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
4
,
No
.
5
,
Octo
b
er
2
0
2
5
,
p
p
.
4
0
9
0
~
4
1
0
0
I
SS
N:
2
2
5
2
-
8
9
3
8
,
DOI
: 1
0
.
1
1
5
9
1
/ijai.v
14
.i
5
.
p
p
4
0
9
0
-
4
1
0
0
4090
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
a
i
.
ia
esco
r
e.
co
m
An impro
v
ed r
ea
l t
ime
dete
ction
tr
a
nsfo
rmer met
ho
d f
o
r re
tail
pro
duct
d
etec
tion
Andi
Wa
hy
u
M
a
ula
na
1,
2
,
Su
ry
o
Adhi
Wibo
wo
1,
2
1
S
c
h
o
o
l
o
f
E
l
e
c
t
r
i
c
a
l
E
n
g
i
n
e
e
r
i
n
g
,
T
e
l
k
o
m
U
n
i
v
e
r
si
t
y
,
B
a
n
d
u
n
g
,
I
n
d
o
n
e
si
a
2
C
e
n
t
e
r
o
f
E
x
c
e
l
l
e
n
c
e
A
r
t
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
f
o
r
Le
a
r
n
i
n
g
a
n
d
O
p
t
i
mi
z
a
t
i
o
n
(
C
o
E
A
I
LO
)
,
Te
l
k
o
m
U
n
i
v
e
r
si
t
y
,
B
a
n
d
u
n
g
,
I
n
d
o
n
e
s
i
a
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
J
u
l 9
,
2
0
2
4
R
ev
is
ed
J
u
l 1
4
,
2
0
2
5
Acc
ep
ted
Au
g
6
,
2
0
2
5
Th
e
m
a
in
p
r
o
b
lem
in
re
tail
p
ro
d
u
c
t
d
e
tec
ti
o
n
is
i
n
tra
-
c
las
s
v
a
riatio
n
,
a
s
so
m
e
p
ro
d
u
c
ts
h
a
v
e
sim
il
a
r
b
u
t
d
isti
n
c
t
c
h
a
ra
c
teristics
.
Th
e
p
rima
ry
g
o
a
l
o
f
th
i
s
stu
d
y
is
to
a
d
d
re
ss
th
e
p
ro
b
lem
o
f
o
b
jec
t
d
e
tec
ti
o
n
o
n
in
tra
-
c
las
s
v
a
riatio
n
i
n
re
tail
e
n
v
iro
n
m
e
n
ts.
As
a
re
su
lt
,
a
n
e
w
a
p
p
r
o
a
c
h
fo
r
o
b
jec
t
d
e
tec
ti
o
n
o
f
re
tail
p
ro
d
u
c
ts
wa
s
d
e
v
e
lo
p
e
d
b
y
m
o
d
ify
in
g
t
h
e
re
a
l
t
ime
d
e
tec
ti
o
n
tr
a
n
sfo
rm
e
r
(RT
-
DETR)
m
o
d
e
l
.
T
o
m
a
n
a
g
e
i
n
tra
-
c
las
s
v
a
riatio
n
m
o
re
su
c
c
e
ss
fu
ll
y
,
t
h
e
RT
-
DETR
m
o
d
e
l
is
u
p
d
a
ted
b
y
m
o
d
if
y
i
n
g
it
s
a
rc
h
it
e
c
t
u
re
.
T
h
e
re
a
re
tw
o
c
o
n
v
o
lu
ti
o
n
s
in
th
e
c
o
n
tex
t
u
a
l
c
ro
ss
-
fe
a
tu
re
m
o
d
u
le
(CCF
M
)
fu
s
io
n
b
lo
c
k
se
c
ti
o
n
,
w
h
ich
is
a
d
j
u
ste
d
b
y
a
d
d
in
g
o
n
e
c
o
n
v
o
l
u
ti
o
n
lay
e
r
t
o
e
a
c
h
CCF
M
fu
sio
n
b
l
o
c
k
.
A
c
u
st
o
m
ize
d
d
a
ta
se
t
wa
s
m
e
ti
c
u
lo
u
sly
c
o
n
str
u
c
ted
to
re
flec
t
th
e
wid
e
ra
n
g
e
o
f
p
ro
d
u
c
ts
fr
e
q
u
e
n
tl
y
se
e
n
i
n
re
tail
o
u
tl
e
ts.
F
o
r
th
e
c
o
n
stru
c
ted
d
a
tas
e
ts,
tes
ts
we
re
ru
n
u
sin
g
t
h
e
m
e
a
n
a
v
e
ra
g
e
p
re
c
is
io
n
(m
AP)
m
e
tri
c
,
wh
ich
h
a
d
a
m
AP
@
0
.
5
o
f
9
9
.
5
%
a
n
d
a
m
AP
@
0
.
5
:
0
.
9
5
o
f
8
8
.
2
%
.
Th
e
u
p
d
a
ted
m
o
d
e
l
is s
u
p
e
rio
r
c
o
m
p
a
re
d
to
o
rig
in
a
l
m
o
d
e
l.
T
h
e
d
if
fe
re
n
c
e
in
m
AP
@
0
.
5
:
0
.
9
5
wa
s
2
.
5
%
,
wh
i
l
e
p
re
c
isio
n
in
c
re
a
se
d
b
y
1
.
3
%
a
n
d
re
c
a
ll
in
c
re
a
se
d
b
y
0
.
1
%
.
Alth
o
u
g
h
th
e
m
AP0
.
5
re
su
lt
s
sta
y
u
n
c
h
a
n
g
e
d
,
th
e
g
a
in
s
in
th
e
o
t
h
e
r
m
e
tri
c
s
su
g
g
e
st
th
a
t
th
e
RT
-
DETR
m
o
d
e
l
m
o
d
ifi
c
a
ti
o
n
s
c
a
n
imp
ro
v
e
o
b
jec
t
d
e
tec
ti
o
n
sk
il
ls,
p
a
rti
c
u
larly
wh
e
n
d
e
a
li
n
g
wit
h
i
n
tra
-
c
las
s
v
a
riatio
n
in
re
tail
m
e
rc
h
a
n
d
ise
.
K
ey
w
o
r
d
s
:
Dete
ctio
n
tr
an
s
f
o
r
m
er
Fu
s
io
n
b
lo
ck
I
n
tr
a
-
class
v
ar
iatio
n
Ob
ject
d
etec
tio
n
R
etail
p
r
o
d
u
ct
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
:
Su
r
y
o
Ad
h
i Wi
b
o
wo
Sch
o
o
l o
f
E
lectr
ical
E
n
g
in
ee
r
in
g
,
T
elk
o
m
Un
iv
e
r
s
ity
B
an
d
u
n
g
,
I
n
d
o
n
esia
E
m
ail: su
r
y
o
ad
h
iwib
o
wo
@
telk
o
m
u
n
i
v
er
s
ity
.
ac
.
id
1.
I
NT
RO
D
UCT
I
O
N
I
n
th
e
er
a
o
f
g
lo
b
aliza
tio
n
an
d
tech
n
o
lo
g
ical
in
n
o
v
atio
n
,
th
e
r
etail
in
d
u
s
tr
y
h
as u
n
d
er
g
o
n
e
s
ig
n
if
ican
t
tr
an
s
f
o
r
m
atio
n
,
d
r
i
v
en
b
y
ch
a
n
g
in
g
c
o
n
s
u
m
er
b
eh
a
v
io
r
s
an
d
in
ten
s
if
y
in
g
m
ar
k
et
co
m
p
eti
tio
n
.
Dee
p
lear
n
in
g
ap
p
r
o
ac
h
es
f
o
r
p
r
o
d
u
ct
item
d
etec
tio
n
h
av
e
em
e
r
g
ed
as
a
cr
itical
tech
n
o
lo
g
ical
s
o
lu
tio
n
[
1
]
,
ad
d
r
ess
in
g
th
e
co
m
p
lex
c
h
allen
g
es
o
f
m
o
d
e
r
n
r
etail
en
v
ir
o
n
m
en
ts
.
Ob
je
ct
d
etec
tio
n
,
a
f
u
n
d
am
en
tal
b
r
an
ch
o
f
c
o
m
p
u
ter
v
is
io
n
,
aim
s
to
id
en
tify
an
d
lo
ca
lize
s
p
ec
if
ic
o
b
jects
with
in
im
ag
es
o
r
v
id
eo
s
[
2
]
,
b
ec
o
m
in
g
in
cr
ea
s
in
g
l
y
cr
u
cial
f
o
r
u
n
d
e
r
s
tan
d
in
g
c
o
n
s
u
m
er
in
ter
ac
tio
n
s
,
o
p
tim
izin
g
s
h
o
p
p
in
g
ex
p
er
ie
n
ce
s
,
an
d
m
a
n
ag
in
g
i
n
v
en
to
r
y
.
T
h
e
r
etail
s
ec
to
r
co
n
tin
u
o
u
s
ly
ev
o
lv
es
[
3
]
t
o
m
ee
t
d
y
n
am
ic
m
ar
k
et
d
em
an
d
s
,
with
tech
n
o
lo
g
ical
ef
f
icien
cy
em
er
g
i
n
g
as
a
k
ey
d
if
f
er
en
tiato
r
.
Sm
ar
t
ca
r
t
tech
n
o
lo
g
ies
r
ep
r
esen
t
a
p
r
o
m
is
in
g
f
r
o
n
tier
in
th
is
tech
n
o
lo
g
ical
r
e
v
o
lu
tio
n
,
o
f
f
e
r
in
g
s
o
lu
tio
n
s
to
s
tr
ea
m
lin
e
t
h
e
s
h
o
p
p
i
n
g
ex
p
er
ien
ce
b
y
le
v
er
ag
in
g
ad
v
a
n
ce
d
o
b
ject
r
ec
o
g
n
itio
n
ca
p
a
b
ilit
ies
[
4
]
.
T
h
ese
i
n
tellig
en
t
s
y
s
tem
s
ca
n
a
u
to
m
atica
lly
i
d
en
tif
y
p
r
o
d
u
cts,
r
ed
u
ce
ch
ec
k
o
u
t
tim
es,
m
i
n
im
ize
p
r
icin
g
er
r
o
r
s
,
a
n
d
p
r
o
v
id
e
r
ea
l
-
tim
e
p
r
o
d
u
ct
in
f
o
r
m
at
io
n
,
f
u
n
d
am
e
n
tally
tr
an
s
f
o
r
m
in
g
tr
ad
itio
n
al
r
etail
in
ter
ac
tio
n
s
[
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
A
n
imp
r
o
ve
d
r
ea
l time
d
etec
tio
n
tr
a
n
s
fo
r
mer m
eth
o
d
fo
r
r
eta
il p
r
o
d
u
ct
d
etec
ti
on
(
A
n
d
i Wa
h
yu
Ma
u
l
a
n
a
)
4091
Ob
ject
d
etec
tio
n
in
s
m
ar
t
ca
r
ts
p
r
o
v
id
es
a
n
u
m
b
er
o
f
cr
u
cial
b
en
ef
its
.
First
o
f
all,
it
in
cr
ea
s
es
s
h
o
p
p
in
g
s
p
ee
d
b
y
s
h
o
r
ten
in
g
th
e
ch
ec
k
-
o
u
t
p
r
o
ce
s
s
[
6
]
.
Se
co
n
d
,
it
r
ed
u
ce
s
th
e
p
o
s
s
ib
ilit
y
o
f
i
n
ac
cu
r
ac
ies
in
r
eg
is
ter
in
g
p
r
o
d
u
cts
an
d
p
r
ici
n
g
.
T
h
ir
d
,
it
allo
ws
b
u
s
in
ess
es
to
im
m
ed
iately
p
r
o
v
i
d
e
p
r
o
m
o
tio
n
al
m
ater
ial
o
r
p
r
o
d
u
ct
r
ec
o
m
m
e
n
d
atio
n
s
to
co
n
s
u
m
er
s
.
I
n
th
e
f
ield
o
f
r
e
tail
s
h
o
p
p
in
g
,
th
e
av
ailab
ilit
y
an
d
ac
cu
r
ac
y
o
f
p
r
o
d
u
ct
in
f
o
r
m
atio
n
is
cr
u
cia
l.
W
h
en
a
c
u
s
to
m
er
u
s
es
a
s
m
ar
t
ca
r
t
to
s
h
o
p
,
an
o
b
jec
t
d
etec
tio
n
s
y
s
tem
r
ec
o
g
n
izes
an
d
r
ec
o
r
d
s
th
e
p
r
o
d
u
cts
th
at
a
r
e
in
s
er
ted
o
r
ta
k
en
o
u
t
f
r
o
m
th
e
ca
r
t.
W
h
ile
p
l
en
ty
o
f
r
esear
ch
h
as
b
ee
n
d
o
n
e
in
th
e
s
u
b
ject
o
f
item
d
etec
tio
n
,
s
p
ec
if
ic
p
r
o
b
lem
s
o
cc
u
r
in
th
e
s
ettin
g
o
f
d
y
n
am
ic
r
etail
en
v
ir
o
n
m
en
ts
,
wh
er
e
v
ar
iatio
n
s
in
p
r
o
d
u
ct
f
o
r
m
s
,
co
lo
r
s
,
a
n
d
g
r
o
u
p
in
g
s
ca
n
b
e
ch
allen
g
in
g
.
I
n
c
o
m
p
a
r
i
s
o
n
w
i
t
h
p
r
i
o
r
r
e
s
e
a
r
c
h
,
S
a
n
t
r
a
e
t
a
l
.
'
s
s
t
u
d
y
[
7
]
u
s
e
d
a
r
e
c
o
n
s
t
r
u
c
t
i
o
n
-
c
l
a
s
s
i
f
i
c
a
t
i
o
n
n
e
t
w
o
r
k
(
R
C
-
N
e
t
)
a
p
p
r
o
a
c
h
,
w
h
i
c
h
c
o
m
b
i
n
e
s
c
l
a
s
s
i
f
i
c
a
t
i
o
n
a
n
d
r
e
c
o
n
s
t
r
u
c
t
i
o
n
t
a
s
k
s
t
o
e
n
h
a
n
c
e
c
l
a
s
s
i
f
i
c
a
t
i
o
n
a
c
c
u
r
a
c
y
.
T
h
e
r
e
c
o
n
s
t
r
u
c
t
i
o
n
s
t
e
p
f
o
c
u
s
e
s
o
n
r
e
d
u
c
i
n
g
n
o
i
s
e
a
n
d
e
n
h
a
n
c
i
n
g
i
m
a
g
e
q
u
a
l
i
t
y
,
w
h
i
l
e
c
l
a
s
s
i
f
i
c
a
t
i
o
n
a
i
m
s
f
o
r
p
r
e
c
i
s
e
r
e
c
o
g
n
i
t
i
o
n
o
f
o
b
j
e
c
t
s
.
R
C
-
N
e
t
h
a
s
p
r
o
v
e
n
e
f
f
e
c
t
i
v
e
i
n
h
a
n
d
l
i
n
g
i
m
a
g
e
q
u
a
l
i
t
y
v
a
r
i
a
t
i
o
n
s
a
n
d
e
n
h
a
n
c
i
n
g
o
v
e
r
a
l
l
c
l
a
s
s
i
f
i
c
a
t
i
o
n
p
e
r
f
o
r
m
a
n
c
e
.
O
n
t
h
e
e
v
a
l
u
a
t
e
d
d
a
t
a
s
e
t
s
,
t
h
e
m
e
t
h
o
d
'
s
a
c
c
u
r
a
c
y
r
a
t
e
w
a
s
a
p
p
r
o
x
i
m
a
t
e
l
y
9
0
%
;
h
o
w
e
v
e
r
,
i
t
s
t
i
l
l
r
e
q
u
i
r
e
s
i
m
p
r
o
v
e
m
e
n
t
t
o
c
o
n
t
i
n
u
o
u
s
l
y
e
x
c
e
e
d
8
0
%
a
c
c
u
r
a
c
y
i
n
a
l
l
c
i
r
c
u
m
s
t
a
n
c
e
s
.
C
o
n
v
e
r
s
e
l
y
,
H
s
i
a
e
t
a
l
.
[
8
]
u
s
e
d
d
a
t
a
a
u
g
m
e
n
t
a
t
i
o
n
i
n
c
o
n
j
u
n
c
t
i
o
n
w
i
t
h
t
h
e
f
a
s
t
e
r
r
e
g
i
o
n
-
b
a
s
e
d
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
u
r
a
l
n
e
t
w
o
r
k
(R
-
C
N
N
)
t
e
c
h
n
i
q
u
e
i
n
t
h
e
i
r
e
x
p
e
r
i
m
e
n
t
.
T
o
i
n
c
r
e
a
s
e
t
h
e
m
o
d
e
l
'
s
e
f
f
i
c
i
e
n
c
y
,
t
h
e
y
u
s
e
d
t
e
c
h
n
i
q
u
e
s
i
n
c
l
u
d
i
n
g
r
o
t
a
t
i
o
n
,
f
l
i
p
p
i
n
g
,
a
n
d
s
c
a
l
i
n
g
t
o
d
i
v
e
r
s
i
f
y
t
h
e
d
a
t
a
s
e
t
.
T
h
e
i
r
f
i
n
d
i
n
g
s
c
o
n
f
i
r
m
e
d
t
h
a
t
t
h
e
s
e
a
u
g
m
e
n
t
a
t
i
o
n
m
e
t
h
o
d
s
s
i
g
n
i
f
i
c
a
n
t
l
y
b
o
o
s
t
e
d
t
h
e
m
o
d
e
l
'
s
a
c
c
u
r
a
c
y
a
n
d
m
a
d
e
i
t
m
o
r
e
r
o
b
u
s
t
a
g
a
i
n
s
t
i
n
p
u
t
v
a
r
i
a
t
i
o
n
s
.
B
e
c
a
u
s
e
o
f
t
h
e
s
m
a
l
l
q
u
a
n
t
i
t
y
o
f
t
h
e
d
a
t
a
s
e
t
,
t
h
e
m
o
d
e
l
h
a
d
t
r
o
u
b
l
e
i
d
e
n
t
i
f
y
i
n
g
v
e
r
y
t
i
n
y
p
r
o
d
u
c
t
d
i
f
f
e
r
e
n
c
e
s
,
e
v
e
n
t
h
o
u
g
h
i
t
a
c
h
i
e
v
e
d
a
h
i
g
h
m
e
a
n
a
v
e
r
a
g
e
p
r
e
c
i
s
i
o
n
(
m
A
P
)
a
c
c
u
r
a
c
y
o
f
9
9
.
2
7
%
.
L
e
e
e
t
a
l
.
[
9
]
s
t
a
t
e
d
,
f
o
r
r
e
t
a
i
l
p
r
o
d
u
c
t
d
e
t
e
c
t
i
o
n
,
t
h
e
y
u
s
e
d
t
h
e
m
o
b
i
l
e
n
e
u
r
a
l
n
e
t
w
o
r
k
v
e
r
s
i
o
n
3
(
M
o
b
i
l
e
N
e
t
V
3
)
a
r
c
h
i
t
e
c
t
u
r
e
i
n
c
o
n
j
u
n
c
t
i
o
n
w
i
t
h
t
h
e
y
o
u
o
n
l
y
l
o
o
k
o
n
c
e
v
e
r
s
i
o
n
5
(
Y
O
L
O
v
5
)
m
o
d
e
l
.
T
h
e
y
e
m
p
l
o
y
e
d
m
e
t
h
o
d
s
l
i
k
e
r
o
t
a
t
i
o
n
,
f
l
i
p
p
i
n
g
,
a
n
d
s
c
a
l
i
n
g
t
o
d
i
v
e
r
s
i
f
y
t
h
e
d
a
t
a
s
e
t
,
b
o
o
s
t
i
n
g
t
h
e
m
o
d
e
l
'
s
e
f
f
e
c
t
i
v
e
n
e
s
s
.
F
i
n
d
i
n
g
s
d
e
m
o
n
s
t
r
a
t
e
d
t
h
a
t
d
a
t
a
a
u
g
m
e
n
t
a
t
i
o
n
n
o
t
a
b
l
y
e
n
h
a
n
c
e
d
b
o
t
h
a
c
c
u
r
a
c
y
a
n
d
r
e
s
i
l
i
e
n
c
e
o
f
t
h
e
m
o
d
e
l
t
o
d
i
v
e
r
s
e
i
n
p
u
t
v
a
r
i
a
t
i
o
n
s
.
D
e
s
p
i
t
e
a
c
h
i
e
v
i
n
g
a
h
i
g
h
m
A
P
a
c
c
u
r
a
c
y
o
f
9
9
.
2
7
%
,
t
h
e
m
o
d
e
l
e
n
c
o
u
n
t
e
r
e
d
c
h
a
l
l
e
n
g
e
s
i
n
d
i
s
t
i
n
g
u
i
s
h
i
n
g
e
x
t
r
e
m
e
l
y
s
u
b
t
l
e
p
r
o
d
u
c
t
d
i
f
f
e
r
e
n
c
e
s
,
p
r
i
m
a
r
i
l
y
d
u
e
t
o
t
h
e
d
a
t
a
s
e
t
'
s
l
i
m
i
t
e
d
s
i
z
e
.
L
e
e
e
t
a
l
.
[
9
]
u
t
i
l
i
z
e
d
t
h
e
Y
O
L
O
v
5
m
o
d
e
l
c
o
m
b
i
n
e
d
w
i
t
h
t
h
e
M
o
b
i
l
e
N
e
t
V
3
a
r
c
h
i
t
e
c
t
u
r
e
f
o
r
r
e
t
a
i
l
p
r
o
d
u
c
t
d
e
t
e
c
t
i
o
n
.
T
h
e
g
o
a
l
o
f
t
h
i
s
c
o
m
b
i
n
a
t
i
o
n
w
a
s
t
o
m
a
x
i
m
i
z
e
d
e
t
e
c
t
i
o
n
e
f
f
i
c
i
e
n
c
y
a
n
d
s
p
e
e
d
w
i
t
h
o
u
t
s
a
c
r
i
f
i
c
i
n
g
a
c
c
u
r
a
c
y
.
A
c
c
o
r
d
i
n
g
t
o
e
x
p
e
r
i
m
e
n
t
a
l
r
e
s
u
l
t
s
,
t
h
i
s
m
o
d
e
l
i
s
p
e
r
f
e
c
t
f
o
r
r
e
a
l
-
t
i
m
e
r
e
t
a
i
l
a
p
p
l
i
c
a
t
i
o
n
s
s
i
n
c
e
i
t
c
a
n
r
e
l
i
a
b
l
y
a
n
d
s
w
i
f
t
l
y
r
e
c
o
g
n
i
z
e
r
e
t
a
i
l
p
r
o
d
u
c
t
s
.
T
h
e
s
t
u
d
y
h
a
d
c
o
n
s
t
r
a
i
n
t
s
b
e
c
a
u
s
e
o
f
t
h
e
r
e
l
a
t
i
v
e
l
y
m
o
d
e
s
t
s
c
a
l
e
o
f
t
h
e
d
a
t
a
s
e
t
s
e
m
p
l
o
y
e
d
,
e
v
e
n
t
h
o
u
g
h
i
t
a
c
h
i
e
v
e
d
a
9
8
.
5
%
m
A
P
a
c
c
u
r
a
c
y
.
B
ased
o
n
p
r
io
r
r
esear
ch
f
i
n
d
i
n
g
s
,
th
is
s
tu
d
y
will
p
r
esen
t
a
m
o
d
if
ied
r
ea
l
tim
e
d
etec
tio
n
t
r
an
s
f
o
r
m
er
(
R
T
-
DE
T
R
)
m
o
d
el
to
im
p
r
o
v
e
m
AP
[
1
0
]
ac
c
u
r
ac
y
u
tili
zin
g
a
s
elf
-
p
r
o
ce
s
s
ed
d
ataset
b
ase
d
o
n
r
etail
p
r
o
d
u
cts
in
th
e
I
n
d
o
n
esian
p
r
o
d
u
ct
b
u
s
in
ess
in
r
ea
l
tim
e.
I
n
a
d
d
itio
n
to
t
h
e
s
elf
-
g
en
er
ated
d
ataset,
th
e
m
o
d
el
will
b
e
test
ed
with
th
r
ee
o
th
e
r
d
atasets
:
th
e
g
r
o
ce
r
y
d
ataset
[
1
1
]
,
wh
ich
f
o
c
u
s
es
o
n
p
r
o
d
u
cts
with
d
if
f
er
en
t
u
n
it
s
izes;
th
e
r
etail
p
r
o
d
u
ct
ch
ec
k
o
u
t
(
R
P
C
)
-
d
ataset
[
1
2
]
,
wh
ich
c
h
a
llen
g
es
p
r
o
d
u
cts
with
s
im
ilar
ch
ar
ac
ter
is
tics
;
an
d
th
e
d
en
s
ely
s
eg
m
en
ted
s
u
p
e
r
m
ar
k
et
(
D2
S)
-
d
ataset
[
1
3
]
,
wh
ich
test
s
d
etec
tio
n
u
n
d
e
r
d
if
f
er
en
t
lig
h
tin
g
co
n
d
itio
n
s
an
d
p
r
o
d
u
ct
s
tack
s
.
T
h
e
p
r
o
p
o
s
ed
ap
p
r
o
ac
h
f
o
c
u
s
es
o
n
ex
am
in
in
g
p
r
o
d
u
ct
v
ar
iatio
n
s
ac
r
o
s
s
s
ize,
co
lo
r
,
an
d
ty
p
e
,
with
a
p
r
im
ar
y
o
b
jectiv
e
o
f
ac
h
iev
i
n
g
a
m
AP a
cc
u
r
ac
y
ex
ce
e
d
in
g
9
0
%.
T
h
e
k
ey
c
o
n
tr
ib
u
tio
n
s
o
f
th
is
s
tu
d
y
in
clu
d
e:
i
)
a
n
o
v
el
ar
ch
itectu
r
al
ad
ap
tatio
n
o
f
th
e
R
T
-
DE
T
R
m
o
d
el
[
1
4
]
,
ii)
d
ev
elo
p
m
en
t o
f
a
co
m
p
r
eh
en
s
iv
e
s
ix
-
class
d
a
taset
r
ep
r
esen
tin
g
I
n
d
o
n
esian
r
etail
p
r
o
d
u
cts,
an
d
iii)
a
r
o
b
u
s
t
m
eth
o
d
o
lo
g
y
f
o
r
d
etec
tin
g
p
r
o
d
u
cts
with
h
i
g
h
ly
s
im
ilar
attr
ib
u
tes
ac
r
o
s
s
d
i
f
f
er
en
t
ca
teg
o
r
ies.
T
h
is
r
esear
ch
ad
v
an
ce
s
r
ea
l
-
ti
m
e
p
r
o
d
u
ct
d
etec
tio
n
ca
p
a
b
ilit
ies.
I
t
s
ee
k
s
to
p
r
o
v
id
e
a
s
o
p
h
is
ticated
s
o
lu
tio
n
th
at
ca
n
s
ig
n
if
ican
tly
e
n
h
an
ce
r
etail
tech
n
o
lo
g
y
'
s
p
r
ec
is
io
n
an
d
ef
f
ec
tiv
e
n
ess
.
2.
M
E
T
H
O
D
R
etail
p
r
o
d
u
ct
d
etec
tio
n
is
u
s
ed
to
id
en
tify
an
d
class
if
y
p
r
o
d
u
cts
f
r
o
m
im
ag
es.
T
h
is
p
r
o
ce
s
s
is
u
s
ef
u
l
f
o
r
a
p
p
licatio
n
s
s
u
c
h
as
s
m
ar
t
ca
r
ts
an
d
in
v
e
n
to
r
y
m
a
n
ag
e
m
en
t
s
y
s
tem
s
.
W
e
p
r
o
p
o
s
e
a
m
eth
o
d
b
ased
o
n
a
m
o
d
if
ied
R
T
-
DE
T
R
m
o
d
el.
T
h
e
m
ain
p
r
o
ce
d
u
r
es
in
th
is
r
esear
ch
in
clu
d
e
d
ata
c
o
llectio
n
u
s
in
g
a
tu
r
n
tab
le
s
etu
p
an
d
a
Fu
jifilm
X
-
T
2
0
ca
m
er
a,
d
ata
au
g
m
e
n
tatio
n
to
en
h
an
ce
d
ataset
d
iv
er
s
ity
,
an
d
m
o
d
el
tr
ain
in
g
with
cu
s
to
m
ized
R
T
-
DE
T
R
lay
er
s
to
im
p
r
o
v
e
d
etec
tio
n
ac
cu
r
ac
y
.
T
h
e
f
in
al
o
u
t
p
u
t
is
a
r
eliab
le
p
r
o
d
u
ct
d
etec
tio
n
s
y
s
tem
.
T
h
e
m
ain
p
r
o
ce
d
u
r
es
ar
e
s
h
o
wn
in
Fig
u
r
e
1
.
2
.
1
.
Da
t
a
c
o
llect
io
n
T
h
e
d
ataset
f
o
r
I
n
d
o
n
esian
r
e
tail
p
r
o
d
u
cts
was
co
llected
u
s
in
g
a
Fu
jifilm
X
-
T
2
0
ca
m
er
a,
with
th
e
im
ag
in
g
p
r
o
ce
s
s
f
ac
ilit
ated
b
y
a
tu
r
n
tab
le
with
p
a
p
er
o
n
it
,
allo
win
g
f
o
r
th
e
r
o
tatio
n
o
f
p
r
o
d
u
cts
to
ca
p
tu
r
e
im
ag
es
f
r
o
m
m
u
ltip
le
an
g
les.
T
h
is
m
eth
o
d
e
n
s
u
r
es
co
m
p
r
eh
en
s
iv
e
v
is
u
al
d
ata
r
e
p
r
esen
tatio
n
f
r
o
m
v
ar
io
u
s
p
er
s
p
ec
tiv
es.
T
h
e
d
ataset
co
n
s
is
ts
o
f
6
p
r
o
d
u
ct
class
es:
B
u
av
ita
Gu
av
a,
C
h
itato
L
ite
Sea
we
ed
,
Or
eo
Or
ig
in
al,
R
ed
B
u
ll
Dr
in
k
,
C
h
o
co
late
W
h
ea
t
E
s
s
en
ce
,
an
d
Selai
Ol
ai
Stra
wb
er
r
y
,
with
ea
ch
clas
s
h
av
in
g
6
5
im
ag
es
tak
en
f
r
o
m
d
if
f
er
e
n
t
an
g
les.
T
h
e
ca
m
er
a
s
ettin
g
s
wer
e
o
p
tim
ized
f
o
r
p
r
o
d
u
ct
p
h
o
to
g
r
ap
h
y
,
in
clu
d
i
n
g
a
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
4
,
No
.
5
,
Octo
b
er
2
0
2
5
:
4
0
9
0
-
4
1
0
0
4092
m
ed
iu
m
ap
er
t
u
r
e
f
o
r
d
ep
t
h
o
f
f
ield
an
d
s
h
ar
p
n
ess
,
an
ad
ju
s
ted
s
h
u
tter
s
p
ee
d
to
av
o
id
m
o
ti
o
n
b
lu
r
,
an
d
a
l
o
w
in
ter
n
atio
n
al
o
r
g
a
n
izatio
n
f
o
r
s
tan
d
ar
d
izatio
n
(
I
SO)
to
m
i
n
im
ize
n
o
is
e.
Pro
p
er
lig
h
tin
g
was
en
s
u
r
ed
u
s
in
g
d
if
f
u
s
ed
lig
h
t
s
o
u
r
ce
s
to
av
o
i
d
h
ar
s
h
s
h
ad
o
ws
an
d
r
ef
lectio
n
s
,
with
b
r
ig
h
tn
ess
an
d
ex
p
o
s
u
r
e
ca
lib
r
ated
f
o
r
n
atu
r
al
co
l
o
r
s
an
d
ad
e
q
u
ate
c
o
n
tr
ast.
T
h
e
tu
r
n
tab
le
allo
wed
f
o
r
s
y
s
tem
atic
an
d
co
n
t
r
o
lled
r
o
tatio
n
,
ty
p
ically
s
et
to
a
f
ix
ed
d
eg
r
ee
in
cr
e
m
e
n
t,
en
s
u
r
in
g
co
n
s
is
ten
t
an
g
les
an
d
in
te
r
v
als
f
o
r
ea
ch
p
r
o
d
u
c
t.
T
h
e
im
ag
es
wer
e
co
m
p
o
s
ed
to
k
ee
p
th
e
p
r
o
d
u
ct
ce
n
ter
ed
an
d
at
a
co
n
s
is
ten
t
d
is
tan
ce
f
r
o
m
th
e
ca
m
er
a,
with
p
lain
b
ac
k
g
r
o
u
n
d
s
to
av
o
id
d
is
tr
ac
tio
n
s
.
T
h
is
ap
p
r
o
ac
h
r
esu
lted
in
a
h
ig
h
-
q
u
ality
,
co
n
s
is
ten
t
d
ataset
s
u
itab
le
f
o
r
v
ar
i
o
u
s
ap
p
licatio
n
s
in
co
m
p
u
ter
v
is
io
n
an
d
d
ee
p
lear
n
in
g
.
T
h
e
r
esu
l
ts
ca
n
b
e
s
ee
n
in
Fig
u
r
e
2
.
Fig
u
r
e
1
.
T
h
e
p
r
o
p
o
s
ed
m
eth
o
d
o
lo
g
y
Fig
u
r
e
2
.
T
h
e
r
esu
lts
o
f
th
e
d
a
taset im
ag
e
ca
p
tu
r
e
2
.
2
.
Da
t
a
a
ug
m
ent
a
t
io
n
I
m
ag
e
au
g
m
en
tatio
n
p
lay
s
a
cr
u
cial
r
o
le
i
n
en
h
an
cin
g
a
d
ataset
b
y
p
r
o
v
id
in
g
a
wid
er
v
ar
iety
o
f
ex
am
p
les
.
T
h
ese
ex
am
p
les h
el
p
a
m
o
d
el
g
en
er
alize
b
etter
,
esp
ec
ially
wh
en
d
ea
lin
g
with
in
t
r
a
-
class
v
ar
iatio
n
s
.
B
y
ar
tific
ially
en
lar
g
in
g
th
e
d
ataset,
th
e
m
o
d
el
b
ec
o
m
es
m
o
r
e
r
o
b
u
s
t
an
d
ca
p
ab
le
o
f
h
an
d
lin
g
d
i
v
er
s
e
s
ce
n
ar
io
s
,
th
u
s
r
ed
u
ci
n
g
th
e
r
i
s
k
o
f
o
v
e
r
f
itti
n
g
a
n
d
im
p
r
o
v
in
g
o
v
er
all
p
er
f
o
r
m
an
ce
.
I
n
th
is
p
r
o
ject,
s
ev
er
al
au
g
m
en
tatio
n
tech
n
iq
u
es
wer
e
ap
p
lied
to
ad
d
r
ess
co
m
m
o
n
ch
allen
g
es
en
co
u
n
ter
e
d
in
r
etail
p
r
o
d
u
ct
d
atasets
,
f
o
r
th
e
au
g
m
en
tatio
n
s
ettin
g
s
ca
n
b
e
s
ee
n
in
T
ab
le
1
.
Au
to
o
r
ien
tatio
n
en
s
u
r
es
th
at
all
im
ag
es
ar
e
p
r
o
p
er
ly
alig
n
e
d
,
wh
ich
is
ess
en
tial
f
o
r
co
n
s
is
ten
t
tr
ain
in
g
.
R
esizin
g
th
e
im
ag
es
to
640
×
6
4
0
p
ix
els
s
tan
d
ar
d
ize
d
im
en
s
io
n
s
,
m
ak
in
g
th
e
d
ata
s
et
u
n
if
o
r
m
a
n
d
r
ed
u
cin
g
c
o
m
p
u
tatio
n
al
l
o
ad
.
Fli
p
p
in
g
im
ag
es
h
o
r
izo
n
tally
an
d
v
e
r
tically
in
tr
o
d
u
ce
s
v
ar
i
atio
n
s
in
p
r
o
d
u
ct
o
r
ien
tatio
n
,
h
elp
in
g
th
e
m
o
d
el
r
ec
o
g
n
ize
item
s
r
e
g
ar
d
less
o
f
th
eir
p
lace
m
en
t.
C
r
o
p
p
in
g
with
a
m
in
im
u
m
zo
o
m
o
f
0
%
an
d
a
m
ax
im
u
m
zo
o
m
o
f
7
0
%
s
im
u
lates
d
if
f
er
e
n
t
d
i
s
tan
ce
s
an
d
p
e
r
s
p
ec
tiv
es,
en
h
an
cin
g
th
e
m
o
d
el'
s
ab
ilit
y
to
d
etec
t
p
r
o
d
u
cts
at
v
ar
io
u
s
zo
o
m
lev
els.
T
ab
le
1
.
Pre
-
p
r
o
ce
s
s
in
g
an
d
a
u
g
m
en
tatio
n
Ty
p
e
P
r
o
c
e
ss
S
e
t
u
p
Pre
-
p
r
o
c
e
ssi
n
g
A
u
t
o
o
r
i
e
n
t
e
d
A
p
p
l
i
e
d
Pre
-
p
r
o
c
e
ssi
n
g
R
e
si
z
e
S
t
r
e
t
c
h
t
o
6
4
0
×
6
4
0
A
u
g
m
e
n
t
a
t
i
o
n
F
l
i
p
H
o
r
i
z
o
n
t
a
l
,
v
e
r
t
i
c
a
l
A
u
g
m
e
n
t
a
t
i
o
n
C
r
o
p
0
%
mi
n
i
m
u
m
z
o
o
m
,
7
0
%
m
a
x
i
m
u
m
z
o
o
m
A
u
g
m
e
n
t
a
t
i
o
n
R
o
t
a
t
i
o
n
B
e
t
w
e
e
n
-
4
5
°
a
n
d
+
4
5
°
A
u
g
m
e
n
t
a
t
i
o
n
S
h
e
a
r
±
1
5
°
h
o
r
i
z
o
n
t
a
l
,
±
1
5
°
v
e
r
t
i
c
a
l
A
u
g
m
e
n
t
a
t
i
o
n
B
r
i
g
h
t
n
e
ss
B
e
t
w
e
e
n
-
3
0
%
a
n
d
+
3
0
%
A
u
g
m
e
n
t
a
t
i
o
n
B
l
u
r
U
p
t
o
2
.
5
px
A
u
g
m
e
n
t
a
t
i
o
n
C
u
t
o
u
t
2
0
b
o
x
e
s w
i
t
h
1
0
%
s
i
z
e
e
a
c
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
A
n
imp
r
o
ve
d
r
ea
l time
d
etec
tio
n
tr
a
n
s
fo
r
mer m
eth
o
d
fo
r
r
eta
il p
r
o
d
u
ct
d
etec
ti
on
(
A
n
d
i Wa
h
yu
Ma
u
l
a
n
a
)
4093
R
o
tatio
n
b
etwe
en
-
45
°
a
n
d
+4
5
°
ac
co
u
n
ts
f
o
r
r
o
tatio
n
al
d
if
f
er
en
ce
s
,
en
s
u
r
in
g
th
e
m
o
d
el
c
an
id
en
tify
p
r
o
d
u
cts
f
r
o
m
v
a
r
io
u
s
a
n
g
les.
Sh
ea
r
in
g
b
o
th
h
o
r
izo
n
tally
a
n
d
v
er
tically
b
y
±
1
5
°
d
is
to
r
ts
th
e
im
a
g
e
s
lig
h
tly
,
m
im
ick
in
g
r
ea
l
-
w
o
r
ld
d
is
to
r
t
io
n
s
an
d
im
p
r
o
v
in
g
r
o
b
u
s
tn
e
s
s
.
B
r
ig
h
tn
ess
ad
ju
s
tm
en
ts
b
etwe
en
-
3
0
%
an
d
+3
0
%
allo
w
t
h
e
m
o
d
el
t
o
p
e
r
f
o
r
m
well
u
n
d
er
v
ar
y
in
g
illu
m
in
atio
n
c
o
n
d
itio
n
s
.
B
lu
r
u
p
to
2
.
5
p
ix
els
a
d
d
s
s
lig
h
t
b
lu
r
r
in
g
to
s
im
u
late
m
o
tio
n
o
r
f
o
cu
s
v
ar
iatio
n
s
,
m
a
k
in
g
t
h
e
m
o
d
el
r
esil
ien
t
to
s
u
ch
is
s
u
es.
Fin
ally
,
cu
to
u
t
with
2
0
b
o
x
es
ea
ch
s
ized
at
1
0
%
co
v
er
s
r
an
d
o
m
p
a
r
ts
o
f
th
e
im
ag
e
to
s
im
u
late
o
cc
lu
s
io
n
,
tr
ain
in
g
th
e
m
o
d
el
to
r
ec
o
g
n
ize
p
r
o
d
u
cts e
v
en
if
p
a
r
tially
o
b
s
cu
r
e
d
.
T
h
e
au
g
m
en
tatio
n
p
r
o
ce
s
s
was
ca
r
r
ied
o
u
t
u
s
in
g
R
o
b
o
f
lo
w
[
1
5
]
,
a
p
latf
o
r
m
th
at
s
im
p
lifie
s
d
ataset
m
o
d
if
icatio
n
.
T
h
is
co
m
p
r
eh
e
n
s
iv
e
au
g
m
e
n
tatio
n
s
tr
ateg
y
r
esu
lted
in
a
tr
ain
i
n
g
d
ataset
o
f
7
8
0
p
h
o
to
s
an
d
a
v
alid
atio
n
d
ataset
o
f
7
5
im
ag
es.
Ad
d
itio
n
ally
,
a
test
in
g
d
ataset
o
f
3
8
im
ag
es
was
g
en
er
ated
,
to
talin
g
8
9
3
im
a
g
es.
2
.
3
.
P
ub
lic
d
a
t
a
s
et
s
T
h
e
s
tu
d
y
em
p
lo
y
s
th
r
ee
p
u
b
lic
d
atasets
to
co
m
p
r
eh
e
n
s
iv
ely
v
alid
ate
th
e
R
T
-
DE
T
R
m
o
d
el'
s
p
er
f
o
r
m
an
ce
in
a
d
d
r
ess
in
g
in
tr
a
-
class
v
ar
iatio
n
ch
all
en
g
es.
T
h
e
g
r
o
ce
r
y
d
atase
t
[
1
1
]
,
co
m
p
r
is
in
g
3
3
,
9
1
9
im
a
g
es
with
n
ea
r
ly
id
en
tical
p
r
o
d
u
ct
f
ea
tu
r
es,
p
r
o
v
id
es
a
r
ig
o
r
o
u
s
test
f
o
r
d
etec
t
in
g
s
u
b
tle
p
r
o
d
u
ct
v
ar
iatio
n
s
.
T
h
e
R
PC
-
d
ataset
[
1
2
]
,
with
its
ex
p
a
n
s
iv
e
2
0
0
p
r
o
d
u
ct
class
es
an
d
8
3
,
6
9
9
im
a
g
es,
o
f
f
er
s
a
lar
g
e
-
s
ca
le
ch
allen
g
e
in
r
etail
p
r
o
d
u
ct
d
etec
tio
n
,
wh
ile
th
e
D2
S
-
d
ataset
[
1
3
]
,
th
o
u
g
h
s
m
aller
with
3
,
7
2
9
im
ag
es,
in
tr
o
d
u
ce
s
co
m
p
le
x
d
etec
tio
n
s
ce
n
ar
io
s
th
r
o
u
g
h
v
a
r
ied
l
ig
h
tin
g
co
n
d
itio
n
s
an
d
p
r
o
d
u
ct
s
tack
in
g
.
T
h
es
e
d
atasets
co
llectiv
ely
r
ep
r
esen
t
a
co
m
p
r
eh
e
n
s
iv
e
ev
alu
atio
n
f
r
am
ewo
r
k
,
en
a
b
lin
g
a
r
o
b
u
s
t
ass
ess
m
en
t
o
f
th
e
m
o
d
el'
s
ca
p
ab
ilit
y
to
ac
cu
r
ate
ly
r
ec
o
g
n
ize
an
d
d
is
tin
g
u
is
h
p
r
o
d
u
cts
with
h
ig
h
ly
s
im
ilar
ch
ar
ac
ter
is
tics
ac
r
o
s
s
d
if
f
er
en
t
co
n
tex
ts
.
T
h
e
d
ataset
p
ar
titi
o
n
in
g
f
o
llo
ws
a
s
tan
d
ar
d
m
ac
h
in
e
lear
n
in
g
ap
p
r
o
ac
h
: th
e
g
r
o
ce
r
y
d
ataset
is
s
p
lit
8
5
%
f
o
r
tr
ain
in
g
,
1
0
%
f
o
r
v
alid
atio
n
,
an
d
5
%
f
o
r
te
s
tin
g
;
th
e
R
P
C
-
Data
s
et
u
s
es
a
7
0
/2
0
/1
0
s
p
lit;
an
d
th
e
D2
S
-
d
ataset
m
ain
tain
s
th
e
s
am
e
7
0
/2
0
/1
0
d
is
tr
ib
u
tio
n
.
T
h
is
s
tr
ateg
ic
s
elec
tio
n
an
d
p
ar
t
itio
n
in
g
o
f
d
atasets
en
s
u
r
e
a
co
m
p
r
eh
en
s
iv
e
v
ali
d
atio
n
o
f
t
h
e
p
r
o
p
o
s
ed
R
T
-
D
E
T
R
m
o
d
el,
ad
d
r
ess
in
g
k
ey
ch
allen
g
es
in
r
etail
p
r
o
d
u
ct
d
etec
tio
n
s
u
c
h
as in
tr
a
-
class
v
ar
iatio
n
,
p
r
o
d
u
ct
s
im
ilar
ity
,
an
d
v
ar
iatio
n
s
in
im
a
g
e
ca
p
tu
r
e
co
n
d
itio
n
s
.
2
.
4
.
Rea
l
-
t
im
e
det
ec
t
io
n t
ra
ns
f
o
rm
er
T
h
e
R
T
-
DE
T
R
[
1
4
]
is
a
r
ea
l
-
tim
e
v
is
io
n
t
r
an
s
f
o
r
m
e
r
(
ViT
)
[
1
6
]
m
o
d
el
m
a
d
e
u
p
o
f
t
h
r
ee
c
o
r
e
co
m
p
o
n
en
ts
:
a
b
ac
k
b
o
n
e,
a
h
y
b
r
id
en
c
o
d
er
,
an
d
a
d
ec
o
d
er
tr
an
s
f
o
r
m
er
th
at
also
in
clu
d
es
an
ex
tr
a
p
r
e
d
ictio
n
h
ea
d
.
Fig
u
r
e
3
s
h
o
ws
th
e
s
y
s
tem
's
s
tr
u
ctu
r
e.
T
h
is
m
o
d
el
u
s
es
th
e
o
u
tp
u
t
f
ea
tu
r
es
f
r
o
m
th
e
f
in
al
th
r
ee
b
ac
k
b
o
n
e
s
tag
es
(
S3
,
S4
,
an
d
S5
)
as
in
p
u
t
f
o
r
th
e
e
n
co
d
e
r
[
1
4
]
.
T
h
r
o
u
g
h
in
tr
a
-
s
ca
le
i
n
ter
ac
tio
n
[
1
7
]
an
d
in
ter
-
s
ca
le
f
u
s
io
n
[
1
8
]
,
th
e
h
y
b
r
id
en
co
d
er
[
1
9
]
c
o
n
v
er
ts
m
u
lti
-
s
ca
le
f
ea
tu
r
es
[
2
0
]
in
to
a
s
e
r
ies
o
f
im
a
g
e
-
lev
el
f
ea
tu
r
es
[
2
1
]
.
T
h
e
n
,
a
n
in
te
r
s
ec
tio
n
o
f
u
n
io
n
(
I
o
U)
-
awa
r
e
q
u
er
y
s
elec
tio
n
m
et
h
o
d
[
2
2
]
i
s
ap
p
lied
t
o
e
x
tr
ac
t
f
ea
tu
r
es
f
r
o
m
th
e
e
n
co
d
e
r
'
s
o
u
tp
u
t
as
th
e
i
n
itial
o
b
ject
q
u
e
r
y
f
o
r
th
e
d
ec
o
d
er
[
2
3
]
.
T
h
e
d
ec
o
d
er
th
en
r
e
f
in
es
th
ese
q
u
er
ies
s
tep
b
y
s
tep
to
p
r
o
d
u
ce
b
o
u
n
d
in
g
b
o
x
es
an
d
co
n
f
i
d
en
ce
s
co
r
es.
T
o
b
o
o
s
t
b
o
th
ac
cu
r
ac
y
a
n
d
ef
f
icien
cy
,
th
e
m
o
d
el
u
s
es
atten
tio
n
-
b
ased
i
n
tr
ascale
f
ea
tu
r
e
in
ter
ac
tio
n
(
AI
FI
)
[
2
4
]
a
n
d
C
NN
-
b
ased
cr
o
s
s
-
s
ca
le
f
ea
tu
r
e
f
u
s
io
n
(
C
C
F
M)
[
2
5
]
.
AI
FI
h
elp
s
cu
t
d
o
wn
r
ed
u
n
d
a
n
cy
at
s
tag
e
S5
wh
ile
s
till
ca
p
tu
r
in
g
th
e
r
elatio
n
s
h
ip
s
b
etwe
en
h
i
g
h
-
le
v
el
s
em
an
tic
f
ea
tu
r
es,
wh
ic
h
s
u
p
p
o
r
ts
o
b
ject
d
etec
tio
n
.
T
h
e
m
o
d
el
also
s
k
ip
s
lo
w
-
lev
el
in
tr
a
-
s
ca
le
in
ter
ac
tio
n
s
b
ec
au
s
e
th
ey
lack
s
em
an
tic
m
ea
n
in
g
an
d
ca
n
ca
u
s
e
d
u
p
l
icatio
n
is
s
u
es
[
2
6
]
.
RT
-
DE
T
R
al
s
o
tack
les
in
co
n
s
is
ten
cies
b
etwe
en
cla
s
s
if
icat
io
n
s
co
r
es
an
d
I
o
U
co
n
f
id
en
ce
d
is
tr
ib
u
tio
n
s
[
2
7
]
.
Du
r
in
g
tr
ain
in
g
,
th
e
m
o
d
el
is
d
esig
n
ed
to
lin
k
h
ig
h
I
o
U
s
co
r
es
to
h
ig
h
class
if
icatio
n
s
co
r
es,
wh
ich
h
elp
s
p
r
ev
en
t
in
ac
cu
r
ate
p
r
ed
ictio
n
s
an
d
av
o
id
s
s
elec
tin
g
b
o
x
es
th
at
h
av
e
lo
w
I
o
U
s
co
r
es
ev
en
if
th
ey
h
av
e
h
ig
h
class
if
icatio
n
s
co
r
es
[
1
4
]
.
T
h
is
o
p
tim
izatio
n
im
p
r
o
v
es
p
er
f
o
r
m
an
ce
b
y
alig
n
in
g
class
if
icatio
n
an
d
lo
ca
tio
n
co
n
f
id
en
ce
ef
f
ec
tiv
ely
.
T
h
e
d
e
tecto
r
o
p
tim
izatio
n
g
o
al
ca
n
b
e
r
ep
h
r
ased
i
n
(
1
)
.
(
̂
,
)
=
(
̂
,
)
+
(
̂
,
̂
,
,
)
=
(
̂
,
)
+
(
̂
,
,
)
(
1
)
W
h
er
e
̂
an
d
d
e
n
o
te
p
r
ed
ictio
n
an
d
g
r
o
u
n
d
tr
u
th
,
̂
={
,
̂
}
an
d
={
̂
,
̂
},
an
d
r
ep
r
esen
t
ca
teg
o
r
i
es
an
d
b
o
u
n
d
in
g
b
o
x
es,
r
esp
ec
tiv
el
y
[
1
4
]
.
2
.
5
.
Rea
l
-
t
im
e
det
ec
t
io
n t
ra
ns
f
o
rm
er
m
o
difica
t
io
n
Fig
u
r
e
4
illu
s
tr
ates
th
e
f
u
s
io
n
b
lo
ck
em
p
lo
y
ed
in
th
e
C
FF
M
f
r
am
ewo
r
k
,
wh
ich
is
s
p
ec
if
ically
d
esig
n
ed
to
en
h
a
n
ce
f
ea
tu
r
e
i
n
ter
ac
tio
n
s
an
d
im
p
r
o
v
e
o
v
e
r
all
m
o
d
el
p
er
f
o
r
m
an
ce
.
As
s
h
o
wn
in
Fig
u
r
e
4
(
a)
,
th
er
e
is
a
f
u
s
io
n
b
lo
ck
th
at
ai
m
s
to
co
m
b
in
e
ad
jace
n
t
f
ea
tu
r
es
in
to
n
ew
f
ea
tu
r
es.
T
h
is
f
u
s
i
o
n
b
lo
ck
co
n
tain
s
n
r
ep
b
lo
ck
s
[
1
4
]
an
d
th
e
o
u
t
p
u
t
s
o
f
two
p
ath
s
ar
e
f
u
s
ed
th
r
o
u
g
h
s
eq
u
en
tial
ad
d
itio
n
o
f
elem
en
ts
.
T
h
e
im
p
r
o
v
ed
f
u
s
io
n
b
lo
ck
is
d
ep
icted
in
Fig
u
r
e
4
(
b
)
,
ea
ch
f
u
s
io
n
b
lo
ck
g
ai
n
s
o
n
e
m
o
r
e
co
n
v
o
l
u
tio
n
lay
er
to
im
p
r
o
v
e
im
a
g
e
o
b
ject
d
etec
tio
n
ac
cu
r
ac
y
.
C
o
n
v
o
lu
tio
n
lay
er
s
ar
e
u
s
ed
in
i
m
ag
e
p
r
o
ce
s
s
in
g
to
r
ec
o
g
n
ize
lo
ca
l
p
atter
n
s
an
d
allo
w
th
e
m
o
d
el
to
u
n
d
er
s
ta
n
d
in
cr
ea
s
in
g
ly
co
m
p
licated
f
ea
tu
r
e
h
ier
ar
c
h
ies.
T
h
e
co
n
v
o
lu
tio
n
lay
er
ca
n
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
4
,
No
.
5
,
Octo
b
er
2
0
2
5
:
4
0
9
0
-
4
1
0
0
4094
ex
tr
ac
t
s
ig
n
if
ican
t
im
ag
e
elem
en
ts
in
clu
d
in
g
ed
g
es,
tex
tu
r
es,
an
d
o
th
er
v
is
u
al
p
atter
n
s
[
2
8
]
.
T
h
is
ass
is
t
s
th
e
m
o
d
el
in
co
m
p
r
eh
en
d
in
g
th
e
d
is
tin
ct
q
u
alities
o
f
th
e
o
b
ject
b
ein
g
r
ec
o
g
n
ized
.
T
h
is
im
p
r
o
v
em
en
t
is
lik
ely
to
aid
in
o
b
ject
d
etec
tio
n
,
p
a
r
ticu
lar
ly
f
o
r
t
h
in
g
s
with
co
m
p
ar
ab
le
b
u
t d
is
tin
ct
v
ar
iatio
n
s
.
Fig
u
r
e
3
.
T
h
e
R
T
-
DE
T
R
ar
ch
i
tectu
r
e
(
a)
(
b
)
Fig
u
r
e
4
.
T
h
e
f
u
s
io
n
b
lo
ck
in
t
h
e
C
C
FM,
wh
ich
is
d
esig
n
ed
to
en
h
a
n
ce
f
ea
tu
r
e
in
ter
ac
tio
n
f
o
r
im
p
r
o
v
ed
m
o
d
el
p
er
f
o
r
m
a
n
ce
co
n
s
is
ts
o
f
two
co
m
p
o
n
en
ts
: (
a)
o
r
ig
in
al
f
u
s
io
n
b
lo
c
k
an
d
(
b
)
m
o
d
if
ica
tio
n
f
u
s
io
n
b
lo
ck
2
.
6
.
P
er
f
o
r
m
a
nce
pa
ra
m
et
e
r
Pre
cisi
o
n
,
r
ec
all,
an
d
m
AP
ar
e
s
o
m
e
f
r
eq
u
e
n
tly
u
s
ed
m
etr
ics
in
ass
e
s
s
in
g
th
e
ef
f
ec
tiv
en
ess
o
f
m
ac
h
in
e
lear
n
in
g
m
o
d
els,
esp
ec
ially
in
th
e
co
n
tex
t
o
f
o
b
j
ec
t
d
etec
tio
n
.
T
h
ese
p
a
r
am
et
er
s
h
elp
u
n
d
er
s
tan
d
h
o
w
well
th
e
m
o
d
el
d
etec
ts
an
d
r
ec
o
g
n
izes
th
e
d
esire
d
o
b
jects.
T
h
e
r
atio
o
f
tr
u
e
p
o
s
itiv
es,
o
r
ac
cu
r
ate
f
o
r
ec
asts
o
f
r
ea
l
o
b
jects,
to
all
p
o
s
itiv
e
p
r
ed
ictio
n
s
,
in
c
lu
d
in
g
in
ac
c
u
r
ate
o
n
es,
is
k
n
o
wn
as
p
r
ec
is
io
n
.
Pre
cisi
o
n
g
au
g
es
h
o
w
ac
cu
r
at
e
th
e
m
o
d
el
is
at
m
ak
in
g
p
r
ed
ictio
n
s
;
th
at
is
,
wh
at
p
r
o
p
o
r
tio
n
o
f
all
th
e
m
o
d
el'
s
p
o
s
itiv
e
p
r
ed
ictio
n
s
ar
e
tr
u
e
in
(
2
)
.
=
(
)
(
)
+
(
)
(
2
)
R
ec
all
is
th
e
r
atio
o
f
t
h
e
n
u
m
b
er
o
f
tr
u
e
p
o
s
itiv
es
to
th
e
to
t
al
n
u
m
b
e
r
o
f
ac
tu
al
o
b
jects
(
t
h
e
s
u
m
o
f
tr
u
e
p
o
s
itiv
es
an
d
f
alse
n
e
g
ativ
es).
R
ec
all
m
ea
s
u
r
es
th
e
m
o
d
el’
s
ab
ilit
y
to
f
in
d
all
i
n
s
tan
ce
s
o
f
th
e
o
b
ject,
a
s
d
ep
icted
in
(
3
)
.
=
(
)
(
)
+
(
)
(
3
)
T
h
e
av
er
a
g
e
o
f
th
e
av
er
a
g
e
p
r
ec
is
io
n
(
AP)
o
v
e
r
all
class
es
is
k
n
o
wn
as
m
ea
n
a
v
er
ag
e
p
r
e
cisi
o
n
,
o
r
m
AP.
Fo
r
ev
er
y
class
,
AP
is
th
e
ar
ea
u
n
d
er
t
h
e
p
r
ec
is
io
n
-
r
ec
all
cu
r
v
e.
T
o
g
iv
e
a
th
o
r
o
u
g
h
p
ictu
r
e
o
f
th
e
m
o
d
el'
s
ef
f
ec
tiv
en
ess
in
id
en
ti
f
y
in
g
o
b
jects
f
r
o
m
all
test
ed
cl
ass
es,
m
A
P
in
teg
r
ates
ac
cu
r
ac
y
an
d
r
ec
all.
AP
is
ca
lcu
lated
f
o
r
ea
c
h
class
an
d
t
h
en
av
er
a
g
ed
as p
a
r
t o
f
t
h
e
co
m
p
u
tatio
n
p
r
o
ce
d
u
r
e.
T
h
e
m
A
P f
o
r
m
u
la
in
(
4
)
.
=
1
∑
=
1
(
4
)
W
h
er
e
is
th
e
n
u
m
b
e
r
o
f
clas
s
es a
n
d
is
th
e
AP f
o
r
class
.
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
A
n
imp
r
o
ve
d
r
ea
l time
d
etec
tio
n
tr
a
n
s
fo
r
mer m
eth
o
d
fo
r
r
eta
il p
r
o
d
u
ct
d
etec
ti
on
(
A
n
d
i Wa
h
yu
Ma
u
l
a
n
a
)
4095
3.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
N
3
.
1
.
E
x
perim
ent
prepa
ra
t
i
o
n
B
ef
o
r
e
tr
ain
in
g
,
n
u
m
e
r
o
u
s
s
e
ttin
g
s
in
th
e
R
T
-
DE
T
R
m
o
d
el
co
n
f
ig
u
r
atio
n
s
ec
tio
n
m
u
s
t
b
e
m
ad
e.
T
h
ese
p
ar
am
eter
s
will
th
en
b
e
ch
ec
k
ed
o
n
all
d
atasets
.
T
ab
le
2
d
is
p
lay
s
th
e
p
ar
am
eter
s
ett
in
g
s
.
T
h
e
NVI
DI
A
DGX
A1
0
0
d
ev
ice
is
u
s
ed
f
o
r
tr
ain
in
g
in
th
is
test
.
T
ab
le
2
.
T
r
ai
n
in
g
p
ar
am
eter
s
etu
p
P
a
r
a
me
t
e
r
s
V
a
l
u
e
Ep
o
c
h
s
1
5
0
B
a
t
c
h
si
z
e
16
O
p
t
i
mi
z
e
r
A
u
t
o
I
n
i
t
i
a
l
l
e
a
r
n
i
n
g
r
a
t
e
1
×
10
−
2
3
.
2
.
O
ur
da
t
a
s
et
re
s
ult
a
nd
dis
cu
s
s
io
n
T
h
is
s
tu
d
y
ad
d
r
ess
es
g
ap
s
i
d
en
tifie
d
in
p
r
ev
i
o
u
s
r
esear
ch
,
s
u
ch
as
th
o
s
e
b
y
San
tr
a
et
a
l
.
[
7
]
,
Hsi
a
et
a
l
.
[
8
]
,
an
d
L
ee
et
a
l
.
[
9
]
,
w
h
o
u
tili
ze
d
v
a
r
io
u
s
ar
ch
itectu
r
es
lik
e
R
C
-
Net,
Fas
ter
R
-
C
NN,
an
d
YOL
Ov
5
f
o
r
r
etail
p
r
o
d
u
ct
d
e
tectio
n
.
W
h
ile
th
ese
m
eth
o
d
s
im
p
r
o
v
e
d
o
b
ject
d
etec
tio
n
ac
c
u
r
ac
y
,
th
ey
d
id
n
o
t
f
u
lly
ad
d
r
ess
ch
allen
g
es
r
elate
d
to
in
tr
a
-
class
v
ar
iatio
n
,
s
u
ch
as
d
is
tin
g
u
is
h
in
g
s
u
b
tle
d
if
f
er
en
ce
s
b
etwe
en
s
im
ilar
p
r
o
d
u
cts.
T
h
is
r
esear
c
h
f
o
cu
s
es
o
n
m
o
d
if
y
in
g
t
h
e
R
T
-
DE
T
R
m
o
d
el
to
b
etter
h
an
d
le
th
ese
ch
allen
g
es,
p
ar
ticu
lar
ly
in
e
n
h
an
ci
n
g
o
b
je
ct
d
etec
tio
n
ac
cu
r
ac
y
b
y
ac
co
u
n
tin
g
f
o
r
in
tr
a
-
class
v
ar
iatio
n
s
.
T
h
e
d
ataset
was
tr
ain
ed
o
n
t
h
e
m
o
d
if
ied
R
T
-
DE
T
R
m
o
d
e
l
f
o
r
1
5
0
ep
o
c
h
s
.
Fig
u
r
e
5
p
r
esen
ts
th
e
p
er
f
o
r
m
an
ce
e
v
alu
atio
n
g
r
a
p
h
s
o
f
th
e
p
r
o
p
o
s
ed
m
o
d
el,
illu
s
tr
atin
g
its
ef
f
ec
tiv
in
ess
ac
r
o
s
s
d
if
f
er
en
t
m
etr
ics.
Sp
ec
ially
,
Fig
u
r
e
5
(
a
)
illu
s
tr
ate
class
if
icatio
n
lo
s
s
an
d
L
1
lo
s
s
.
C
la
s
s
if
icat
io
n
lo
s
s
esti
m
ate
s
th
e
co
r
r
ec
t
o
b
ject
ca
teg
o
r
y
in
th
e
b
o
u
n
d
in
g
b
o
x
,
wh
ile
L
1
lo
s
s
ca
lcu
lates
th
e
ab
s
o
lu
te
d
if
f
er
en
ce
b
etwe
en
e
x
p
ec
ted
a
n
d
tar
g
et
v
alu
es.
T
h
e
lo
s
s
r
esu
lts
f
o
r
c
lass
if
icatio
n
an
d
L
1
ar
e
s
tab
le
d
u
r
i
n
g
tr
ai
n
in
g
b
u
t
u
n
s
tab
l
e
d
u
r
i
n
g
v
alid
atio
n
,
s
tab
ilizin
g
af
ter
1
0
0
ep
o
c
h
s
.
T
h
e
d
ataset
u
s
ed
is
s
elf
-
g
en
er
ated
,
an
d
au
g
m
en
tatio
n
ca
u
s
e
d
s
o
m
e
in
s
tab
ilit
y
,
p
ar
ticu
lar
ly
in
g
en
er
alizin
g
d
ata
u
n
s
ee
n
d
u
r
in
g
tr
ain
i
n
g
.
A
cc
u
r
ac
y
r
esu
lts
o
f
th
e
u
p
d
ate
d
R
T
-
DE
T
R
m
o
d
el,
in
clu
d
in
g
p
r
ec
is
io
n
,
r
ec
all,
m
AP@
0
.
5
,
an
d
m
AP@
0
.
5
:0
.
9
5
,
ar
e
s
h
o
wn
in
Fig
u
r
e
5
(
b
)
.
T
h
ese
r
esu
lts
ar
e
g
o
o
d
,
b
u
t
lear
n
in
g
iter
atio
n
s
tab
ilit
y
im
p
r
o
v
es
af
ter
1
0
0
e
p
o
ch
s
d
u
e
to
th
e
d
ataset'
s
cu
to
u
t,
allo
win
g
b
etter
r
ec
o
g
n
itio
n
o
f
p
r
o
d
u
cts
an
d
h
an
d
lin
g
o
f
in
tr
a
-
class
v
ar
iatio
n
.
T
ab
le
3
d
is
p
lay
s
th
e
r
esu
lt
s
o
f
tr
ain
in
g
th
r
ee
m
o
d
els:
YOL
Ov
8
,
R
T
-
DE
T
R
,
an
d
th
e
m
o
d
if
ied
R
T
-
DE
T
R
.
T
h
e
r
esu
lts
s
h
o
w
th
at
wh
en
th
e
d
ataset
is
r
u
n
with
th
e
YOL
Ov
8
m
o
d
el,
p
r
e
cisi
o
n
ac
cu
r
ac
y
r
ea
ch
es
9
1
.
1
%,
r
ec
all
is
ap
p
r
o
x
im
ately
9
4
.
0
7
%,
m
AP@
0
.
5
is
h
ig
h
at
9
8
.
7
%,
b
u
t
m
AP@
0
.
5
:0
.
9
5
is
r
elativ
el
y
lo
w
at
7
7
.
4
%.
T
h
e
R
T
-
DE
T
R
m
o
d
el
o
u
t
p
er
f
o
r
m
s
YOL
Ov
8
with
a
p
r
ec
is
io
n
o
f
9
7
.
4
%,
r
ec
all
o
f
9
9
.
5
%,
m
AP@
0
.
5
o
f
9
9
.
5
%,
an
d
m
AP@
0
.
5
:0
.
9
5
o
f
8
5
.
7
%.
T
h
e
h
i
g
h
est
p
er
f
o
r
m
an
ce
is
ac
h
iev
e
d
b
y
t
h
e
m
o
d
i
f
ied
R
T
-
DE
T
R
m
o
d
el,
wh
ich
h
as
a
p
r
ec
is
io
n
o
f
9
8
.
7
%,
r
ec
all
o
f
9
9
.
6
%
(
a
0
.
1
%
im
p
r
o
v
em
e
n
t
o
v
e
r
R
T
-
DE
T
R
)
,
m
AP@
0
.
5
o
f
9
9
.
5
%
(
m
atch
in
g
th
e
o
r
ig
in
al
R
T
-
DE
T
R
)
,
an
d
m
AP@
0
.
5
:0
.
9
5
o
f
8
8
.
2
%,
s
u
r
p
ass
in
g
th
e
o
r
ig
in
al
R
T
-
DE
T
R
m
o
d
el.
(
a)
(
b
)
Fig
u
r
e
5
.
Per
f
o
r
m
an
c
e
ev
alu
at
io
n
g
r
a
p
h
s
o
f
t
h
e
p
r
o
p
o
s
ed
m
o
d
el:
(
a)
class
if
icatio
n
an
d
L
1
l
o
s
s
g
r
ap
h
f
o
r
tr
ain
an
d
v
alid
atio
n
an
d
(
b
)
ac
cu
r
ac
y
g
r
ap
h
o
f
p
r
ec
is
io
n
,
r
ec
all,
m
AP
@
0
.
5
an
d
m
AP
@
0
.
5
:0
.
9
5
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
4
,
No
.
5
,
Octo
b
er
2
0
2
5
:
4
0
9
0
-
4
1
0
0
4096
T
ab
le
3
.
R
esu
lt a
n
d
co
m
p
ar
is
o
n
in
o
u
r
d
ataset
M
e
t
h
o
d
P
r
e
c
i
ss
i
o
n
(
%)
R
e
c
a
l
l
(
%)
mA
P
@
0
.
5
(
%)
mA
P
@
0
.
5
:
0
.
9
5
(
%)
Y
O
LO
v
8
[
2
9
]
9
1
.
1
9
4
.
0
7
98
.
7
7
7
.
4
RT
-
D
ET
R
[
2
9
]
9
7
.
4
9
9
.
5
9
9
.
5
8
5
.
7
RT
-
D
ET
R
M
o
d
9
8
.
7
9
9
.
6
9
9
.
5
8
8
.
2
T
h
e
m
o
d
if
ied
R
T
-
DE
T
R
m
o
d
el
p
r
o
v
es
th
at
it
ca
n
o
u
tp
er
f
o
r
m
th
e
o
r
ig
i
n
al
m
o
d
el
in
th
e
f
ield
o
f
o
b
ject
d
etec
tio
n
th
at
p
ay
s
atten
tio
n
to
th
e
in
tr
a
-
class
v
a
r
iatio
n
p
ar
t.
Fig
u
r
e
6
s
h
o
ws
th
e
v
alid
atio
n
r
esu
lts
o
f
th
e
m
o
d
if
ied
R
T
-
DE
T
R
m
o
d
el.
T
h
e
v
alid
atio
n
r
esu
lts
ar
e
s
h
o
wn
in
Fig
u
r
e
6
(
a)
.
T
h
er
e
ar
e
s
ix
class
o
b
jects
f
o
u
n
d
,
an
d
th
e
p
r
e
d
ictio
n
s
ar
e
all
co
r
r
ec
t.
T
h
e
co
m
p
ar
is
o
n
f
in
d
in
g
s
o
f
ea
ch
class
f
r
o
m
th
e
d
ataset
ar
e
d
is
p
lay
ed
in
T
ab
le
4
.
I
n
ter
m
s
o
f
m
AP
@
0
.
5
ac
cu
r
ac
y
,
it
ap
p
ea
r
s
to
b
e
9
9
.
5
%
ac
r
o
s
s
all
class
e
s
.
A
clea
r
er
co
m
p
ar
is
o
n
ca
n
b
e
s
ee
n
in
th
e
ac
cu
r
ac
y
o
f
m
AP
@
0
.
5
:0
.
9
5
,
in
wh
ich
ea
ch
class
is
ex
ce
ed
ed
b
y
t
h
e
m
o
d
if
ied
RT
-
DE
T
R
m
o
d
el.
I
t
is
a
p
p
ar
e
n
t
th
at
th
e
im
p
r
o
v
ed
R
T
-
DE
T
R
m
o
d
el
p
r
o
d
u
ce
s
b
etter
r
esu
lts
th
an
th
e
o
r
ig
in
a
l
m
o
d
el,
an
d
th
is
m
o
d
if
icatio
n
h
as
p
r
o
v
e
n
th
at
t
h
e
ef
f
ec
t
o
f
ad
d
in
g
c
o
n
v
o
lu
tio
n
lay
er
s
to
th
e
C
C
FM
f
u
s
io
n
b
lo
ck
ca
n
in
cr
ea
s
e
m
o
d
el
p
er
f
o
r
m
an
ce
ac
c
u
r
ac
y
.
T
ab
le
4
.
C
o
m
p
a
r
is
o
n
o
f
th
e
r
e
s
u
lts
f
o
r
ea
ch
class
f
o
r
th
e
R
T
-
DE
T
R
an
d
m
o
d
if
ie
d
R
T
-
DE
T
R
m
o
d
els
C
l
a
s
s
RT
-
D
ET
R
RT
-
D
ET
R
m
o
d
i
f
i
e
d
mA
P
@
0
.
5
(
%)
mA
P
@
0
.
5
:
0
.
9
5
(
%)
mA
P
@
0
.
5
(
%)
mA
P
@
0
.
5
:
0
.
9
5
(
%)
B
u
a
v
i
t
a
Ja
mb
u
9
9
.
5
7
4
.
4
9
9
.
5
7
9
.
4
C
h
i
t
a
t
o
L
i
t
e
R
u
m
p
u
t
La
u
t
9
9
.
5
9
0
.
3
9
9
.
5
9
2
.
4
O
r
e
o
O
r
i
g
i
n
a
l
9
9
.
5
8
8
.
3
9
9
.
5
9
1
.
3
R
e
d
B
u
l
l
D
r
i
n
k
9
9
.
5
8
8
.
2
9
9
.
5
9
0
.
6
S
a
r
i
G
a
n
d
u
m
C
o
k
e
l
a
t
9
9
.
5
8
6
.
1
9
9
.
5
8
6
.
6
S
l
a
i
O
l
a
i
S
t
r
o
b
e
r
i
9
9
.
5
8
6
.
9
9
9
.
5
8
9
.
1
3
.
3
.
O
t
her
da
t
a
s
et
re
s
ult
a
n
d dis
cu
s
s
io
n
3.
3
.
1
.
G
ro
ce
r
y
da
t
a
s
et
re
s
ult
a
nd
dis
cu
s
s
io
n
T
ab
le
5
p
r
esen
ts
th
e
r
esu
lts
o
f
tr
ain
in
g
YOL
Ov
8
,
R
T
-
DE
T
R
an
d
m
o
d
if
ied
R
T
-
DE
T
R
m
o
d
els.
T
h
e
d
ataset
is
r
u
n
with
th
e
YOL
Ov
8
m
o
d
el
th
e
r
esu
lt
o
f
p
r
ec
is
io
n
ac
cu
r
ac
y
is
9
9
.
8
%,
th
e
r
es
u
lt
o
f
r
ec
all
r
ea
ch
es
ar
o
u
n
d
9
9
.
8
%,
m
AP@
0
.
5
is
9
9
.
4
%
an
d
f
o
r
m
AP@
0
.
5
:0
.
9
5
is
8
2
.
1
%.
T
h
e
n
th
e
R
T
-
DE
T
R
m
o
d
el
p
r
o
d
u
ce
s
a
p
r
ec
is
io
n
v
alu
e
o
f
9
9
.
8
%
th
e
s
am
e
as
th
e
YOL
O
v
8
m
o
d
el,
th
e
r
ec
all
v
alu
e
r
ea
c
h
es
9
9
.
9
%,
th
e
m
AP@
0
.
5
v
alu
e
r
ea
ch
es
9
9
.
5
%
an
d
m
A
P@
0
.
5
:0
.
9
5
r
ea
ch
es
8
3
.
7
%
in
d
icatin
g
th
at
th
e
R
T
-
DE
T
R
m
o
d
el
is
s
u
p
er
io
r
to
th
e
YOL
Ov
8
m
o
d
el.
T
h
e
h
ig
h
est
p
er
f
o
r
m
an
ce
is
ac
h
iev
ed
b
y
th
e
m
o
d
if
ied
R
T
-
DE
T
R
m
o
d
el,
wh
ich
h
as
a
p
r
ec
is
io
n
o
f
9
9
.
9
%
(
a
0
.
1
%
im
p
r
o
v
em
en
t
o
v
er
t
h
e
o
r
i
g
in
al
R
T
-
DE
T
R
)
,
r
ec
all
o
f
9
9
.
9
%
(
m
atch
in
g
th
e
o
r
ig
in
al
R
T
-
DE
T
R
)
,
m
AP@
0
.
5
o
f
9
9
.
5
%
(
th
e
s
am
e
as
th
e
o
r
ig
in
al
R
T
-
DE
T
R
)
,
an
d
m
AP@
0
.
5
:0
.
9
5
o
f
8
4
.
2
%,
s
u
r
p
ass
in
g
th
e
o
r
ig
in
al
m
o
d
el.
T
h
e
m
o
d
if
ie
d
R
T
-
DE
T
R
m
o
d
el
p
r
o
v
es
th
at
it
ca
n
o
u
tp
er
f
o
r
m
th
e
o
r
ig
in
al
m
o
d
el
in
th
e
f
ield
o
f
d
etec
tio
n
o
b
jects
th
at
p
ay
at
ten
tio
n
to
th
e
in
tr
a
-
class
v
ar
i
atio
n
p
ar
t.
Fo
r
t
h
e
v
alid
atio
n
r
esu
lts
ca
n
b
e
s
ee
n
in
Fig
u
r
e
6
(
b
)
.
3.
3
.
2
.
R
et
a
il pro
du
ct
chec
k
o
ut
da
t
a
s
et
re
s
ult
a
nd
dis
cu
s
s
i
o
n
T
ab
le
6
p
r
esen
ts
th
e
r
esu
lts
o
f
tr
ain
in
g
YOL
Ov
8
,
R
T
-
DE
T
R
an
d
m
o
d
if
ied
R
T
-
DE
T
R
m
o
d
els.
T
h
e
d
ataset
is
r
u
n
with
th
e
YOL
Ov
8
m
o
d
el
th
e
r
esu
lt
o
f
p
r
ec
is
io
n
ac
cu
r
ac
y
is
9
9
.
8
%,
th
e
r
es
u
lt
o
f
r
ec
all
r
ea
ch
es
ar
o
u
n
d
9
9
.
8
%,
m
AP@
0
.
5
is
9
9
.
2
%
an
d
f
o
r
m
AP@
0
.
5
:0
.
9
5
is
8
6
.
4
%.
T
h
e
n
th
e
R
T
-
DE
T
R
m
o
d
el
p
r
o
d
u
ce
s
a
p
r
ec
is
io
n
v
alu
e
o
f
9
9
.
8
%
th
e
s
am
e
as
th
e
YOL
OV8
m
o
d
el
,
th
e
r
ec
all
v
alu
e
r
ea
ch
es
9
9
.
8
%,
th
e
m
AP@
0
.
5
v
alu
e
r
ea
ch
es
9
9
.
4
%
an
d
m
AP@
0
.
5
:0
.
9
5
r
ea
ch
es
8
8
.
0
3
%
with
th
is
s
tatin
g
th
at
th
e
R
T
-
DE
T
R
m
o
d
el
i
s
s
u
p
er
io
r
t
o
th
e
YOL
Ov
8
m
o
d
el.
T
h
e
h
ig
h
est
p
er
f
o
r
m
a
n
ce
is
ac
h
iev
ed
b
y
th
e
m
o
d
if
ied
R
T
-
DE
T
R
m
o
d
el,
wh
ich
h
as
a
p
r
ec
is
io
n
o
f
9
9
.
9
%
(
a
0
.
1
%
im
p
r
o
v
e
m
en
t
o
v
er
b
o
th
t
h
e
o
r
ig
in
al
R
T
-
DE
T
R
an
d
YOL
Ov
8
m
o
d
els),
r
ec
all
o
f
9
9
.
8
%
(
th
e
s
am
e
as
th
e
o
r
ig
in
al
R
T
-
D
E
T
R
an
d
YOL
Ov
8
)
,
m
AP@
0
.
5
o
f
9
9
.
5
%,
an
d
m
AP@
0
.
5
:0
.
9
5
o
f
8
8
.
2
%,
s
u
r
p
ass
in
g
th
e
o
r
ig
in
al
R
T
-
DE
T
R
m
o
d
el.
T
h
e
m
o
d
i
f
ied
R
T
-
D
E
T
R
m
o
d
el
p
r
o
v
es
th
at
it
ca
n
o
u
tp
er
f
o
r
m
th
e
o
r
i
g
in
al
m
o
d
el
in
th
e
f
ield
o
f
d
et
ec
tio
n
o
b
jects
th
at
p
ay
atten
tio
n
to
th
e
in
tr
a
-
class
v
ar
iatio
n
p
a
r
t.
Fo
r
th
e
v
alid
ati
o
n
r
esu
lts
ca
n
b
e
s
ee
n
in
Fig
u
r
e
6
(
c)
.
T
ab
le
5
.
R
esu
lt a
n
d
co
m
p
ar
is
o
n
in
Gr
o
ce
r
y
d
ataset
M
e
t
h
o
d
P
r
e
c
i
ss
i
o
n
(
%)
R
e
c
a
l
l
(
%)
mA
P
@
0
.
5
(
%)
mA
P
@
0
.
5
:
0
.
9
5
(
%)
Y
O
LO
v
8
[
2
9
]
9
9
.
8
9
9
.
8
99
.
4
8
2
.
1
RT
-
D
ET
R
[
2
9
]
9
9
.
8
9
9
.
9
9
9
.
5
8
3
.
7
RT
-
D
ET
R
M
o
d
9
9
.
9
9
9
.
9
9
9
.
5
8
4
.
2
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
A
n
imp
r
o
ve
d
r
ea
l time
d
etec
tio
n
tr
a
n
s
fo
r
mer m
eth
o
d
fo
r
r
eta
il p
r
o
d
u
ct
d
etec
ti
on
(
A
n
d
i Wa
h
yu
Ma
u
l
a
n
a
)
4097
T
ab
le
6
.
R
esu
lt a
n
d
co
m
p
ar
is
o
n
in
R
PC
d
ataset
M
e
t
h
o
d
P
r
e
c
i
ss
i
o
n
(
%)
R
e
c
a
l
l
(
%)
mA
P
@
0
.
5
(
%)
mA
P
@
0
.
5
:
0
.
9
5
(
%)
Y
O
LO
v
8
[
2
9
]
9
9
.
8
9
9
.
8
99
.
2
8
6
.
4
RT
-
D
ET
R
[
2
9
]
9
9
.
8
9
9
.
8
9
9
.
4
8
8
.
0
3
RT
-
D
ET
R
M
o
d
9
9
.
9
9
9
.
8
9
9
.
5
8
8
.
2
3.
3
.
3
.
D
ens
ely
s
eg
m
ent
ed
s
u
perm
a
rk
et
da
t
a
s
et
re
s
ult
a
n
d dis
cu
s
s
io
n
T
ab
le
7
p
r
esen
ts
th
e
r
esu
lts
o
f
tr
ain
in
g
YOL
Ov
8
,
R
T
-
DE
T
R
an
d
m
o
d
if
ied
R
T
-
DE
T
R
m
o
d
els.
T
h
e
d
ataset
is
r
u
n
with
th
e
YOL
Ov
8
m
o
d
el
th
e
r
esu
lt
o
f
p
r
ec
is
io
n
ac
cu
r
ac
y
is
9
1
.
8
%,
th
e
r
es
u
lt
o
f
r
ec
all
r
ea
ch
es
ar
o
u
n
d
9
0
.
9
%,
m
AP@
0
.
5
is
8
1
.
9
%
an
d
f
o
r
m
AP@
0
.
5
:0
.
9
5
is
5
8
.
2
%.
T
h
e
n
th
e
R
T
-
DE
T
R
m
o
d
el
p
r
o
d
u
ce
s
a
p
r
ec
is
io
n
v
alu
e
o
f
9
3
.
9
%,
t
h
e
r
ec
all
v
alu
e
r
ea
ch
es
8
4
.
0
5
%,
th
e
m
AP@
0
.
5
v
alu
e
r
ea
ch
es
9
1
.
8
%
an
d
m
AP@
0
.
5
:0
.
9
5
r
ea
ch
es 7
2
.
0
3
% with
th
is
s
tatin
g
th
at
th
e
R
T
-
DE
T
R
m
o
d
el
is
s
u
p
er
io
r
to
th
e
YOL
Ov
8
m
o
d
el.
T
h
e
h
ig
h
est
p
er
f
o
r
m
a
n
ce
is
ac
h
iev
ed
b
y
th
e
m
o
d
if
ied
R
T
-
DE
T
R
m
o
d
el,
wh
ich
h
as
a
p
r
ec
is
io
n
o
f
9
4
.
1
%
(
0
.
2
%
h
i
g
h
er
th
an
th
e
o
r
ig
in
al
R
T
-
DE
T
R
)
,
r
ec
all
o
f
8
5
.
5
%,
m
AP@
0
.
5
o
f
9
2
.
2
%,
an
d
m
AP@
0
.
5
:0
.
9
5
o
f
7
0
.
6
%.
Alth
o
u
g
h
t
h
e
m
AP@
0
.
5
:0
.
9
5
is
s
lig
h
tly
lo
wer
th
an
th
e
o
r
ig
in
al
m
o
d
el,
th
e
m
o
d
if
ied
R
T
-
DE
T
R
o
u
tp
er
f
o
r
m
s
th
e
o
r
ig
in
al
R
T
-
DE
T
R
in
p
r
ec
is
io
n
,
r
ec
all,
an
d
m
AP@
0
.
5
,
m
ak
in
g
it
th
e
s
u
p
er
io
r
m
o
d
el
o
v
er
all.
T
h
e
s
lig
h
tly
lo
wer
ac
cu
r
ac
y
in
th
is
d
ataset
co
m
p
ar
ed
to
p
r
ev
io
u
s
test
s
is
d
u
e
to
th
e
d
ataset
co
n
tain
in
g
p
r
o
d
u
cts
with
v
ar
iatio
n
s
in
li
g
h
tin
g
,
wh
ich
af
f
ec
te
d
th
e
tr
ain
in
g
p
r
o
ce
s
s
.
T
h
e
m
o
d
if
ie
d
R
T
-
DE
T
R
m
o
d
el
p
r
o
v
es
th
at
it
ca
n
o
u
tp
er
f
o
r
m
th
e
o
r
ig
in
al
m
o
d
el
in
th
e
f
ie
ld
o
f
d
etec
tio
n
o
b
jects
th
at
p
ay
atten
tio
n
to
th
e
in
tr
a
-
class
v
ar
iatio
n
p
ar
t.
T
h
e
v
alid
atio
n
r
esu
lts
ca
n
b
e
s
ee
n
in
Fig
u
r
e
6
(
d
)
.
T
ab
le
7
.
R
esu
lt a
n
d
co
m
p
ar
is
o
n
in
D
2S
d
ataset
M
e
t
h
o
d
P
r
e
c
i
ss
i
o
n
(
%)
R
e
c
a
l
l
(
%)
mA
P
@
0
.
5
(
%)
mA
P
@
0
.
5
:
0
.
9
5
(
%)
Y
O
LO
v
8
[
2
9
]
9
1
.
8
9
0
.
9
8
1
.
9
5
8
.
2
RT
-
D
ET
R
[
2
9
]
9
3
.
9
8
4
.
0
5
9
1
.
8
7
2
.
0
3
RT
-
D
ET
R
M
o
d
9
4
.
1
8
5
.
5
9
2
.
2
7
0
.
6
(
a)
(
b
)
(
c)
(
d
)
Fig
u
r
e
6
.
Valid
atio
n
r
esu
lts
o
f
th
e
R
T
-
DE
T
R
m
o
d
if
icatio
n
m
o
d
el
o
n
(
a)
o
u
r
d
ataset,
(
b
)
Gr
o
ce
r
y
d
ataset,
(
c)
R
PC
-
d
ataset,
an
d
(
d
)
D2
S
-
d
ataset
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
4
,
No
.
5
,
Octo
b
er
2
0
2
5
:
4
0
9
0
-
4
1
0
0
4098
4.
CO
NCLU
SI
O
N
B
a
s
e
d
o
n
t
h
e
r
e
s
u
l
t
s
o
f
t
h
i
s
s
t
u
d
y
,
i
t
i
s
p
o
s
s
i
b
l
e
to
c
o
n
c
lu
d
e
th
a
t
t
h
e
R
T
-
DE
T
R
m
o
d
e
l
c
a
n
b
e
a
d
j
u
s
t
ed
to
in
c
r
ea
s
e
th
e
m
o
d
el
’
s
p
er
f
o
r
m
a
n
c
e
a
cc
u
r
a
c
y
.
W
e
a
d
d
ed
a
co
n
v
o
lu
t
i
o
n
l
ay
e
r
to
e
ac
h
f
u
s
i
o
n
b
lo
ck
i
n
t
h
e
C
C
F
M
f
u
s
i
o
n
b
lo
ck
t
o
i
m
p
r
o
v
e
th
e
a
cc
u
r
a
c
y
o
f
im
ag
e
o
b
j
e
c
t
d
e
t
e
c
t
io
n
.
C
o
n
v
o
l
u
ti
o
n
l
ay
e
r
s
ar
e
u
s
ed
i
n
i
m
ag
e
p
r
o
ce
s
s
i
n
g
to
r
ec
o
g
n
i
z
e
lo
c
a
l
p
a
t
t
e
r
n
s
a
n
d
a
l
l
o
w
t
h
e
m
o
d
e
l
to
u
n
d
er
s
t
a
n
d
in
c
r
e
a
s
i
n
g
l
y
c
o
m
p
l
i
c
a
t
ed
f
e
at
u
r
e
h
i
er
a
r
c
h
ie
s
.
C
o
n
v
o
l
u
t
io
n
l
a
y
e
r
s
c
an
ex
t
r
a
c
t
s
i
g
n
i
f
i
c
an
t
in
f
o
r
m
a
t
i
o
n
f
r
o
m
im
a
g
e
s
l
i
k
e
a
s
e
d
g
e
s
,
t
ex
t
u
r
e
s
,
an
d
o
th
er
v
i
s
u
a
l
p
a
t
t
er
n
s
,
a
s
s
i
s
t
i
n
g
t
h
e
m
o
d
e
l
i
n
u
n
d
e
r
s
t
a
n
d
i
n
g
t
h
e
d
i
s
t
in
c
t
q
u
a
l
i
t
i
e
s
o
f
th
e
i
t
e
m
b
e
in
g
r
e
co
g
n
i
z
ed
.
T
e
s
t
in
g
o
u
r
o
wn
d
a
t
a
s
e
t
a
s
w
e
l
l
a
s
t
h
r
ee
o
th
e
r
d
a
ta
s
e
t
s
d
e
m
o
n
s
t
r
a
te
d
th
a
t
o
u
r
m
o
d
if
i
e
d
R
T
-
D
E
T
R
m
o
d
e
l
m
a
y
i
n
c
r
e
a
s
e
th
e
ac
c
u
r
a
c
y
o
f
p
r
o
d
u
c
t
o
b
j
ec
t
d
e
te
c
t
i
o
n
an
d
ai
d
i
n
th
e
d
e
t
e
c
t
io
n
o
f
p
r
o
d
u
c
t
v
ar
i
a
t
i
o
n
s
.
T
h
e
i
m
p
r
o
v
ed
r
e
s
u
l
t
s
o
f
m
A
P
@
0
.
5
f
o
r
th
e
s
e
l
f
-
p
r
o
v
id
e
d
d
at
a
s
e
t
r
e
a
c
h
ed
9
9
.
5
%
an
d
m
A
P
@
0
.
5
:
0
.
9
5
r
e
a
c
h
e
d
8
8
.
2
%
a
s
a
r
e
s
u
l
t
o
f
o
u
r
m
o
d
if
i
ed
R
T
-
D
E
T
R
m
o
d
e
l
,
w
h
i
ch
al
s
o
a
p
p
l
i
e
s
to
t
h
e
o
t
h
er
t
h
r
e
e
t
y
p
e
s
o
f
d
a
t
a
s
e
ts
t
h
a
t
o
u
tp
e
r
f
o
r
m
ed
t
h
e
o
r
ig
i
n
a
l
m
o
d
e
l
an
d
Y
O
L
Ov
8
.
H
o
w
e
v
er
,
t
h
e
r
e
ar
e
l
i
m
i
t
a
t
i
o
n
s
t
o
t
h
i
s
s
t
u
d
y
.
T
h
e
i
m
p
r
o
v
e
m
e
n
t
s
o
b
s
e
r
v
e
d
i
n
t
h
e
m
o
d
if
i
e
d
R
T
-
D
E
T
R
m
o
d
e
l
m
a
y
n
o
t
b
e
c
o
n
s
i
s
t
e
n
t
a
cr
o
s
s
a
l
l
ty
p
e
s
o
f
d
a
t
a
s
e
t
s
,
p
a
r
t
i
cu
l
a
r
ly
th
o
s
e
w
i
t
h
m
o
r
e
co
m
p
le
x
i
n
tr
a
-
c
l
a
s
s
v
ar
i
a
t
i
o
n
s
o
r
e
x
t
r
em
e
l
i
g
h
t
i
n
g
c
o
n
d
i
t
io
n
s
.
F
u
r
t
h
er
m
o
r
e,
th
e
m
o
d
e
l
's
p
e
r
f
o
r
m
a
n
c
e
c
o
u
l
d
b
e
c
o
n
s
t
r
a
i
n
e
d
b
y
th
e
s
i
z
e
a
n
d
d
i
v
er
s
i
t
y
o
f
t
h
e
d
a
t
a
s
e
t
u
s
e
d
f
o
r
t
r
a
i
n
in
g
.
Fo
r
f
u
tu
r
e
wo
r
k
,
w
e
s
u
g
g
e
s
t
e
x
p
l
o
r
in
g
a
d
d
i
t
i
o
n
a
l
m
o
d
i
f
ic
a
t
i
o
n
s
t
o
t
h
e
R
T
-
D
E
T
R
a
r
ch
i
t
e
c
tu
r
e
b
ey
o
n
d
t
h
e
f
u
s
i
o
n
b
lo
ck
,
s
u
ch
as
i
n
c
o
r
p
o
r
a
t
i
n
g
a
d
v
a
n
c
e
d
a
t
ten
t
i
o
n
m
e
ch
a
n
i
s
m
s
o
r
ex
p
er
i
m
en
t
i
n
g
w
i
t
h
o
t
h
er
ty
p
e
s
o
f
co
n
v
o
lu
t
i
o
n
a
l
l
a
y
e
r
s
.
A
d
d
i
t
io
n
a
l
ly
,
ex
p
an
d
in
g
th
e
d
a
t
a
s
e
t
w
i
t
h
m
o
r
e
d
i
v
er
s
e
p
r
o
d
u
c
t
c
a
te
g
o
r
i
e
s
a
n
d
c
h
a
l
le
n
g
i
n
g
en
v
ir
o
n
m
en
t
s
c
o
u
l
d
h
el
p
f
u
r
t
h
er
e
n
h
a
n
c
e
th
e
m
o
d
e
l
's
r
o
b
u
s
t
n
e
s
s
a
n
d
a
cc
u
r
a
c
y
.
I
n
v
e
s
t
i
g
a
t
i
n
g
t
h
e
im
p
ac
t
o
f
d
i
f
f
er
e
n
t
au
g
m
e
n
t
at
i
o
n
s
t
r
a
t
eg
i
e
s
a
n
d
o
p
t
im
i
z
i
n
g
t
h
e
t
r
a
i
n
in
g
p
r
o
c
e
s
s
co
u
ld
a
l
s
o
l
ea
d
to
b
e
t
t
e
r
g
en
e
r
a
l
iz
a
t
i
o
n
a
cr
o
s
s
v
a
r
io
u
s
r
e
t
a
i
l
s
c
e
n
ar
i
o
s
.
ACK
NO
WL
E
DG
E
M
E
NT
S
T
h
is
wo
r
k
was su
p
p
o
r
ted
b
y
T
elk
o
m
Un
iv
er
s
ity
.
F
UNDING
I
NF
O
R
M
A
T
I
O
N
T
h
is
wo
r
k
was
f
u
n
d
ed
b
y
t
h
e
R
I
I
M
B
atch
3
p
r
o
g
r
am
g
r
an
t
f
r
o
m
L
PDP,
Min
is
tr
y
o
f
Fin
an
ce
,
R
ep
u
b
lic
o
f
I
n
d
o
n
esia,
an
d
B
R
I
N
(
No
.
8
1
/I
V/KS/0
5
/2
0
2
3
)
.
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
An
d
i Wah
y
u
Ma
u
lan
a
✓
✓
✓
✓
✓
✓
✓
✓
✓
Su
r
y
o
Ad
h
i Wi
b
o
wo
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
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
t
h
o
r
s
d
ec
lar
e
th
at
th
e
y
h
av
e
n
o
k
n
o
wn
c
o
m
p
etin
g
f
in
an
cial
in
ter
ests
o
r
p
er
s
o
n
al
r
el
atio
n
s
h
ip
s
th
at
co
u
ld
h
av
e
ap
p
ea
r
ed
t
o
in
f
lu
en
ce
th
e
wo
r
k
r
e
p
o
r
te
d
in
t
h
is
p
ap
er
.
I
NF
O
RM
E
D
CO
NS
E
N
T
W
e
h
av
e
o
b
tain
ed
in
f
o
r
m
ed
c
o
n
s
en
t f
r
o
m
all
in
d
iv
id
u
als in
c
lu
d
ed
in
t
h
is
s
tu
d
y
.
E
T
H
I
CAL AP
P
RO
V
AL
No
t a
p
p
licab
le.
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
A
n
imp
r
o
ve
d
r
ea
l time
d
etec
tio
n
tr
a
n
s
fo
r
mer m
eth
o
d
fo
r
r
eta
il p
r
o
d
u
ct
d
etec
ti
on
(
A
n
d
i Wa
h
yu
Ma
u
l
a
n
a
)
4099
DATA AV
AI
L
AB
I
L
I
T
Y
T
h
e
d
ata
th
at
s
u
p
p
o
r
t th
e
f
in
d
i
n
g
s
o
f
th
is
s
tu
d
y
a
r
e
o
p
en
ly
a
v
ailab
le
in
:
‒
R
o
b
o
f
lo
w
at
h
ttp
s
://u
n
iv
er
s
e.
r
o
b
o
f
lo
w.
c
o
m
/n
ew
-
wo
r
k
s
p
ac
e
-
wf
zw3
/g
r
o
ce
r
y
-
d
ataset
-
q
9
f
j2
‒
R
P
C
Data
s
et
Gi
th
u
b
at
h
ttp
s
://rp
c
-
d
ataset.
g
ith
u
b
.
io
/
‒
Mv
tec
So
f
twar
e
at
h
ttp
s
://www.
m
v
tec.
co
m
/co
m
p
an
y
/r
esear
ch
/d
atasets
/m
v
tec
-
d2s
RE
F
E
R
E
NC
E
S
[
1
]
J
.
Th
ø
g
e
r
sen
,
“
C
o
n
su
me
r
b
e
h
a
v
i
o
r
a
n
d
c
l
i
ma
t
e
c
h
a
n
g
e
:
c
o
n
su
me
r
s
n
e
e
d
c
o
n
si
d
e
r
a
b
l
e
a
s
s
i
s
t
a
n
c
e
,
”
C
u
rr
e
n
t
O
p
i
n
i
o
n
i
n
B
e
h
a
v
i
o
ra
l
S
c
i
e
n
c
e
s
,
v
o
l
.
4
2
,
p
p
.
9
–
1
4
,
D
e
c
.
2
0
2
1
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
c
o
b
e
h
a
.
2
0
2
1
.
0
2
.
0
0
8
.
[
2
]
Y
.
A
mi
t
,
P
.
F
e
l
z
e
n
sz
w
a
l
b
,
a
n
d
R
.
G
i
r
sh
i
c
k
,
“
O
b
j
e
c
t
d
e
t
e
c
t
i
o
n
,
”
i
n
C
o
m
p
u
t
e
r
Vi
si
o
n
,
v
o
l
.
5
,
n
o
.
1
,
C
h
a
m:
S
p
r
i
n
g
e
r
I
n
t
e
r
n
a
t
i
o
n
a
l
P
u
b
l
i
sh
i
n
g
,
2
0
2
0
,
p
p
.
1
–
9
,
d
o
i
:
1
0
.
1
0
0
7
/
9
7
8
-
3
-
030
-
0
3
2
4
3
-
2
_
6
6
0
-
1.
[
3
]
W
.
M
.
Li
m,
S
.
K
u
ma
r
,
N
.
P
a
n
d
e
y
,
D
.
V
e
r
ma
,
a
n
d
D
.
K
u
m
a
r
,
“
E
v
o
l
u
t
i
o
n
a
n
d
t
r
e
n
d
s
i
n
c
o
n
su
me
r
b
e
h
a
v
i
o
u
r
:
i
n
si
g
h
t
s
f
r
o
m
j
o
u
r
n
a
l
o
f
c
o
n
su
m
e
r
b
e
h
a
v
i
o
u
r
,
”
J
o
u
rn
a
l
o
f
C
o
n
s
u
m
e
r
B
e
h
a
v
i
o
u
r
,
v
o
l
.
2
2
,
n
o
.
1
,
p
p
.
2
1
7
–
2
3
2
,
J
a
n
.
2
0
2
3
,
d
o
i
:
1
0
.
1
0
0
2
/
c
b
.
2
1
1
8
.
[
4
]
D
.
G
r
e
w
a
l
,
S
.
B
e
n
o
i
t
,
S
.
M
.
N
o
b
l
e
,
A
.
G
u
h
a
,
C
.
P
.
A
h
l
b
o
m,
a
n
d
J
.
N
o
r
d
f
ä
l
t
,
“
Le
v
e
r
a
g
i
n
g
i
n
-
st
o
r
e
t
e
c
h
n
o
l
o
g
y
a
n
d
A
I
:
i
n
c
r
e
a
si
n
g
c
u
st
o
m
e
r
a
n
d
e
mp
l
o
y
e
e
e
f
f
i
c
i
e
n
c
y
a
n
d
e
n
h
a
n
c
i
n
g
t
h
e
i
r
e
x
p
e
r
i
e
n
c
e
s
,
”
J
o
u
rn
a
l
o
f
Re
t
a
i
l
i
n
g
,
v
o
l
.
9
9
,
n
o
.
4
,
p
p
.
4
8
7
–
5
0
4
,
D
e
c
.
2
0
2
3
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
j
r
e
t
a
i
.
2
0
2
3
.
1
0
.
0
0
2
.
[
5
]
H
.
B
.
-
S
a
l
a
u
,
A
.
J.
O
n
u
man
y
i
,
D
.
M
i
c
h
a
e
l
,
R
.
I
s
a
,
C
.
O
.
A
l
e
n
o
g
h
e
n
a
,
a
n
d
H
.
O
h
i
z
e
,
“
A
n
e
w
a
u
t
o
m
a
t
e
d
s
ma
r
t
c
a
r
t
sy
s
t
e
m
f
o
r
m
o
d
e
r
n
sh
o
p
p
i
n
g
c
e
n
t
r
e
s,
”
Bu
l
l
e
t
i
n
o
f
El
e
c
t
ri
c
a
l
En
g
i
n
e
e
ri
n
g
a
n
d
I
n
f
o
r
m
a
t
i
c
s
,
v
o
l
.
1
0
,
n
o
.
4
,
p
p
.
2
0
2
8
–
2
0
3
6
,
A
u
g
.
2
0
2
1
,
d
o
i
:
1
0
.
1
1
5
9
1
/
E
E
I
.
V
1
0
I
4
.
2
7
6
2
.
[
6
]
N
.
X
.
Ji
e
a
n
d
I
.
F
.
B
.
K
a
m
si
n
,
“
S
e
l
f
-
c
h
e
c
k
o
u
t
se
r
v
i
c
e
w
i
t
h
R
F
I
D
t
e
c
h
n
o
l
o
g
y
i
n
su
p
e
r
m
a
r
k
e
t
,
”
i
n
P
ro
c
e
e
d
i
n
g
s
o
f
t
h
e
3
rd
I
n
t
e
rn
a
t
i
o
n
a
l
C
o
n
f
e
ren
c
e
o
n
I
n
t
e
g
r
a
t
e
d
I
n
t
e
l
l
i
g
e
n
t
C
o
m
p
u
t
i
n
g
C
o
m
m
u
n
i
c
a
t
i
o
n
&
S
e
c
u
ri
t
y
(
I
C
I
I
C
2
0
2
1
)
,
2
0
2
1
,
v
o
l
.
4
,
d
o
i
:
1
0
.
2
9
9
1
/
a
h
i
s
.
k
.
2
1
0
9
1
3
.
0
6
2
.
[
7
]
B
.
S
a
n
t
r
a
,
A
.
K
.
S
h
a
w
,
a
n
d
D
.
P
.
M
u
k
h
e
r
j
e
e
,
“
P
a
r
t
-
b
a
s
e
d
a
n
n
o
t
a
t
i
o
n
-
f
r
e
e
f
i
n
e
-
g
r
a
i
n
e
d
c
l
a
ss
i
f
i
c
a
t
i
o
n
o
f
i
ma
g
e
s
o
f
r
e
t
a
i
l
p
r
o
d
u
c
t
s
,
”
Pa
t
t
e
r
n
Re
c
o
g
n
i
t
i
o
n
,
v
o
l
.
1
2
1
,
J
a
n
.
2
0
2
2
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
p
a
t
c
o
g
.
2
0
2
1
.
1
0
8
2
5
7
.
[
8
]
C
.
H
.
H
si
a
,
T.
H
.
W
.
C
h
a
n
g
,
C
.
Y
.
C
h
i
a
n
g
,
a
n
d
H
.
T.
C
h
a
n
,
“
M
a
sk
R
-
C
N
N
w
i
t
h
n
e
w
d
a
t
a
a
u
g
men
t
a
t
i
o
n
f
e
a
t
u
r
e
s
f
o
r
s
ma
r
t
d
e
t
e
c
t
i
o
n
o
f
r
e
t
a
i
l
p
r
o
d
u
c
t
s,
”
Ap
p
l
i
e
d
S
c
i
e
n
c
e
s
,
v
o
l
.
1
2
,
n
o
.
6
,
M
a
r
.
2
0
2
2
,
d
o
i
:
1
0
.
3
3
9
0
/
a
p
p
1
2
0
6
2
9
0
2
.
[
9
]
R
.
Y
.
L
e
e
,
S
.
Y
.
C
h
u
a
,
Y
.
L
.
La
i
,
T.
Y
.
C
h
a
i
,
S
.
Y
.
W
a
i
,
a
n
d
S
.
C
.
H
a
w
,
“
C
a
sh
i
e
r
l
e
s
s
c
h
e
c
k
o
u
t
v
i
si
o
n
sy
s
t
e
m
f
o
r
s
ma
r
t
r
e
t
a
i
l
u
si
n
g
d
e
e
p
l
e
a
r
n
i
n
g
,
”
J
o
u
rn
a
l
o
f
S
y
st
e
m
a
n
d
Ma
n
a
g
e
m
e
n
t
S
c
i
e
n
c
e
s
,
v
o
l
.
1
2
,
n
o
.
4
,
p
p
.
2
3
2
–
2
5
0
,
A
u
g
.
2
0
2
2
,
d
o
i
:
1
0
.
3
3
1
6
8
/
J
S
M
S
.
2
0
2
2
.
0
4
1
5
.
[
1
0
]
Z
.
Q
.
Zh
a
o
,
P
.
Zh
e
n
g
,
S
.
T
.
X
u
,
a
n
d
X
.
W
u
,
“
O
b
j
e
c
t
d
e
t
e
c
t
i
o
n
w
i
t
h
d
e
e
p
l
e
a
r
n
i
n
g
:
a
r
e
v
i
e
w
,
”
I
E
EE
T
ra
n
s
a
c
t
i
o
n
s
o
n
N
e
u
ra
l
N
e
t
w
o
r
k
s a
n
d
L
e
a
r
n
i
n
g
S
y
st
e
m
s
,
v
o
l
.
3
0
,
n
o
.
1
1
,
p
p
.
3
2
1
2
–
3
2
3
2
,
N
o
v
.
2
0
1
9
,
d
o
i
:
1
0
.
1
1
0
9
/
T
N
N
L
S
.
2
0
1
8
.
2
8
7
6
8
6
5
.
[
1
1
]
n
e
w
-
w
o
r
k
sp
a
c
e
-
w
f
z
w
3
,
“
G
r
o
c
e
r
y
d
a
t
a
se
t
c
o
mp
u
t
e
r
v
i
si
o
n
mo
d
e
l
,
”
Ro
b
o
f
l
o
w
U
n
i
v
e
rs
e
,
2
0
2
2
.
A
c
c
e
s
sed
:
F
e
b
.
7
,
2
0
2
4
.
[
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/
n
e
w
-
w
o
r
k
sp
a
c
e
-
w
f
z
w
3
/
g
r
o
c
e
r
y
-
d
a
t
a
s
e
t
-
q
9
f
j
2
[
1
2
]
X.
-
S
.
W
e
i
,
Q
.
C
u
i
,
L
.
Y
a
n
g
,
P
.
W
a
n
g
,
a
n
d
L
.
L
i
u
,
“
R
P
C
:
a
l
a
r
g
e
-
s
c
a
l
e
r
e
t
a
i
l
p
r
o
d
u
c
t
c
h
e
c
k
o
u
t
d
a
t
a
s
e
t
,
”
a
rX
i
v
-
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
,
2
0
1
9
,
[
O
n
l
i
n
e
]
.
A
v
a
i
l
a
b
l
e
:
h
t
t
p
:
/
/
a
r
x
i
v
.
o
r
g
/
a
b
s/
1
9
0
1
.
0
7
2
4
9
[
1
3
]
P
.
F
o
l
l
m
a
n
n
,
T
.
B
ö
t
t
g
e
r
,
P
.
H
ä
r
t
i
n
g
e
r
,
R
.
K
ö
n
i
g
,
a
n
d
M
.
U
l
r
i
c
h
,
“
M
V
Te
c
D
2
S
:
d
e
n
s
e
l
y
s
e
g
men
t
e
d
su
p
e
r
ma
r
k
e
t
d
a
t
a
se
t
,
”
i
n
L
e
c
t
u
r
e
N
o
t
e
s
i
n
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
(
i
n
c
l
u
d
i
n
g
su
b
s
e
ri
e
s
L
e
c
t
u
re
N
o
t
e
s
i
n
A
rt
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
a
n
d
L
e
c
t
u
r
e
N
o
t
e
s
i
n
B
i
o
i
n
f
o
r
m
a
t
i
c
s
)
,
v
o
l
.
1
1
2
1
4
L
N
C
S
,
2
0
1
8
,
p
p
.
5
8
1
–
5
9
7
,
d
o
i
:
1
0
.
1
0
0
7
/
9
7
8
-
3
-
0
3
0
-
0
1
2
4
9
-
6
_
3
5
.
[
1
4
]
Y
.
Zh
a
o
e
t
a
l
.
,
“
D
ETR
s
b
e
a
t
Y
O
L
O
s
o
n
r
e
a
l
-
t
i
me
o
b
j
e
c
t
d
e
t
e
c
t
i
o
n
,
”
P
ro
c
e
e
d
i
n
g
s
o
f
t
h
e
I
EE
E
C
o
m
p
u
t
e
r
S
o
c
i
e
t
y
C
o
n
f
e
r
e
n
c
e
o
n
C
o
m
p
u
t
e
r
V
i
s
i
o
n
a
n
d
Pa
t
t
e
rn
Re
c
o
g
n
i
t
i
o
n
,
p
p
.
1
6
9
6
5
–
1
6
9
7
4
,
2
0
2
4
,
d
o
i
:
1
0
.
1
1
0
9
/
C
V
P
R
5
2
7
3
3
.
2
0
2
4
.
0
1
6
0
5
.
[
1
5
]
Q
.
Li
n
,
G
.
Y
e
,
J
.
W
a
n
g
,
a
n
d
H
.
L
i
u
,
“
R
o
b
o
F
l
o
w
:
a
d
a
t
a
-
c
e
n
t
r
i
c
w
o
r
k
f
l
o
w
m
a
n
a
g
e
men
t
sy
st
e
m
f
o
r
d
e
v
e
l
o
p
i
n
g
A
I
-
e
n
h
a
n
c
e
d
r
o
b
o
t
s,
”
Pr
o
c
e
e
d
i
n
g
s o
f
M
a
c
h
i
n
e
L
e
a
rn
i
n
g
Re
s
e
a
r
c
h
,
v
o
l
.
1
6
4
,
p
p
.
1
7
8
9
–
1
7
9
4
,
2
0
2
1
.
[
1
6
]
K
.
I
s
l
a
m,
“
R
e
c
e
n
t
a
d
v
a
n
c
e
s
i
n
v
i
si
o
n
t
r
a
n
s
f
o
r
m
e
r
:
a
su
r
v
e
y
a
n
d
o
u
t
l
o
o
k
o
f
r
e
c
e
n
t
w
o
r
k
,
”
a
r
Xi
v
-
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
,
2
0
2
3
.
[
O
n
l
i
n
e
]
.
A
v
a
i
l
a
b
l
e
:
h
t
t
p
s:
/
/
a
r
x
i
v
.
o
r
g
/
a
b
s/
2
2
0
3
.
0
1
5
3
6
[
1
7
]
J
.
Li
n
,
X
.
M
a
o
,
Y
.
C
h
e
n
,
L.
X
u
,
Y
.
H
e
,
a
n
d
H
.
X
u
e
,
“
D
^
2
E
TR
:
d
e
c
o
d
e
r
-
o
n
l
y
D
E
TR
w
i
t
h
c
o
mp
u
t
a
t
i
o
n
a
l
l
y
e
f
f
i
c
i
e
n
t
c
r
o
s
s
-
sc
a
l
e
a
t
t
e
n
t
i
o
n
,
”
a
r
Xi
v
-
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
,
2
0
2
2
.
[
O
n
l
i
n
e
]
.
A
v
a
i
l
a
b
l
e
:
h
t
t
p
s:
/
/
a
r
x
i
v
.
o
r
g
/
a
b
s/
2
2
0
3
.
0
0
8
6
0
[
1
8
]
C
.
W
a
n
g
,
X
.
X
i
n
g
,
Y
.
W
u
,
Z
.
S
u
,
a
n
d
J
.
C
h
e
n
,
“
D
C
S
F
N
:
d
e
e
p
c
r
o
s
s
-
s
c
a
l
e
f
u
s
i
o
n
n
e
t
w
o
r
k
f
o
r
si
n
g
l
e
i
mag
e
r
a
i
n
r
e
m
o
v
a
l
,
”
i
n
M
M
2
0
2
0
-
P
ro
c
e
e
d
i
n
g
s
o
f
t
h
e
2
8
t
h
AC
M
I
n
t
e
rn
a
t
i
o
n
a
l
C
o
n
f
e
ren
c
e
o
n
Mu
l
t
i
m
e
d
i
a
,
O
c
t
.
2
0
2
0
,
p
p
.
1
6
4
3
–
1
6
5
1
,
d
o
i
:
1
0
.
1
1
4
5
/
3
3
9
4
1
7
1
.
3
4
1
3
8
2
0
.
[
1
9
]
L
.
A
n
,
L.
W
a
n
g
,
a
n
d
Y
.
Li
,
“
H
E
A
-
N
e
t
:
a
t
t
e
n
t
i
o
n
a
n
d
M
L
P
h
y
b
r
i
d
e
n
c
o
d
e
r
a
r
c
h
i
t
e
c
t
u
r
e
f
o
r
m
e
d
i
c
a
l
i
m
a
g
e
s
e
g
men
t
a
t
i
o
n
,
”
S
e
n
so
r
s
,
v
o
l
.
2
2
,
n
o
.
1
8
,
S
e
p
.
2
0
2
2
,
d
o
i
:
1
0
.
3
3
9
0
/
s2
2
1
8
7
0
2
4
.
[
2
0
]
L
.
A
.
Li
m
a
n
d
H
.
Y
.
K
e
l
e
s,
“
L
e
a
r
n
i
n
g
mu
l
t
i
-
sc
a
l
e
f
e
a
t
u
r
e
s
f
o
r
f
o
r
e
g
r
o
u
n
d
s
e
g
m
e
n
t
a
t
i
o
n
,
”
Pa
t
t
e
r
n
An
a
l
y
si
s
a
n
d
Ap
p
l
i
c
a
t
i
o
n
s
,
v
o
l
.
2
3
,
n
o
.
3
,
p
p
.
1
3
6
9
–
1
3
8
0
,
2
0
2
0
,
d
o
i
:
1
0
.
1
0
0
7
/
s1
0
0
4
4
-
0
1
9
-
0
0
8
4
5
-
9.
[
2
1
]
S
.
Zh
e
n
g
e
t
a
l
.
,
“
R
e
t
h
i
n
k
i
n
g
se
man
t
i
c
s
e
g
men
t
a
t
i
o
n
f
r
o
m
a
s
e
q
u
e
n
c
e
-
to
-
se
q
u
e
n
c
e
p
e
r
sp
e
c
t
i
v
e
w
i
t
h
t
r
a
n
s
f
o
r
mer
s
,
”
P
ro
c
e
e
d
i
n
g
s
o
f
t
h
e
I
E
EE
C
o
m
p
u
t
e
r
S
o
c
i
e
t
y
C
o
n
f
e
r
e
n
c
e
o
n
C
o
m
p
u
t
e
r
Vi
s
i
o
n
a
n
d
Pa
t
t
e
rn
Re
c
o
g
n
i
t
i
o
n
,
p
p
.
6
8
7
7
–
6
8
8
6
,
2
0
2
1
,
d
o
i
:
1
0
.
1
1
0
9
/
C
V
P
R
4
6
4
3
7
.
2
0
2
1
.
0
0
6
8
1
.
[
2
2
]
H
.
R
e
z
a
t
o
f
i
g
h
i
,
N
.
T
so
i
,
J.
G
w
a
k
,
A
.
S
a
d
e
g
h
i
a
n
,
I
.
R
e
i
d
,
a
n
d
S
.
S
a
v
a
r
e
s
e
,
“
G
e
n
e
r
a
l
i
z
e
d
i
n
t
e
r
s
e
c
t
i
o
n
o
v
e
r
u
n
i
o
n
:
a
me
t
r
i
c
a
n
d
a
l
o
s
s
f
o
r
b
o
u
n
d
i
n
g
b
o
x
r
e
g
r
e
ss
i
o
n
,
”
P
r
o
c
e
e
d
i
n
g
s
o
f
t
h
e
I
E
EE
C
o
m
p
u
t
e
r
S
o
c
i
e
t
y
C
o
n
f
e
r
e
n
c
e
o
n
C
o
m
p
u
t
e
r
Vi
s
i
o
n
a
n
d
Pa
t
t
e
rn
R
e
c
o
g
n
i
t
i
o
n
,
p
p
.
6
5
8
–
6
6
6
,
2
0
1
9
,
d
o
i
:
1
0
.
1
1
0
9
/
C
V
P
R
.
2
0
1
9
.
0
0
0
7
5
.
[
2
3
]
Z
.
Y
a
o
,
J.
A
i
,
B
.
Li
,
a
n
d
C
.
Zh
a
n
g
,
“
E
f
f
i
c
i
e
n
t
D
E
TR
:
i
mp
r
o
v
i
n
g
e
n
d
-
to
-
e
n
d
o
b
j
e
c
t
d
e
t
e
c
t
o
r
w
i
t
h
d
e
n
s
e
p
r
i
o
r
,
”
a
rX
i
v
-
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
,
2
0
2
1
.
h
t
t
p
s:
/
/
a
r
x
i
v
.
o
r
g
/
a
b
s/
2
1
0
4
.
0
1
3
1
8
.
[
2
4
]
Y
.
F
e
n
g
,
H
.
X
u
,
J.
Ji
a
n
g
,
H
.
L
i
u
,
a
n
d
J.
Zh
e
n
g
,
“
I
C
I
F
-
N
e
t
:
i
n
t
r
a
-
sc
a
l
e
c
r
o
s
s
-
i
n
t
e
r
a
c
t
i
o
n
a
n
d
i
n
t
e
r
-
sca
l
e
f
e
a
t
u
r
e
f
u
s
i
o
n
n
e
t
w
o
r
k
f
o
r
b
i
t
e
mp
o
r
a
l
r
e
m
o
t
e
se
n
s
i
n
g
i
m
a
g
e
s
c
h
a
n
g
e
d
e
t
e
c
t
i
o
n
,
”
I
EE
E
T
r
a
n
sa
c
t
i
o
n
s
o
n
G
e
o
sci
e
n
c
e
a
n
d
Re
m
o
t
e
S
e
n
si
n
g
,
v
o
l
.
6
0
,
p
p
.
1
–
1
3
,
2
0
2
2
,
d
o
i
:
1
0
.
1
1
0
9
/
T
G
R
S
.
2
0
2
2
.
3
1
6
8
3
3
1
.
[
2
5
]
W
.
H
u
a
n
g
,
G
.
L
i
,
Q
.
C
h
e
n
,
M
.
J
u
,
a
n
d
J
.
Q
u
,
“
C
F
2
P
N
:
a
c
r
o
s
s
-
sc
a
l
e
f
e
a
t
u
r
e
f
u
si
o
n
p
y
r
a
mi
d
n
e
t
w
o
r
k
b
a
sed
r
e
mo
t
e
sen
s
i
n
g
t
a
r
g
e
t
d
e
t
e
c
t
i
o
n
,
”
R
e
m
o
t
e
S
e
n
si
n
g
,
v
o
l
.
1
3
,
n
o
.
5
,
p
p
.
1
–
2
3
,
F
e
b
.
2
0
2
1
,
d
o
i
:
1
0
.
3
3
9
0
/
r
s1
3
0
5
0
8
4
7
.
[
2
6
]
Y
.
H
u
a
n
g
,
Q
.
W
u
,
C
.
S
o
n
g
,
a
n
d
L
.
W
a
n
g
,
“
L
e
a
r
n
i
n
g
se
man
t
i
c
c
o
n
c
e
p
t
s
a
n
d
o
r
d
e
r
f
o
r
i
m
a
g
e
a
n
d
s
e
n
t
e
n
c
e
ma
t
c
h
i
n
g
,
”
P
r
o
c
e
e
d
i
n
g
s
o
f
t
h
e
I
EE
E
C
o
m
p
u
t
e
r
S
o
c
i
e
t
y
C
o
n
f
e
ren
c
e
o
n
C
o
m
p
u
t
e
r
V
i
s
i
o
n
a
n
d
Pa
t
t
e
rn
R
e
c
o
g
n
i
t
i
o
n
,
p
p
.
6
1
6
3
–
6
1
7
1
,
2
0
1
8
,
d
o
i
:
1
0
.
1
1
0
9
/
C
V
P
R
.
2
0
1
8
.
0
0
6
4
5
.
Evaluation Warning : The document was created with Spire.PDF for Python.