I
AE
S In
t
er
na
t
io
na
l J
o
urna
l o
f
Art
if
icia
l In
t
ellig
ence
(
I
J
-
AI
)
Vo
l.
1
5
,
No
.
1
,
Feb
r
u
ar
y
2
0
2
6
,
p
p
.
744
~
755
I
SS
N:
2
2
5
2
-
8
9
3
8
,
DOI
: 1
0
.
1
1
5
9
1
/ijai.v
15
.i
1
.
p
p
7
4
4
-
7
5
5
744
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
a
i
.
ia
esco
r
e.
co
m
YO
L
O
v
5
:
a
n imp
ro
v
ed alg
o
rithm f
o
r re
a
l
-
time
det
ec
tion o
f
indus
trial de
fec
ti
v
e pieces
Abdela
ziz
E
l
ba
g
hd
a
di
1
,
Ya
s
s
ine Y
a
zid
1,
2
,
Ahm
ed
E
l O
ua
l
k
a
di
1
,
Ant
o
nio
G
uerr
er
o
-
G
o
nzá
lez
2
,
So
ufia
ne
M
ez
ro
ui
3
1
La
b
o
r
a
t
o
r
y
o
f
I
n
n
o
v
a
t
i
v
e
S
y
st
e
ms
En
g
i
n
e
e
r
i
n
g
,
N
a
t
i
o
n
a
l
S
c
h
o
o
l
o
f
A
p
p
l
i
e
d
S
c
i
e
n
c
e
s
o
f
T
e
t
u
a
n
,
A
b
d
e
l
ma
l
e
k
Essa
a
d
i
U
n
i
v
e
r
si
t
y
,
Té
t
o
u
a
n
,
M
o
r
o
c
c
o
2
D
e
p
a
r
t
me
n
t
o
f
A
u
t
o
m
a
t
i
o
n
,
El
e
c
t
r
i
c
a
l
E
n
g
i
n
e
e
r
i
n
g
a
n
d
El
e
c
t
r
o
n
i
c
Te
c
h
n
o
l
o
g
y
,
U
n
i
v
e
r
si
d
a
d
P
o
l
i
t
é
c
n
i
c
a
d
e
C
a
r
t
a
g
e
n
a
,
C
a
r
t
a
g
e
n
a
,
S
p
a
i
n
3
La
b
o
r
a
t
o
r
y
o
f
M
a
t
h
e
ma
t
i
c
s a
n
d
I
n
t
e
l
l
i
g
e
n
t
S
y
st
e
ms,
N
a
t
i
o
n
a
l
S
c
h
o
o
l
o
f
A
p
p
l
i
e
d
S
c
i
e
n
c
e
s
o
f
T
a
n
g
i
e
r
,
A
b
d
e
l
m
a
l
e
k
Ess
a
a
d
i
U
n
i
v
e
r
s
i
t
y
,
Ta
n
g
i
e
r
,
M
o
r
o
c
c
o
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Ap
r
2
2
,
2
0
2
5
R
ev
is
ed
No
v
2
4
,
2
0
2
5
Acc
ep
ted
Dec
1
5
,
2
0
2
5
Th
e
ra
p
i
d
a
d
v
a
n
c
e
m
e
n
t
o
f
c
o
m
m
u
n
ica
ti
o
n
tec
h
n
o
l
o
g
ies
a
n
d
th
e
g
ro
wi
n
g
d
e
m
a
n
d
f
o
r
a
rti
f
icia
l
i
n
telli
g
e
n
c
e
a
re
tran
sfo
rm
in
g
tra
d
it
i
o
n
a
l
m
a
n
u
fa
c
tu
rin
g
in
to
sm
a
rt
in
d
u
stries
.
R
o
b
o
ti
c
a
rm
s
a
n
d
sm
a
rt
v
isio
n
c
a
m
e
ra
s
a
re
wid
e
ly
a
d
o
p
te
d
t
o
su
p
p
o
rt
i
n
d
u
strial
i
n
te
rn
e
t
o
f
th
i
n
g
s
(IIo
T)
a
p
p
l
ica
ti
o
n
s
.
Be
y
o
n
d
e
n
h
a
n
c
in
g
p
r
o
d
u
c
ti
o
n
e
fficie
n
c
y
a
n
d
q
u
a
l
it
y
,
t
h
e
se
tec
h
n
o
l
o
g
ie
s
p
lay
a
c
ru
c
ial
ro
le
in
c
o
st
re
d
u
c
ti
o
n
,
e
n
e
rg
y
sa
v
i
n
g
s,
a
n
d
imp
ro
v
in
g
o
p
e
ra
to
r
sa
fe
ty
.
In
th
is
a
rti
c
le,
we
p
r
o
p
o
se
a
n
i
n
te
ll
ig
e
n
t
in
d
u
strial
sy
ste
m
u
si
n
g
a
n
imp
ro
v
e
d
v
e
rsio
n
o
f
t
h
e
y
o
u
o
n
l
y
lo
o
k
o
n
c
e
(YO
LO)
a
lg
o
rit
h
m
fo
r
d
e
fe
c
t
d
e
tec
ti
o
n
o
n
p
ro
d
u
c
ti
o
n
l
in
e
s
.
T
h
e
sy
ste
m
in
teg
ra
tes
ro
b
o
ts
a
n
d
c
a
m
e
ra
s
to
a
u
to
m
a
te
d
e
fe
c
t
in
sp
e
c
ti
o
n
a
n
d
c
las
sifica
t
io
n
o
f
m
a
n
u
fa
c
tu
re
d
p
iec
e
s.
An
u
p
d
a
ted
YO
LOv
5
m
o
d
e
l
is
d
e
si
g
n
e
d
a
s
a
n
e
n
d
-
to
-
e
n
d
so
lu
ti
o
n
fo
r
d
e
tec
ti
n
g
su
rfa
c
e
d
e
fe
c
ts
in
th
re
e
sp
e
c
ifi
c
re
g
io
n
s.
We
train
e
d
a
n
d
e
v
a
lu
a
ted
th
e
m
o
d
e
l
u
sin
g
c
u
sto
m
d
a
ta
tailo
re
d
to
th
e
in
s
p
e
c
ted
p
iec
e
s.
T
h
e
sy
ste
m
a
c
h
ie
v
e
d
a
9
9
%
m
e
a
n
a
v
e
ra
g
e
p
re
c
isi
o
n
(m
AP)
a
n
d
a
n
8
0
%
re
c
a
ll
ra
te.
Ad
d
it
io
n
a
ll
y
,
it
d
e
li
v
e
rs
a
9
9
%
d
e
tec
ti
o
n
ra
te
a
t
h
ig
h
sp
e
e
d
,
e
n
a
b
li
n
g
re
a
l
-
ti
m
e
su
rf
a
c
e
d
e
fe
c
t
d
e
tec
ti
o
n
.
Th
is
m
e
th
o
d
n
o
t
o
n
ly
a
c
c
u
ra
tely
p
re
d
icts
d
e
fe
c
ti
v
e
lo
c
a
ti
o
n
s
b
u
t
a
lso
p
ro
v
id
e
s
siz
e
in
fo
rm
a
ti
o
n
,
wh
ich
is
c
rit
ica
l
fo
r
a
ss
e
ss
in
g
th
e
q
u
a
li
ty
o
f
n
e
wly
p
r
o
d
u
c
e
d
p
iec
e
s.
K
ey
w
o
r
d
s
:
Ar
tific
ial
in
tellig
en
ce
Def
ec
t d
etec
tio
n
I
n
d
u
s
tr
ial
in
ter
n
et
o
f
th
in
g
s
Sm
ar
t v
is
io
n
YOL
Ov
5
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
:
Ab
d
elaz
iz
E
lb
ag
h
d
ad
i
L
ab
o
r
ato
r
y
o
f
I
n
n
o
v
ativ
e
Sy
s
t
em
s
E
n
g
in
ee
r
in
g
,
Natio
n
al
Sc
h
o
o
l o
f
Ap
p
lied
Scien
ce
s
o
f
T
etu
an
Ab
d
elm
alek
E
s
s
aa
d
i U
n
iv
er
s
ity
B
P 2
2
2
2
,
M’
h
a
n
n
ec
h
I
I
T
éto
u
an
,
Mo
r
o
cc
o
E
m
ail:
ab
d
elaz
iz.
elb
ag
h
d
ad
i@
etu
.
u
ae
.
ac
.
m
a
1.
I
NT
RO
D
UCT
I
O
N
I
n
d
u
s
tr
ial
d
ef
ec
t
d
etec
tio
n
h
as
b
ee
n
s
tu
d
ied
f
o
r
d
ec
ad
es
,
p
ar
ticu
lar
l
y
in
tr
ad
itio
n
al
m
ac
h
in
in
g
p
r
o
ce
s
s
es.
Mo
s
t
o
f
th
e
p
r
e
v
io
u
s
r
esear
ch
h
as
co
n
ce
n
tr
ated
o
n
m
eth
o
d
s
f
o
r
d
etec
tin
g
d
ef
ec
t
s
in
p
o
s
t
-
p
r
o
ce
s
s
ed
p
ar
ts
.
I
n
th
e
co
m
p
o
s
ite’
s
co
m
m
u
n
ity
,
f
o
r
ex
am
p
le,
r
esear
ch
er
s
h
av
e
u
s
ed
d
if
f
er
en
t
tr
ad
iti
o
n
al
m
eth
o
d
s
s
u
ch
as
u
ltra
s
o
n
ic
test
in
g
to
d
etec
t
d
am
ag
ed
s
u
r
f
ac
es
in
ca
r
b
o
n
f
ib
er
-
r
ein
f
o
r
ce
d
p
last
ics
[
1
]
,
o
s
m
o
s
is
tes
tin
g
[
2
]
,
X
-
r
ay
test
in
g
[
3
]
,
a
n
d
tr
a
d
itio
n
al
m
ac
h
in
e
v
is
io
n
d
etec
tio
n
[
4
]
.
T
h
er
ef
o
r
e,
s
ev
er
al
m
eth
o
d
s
h
av
e
b
ee
n
u
s
ed
i
n
in
d
u
s
tr
y
f
o
r
d
ef
ec
t
d
etec
tio
n
th
at
p
r
o
ce
e
d
in
two
s
tag
es:
f
ea
tu
r
e
ex
tr
ac
tio
n
an
d
d
e
f
ec
t
i
d
en
tific
atio
n
.
Fr
o
m
s
u
r
f
ac
e
d
ef
ec
t
d
etec
tio
n
,
f
ea
tu
r
e
ex
tr
ac
tio
n
,
i
d
en
tific
atio
n
,
a
n
d
p
er
s
p
ec
tiv
es,
th
e
y
ca
n
b
e
ca
teg
o
r
ized
m
ain
l
y
in
to
s
tatis
tical
m
eth
o
d
s
,
s
p
ec
t
r
al
m
eth
o
d
s
,
m
o
d
el
-
b
ased
m
e
th
o
d
s
,
a
n
d
lea
r
n
in
g
-
b
ased
m
et
h
o
d
s
.
Desp
ite
t
h
eir
r
o
b
u
s
tn
ess
an
d
ac
cu
r
ac
y
,
ea
s
e
o
f
u
s
e,
a
n
d
in
teg
r
atio
n
,
th
ese
m
eth
o
d
s
ca
n
n
o
t
b
e
im
p
lem
e
n
t
ed
,
f
o
r
ex
a
m
p
le
f
o
r
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
YOLOv5
:
a
n
imp
r
o
ve
d
a
lg
o
r
ith
m
fo
r
r
ea
l
-
time
d
etec
tio
n
o
f i
n
d
u
s
tr
ia
l d
efec
tive
… (
A
b
d
ela
z
iz
E
lb
a
g
h
d
a
d
i
)
745
o
n
lin
e
task
s
o
r
m
ak
in
g
r
o
b
o
t
ic
au
to
m
atio
n
an
o
p
e
n
-
lo
o
p
p
r
o
ce
s
s
.
B
esid
es,
lear
n
in
g
-
b
ased
m
eth
o
d
s
h
av
e
s
h
o
wn
m
u
c
h
i
n
ter
est
r
ec
en
tl
y
as
th
e
y
h
a
v
e
s
h
o
wn
i
n
ter
e
s
tin
g
p
er
f
o
r
m
an
ce
f
o
r
b
o
th
s
im
u
latio
n
s
an
d
r
ea
l
ap
p
licatio
n
s
.
Gen
etic
alg
o
r
ith
m
s
in
s
u
p
p
o
r
t
v
ec
t
o
r
m
ac
h
in
e
(
SVM)
[
5
]
,
ar
tific
ial
n
eu
r
al
n
etwo
r
k
(
ANN)
[
6
]
,
k
-
n
ea
r
est
n
eig
h
b
o
r
(K
-
NN)
[
7
]
,
r
an
d
o
m
f
o
r
est
[
8
]
,
g
en
etic
alg
o
r
ith
m
s
[
9
]
,
a
n
d
clu
s
ter
in
g
m
eth
o
d
s
[
1
0
]
,
ar
e
ap
p
lied
f
r
e
q
u
en
tly
f
o
r
d
ef
ec
ti
o
n
d
etec
tio
n
a
n
d
class
if
icatio
n
f
o
r
m
o
d
er
n
in
d
u
s
tr
ial
ap
p
lic
atio
n
s
.
Fu
r
th
er
m
o
r
e,
u
s
in
g
o
n
ly
th
e
n
ec
ess
ar
y
n
u
m
b
er
o
f
tr
ain
in
g
im
ag
es,
d
ata
-
d
r
iv
en
,
an
d
m
ac
h
in
e
-
lea
r
n
in
g
ap
p
r
o
ac
h
es
ca
n
b
e
q
u
ick
ly
a
d
ap
ted
to
n
ew
ty
p
es
o
f
p
r
o
d
u
cts an
d
s
u
r
f
ac
e
d
ef
ec
t
s
.
No
wad
ay
s
,
in
d
u
s
tr
y
is
ex
p
er
i
en
cin
g
a
m
ass
iv
e
r
ev
o
lu
tio
n
d
u
e
to
th
e
im
p
r
ess
iv
e
ev
o
l
u
tio
n
an
d
th
e
ad
o
p
tio
n
o
f
n
ew
tech
n
o
lo
g
ies
an
d
m
ater
ials
o
n
t
h
e
o
n
e
h
an
d
an
d
th
e
f
ast
m
ar
k
et
r
eq
u
ir
e
m
en
ts
o
n
th
e
o
th
er
h
an
d
.
All
th
is
is
to
im
p
r
o
v
e
t
h
e
m
an
u
f
ac
tu
r
er
s
’
b
u
s
in
ess
an
d
r
esp
o
n
d
to
c
u
s
to
m
s
n
ee
d
s
.
Ma
n
y
in
d
u
s
tr
ies
ar
e
n
o
w
u
s
in
g
d
iv
e
r
s
e
h
ig
h
-
lev
el
d
ev
ices
th
at
ar
e
co
n
n
ec
ted
t
o
th
e
in
ter
n
et
lik
e
s
en
s
o
r
s
,
ca
m
e
r
as,
an
d
r
o
b
o
ts
th
at
g
en
er
ate
h
u
g
e
am
o
u
n
ts
o
f
d
at
a
wh
ich
ca
n
b
e
ex
p
lo
ited
to
m
ak
e
th
e
in
d
u
s
tr
ial
ap
p
licatio
n
m
o
r
e
in
tellig
en
t
an
d
r
ed
u
ce
h
u
m
a
n
in
ter
v
en
tio
n
s
.
T
h
e
in
d
u
s
tr
ial
in
ter
n
et
o
f
t
h
in
g
s
(
I
I
o
T
)
is
m
ak
in
g
s
ig
n
if
i
ca
n
t
ch
an
g
es
as
it
p
r
o
v
id
es
s
m
ar
t
s
o
l
u
tio
n
s
ex
p
l
o
itin
g
th
e
i
n
ter
ac
tiv
e
d
ata
b
et
wee
n
m
ac
h
in
es,
s
en
s
o
r
s
,
a
n
d
ac
tu
atio
n
r
ely
in
g
o
n
em
er
g
in
g
m
ac
h
in
e
lear
n
in
g
m
eth
o
d
s
[
1
1
]
.
Gen
er
ally
,
d
ee
p
lear
n
in
g
m
eth
o
d
s
ca
n
ac
h
iev
e
ex
ce
llen
t
r
esu
lts
wh
en
ap
p
lied
to
th
e
p
r
o
b
le
m
o
f
s
u
r
f
ac
e
-
q
u
ality
co
n
tr
o
l.
W
h
en
co
m
p
ar
ed
to
tr
a
d
itio
n
al
m
ac
h
in
e
-
v
is
io
n
m
eth
o
d
s
,
d
ee
p
lear
n
in
g
ca
n
d
i
r
ec
tly
lear
n
f
ea
t
u
r
es
f
r
o
m
lo
w
-
lev
el
d
ata
an
d
h
as
a
g
r
ea
ter
c
ap
ac
ity
to
r
e
p
r
esen
t
co
m
p
lex
s
tr
u
ctu
r
es.
T
h
is
lead
s
to
r
ep
lacin
g
s
o
m
e
h
u
m
an
o
p
er
ato
r
s
’
m
o
n
ito
r
e
d
task
s
with
au
to
m
ated
lea
r
n
in
g
o
n
es.
Dee
p
lear
n
i
n
g
-
b
ased
m
e
th
o
d
s
h
av
e
b
ec
o
m
e
v
er
y
s
u
itab
le
f
o
r
n
ewly
em
er
g
ed
I
n
d
u
s
t
r
y
4
.
0
ap
p
licatio
n
s
.
Ma
n
y
r
esear
ch
er
s
h
a
v
e
ad
o
p
te
d
m
ac
h
in
e
lear
n
in
g
m
eth
o
d
s
to
r
em
ed
y
s
o
m
e
is
s
u
es in
d
if
f
er
en
t c
ase
s
ce
n
ar
io
s
.
R
ec
en
t
ad
v
an
ce
m
en
ts
in
d
ee
p
lear
n
in
g
h
av
e
led
to
s
ig
n
if
ic
an
t
im
p
r
o
v
em
e
n
ts
ac
r
o
s
s
v
ar
io
u
s
f
ield
s
.
T
h
e
"Du
alVitOA"
m
o
d
el
u
tili
ze
s
d
u
al
v
is
io
n
tr
a
n
s
f
o
r
m
er
s
to
ac
cu
r
ately
g
r
ad
e
k
n
ee
o
s
teo
ar
th
r
itis
f
r
o
m
X
-
r
a
y
im
ag
es,
ac
h
iev
in
g
7
8
.
4
%
ac
cu
r
ac
y
an
d
r
ed
u
cin
g
d
iag
n
o
s
tic
s
u
b
jectiv
ity
[
1
2
]
.
I
n
s
m
ar
t
cities,
y
o
u
o
n
ly
lo
o
k
o
n
ce
(
YOL
O)
h
as
b
ee
n
em
p
lo
y
ed
f
o
r
r
ea
l
-
tim
e
d
etec
tio
n
o
f
leth
al
wea
p
o
n
s
,
ac
h
iev
in
g
8
9
.
5
6
%
ac
cu
r
ac
y
[
1
3
]
.
Pre
cisi
o
n
ag
r
icu
ltu
r
e
b
e
n
ef
its
f
r
o
m
a
h
y
b
r
id
ap
p
r
o
ac
h
in
teg
r
atin
g
YOL
O
with
a
r
ice
f
ield
s
id
ewa
lk
d
etec
tio
n
alg
o
r
ith
m
,
en
s
u
r
in
g
h
ig
h
ac
c
u
r
ac
y
i
n
o
b
ject
d
etec
tio
n
an
d
d
is
tan
ce
m
ea
s
u
r
em
e
n
t
[
1
4
]
.
Ad
d
itio
n
ally
,
d
ee
p
lear
n
in
g
m
o
d
els
h
av
e
b
ee
n
i
n
teg
r
ated
f
o
r
v
id
e
o
e
n
h
an
ce
m
en
t,
co
m
p
r
ess
io
n
,
an
d
r
esto
r
atio
n
,
r
esu
ltin
g
in
n
o
tab
le
im
p
r
o
v
em
e
n
ts
in
v
id
e
o
clar
ity
an
d
b
itra
te
r
e
d
u
ctio
n
[
1
5
]
.
C
o
n
v
o
l
u
ti
o
n
a
l
n
eu
r
a
l
n
etw
o
r
k
s
(
C
NNs)
h
av
e
f
u
e
le
d
s
i
g
n
if
i
ca
n
t
a
d
v
a
n
c
es
i
n
t
h
is
f
iel
d
o
f
c
o
m
p
u
t
e
r
v
is
i
o
n
,
r
es
u
lti
n
g
i
n
s
i
g
n
i
f
ic
a
n
t
b
r
ea
k
t
h
r
o
u
g
h
s
i
n
im
a
g
e
cl
a
s
s
if
i
ca
t
io
n
[
1
6
]
a
n
d
o
b
j
ec
t
d
e
tect
io
n
[
1
7
]
.
Ma
n
y
r
es
ea
r
c
h
e
r
s
h
a
v
e
a
p
p
li
ed
C
NN
s
to
i
n
d
u
s
t
r
i
al
in
s
p
e
cti
o
n
s
y
s
t
e
m
s
t
o
i
n
cr
ea
s
e
t
h
e
ir
p
r
a
cti
ca
l
v
alu
e,
p
a
r
ti
cu
la
r
l
y
i
n
d
e
f
e
ct
d
e
tec
ti
o
n
.
T
h
e
y
h
a
v
e
b
e
g
u
n
t
o
i
n
c
o
r
p
o
r
at
e
i
n
d
u
s
tr
i
al
p
r
o
d
u
cti
o
n
w
it
h
o
b
jec
t
d
et
ec
ti
o
n
a
lg
o
r
it
h
m
s
t
o
d
i
r
e
ctl
y
o
b
t
ai
n
t
h
e
l
o
ca
ti
o
n
s
a
n
d
ty
p
es
o
f
d
e
f
ec
ts
.
T
o
s
eg
m
e
n
t
a
n
d
l
o
c
ali
ze
d
e
f
ec
ts
o
n
a
m
etal
lic
s
u
r
f
ac
e.
F
o
r
in
s
t
an
ce
,
T
ao
et
a
l.
[
1
8
]
p
r
o
p
o
s
ed
a
n
o
v
el
c
asc
ad
e
d
a
u
t
o
-
en
c
o
d
e
r
a
r
c
h
it
ec
t
u
r
e
.
T
h
is
m
et
h
o
d
,
h
o
we
v
e
r
,
ca
n
n
o
t
d
i
f
f
er
e
n
ti
ate
b
etw
ee
n
d
i
f
f
e
r
e
n
t
i
n
d
iv
id
u
als
o
f
t
h
e
i
d
e
n
ti
ca
l
ty
p
e
,
a
n
d
t
h
e
b
ac
k
g
r
o
u
n
d
o
f
t
h
e
o
b
j
ec
ts
b
ei
n
g
d
et
ec
te
d
.
F
aste
r
r
e
g
i
o
n
-
b
as
ed
co
n
v
o
lu
ti
o
n
al
n
e
u
r
al
n
et
wo
r
k
(
Fas
t
er
-
R
C
NN
)
[
1
9
]
h
as
b
ee
n
a
d
o
p
te
d
f
o
r
d
ef
ec
t
d
et
ec
ti
o
n
o
n
a
s
t
ee
l
s
u
r
f
ac
e
[
2
0
]
,
[
2
1
]
.
E
v
e
n
t
h
o
u
g
h
F
aste
r
-
R
C
NN
is
a
tw
o
-
s
ta
g
e
m
o
d
e
l,
it
is
c
o
n
s
t
r
ai
n
e
d
i
n
ter
m
s
o
f
t
h
e
p
r
o
c
ess
i
n
g
s
p
e
e
d
o
f
im
ag
es
w
h
ic
h
s
ev
er
el
y
li
m
it
s
its
a
p
p
lic
ati
o
n
i
n
r
ea
l
-
t
im
e
i
n
d
u
s
tr
ial
in
s
p
e
cti
o
n
.
Pia
o
et
a
l
.
[
2
2
]
p
r
o
p
o
s
e
d
a
d
ec
is
io
n
t
r
ee
e
n
s
em
b
l
e
l
ea
r
n
i
n
g
-
b
ase
d
m
et
h
o
d
f
o
r
d
et
ec
t
in
g
wa
f
er
m
ap
f
ai
lu
r
es
.
Sin
ce
it
h
as
b
ee
n
in
tr
o
d
u
ce
d
,
t
h
e
YO
L
O
al
g
o
r
it
h
m
wi
th
it
s
a
m
el
io
r
a
te
d
v
er
s
io
n
s
h
as
g
a
in
e
d
h
i
g
h
in
t
er
est
as
it
s
a
v
es
ti
m
e
at
t
h
e
e
x
p
e
n
s
e
o
f
a
s
li
g
h
t
lo
s
s
o
f
a
c
cu
r
a
cy
w
h
e
n
c
o
m
p
ar
ed
t
o
t
h
e
tw
o
-
s
ta
g
e
m
et
h
o
d
.
Fo
r
e
x
am
p
l
e,
t
h
e
YO
L
Ov
3
[
2
3
]
h
as
b
ee
n
u
s
ed
t
o
d
ete
ct
d
ef
ec
ts
i
n
r
a
il s
u
r
f
ac
es
an
d
ac
h
i
e
v
e
d
a
9
7
%
r
e
c
o
g
n
i
ti
o
n
r
at
e.
L
i
et
a
l.
[
2
4
]
h
a
v
e
p
r
es
e
n
t
ed
an
a
u
t
o
m
ati
c
d
ef
ec
t
d
ete
cti
o
n
s
o
l
u
ti
o
n
b
as
e
d
o
n
Y
OL
O
v
4
,
t
h
at
ac
h
i
ev
e
d
b
o
t
h
f
ast
a
n
d
a
cc
u
r
at
e
d
e
f
e
ct
d
e
te
ct
io
n
f
o
r
wi
r
e
an
d
a
r
c
a
d
d
iti
v
e
m
a
n
u
f
ac
t
u
r
i
n
g
(
W
A
AM
)
.
I
n
a
d
d
iti
o
n
,
a
YOL
Ov
5
was
u
s
e
d
o
n
a
r
m
r
o
b
o
t
t
o
ef
f
ec
tiv
el
y
d
e
te
ct
d
ef
ec
t
d
et
ec
ti
o
n
o
n
c
u
s
to
m
i
n
d
u
s
t
r
i
al
a
p
p
lic
ati
o
n
[
2
5
]
.
T
h
e
m
o
d
el
YOL
O
h
as
s
h
o
w
n
a
cc
u
r
at
e
e
n
h
a
n
c
em
en
t
o
n
t
h
r
ee
ex
is
ti
n
g
o
b
jec
t
d
et
ec
ti
o
n
m
o
d
e
ls
:
ch
an
n
el
-
wis
e
att
e
n
ti
o
n
,
m
u
lti
p
le
s
p
ati
al
p
y
r
am
id
p
o
o
li
n
g
,
an
d
e
x
p
o
n
en
tia
l
m
o
v
i
n
g
a
v
e
r
a
g
e
[
2
6
]
.
I
n
a
d
d
it
io
n
t
o
its
ea
s
e
o
f
a
d
o
p
ti
o
n
a
n
d
s
i
m
p
li
cit
y
,
Y
OL
O
v
5
p
r
o
v
i
d
es
h
i
g
h
q
u
al
it
y
i
n
te
r
m
s
o
f
p
r
e
d
i
cti
o
n
a
n
d
p
r
o
c
ess
i
n
g
s
p
e
e
d
wh
i
ch
is
h
ig
h
l
y
r
e
q
u
i
r
e
d
i
n
s
e
v
e
r
a
l
e
m
e
r
g
in
g
i
n
d
u
s
t
r
ia
l
a
p
p
li
ca
ti
o
n
s
c
o
m
p
ar
e
d
t
o
s
ta
te
-
of
-
t
h
e
-
a
r
t
m
et
h
o
d
s
.
F
o
r
th
es
e
tw
o
r
ea
s
o
n
s
,
w
e
h
av
e
s
el
ec
t
ed
t
o
u
s
e
YO
L
O
v
5
in
c
o
n
c
r
ete
i
n
d
u
s
tr
ial
t
ask
s
t
o
i
m
p
r
o
v
e
th
e
p
r
o
d
u
c
ti
o
n
r
at
e
an
d
r
ed
u
ce
h
u
m
a
n
in
te
r
v
e
n
ti
o
n
s
i
n
t
h
e
p
r
o
ce
s
s
o
f
s
p
ec
if
ic
el
ec
tr
o
n
ic
ci
r
c
u
i
t
ca
r
d
s
m
a
n
u
f
ac
t
u
r
in
g
.
Mo
tiv
ated
b
y
th
e
im
p
r
ess
iv
e
r
ep
u
tatio
n
o
f
d
ee
p
lear
n
in
g
im
p
lem
en
tatio
n
in
d
iv
e
r
s
e
d
o
m
ain
s
an
d
th
eir
h
ig
h
p
e
r
f
o
r
m
an
ce
in
d
iv
er
s
e
f
ield
s
lik
e
in
d
u
s
tr
ial
s
u
p
er
v
is
io
n
,
d
r
o
n
es,
r
o
b
o
tics
[
2
7
]
.
B
esid
es
th
e
o
u
tp
er
f
o
r
m
an
ce
o
f
d
ata
-
d
r
iv
e
n
-
b
ased
a
p
p
licatio
n
s
o
n
class
ic
m
ac
h
in
e
v
is
io
n
m
et
h
o
d
s
in
ter
m
s
o
f
ac
c
u
r
ac
y
,
r
eliab
ilit
y
,
f
lex
ib
ilit
y
,
an
d
laten
cy
.
I
n
t
h
is
p
ap
er
,
we
p
r
o
v
id
e
an
im
p
r
o
v
ed
v
er
s
io
n
o
f
a
m
ac
h
in
e
lear
n
in
g
m
eth
o
d
f
o
r
d
etec
tin
g
v
is
u
al
s
u
r
f
ac
e
d
ef
ec
ts
.
I
t
e
m
p
h
asizes
th
e
u
s
e
o
f
th
e
YOL
O
alg
o
r
ith
m
,
wh
ich
h
as
g
ain
e
d
a
lo
t
o
f
in
te
r
est
in
co
m
p
u
ter
v
is
io
n
in
r
ec
e
n
t
y
ea
r
s
.
Mo
r
e
p
r
ec
is
ely
,
we
p
r
o
v
id
e
a
n
i
n
tellig
en
t
in
d
u
s
tr
ial
ap
p
licatio
n
m
eth
o
d
th
at
d
etec
ts
an
o
m
alies
f
r
o
m
p
r
o
d
u
ce
d
elec
tr
o
n
ic
p
iece
s
f
o
r
th
e
p
u
r
p
o
s
e
o
f
q
u
alit
y
au
g
m
en
tatio
n
b
e
f
o
r
e
th
e
m
ar
k
et.
W
e
p
r
o
v
id
e
a
co
n
cr
ete
s
m
a
r
t
in
d
u
s
tr
ial
a
p
p
licatio
n
th
at
e
n
ab
les
b
o
th
a
s
m
ar
t
v
is
io
n
s
y
s
tem
f
o
r
q
u
ality
class
if
icatio
n
an
d
an
ar
m
r
o
b
o
t
f
o
r
th
e
p
r
o
ce
s
s
o
f
r
e
d
u
cin
g
h
u
m
an
in
ter
v
e
n
tio
n
,
f
o
r
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
1
5
,
No
.
1
,
Feb
r
u
ar
y
2
0
2
6
:
744
-
7
5
5
746
ex
am
p
le
in
in
d
u
s
tr
ial
en
v
ir
o
n
m
en
ts
.
T
h
e
co
m
p
u
ter
v
is
io
n
s
y
s
tem
r
elies
o
n
th
e
in
tellig
en
ce
o
f
th
e
YOL
O
alg
o
r
ith
m
to
im
p
r
o
v
e
th
e
in
s
p
ec
tio
n
r
ate
as
it
p
r
o
v
id
es
h
ig
h
p
er
f
o
r
m
an
ce
a
n
d
ac
c
u
r
ate
l
aten
cy
.
T
h
e
s
y
s
tem
ch
ec
k
s
th
e
q
u
ality
o
f
n
ewly
m
an
u
f
ac
tu
r
e
d
p
iece
s
in
tellig
en
tl
y
in
th
e
a
b
s
en
ce
o
f
h
u
m
a
n
s
.
T
h
e
co
llab
o
r
atio
n
b
etwe
en
th
e
ca
m
er
a
an
d
ar
m
r
o
b
o
t
s
er
v
es
to
m
o
n
ito
r
th
e
s
u
r
f
ac
e
o
f
th
e
co
n
ce
r
ted
p
iece
s
.
T
h
e
p
iece
m
a
y
co
n
tain
an
o
m
alies
in
th
r
ee
d
is
tin
ct
ar
ea
s
,
f
o
r
t
h
is
aim
,
we
h
av
e
co
llected
d
ata
o
n
d
if
f
er
en
t
p
o
s
s
ib
le
d
ef
ec
t
ca
s
es
to
tr
ain
o
u
r
m
o
d
el.
T
h
e
d
ef
ec
t
m
ay
b
e
p
r
esen
ted
in
o
n
e
ar
e
a,
two
,
o
r
th
r
ee
.
T
o
g
ain
tim
e
if
th
e
s
y
s
tem
d
etec
ts
o
n
ly
o
n
e
ar
ea
as
d
ef
ec
tiv
e
th
e
s
y
s
tem
d
ec
lar
es
th
e
p
iec
e
as
n
o
n
-
v
alid
an
d
s
h
o
u
ld
b
e
r
e
m
o
v
e
d
f
r
o
m
th
e
p
r
o
d
u
ctio
n
lin
e.
T
h
e
r
est
o
f
th
is
p
ap
e
r
is
s
tr
u
ctu
r
ed
as
f
o
llo
ws.
Sectio
n
2
p
r
o
v
id
es
th
e
p
r
o
p
o
s
ed
s
y
s
tem
m
o
d
el.
Sectio
n
s
3
p
r
esen
t
r
esp
ec
tiv
ely
th
e
f
o
u
n
d
r
esu
l
ts
an
d
d
is
cu
s
s
io
n
s
.
Fin
ally
,
s
ec
tio
n
4
p
r
o
v
id
es a
c
o
n
clu
s
io
n
.
2.
SYST
E
M
M
O
D
E
L
Vis
u
al
in
s
p
ec
tio
n
is
o
n
e
o
f
t
h
e
im
p
o
r
tan
t
p
illar
s
to
im
p
r
o
v
e
th
e
s
p
ee
d
,
q
u
ality
,
an
d
f
le
x
ib
ilit
y
i
n
I
n
d
u
s
tr
ial
4
.
0
,
esp
ec
ially
f
o
r
d
ef
ec
t
r
ec
o
g
n
itio
n
o
n
p
r
o
d
u
c
tio
n
lin
es.
I
n
th
e
p
r
esen
t
wo
r
k
,
we
p
r
o
p
o
s
e
a
n
am
elio
r
ated
s
y
s
tem
th
at
aim
s
to
im
p
r
o
v
e
t
h
e
q
u
ality
an
d
au
to
m
ate
v
is
u
al
in
s
p
ec
tio
n
in
a
s
p
ec
if
ic
in
d
u
s
tr
ial
ap
p
licatio
n
.
T
h
e
p
r
o
p
o
s
ed
s
y
s
tem
in
clu
d
es
a
c
o
llab
o
r
ativ
e
r
o
b
o
t,
ca
m
er
a,
a
n
d
d
ata
p
r
o
ce
s
s
in
g
u
n
it
th
at
r
elies
o
n
th
e
ef
f
ec
tiv
e
n
ess
o
f
th
e
ar
t
if
icial
in
tellig
en
ce
m
eth
o
d
to
m
ak
e
r
eliab
le,
ef
f
icien
t,
an
d
a
u
to
m
atic
in
s
p
ec
tio
n
d
ec
is
io
n
s
.
An
am
elio
r
ated
v
er
s
io
n
o
f
th
e
YOL
O
alg
o
r
ith
m
is
u
s
ed
to
cla
s
s
if
y
n
ewly
p
r
o
d
u
ce
d
elec
tr
o
n
i
c
p
iece
s
in
to
d
ef
ec
tiv
e
(
non
-
OK
)
o
r
v
alid
(
OK
)
o
n
es.
T
h
e
m
o
n
ito
r
ed
p
iece
s
ca
n
ca
r
r
y
d
ef
ec
ts
o
n
th
e
s
u
r
f
ac
e
in
th
r
ee
d
is
tin
ct
ar
ea
s
.
T
h
e
s
y
s
tem
class
if
ies
th
e
p
iece
s
as
d
ef
ec
tiv
e
if
at
least
o
n
e
ar
ea
is
in
v
alid
.
B
esid
es,
an
ar
m
r
o
b
o
t
is
lin
k
ed
to
th
e
v
is
io
n
in
s
p
ec
tio
n
s
y
s
tem
an
d
n
etwo
r
k
co
m
m
u
n
icatio
n
co
lla
b
o
r
ates
in
tellig
en
tly
to
d
ec
id
e
th
e
q
u
ality
o
f
th
e
p
r
o
d
u
ce
d
p
iece
(
i.e
.
,
Par
t)
.
Fig
u
r
e
1
s
h
o
ws
th
e
p
r
o
v
i
d
ed
o
v
er
all
s
y
s
tem
f
u
n
ctio
n
in
g
cy
cle.
Firstl
y
,
th
e
n
ewly
p
r
o
d
u
ce
d
p
iece
s
m
en
tio
n
ed
as Par
t m
o
v
e
o
n
th
e
co
n
v
e
y
o
r
with
a
s
p
ec
if
ic
d
is
tan
ce
s
ep
ar
atio
n
.
T
h
en
a
s
en
s
o
r
d
etec
ts
th
e
p
r
esen
ce
o
f
th
e
p
iece
u
n
d
er
th
e
r
o
b
o
t.
Af
ter
th
at
an
o
r
d
e
r
is
s
en
t,
an
d
th
e
r
o
b
o
t r
etr
iev
es
th
e
p
a
r
t
an
d
m
o
v
es
to
th
e
ca
m
e
r
a
to
tak
e
p
ict
u
r
es
o
f
ea
ch
s
id
e
o
f
t
h
e
p
ar
t.
Fin
ally
,
th
e
ar
tific
ial
in
tellig
en
ce
tech
n
iq
u
e
in
ter
v
e
n
es
to
m
ak
e
an
a
d
eq
u
ate
d
ec
is
io
n
ab
o
u
t
th
e
s
tate
o
f
th
e
p
ick
ed
p
iece
.
I
f
th
e
p
iece
is
class
if
ied
as
d
ef
ec
tiv
e
,
it
is
r
em
o
v
ed
f
r
o
m
th
e
c
o
n
v
e
y
o
r
.
T
ab
le
1
p
r
esen
ts
th
e
u
s
ed
h
ar
d
war
e
elem
e
n
ts
o
f
th
e
p
r
o
p
o
s
ed
s
y
s
tem
.
Fig
u
r
e
1
.
T
h
e
p
r
o
p
o
s
ed
s
y
s
te
m
f
lo
wch
ar
t
T
ab
le
1
.
T
h
e
h
ar
d
war
e
o
f
th
e
p
r
o
p
o
s
ed
s
y
s
tem
C
o
m
p
o
n
e
n
t
D
e
scri
p
t
i
o
n
R
o
b
o
t
Th
e
r
o
b
o
t
h
a
s
t
h
r
e
e
f
u
n
c
t
i
o
n
s;
t
h
e
f
i
r
st
o
n
e
i
s
t
o
r
e
t
r
i
e
v
e
t
h
e
p
i
e
c
e
w
h
e
n
r
e
c
e
i
v
i
n
g
a
s
i
g
n
a
l
f
r
o
m
t
h
e
se
n
so
r
(
i
n
d
i
c
a
t
e
t
h
a
t
t
h
e
p
a
r
t
i
s
c
o
mi
n
g
)
.
T
h
e
sec
o
n
d
o
n
e
i
s
t
o
t
r
i
g
g
e
r
t
h
e
c
a
mera
t
o
t
a
k
e
p
i
c
t
u
r
e
s
(
se
n
d
t
o
2
4
V
t
o
t
h
e
c
a
mer
a
i
n
t
h
e
t
r
i
g
g
e
r
l
i
n
e
)
.
Th
e
t
h
i
r
d
f
u
n
c
t
i
o
n
i
s re
c
e
i
v
i
n
g
t
h
e
s
t
a
t
u
s fr
o
m
t
h
e
P
C
a
s
i
n
p
u
t
.
PC
Th
e
c
o
m
p
u
t
e
r
c
o
n
t
a
i
n
s
t
h
e
v
i
si
o
n
s
y
s
t
e
m
a
n
d
w
h
e
n
t
h
e
a
r
t
i
f
i
c
i
a
l
i
n
t
e
l
l
i
g
e
n
c
e
m
o
d
e
l
se
n
d
s
t
h
e
s
t
a
t
u
s
o
f
t
h
e
p
a
r
t
(
OK
o
r
n
o
n
-
OK
)
.
T
h
e
P
C
s
e
n
d
s
t
h
e
s
t
a
t
u
s t
o
t
h
e
r
o
b
o
t
C
a
mer
a
Th
e
I
P
c
a
m
e
r
a
c
o
n
t
a
i
n
s
t
w
o
l
i
n
e
s;
o
n
e
i
s
u
se
d
f
o
r
p
o
w
e
r
i
n
g
a
n
d
t
h
e
se
c
o
n
d
o
n
e
f
o
r
t
r
i
g
g
e
r
i
n
g
t
h
e
c
a
m
e
r
a
t
o
t
a
k
e
t
h
e
p
i
c
t
u
r
e
s.
S
e
n
s
o
r
Th
e
u
s
e
d
se
n
s
o
r
i
s
a
n
u
l
t
r
a
s
o
n
i
c
o
n
e
t
h
a
t
w
o
r
k
s o
n
t
h
e
p
r
i
n
c
i
p
l
e
o
f
e
mi
t
t
i
n
g
sh
o
r
t
h
i
g
h
-
f
r
e
q
u
e
n
c
y
so
u
n
d
p
u
l
s
e
s o
n
t
h
e
c
o
n
c
e
p
t
o
f
e
q
u
a
l
i
n
t
e
r
v
a
l
.
W
h
e
n
t
h
e
se
p
u
l
ses
e
n
c
o
u
n
t
e
r
a
n
o
b
s
t
a
c
l
e
,
t
h
e
y
a
r
e
e
c
h
o
e
d
a
n
d
r
e
f
l
e
c
t
e
d
t
o
t
h
e
sen
s
o
r
a
s
a
r
e
f
l
e
c
t
i
o
n
.
T
h
i
s
i
s w
h
a
t
m
a
k
e
s
i
t
p
o
ss
i
b
l
e
t
o
c
a
l
c
u
l
a
t
e
t
h
e
d
i
st
a
n
c
e
b
e
t
w
e
e
n
t
h
e
se
n
s
o
r
a
n
d
t
h
e
o
b
j
e
c
t
.
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
YOLOv5
:
a
n
imp
r
o
ve
d
a
lg
o
r
ith
m
fo
r
r
ea
l
-
time
d
etec
tio
n
o
f i
n
d
u
s
tr
ia
l d
efec
tive
… (
A
b
d
ela
z
iz
E
lb
a
g
h
d
a
d
i
)
747
2
.
1
.
Da
t
a
s
et
T
h
e
d
ataset
is
co
llected
f
r
o
m
a
r
ea
l
p
r
o
d
u
ctio
n
lin
e;
it
co
n
tain
s
4
0
0
im
ag
es.
T
h
e
im
ag
es
ar
e
lab
eled
u
s
in
g
lab
el
s
tu
d
io
to
o
ls
.
T
h
e
d
ef
ec
t
m
ay
ap
p
ea
r
o
n
t
h
e
p
ie
ce
’
s
s
u
r
f
ac
e
in
th
r
ee
d
is
tin
ct
ar
ea
s
.
T
h
e
d
ataset
is
d
iv
id
ed
in
to
3
0
%
f
o
r
th
e
test
an
d
7
0
%
f
o
r
th
e
tr
ai
n
.
As
th
e
s
t
u
d
ied
p
iece
is
co
n
f
id
en
tial,
w
e
s
h
o
w
in
Fig
u
r
e
2
its
ap
p
r
o
x
im
atio
n
illu
s
tr
atio
n
.
As
s
h
o
wn
,
th
e
o
b
jectiv
e
o
f
u
s
in
g
th
e
co
m
p
u
ter
v
is
io
n
m
et
h
o
d
s
is
to
c
h
ec
k
th
e
v
alid
ity
o
f
th
e
p
iece
,
in
o
th
er
wo
r
d
s
,
it
c
o
n
tr
o
ls
if
th
e
th
r
ee
ar
ea
s
illu
s
tr
ated
d
o
n
o
t
ca
r
r
y
a
n
y
d
ef
ec
t.
T
h
e
p
iece
is
co
n
s
id
er
ed
n
o
n
-
v
alid
if
at
le
ast o
n
e
ar
ea
is
d
ef
ec
tiv
e.
Fig
u
r
e
2
.
T
h
e
p
iece
illu
s
tr
atio
n
2
.
2
.
M
a
chine le
a
rning
m
et
ri
cs
T
h
is
s
e
cti
o
n
d
es
cr
ib
es
th
e
ev
alu
ati
o
n
m
e
tr
ics
a
d
o
p
t
ed
t
o
a
s
s
ess
t
h
e
p
er
f
o
r
m
an
ce
o
f
th
e
p
r
o
p
o
s
e
d
m
et
h
o
d
.
T
h
ese
m
e
tr
ics
r
el
y
o
n
th
e
a
n
al
y
s
is
o
f
t
r
u
e
an
d
f
a
ls
e
p
r
e
d
i
cti
o
n
o
u
t
c
o
m
es t
o
d
et
er
m
i
n
e
th
e
e
f
f
ec
ti
v
e
n
ess
o
f
t
h
e
m
o
d
e
l.
S
p
ec
if
ica
ll
y
,
t
h
e
e
v
al
u
ati
o
n
is
b
as
e
d
o
n
f
o
u
r
f
u
n
d
am
e
n
ta
l
q
u
a
n
ti
ties
:
t
r
u
e
n
e
g
a
ti
v
e
(
T
N
)
,
t
r
u
e
p
o
s
iti
v
e
(
T
P)
,
f
als
e
n
e
g
at
iv
e
(
FN
)
,
an
d
f
a
ls
e
p
o
s
iti
v
e
(
FP
)
.
T
N
c
o
r
r
es
p
o
n
d
s
to
n
e
g
ati
v
e
s
a
m
p
les
c
o
r
r
e
ctl
y
class
i
f
i
ed
as
n
e
g
ati
v
e
,
T
P
r
ef
er
s
t
o
p
o
s
it
iv
e
s
a
m
p
les
c
o
r
r
ec
tly
i
d
e
n
t
if
ie
d
as
p
o
s
it
i
v
e
,
FN
r
e
p
r
ese
n
ts
p
o
s
iti
v
e
s
am
p
les
i
n
c
o
r
r
ec
tl
y
class
if
ie
d
as
n
e
g
a
ti
v
e
,
an
d
FP
d
en
o
t
es
n
e
g
at
iv
e
s
am
p
l
es
in
c
o
r
r
e
ctl
y
p
r
e
d
ic
te
d
as
p
o
s
it
iv
e.
Pre
cisi
o
n
is
th
e
f
ir
s
t
ev
alu
atio
n
m
etr
ic
an
d
m
ea
s
u
r
es
th
e
p
r
o
p
o
r
tio
n
o
f
c
o
r
r
ec
tly
p
r
ed
icte
d
p
o
s
itiv
e
in
s
tan
ce
s
am
o
n
g
all
s
am
p
les c
lass
if
ied
as p
o
s
itiv
e.
I
t is co
m
p
u
ted
as
(
1
)
.
=
+
(
1
)
R
ec
all
is
th
e
s
ec
o
n
d
m
etr
ic
an
d
ev
al
u
ates
th
e
m
o
d
el’
s
a
b
ilit
y
to
co
r
r
ec
tly
id
e
n
tify
p
o
s
itiv
e
s
am
p
les.
I
t
r
ep
r
esen
ts
th
e
r
atio
o
f
co
r
r
ec
tl
y
d
etec
ted
p
o
s
itiv
e
in
s
tan
ce
s
to
th
e
to
tal
n
u
m
b
er
o
f
ac
tu
al
p
o
s
itiv
e
s
am
p
les
an
d
is
d
ef
in
ed
as
(
2
)
.
=
+
(
2
)
Acc
u
r
ac
y
is
th
e
t
h
ir
d
m
etr
ic
u
s
ed
to
ev
alu
ate
th
e
o
v
er
all
p
e
r
f
o
r
m
an
ce
o
f
th
e
class
if
icatio
n
m
o
d
el.
I
t
in
d
icate
s
th
e
f
r
ac
tio
n
o
f
co
r
r
ec
tly
class
if
ied
s
am
p
les am
o
n
g
all
e
v
alu
at
ed
in
s
tan
ce
s
an
d
is
f
o
r
m
ally
e
x
p
r
ess
ed
as
(
3
)
.
=
+
+
+
+
(
3
)
T
h
e
f
in
al
m
etr
ic
m
ea
s
u
r
es
th
e
m
ea
n
av
er
a
g
e
p
r
ec
is
io
n
(
m
AP)
r
ate
o
f
th
e
o
b
ject
d
etec
tio
n
a
lg
o
r
ith
m
s
in
clu
d
in
g
f
ast
cu
r
r
en
t
n
e
u
r
al
n
etwo
r
k
s
an
d
f
aster
co
r
o
n
e
r
n
etwo
r
k
s
.
T
h
e
YOL
O
alg
o
r
ith
m
s
o
lv
es
two
m
ajo
r
p
r
o
b
lem
s
: c
lass
if
icatio
n
an
d
lo
ca
lizatio
n
[
2
8
]
.
T
h
e
class
if
icatio
n
id
en
tifie
s
if
th
e
o
b
ject
an
d
its
clas
s
ar
e
o
n
th
e
im
ag
e.
T
h
e
lo
ca
lizatio
n
p
r
ed
i
cts
th
e
co
o
r
d
in
ates
o
f
th
e
b
o
u
n
d
in
g
b
o
x
ar
o
u
n
d
.
Fu
r
th
er
m
o
r
e,
th
e
co
n
ce
p
t
o
f
in
ter
s
ec
tio
n
o
v
er
u
n
i
o
n
(
I
o
U)
is
u
s
ed
b
y
th
e
YOL
O
alg
o
r
ith
m
to
d
en
o
te
h
o
w
m
u
ch
t
h
e
p
r
ed
icted
b
o
u
n
d
ar
y
o
v
er
lap
s
th
e
b
o
u
n
d
in
g
b
o
x
.
T
h
e
I
o
U
is
ca
lcu
lated
b
y
(
4
)
.
=
(
4
)
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
1
5
,
No
.
1
,
Feb
r
u
ar
y
2
0
2
6
:
744
-
7
5
5
748
T
h
e
I
o
U
m
etr
ic
f
o
llo
ws
ce
r
tai
n
co
n
d
itio
n
s
.
I
f
I
o
U
is
s
u
p
e
r
io
r
to
o
r
eq
u
al
to
0
.
5
,
th
e
o
b
ject
d
etec
tio
n
is
class
if
ied
as
T
P.
E
ls
e,
th
en
it
is
co
n
s
id
er
ed
as
a
wr
o
n
g
d
etec
tio
n
an
d
class
if
ied
as
FP
.
Fu
r
th
er
m
o
r
e,
th
e
m
AP is ca
lcu
lated
u
s
in
g
(
5
)
.
=
1
∑
=
1
(
5
)
W
h
er
e
AP
c
is
th
e
f
in
d
in
g
av
er
a
g
e
p
r
ec
is
io
n
o
f
ea
c
h
class
.
2
.
3
.
YO
L
O
m
o
del
T
h
e
aim
o
f
u
s
in
g
th
e
YOL
Ov
5
alg
o
r
ith
m
is
to
d
etec
t
th
e
d
ef
ec
t
o
n
th
e
p
iece
.
T
h
r
ee
d
ef
ec
ts
ar
e
d
iv
id
ed
i
n
to
6
o
b
jects,
wh
e
r
e
two
o
b
jects
ca
n
ap
p
ea
r
in
ea
ch
ar
ea
,
a
n
d
two
o
b
jects
ca
n
ap
p
ea
r
in
th
e
f
ir
s
t,
s
ec
o
n
d
,
an
d
th
ir
d
ar
ea
.
T
h
e
f
i
r
s
t
an
d
s
ec
o
n
d
o
b
ject
ca
n
ap
p
ea
r
in
th
e
f
ir
s
t
ar
ea
,
wh
er
e
t
h
e
f
ir
s
t
o
b
ject
is
OK
an
d
th
e
s
ec
o
n
d
is
n
o
n
-
OK
,
w
h
ich
is
s
im
ilar
ly
ap
p
lied
to
th
e
r
est
ar
ea
s
.
T
o
tr
ain
th
e
Y
OL
O
alg
o
r
ith
m
,
we
d
iv
ed
th
e
im
a
g
e
in
to
3
×
3
b
o
u
n
d
in
g
b
o
x
es.
T
h
e
ta
r
g
et
Y
is
d
e
f
in
ed
as d
escr
ib
ed
in
(
6
)
.
(
ℎ
1
2
3
4
5
6
)
(
6
)
W
h
er
e
P
c
in
d
icate
s
if
th
e
o
b
ject
ex
is
ts
in
th
e
b
o
u
n
d
in
g
b
o
x
o
r
n
o
t,
if
it
is
th
e
ca
s
e,
it
g
ets
1
else
0
.
B
x
an
d
B
y
in
d
icate
th
e
co
o
r
d
in
ates o
f
th
e
o
b
ject.
C
1
,
C
2
,
C
3
,
C
4
,
C
5
, a
nd
C
6
p
r
esen
t th
e
p
o
s
s
ib
le
class
es o
f
th
e
o
b
ject.
2
.
4
.
E
f
f
icient
Net
m
o
del
E
f
f
icien
tNet
is
C
NN
ar
ch
itec
tu
r
e
an
d
s
ca
lin
g
m
eth
o
d
[
2
9
]
.
T
h
e
aim
s
o
f
th
is
ar
ch
itect
u
r
e
ar
e
to
in
cr
ea
s
e
th
e
d
im
en
s
io
n
o
f
th
e
d
ep
th
/wid
th
/r
eso
lu
tio
n
.
T
h
e
d
ep
th
is
to
ad
d
m
o
r
e
lay
er
s
,
th
e
wid
th
is
to
in
cr
ea
s
e
th
e
n
u
m
b
e
r
o
f
f
ea
tu
r
es,
an
d
th
e
r
eso
lu
tio
n
m
ea
n
s
to
in
c
r
ea
s
e
th
e
r
eso
lu
tio
n
o
f
th
e
im
ag
es.
Fig
u
r
e
3
s
h
o
ws
th
e
ac
cu
r
ac
y
ac
c
o
r
d
in
g
to
th
e
n
u
m
b
er
s
o
f
th
e
p
ar
am
eter
s
.
T
h
e
f
ig
u
r
e
i
n
d
icate
s
th
at
E
f
f
icien
t
NetB7
r
ea
ch
es
8
4
.
3
%
with
6
6
M
p
ar
a
m
eter
s
.
I
n
t
h
is
p
r
o
p
o
s
ed
ap
p
r
o
ac
h
,
we
u
s
e
E
f
f
icien
tNetB
0
.
Fig
u
r
e
3
.
E
f
f
icien
tB
0
T
o
B
7
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
YOLOv5
:
a
n
imp
r
o
ve
d
a
lg
o
r
ith
m
fo
r
r
ea
l
-
time
d
etec
tio
n
o
f i
n
d
u
s
tr
ia
l d
efec
tive
… (
A
b
d
ela
z
iz
E
lb
a
g
h
d
a
d
i
)
749
3.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
N
All
ex
p
er
im
en
ts
h
eld
i
n
th
is
p
ap
er
wer
e
ca
r
r
ied
o
u
t
u
s
in
g
th
e
co
d
in
g
la
n
g
u
a
g
e
Py
th
o
n
an
d
th
e
Py
to
r
ch
d
ee
p
lear
n
in
g
f
r
am
e
wo
r
k
.
I
n
ad
d
itio
n
to
x
AR
M
6
R
o
b
o
t,
th
e
h
ar
d
war
e
in
cl
u
d
es
an
I
n
tel
(
R
)
C
o
r
e
(
T
M)
i7
-
1
0
7
5
0
H
C
PU
r
u
n
n
in
g
at
2
.
6
0
GHz
,
1
6
GB
o
f
R
AM
,
an
NVI
DI
A
GeFo
r
ce
G
T
X
1
6
5
0
T
i
(
4
GB
)
GPU,
an
d
W
in
d
o
ws
1
0
(
6
4
-
b
i
t)
.
I
n
o
u
r
s
im
u
latio
n
,
we
u
tili
ze
d
th
e
ca
teg
o
r
ical
cr
o
s
s
-
en
tr
o
p
y
lo
s
s
f
u
n
ctio
n
to
tr
ain
th
e
E
f
f
icien
tB
0
alg
o
r
ith
m
u
s
in
g
a
s
p
ec
if
ic
d
ataset.
T
h
e
lo
s
s
f
u
n
ctio
n
m
ea
s
u
r
es
th
e
d
is
s
im
ilar
ity
b
etwe
en
th
e
p
r
e
d
icted
p
r
o
b
ab
ilit
ies
an
d
th
e
tr
u
e
lab
els
f
o
r
ea
c
h
clas
s
in
a
m
u
lti
-
class
class
if
icatio
n
p
r
o
b
lem
.
T
ab
le
2
s
u
m
m
ar
izes th
e
p
r
im
ar
y
tr
ain
i
n
g
p
ar
a
m
eter
s
.
T
ab
le
2
.
Simu
latio
n
s
ettin
g
s
a
n
d
th
eir
v
alu
es
V
a
r
i
a
b
l
e
S
p
e
e
d
(
r
p
m)
I
n
i
t
i
a
l
l
e
a
r
n
i
n
g
r
a
t
e
l
r
0
=
0
.
0
1
f
i
n
a
l
O
n
e
C
y
c
l
e
LR
l
e
a
r
n
i
n
g
r
a
t
e
l
r
f
:
0
.
1
mo
me
n
t
u
m
0
.
9
3
7
w
e
i
g
h
t
_
d
e
c
a
y
0
.
0
0
0
5
W
a
r
mu
p
e
p
o
c
h
s
3
.
0
w
a
r
mu
p
_
m
o
m
e
n
t
u
m
0
.
8
w
a
r
mu
p
_
b
i
a
s_
l
r
0
.
1
b
o
x
0
.
0
5
c
l
s
0
.
3
c
l
s
_
p
w
1
.
0
c
l
s
_
p
w
0
.
7
o
b
j
_
p
w
1
.
0
i
o
u
_
t
0
.
2
0
a
n
c
h
o
r
_
t
4
.
0
f
l
_
g
a
mm
a
0
.
0
h
sv
_
h
0
.
0
1
5
h
sv
_
s
0
.
7
3
.
1
.
M
o
del per
f
o
rma
nce
ev
a
lua
t
io
n
Fig
u
r
e
4
s
h
o
ws
th
e
co
n
f
u
s
io
n
m
atr
ix
,
wh
ich
ev
alu
ates
th
e
q
u
ality
o
f
th
e
class
if
ier
’
s
o
u
tp
u
t
o
n
th
e
g
iv
en
d
ata
s
et.
T
h
e
d
iag
o
n
al
elem
en
ts
r
ef
lect
th
e
n
u
m
b
er
o
f
p
o
in
ts
f
o
r
wh
ich
th
e
p
r
e
d
icted
lab
el
eq
u
als
th
e
r
ea
l
lab
el,
a
n
d
o
f
f
-
d
iag
o
n
al
el
em
en
ts
ar
e
th
o
s
e
f
o
r
wh
ich
th
e
class
if
ier
m
is
lab
eled
.
T
h
e
g
r
ea
ter
th
e
d
iag
o
n
al
v
alu
es
o
f
th
e
co
n
f
u
s
io
n
m
atr
i
x
,
th
e
r
ig
h
ter
p
r
e
d
ictio
n
s
th
er
e
ar
e.
T
h
is
ty
p
e
o
f
p
r
esen
tatio
n
ca
n
b
e
u
s
ef
u
l
f
o
r
p
r
o
v
id
i
n
g
a
m
o
r
e
v
is
u
al
v
iew
o
f
wh
ich
class
es
ar
e
b
ei
n
g
m
is
class
if
ied
.
Fro
m
th
is
f
ig
u
r
e,
we
g
et
th
e
r
esu
lts
s
h
o
wn
in
T
ab
le
3
.
Fro
m
th
is
tab
le,
we
n
o
tice
th
at
th
e
s
en
s
itiv
ity
r
ea
ch
es,
a
r
ec
all
o
f
0
.
9
8
,
a
s
p
ec
if
icity
o
f
0
.
9
9
,
a
p
r
ec
is
io
n
o
f
0
.
9
8
,
an
d
an
F
1
-
s
co
r
e
o
f
0
.
9
8
.
T
h
ese
p
er
f
o
r
m
an
ce
s
s
h
o
w
p
o
s
itiv
e
f
ee
d
b
a
ck
ab
o
u
t
t
h
e
v
is
u
al
in
ter
p
r
etatio
n
o
f
th
e
a
p
p
r
o
ac
h
th
at
th
e
class
es a
r
e
b
ein
g
well
class
if
ied
.
Fig
u
r
e
4
.
C
o
n
f
u
s
io
n
m
atr
i
x
p
e
r
f
o
r
m
a
n
ce
r
ep
r
esen
tatio
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
5
,
No
.
1
,
Feb
r
u
ar
y
2
0
2
6
:
744
-
7
5
5
750
T
ab
le
3
.
C
o
n
f
u
s
io
n
m
atr
ix
o
u
t
p
u
ts
m
ea
s
u
r
em
en
ts
M
e
a
su
r
e
S
p
e
e
d
(
r
p
m)
R
e
c
a
l
l
0
.
9
8
S
p
e
c
i
f
i
c
a
t
i
o
n
0
.
9
9
P
r
e
c
i
s
i
o
n
0
.
9
8
9
9
N
e
g
a
t
i
v
e
p
r
e
d
i
c
t
i
v
e
v
a
l
u
e
0
.
9
8
0
2
F
a
l
se
p
o
si
t
i
v
e
r
a
t
e
0
.
0
1
F
a
l
se
d
i
sc
o
v
e
r
y
r
a
t
e
0
.
0
1
F
a
l
se
n
e
g
a
t
i
v
e
r
a
t
e
0
.
0
2
A
c
c
u
r
a
c
y
0
.
9
8
5
0
F1
-
sc
o
r
e
0
.
9
8
4
9
T
h
e
b
alan
ce
b
etwe
en
ac
cu
r
ac
y
an
d
r
ec
all
m
ay
b
e
illu
s
tr
ated
u
s
in
g
th
e
F1
-
s
co
r
e
cu
r
v
e
,
an
d
a
d
esig
n
p
o
in
t
ca
n
b
e
estab
lis
h
ed
u
s
in
g
Fig
u
r
e
5
.
T
h
e
co
n
f
i
d
en
ce
v
alu
e
th
at
im
p
r
o
v
es
ac
cu
r
ac
y
a
n
d
r
ec
all
is
0
.
3
2
1
,
as
illu
s
tr
ated
in
th
e
F1
-
s
co
r
e
c
u
r
v
e.
Gr
ea
ter
co
n
f
id
e
n
ce
v
alu
e
i
s
p
r
ef
er
r
ed
in
m
a
n
y
cir
cu
m
s
tan
ce
s
.
I
n
th
e
ca
s
e
o
f
th
is
m
o
d
el,
a
co
n
f
id
en
ce
o
f
0
.
4
m
ay
b
e
ap
p
r
o
p
r
iate
b
ec
a
u
s
e
th
e
F1
-
s
co
r
e
v
alu
e
lo
o
k
s
to
b
e
ab
o
u
t
0
.
6
3
,
wh
ich
r
em
ain
s
clo
s
e
to
th
e
m
ax
im
u
m
v
alu
e
o
f
0
.
9
0
.
Ob
s
er
v
i
n
g
t
h
e
ac
cu
r
ac
y
an
d
r
ec
all
n
u
m
b
er
s
at
a
co
n
f
id
en
ce
lev
el
o
f
0
.
4
s
u
g
g
ests
th
at
th
is
m
ig
h
t
b
e
a
g
o
o
d
d
esig
n
p
o
in
t.
T
h
e
r
ec
all
v
alu
e
b
eg
in
s
to
d
eg
r
ad
e
ar
o
u
n
d
0
.
4
,
alth
o
u
g
h
th
e
ac
cu
r
ac
y
v
al
u
e
r
e
m
ain
s
co
n
s
tan
t.
Fig
u
r
es
6
a
n
d
7
b
o
th
s
h
o
w
th
e
p
er
f
o
r
m
an
ce
o
f
t
h
e
E
f
f
i
c
ien
tB
0
al
g
o
r
it
h
m
in
te
r
m
s
o
f
ac
c
u
r
a
cy
th
r
o
u
g
h
o
u
t
t
h
e
v
a
li
d
at
io
n
p
r
o
c
ess
.
Fi
g
u
r
e
6
p
r
es
e
n
ts
th
e
lo
s
s
p
er
f
o
r
m
an
ce
o
f
t
h
e
a
lg
o
r
it
h
m
f
o
r
test
in
g
r
ea
ch
es
ze
r
o
a
f
t
er
a
b
o
u
t
2
7
ep
o
ch
s
.
I
n
o
t
h
e
r
s
id
e,
Fi
g
u
r
e
7
s
h
o
ws
t
h
a
t
a
f
te
r
t
h
e
2
1
st
ep
o
ch
,
th
e
ac
cu
r
ac
y
o
f
t
h
e
al
g
o
r
it
h
m
o
n
t
h
e
d
at
ase
t
r
e
ac
h
es
7
4
%
.
T
h
is
ac
c
u
r
ac
y
v
al
u
e
i
n
d
ica
tes
t
h
e
p
e
r
c
en
ta
g
e
o
f
c
o
r
r
ec
t
ly
c
lass
if
i
ed
s
a
m
p
les
f
r
o
m
th
e
t
o
t
al
n
u
m
b
er
o
f
s
am
p
l
es
in
th
e
d
at
aset
.
R
ea
ch
in
g
a
7
4
%
a
cc
u
r
ac
y
af
te
r
e
p
o
ch
2
1
s
u
g
g
es
t
s
th
at
t
h
e
al
g
o
r
it
h
m
h
as
u
n
d
e
r
g
o
n
e
s
ev
e
r
al
t
r
ai
n
i
n
g
it
er
ati
o
n
s
an
d
h
as
le
ar
n
e
d
t
o
m
a
k
e
a
cc
u
r
at
e
p
r
e
d
ic
ti
o
n
s
o
n
t
h
e
g
i
v
e
n
d
a
tas
et.
Fig
u
r
e
5
.
F1
-
s
co
r
e
cu
r
v
e
r
ep
r
e
s
en
tatio
n
Fig
u
r
e
6
.
L
o
s
s
p
er
f
o
r
m
an
ce
o
f
tr
ain
an
d
test
o
v
er
ep
o
ch
s
Fig
u
r
e
7
.
A
cc
u
r
ac
y
o
f
tr
ain
o
v
er
ep
o
ch
s
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
YOLOv5
:
a
n
imp
r
o
ve
d
a
lg
o
r
ith
m
fo
r
r
ea
l
-
time
d
etec
tio
n
o
f i
n
d
u
s
tr
ia
l d
efec
tive
… (
A
b
d
ela
z
iz
E
lb
a
g
h
d
a
d
i
)
751
Fig
u
r
e
8
illu
s
tr
ates
th
e
r
ec
all
m
etr
ic,
s
ty
led
b
y
th
e
f
ir
s
t
m
etr
ic,
in
cr
ea
s
in
g
g
r
ad
u
ally
o
v
er
th
e
f
ir
s
t
4
0
ep
o
c
h
s
,
r
em
ain
in
g
u
n
d
er
0
.
8
.
Af
ter
ep
o
ch
4
0
,
th
e
r
ec
all
m
etr
ic
ex
p
er
ien
ce
s
a
s
ig
n
if
ica
n
t
r
is
e,
u
ltima
tely
r
ea
ch
in
g
0
.
9
9
.
T
h
is
s
u
g
g
ests
th
at
th
e
m
o
d
el
im
p
r
o
v
es
its
ab
ilit
y
to
id
en
tify
r
elev
an
t
in
s
tan
ce
s
as
tr
ain
in
g
p
r
o
g
r
ess
es.
Fig
u
r
e
9
p
r
esen
ts
th
e
lo
s
s
v
alu
es
f
o
r
b
o
x
lo
s
s
,
o
b
ject
lo
s
s
,
an
d
class
if
icatio
n
lo
s
s
,
wh
ich
co
n
v
e
r
g
e
to
0
.
0
4
,
0
.
0
1
,
an
d
0
.
0
0
3
7
5
,
r
esp
ec
tiv
ely
.
T
h
ese
r
esu
lts
in
d
icate
th
at
as
th
e
tr
ain
in
g
ad
v
an
ce
s
,
th
e
m
o
d
el
ef
f
ec
tiv
ely
lear
n
s
to
e
x
tr
ac
t
f
e
atu
r
es
an
d
ac
cu
r
ately
lo
ca
lize
o
b
jects
with
in
th
e
in
s
p
ec
ted
i
m
ag
es.
A
lo
wer
lo
s
s
s
ig
n
if
ies b
etter
g
en
er
aliza
tio
n
,
m
ea
n
in
g
t
h
e
m
o
d
el
s
u
cc
ess
f
u
lly
m
in
im
izes e
r
r
o
r
s
in
d
etec
tin
g
d
ef
ec
ts
.
Fig
u
r
e
8
.
Pre
cisi
o
n
a
n
d
r
ec
all
m
etr
ics
Fig
u
r
e
9
.
Valid
atio
n
lo
s
s
p
er
f
o
r
m
an
ce
v
e
r
s
u
s
ea
ch
ep
o
c
h
Fig
u
r
e
1
0
illu
s
tr
ates
th
at
th
e
er
r
o
r
s
f
o
r
b
o
x
,
o
b
ject,
an
d
class
if
icatio
n
in
cr
ea
s
e
to
0
.
0
2
,
0
.
0
1
6
,
an
d
0
.
0
0
2
7
,
r
esp
ec
tiv
ely
.
Ad
d
itio
n
ally
,
it
co
n
f
ir
m
s
th
at
o
u
r
p
r
o
p
o
s
ed
ap
p
r
o
ac
h
ef
f
ec
tiv
el
y
d
etec
ts
b
o
th
th
e
b
o
x
an
d
th
e
o
b
ject
in
th
e
in
s
p
ec
ted
im
ag
e.
I
n
t
h
is
ex
p
er
im
en
t,
we
s
et
two
I
o
U
th
r
esh
o
ld
s
:
0
.
5
a
n
d
0
.
5
-
0
.
9
5
.
W
h
e
n
I
o
U
is
0
.
5
,
th
e
m
AP
r
ea
c
h
es
0
.
5
5
,
wh
er
ea
s
it
d
r
o
p
s
to
0
.
2
5
f
o
r
a
n
I
o
U
r
an
g
e
o
f
0
.
5
-
0
.
9
5
.
Fig
u
r
e
1
1
d
is
p
lay
s
th
e
s
im
u
latio
n
b
atch
es
o
f
th
e
in
s
p
ec
ted
p
iece
s
,
wh
er
e
d
ef
e
ctiv
e
ar
ea
s
ar
e
h
ig
h
lig
h
ted
with
a
b
o
u
n
d
in
g
b
o
x
lab
eled
'
N
-
OK
'
an
d
n
o
n
-
d
e
f
ec
tiv
e
o
n
es
with
'
OK
'
.
T
h
i
s
v
is
u
aliza
tio
n
r
ep
r
esen
ts
a
r
an
d
o
m
l
y
s
h
u
f
f
led
s
am
p
le
o
f
th
e
tr
ai
n
ed
d
ataset,
ca
teg
o
r
ized
in
to
two
class
es.
As
s
h
o
wn
,
th
e
alg
o
r
ith
m
d
em
o
n
s
tr
at
es
h
ig
h
p
r
ec
is
io
n
in
d
is
tin
g
u
is
h
in
g
d
ef
ec
tiv
e
r
eg
io
n
s
f
r
o
m
n
o
n
-
d
ef
ec
tiv
e
o
n
es.
Fig
u
r
e
1
0
.
T
h
e
m
AP m
etr
ics p
er
f
o
r
m
a
n
ce
ev
alu
atio
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
5
,
No
.
1
,
Feb
r
u
ar
y
2
0
2
6
:
744
-
7
5
5
752
Fig
u
r
e
1
1
.
T
h
e
s
im
u
latio
n
b
at
ch
es o
f
s
tu
d
ied
p
iece
s
3
.
2
.
Dis
cus
s
io
ns
Ma
n
y
s
cien
tific
p
ap
er
s
h
av
e
t
r
ea
ted
v
is
u
al
in
s
p
ec
tio
n
u
s
in
g
m
ac
h
in
e
lear
n
in
g
.
I
n
ad
d
itio
n
,
au
to
m
ated
d
ef
ec
t
d
etec
tio
n
u
s
in
g
co
m
p
u
ter
v
is
io
n
an
d
m
ac
h
in
e
lear
n
i
n
g
h
as
b
ec
o
m
e
a
p
r
o
m
is
in
g
r
esear
ch
ar
ea
with
a
h
ig
h
an
d
d
ir
ec
t
im
p
ac
t
o
n
th
e
f
ield
s
o
f
v
is
u
al
in
s
p
ec
tio
n
.
I
s
m
ail
an
d
Ma
lik
[
3
0
]
co
m
p
ar
e
th
e
p
er
f
o
r
m
an
ce
o
f
m
ac
h
in
e
lear
n
in
g
alg
o
r
ith
m
s
,
in
clu
d
in
g
R
esNet,
Den
s
eNe
t,
Mo
b
ileNetV2
,
NASNet,
an
d
E
f
f
icien
tNet,
in
wh
ich
v
is
u
al
in
s
p
ec
tio
n
is
ap
p
lied
in
th
e
f
ield
s
o
f
ag
r
icu
ltu
r
e
.
T
h
e
r
e
v
ea
lin
g
r
esu
lts
o
f
th
is
wo
r
k
s
h
o
w
th
at
th
e
ac
cu
r
ac
y
r
ea
ch
es
9
9
.
2
an
d
9
8
.
6
%
u
s
in
g
th
e
E
f
f
icien
tNet
m
o
d
el.
I
n
[
3
1
]
,
d
ee
p
lear
n
in
g
is
p
r
o
p
o
s
ed
f
o
r
th
e
q
u
an
titativ
e
ass
ess
m
en
t
o
f
v
is
u
al
d
etec
tab
ilit
y
o
f
d
if
f
er
en
t
t
y
p
es
o
f
in
-
s
er
v
ice
d
e
f
ec
ts
in
l
am
in
ated
co
m
p
o
s
ite
s
tr
u
ctu
r
es.
T
h
e
r
esu
lts
s
h
o
w
th
at
em
p
lo
y
i
n
g
Alex
Net
n
etwo
r
k
,
u
s
in
g
th
e
r
elativ
ely
s
m
all
im
ag
e
d
ataset,
p
r
o
v
id
e
d
th
e
h
i
g
h
est
ac
cu
r
ac
y
lev
el
(
8
7
-
9
6
%)
f
o
r
id
en
tif
y
in
g
th
e
d
am
a
g
e
s
ev
er
ity
an
d
ty
p
es
in
a
r
ea
s
o
n
ab
le
co
m
p
u
tatio
n
al
tim
e.
I
n
s
tead
o
f
u
s
in
g
th
e
m
en
tio
n
ed
d
ata
s
et,
in
o
u
r
p
r
o
p
o
s
ed
ap
p
r
o
ac
h
w
e
h
av
e
u
s
ed
cu
s
to
m
d
ata
th
at
tar
g
ets th
e
im
p
r
o
v
em
en
t o
f
s
m
ar
t
I
n
d
u
s
tr
ial
4
.
0
ap
p
licatio
n
s
th
at
co
n
ce
r
n
s
th
e
r
ea
l p
r
o
d
u
ctio
n
lin
es.
T
h
e
p
r
o
p
o
s
ed
m
eth
o
d
’
s
p
er
f
o
r
m
an
ce
was
ac
h
iev
ed
b
y
lear
n
i
n
g
f
r
o
m
two
s
ets
o
f
im
ag
es
o
f
d
ef
ec
tiv
e
an
d
n
o
n
-
d
e
f
ec
tiv
e
s
am
p
les.
F
u
r
th
er
m
o
r
e,
u
s
in
g
o
n
ly
h
alf
o
f
th
e
d
e
f
ec
tiv
e
s
am
p
les
d
em
o
n
s
tr
ated
th
at
g
o
o
d
p
er
f
o
r
m
an
ce
c
o
u
ld
s
till
b
e
a
ch
iev
ed
,
wh
e
r
ea
s
r
elate
d
m
et
h
o
d
s
p
r
o
d
u
ce
d
wo
r
s
e
r
esu
lts
in
th
is
ca
s
e.
T
h
is
in
d
icate
s
th
at
th
e
YOL
O
a
p
p
r
o
ac
h
u
s
ed
is
a
p
p
r
o
p
r
iate
f
o
r
t
h
e
in
v
esti
g
ated
in
d
u
s
tr
ial
ap
p
l
icatio
n
,
d
esp
ite
th
e
lim
ited
n
u
m
b
er
o
f
d
e
f
ec
tiv
e
s
am
p
les av
ailab
le.
Fu
r
th
er
m
o
r
e,
th
r
ee
im
p
o
r
tan
t c
h
ar
ac
ter
is
tics
wer
e
ev
alu
ated
to
f
u
r
th
er
c
o
n
s
id
er
a
p
p
licatio
n
s
f
o
r
th
e
i
n
d
u
s
tr
ial
en
v
ir
o
n
m
en
t:
th
e
p
er
f
o
r
m
an
ce
to
ac
h
iev
e
a
1
0
0
%
d
etec
tio
n
r
ate,
an
n
o
tatio
n
d
etails,
an
d
co
m
p
u
tatio
n
al
tim
e.
T
h
e
YOL
Ov
5
alg
o
r
ith
m
h
as
b
ee
n
u
s
ed
to
f
in
d
th
e
lo
ca
lizatio
n
o
f
th
e
o
b
ject
an
d
its
clas
s
in
th
e
im
ag
e
to
d
etec
t
d
ef
ec
tiv
e
p
ar
ts
.
T
h
e
r
esu
lts
s
h
o
w
p
r
ec
is
io
n
a
n
d
r
ec
all
r
ea
ch
r
esp
ec
tiv
ely
9
9
an
d
8
0
%
a
n
d
ta
n
im
a
g
e
r
esu
lts
in
d
icate
th
at
o
n
e
alg
o
r
it
h
m
ca
n
id
en
tif
y
a
n
d
lo
ca
lize
th
e
o
b
ject
in
th
e
im
a
g
e.
I
n
a
d
d
itio
n
to
th
e
h
ig
h
p
r
o
v
id
ed
p
e
r
f
o
r
m
an
ce
o
f
th
e
u
s
e
d
YOL
O
v
er
s
io
n
in
ter
m
s
o
f
p
r
ec
is
io
n
an
d
r
ec
all,
th
is
alg
o
r
ith
m
is
v
er
y
s
m
o
o
t
h
an
d
f
ast
wh
ich
is
m
a
n
d
ato
r
y
t
o
d
ea
l
with
a
h
ig
h
am
o
u
n
t
o
f
d
ata
a
n
d
laten
cy
.
T
h
er
ef
o
r
e,
o
u
r
r
ea
l
d
em
o
n
s
tr
atio
n
b
y
th
e
r
o
b
o
t
s
y
s
tem
h
a
s
s
h
o
wn
in
ter
esti
n
g
r
ea
ctio
n
s
in
ter
m
s
o
f
tim
e
t
o
d
etec
tio
n
an
d
r
ea
ctio
n
.
4.
CO
NCLU
SI
O
N
Vis
u
al
in
s
p
ec
tio
n
,
au
to
m
atio
n
,
an
d
ar
tific
ial
in
tellig
en
ce
ar
e
in
cr
ea
s
in
g
ly
b
ein
g
u
s
ed
to
en
h
a
n
ce
p
r
o
d
u
ct
q
u
ality
an
d
au
t
o
m
ate
v
ar
io
u
s
task
s
in
th
e
in
d
u
s
tr
y
.
I
n
t
h
is
p
ap
er
,
we
p
r
o
p
o
s
e
an
in
tellig
en
t
s
y
s
tem
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
YOLOv5
:
a
n
imp
r
o
ve
d
a
lg
o
r
ith
m
fo
r
r
ea
l
-
time
d
etec
tio
n
o
f i
n
d
u
s
tr
ia
l d
efec
tive
… (
A
b
d
ela
z
iz
E
lb
a
g
h
d
a
d
i
)
753
th
at
in
teg
r
ates
r
o
b
o
tics
,
n
et
wo
r
k
co
m
m
u
n
icatio
n
,
an
d
a
r
tific
ial
in
tellig
en
ce
to
estab
lis
h
an
I
n
d
u
s
tr
y
4
.
0
s
o
lu
tio
n
f
o
r
d
ef
ec
t
in
s
p
ec
tio
n
.
T
h
e
f
in
d
in
g
s
o
f
th
is
s
tu
d
y
d
em
o
n
s
tr
ate
th
at
th
e
m
o
d
i
f
ied
v
er
s
io
n
o
f
th
e
YOL
Ov
5
alg
o
r
ith
m
d
eliv
er
s
p
r
o
m
is
in
g
p
er
f
o
r
m
an
ce
in
d
e
f
e
ct
class
if
icatio
n
an
d
lo
ca
lizatio
n
with
in
in
s
p
ec
ted
im
ag
es.
Fu
r
th
er
m
o
r
e,
th
e
s
y
s
tem
ac
h
iev
es
a
p
r
ec
is
io
n
o
f
9
9
%
an
d
a
r
ec
all
o
f
8
0
%,
ef
f
ec
tiv
ely
d
etec
tin
g
d
ef
ec
tiv
e
p
iece
s
o
n
th
e
p
r
o
d
u
ctio
n
lin
e.
T
h
e
s
y
n
er
g
y
b
etw
ee
n
th
e
ac
cu
r
ac
y
o
f
YOL
Ov
5
an
d
th
e
f
lex
ib
ilit
y
an
d
s
p
ee
d
o
f
th
e
x
AR
M
-
6
r
o
b
o
t
co
n
tr
ib
u
tes
to
an
ac
ce
ler
ate
d
p
r
o
d
u
ctio
n
r
ate
a
n
d
im
p
r
o
v
e
d
q
u
ality
o
f
n
ewly
m
an
u
f
ac
tu
r
ed
p
iece
s
.
As
a
f
u
tu
r
e
p
er
s
p
ec
tiv
e,
we
aim
to
ex
p
lo
r
e
en
h
an
ce
d
v
e
r
s
io
n
s
o
f
YOL
O
f
o
r
an
o
th
er
ca
s
e
s
tu
d
y
,
p
ar
ticu
lar
ly
to
an
aly
ze
its
p
er
f
o
r
m
an
ce
in
d
ete
ctin
g
co
m
p
lex
d
ef
ec
ts
th
at
p
o
s
e
ch
allen
g
es
f
o
r
h
u
m
an
o
p
er
ato
r
s
.
Ad
d
itio
n
ally
,
we
p
lan
to
in
co
r
p
o
r
ate
ed
g
e
co
m
p
u
tin
g
an
d
clo
u
d
co
m
p
u
tin
g
p
latf
o
r
m
s
to
o
p
tim
ize
laten
cy
an
d
ac
cu
r
ac
y
.
F
UNDING
I
NF
O
R
M
A
T
I
O
N
T
h
is
wo
r
k
was
n
o
t
s
u
p
p
o
r
ted
b
y
an
y
s
p
ec
if
ic
g
r
an
t
f
r
o
m
f
u
n
d
in
g
ag
e
n
cies
in
th
e
p
u
b
lic,
co
m
m
er
cial,
o
r
n
o
t
-
f
o
r
-
p
r
o
f
it
s
ec
to
r
s
.
T
h
e
s
tu
d
y
was
ca
r
r
ied
o
u
t
with
th
e
s
u
p
p
o
r
t
o
f
th
e
au
th
o
r
s
’
in
s
titu
tio
n
s
,
in
clu
d
in
g
E
NSA
T
éto
u
an
,
Un
iv
e
r
s
id
ad
Po
litécn
ica
d
e
C
ar
tag
en
a,
a
n
d
th
e
L
ab
o
r
ato
r
y
o
f
Ma
th
em
at
ics
an
d
I
n
tellig
en
t
Sy
s
tem
s
,
E
NSA
T
an
g
ier
.
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
Ab
d
elaz
iz
E
lb
ag
h
d
ad
i
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
Yass
in
e
Yaz
id
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
Ah
m
ed
E
l O
u
alk
a
d
i
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
An
to
n
io
Gu
er
r
er
o
-
Go
n
zá
lez
✓
✓
✓
✓
✓
✓
✓
✓
✓
So
u
f
ian
e
Me
zr
o
u
i
✓
✓
✓
✓
✓
✓
✓
✓
✓
C
:
C
o
n
c
e
p
t
u
a
l
i
z
a
t
i
o
n
M
:
M
e
t
h
o
d
o
l
o
g
y
So
:
So
f
t
w
a
r
e
Va
:
Va
l
i
d
a
t
i
o
n
Fo
:
Fo
r
mal
a
n
a
l
y
s
i
s
I
:
I
n
v
e
s
t
i
g
a
t
i
o
n
R
:
R
e
so
u
r
c
e
s
D
:
D
a
t
a
C
u
r
a
t
i
o
n
O
:
W
r
i
t
i
n
g
-
O
r
i
g
i
n
a
l
D
r
a
f
t
E
:
W
r
i
t
i
n
g
-
R
e
v
i
e
w
&
E
d
i
t
i
n
g
Vi
:
Vi
su
a
l
i
z
a
t
i
o
n
Su
:
Su
p
e
r
v
i
s
i
o
n
P
:
P
r
o
j
e
c
t
a
d
mi
n
i
st
r
a
t
i
o
n
Fu
:
Fu
n
d
i
n
g
a
c
q
u
i
si
t
i
o
n
CO
NF
L
I
C
T
O
F
I
N
T
E
R
E
S
T
ST
A
T
E
M
E
NT
T
h
e
au
th
o
r
s
d
ec
lar
e
th
at
th
e
r
e
ar
e
n
o
k
n
o
wn
f
i
n
an
cial
o
r
n
o
n
-
f
in
an
cial
co
m
p
etin
g
in
ter
ests
th
at
co
u
ld
h
av
e
in
f
lu
e
n
ce
d
th
e
wo
r
k
r
ep
o
r
ted
in
th
is
p
ap
e
r
.
T
h
e
au
th
o
r
s
s
tate
n
o
co
n
f
lict o
f
in
ter
est.
DATA AV
AI
L
AB
I
L
I
T
Y
T
h
e
d
ata
is
co
llected
in
r
ea
l
in
d
u
s
tr
y
p
r
o
d
u
ctio
n
.
T
h
e
d
ata
th
at
s
u
p
p
o
r
t
th
e
f
in
d
in
g
s
o
f
th
is
s
tu
d
y
will
b
e
av
ailab
le
in
GitHu
b
at
h
ttp
s
://g
ith
u
b
.
co
m
/ab
d
el
-
az
f
o
llo
win
g
a
3
-
m
o
n
t
h
f
r
o
m
th
e
d
ate
o
f
p
u
b
licatio
n
.
RE
F
E
R
E
NC
E
S
[
1
]
H
.
Y
a
n
g
a
n
d
L.
Y
u
,
“
F
e
a
t
u
r
e
e
x
t
r
a
c
t
i
o
n
o
f
w
o
o
d
-
h
o
l
e
d
e
f
e
c
t
s
u
si
n
g
w
a
v
e
l
e
t
-
b
a
se
d
u
l
t
r
a
so
n
i
c
t
e
st
i
n
g
,
”
J
o
u
rn
a
l
o
f
F
o
res
t
ry
Re
se
a
rc
h
,
v
o
l
.
2
8
,
n
o
.
2
,
p
p
.
3
9
5
–
4
0
2
,
2
0
1
6
,
d
o
i
:
1
0
.
1
0
0
7
/
s
1
1
6
7
6
-
0
1
6
-
0
2
9
7
-
z.
[
2
]
L.
Z
h
a
o
,
L.
K
a
n
g
,
a
n
d
S
.
Y
a
o
,
“
R
e
se
a
r
c
h
a
n
d
a
p
p
l
i
c
a
t
i
o
n
o
f
a
c
o
u
st
i
c
e
m
i
ssi
o
n
si
g
n
a
l
p
r
o
c
e
ssi
n
g
t
e
c
h
n
o
l
o
g
y
,
”
I
E
EE
Ac
c
e
ss
,
v
o
l
.
7
,
p
p
.
9
8
4
–
9
9
3
,
2
0
1
9
,
d
o
i
:
1
0
.
1
1
0
9
/
A
C
C
ESS
.
2
0
1
8
.
2
8
8
6
0
9
5
.
[
3
]
A
.
d
u
P
l
e
ssi
s
a
n
d
S
.
G
.
l
e
R
o
u
x
,
“
S
t
a
n
d
a
r
d
i
z
e
d
X
-
r
a
y
t
o
m
o
g
r
a
p
h
y
t
e
st
i
n
g
o
f
a
d
d
i
t
i
v
e
l
y
ma
n
u
f
a
c
t
u
r
e
d
p
a
r
t
s:
a
r
o
u
n
d
r
o
b
i
n
t
e
s
t
,
”
Ad
d
i
t
i
v
e
Ma
n
u
f
a
c
t
u
r
i
n
g
,
v
o
l
.
2
4
,
p
p
.
1
2
5
–
1
3
6
,
2
0
1
8
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
a
d
d
ma.
2
0
1
8
.
0
9
.
0
1
4
.
[
4
]
F
.
Ji
a
,
Y
.
L
e
i
,
J.
L
i
n
,
X
.
Z
h
o
u
,
a
n
d
N
.
L
u
,
“
D
e
e
p
n
e
u
r
a
l
n
e
t
w
o
r
k
s
:
a
p
r
o
m
i
si
n
g
t
o
o
l
f
o
r
f
a
u
l
t
c
h
a
r
a
c
t
e
r
i
s
t
i
c
m
i
n
i
n
g
a
n
d
i
n
t
e
l
l
i
g
e
n
t
d
i
a
g
n
o
si
s
o
f
r
o
t
a
t
i
n
g
m
a
c
h
i
n
e
r
y
w
i
t
h
massi
v
e
d
a
t
a
,
”
M
e
c
h
a
n
i
c
a
l
S
y
s
t
e
m
s
a
n
d
S
i
g
n
a
l
Pr
o
c
e
ssi
n
g
,
v
o
l
.
7
2
–
7
3
,
p
p
.
3
0
3
–
3
1
5
,
2
0
1
6
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
y
mss
p
.
2
0
1
5
.
1
0
.
0
2
5
.
[
5
]
S
.
S
u
t
h
a
h
a
r
a
n
,
Ma
c
h
i
n
e
l
e
a
r
n
i
n
g
m
o
d
e
l
s
a
n
d
a
l
g
o
ri
t
h
m
s
f
o
r
b
i
g
d
a
t
a
c
l
a
ssi
f
i
c
a
t
i
o
n
.
N
e
w
Y
o
r
k
,
U
n
i
t
e
d
S
t
a
t
e
s
:
S
p
r
i
n
g
e
r
,
2
0
1
6
,
d
o
i
:
1
0
.
1
0
0
7
/
9
7
8
-
1
-
4
8
9
9
-
7
6
4
1
-
3
_
3
.
[
6
]
S
.
A
.
-
K
u
s
t
r
i
n
a
n
d
R
.
B
e
r
e
sf
o
r
d
,
“
B
a
si
c
c
o
n
c
e
p
t
s
o
f
a
r
t
i
f
i
c
i
a
l
n
e
u
r
a
l
n
e
t
w
o
r
k
(
A
N
N
)
m
o
d
e
l
i
n
g
a
n
d
i
t
s
a
p
p
l
i
c
a
t
i
o
n
i
n
p
h
a
r
mac
e
u
t
i
c
a
l
r
e
se
a
r
c
h
,
”
J
o
u
rn
a
l
o
f
P
h
a
rm
a
c
e
u
t
i
c
a
l
a
n
d
Bi
o
m
e
d
i
c
a
l
A
n
a
l
y
s
i
s
,
v
o
l
.
2
2
,
n
o
.
5
,
p
p
.
7
1
7
–
7
2
7
,
2
0
0
0
,
d
o
i
:
1
0
.
1
0
1
6
/
s
0
7
3
1
-
7
0
8
5
(
9
9
)
0
0
2
7
2
-
1.
Evaluation Warning : The document was created with Spire.PDF for Python.