
e
r
n
o
b
je
c
t
d
e
t
e
c
t
i
o
n
w
o
r
k
l
o
a
d
s
w
it
h
o
u
t
s
u
b
s
t
an
t
i
a
l
a
c
c
u
r
ac
y
l
o
s
s
,
s
u
p
p
o
r
t
in
g
c
o
s
t
-
e
f
f
e
ct
iv
e
d
e
p
l
o
y
m
en
t
in
s
u
r
v
e
i
ll
a
n
ce
,
r
o
b
o
t
i
c
s
,
a
n
d
I
o
T
a
p
p
l
i
ca
t
i
o
n
s
[
1
5
]
.
2.
MET
H
O
D
YOL
O1
1
i
n
co
r
p
o
r
ates
s
ev
er
al
n
e
w
m
o
d
u
le
s
,
in
c
lu
d
i
n
g
cr
o
s
s
s
tag
e
p
ar
tial
w
it
h
k
er
n
el
s
ize
2
(
C
3
K2
)
,
s
p
atial
p
y
r
a
m
id
p
o
o
lin
g
-
f
ast
(
SP
P
F)
[
1
6
]
,
[
1
7
]
,
an
d
th
e
co
n
v
o
l
u
tio
n
al
b
lo
c
k
w
it
h
p
ar
al
lel
s
p
atial
atte
n
tio
n
(
C
2
P
SA
)
.
T
h
ese
co
m
p
o
n
e
n
ts
s
tr
ea
m
li
n
e
co
m
p
u
tatio
n
an
d
ac
ce
ler
ate
d
etec
tio
n
w
it
h
o
u
t
r
eq
u
ir
in
g
h
i
g
h
-
e
n
d
p
r
o
ce
s
s
o
r
s
.
B
u
ild
in
g
o
n
th
e
s
e
im
p
r
o
v
e
m
e
n
t
s
,
th
e
p
r
esen
t
s
t
u
d
y
e
x
a
m
i
n
es
YO
L
O1
1
’
s
p
er
f
o
r
m
an
ce
w
h
e
n
p
air
ed
w
it
h
a
n
Or
an
g
e
P
i
3
B
Fig
u
r
e
1
an
d
th
e
R
o
ck
c
h
ip
R
K3
5
6
6
NP
U,
w
i
th
p
ar
ticu
lar
e
m
p
h
as
is
o
n
s
p
ee
d
an
d
en
er
g
y
co
n
s
u
m
p
tio
n
.
T
h
e
s
tu
d
y
also
in
tr
o
d
u
ce
s
a
d
ed
icate
d
YOL
O
1
1
p
o
r
t
f
o
r
th
e
R
K3
5
6
6
p
latf
o
r
m
,
d
e
m
o
n
s
tr
ati
n
g
th
at
co
m
p
ac
t,
a
f
f
o
r
d
ab
le
h
ar
d
w
ar
e
ca
n
s
til
l
d
eliv
er
p
r
ac
tical
,
lo
w
-
late
n
c
y
i
m
a
g
e
i
n
f
er
e
n
ce
p
er
f
o
r
m
a
n
ce
at
t
h
e
ed
g
e.
Fig
u
r
e
1
.
Or
an
g
e
P
i 3
B
w
ith
R
o
ck
ch
ip
NP
U
R
K3
5
6
6
As
s
h
o
w
n
i
n
Fi
g
u
r
e
2
,
th
ese
n
e
w
m
o
d
u
les
w
er
e
in
te
g
r
ated
in
to
th
e
YO
L
O1
1
p
ip
elin
e
to
r
ef
in
e
f
ea
tu
r
e
ex
tr
ac
tio
n
,
i
m
p
r
o
v
e
s
e
n
s
iti
v
it
y
ac
r
o
s
s
o
b
j
ec
t
s
ca
les,
an
d
en
h
an
ce
s
p
atial
atte
n
tio
n
.
T
h
e
C
3
K2
b
lo
ck
r
e
p
lace
s
h
ea
v
ier
la
y
er
s
w
it
h
l
ig
h
t
w
ei
g
h
t
3
×3
co
n
v
o
lu
t
io
n
s
,
en
ab
li
n
g
th
e
m
o
d
el
to
m
ai
n
tai
n
d
etec
t
io
n
ac
cu
r
ac
y
w
h
il
e
r
ed
u
cin
g
co
m
p
u
ta
tio
n
al
lo
ad
an
d
ac
ce
ler
atin
g
th
e
o
v
er
al
l p
r
o
ce
s
s
.
T
h
e
SP
P
F
m
o
d
u
le
s
tr
ea
m
li
n
e
s
t
h
e
p
o
o
lin
g
s
ta
g
e
to
ca
p
t
u
r
e
in
f
o
r
m
atio
n
f
r
o
m
m
u
ltip
le
s
c
ales
m
o
r
e
ef
f
icien
tl
y
,
t
h
er
eb
y
i
m
p
r
o
v
i
n
g
th
e
d
etec
tio
n
o
f
b
o
th
s
m
all
an
d
lar
g
e
o
b
j
ec
ts
.
I
n
p
ar
allel
,
th
e
C
2
P
SA
b
lo
c
k
ap
p
lies
d
u
al
s
p
atial
a
tten
tio
n
p
ath
s
th
at
h
i
g
h
li
g
h
t
th
e
m
o
s
t
in
f
o
r
m
ati
v
e
r
eg
io
n
s
,
e
n
ab
li
n
g
th
e
n
et
w
o
r
k
to
r
e
f
i
n
e
o
b
j
ec
t b
o
u
n
d
ar
ies an
d
i
m
p
r
o
v
e
lo
ca
lizatio
n
ac
cu
r
ac
y
.
I
n
ad
d
itio
n
to
th
ese
m
o
d
u
le
s
,
YOL
O1
1
u
p
d
ates it
s
b
ac
k
b
o
n
e
w
it
h
li
g
h
ter
b
lo
ck
s
,
s
u
ch
a
s
C
3
K2
an
d
a
cr
o
s
s
-
s
tag
e
p
ar
tial
b
lo
ck
w
i
th
a
3
×3
k
er
n
el
(
C
3
K)
.
C
o
m
p
ar
ed
w
it
h
t
h
e
ea
r
lier
co
ar
s
e
-
to
-
f
in
e
(
C
2
F)
d
esig
n
,
t
h
es
e
s
tr
u
ct
u
r
es
r
ed
u
ce
r
ed
u
n
d
a
n
t
o
p
er
atio
n
s
w
h
ile
p
r
eser
v
in
g
f
ea
tu
r
e
r
ich
n
ess
,
g
iv
in
g
t
h
e
n
et
wo
r
k
a
lean
er
p
ath
f
o
r
tr
ain
i
n
g
a
n
d
in
f
er
e
n
ce
.
T
o
em
p
h
asize
t
h
i
s
r
ef
i
n
e
m
en
t,
Fi
g
u
r
e
3
d
ir
ec
tly
co
m
p
ar
e
s
t
h
e
C
2
F
b
l
o
ck
w
ith
th
e
n
e
w
er
C
3
K2
/C
3
K
s
tr
u
ctu
r
e
s
,
illu
s
tr
at
in
g
h
o
w
ef
f
icie
n
c
y
is
i
m
p
r
o
v
e
d
w
it
h
o
u
t sacr
i
f
ici
n
g
r
ep
r
esen
t
atio
n
al
p
o
w
er
.
An
NP
U
is
a
s
p
ec
ialized
ac
ce
l
er
ato
r
f
o
r
d
ee
p
-
lear
n
i
n
g
w
o
r
k
l
o
ad
s
[
1
8
]
-
[
2
0
]
,
d
esig
n
ed
to
p
er
f
o
r
m
lar
g
e
m
atr
i
x
o
p
er
atio
n
s
m
o
r
e
e
f
f
icie
n
tl
y
t
h
an
C
P
Us
o
r
g
r
ap
h
ic
s
p
r
o
ce
s
s
in
g
u
n
its
(
GP
Us).
I
ts
e
f
f
i
cien
c
y
d
er
iv
es f
r
o
m
p
ar
allel
d
ataf
lo
w
ar
ch
itectu
r
e
s
,
o
n
-
c
h
ip
m
e
m
o
r
y
,
an
d
lo
w
-
p
o
w
er
cir
cu
itr
y
,
m
ak
in
g
it
s
u
itab
le
f
o
r
en
er
g
y
-
co
n
s
tr
ain
ed
ed
g
e
d
ev
ices s
u
c
h
as s
m
ar
tp
h
o
n
es a
n
d
s
i
n
g
le
-
b
o
ar
d
co
m
p
u
ter
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
1
6
9
3
-
6930
T
E
L
KOM
NI
K
A
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l
C
o
n
tr
o
l
,
Vo
l.
24
,
No
.
1
,
Feb
r
u
ar
y
20
26
:
1
26
-
1
41
128
Fig
u
r
e
2
.
YOL
O1
1
a
r
ch
itect
u
r
e
Fig
u
r
e
3
.
C
o
m
p
ar
is
o
n
o
f
C
2
F
an
d
C
3
K2
b
lo
ck
s
T
h
is
s
tu
d
y
e
v
alu
a
ted
YOL
O1
1
p
er
f
o
r
m
a
n
ce
o
n
th
e
Or
an
g
e
P
i
3
B
b
y
co
m
p
ar
in
g
C
P
U
p
r
o
ce
s
s
in
g
w
it
h
NP
U
ac
ce
ler
atio
n
f
r
o
m
th
e
R
K3
5
6
6
ch
ip
.
T
h
e
an
aly
s
is
co
n
s
i
d
er
ed
th
r
ee
s
tag
es
o
f
d
etec
ti
o
n
.
I
n
p
r
ep
r
o
ce
s
s
in
g
,
r
a
w
i
m
a
g
es
w
er
e
r
es
ized
,
n
o
r
m
alize
d
,
an
d
r
e
f
o
r
m
atted
to
m
ee
t
in
p
u
t
r
eq
u
ir
e
m
e
n
t
s
.
I
n
f
er
en
ce
th
e
n
ex
ec
u
ted
t
h
e
tr
ain
ed
YO
L
O1
1
n
et
w
o
r
k
,
s
c
an
n
ed
ea
c
h
i
m
a
g
e,
ex
tr
ac
ted
f
ea
tu
r
es,
an
d
p
r
ed
icted
b
o
u
n
d
in
g
b
o
x
e
s
w
it
h
clas
s
lab
els,
a
p
r
o
ce
s
s
d
o
m
i
n
ated
b
y
lar
g
e
m
atr
ix
m
u
ltip
licat
io
n
s
co
n
s
tit
u
ti
n
g
m
o
s
t
o
f
th
e
co
m
p
u
tatio
n
a
l
lo
ad
[
2
1
]
.
Fin
all
y
,
p
o
s
tp
r
o
ce
s
s
i
n
g
ap
p
lie
d
n
o
n
-
m
a
x
i
m
u
m
s
u
p
p
r
ess
io
n
(
NM
S)
to
f
ilter
o
v
er
lap
s
a
n
d
g
en
er
ate
f
i
n
al
o
b
j
ec
t
class
es a
n
d
co
n
f
id
en
ce
s
co
r
es.
2
.
1
.
H
a
rdwa
re
pla
t
f
o
rm
T
h
e
Or
an
g
e
P
i
3
B
s
in
g
le
-
b
o
ar
d
co
m
p
u
ter
is
b
u
ilt
o
n
t
h
e
R
o
ck
ch
ip
R
K3
5
6
6
s
y
s
te
m
-
on
-
c
h
ip
(
So
C
)
[
2
2
]
,
a
lo
w
-
p
o
w
er
AR
M
-
b
ase
d
p
latf
o
r
m
d
esi
g
n
ed
f
o
r
e
m
b
ed
d
ed
an
d
ar
tif
icial
in
telli
g
e
n
c
e
o
f
th
i
n
g
s
(
A
I
o
T
)
ap
p
licatio
n
s
.
T
h
e
So
C
in
te
g
r
at
es
a
q
u
ad
-
co
r
e
AR
M
C
o
r
tex
-
A
5
5
C
P
U
[
2
3
]
,
a
Ma
li
-
G5
2
GP
U,
an
d
an
R
K3
5
6
6
NP
U
o
f
f
iciall
y
r
ated
at
0
.
8
–
1
t
er
a
o
p
e
r
atio
n
s
p
er
s
e
co
n
d
(
T
OP
S)
[
2
4
]
,
[
2
5
]
,
d
ep
en
d
in
g
o
n
ar
ith
m
etic
p
r
ec
is
io
n
.
I
n
p
r
ac
tice,
th
is
co
r
r
esp
o
n
d
s
to
ap
p
r
o
x
i
m
atel
y
0
.
8
T
O
P
S
f
o
r
1
6
-
b
it
in
teg
er
(
I
NT
1
6
)
p
r
ec
is
i
o
n
[
2
6
]
an
d
1
T
O
P
S
f
o
r
8
-
b
it
i
n
te
g
er
(
I
NT
8
)
p
r
ec
is
io
n
[
2
7
]
.
Ho
w
ev
er
,
s
u
c
h
p
ea
k
r
atin
g
s
ar
e
t
h
eo
r
etica
l
a
n
d
d
o
n
o
t
al
w
a
y
s
r
ef
lec
t
r
ea
l
p
er
f
o
r
m
an
ce
,
p
ar
ticu
lar
l
y
u
n
d
er
th
e
f
lo
ati
n
g
-
p
o
in
t
(
FP
3
2
)
[
2
8
]
w
o
r
k
lo
ad
s
u
s
e
d
in
th
i
s
s
tu
d
y
.
T
h
e
b
o
ar
d
in
cl
u
d
es
L
P
DD
R
4
m
e
m
o
r
y
a
n
d
a
2
5
6
G
B
n
o
n
-
v
o
latile
m
e
m
o
r
y
ex
p
r
es
s
(
NVM
e)
M.
2
s
o
lid
-
s
tate
d
r
iv
e
(
SS
D)
as
p
r
i
m
ar
y
s
to
r
a
g
e.
T
h
ese
s
p
ec
if
icatio
n
s
o
f
f
er
s
u
f
f
ic
ien
t
b
a
n
d
w
id
t
h
an
d
ca
p
ac
i
t
y
f
o
r
m
o
d
er
n
o
b
j
ec
t
d
etec
tio
n
m
o
d
el
s
w
h
i
le
m
ai
n
tai
n
i
n
g
a
c
o
m
p
ac
t
d
esi
g
n
an
d
lo
w
p
o
w
e
r
co
n
s
u
m
p
t
io
n
,
m
a
k
i
n
g
t
h
e
p
latf
o
r
m
s
u
itab
le
f
o
r
ed
g
e
ar
tif
icial
i
n
tel
lig
e
n
ce
(
AI
)
ex
p
er
i
m
en
t
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KOM
NI
K
A
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l
C
o
n
tr
o
l
P
erfo
r
ma
n
ce
en
h
a
n
ce
men
t o
f
emb
ed
d
ed
o
b
ject
d
etec
tio
n
via
n
eu
r
a
l h
a
r
d
w
a
r
e
..
(
A
lw
in
Ha
r
t
o
n
o
Lima
r
a
n
)
129
2
.
2
.
So
f
t
w
a
re
env
iro
nm
e
nt
A
ll
e
x
p
er
i
m
e
n
ts
w
er
e
co
n
d
u
ct
ed
o
n
Or
an
g
e
P
i
OS
1
.
0
.
8
(
U
b
u
n
t
u
J
a
m
m
y
)
w
it
h
th
e
L
in
u
x
5
.
1
0
.
1
6
0
-
R
o
ck
c
h
ip
-
RK
356
X
k
er
n
e
l,
r
u
n
n
i
n
g
o
n
a
6
4
-
b
it
AR
M
ar
ch
it
ec
tu
r
e.
T
h
e
w
o
r
k
f
lo
w
u
s
ed
P
y
th
o
n
3
.
1
0
.
1
2
,
w
i
th
P
y
T
o
r
ch
2
.
8
.
0
an
d
T
o
r
ch
Vis
io
n
0
.
2
3
.
0
as
th
e
p
r
im
ar
y
d
ee
p
-
le
ar
n
i
n
g
f
r
a
m
e
w
o
r
k
s
.
YOL
O1
1
w
a
s
d
ep
lo
y
ed
u
s
in
g
th
e
Ultr
al
y
tics
lib
r
ar
y
[
2
9
]
(
v
er
s
io
n
8
.
3
.
2
0
2
)
.
T
h
e
p
y
co
co
to
o
ls
p
ac
k
ag
e
(
v
er
s
io
n
2
.
0
.
1
0
)
e
n
s
u
r
ed
co
m
p
lia
n
ce
w
it
h
th
e
C
OC
O
ev
a
lu
a
t
io
n
p
r
o
to
co
l.
Su
p
p
o
r
tin
g
lib
r
ar
ies
in
clu
d
ed
Nu
m
P
y
2
.
2
.
6
,
Sci
P
y
1
.
1
5
.
3
,
P
illo
w
1
1
.
3
.
0
,
an
d
Op
en
C
V
4
.
1
2
.
0
f
o
r
p
r
e
p
r
o
ce
s
s
in
g
an
d
n
u
m
er
ical
co
m
p
u
tatio
n
.
Dep
lo
y
m
e
n
t
o
n
t
h
e
R
K3
5
6
6
NP
U
u
s
ed
R
o
ck
c
h
ip
’
s
r
o
ck
ch
ip
n
e
u
r
al
n
e
t
w
o
r
k
(
R
KNN
)
-
T
o
o
lk
it
L
ite2
(
v
er
s
io
n
1
.
6
.
0
)
,
in
s
talled
f
r
o
m
t
h
e
o
f
f
icial
aa
r
ch
6
4
w
h
ee
l,
s
u
p
p
o
r
tin
g
p
r
etr
ain
ed
m
o
d
el
i
m
p
o
r
t
f
r
o
m
P
y
T
o
r
ch
,
T
en
s
o
r
F
lo
w
,
an
d
o
p
en
n
e
u
r
al
n
et
w
o
r
k
ex
c
h
a
n
g
e
(
ONNX
)
f
o
r
m
at.
2
.
3
.
M
o
del c
o
nv
er
s
io
n wo
rkf
lo
w
YOL
O1
1
m
o
d
els
w
er
e
o
b
tain
ed
as o
f
f
icia
l p
r
etr
ain
ed
w
e
ig
h
ts
f
r
o
m
Ultr
al
y
tics
,
tr
ai
n
ed
o
n
th
e
C
OC
O
d
ataset.
C
o
n
v
er
s
io
n
u
s
ed
th
e
Ultr
al
y
tic
s
ex
p
o
r
t
u
tili
t
y
w
it
h
th
e
co
m
m
an
d
“y
o
lo
ex
p
o
r
t
m
o
d
el=
m
o
d
el_
n
a
m
e.
p
t
f
o
r
m
at=
r
k
n
n
n
a
m
e=
r
k
3
5
6
6
,
”
w
h
ic
h
f
ir
s
t g
en
er
ated
an
ONN
X
m
o
d
el
b
ef
o
r
e
p
r
o
d
u
cin
g
th
e
R
KNN
f
o
r
m
a
t.
T
h
e
co
m
m
a
n
d
ex
p
o
r
ted
th
e
P
y
T
o
r
ch
m
o
d
el
to
ONNX
a
n
d
th
e
n
co
n
v
er
ted
it to
R
KNN
u
s
in
g
R
KNN
-
T
o
o
lk
it L
ite2
.
T
h
is
in
ter
m
ed
iate
s
tep
w
as a
u
t
o
m
a
ted
w
i
th
i
n
t
h
e
Ultr
al
y
tics
l
ib
r
ar
y
,
en
ab
li
n
g
ex
ec
u
tio
n
a
s
a
s
in
g
le
co
m
m
a
n
d
.
A
ll
m
o
d
el
s
w
er
e
m
ai
n
tai
n
ed
at
FP
3
2
p
r
ec
is
io
n
w
it
h
o
u
t
p
r
u
n
in
g
o
r
q
u
an
ti
za
tio
n
to
p
r
eser
v
e
b
aselin
e
ac
cu
r
ac
y
.
Dete
c
tio
n
o
u
tp
u
t
s
w
er
e
r
e
f
in
ed
u
s
in
g
cla
s
s
-
a
g
n
o
s
tic
NM
S
a
t
an
I
o
U
th
r
es
h
o
ld
o
f
0
.
4
5
.
T
h
e
ex
p
er
i
m
e
n
ts
w
er
e
r
u
n
w
ith
a
b
atch
s
ize
o
f
1
,
an
d
th
e
m
a
x
i
m
u
m
d
etec
t
io
n
s
(
m
ax
_
d
et)
w
er
e
l
i
m
ited
b
y
th
e
8
GB
s
y
s
te
m
R
A
M.
A
lt
h
o
u
g
h
t
h
e
R
K3
5
6
6
N
P
U
s
u
p
p
o
r
ts
I
N
T
8
an
d
I
N
T
1
6
in
f
e
r
en
ce
,
s
ev
er
al
I
NT
8
co
n
v
er
s
io
n
tr
ials
o
n
C
OC
O
Val2
0
1
7
(
5
0
0
–
5
0
0
0
i
m
ag
e
s
)
p
r
o
d
u
ce
d
in
v
alid
r
esu
lts
d
u
e
to
q
u
an
tizat
io
n
n
o
is
e.
P
r
u
n
in
g
w
a
s
also
ex
clu
d
ed
to
av
o
id
co
n
f
o
u
n
d
i
n
g
ef
f
ec
ts
.
FP
3
2
p
r
ec
is
io
n
w
a
s
th
er
ef
o
r
e
m
ai
n
t
ai
n
ed
th
r
o
u
g
h
o
u
t to
en
s
u
r
e
th
at
t
h
e
m
ea
s
u
r
ed
C
P
U
–
NP
U
d
if
f
er
en
ce
s
r
ef
lecte
d
th
e
h
ar
d
w
ar
e
e
f
f
i
cien
c
y
a
lo
n
e.
2
.
4
.
Da
t
a
s
et
T
h
e
Mic
r
o
s
o
f
t
C
OC
O
2
0
1
7
d
ataset
w
as
u
s
ed
to
ev
alu
ate
d
etec
tio
n
ac
cu
r
ac
y
.
I
t
d
ef
in
es
8
0
o
b
j
ec
t
ca
teg
o
r
ies
a
n
d
p
r
o
v
id
es
1
1
8
,
2
8
7
tr
ain
in
g
i
m
a
g
es
(
tr
ain
2
0
1
7
)
,
5
,
0
0
0
v
alid
atio
n
i
m
a
g
es
(
v
a
l2
0
1
7
)
,
an
d
4
0
,
6
7
0
u
n
lab
eled
tes
t i
m
a
g
es d
i
v
id
ed
in
to
test
-
d
ev
a
n
d
test
-
ch
alle
n
g
e
[
3
0
]
,
[
3
1
]
.
T
h
e
v
alid
atio
n
s
et
w
as
u
s
ed
f
o
r
m
A
P
ev
alu
a
tio
n
,
co
n
s
i
s
te
n
t
w
it
h
th
e
Ultr
al
y
tics
YO
L
O1
1
b
en
ch
m
ar
k
.
A
cc
u
r
ac
y
w
as
co
m
p
u
te
d
u
s
in
g
p
y
co
co
to
o
ls
an
d
th
e
C
OC
O
p
r
o
to
co
l
at
a
co
n
f
id
en
ce
t
h
r
es
h
o
ld
o
f
0
.
0
0
1
.
F
o
r
Or
an
g
e
P
i
3
B
in
f
er
en
ce
,
a
0
.
2
5
th
r
esh
o
ld
f
i
lter
ed
lo
w
-
co
n
f
id
en
ce
p
r
ed
ictio
n
s
.
L
aten
c
y
an
d
e
n
er
g
y
b
en
c
h
m
ar
k
i
n
g
w
er
e
co
n
d
u
cted
o
n
f
i
v
e
r
a
n
d
o
m
I
n
ter
n
et
i
m
a
g
e
s
r
ep
r
esen
tin
g
d
i
v
er
s
e
v
is
u
al
co
n
d
itio
n
s
.
T
h
is
e
n
s
u
r
ed
ac
cu
r
a
c
y
ali
g
n
ed
w
it
h
t
h
e
o
f
f
icia
l
b
en
ch
m
ar
k
,
w
h
er
ea
s
h
ar
d
w
ar
e
p
er
f
o
r
m
a
n
ce
w
as a
s
s
es
s
ed
i
n
d
ep
en
d
en
tl
y
o
f
lar
g
e
-
s
ca
le
d
ataset
p
r
o
ce
s
s
in
g
.
2
.
5
.
P
r
o
ce
s
s
ing
pip
eli
ne
T
h
e
w
o
r
k
f
lo
w
w
as
m
ea
s
u
r
ed
in
th
r
ee
s
tag
e
s
Fi
g
u
r
e
4
:
p
r
ep
r
o
ce
s
s
i
n
g
(
r
esize,
n
o
r
m
aliza
t
io
n
,
an
d
ten
s
o
r
r
ef
o
r
m
atti
n
g
)
,
in
f
er
en
ce
,
a
n
d
p
o
s
tp
r
o
ce
s
s
in
g
(
NM
S).
I
n
t
h
e
C
P
U
b
aselin
e,
al
l
s
ta
g
es
r
a
n
o
n
th
e
C
P
U.
I
n
t
h
e
NP
U
co
n
f
ig
u
r
atio
n
,
in
f
er
en
ce
r
an
o
n
th
e
R
K3
5
6
6
NP
U
(
d
e
v
ice)
,
w
h
er
ea
s
p
r
ep
r
o
ce
s
s
in
g
an
d
p
o
s
tp
r
o
ce
s
s
in
g
r
an
o
n
th
e
C
P
U
(
h
o
s
t)
;
h
o
s
t→
NP
U
an
d
NP
U→h
o
s
t
te
n
s
o
r
tr
an
s
f
er
s
w
er
e
in
cl
u
d
ed
in
t
h
e
laten
c
y
an
d
en
er
g
y
to
tals
.
Stag
e
ti
m
e
s
w
er
e
s
u
m
m
ed
f
o
r
en
d
-
t
o
-
e
n
d
laten
c
y
,
an
d
p
o
w
er
d
r
a
w
w
a
s
lo
g
g
ed
co
n
tin
u
o
u
s
l
y
w
it
h
a
d
ig
ital
DC
a
m
m
eter
to
co
m
p
u
t
e
en
er
g
y
p
er
in
f
er
en
ce
.
Static,
s
in
g
le
-
i
m
ag
e
i
n
p
u
t
s
w
er
e
u
s
ed
in
th
e
ev
alu
atio
n
s
to
is
o
late
th
e
C
P
U
v
s
.
NP
U
ef
f
ec
ts
at
m
atc
h
ed
ac
cu
r
ac
y
.
2
.
6
.
E
v
a
lua
t
i
o
n
m
et
rics
YOL
O
1
1
o
n
t
h
e
Or
a
n
g
e
P
i
3
B
w
as
e
v
al
u
ated
u
s
in
g
t
h
r
ee
in
d
icato
r
s
:
p
r
o
ce
s
s
in
g
t
i
m
e,
p
o
w
er
co
n
s
u
m
p
tio
n
,
an
d
d
etec
tio
n
ac
cu
r
ac
y
.
P
r
o
ce
s
s
in
g
ti
m
e
w
as
o
b
tain
ed
f
r
o
m
m
illi
s
ec
o
n
d
-
le
v
el
ti
m
e
s
ta
m
p
s
f
o
r
p
r
ep
r
o
ce
s
s
in
g
,
i
n
f
er
e
n
ce
,
an
d
p
o
s
tp
r
o
ce
s
s
in
g
,
w
i
th
th
e
to
tal
as
th
ei
r
s
u
m
.
P
o
w
er
co
n
s
u
m
p
t
io
n
w
as
m
ea
s
u
r
ed
in
li
n
e
o
n
th
e
5
V
USB
-
C
s
u
p
p
ly
u
s
i
n
g
a
HiDa
n
ce
t
y
p
e
-
C
p
o
w
er
m
eter
(
D
C
4
.
5
–
5
0
V,
0
–
6
A
co
n
ti
n
u
o
u
s
)
.
Vo
ltag
e
an
d
cu
r
r
en
t
w
er
e
lo
g
g
ed
in
r
ea
l
ti
m
e,
an
d
in
s
ta
n
ta
n
eo
u
s
p
o
w
er
w
as
ca
lc
u
lated
as
th
eir
p
r
o
d
u
ct.
P
er
-
i
m
a
g
e
m
e
a
n
p
o
w
er
w
as
d
er
iv
e
d
b
y
av
er
ag
i
n
g
r
ea
d
in
g
s
ac
r
o
s
s
ea
ch
in
f
er
en
ce
in
ter
v
al.
Dete
ctio
n
ac
cu
r
ac
y
w
a
s
r
ep
o
r
ted
as
m
A
P
@
[
0
.
5
:0
.
9
5
]
u
n
d
er
th
e
C
OC
O
p
r
o
to
co
l,
w
ith
m
A
P
at
a
0
.
0
0
1
co
n
f
id
en
ce
th
r
esh
o
ld
(
as
in
Ultr
al
y
tic
s
YO
L
O)
an
d
d
ep
lo
y
m
en
t in
f
er
en
ce
at
0
.
2
5
.
A
d
d
iti
o
n
al
in
d
icato
r
s
i
n
cl
u
d
e
d
th
r
o
u
g
h
p
u
t
(
i
m
a
g
es
p
er
s
ec
o
n
d
at
b
atch
s
ize
1
)
,
en
er
g
y
p
er
im
a
g
e
(
av
er
ag
e
p
o
w
er
×
in
f
er
e
n
ce
d
u
r
atio
n
)
,
p
er
f
o
r
m
a
n
ce
p
er
w
at
t,
an
d
C
P
U
–
NP
U
s
p
ee
d
u
p
an
d
en
er
g
y
-
r
ed
u
c
tio
n
r
atio
s
.
T
h
e
r
esu
lts
ar
e
r
ep
o
r
ted
as p
er
-
i
m
a
g
e
av
er
a
g
es,
w
it
h
d
is
t
r
ib
u
tio
n
s
p
r
o
v
id
ed
w
h
er
e
r
elev
an
t.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
1
6
9
3
-
6930
T
E
L
KOM
NI
K
A
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l
C
o
n
tr
o
l
,
Vo
l.
24
,
No
.
1
,
Feb
r
u
ar
y
20
26
:
1
26
-
1
41
130
Fig
u
r
e
4
.
C
P
U
v
s
.
NP
U
p
r
o
ce
s
s
i
n
g
p
ip
eli
n
e
2
.
7
.
E
x
peri
m
ent
a
l scena
rio
s
Fiv
e
p
r
ac
tical
s
ce
n
ar
io
s
w
er
e
d
ev
elo
p
ed
to
ev
alu
ate
h
o
w
Y
OL
O1
1
ad
ap
ts
to
d
if
f
er
en
t
en
v
ir
o
n
m
en
tal
co
n
d
itio
n
s
o
n
t
h
e
Or
an
g
e
P
i
3
B
.
T
h
e
r
ep
r
esen
tat
iv
e
f
r
a
m
e
s
s
h
o
w
n
i
n
Fig
u
r
e
s
5
(
a)
–
(
e)
ca
p
tu
r
e
v
ar
iatio
n
s
,
in
cl
u
d
in
g
d
iv
er
s
e
li
g
h
ti
n
g
co
n
d
itio
n
s
,
m
o
tio
n
lev
e
ls
,
an
d
o
b
ject
s
izes.
T
h
ese
s
ce
n
ar
io
s
w
er
e
s
p
ec
i
f
ica
ll
y
d
esig
n
ed
to
p
r
o
b
e
YOL
O1
1
’
s
p
er
f
o
r
m
a
n
ce
u
n
d
er
v
ar
io
u
s
r
ea
l
-
w
o
r
ld
co
n
d
itio
n
s
.
−
C
las
s
r
o
o
m
(
I
m
a
g
e
1
;
Fig
u
r
e
5
(
a)
-
q
u
iet
class
r
o
o
m
w
it
h
eig
h
t
s
tan
d
in
g
s
u
b
j
ec
ts
u
n
d
er
u
n
if
o
r
m
li
g
h
ti
n
g
;
it
ev
alu
a
tes t
h
e
s
tab
ilit
y
o
f
m
u
lti
-
p
er
s
o
n
d
etec
tio
n
o
n
a
s
in
g
le
f
r
a
m
e.
−
Ur
b
an
s
tr
ee
t
cr
o
s
s
in
g
(
I
m
a
g
e
2
;
Fig
u
r
e
5
(
b)
-
b
u
s
y
cr
o
s
s
w
al
k
w
it
h
p
ed
estrian
s
an
d
cy
clis
ts
;
it
te
s
ts
d
etec
tio
n
u
n
d
er
cr
o
w
d
i
n
g
an
d
o
cc
lu
s
io
n
.
−
W
ar
eh
o
u
s
e
o
p
er
atio
n
s
(
I
m
a
g
e
3
;
Fig
u
r
e
5
(
c)
-
lo
ad
in
g
-
d
o
ck
C
C
T
V
w
it
h
aisl
es
an
d
clu
t
t
er
;
it
ass
ess
es
r
o
b
u
s
tn
es
s
ag
a
in
s
t o
cc
lu
s
io
n
a
n
d
v
is
u
all
y
s
i
m
i
lar
class
e
s
.
−
P
ar
k
in
g
lo
t
m
o
n
ito
r
i
n
g
(
I
m
a
g
e
4
;
Fig
u
r
e
5
(
d)
-
m
id
-
a
f
t
er
n
o
o
n
lo
t
w
i
th
v
e
h
icle
s
an
d
p
ed
e
s
tr
ian
s
;
s
tr
o
n
g
s
h
ad
o
w
s
a
n
d
h
i
g
h
co
n
tr
ast ch
a
llen
g
e
th
e
s
ep
ar
atio
n
o
f
ad
j
ac
e
n
t o
b
j
ec
ts
.
−
Nig
h
tti
m
e
s
tr
ee
t
s
ce
n
e
(
I
m
ag
e
5
;
Fig
u
r
e
5
(
e)
-
d
i
m
u
r
b
an
s
tr
ee
t
w
i
th
la
m
p
s
a
n
d
h
ea
d
li
g
h
ts
;
it
p
r
o
b
es
lo
w
lig
h
t,
g
lar
e,
an
d
n
o
is
e
ef
f
ec
t
s
o
n
d
etec
tio
n
.
(
a)
(
b
)
(
c)
(
d
)
(
e)
Fig
u
r
e
5
.
B
en
ch
m
ar
k
i
n
p
u
t sc
en
es
u
s
ed
f
o
r
test
i
n
g
YO
L
O1
1
p
er
f
o
r
m
a
n
ce
ac
r
o
s
s
d
iv
er
s
e
e
n
v
ir
o
n
m
e
n
t
s
:
(
a)
class
r
o
o
m
s
etti
n
g
,
(
b
)
u
r
b
an
s
tr
ee
t c
r
o
s
s
i
n
g
,
(
c)
w
ar
eh
o
u
s
e
o
p
er
atio
n
s
,
(
d
)
p
ar
k
in
g
lo
t
m
o
n
ito
r
in
g
,
an
d
(
e)
n
ig
h
tti
m
e
s
tr
ee
t sce
n
e
T
h
ese
s
ce
n
ar
io
s
p
r
o
v
id
ed
co
m
p
le
m
en
tar
y
b
en
c
h
m
ar
k
s
f
o
r
co
m
p
ar
i
n
g
C
P
U
-
o
n
l
y
a
n
d
NP
U
-
a
cc
eler
ated
in
f
er
en
ce
.
T
h
e
m
ea
s
u
r
e
m
en
ts
in
cl
u
d
ed
p
r
ep
r
o
ce
s
s
in
g
,
in
f
er
en
ce
,
p
o
s
tp
r
o
ce
s
s
in
g
,
an
d
to
t
al
p
er
-
i
m
a
g
e
ti
m
es.
Fiv
e
s
ce
n
ar
io
s
ca
p
tu
r
ed
r
ea
l
-
w
o
r
ld
co
n
d
itio
n
s
:
(
i)
li
g
h
tin
g
f
r
o
m
b
r
ig
h
t
o
u
td
o
o
r
to
lo
w
-
li
g
h
t
n
ig
h
tt
i
m
e,
(
ii)
m
o
tio
n
f
r
o
m
s
tat
ic
in
d
o
o
r
to
cr
o
w
d
ed
s
tr
ee
ts
,
an
d
(
iii)
o
b
j
ec
t
s
i
ze
s
an
d
d
en
s
ities
,
i
n
c
lu
d
in
g
p
ed
estrian
s
,
v
eh
ic
les,
an
d
w
ar
e
h
o
u
s
e
g
o
o
d
s
.
So
u
r
ce
im
a
g
es
also
v
ar
ie
d
in
f
ile
s
iz
e
(
h
u
n
d
r
ed
s
o
f
k
ilo
b
y
te
s
to
s
ev
er
al
m
eg
ab
y
tes)
a
n
d
r
eso
lu
tio
n
(
7
2
–
3
0
0
D
P
I
)
,
en
h
an
ci
n
g
in
p
u
t d
i
v
er
s
it
y
.
T
o
g
eth
er
,
th
ese
s
elec
t
io
n
s
e
n
s
u
r
ed
th
at
t
h
e
e
v
alu
a
tio
n
r
e
f
l
ec
ted
r
ea
lis
tic,
ch
alle
n
g
in
g
d
ep
lo
y
m
e
n
t
w
it
h
o
u
t
ad
d
itio
n
al
d
ata
s
ets.
T
h
e
m
eth
o
d
o
lo
g
y
also
o
u
t
li
n
ed
th
e
Or
a
n
g
e
P
i
3
B
h
ar
d
w
ar
e,
C
OC
O
2
0
1
7
d
ataset
w
it
h
lo
ca
l
v
alid
atio
n
s
u
b
s
et
s
,
m
o
d
el
co
n
v
er
s
io
n
,
p
ar
am
eter
s
etti
n
g
s
,
an
d
p
er
f
o
r
m
a
n
ce
m
et
r
ics.
A
cc
u
r
ac
y
an
d
ef
f
icien
c
y
w
er
e
r
ep
o
r
ted
th
r
o
u
g
h
m
AP
s
co
r
es,
p
r
o
ce
s
s
in
g
ti
m
es,
a
n
d
p
o
w
er
m
ea
s
u
r
e
m
en
t
s
f
o
r
C
P
U
an
d
NP
U
r
u
n
s
.
W
ith
t
h
e
p
r
o
ce
d
u
r
es
estab
lis
h
ed
,
th
e
n
ex
t
s
ec
tio
n
p
r
esen
ts
t
h
e
e
x
p
er
i
m
e
n
tal
r
e
s
u
lt
s
a
n
d
d
is
cu
s
s
es
YOL
O1
1
p
er
f
o
r
m
a
n
ce
u
n
d
er
e
m
b
ed
d
ed
ed
g
e
h
ar
d
w
ar
e
co
n
s
tr
ain
t
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KOM
NI
K
A
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l
C
o
n
tr
o
l
P
erfo
r
ma
n
ce
en
h
a
n
ce
men
t o
f
emb
ed
d
ed
o
b
ject
d
etec
tio
n
via
n
eu
r
a
l h
a
r
d
w
a
r
e
..
(
A
lw
in
Ha
r
t
o
n
o
Lima
r
a
n
)
131
3.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
N
T
h
is
s
ec
tio
n
co
m
p
ar
es
YO
L
O
1
1
o
n
th
e
C
P
U
a
n
d
o
n
t
h
e
R
K3
5
6
6
NP
U
u
n
d
er
m
atc
h
ed
-
ac
c
u
r
ac
y
,
s
tat
ic
-
i
m
a
g
e
co
n
d
itio
n
s
.
W
e
ev
al
u
ate
p
er
-
i
m
ag
e
late
n
c
y
—
d
ec
o
m
p
o
s
ed
in
to
p
r
ep
r
o
ce
s
s
in
g
,
in
f
er
en
ce
,
an
d
p
o
s
tp
r
o
ce
s
s
in
g
a
n
d
en
e
r
g
y
,
a
n
d
q
u
an
ti
f
y
e
f
f
icie
n
c
y
a
s
p
er
ce
n
tag
e
c
h
a
n
g
e.
T
ab
le
1
co
n
s
o
lid
ates
th
e
r
es
u
lts
b
y
j
u
x
tap
o
s
in
g
C
P
U
T
ab
le
1
(
a
)
an
d
NP
U
T
ab
le
1
(
b
)
w
all
-
clo
ck
ti
m
es
an
d
r
ep
o
r
tin
g
ac
ce
ler
atio
n
as
p
er
ce
n
tag
e
r
ed
u
ctio
n
.
T
ab
le
1
.
Vis
u
al
co
m
p
ar
is
o
n
o
f
YOL
O1
1
d
etec
tio
n
p
er
f
o
r
m
a
n
ce
ac
r
o
s
s
t
w
o
p
latf
o
r
m
s
(
a)
C
P
U
-
b
ased
in
f
er
e
n
ce
u
s
i
n
g
P
y
T
o
r
ch
M
o
d
e
l
P
i
c
t
u
r
e
P
r
e
p
r
o
c
e
ss
(
ms)
I
n
f
e
r
e
n
c
e
(
ms)
P
o
st
p
r
o
c
e
ss
(
ms)
T
o
t
a
l
t
i
me
(
ms)
D
e
t
e
c
t
i
o
n
r
e
su
l
t
Y
O
L
O
1
1
n
I
mag
e
1
4
2
.
1
0
1
,
5
9
7
.
2
0
1
9
.
8
0
1
,
6
5
9
.
1
0
8
p
e
r
so
n
s
Y
O
L
O
1
1
n
I
mag
e
2
4
4
.
3
0
1
,
6
2
2
.
5
0
1
0
.
6
0
1
,
6
7
7
.
4
0
1
2
p
e
r
so
n
s,
3
c
a
r
s
,
a
n
d
3
t
r
a
f
f
i
c
l
i
g
h
t
s
Y
O
L
O
1
1
n
I
mag
e
3
2
8
.
4
0
1
,
4
2
8
.
1
0
9
.
8
0
1
,
4
6
6
.
3
0
4
p
e
r
so
n
s
a
n
d
2
t
r
u
c
k
s
Y
O
L
O
1
1
n
I
mag
e
4
2
7
.
2
0
1
,
4
3
1
.
8
0
9
.
8
0
1
,
4
6
8
.
8
0
1
9
c
a
r
s
a
n
d
7
t
r
u
c
k
s
Y
O
L
O
1
1
n
I
mag
e
5
1
8
.
4
0
1
,
6
3
5
.
2
0
1
0
.
5
0
1
,
6
6
4
.
1
0
6
p
e
r
so
n
s
a
n
d
4
c
a
r
s
Y
O
L
O
1
1
s
I
mag
e
1
3
6
.
4
0
5
,
1
2
2
.
2
0
1
0
.
2
0
5
,
1
6
8
.
8
0
8
p
e
r
so
n
s
Y
O
L
O
1
1
s
I
mag
e
2
4
1
.
3
0
5
,
1
1
0
.
2
0
1
0
.
6
0
5
,
1
6
2
.
1
0
1
3
p
e
r
so
n
s,
2
c
a
r
s
,
2
t
r
a
f
f
i
c
l
i
g
h
t
s
,
3
h
a
n
d
b
a
g
s,
1
t
i
e
,
a
n
d
1
c
u
p
Y
O
L
O
1
1
s
I
mag
e
3
2
5
.
9
0
4
,
4
8
4
.
8
0
9
.
9
0
4
,
5
2
0
.
6
0
8
p
e
r
so
n
s
a
n
d
1
t
r
u
c
k
Y
O
L
O
1
1
s
I
mag
e
4
2
5
.
0
0
4
,
4
4
5
.
0
0
1
0
.
3
0
4
,
4
8
0
.
3
0
1
p
e
r
so
n
,
3
0
c
a
r
s
,
a
n
d
5
t
r
u
c
k
s
Y
O
L
O
1
1
s
I
mag
e
5
1
6
.
0
0
5
,
1
3
0
.
8
0
1
0
.
2
0
5
,
1
5
7
.
0
0
7
p
e
r
so
n
s,
2
c
a
r
s,
a
n
d
1
h
a
n
d
b
a
g
Y
O
L
O
1
1
m
I
mag
e
1
3
6
.
2
0
1
6
,
9
5
2
.
7
0
1
0
.
6
0
1
6
,
9
9
9
.
5
0
8
p
e
r
so
n
s
a
n
d
1
t
i
e
Y
O
L
O
1
1
m
I
mag
e
2
4
0
.
8
0
1
6
,
8
6
8
.
2
0
1
0
.
9
0
1
6
,
9
1
9
.
9
0
1
5
p
e
r
so
n
s,
3
c
a
r
s
,
3
t
r
a
f
f
i
c
l
i
g
h
t
s
,
a
n
d
5
h
a
n
d
b
a
g
s
Y
O
L
O
1
1
m
I
mag
e
3
2
5
.
0
0
1
4
,
6
2
5
.
4
0
9
.
8
0
1
4
,
6
6
0
.
2
0
5
p
e
r
so
ns
Y
O
L
O
1
1
m
I
mag
e
4
2
5
.
3
0
1
4
,
7
7
1
.
4
0
1
0
.
8
0
1
4
,
8
0
7
.
5
0
1
p
e
r
so
n
,
2
6
c
a
r
s
,
a
n
d
5
t
r
u
c
k
s
Y
O
L
O
1
1
m
I
mag
e
5
1
6
.
7
0
1
6
,
8
8
5
.
9
0
1
0
.
8
0
1
6
,
9
1
3
.
4
0
7
p
e
r
so
n
s,
2
c
a
r
s,
1
b
a
c
k
p
a
c
k
,
a
n
d
1
b
a
se
b
a
l
l
g
l
o
v
e
Y
O
L
O
1
1
l
I
mag
e
1
3
5
.
5
0
2
1
,
2
0
1
.
7
0
1
0
.
3
0
2
1
,
2
4
7
.
5
0
8
p
e
r
so
n
s
a
n
d
1
t
i
e
Y
O
L
O
1
1
l
I
mag
e
2
3
9
.
2
0
2
1
,
3
2
6
.
4
0
1
1
.
1
0
2
1
,
3
7
6
.
7
0
1
6
p
e
r
so
n
s,
3
c
a
r
s
,
3
t
r
a
f
f
i
c
l
i
g
h
t
s
,
a
n
d
5
h
a
n
d
b
a
g
s
Y
O
L
O
1
1
l
I
mag
e
3
2
5
.
5
0
1
8
,
5
0
0
.
1
0
9
.
9
0
1
8
,
5
3
5
.
5
0
6
p
e
r
so
n
s
a
n
d
1
b
a
c
k
p
a
c
k
Y
O
L
O
1
1
l
I
mag
e
4
2
5
.
6
0
1
8
,
4
0
6
.
1
0
1
0
.
4
0
1
8
,
4
4
2
.
1
0
1
p
e
r
so
n
,
2
5
c
a
r
s
,
6
t
r
u
c
k
s,
a
n
d
2
t
r
a
f
f
i
c
l
i
g
h
t
s
Y
O
L
O
1
1
l
I
mag
e
5
1
7
.
0
0
2
1
,
2
5
7
.
1
0
1
0
.
4
0
2
1
,
2
8
4
.
5
0
7
p
e
r
so
n
s
a
n
d
2
c
a
r
s
Y
O
L
O
1
1
x
I
mag
e
1
3
5
.
8
0
4
2
,
6
7
2
.
2
0
1
0
.
3
0
4
2
,
7
1
8
.
3
0
8
p
e
r
so
n
s
a
n
d
1
t
i
e
Y
O
L
O
1
1
x
I
mag
e
2
3
6
.
6
0
4
2
,
7
0
1
.
1
0
1
0
.
8
0
4
2
,
7
4
8
.
5
0
1
4
p
e
r
s
o
n
s
,
3
c
a
r
s
,
1
t
r
a
f
f
i
c
l
i
g
h
t
.
4
h
a
n
d
b
a
g
s
, an
d
1
c
u
p
Y
O
L
O
1
1
x
I
mag
e
3
2
5
.
6
0
3
6
,
7
4
1
.
2
0
9
.
7
0
3
6
,
7
7
6
.
5
0
6
p
e
r
so
n
s,
1
t
r
u
c
k
,
a
n
d
1
d
i
n
i
n
g
t
a
b
l
e
Y
O
L
O
1
1
x
I
mag
e
4
2
4
.
9
0
3
6
,
7
5
7
.
6
0
1
0
.
0
0
3
6
,
7
9
2
.
5
0
1
p
e
r
so
n
,
2
1
c
a
r
s
,
6
t
r
u
c
k
s,
a
n
d
2
t
r
a
f
f
i
c
l
i
g
h
t
s
Y
O
L
O
1
1
x
I
mag
e
5
1
6
.
5
0
4
3
,
1
4
4
.
6
0
1
0
.
3
0
4
3
,
1
7
1
.
4
0
7
p
e
r
so
n
s,
2
c
a
r
s,
1
a
i
r
p
l
a
n
e
,
a
n
d
1
t
i
e
(
b
)
R
K3
5
6
6
NP
U
-
b
ased
in
f
er
e
n
ce
u
s
in
g
R
KNN
M
o
d
e
l
P
i
c
t
u
r
e
P
r
e
p
r
o
c
e
ss
(
ms)
I
n
f
e
r
e
n
c
e
(
ms)
P
o
st
p
r
o
c
e
ss
(
ms)
T
o
t
a
l
t
i
me
(
ms)
D
e
t
e
c
t
i
o
n
r
e
su
l
t
Y
O
L
O
1
1
n
I
mag
e
1
3
5
.
9
0
2
6
6
.
1
0
1
2
.
9
0
3
1
4
.
9
0
8
p
e
r
s
o
n
s
Y
O
L
O
1
1
n
I
mag
e
2
4
9
.
0
0
2
7
0
.
1
0
1
3
.
1
0
3
3
2
.
2
0
1
2
p
e
r
s
o
n
s
,
3
c
a
r
s
,
a
n
d
4
t
r
a
f
f
i
c
l
i
g
h
t
s
Y
O
L
O
1
1
n
I
mag
e
3
3
3
.
9
0
2
7
3
.
5
0
1
4
.
5
0
3
2
1
.
9
0
6
p
e
r
s
o
n
s
a
n
d
2
t
r
u
c
k
s
Y
O
L
O
1
1
n
I
mag
e
4
3
3
.
9
0
2
7
1
.
7
0
1
3
.
4
0
3
1
9
.
0
0
2
2
c
a
r
s
a
n
d
7
t
r
u
c
k
s
Y
O
L
O
1
1
n
I
mag
e
5
2
1
.
2
0
2
8
0
.
1
0
1
4
.
4
0
3
1
5
.
7
0
7
p
e
r
s
o
n
s
a
n
d
4
c
a
r
s
Y
O
L
O
1
1
n
I
mag
e
1
4
2
.
5
0
4
2
9
.
4
0
1
3
.
0
0
4
8
4
.
9
0
8
p
e
r
s
o
n
s
Y
O
L
O
1
1
n
I
mag
e
2
4
8
.
5
0
4
3
4
.
3
0
1
3
.
4
0
4
9
6
.
2
0
1
2
p
e
r
s
o
n
s
,
2
c
a
r
s
,
1
t
r
a
f
f
i
c
l
i
g
h
t
,
3
h
a
n
d
b
a
g
s
,
1
t
i
e
,
a
n
d
1
c
u
p
Y
O
L
O
1
1
n
I
mag
e
3
3
3
.
6
0
4
3
7
.
4
0
1
3
.
2
0
4
8
4
.
2
0
8
p
e
r
s
o
n
s
a
n
d
2
t
r
u
c
k
s
Y
O
L
O
1
1
n
I
mag
e
4
3
4
.
1
0
4
3
5
.
9
0
1
3
.
9
0
4
8
3
.
9
0
1
p
e
r
s
o
n
,
2
9
c
a
r
s
,
a
n
d
4
t
r
u
c
k
s
Y
O
L
O
1
1
n
I
mag
e
5
2
1
.
4
0
4
4
4
.
1
0
1
3
.
1
0
4
7
8
.
6
0
7
p
e
r
s
o
n
s
,
2
c
a
r
s
,
a
n
d
1
h
a
n
d
b
a
g
Y
O
L
O
1
1
m
I
mag
e
1
3
7
.
4
0
9
4
6
.
6
0
1
3
.
0
0
9
9
7
.
0
0
8
p
e
r
s
o
n
s
a
n
d
1
t
i
e
Y
O
L
O
1
1
m
I
mag
e
2
4
9
.
0
0
9
5
2
.
2
0
1
3
.
9
0
1
,
0
1
5
.
1
0
1
7
p
e
r
s
o
n
s
,
3
c
a
r
s
,
4
t
r
a
f
f
i
c
l
i
g
h
t
s
,
a
n
d
5
h
a
n
d
b
a
g
s
Y
O
L
O
1
1
m
I
mag
e
3
3
3
.
4
0
9
5
5
.
4
0
1
3
.
0
0
1
,
0
0
1
.
8
0
5
p
e
r
s
o
n
s
Y
O
L
O
1
1
m
I
mag
e
4
3
4
.
1
0
9
5
6
.
1
0
1
4
.
1
0
1
,
0
0
4
.
3
0
1
p
e
r
s
o
n
,
2
9
c
a
r
s
,
a
n
d
5
t
r
u
c
k
s
Y
O
L
O
1
1
m
I
mag
e
5
2
0
.
7
0
9
5
9
.
9
0
1
3
.
0
0
9
9
3
.
6
0
7
p
e
r
s
o
n
s
,
2
c
a
r
s
,
1
b
a
c
k
p
a
c
k
,
a
n
d
1
b
a
s
e
b
a
l
l
g
l
o
v
e
Y
O
L
O
1
1
l
I
mag
e
1
3
5
.
8
0
1
,
1
5
3
.
1
0
1
3
.
0
0
1
,
2
0
1
.
9
0
8
p
e
r
s
o
n
s
a
n
d
1
t
i
e
Y
O
L
O
1
1
l
I
mag
e
2
4
9
.
0
0
1
,
1
5
8
.
5
0
1
3
.
7
0
1
,
2
2
1
.
2
0
1
7
p
e
r
s
o
n
s
,
4
c
a
r
s
,
3
t
r
a
f
f
i
c
l
i
g
h
t
s
,
a
n
d
5
h
a
n
d
b
a
g
s
Y
O
L
O
1
1
l
I
mag
e
3
3
3
.
6
0
1
,
1
6
0
.
5
0
1
4
.
1
0
1
,
2
0
8
.
2
0
7
p
e
r
s
o
n
s
Y
O
L
O
1
1
l
I
mag
e
4
3
3
.
9
0
1
,
1
6
0
.
4
0
1
3
.
7
0
1
,
2
0
8
.
0
0
1
p
e
r
s
o
n
,
2
9
c
a
r
s
,
7
t
r
u
c
k
s
,
a
n
d
2
t
r
a
f
f
i
c
l
i
g
h
t
s
Y
O
L
O
1
1
l
I
mag
e
5
2
1
.
1
0
1
,
1
6
8
.
8
0
1
3
.
2
0
1
,
2
0
3
.
1
0
7
p
e
r
s
o
n
s
a
n
d
2
c
a
r
s
Y
O
L
O
1
1
x
I
mag
e
1
4
2
.
5
0
2
,
5
2
5
.
1
0
1
3
.
0
0
2
,
5
8
0
.
6
0
8
p
e
r
s
o
n
s
a
n
d
1
t
i
e
Y
O
L
O
1
1
x
I
mag
e
2
4
9
.
1
0
2
,
6
4
2
.
0
0
1
3
.
5
0
2
,
7
0
4
.
6
0
1
4
p
e
r
s
o
n
s
,
3
c
a
r
s
,
1
t
r
a
f
f
i
c
l
i
g
h
t
,
a
n
d
5
h
a
n
d
b
a
g
s
Y
O
L
O
1
1
x
I
mag
e
3
3
3
.
8
0
2
,
6
4
6
.
8
0
1
3
.
2
0
2
,
6
9
3
.
8
0
6
p
e
r
s
o
n
s
,
1
t
r
u
c
k
,
a
n
d
1
d
i
n
i
n
g
t
a
b
l
e
Y
O
L
O
1
1
x
I
mag
e
4
3
3
.
7
0
2
,
6
4
4
.
2
0
1
3
.
5
0
2
,
6
9
1
.
4
0
1
p
e
r
s
o
n
,
2
2
c
a
r
s
,
6
t
r
u
c
k
s
,
a
n
d
2
t
r
a
f
f
i
c
l
i
g
h
t
s
Y
O
L
O
1
1
x
I
mag
e
5
2
1
.
4
0
2
,
6
1
2
.
4
0
1
3
.
0
0
2
,
6
4
6
.
8
0
7
p
e
r
s
o
n
s
a
n
d
2
c
a
r
s
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
1
6
9
3
-
6930
T
E
L
KOM
NI
K
A
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l
C
o
n
tr
o
l
,
Vo
l.
24
,
No
.
1
,
Feb
r
u
ar
y
20
26
:
1
26
-
1
41
132
Af
ter
r
ec
o
r
d
in
g
p
r
ep
r
o
ce
s
s
in
g
,
in
f
er
en
ce
,
p
o
s
tp
r
o
ce
s
s
i
n
g
,
an
d
to
tal
elap
s
ed
ti
m
e
f
o
r
ea
c
h
i
m
ag
e,
we
co
m
p
u
ted
th
e
m
ea
n
f
o
r
ea
ch
Y
OL
O1
1
v
ar
ia
n
t.
T
h
e
r
esu
l
ts
ar
e
r
ep
o
r
ted
f
o
r
b
o
th
C
P
U
an
d
N
P
U
ex
ec
u
tio
n
s
.
T
h
e
d
ata
w
er
e
o
r
g
an
ized
b
y
m
o
d
el
ty
p
e
a
n
d
p
r
o
ce
s
s
in
g
p
at
h
,
p
r
o
v
id
in
g
an
i
n
ter
p
r
etab
le
v
ie
w
o
f
wo
r
k
lo
ad
tr
an
s
latio
n
f
r
o
m
g
en
er
al
-
p
u
r
p
o
s
e
to
h
ar
d
w
ar
e
-
ac
ce
ler
ated
en
v
ir
o
n
m
en
t
s
.
T
h
e
ar
it
h
m
etic
m
ea
n
f
o
r
ea
ch
p
ip
elin
e
s
ta
g
e
w
as
ca
lc
u
lated
u
s
i
n
g
th
e
s
tan
d
ar
d
d
ef
i
n
itio
n
,
ex
p
r
ess
ed
(
1
)
-
(
4
)
:
.
=
∑
(
)
(
1
)
.
=
∑
(
)
(
2
)
.
=
∑
(
)
(
3
)
.
=
∑
(
)
(
4
)
T
h
e
ev
a
lu
a
ti
o
n
r
e
s
u
l
t
s
o
f
YOL
O
1
1
o
n
C
PU
an
d
N
PU
a
r
e
p
r
e
s
en
t
e
d
in
T
a
b
l
e
2
an
d
F
ig
u
r
e
s
6
(
a
)
–
(
d
)
,
w
h
ic
h
c
o
m
p
a
r
e
ex
e
cu
ti
o
n
tim
es
f
o
r
p
r
e
p
r
o
c
e
s
s
in
g
[
3
2
]
,
i
n
f
e
r
en
c
e
[
3
3
]
,
[
3
4
]
,
a
n
d
p
o
s
t
p
r
o
c
e
s
s
i
n
g
[
3
5
]
.
T
h
e
N
P
U
c
o
n
s
i
s
t
e
n
t
l
y
o
u
t
p
e
r
f
o
r
m
e
d
t
h
e
C
P
U
,
p
a
r
t
i
c
u
l
a
r
l
y
d
u
r
i
n
g
i
n
f
e
r
e
n
c
e
,
w
h
e
r
e
h
a
r
d
w
a
r
e
a
c
c
e
l
e
r
a
t
i
o
n
p
l
a
y
e
d
t
h
e
d
o
m
i
n
a
n
t
r
o
l
e
.
F
o
r
Y
O
L
O
1
1
n
,
i
n
f
e
r
e
n
c
e
t
i
m
e
d
e
c
r
e
a
s
e
d
f
r
o
m
1
,
5
4
2
.
9
6
m
s
o
n
t
h
e
C
P
U
t
o
2
7
2
.
3
0
m
s
o
n
t
h
e
N
P
U
,
r
e
p
r
e
s
e
n
t
i
n
g
a
n
8
2
.
4
%
r
e
d
u
c
t
i
o
n
.
R
eg
ar
d
i
n
g
lar
g
er
m
o
d
els,
f
o
r
YO
L
O1
1
l
i
t
d
r
o
p
p
ed
f
r
o
m
2
0
,
1
3
8
.
2
8
to
1
,
1
6
0
.
2
6
m
s
(
-
9
4
.
2
%),
w
h
er
ea
s
f
o
r
YOL
O1
1
x
,
t
h
e
la
r
g
est
v
ar
ian
t,
it
d
ec
r
ea
s
ed
f
r
o
m
4
0
,
4
0
3
.
3
4
to
2
,
6
1
4
.
1
0
m
s
(
-
9
3
.
5
%).
E
n
d
-
to
-
en
d
to
tal
ti
m
e
also
f
ell
s
h
ar
p
l
y
:
f
o
r
YOL
O1
1
s
,
it
d
ec
r
ea
s
ed
f
r
o
m
4
,
8
9
7
.
7
6
to
4
8
5
.
5
6
m
s
(
-
9
0
.
1
%),
an
d
f
o
r
YOL
O1
1
m
,
i
t
d
r
o
p
p
e
d
f
r
o
m
1
6
,
0
6
0
.
1
0
to
1
,
0
0
2
.
3
6
m
s
(
-
9
3
.
8
%).
L
aten
c
y
d
i
f
f
er
e
n
ce
s
i
n
p
r
ep
r
o
ce
s
s
in
g
an
d
p
o
s
tp
r
o
ce
s
s
in
g
w
e
r
e
r
elati
v
el
y
s
m
all,
i
n
d
icati
n
g
t
h
at
i
n
ter
f
ac
e
o
v
er
h
ea
d
an
d
o
p
ti
m
izatio
n
r
o
u
tin
e
s
h
ad
m
in
i
m
al
i
m
p
ac
t
o
n
t
h
ese
p
h
a
s
es.
I
n
f
er
e
n
ce
th
u
s
r
e
m
a
i
n
ed
th
e
d
o
m
i
n
a
n
t
co
m
p
o
n
e
n
t
o
f
o
v
er
all
laten
c
y
an
d
b
en
e
f
ited
m
o
s
t
f
r
o
m
NP
U
ac
ce
ler
atio
n
.
F
ig
u
r
e
6
(
d
)
p
r
o
v
i
d
e
s
a
h
e
atm
a
p
o
f
ex
e
cu
ti
o
n
t
im
e
p
a
tt
e
r
n
s
ac
r
o
s
s
a
ll
v
a
r
i
an
t
s
(
,
,
,
,
a
n
d
)
,
w
h
e
r
e
d
a
r
k
e
r
s
h
a
d
es
c
o
r
r
e
s
p
o
n
d
t
o
s
h
o
r
te
r
d
u
r
a
t
i
o
n
s
,
p
ar
t
i
cu
l
a
r
ly
in
N
PU
r
u
n
s
.
N
PUs
en
a
b
l
e
a
c
c
el
e
r
a
t
i
o
n
o
f
u
p
t
o
7
2
4
×
c
o
m
p
a
r
e
d
w
ith
s
o
f
tw
a
r
e
-
o
n
ly
m
ic
r
o
c
o
n
t
r
o
l
le
r
im
p
l
em
en
t
at
i
o
n
s
,
a
s
r
e
p
o
r
te
d
b
y
M
an
o
r
an
d
G
r
e
e
n
b
e
r
g
[
3
6
]
;
o
u
r
r
e
s
u
l
ts
c
o
r
r
o
b
o
r
a
t
e
t
h
is
,
d
em
o
n
s
t
r
a
tin
g
s
u
b
s
t
an
t
i
al
in
f
e
r
en
c
e
l
a
te
n
cy
r
ed
u
c
t
i
o
n
a
t
th
e
e
d
g
e
.
E
n
e
r
g
y
c
o
n
s
u
m
p
ti
o
n
w
as
m
ea
s
u
r
e
d
in
r
e
a
l
t
im
e
w
ith
a
d
ig
i
t
al
a
n
a
ly
ze
r
th
at
l
o
g
g
e
d
i
n
s
t
an
t
an
eo
u
s
v
o
l
t
ag
e
,
cu
r
r
e
n
t
,
a
n
d
p
o
w
e
r
d
u
r
in
g
m
o
d
el
ex
e
cu
t
i
o
n
o
n
th
e
O
r
an
g
e
Pi
3
B
(
R
K
3
5
6
6
)
,
e
n
s
u
r
in
g
a
c
cu
r
a
te
c
a
l
cu
l
at
i
o
n
s
.
T
ab
le
2
.
A
v
er
ag
e
ti
m
e
ca
lc
u
la
tio
n
r
esu
l
ts
P
r
o
c
e
ssi
n
g
s
t
a
g
e
Y
O
L
O
1
1
n
Y
O
L
O
1
1
s
Y
O
L
O
1
1
m
Y
O
L
O
1
1
l
Y
O
L
O
1
1
x
C
P
U
N
P
U
C
P
U
N
P
U
C
P
U
N
P
U
C
P
U
N
P
U
C
P
U
N
P
U
P
r
e
p
r
o
c
e
ss (m
s)
3
2
.
0
8
3
4
.
7
8
2
8
.
9
2
3
6
.
0
2
2
8
.
8
0
3
4
.
9
2
2
8
.
5
6
3
4
.
6
8
2
7
.
8
8
3
6
.
1
0
I
n
f
e
r
e
n
c
e
(
ms)
1
,
5
4
2
.
9
6
2
7
2
.
3
0
4
,
8
5
8
.
6
0
4
3
6
.
2
2
1
6
,
0
2
0
.
7
2
9
5
4
.
0
4
2
0
,
1
3
8
.
2
8
1
,
1
6
0
.
2
6
4
0
,
4
0
3
.
3
4
2
,
6
1
4
.
1
0
P
o
st
p
r
o
c
e
ss
(
ms)
1
2
.
1
0
1
3
.
6
6
1
0
.
2
4
1
3
.
3
2
1
0
.
5
8
1
3
.
4
0
1
0
.
4
2
1
3
.
5
4
1
0
.
2
2
1
3
.
2
4
T
o
t
a
l
t
i
me
(
ms)
1
,
5
8
7
.
1
4
3
2
0
.
7
4
4
,
8
9
7
.
7
6
4
8
5
.
5
6
1
6
,
0
6
0
.
1
0
1
,
0
0
2
.
3
6
2
0
,
1
7
7
.
2
6
1
,
2
0
8
.
4
8
4
0
,
4
4
1
.
4
4
2
,
6
6
3
.
4
4
E
lap
s
ed
ti
m
e
w
a
s
m
ea
s
u
r
ed
with
i
n
t
h
e
e
x
ec
u
tio
n
s
cr
ip
t,
w
i
th
ea
ch
co
n
d
itio
n
r
ep
ea
ted
s
ev
er
al
ti
m
e
s
to
co
n
f
ir
m
r
ep
ea
tab
ilit
y
a
n
d
ac
c
u
r
ac
y
.
T
h
e
r
esu
lts
ar
e
s
u
m
m
a
r
ized
in
co
m
p
ar
ativ
e
tab
le
s
c
o
n
tr
asti
n
g
C
P
U
an
d
NP
U
ex
ec
u
tio
n
s
,
f
r
o
m
w
h
ich
m
ea
n
v
al
u
es
w
er
e
e
x
tr
ac
ted
,
h
i
g
h
l
ig
h
ti
n
g
th
e
s
p
ee
d
g
ain
s
f
r
o
m
NP
U
ac
ce
ler
atio
n
an
d
r
ed
u
c
tio
n
s
i
n
to
ta
l p
r
o
ce
s
s
in
g
t
i
m
e.
A
lt
h
o
u
g
h
NP
U
ac
ce
ler
atio
n
d
r
asti
ca
ll
y
r
ed
u
ce
d
in
f
er
en
ce
ti
m
e,
p
r
e
p
r
o
ce
s
s
in
g
[
3
7
]
an
d
p
o
s
tp
r
o
ce
s
s
in
g
laten
cie
s
in
cr
ea
s
ed
m
o
d
estl
y
(
ap
p
r
o
x
i
m
atel
y
8
–
3
0
%)
co
m
p
ar
ed
w
ith
t
h
o
s
e
i
n
C
P
U
-
o
n
l
y
e
x
ec
u
tio
n
.
P
o
s
tp
r
o
ce
s
s
in
g
,
d
o
m
i
n
ated
b
y
C
P
U
-
e
x
ec
u
ted
NM
S
w
it
h
q
u
ad
r
a
tic
co
m
p
lex
it
y
,
r
e
m
ai
n
e
d
s
m
all
i
n
ab
s
o
l
u
te
ter
m
s
r
elati
v
e
to
th
e
i
n
f
er
en
ce
s
p
ee
d
u
p
.
T
h
e
in
cr
ea
s
e
ar
is
es
b
ec
au
s
e
o
n
l
y
i
n
f
er
en
ce
is
o
f
f
lo
ad
ed
to
th
e
R
K3
5
6
6
NP
U,
w
h
er
ea
s
p
r
ep
r
o
ce
s
s
in
g
(
r
esize,
n
o
r
m
aliza
tio
n
,
r
ef
o
r
m
atti
n
g
)
[
3
7
]
an
d
p
o
s
tp
r
o
ce
s
s
i
n
g
(
n
o
n
-
m
ax
i
m
u
m
s
u
p
p
r
ess
io
n
,
b
o
u
n
d
in
g
-
b
o
x
ad
j
u
s
t
m
e
n
t)
[
3
8
]
-
[
4
0
]
r
em
ai
n
o
n
th
e
C
P
U.
Ho
s
t
-
de
v
ice
te
n
s
o
r
t
r
an
s
f
er
s
(
C
P
U↔NP
U)
in
cl
u
d
in
g
i
n
p
u
t
u
p
lo
ad
an
d
o
u
tp
u
t
d
o
w
n
lo
ad
in
tr
o
d
u
ce
in
ter
f
ac
e
la
ten
c
y
,
m
e
m
o
r
y
co
p
ies,
an
d
f
o
r
m
at
co
n
v
er
s
io
n
s
t
h
at
ac
co
u
n
t
f
o
r
m
o
s
t
o
f
th
e
r
esid
u
al
o
v
er
h
ea
d
(
co
s
t)
.
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KOM
NI
K
A
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l
C
o
n
tr
o
l
P
erfo
r
ma
n
ce
en
h
a
n
ce
men
t o
f
emb
ed
d
ed
o
b
ject
d
etec
tio
n
via
n
eu
r
a
l h
a
r
d
w
a
r
e
..
(
A
lw
in
Ha
r
t
o
n
o
Lima
r
a
n
)
133
D
e
s
p
i
te
th
es
e
t
r
a
d
e
-
o
f
f
s
,
d
e
t
ec
ti
o
n
q
u
a
li
ty
is
p
r
es
e
r
v
e
d
.
T
a
b
l
e
3
c
o
m
p
a
r
es
m
A
P
@
0
.
5
:
0
.
9
5
b
e
tw
e
en
th
e
o
f
f
i
c
ia
l
b
en
ch
m
a
r
k
an
d
th
e
O
r
an
g
e
Pi
3
B
r
e
s
u
lt
s
.
T
h
e
5
0
0
-
im
ag
e
s
u
b
s
e
t sh
o
w
s
l
a
r
g
e
r
f
lu
ctu
a
ti
o
n
s
(
Δ
≈
-
0
.
0
1
0
t
o
-
0
.
0
7
4
)
,
w
h
e
r
ea
s
th
e
f
u
ll
5
,
0
0
0
-
im
ag
e
ev
alu
a
t
i
o
n
is
c
o
n
s
is
t
en
t,
w
ith
d
e
v
i
a
t
i
o
n
s
o
f
a
p
p
r
o
x
im
a
te
ly
-
0
.
0
2
6
t
o
-
0
.
0
3
0
acr
o
s
s
Y
OL
O
1
1
v
a
r
i
an
t
s
.
T
h
es
e
f
in
d
in
g
s
c
o
n
f
i
r
m
th
at
R
K
NN
c
o
n
v
e
r
s
i
o
n
p
r
e
s
e
r
v
e
d
b
as
e
l
in
e
a
c
cu
r
a
cy
w
i
th
in
a
n
a
r
r
o
w
m
a
r
g
in
o
f
th
e
Ul
t
r
aly
t
i
cs
b
e
n
c
h
m
a
r
k
w
h
i
l
e
d
el
iv
e
r
in
g
s
u
b
s
t
an
t
i
al
ef
f
i
c
ie
n
cy
im
p
r
o
v
em
e
n
ts
[
4
1
]
.
(
a)
(
b
)
(
c)
(
d
)
Fig
u
r
e
6
.
C
P
U
v
s
.
NP
U
p
er
f
o
r
m
an
ce
co
m
p
ar
i
s
o
n
: (
a)
p
r
ep
r
o
ce
s
s
i
n
g
,
(
b
)
in
f
er
en
ce
,
(
c)
p
o
s
tp
r
o
ce
s
s
in
g
,
a
n
d
(
d
)
ex
ec
u
tio
n
ti
m
e
h
ea
t
m
ap
ac
r
o
s
s
YOL
O1
1
v
ar
ian
t
s
.
X
-
a
x
is
=
m
o
d
el
v
ar
ia
n
t (
YO
L
O1
1
n
–
YOL
O1
1
x
)
;
Y
-
a
x
is
=
e
x
ec
u
tio
n
ti
m
e
(
m
s
)
T
ab
le
3
.
C
o
m
p
ar
is
o
n
o
f
YO
L
O1
1
m
A
P
r
esu
lts
(
o
f
f
icial
U
ltra
l
y
tic
s
v
s
.
Or
a
n
g
e
P
i 3
B
,
C
OC
O2
0
1
7
v
al2
0
1
7
)
M
o
d
e
l
O
f
f
i
c
i
a
l
mA
P
5
0
–
95
O
n
-
d
e
v
i
c
e
mA
P
(
5
0
0
)
O
n
-
d
e
v
i
c
e
mA
P
(
5
,
0
0
0
)
Δ o
n
-
d
e
v
i
c
e
-
500
Δ o
n
-
d
e
v
i
c
e
-
5
,
0
0
0
Y
O
L
O
1
1
n
0
.
3
9
5
0
.
3
6
1
0
.
3
6
9
–
0
.
0
3
4
–
0
.
0
2
6
Y
O
L
O
1
1
s
0
.
4
7
0
0
.
4
2
7
0
.
4
4
1
–
0
.
0
4
3
–
0
.
0
2
9
Y
O
L
O
1
1
m
0
.
5
1
5
0
.
5
0
5
0
.
4
8
7
–
0
.
0
1
0
–
0
.
0
2
8
Y
O
L
O
1
1
l
0
.
5
3
4
0
.
4
8
8
0
.
5
0
4
–
0
.
0
4
6
–
0
.
0
3
0
Y
O
L
O
1
1
x
0
.
5
4
7
0
.
4
7
3
0
.
5
1
7
–
0
.
0
7
4
–
0
.
0
3
0
N
o
t
e
:
o
n
-
d
e
v
i
c
e
mA
P
r
e
f
e
r
s
t
o
t
h
e
a
c
c
u
r
a
c
y
me
a
su
r
e
d
d
i
r
e
c
t
l
y
o
n
t
h
e
O
r
a
n
g
e
P
i
3
B
p
l
a
t
f
o
r
m,
w
h
e
r
e
a
s
t
h
e
o
f
f
i
c
i
a
l
mA
P
d
e
n
o
t
e
s
t
h
e
b
e
n
c
h
mar
k
r
e
p
o
r
t
e
d
b
y
U
l
t
r
a
l
y
t
i
c
s o
n
t
h
e
C
O
C
O
2
0
1
7
v
a
l
i
d
a
t
i
o
n
se
t
.
Af
ter
co
n
f
ir
m
i
n
g
co
n
s
i
s
te
n
t
d
etec
tio
n
ac
cu
r
ac
y
ac
r
o
s
s
YOL
O1
1
v
ar
ia
n
ts
w
it
h
m
in
i
m
al
m
AP
d
ev
iatio
n
s
(
-
0
.
0
2
6
to
-
0
.
0
3
0
)
,
w
e
e
x
a
m
in
ed
e
n
er
g
y
ef
f
icie
n
c
y
.
T
ab
le
4
p
r
esen
ts
n
ea
r
l
y
id
en
tical
av
er
ag
e
p
o
w
er
co
n
s
u
m
p
tio
n
f
o
r
C
P
U
(
3
.
6
0
W
)
an
d
NP
U
(
3
.
5
9
W
)
,
d
if
f
er
in
g
b
y
le
s
s
th
a
n
0
.
3
%.
Ho
w
e
v
er
,
th
e
m
u
c
h
lo
w
er
in
f
er
en
ce
late
n
c
y
o
f
t
h
e
NP
U
(
8
2
–
9
4
%
f
aster
)
s
h
ar
p
l
y
r
ed
u
ce
d
en
er
g
y
p
er
in
f
er
en
ce
:
YOL
O1
1
x
co
n
s
u
m
ed
1
4
5
.
5
9
J
o
n
th
e
C
P
U
v
er
s
u
s
9
.
5
6
J
o
n
th
e
NP
U,
a
9
3
.
4
%
r
ed
u
ctio
n
.
T
h
ese
r
es
u
lt
s
co
n
f
ir
m
t
h
e
NP
U
as
th
e
m
o
s
t
en
er
g
y
-
e
f
f
icie
n
t p
r
o
ce
s
s
o
r
f
o
r
YOL
O1
1
w
o
r
k
lo
ad
s
o
n
t
h
e
Or
an
g
e
P
i 3
B
.
NP
Us
allo
w
ed
g
e
d
e
v
ice
s
s
u
c
h
as
t
h
e
Or
a
n
g
e
P
i
3
B
to
p
er
f
o
r
m
s
u
b
s
tan
t
iall
y
m
o
r
e
i
n
f
er
e
n
ce
p
ass
e
s
w
it
h
i
n
t
h
e
s
a
m
e
p
o
w
er
b
u
d
g
et
,
an
ad
v
an
ta
g
e
f
o
r
p
latf
o
r
m
s
p
o
w
er
ed
b
y
b
atter
ies o
r
o
th
er
lo
w
-
w
atta
g
e
s
o
u
r
ce
s
.
T
h
e
b
en
ef
it
co
m
es
n
o
t
o
n
l
y
f
r
o
m
t
h
r
o
u
g
h
p
u
t
b
u
t
al
s
o
f
r
o
m
s
h
ar
p
l
y
r
ed
u
c
ed
laten
c
y
at
n
ea
r
l
y
co
n
s
ta
n
t
a
v
er
ag
e
p
o
w
er
.
T
h
is
en
h
a
n
ce
s
t
h
e
f
e
asib
ilit
y
o
f
en
er
g
y
-
i
n
ten
s
i
v
e
A
I
w
o
r
k
lo
ad
s
an
d
s
u
p
p
o
r
ts
n
ea
r
-
r
ea
l
-
ti
m
e
d
ee
p
lear
n
in
g
i
n
lo
w
-
p
o
w
er
n
et
w
o
r
k
s
.
Ver
if
icatio
n
co
n
f
ir
m
ed
th
at
ac
ce
ler
atio
n
o
n
th
e
R
K3
5
6
6
NPU
d
id
n
o
t
d
eg
r
ad
e
d
et
ec
tio
n
q
u
alit
y
[
4
2
]
.
E
v
alu
a
tio
n
s
u
s
ed
t
h
e
C
OC
O2
0
1
7
v
alid
atio
n
s
et
(
v
al2
0
1
7
)
.
T
h
e
o
f
f
icia
l
YO
L
O1
1
r
es
u
lt
s
wer
e
b
ased
o
n
t
h
e
f
u
ll
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
1
6
9
3
-
6930
T
E
L
KOM
NI
K
A
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l
C
o
n
tr
o
l
,
Vo
l.
24
,
No
.
1
,
Feb
r
u
ar
y
20
26
:
1
26
-
1
41
134
5
,
0
0
0
-
im
a
g
e
s
et,
w
h
er
ea
s
Or
a
n
g
e
P
i
3
B
test
s
e
m
p
lo
y
ed
b
o
t
h
a
r
a
n
d
o
m
5
0
0
-
i
m
ag
e
s
u
b
s
et
an
d
t
h
e
f
u
ll
s
e
t
o
f
5
,
0
0
0
im
a
g
es to
e
n
s
u
r
e
co
n
s
i
s
t
en
c
y
.
T
ab
les
4
(
a)
an
d
(
b
)
d
etail
th
e
r
ec
o
r
d
ed
v
o
ltag
e,
cu
r
r
en
t
,
an
d
p
o
w
er
d
u
r
in
g
d
etec
t
io
n
.
T
ab
le
5
th
e
n
co
n
s
o
lid
ates
late
n
c
y
,
s
p
ee
d
u
p
,
p
er
-
in
f
er
e
n
ce
en
er
g
y
,
a
n
d
ac
cu
r
ac
y
d
ev
ia
tio
n
ac
r
o
s
s
YO
L
O1
1
v
ar
ian
ts
.
T
h
e
r
esu
lt
s
s
h
o
w
t
h
at
av
er
a
g
e
p
o
w
er
co
n
s
u
m
p
tio
n
w
a
s
n
ea
r
l
y
id
en
tical
b
et
w
ee
n
C
P
U
an
d
NP
U
ex
ec
u
tio
n
s
.
Ho
w
e
v
er
,
th
e
NP
U
co
n
s
i
s
te
n
t
l
y
ac
h
ie
v
ed
5
–
1
7
×
f
aster
in
f
e
r
en
ce
an
d
~8
0
–
9
4
%
lo
w
er
en
er
g
y
p
er
in
f
er
en
ce
[
4
3
]
,
[
4
4
]
,
w
h
i
le
m
ai
n
tai
n
i
n
g
ac
cu
r
ac
y
w
it
h
i
n
-
0
.
0
3
m
A
P
o
f
th
e
o
f
f
icial
b
aseli
n
e.
E
n
er
g
y
p
er
in
f
er
en
ce
w
a
s
ca
lcu
lated
as a
v
er
ag
e
p
o
w
er
m
u
ltip
lied
b
y
to
tal
late
n
c
y
.
T
ab
le
4
.
R
ec
o
r
d
e
d
v
o
ltag
e,
cu
r
r
en
t
,
an
d
p
o
w
er
co
n
s
u
m
p
tio
n
d
u
r
in
g
th
e
YO
L
O1
1
ex
ec
u
tio
n
o
n
t
h
e
t
w
o
p
latf
o
r
m
s
(
a)
C
P
U
-
b
ased
(
b
)
NP
U
-
b
ased
M
o
d
e
l
P
i
c
t
u
r
e
A
v
e
r
a
g
e
p
e
r
p
i
c
t
u
r
e
A
v
e
r
a
g
e
p
e
r
mo
d
e
l
V
o
l
t
a
g
e
(V)
C
u
r
r
e
n
t
(
A
)
P
o
w
e
r
(
W
)
V
o
l
t
a
g
e
(
V
)
C
u
r
r
e
n
t
(A)
P
o
w
e
r
(
W
)
Y
O
L
O
1
1
n
I
mag
e
1
5
.
3
7
0
.
6
8
3
.
6
7
5
.
3
8
0
.
6
6
3
.
5
7
I
mag
e
2
5
.
3
8
0
.
6
6
3
.
5
9
I
mag
e
3
5
.
3
8
0
.
6
5
3
.
5
4
I
mag
e
4
5
.
3
8
0
.
6
5
3
.
5
3
I
mag
e
5
5
.
3
8
0
.
6
5
3
.
5
0
Y
O
L
O
1
1
s
I
mag
e
1
5
.
3
8
0
.
6
6
3
.
5
6
5
.
3
8
0
.
6
5
3
.
6
1
I
mag
e
2
5
.
3
8
0
.
6
6
3
.
7
6
I
mag
e
3
5
.
3
8
0
.
6
5
3
.
7
2
I
mag
e
4
5
.
3
8
0
.
6
5
3
.
5
2
I
mag
e
5
5
.
3
8
0
.
6
4
3
.
4
8
Y
O
L
O
1
1
m
I
mag
e
1
5
.
3
8
0
.
6
8
3
.
6
7
5
.
3
8
0
.
6
6
3
.
5
4
I
mag
e
2
5
.
3
9
0
.
6
4
3
.
4
7
I
mag
e
3
5
.
3
8
0
.
6
5
3
.
5
3
I
mag
e
4
5
.
3
9
0
.
6
4
3
.
4
8
I
mag
e
5
5
.
3
8
0
.
6
6
3
.
5
7
Y
O
L
O
1
1
l
I
mag
e
1
5
.
3
8
0
.
6
7
3
.
6
0
5
.
3
8
0
.
6
5
3
.
5
0
I
mag
e
2
5
.
3
8
0
.
6
5
3
.
4
8
I
mag
e
3
5
.
3
8
0
.
6
4
3
.
4
5
I
mag
e
4
5
.
3
9
0
.
6
4
3
.
4
6
I
mag
e
5
5
.
3
9
0
.
6
5
3
.
5
1
Y
O
L
O
1
1
x
I
mag
e
1
5
.
3
8
0
.
6
7
3
.
6
0
5
.
3
8
0
.
6
6
3
.
5
7
I
mag
e
2
5.
38
0
.
6
6
3
.
6
1
I
mag
e
3
5
.
3
9
0
.
6
6
3
.
5
5
I
mag
e
4
5
.
3
8
0
.
6
5
3
.
5
0
I
mag
e
5
5
.
3
8
0
.
6
6
3
.
5
6
M
o
d
e
l
P
i
c
t
u
r
e
A
v
e
r
a
g
e
p
e
r
p
i
c
t
u
r
e
A
v
e
r
a
g
e
p
e
r
mo
d
e
l
V
o
l
t
a
g
e
(V)
C
u
r
r
e
n
t
(A)
P
o
w
e
r
(
W
)
V
o
l
t
a
g
e
(V)
C
u
r
r
e
n
t
(A)
P
o
w
e
r
(
W
)
Y
O
L
O
1
1
n
I
mag
e
1
5
.
3
5
0
.
6
8
3
.
6
9
5
.
3
5
0
.
6
6
3
.
5
8
I
mag
e
2
5
.
3
5
0
.
6
5
3
.
5
3
I
mag
e
3
5
.
3
6
0
.
6
5
3
.
5
1
I
mag
e
4
5
.
3
5
0
.
6
6
3
.
5
4
I
mag
e
5
5
.
3
5
0
.
6
7
3
.
6
5
Y
O
L
O
1
1
s
I
mag
e
1
5
.
3
4
0
.
7
1
3
.
8
1
5
.
3
5
0
.
6
9
3
.
7
2
I
mag
e
2
5
.
3
5
0
.
7
0
3
.
7
5
I
mag
e
3
5
.
3
6
0
.
7
0
3
.
7
6
I
mag
e
4
5
.
3
5
0
.
6
6
3
.
5
8
I
mag
e
5
5
.
3
5
0
.
6
8
3
.
6
9
Y
O
L
O
1
1
m
I
mag
e
1
5
.
3
6
0
.
6
6
3
.
5
9
5
.
3
5
0
.
6
8
3
.
6
7
I
mag
e
2
5
.
3
4
0
.
6
9
3
.
7
2
I
mag
e
3
5
.
3
5
0
.
6
8
3
.
6
8
I
mag
e
4
5
.
3
5
0
.
6
7
3
.
6
4
I
mag
e
5
5
.
3
5
0
.
6
9
3
.
7
1
Y
O
L
O
1
1
l
I
mag
e
1
5
.
3
5
0
.
6
8
3
.
6
9
5
.
3
5
0
.
6
7
3
.
6
2
I
mag
e
2
5
.
3
5
0
.
6
9
3
.
7
3
I
mag
e
3
5
.
3
6
0
.
6
4
3
.
4
9
I
mag
e
4
5
.
3
5
0
.
6
6
3
.
5
9
I
mag
e
5
5
.
3
6
0
.
6
6
3
.
5
8
Y
O
L
O
1
1
x
I
mag
e
1
5
.
3
3
0
.
6
7
3
.
5
9
5
.
3
5
0
.
6
7
3
.
6
2
I
mag
e
2
5
.
3
6
0
.
6
6
3
.
5
9
I
mag
e
3
5
.
3
5
0
.
6
7
3
.
6
2
I
mag
e
4
5
.
3
5
0
.
6
8
3
.
6
6
I
mag
e
5
5
.
3
5
0
.
6
7
3
.
6
3
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KOM
NI
K
A
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l
C
o
n
tr
o
l
P
erfo
r
ma
n
ce
en
h
a
n
ce
men
t o
f
emb
ed
d
ed
o
b
ject
d
etec
tio
n
via
n
eu
r
a
l h
a
r
d
w
a
r
e
..
(
A
lw
in
Ha
r
t
o
n
o
Lima
r
a
n
)
135
T
ab
le
5
.
Un
if
ied
s
u
m
m
ar
y
o
f
C
P
U
v
s
.
NP
U
p
er
f
o
r
m
an
ce
f
o
r
YOL
O1
1
v
ar
ian
ts
,
i
n
clu
d
i
n
g
laten
c
y
,
e
n
er
g
y
p
er
in
f
er
en
ce
,
an
d
ac
cu
r
ac
y
M
o
d
e
l
C
P
U
t
i
me
(
ms)
N
P
U
t
im
e
(
ms)
S
p
e
e
d
u
p
(
×
)
C
P
U
p
o
w
e
r
(
W
)
N
P
U
p
o
w
e
r
(
W
)
C
P
U
e
n
e
r
g
y
(
J)
N
P
U
e
n
e
r
g
y
(
J)
En
e
r
g
y
s
a
v
i
n
g
(
%)
O
f
f
i
c
i
a
l
Y
O
L
O
1
1
mA
P
O
n
-
d
e
v
i
c
e
mA
P
(
5
,
0
0
0
)
Δ mA
P
Y
O
L
O
1
1
n
1
,
5
8
7
.
1
4
3
2
0
.
7
4
4
.
9
5
3
.
6
0
3
.
5
9
5
.
7
1
1
.
1
5
7
9
.
8
0
.
3
9
5
0
.
3
6
9
-
0
.
0
2
6
Y
O
L
O
1
1
s
4
,
8
9
7
.
7
6
4
8
5
.
5
6
1
0
.
0
9
3
.
6
0
3
.
5
9
1
7
.
6
3
1
.
7
4
9
0
.
1
0
.
4
7
0
0
.
4
4
1
-
0
.
0
2
9
Y
O
L
O
1
1
m
1
6
,
0
6
0
.
1
0
1
,
0
0
2
.
3
6
1
6
.
0
2
3
.
6
0
3
.
5
9
5
7
.
8
2
3
.
6
0
9
3
.
8
0
.
5
1
5
0
.
4
8
7
-
0
.
0
2
8
Y
O
L
O
1
1
l
2
0
,
1
7
7
.
2
6
1
,
2
0
8
.
4
8
1
6
.
7
0
3
.
6
0
3
.
5
9
7
2
.
6
4
4
.
3
4
9
4
.
0
0
.
5
3
4
0
.
5
0
4
-
0
.
0
3
0
Y
O
L
O
1
1
x
4
0
,
4
4
1
.
4
4
2
,
6
6
3
.
4
4
1
5
.
1
8
3
.
6
0
3
.
5
9
1
4
5
.
5
9
9
.
5
6
9
3
.
4
0
.
5
4
7
0
.
5
1
7
-
0
.
0
3
0
T
h
ese
f
in
d
in
g
s
s
u
p
p
o
r
t
r
ea
l
-
ti
m
e
o
r
n
ea
r
-
r
ea
l
-
ti
m
e
d
ep
lo
y
m
e
n
t
o
f
m
o
d
er
n
o
b
j
ec
t
d
et
ec
tio
n
o
n
af
f
o
r
d
ab
le
e
m
b
ed
d
ed
p
latf
o
r
m
s
,
w
h
ic
h
h
a
s
p
r
ac
tical
s
i
g
n
i
f
ica
n
ce
.
T
h
e
R
K3
5
6
6
NP
U
e
n
ab
les
5
–
1
7
×
f
aster
in
f
er
en
ce
at
co
m
p
ar
ab
le
ac
cu
r
ac
y
(
Δ
m
AP
≈
-
0
.
0
3
)
w
i
th
~8
0
–
9
4
%
lo
w
er
en
er
g
y
p
er
i
n
f
er
en
ce
,
in
cr
ea
s
in
g
t
h
e
f
ea
s
ib
ili
t
y
o
f
ap
p
licatio
n
s
i
n
s
u
r
v
eil
lan
ce
,
r
o
b
o
tics
,
an
d
I
o
T
ed
g
e
w
o
r
k
lo
ad
s
.
(
×
)
=
(
5
)
(
)
=
1000
(
6
)
(
)
=
1000
(
7
)
(
%
)
(
%
)
=
−
100
(
8
)
=
−
5
,
000
−
(
9)
T
o
ex
p
lain
th
e
o
r
i
g
in
o
f
late
n
c
y
r
ed
u
ctio
n
s
,
T
ab
le
6
r
ep
o
r
ts
th
e
r
elat
i
v
e
ch
a
n
g
es
i
n
p
r
ep
r
o
ce
s
s
in
g
,
in
f
er
en
ce
,
an
d
p
o
s
tp
r
o
ce
s
s
in
g
ti
m
es
w
h
en
s
h
i
f
ti
n
g
f
r
o
m
C
P
U
to
N
P
U,
s
h
o
w
i
n
g
t
h
at
m
o
d
est
C
P
U
-
b
o
u
n
d
o
v
er
h
ea
d
s
ar
e
o
u
t
w
ei
g
h
ed
b
y
d
o
m
i
n
a
n
t in
f
er
e
n
ce
ac
ce
ler
ati
o
n
.
T
ab
le
6
.
R
elativ
e
ch
a
n
g
e
i
n
p
r
o
ce
s
s
in
g
ti
m
e
w
h
en
s
h
i
f
ti
n
g
f
r
o
m
C
P
U
to
NP
U
(
Δ
t
i
m
e
in
%
,
r
elativ
e
to
C
P
U
ex
ec
u
t
io
n
)
M
o
d
e
l
P
r
e
p
r
o
c
e
ss Δ
t
i
me
(
%)
I
n
f
e
r
e
n
c
e
Δ
t
i
me
(
%)
P
o
st
p
r
o
c
e
ss Δ
t
i
me
(
%)
T
o
t
a
l
t
i
me
Δ (%)
Y
O
L
O
1
1
n
+
8
.
4
0
-
8
2
.
3
+
1
2
.
9
-
7
9
.
8
Y
O
L
O
1
1
s
+
2
4
.
5
-
9
1
.
0
+
3
0
.
1
-
9
0
.
1
Y
O
L
O
1
1
m
+
2
1
.
3
-
9
4
.
0
+
2
6
.
7
-
9
3
.
8
Y
O
L
O
1
1
l
+
2
1
.
4
-
9
4
.
2
+
3
0
.
0
-
9
4
.
0
Y
O
L
O
1
1
x
+
2
9
.
5
-
9
3
.
5
+
2
9
.
6
-
9
3
.
4
T
ab
le
5
s
u
m
m
ar
izes
th
e
a
g
g
r
eg
ate
p
er
f
o
r
m
a
n
ce
late
n
c
y
,
e
n
er
g
y
p
er
in
f
er
en
ce
,
a
n
d
ac
cu
r
ac
y
ac
r
o
s
s
YOL
O1
1
v
ar
ian
ts
.
Ho
w
e
v
er
,
th
ese
m
etr
ic
s
d
o
n
o
t
ca
p
tu
r
e
s
ce
n
e
-
s
p
ec
if
ic
b
eh
av
io
r
s
.
T
ab
le
6
p
r
o
v
id
es
a
s
tag
e
-
w
i
s
e
b
r
ea
k
d
o
w
n
,
s
h
o
w
i
n
g
t
h
a
t
p
r
ep
r
o
ce
s
s
in
g
a
n
d
p
o
s
tp
r
o
ce
s
s
i
n
g
i
n
c
u
r
m
o
d
est
i
n
cr
ea
s
e
s
(
8
–
3
0
%),
w
h
er
ea
s
in
f
er
en
ce
d
o
m
i
n
ates
ac
ce
ler
ati
o
n
w
it
h
r
ed
u
ctio
n
s
o
f
8
2
–
9
4
%.
T
h
is
v
ie
w
clar
i
f
ies
t
h
e
s
o
u
r
ce
o
f
laten
c
y
g
ai
n
s
in
T
ab
le
5
.
T
o
c
o
n
tex
t
u
alize
t
h
ese
q
u
a
n
ti
t
ativ
e
r
es
u
lt
s
,
Fi
g
u
r
e
6
p
r
esen
t
s
r
ep
r
esen
tativ
e
d
etec
tio
n
s
ce
n
ar
io
s
.
T
h
ese
q
u
alitati
v
e
ca
s
e
s
tu
d
ie
s
ill
u
s
tr
ate
h
o
w
t
h
e
late
n
c
y
a
n
d
ac
cu
r
ac
y
p
atter
n
s
i
n
T
ab
le
5
ap
p
e
ar
in
r
ea
l
i
m
ag
e
s
,
h
ig
h
li
g
h
ti
n
g
lo
ca
lized
v
ar
iatio
n
s
u
n
d
er
d
iv
er
s
e
co
n
d
itio
n
s
.
I
n
t
h
e
w
ar
e
h
o
u
s
e
s
ce
n
e,
as
s
h
o
w
n
i
n
Fi
g
u
r
e
7
(
a)
,
C
P
U
in
f
er
e
n
ce
r
eq
u
ir
e
d
1
,
4
6
6
.
3
m
s
a
n
d
d
etec
ted
f
o
u
r
p
er
s
o
n
s
an
d
t
w
o
tr
u
c
k
s
,
w
h
er
ea
s
th
e
NP
U
co
m
p
leted
in
f
er
en
ce
in
3
2
1
.
9
m
s
an
d
d
etec
ted
s
ix
p
er
s
o
n
s
an
d
t
w
o
tr
u
ck
s
,
in
clu
d
i
n
g
o
n
e
ad
d
itio
n
al
p
er
s
o
n
at
a
co
n
f
i
d
en
ce
o
f
0
.
2
8
.
B
o
th
co
n
f
i
g
u
r
atio
n
s
o
cc
asio
n
all
y
m
is
cla
s
s
i
f
ied
s
tac
k
ed
b
o
x
es a
s
tr
u
ck
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.