I
A
E
S
I
n
t
e
r
n
at
io
n
al
Jou
r
n
al
of
A
r
t
if
ic
ia
l
I
n
t
e
ll
ig
e
n
c
e
(
I
J
-
AI
)
V
ol
.
15
, N
o.
1
,
F
e
br
ua
r
y
2026
, pp.
116
~
128
I
S
S
N
:
2252
-
8938
,
D
O
I
:
10.11591/
ij
a
i.
v
15
.i
1
.pp
116
-
128
116
Jou
r
n
al
h
om
e
page
:
ht
tp
:
//
ij
ai
.
ia
e
s
c
or
e
.c
om
R
e
al
-
t
i
m
e
ob
j
e
c
t
d
e
t
e
c
t
i
on
t
o c
l
ass
i
f
y
e
xp
or
t
q
u
al
i
t
y of
m
an
gost
e
e
n
u
si
n
g var
i
an
t
s
of
you
on
l
y l
ook
on
c
e
ve
r
si
on
8
D
ia
n
S
a’
ad
il
la
h
M
ayl
aw
at
i
1
, M
i’
r
aj
F
u
ad
i
2
, K
u
r
n
ia
w
an
Y
u
n
ia
r
t
o
2
, Y
u
h
e
n
d
r
a A
P
2
,
R
iz
k
y R
ah
m
at
N
u
gr
ah
a
1
, A
k
b
ar
H
i
d
ayat
u
ll
a
h
H
ar
ah
ap
1
, A
gu
n
g Wah
an
a
1
1
D
e
pa
r
t
m
e
nt
of
I
nf
or
m
a
t
i
c
s
, F
a
c
ul
t
y of
S
c
i
e
nc
e
a
nd T
e
c
hnol
ogy,
U
ni
ve
r
s
i
t
a
s
I
s
l
a
m
N
e
ge
r
i
S
una
n G
unung D
j
a
t
i
B
a
ndung
,
B
a
ndung,
I
ndone
s
i
a
2
D
e
pa
r
t
m
e
nt
of
A
gr
i
c
ul
t
ur
a
l
E
ngi
ne
e
r
i
ng, F
a
c
ul
t
y of
F
ood
T
e
c
hnol
ogy a
nd A
gr
oi
ndus
t
r
y, U
ni
ve
r
s
i
t
y of
M
a
t
a
r
a
m
,
M
a
t
a
r
a
m
,
I
ndone
s
i
a
A
r
t
ic
le
I
n
f
o
A
B
S
T
R
A
C
T
A
r
ti
c
le
h
is
to
r
y
:
R
e
c
e
iv
e
d
D
e
c
24, 2024
R
e
vi
s
e
d
N
ov 10, 2025
A
c
c
e
pt
e
d
J
a
n 10, 2026
Mangosteen
is
one
of
the
leading
export
commodities
from
Ind
onesia.
Despite
its
great
economic
potential,
only
about
25%
of
Indo
nesian
mangosteens
meet
export
standards,
mainly
due
to
visual
defects
such
as
yellow
sap
and
spots
on
the
skin
of
the
fruit.
The
process
of
sorting
export
-
worthy
mangosteens
has
been
done
manually,
which
tends
to
be
time
-
consumi
ng
and
inconsi
stent.
Therefore,
this
study
aims
to
utili
ze
ar
tificial
intelligence
technology
in
building
a
real
-
time
image
recognition
m
odel
to
improve
th
e
efficiency
and
accuracy
of
the
export
-
quality
mang
osteen
sorting
process.
This
study
uses
you
only
look
once
version
8
(
YOLO
v8
)
as
an
image
recognition
model
with
YOLOv8
variants,
including
nano,
small,
medium,
large,
and
extra
large
variants.
The
results
of
the
study
using
4,014
primary
and
255
secondary
data
of
mangosteen,
the
highest
perform
ance
is
reached
by
YOLOv8
m
edium
82%
of
accuracy,
0.856
of
mean
a
verage
precision
(mAP)
50,
and
0.616
of
mAP50
-
95.
This
r
esult
is
obtaine
d
from
70%
training,
20%
validation,
and
10%
testing
data
with
epoch
st
op
85.
These
results
indicate
that
the
model
can
provide
good
performa
nce
in
mangostee
n
export
quality
classification.
This
research
contributes
to
the
fields
of
agricultural
technology
and
artificial
intell
igence
by
offer
ing
an
innovative
solution
to
a
practica
l
problem,
enhancing
ef
ficiency,
acc
uracy,
and scalabi
lity i
n export
-
quality mangosteen sorting.
K
e
y
w
o
r
d
s
:
D
e
e
p l
e
a
r
ni
ng
E
xpor
t
qua
li
ty
M
a
ngos
te
e
n
O
bj
e
c
t
de
te
c
ti
on
Y
ou only l
ook onc
e
This is an
open
acce
ss artic
le unde
r the
CC BY
-
SA
license.
C
or
r
e
s
pon
di
n
g A
u
th
or
:
D
ia
n S
a
’
a
di
ll
a
h M
a
yl
a
w
a
ti
D
e
pa
r
tm
e
nt
of
I
nf
or
m
a
ti
c
s
, F
a
c
ul
ty
of
S
c
ie
nc
e
a
nd T
e
c
hnol
ogy
U
ni
ve
r
s
it
a
s
I
s
la
m
N
e
g
e
r
i
S
una
n G
unung Dja
ti
B
a
ndung
St
. A
. H
. N
a
s
ut
io
n N
o. 105, C
ib
ir
u, B
a
ndung
40614
, I
ndone
s
ia
E
m
a
il
:
di
a
ns
m
@
ui
ns
gd.a
c
.i
d
1.
I
N
T
R
O
D
U
C
T
I
O
N
M
a
ngos
te
e
n
f
r
ui
t,
known
a
s
th
e
que
e
n
of
tr
opi
c
a
l
f
r
ui
t,
is
one
of
th
e
f
r
ui
ts
w
id
e
ly
c
ul
ti
va
te
d
in
th
e
a
gr
ic
ul
tu
r
a
l
s
e
c
to
r
.
A
s
one
of
th
e
m
o
s
t
im
por
ta
nt
e
xpor
t
c
om
m
odi
ti
e
s
in
I
ndone
s
ia
,
m
a
ngos
t
e
e
n
ge
n
e
r
a
te
s
in
c
om
e
f
or
th
e
c
ount
r
y
a
nd
f
a
r
m
e
r
s
[
1]
–
[
3]
.
H
ow
e
ve
r
,
on
e
of
th
e
m
a
in
pr
obl
e
m
s
in
th
e
m
a
ngos
te
e
n
pr
oduc
ti
on
s
ys
te
m
is
lo
w
qua
li
ty
due
to
ye
ll
ow
s
a
p
di
s
e
a
s
e
a
n
d
f
r
ui
t
s
ki
n
s
pot
s
th
a
t
a
r
e
s
ti
ll
not
in
c
lu
d
e
d
a
s
qua
li
ty
f
r
ui
t
f
or
e
xpor
t.
B
e
c
a
us
e
of
th
e
d
e
m
a
nds
of
th
e
in
t
e
r
na
ti
ona
l
m
a
r
ke
t
s
ha
r
e
f
or
hi
gh
-
qua
li
ty
a
nd
s
a
f
e
-
to
-
c
ons
um
e
pr
oduc
ts
,
pr
oduc
e
r
s
m
us
t
be
a
bl
e
to
pr
oduc
e
hi
gh
-
qua
li
ty
m
a
ngos
te
e
n
f
r
ui
t
[
4]
–
[
7
]
.
T
he
m
a
in
f
a
c
to
r
c
a
us
in
g
lo
w
m
a
ngos
te
e
n
e
xpor
ts
is
it
s
lo
w
qua
li
t
y.
O
nl
y
25%
of
I
ndone
s
ia
n
m
a
ngos
te
e
n
f
r
ui
t
m
e
e
ts
e
xpor
t
s
t
a
nda
r
ds
.
E
xpor
t
-
w
or
th
y
qua
li
ty
is
de
te
r
m
in
e
d
b
y
th
e
le
ve
l
of
s
m
oot
hne
s
s
of
th
e
f
r
ui
t
(
s
ki
n
a
nd
ye
ll
ow
s
a
p
on
th
e
f
r
ui
t)
a
nd
th
e
s
iz
e
of
th
e
f
r
ui
t.
Y
e
ll
ow
s
a
p
c
ont
a
m
in
a
te
s
th
e
f
le
s
h
a
nd
s
ki
n
of
th
e
f
r
ui
t.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
R
e
al
-
ti
m
e
obj
e
c
t
de
te
c
ti
on t
o c
la
s
s
if
y
e
x
por
t
qual
it
y
of
m
angos
t
e
e
n …
(
D
ia
n Sa’
adi
ll
ah M
ay
la
w
at
i)
117
B
e
c
a
us
e
it
s
ta
in
s
th
e
s
ki
n
of
th
e
f
r
ui
t,
it
c
a
us
e
s
th
e
s
ki
n
of
t
he
f
r
ui
t
to
be
c
om
e
dul
l
a
nd
una
tt
r
a
c
ti
ve
,
a
nd
c
a
us
e
s
a
bi
tt
e
r
t
a
s
te
.
T
he
pr
im
a
r
y
c
ha
ll
e
nge
in
th
e
m
a
ngos
te
e
n
e
xpor
t
c
ha
in
is
a
la
c
k
of
tr
a
ns
pa
r
e
nc
y
in
qua
li
ty
s
ta
nda
r
ds
by
f
a
r
m
e
r
s
.
M
a
ngo
s
te
e
n
f
r
ui
t
is
w
or
th
y
of
e
xpor
t
is
a
f
r
ui
t
th
a
t
pa
s
s
e
s
s
e
v
e
r
a
l
s
t
a
ge
s
of
f
a
r
m
e
r
s
or
ti
ng,
w
a
r
e
hous
e
s
or
ti
ng,
a
nd
e
xpor
te
r
s
or
ti
ng.
T
he
s
or
ti
ng
pr
oc
e
s
s
ha
s
b
e
e
n
done
m
a
nua
ll
y
by
lo
oki
ng
a
t
th
e
s
ur
f
a
c
e
of
t
he
f
r
ui
t
a
nd t
he
f
r
ui
t
s
ta
lk
. S
or
ti
ng
t
ha
t
is
done
m
a
nua
ll
y f
oc
us
e
s
on t
he
s
iz
e
, di
a
m
e
te
r
, c
ol
or
, s
ta
lk
,
te
xt
ur
e
,
a
nd
pa
tt
e
r
n
of
th
e
f
r
ui
t.
T
he
s
ur
f
a
c
e
of
th
e
f
r
ui
t
th
a
t
do
e
s
not
h
a
ve
w
hi
te
s
pot
s
(
s
now
di
s
e
a
s
e
)
,
ye
ll
ow
s
pot
s
(
ye
ll
ow
s
a
p)
,
or
e
ve
n
c
r
a
c
ki
ng
is
one
of
th
e
r
e
a
s
ons
f
or
qua
li
ty
f
r
ui
t.
I
n
a
ddi
ti
on
to
th
e
s
ur
f
a
c
e
of
th
e
f
r
ui
t,
th
e
s
ta
lk
is
a
l
s
o
one
of
th
e
be
nc
hm
a
r
k
s
f
or
th
e
qua
li
ty
of
th
e
f
r
ui
t.
O
f
c
our
s
e
,
th
is
s
or
ti
ng
pr
oc
e
s
s
ta
ke
s
lo
nge
r
be
c
a
us
e
it
h
a
s
to
be
c
he
c
k
e
d
one
by
one
on
th
e
f
r
ui
t
a
nd i
s
done
by
f
a
r
m
e
r
s
.
T
he
r
a
pi
d
de
ve
lo
pm
e
nt
of
a
r
ti
f
ic
ia
l
in
te
ll
ig
e
nc
e
te
c
hnol
ogy
f
o
r
im
a
ge
r
e
c
ogni
ti
on
c
a
n
be
us
e
d
to
he
lp
f
a
r
m
e
r
s
in
th
e
pr
oc
e
s
s
of
s
or
ti
ng
m
a
ngos
te
e
n
f
r
ui
t
a
ut
om
a
ti
c
a
ll
y.
T
he
im
a
ge
r
e
c
ogni
ti
on
m
ode
l
c
a
n
c
e
r
ta
in
ly
be
e
m
be
dde
d
in
va
r
io
us
de
vi
c
e
s
th
a
t
he
lp
th
e
s
or
ti
ng
pr
oc
e
s
s
m
or
e
e
f
f
ic
ie
nt
ly
.
I
m
a
ge
r
e
c
ogni
ti
on
is
a
te
c
hnol
ogy
us
e
d
in
v
a
r
io
us
in
dus
tr
ie
s
s
uc
h
a
s
he
a
lt
h,
s
e
c
ur
it
y,
s
a
le
s
,
a
nd
te
c
hnol
ogy
to
r
e
c
ogni
z
e
o
bj
e
c
ts
,
pe
opl
e
,
w
r
it
in
g,
a
nd
a
c
ti
ons
in
di
gi
ta
l
im
a
ge
s
[
8]
.
P
r
e
vi
ous
r
e
s
e
a
r
c
h
ha
s
pr
ove
n
th
e
r
ol
e
of
a
r
ti
f
ic
ia
l
in
te
ll
ig
e
nc
e
in
im
a
ge
r
e
c
ogni
ti
on
in
va
r
io
us
f
ie
ld
s
,
s
uc
h
a
s
he
a
lt
hc
a
r
e
a
ppl
ic
a
ti
ons
f
or
m
e
di
c
a
l
im
a
ge
a
na
l
ys
is
th
a
t
im
pr
ove
th
e
s
pe
e
d
a
nd
a
c
c
ur
a
c
y
of
in
te
r
pr
e
ti
ng
m
e
di
c
a
l
im
a
ge
s
,
a
id
in
g
in
di
a
gno
s
ti
c
s
a
nd
tr
e
a
tm
e
n
t
pl
a
nni
ng
[
9]
.
A
de
c
is
io
n
s
uppor
t
s
ys
te
m
th
a
t
pr
ovi
de
s
c
om
pr
e
he
ns
iv
e
in
f
or
m
a
ti
on
to
he
a
lt
hc
a
r
e
pr
o
f
e
s
s
io
na
ls
,
e
nha
nc
in
g
de
c
is
io
n
-
m
a
ki
ng
c
a
pa
bi
li
ti
e
s
[
9]
.
T
he
n,
in
in
te
ll
ig
e
nt
dr
iv
in
g
s
y
s
te
m
s
(
a
ut
onomou
s
ve
hi
c
le
s
)
,
im
a
ge
r
e
c
ogni
ti
on
e
na
bl
e
s
ve
hi
c
l
e
s
to
und
e
r
s
ta
nd
th
e
ir
s
ur
r
oundings
in
r
e
a
l
-
ti
m
e
,
w
hi
c
h
is
c
r
uc
ia
l
f
or
s
a
f
e
ty
a
nd
a
ut
om
a
ti
on
in
in
te
ll
ig
e
nt
dr
iv
in
g
s
ys
te
m
s
.
I
n
th
e
s
e
c
ur
it
y s
e
c
to
r
s
, f
a
c
ia
l
r
e
c
ogni
ti
on a
nd obje
c
t
de
t
e
c
ti
on c
a
n
e
ns
ur
e
a
nd e
nha
nc
e
m
oni
to
r
in
g a
nd s
a
f
e
ty
[
9]
.
I
n
r
e
ta
il
,
im
a
ge
r
e
c
ogni
ti
on
s
tr
e
a
m
li
ne
s
in
ve
nt
or
y
m
a
na
ge
m
e
nt
a
nd
e
nh
a
nc
e
s
c
u
s
to
m
e
r
e
xpe
r
ie
nc
e
by
a
c
c
ur
a
te
ly
i
de
nt
if
yi
ng pr
oduc
ts
, e
ve
n w
it
h s
ubt
le
pa
c
ka
gi
ng dif
f
e
r
e
nc
e
s
[
10]
.
I
n
a
gr
ic
ul
tu
r
e
,
im
a
ge
r
e
c
ogni
ti
on
ha
s
e
m
e
r
ge
d
a
s
a
br
e
a
kt
hr
ough
te
c
hni
que
th
a
t
im
pr
ove
s
c
r
op
m
a
na
ge
m
e
nt
a
nd
di
s
e
a
s
e
de
te
c
ti
on.
F
a
r
m
e
r
s
m
a
y
us
e
a
dva
nc
e
d
a
lg
or
it
hm
s
to
id
e
nt
if
y
pl
a
nt
i
ll
ne
s
s
e
s
,
c
la
s
s
if
y
s
oi
l
ty
pe
s
,
a
nd
m
oni
to
r
in
s
e
c
t
in
f
e
s
ta
ti
ons
,
th
e
r
e
by
in
c
r
e
a
s
in
g
a
gr
ic
ul
tu
r
a
l
out
put
a
nd
s
us
ta
in
a
bi
li
ty
.
V
a
r
io
us
im
a
ge
pr
oc
e
s
s
in
g
te
c
hni
que
s
,
in
c
lu
di
ng
c
onvolut
io
na
l
ne
ur
a
l
ne
twor
ks
(
C
N
N
s
)
a
nd
s
uppor
t
ve
c
to
r
m
a
c
hi
ne
s
(
S
V
M
s
)
,
ha
ve
s
how
n
hi
gh
a
c
c
ur
a
c
y
in
id
e
nt
if
yi
ng
c
r
op
di
s
e
a
s
e
s
[
11]
–
[
14]
.
I
m
a
ge
r
e
c
ogni
ti
on
to
o
ls
c
a
n
c
la
s
s
if
y
s
oi
l
ty
pe
s
a
nd
r
e
c
om
m
e
nd
s
ui
ta
bl
e
c
r
ops
,
c
ul
ti
va
ti
on
s
,
a
nd
f
e
r
ti
li
z
e
r
s
ba
s
e
d
on
pa
r
a
m
e
te
r
s
li
k
e
s
oi
l
c
ol
or
a
nd l
oc
a
ti
on
[
15]
. T
he
n, hybr
id
m
ode
ls
c
om
bi
ni
ng l
ogi
s
ti
c
r
e
gr
e
s
s
io
n w
it
h de
c
is
io
n t
r
e
e
s
ha
ve
a
ls
o b
e
e
n
e
f
f
e
c
ti
ve
in
im
pr
ovi
ng
r
e
c
ogni
ti
on
a
c
c
ur
a
c
y
[
12]
.
I
n
ot
he
r
r
e
s
e
a
r
c
h,
a
ut
om
a
t
e
d
s
y
s
te
m
s
ut
il
iz
in
g
im
a
ge
r
e
c
ogni
ti
on
c
a
n
di
a
gnos
e
pe
s
t
s
a
nd
di
s
e
a
s
e
s
,
f
a
c
il
it
a
ti
ng
pr
e
c
i
s
io
n
m
oni
to
r
in
g
a
nd
r
e
duc
in
g
pe
s
ti
c
id
e
us
a
ge
[
16]
.
D
e
e
p
le
a
r
ni
ng
a
lg
or
it
h
m
s
,
s
uc
h
a
s
you
onl
y
lo
ok
on
c
e
(
Y
O
L
O
)
,
ha
ve
be
e
n
a
ppl
ie
d
to
va
r
io
us
a
gr
ic
ul
tu
r
a
l
s
c
e
na
r
io
s
, de
m
ons
tr
a
ti
ng t
he
ir
ve
r
s
a
ti
li
ty
a
nd e
f
f
e
c
t
iv
e
ne
s
s
[
17]
.
Y
O
L
O
ha
s
a
pe
r
f
o
r
m
a
nc
e
t
ha
t
is
w
o
r
t
h
c
ons
id
e
r
in
g
f
o
r
o
bj
e
c
t
de
te
c
t
io
n
.
Y
O
L
O
is
ou
tp
e
r
f
o
r
m
c
om
p
a
r
in
g
w
i
th
s
in
gl
e
s
ho
t
de
te
c
t
io
n
(
S
S
D
)
a
n
d
f
a
s
te
r
r
e
gi
o
n
-
ba
s
e
d
c
o
nvo
lu
ti
ona
l
ne
ur
a
l
ne
tw
o
r
ks
(
F
a
s
te
r
R
-
C
N
N
)
i
n
a
n
id
e
nt
ic
a
l
te
s
ti
ng
e
n
vi
r
onm
e
n
t
[
18
]
.
O
th
e
r
a
ppl
ie
d
s
tu
di
e
s
r
e
po
r
t
Y
O
L
O
v
8
y
ie
l
ds
hi
g
he
r
in
f
e
r
e
n
c
e
s
pe
e
d
a
nd
c
om
pe
t
it
iv
e
o
r
s
u
pe
r
io
r
m
e
a
n
a
v
e
r
a
g
e
pr
e
c
is
io
n
(
m
A
P
)
c
om
p
a
r
e
d
w
i
th
S
S
D
,
F
a
s
te
r
R
-
C
N
N
,
a
nd
e
f
f
i
c
ie
nt
o
bj
e
c
t
de
t
e
c
t
io
n
(
E
f
f
ic
ie
nt
D
e
t
)
i
n
t
r
a
f
f
ic
a
n
d
s
ur
ve
i
ll
a
nc
e
ta
s
ks
[
1
9
]
,
[
20
]
.
I
ts
opt
im
iz
a
ti
on
a
ll
o
w
s
f
or
de
p
lo
ym
e
nt
i
n
a
g
r
ic
u
lt
ur
a
l
s
ys
te
m
s
,
e
na
bl
in
g
r
e
a
l
-
t
im
e
g
r
a
d
in
g
a
nd
s
o
r
ti
ng
w
it
hou
t
hi
g
h
-
e
nd
ha
r
dw
a
r
e
.
T
h
e
r
e
f
or
e
,
Y
O
L
O
v8
c
a
n
of
f
e
r
t
he
op
ti
m
a
l
s
pe
e
d,
de
te
c
ti
on
a
c
c
ur
a
c
y
, a
nd
de
pl
oya
bi
li
ty
f
o
r
p
r
a
c
ti
c
a
l
a
pp
li
c
a
t
io
ns
s
uc
h
a
s
m
a
ng
os
te
e
n
qua
li
ty
in
s
pe
c
ti
o
n.
O
ne
of
t
he
i
m
a
g
e
r
e
c
o
gni
ti
on
m
e
th
o
ds
th
a
t
s
ho
w
s
go
od
r
e
s
u
lt
s
f
r
om
va
r
io
us
s
t
ud
ie
s
is
d
e
e
p
l
e
a
r
ni
n
g,
w
h
ic
h
a
c
c
e
l
e
r
a
te
s
th
e
le
a
r
ni
ng
pr
oc
e
s
s
i
n
ne
u
r
a
l
n
e
tw
o
r
ks
w
i
th
m
a
ny
la
ye
r
s
o
f
c
o
m
pu
ta
ti
on.
O
ne
o
f
th
e
m
a
ny
d
e
e
p
l
e
a
r
ni
ng
a
l
go
r
it
h
m
s
is
t
he
C
N
N
a
lg
o
r
it
hm
o
n
th
e
Y
O
L
O
a
r
c
hi
te
c
t
u
r
e
[
21
]
.
Y
O
L
O
'
s
a
r
c
h
it
e
c
tu
r
e
e
na
bl
e
s
i
t
to
f
o
r
e
c
a
s
t
bo
und
in
g
bo
xe
s
a
nd
c
la
s
s
p
r
o
ba
b
il
it
ie
s
s
t
r
a
ig
ht
f
r
om
ph
ot
os
,
m
a
ki
ng
it
id
e
a
l
f
o
r
a
va
r
ie
t
y
o
f
a
p
pl
ic
a
ti
ons
s
uc
h
a
s
s
e
r
v
ic
e
r
o
bot
s
,
r
e
m
ot
e
s
e
ns
i
ng,
a
nd
li
v
e
ob
je
c
t
r
e
c
og
ni
ti
o
n.
V
a
r
ia
n
ts
s
uc
h
a
s
Y
O
L
O
v4
,
Y
O
L
O
v5
,
a
nd
Y
O
L
O
v
8
ha
v
e
be
e
n
tu
ne
d
f
or
pe
r
f
o
r
m
a
nc
e
,
w
i
th
Y
O
L
O
v8
de
m
ons
t
r
a
t
in
g
s
i
gni
f
ic
a
n
t
in
c
r
e
a
s
e
s
in
de
te
c
t
io
n
pr
e
c
is
io
n,
pa
r
ti
c
ul
a
r
ly
f
o
r
s
m
a
ll
ta
r
ge
ts
[
22
]
.
B
a
s
e
d
o
n
Y
O
L
O
v8
’
s
a
nc
ho
r
‑
f
r
e
e
de
s
ig
n,
u
pda
te
d
ba
c
k
bo
ne
o
r
ne
c
k,
m
ode
r
n
lo
s
s
a
nd
t
r
a
in
i
ng
s
c
he
m
e
p
r
od
uc
e
i
m
p
r
ov
e
d
m
A
P
a
nd
in
f
e
r
e
nc
e
e
f
f
ic
ie
nc
y
ve
r
s
us
m
a
n
y
e
a
r
li
e
r
Y
O
L
O
r
e
le
a
s
e
s
(
s
uc
h
a
s
Y
O
L
O
v
5)
,
a
n
d
s
t
udi
e
s
s
pe
c
if
ic
a
ll
y
r
e
p
or
t
Y
O
L
O
v8
ou
tp
e
r
f
o
r
m
in
g
Y
O
L
O
v5
in
tr
a
f
f
ic
s
i
gn
a
n
d ge
ne
r
a
l
de
te
c
ti
on
te
s
ts
[
23
]
–
[
25
]
.
A
m
ong the
us
e
s
of
de
e
p l
e
a
r
ni
ng i
n a
gr
ic
ul
tu
r
a
l
i
m
a
ge
r
e
c
ogni
t
io
n r
e
s
e
a
r
c
h, t
he
r
e
i
s
l
im
it
e
d r
e
s
e
a
r
c
h
th
a
t
ha
s
f
oc
us
e
d
on
a
ppl
yi
ng
a
dva
nc
e
d
obj
e
c
t
de
te
c
ti
on
m
o
de
ls
to
m
a
ngos
te
e
n
e
xpor
t
qua
li
ty
.
E
xi
s
ti
ng
s
tu
di
e
s
ha
ve
m
os
tl
y
us
e
d
tr
a
di
ti
ona
l
im
a
ge
pr
oc
e
s
s
in
g
or
c
la
s
s
ic
a
l
m
a
c
hi
ne
le
a
r
ni
ng
m
e
th
ods
.
B
y
a
ppl
yi
ng
Y
O
L
O
v8
to
th
is
nove
l
a
gr
ic
ul
tu
r
a
l
c
ont
e
xt
,
th
is
r
e
s
e
a
r
c
h
not
onl
y
c
ont
r
ib
ut
e
s
to
a
dva
nc
in
g
pr
e
c
is
io
n
a
gr
ic
ul
tu
r
e
in
I
ndone
s
ia
but
a
ls
o
pr
ovi
de
s
a
tr
a
ns
f
e
r
a
bl
e
m
od
e
l
f
or
ot
he
r
e
xpor
t
c
om
m
odi
ti
e
s
w
it
h
s
im
il
a
r
gr
a
di
ng c
ha
ll
e
nge
s
. T
h
e
r
e
f
or
e
, t
hi
s
s
tu
dy a
im
s
t
o a
s
s
is
t
s
or
ti
ng i
n t
he
m
a
ngos
te
e
n i
ndus
tr
y by us
in
g one
of
t
he
de
e
p
le
a
r
ni
ng
m
e
th
ods
,
na
m
e
ly
im
a
ge
r
e
c
ogni
ti
on
us
in
g
Y
O
L
O
v8,
a
s
w
e
ll
a
s
to
d
e
te
r
m
in
e
th
e
pe
r
f
or
m
a
nc
e
of
t
he
m
ode
l
bui
lt
w
it
h va
r
io
us
Y
O
L
O
v8 va
r
ia
nt
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
, V
ol
.
15
, N
o.
1
,
F
e
br
ua
r
y
20
26
:
116
-
128
118
2.
M
E
T
H
O
D
2.1. Re
s
e
ar
c
h
ac
t
iv
it
y
T
hi
s
r
e
s
e
a
r
c
h a
d
a
pt
s
t
he
c
r
os
s
-
in
dus
tr
y s
ta
nda
r
d pr
oc
e
s
s
f
or
da
ta
m
in
in
g
(
C
R
I
S
P
-
DM
)
m
e
th
odol
ogy,
w
hi
c
h
is
w
id
e
ly
us
e
d
in
da
ta
m
in
in
g
a
nd
m
a
c
hi
n
e
le
a
r
ni
ng
pr
oj
e
c
ts
.
T
hi
s
m
e
th
odol
ogy
ha
s
s
ix
m
a
in
s
ta
ge
s
,
na
m
e
ly
bus
in
e
s
s
unde
r
s
ta
ndi
ng,
da
ta
unde
r
s
ta
ndi
ng,
da
t
a
pr
e
pa
r
a
ti
on,
m
ode
li
ng,
e
va
lu
a
ti
on,
a
nd
de
pl
oym
e
nt
.
F
ig
ur
e
1 s
how
s
r
e
s
e
a
r
c
h a
c
ti
vi
ti
e
s
t
ha
t
a
da
pt
t
he
s
ta
ge
s
of
C
R
I
S
P
-
D
M
. A
t
th
e
bus
in
e
s
s
unde
r
s
ta
ndi
ng s
ta
ge
, i
t
w
a
s
f
ound
th
a
t
th
e
pr
oc
e
s
s
of
s
or
ti
ng
e
xpor
t
-
qua
li
ty
m
a
ngos
te
e
ns
to
ok
qui
te
a
lo
ng
ti
m
e
due
to
th
e
li
m
it
e
d
num
be
r
of
f
a
r
m
e
r
s
.
T
he
r
e
f
or
e
,
by
ut
il
iz
in
g
a
r
ti
f
ic
ia
l
in
te
l
li
ge
nc
e
te
c
hnol
ogy,
th
e
pr
oc
e
s
s
of
s
or
ti
ng
e
xpor
t
-
w
or
th
y
m
a
ngos
te
e
ns
c
a
n
be
m
or
e
e
f
f
ic
ie
nt
w
it
h
im
a
ge
r
e
c
ogni
ti
on
te
c
hni
que
s
.
F
ur
th
e
r
m
or
e
,
a
t
th
e
da
ta
unde
r
s
ta
ndi
ng s
ta
ge
, t
hi
s
s
tu
dy r
e
qui
r
e
s
i
nf
or
m
a
ti
on on the
c
ha
r
a
c
te
r
is
ti
c
s
of
e
xpor
t
-
w
or
th
y m
a
ngos
te
e
ns
a
nd c
ol
le
c
ts
m
a
ngos
te
e
n da
ta
f
r
om
m
a
ngos
te
e
n pl
a
nt
a
ti
ons
a
nd s
or
ti
ng pla
c
e
s
f
or
e
xpor
t.
T
he
da
ta
ne
e
de
d a
r
e
im
a
ge
s
of
m
a
ngo
s
te
e
n
s
in
di
vi
dua
ll
y
or
m
a
ny
m
a
ngos
t
e
e
ns
in
one
im
a
ge
.
I
n
a
ddi
ti
on,
th
is
s
tu
dy
a
ls
o
us
e
s
vi
de
o da
ta
t
o t
e
s
t
th
e
r
e
a
l
-
ti
m
e
de
te
c
ti
on mode
l.
F
ur
th
e
r
m
or
e
,
f
o
r
th
e
m
ode
li
ng
pr
oc
e
s
s
,
15
e
xpe
r
im
e
nt
s
c
e
na
r
io
s
w
e
r
e
de
s
ig
ne
d
ba
s
e
d
on
a
c
om
bi
na
ti
on
of
th
r
e
e
da
ta
s
pl
it
ti
ng
r
a
ti
os
f
or
tr
a
in
in
g,
va
li
da
ti
on,
a
nd
te
s
ti
ng,
na
m
e
ly
80:
10:
10,
70:
20:
10,
a
nd
60:
20:
20,
w
it
h
a
c
om
bi
na
ti
on
o
f
5
a
va
il
a
bl
e
Y
O
L
O
v8
a
r
c
hi
te
c
tu
r
e
va
r
ia
nt
s
,
na
m
e
ly
na
no,
s
m
a
ll
,
m
e
di
um
,
la
r
ge
,
a
nd
e
xt
r
a
l
a
r
ge
.
T
h
e
n,
th
e
e
va
lu
a
ti
on
w
a
s
c
onduc
te
d
us
in
g
a
c
onf
us
io
n
m
a
tr
ix
.
L
a
s
tl
y,
a
s
im
pl
e
de
pl
oym
e
nt
pr
oc
e
s
s
is
c
a
r
r
ie
d
out
by
im
pl
e
m
e
nt
in
g
th
e
be
s
t
m
ode
l
f
r
om
th
e
m
ode
l
de
ve
lo
pm
e
nt
r
e
s
ul
ts
in
to
th
e
m
a
ngos
te
e
n
d
e
te
c
ti
on
a
ppl
ic
a
ti
on. T
he
r
e
s
ul
t
s
of
th
e
de
e
p
l
e
a
r
ni
ng
m
ode
l
f
or
s
or
ti
ng
th
e
e
xpor
t
e
li
gi
bi
li
ty
of
m
a
ngos
te
e
n
us
in
g
Y
O
L
O
v8
ha
ve
gr
e
a
t
pot
e
nt
ia
l
to
be
im
pl
e
m
e
nt
e
d
in
th
e
f
ut
ur
e
.
T
hi
s
m
ode
l
c
a
n
be
a
ppl
ie
d
to
be
tt
e
r
te
c
hnol
ogy
-
ba
s
e
d
s
y
s
te
m
s
,
s
uc
h
a
s
w
e
b
a
ppl
ic
a
ti
ons
,
m
obi
le
de
vi
c
e
s
,
or
in
te
r
ne
t
of
th
in
g
(
I
oT
)
-
ba
s
e
d
a
ut
om
a
te
d
s
y
s
te
m
s
to
a
s
s
i
s
t
th
e
f
r
ui
t
s
or
ti
ng
pr
oc
e
s
s
in
r
e
a
l
ti
m
e
.
B
y
in
te
gr
a
ti
ng
th
e
m
ode
l
in
to
th
e
pr
oduc
ti
on
s
ys
te
m
,
f
a
r
m
e
r
s
a
nd
e
xpor
te
r
s
c
a
n
in
c
r
e
a
s
e
th
e
e
f
f
ic
ie
nc
y
a
nd
a
c
c
ur
a
c
y
of
th
e
s
or
ti
ng
pr
oc
e
s
s
,
th
e
r
e
by i
nc
r
e
a
s
in
g t
he
c
ha
n
c
e
s
of
I
ndone
s
ia
n m
a
ngos
te
e
n
s
m
e
e
t
in
g e
xpor
t
s
ta
nda
r
ds
.
F
ig
ur
e
1. C
R
I
P
S
-
D
M
a
dopt
io
n f
or
r
e
s
e
a
r
c
h a
c
ti
vi
ti
e
s
2.2. M
an
gos
t
e
e
n
d
at
as
e
t
s
E
xpor
t
-
w
or
th
y
m
a
ngos
te
e
n
f
r
ui
t
c
a
n
be
id
e
nt
if
ie
d
f
r
om
th
e
s
ur
f
a
c
e
or
s
ki
n
of
th
e
f
r
ui
t,
c
ol
or
,
s
ta
lk
,
a
nd
di
s
e
a
s
e
.
T
h
e
s
ur
f
a
c
e
w
il
l
gr
e
a
tl
y
a
f
f
e
c
t
th
e
s
ta
nd
a
r
d
of
e
li
gi
bi
li
ty
of
m
a
ngos
te
e
n
f
r
ui
t,
a
s
w
e
ll
a
s
th
e
c
ol
or
to
m
e
a
s
ur
e
t
he
r
ip
e
ne
s
s
of
t
he
m
a
ngos
te
e
n f
r
ui
t.
T
he
n, t
he
qua
li
ty
of
t
he
s
ta
lk
w
il
l
a
lwa
ys
a
f
f
e
c
t
th
e
qua
li
ty
o
f
th
e
m
a
ngos
te
e
n
f
r
ui
t
[
26]
.
I
n
a
ddi
ti
on,
m
a
ngos
te
e
n
f
r
ui
t
th
a
t
h
a
s
a
di
s
e
a
s
e
(
th
e
r
e
a
r
e
s
a
p
or
ye
ll
ow
s
pot
s
on
th
e
s
ur
f
a
c
e
)
i
s
c
a
te
gor
iz
e
d
a
s
m
a
ngo
s
te
e
n t
ha
t
is
not
w
or
th
y of
e
xpor
t.
T
hi
s
s
tu
dy
ha
s
two
ty
pe
s
of
da
ta
,
di
vi
de
d
in
to
pr
im
a
r
y
a
nd
s
e
c
onda
r
y
da
ta
.
D
a
ta
in
th
e
f
or
m
of
m
a
ngos
te
e
n
f
r
ui
t
im
a
ge
s
,
e
it
he
r
one
m
a
ngos
te
e
n
or
m
a
ny
m
a
ngos
te
e
ns
.
F
ig
ur
e
2
s
how
s
a
n
e
xa
m
pl
e
of
a
m
a
ngos
te
e
n
im
a
ge
u
s
e
d
in
th
is
s
tu
dy.
P
r
im
a
r
y
da
ta
w
a
s
c
ol
le
c
te
d
f
r
om
m
a
ngos
te
e
n
pl
a
nt
a
ti
ons
a
nd
s
or
ti
ng
lo
c
a
ti
ons
in
th
e
M
a
t
a
r
a
m
a
r
e
a
,
W
e
s
t
N
us
a
T
e
ngga
r
a
,
on
e
of
I
ndone
s
ia
'
s
l
a
r
ge
s
t
m
a
ngo
s
te
e
n
pr
oduc
e
r
s
a
nd
s
or
ti
ng
c
e
nt
e
r
s
.
P
r
im
a
r
y
da
ta
w
a
s
c
ol
le
c
te
d
on
D
e
c
e
m
be
r
18,
2023,
D
e
c
e
m
be
r
6,
2024,
a
nd
D
e
c
e
m
be
r
18,
2024,
us
in
g
th
e
iP
hone
X
R
,
iP
hone
X
S
M
a
x,
R
e
dm
i
N
ot
e
12
P
r
o,
a
nd
S
a
m
s
ung
A
35
de
vi
c
e
s
.
T
hi
s
de
vi
c
e
c
e
r
ta
in
ly
a
f
f
e
c
ts
t
he
l
ig
ht
in
g a
nd i
m
a
ge
qua
li
ty
.
T
he
s
e
c
onda
r
y
d
a
ta
s
e
t
w
a
s
obt
a
in
e
d
f
r
om
th
e
in
te
r
ne
t
w
it
h
e
xi
s
ti
ng
s
e
a
r
c
h
e
ngi
n
e
s
,
s
uc
h
a
s
G
oogl
e
a
nd
K
a
ggl
e
a
t
ht
tp
s
:/
/
w
w
w
.
ka
gg
le
.
c
om
/d
a
ta
s
e
t
s
/m
a
r
ia
m
e
r
e
s
/d
e
e
p
-
le
a
r
ni
ng
-
kl
a
s
if
i
ka
s
i
-
je
ni
s
-
bu
a
h
-
m
a
nggi
s
,
to
in
c
r
e
a
s
e
d
a
ta
va
r
i
a
ti
on
a
n
d
b
e
a
bl
e
to
r
e
c
ogni
z
e
va
r
io
u
s
c
on
d
it
io
ns
or
a
pp
e
a
r
a
nc
e
s
of
m
a
ngo
s
te
e
n,
s
uc
h
a
s
li
ght
in
g
va
r
i
a
ti
on
s
,
s
ho
ot
in
g
a
ngl
e
s
,
a
nd
im
a
ge
qu
a
li
ty
,
us
in
g
m
a
ngo
s
te
e
n
im
a
g
e
s
ta
k
e
n
f
r
om
va
r
io
us
s
o
ur
c
e
s
on
th
e
in
t
e
r
ne
t
.
T
he
c
o
m
bi
na
ti
on
of
th
e
s
e
t
w
o
d
a
ta
s
e
ts
w
a
s
c
a
r
r
ie
d
o
ut
to
e
nr
i
c
h
th
e
m
ode
l
in
r
e
c
ogni
z
in
g
a
nd
c
la
s
s
if
y
in
g
m
a
ngo
s
te
e
n
f
r
ui
t
c
or
r
e
c
tl
y
i
n
va
r
i
ous
c
o
ndi
ti
on
s
. T
h
e
m
a
n
gos
t
e
e
n
im
a
ge
d
a
ta
th
a
t
w
a
s
s
uc
c
e
s
s
f
ul
ly
c
ol
le
c
te
d
i
n t
h
is
s
tu
d
y w
e
r
e
2
,717
pr
im
a
r
y
da
t
a
(
ht
tp
s
:/
/
s
.i
d/
m
a
n
gos
t
e
e
n_
da
t
a
s
e
t)
, a
nd 1,
297
s
e
c
ond
a
r
y d
a
ta
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
R
e
al
-
ti
m
e
obj
e
c
t
de
te
c
ti
on t
o c
la
s
s
if
y
e
x
por
t
qual
it
y
of
m
angos
t
e
e
n …
(
D
ia
n Sa’
adi
ll
ah M
ay
la
w
at
i)
119
F
ig
ur
e
2. T
he
pr
im
a
r
y m
a
ngos
te
e
n da
ta
s
e
ts
f
r
om
pl
a
nt
a
ti
on f
ie
ld
2.3. You
on
ly
l
ook
on
c
e
v
e
r
s
io
n
8
I
m
a
ge
id
e
nt
if
ic
a
ti
on
ut
il
iz
in
g
th
e
Y
O
L
O
a
lg
or
it
hm
ha
s
ga
in
e
d
popula
r
it
y
due
to
it
s
e
f
f
ic
ie
nc
y
a
nd
a
c
c
ur
a
c
y
in
r
e
a
l
-
ti
m
e
it
e
m
de
te
c
ti
on.
Y
O
L
O
'
s
a
r
c
hi
te
c
tu
r
e
e
n
a
bl
e
s
it
to
f
or
e
c
a
s
t
bounding
boxe
s
a
nd
c
la
s
s
pr
oba
bi
li
ti
e
s
s
tr
a
ig
ht
f
r
om
pho
to
s
,
m
a
ki
ng
it
id
e
a
l
f
or
a
va
r
ie
ty
of
a
ppl
ic
a
ti
ons
s
uc
h
a
s
s
e
r
vi
c
e
r
obot
s
,
r
e
m
ot
e
s
e
ns
in
g,
a
nd
li
ve
obj
e
c
t
r
e
c
ogni
ti
on.
Y
O
L
O
th
r
iv
e
s
in
r
e
a
l
-
ti
m
e
a
ppl
ic
a
ti
ons
,
w
it
h
hi
gh
a
c
c
ur
a
c
y
r
a
te
s
a
nd
lo
w
la
te
nc
y,
w
hi
c
h
is
e
s
s
e
nt
ia
l
f
or
a
c
ti
vi
ti
e
s
th
a
t
r
e
qui
r
e
qui
c
k
f
e
e
dba
c
k.
V
a
r
ia
nt
s
s
uc
h
a
s
Y
O
L
O
v4,
Y
O
L
O
v5, a
nd Y
O
L
O
v8 ha
ve
be
e
n t
une
d f
or
pe
r
f
o
r
m
a
nc
e
, w
it
h Y
O
L
O
v8 de
m
ons
tr
a
ti
ng s
ig
ni
f
ic
a
nt
i
nc
r
e
a
s
e
s
in
de
te
c
ti
on
pr
e
c
is
io
n,
pa
r
ti
c
ul
a
r
ly
f
or
s
m
a
ll
ta
r
ge
ts
[
22]
.
Y
O
L
O
im
pr
ove
s
vi
s
ua
l
id
e
nt
if
ic
a
ti
on
c
a
pa
c
it
ie
s
in
c
om
pl
e
x
s
it
ua
ti
ons
[
27]
a
nd
unde
r
di
f
f
e
r
e
nt
c
ondi
ti
ons
[
28
]
.
D
e
s
pi
te
it
s
be
ne
f
it
s
,
Y
O
L
O
s
tr
uggl
e
s
w
it
h
s
m
a
ll
obj
e
c
t
de
te
c
ti
on a
nd mul
ti
-
obj
e
c
t
s
e
tt
in
gs
. T
o
s
ol
ve
t
he
s
e
di
f
f
ic
ul
ti
e
s
, i
nnova
ti
ons
l
ik
e
m
odi
f
ie
d l
os
s
f
unc
ti
ons
a
nd
m
ul
ti
-
s
c
a
le
tr
a
in
in
g
ha
ve
be
e
n
de
ve
lo
pe
d
[
28]
.
T
he
us
e
of
a
dva
nc
e
d
a
ppr
oa
c
he
s
,
s
uc
h
a
s
nor
m
a
li
z
e
d
W
a
s
s
e
r
s
te
in
di
s
ta
nc
e
l
os
s
, h
a
s
i
nc
r
e
a
s
e
d de
te
c
ti
on a
c
c
ur
a
c
y i
n
s
pe
c
ia
li
z
e
d a
ppl
ic
a
ti
on
s
[
22]
.
Y
O
L
O
us
e
s
th
e
C
N
N
m
e
th
od,
w
hi
c
h
is
w
id
e
ly
a
ppl
ie
d
to
im
a
ge
da
ta
,
a
m
e
th
od
f
or
de
te
c
ti
ng
obj
e
c
ts
.
Y
O
L
O
pr
oc
e
s
s
e
s
im
a
ge
s
in
r
e
a
l
-
ti
m
e
a
t
f
or
ty
-
f
iv
e
(
4
5)
f
r
a
m
e
s
pe
r
s
e
c
ond
[
29]
–
[
32]
.
T
he
Y
O
L
O
a
r
c
hi
te
c
tu
r
e
a
s
s
how
n
in
F
ig
ur
e
3,
in
c
lu
di
ng
th
e
Y
O
L
O
v8
ve
r
s
io
n,
c
ons
i
s
ts
of
th
r
e
e
m
a
in
c
om
pone
nt
s
:
ba
c
kbone
,
ne
c
k,
a
nd
he
a
d,
w
hi
c
h
w
or
k
s
yne
r
gi
s
ti
c
a
ll
y
to
d
e
te
c
t
a
nd
c
la
s
s
if
y
obj
e
c
ts
.
I
n
Y
O
L
O
v8,
th
e
ba
c
kbone
us
e
s
th
e
c
r
os
s
-
s
ta
ge
p
a
r
ti
a
l
da
r
kne
t
(
C
S
P
D
a
r
kne
t)
a
r
c
hi
te
c
tu
r
e
de
s
ig
ne
d
to
e
f
f
ic
ie
nt
ly
c
a
pt
ur
e
pa
tt
e
r
ns
a
nd
vi
s
ua
l
c
ha
r
a
c
te
r
is
ti
c
s
of
im
a
ge
s
.
C
S
P
D
a
r
kne
t
ut
il
iz
e
s
r
e
s
id
ua
l
c
onne
c
ti
ons
to
im
pr
ove
f
e
a
tu
r
e
le
a
r
ni
ng
w
it
hout
lo
s
in
g
in
f
or
m
a
ti
on
du
r
in
g
da
ta
pr
opa
ga
ti
on.
T
hr
ough
s
e
ve
r
a
l
la
ye
r
s
of
c
onvolut
io
n
a
nd
a
c
ti
va
ti
on
f
unc
ti
ons
,
th
e
b
a
c
kbone
pr
oduc
e
s
a
r
ic
h
f
e
a
tu
r
e
r
e
pr
e
s
e
nt
a
ti
on,
w
hi
c
h
i
s
th
e
n
f
or
w
a
r
de
d
to
th
e
ne
xt
s
e
c
ti
on.
T
he
ne
c
k
is
a
c
om
pon
e
nt
th
a
t
c
onn
e
c
ts
th
e
ba
c
kbon
e
to
th
e
he
a
d
a
nd
is
r
e
s
pons
ib
le
f
or
c
om
bi
ni
ng
f
e
a
tu
r
e
s
f
r
om
va
r
io
us
s
c
a
le
s
(
m
ul
ti
s
c
a
le
f
e
a
tu
r
e
a
ggr
e
ga
ti
on)
. I
n Y
O
L
O
v8, t
he
ne
c
k us
e
s
t
he
pa
th
a
ggr
e
ga
ti
on
ne
twor
k
(
P
A
N
e
t)
m
e
c
ha
ni
s
m
,
w
hi
c
h
im
pr
ove
s
th
e
f
us
io
n
of
f
e
a
tu
r
e
s
f
r
om
de
e
pe
r
a
nd
s
ha
ll
ow
e
r
la
ye
r
s
.
P
A
N
e
t
he
lp
s
th
e
m
ode
l
unde
r
s
ta
nd
th
e
c
ont
e
xt
o
f
obj
e
c
ts
of
d
if
f
e
r
e
nt
s
iz
e
s
,
th
us
in
c
r
e
a
s
in
g
th
e
a
c
c
ur
a
c
y
in
de
te
c
ti
ng
s
m
a
ll
a
nd
la
r
ge
obj
e
c
ts
.
M
e
a
nw
hi
le
,
th
e
he
a
d
i
s
th
e
la
s
t
pa
r
t
of
th
e
Y
O
L
O
a
r
c
hi
te
c
tu
r
e
th
a
t
i
s
r
e
s
pons
ib
le
f
or
m
a
ki
ng
pr
e
di
c
ti
ons
,
na
m
e
ly
de
te
r
m
in
in
g
th
e
pr
e
s
e
nc
e
of
obj
e
c
ts
,
obj
e
c
t
ty
pe
s
,
a
nd
th
e
pos
it
io
n
(
bounding
box)
of
th
e
obj
e
c
t
in
th
e
im
a
ge
.
Y
O
L
O
v8
us
e
s
a
de
c
oupl
e
d
he
a
d
th
a
t
s
e
p
a
r
a
te
s
th
e
ta
s
ks
of
obj
e
c
t
lo
c
a
ti
on
r
e
gr
e
s
s
io
n
(
bounding
box
r
e
gr
e
s
s
io
n)
a
nd
obj
e
c
t
la
be
l
c
la
s
s
if
ic
a
ti
on.
T
a
bl
e
1
s
how
s
th
e
c
ha
r
a
c
te
r
is
ti
c
s
of
e
a
c
h Y
O
L
O
v8 va
r
ia
ti
on t
ha
t
w
il
l
be
us
e
d i
n t
hi
s
s
tu
dy
[
33]
.
T
a
bl
e
1. C
h
a
r
a
c
te
r
is
ti
c
s
of
Y
O
L
O
v8
V
a
r
i
a
nt
M
ode
l
s
i
z
e
N
um
be
r
of
pa
r
a
m
e
t
e
r
s
(
m
i
l
l
i
on)
S
pe
e
d
Y
O
L
O
v8n (
na
no
)
V
e
r
y s
m
a
l
l
3.2
V
e
r
y f
a
s
t
Y
O
L
O
v8s
(
s
m
a
l
l
)
S
m
a
l
l
7.2
F
a
s
t
Y
O
L
O
v8m
(
m
e
di
um
)
M
e
di
um
21.2
M
e
di
um
Y
O
L
O
v8l
(
l
a
r
ge
)
L
a
r
ge
46.5
S
l
ow
Y
O
L
O
v8x (
e
xt
r
a
l
a
r
ge
)
V
e
r
y l
a
r
ge
87.7
V
e
r
y s
l
ow
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
, V
ol
.
15
, N
o.
1
,
F
e
br
ua
r
y
20
26
:
116
-
128
120
F
ig
ur
e
3. Y
O
L
O
v8
a
r
c
hi
te
c
tu
r
e
[
34]
3.
R
E
S
U
L
T
S
A
N
D
D
I
S
C
U
S
S
I
O
N
T
hi
s
s
e
c
ti
on pr
e
s
e
nt
s
t
he
r
e
s
e
a
r
c
h r
e
s
ul
t
s
, s
ta
r
ti
ng w
it
h da
ta
pr
e
pa
r
a
ti
on. I
t
th
e
n c
ove
r
s
t
he
m
ode
li
ng
,
e
va
lu
a
ti
on, a
nd de
pl
oym
e
nt
pha
s
e
s
. F
in
a
ll
y, a
c
om
pr
e
he
n
s
iv
e
d
is
c
us
s
io
n i
s
pr
ovi
de
d a
t
th
e
e
nd of
t
he
s
e
c
ti
on.
3.1. Dat
a
p
r
e
p
ar
at
io
n
T
he
r
e
a
r
e
4
s
ta
ge
s
of
da
ta
pr
e
pa
r
a
ti
on
in
th
is
s
tu
dy,
in
c
lu
d
in
g
da
ta
pr
e
-
pr
oc
e
s
s
in
g,
a
nnot
a
ti
on,
s
pl
it
ti
ng,
a
nd
da
ta
a
ugm
e
nt
a
ti
on.
B
a
s
e
d
on
th
e
unde
r
s
ta
ndi
ng
of
th
e
da
ta
th
a
t
ha
s
be
e
n
c
ol
le
c
te
d,
th
e
im
a
ge
e
xt
e
ns
io
ns
va
r
y
f
r
om
J
P
G
,
J
P
E
G
,
a
nd
P
N
G
,
s
o
a
ll
a
r
e
s
ta
nda
r
di
z
e
d
in
to
J
P
G
.
T
he
n
th
e
va
r
io
us
s
iz
e
s
of
th
e
im
a
ge
s
r
a
nge
f
r
om
ki
lo
byt
e
s
to
m
e
ga
byt
e
s
,
w
it
h
a
n
a
ve
r
a
ge
s
i
z
e
of
2.6
m
e
ga
byt
e
s
(
3,024
px
×
4,032
px)
f
or
th
e
pr
im
a
r
y
da
ta
s
e
t
a
nd
500
ki
lo
byt
e
s
(
718
px
×
420
px)
f
or
th
e
s
e
c
onda
r
y
da
ta
s
e
t,
s
o
th
a
t
a
ll
im
a
ge
s
iz
e
s
a
r
e
s
ta
nda
r
di
z
e
d
to
608
px
×
608
px.
T
he
a
nnot
a
ti
on
pr
oc
e
s
s
is
t
he
n
c
a
r
r
ie
d
out
us
in
g
R
obof
lo
w
to
m
a
r
k
th
e
m
a
ngos
te
e
n
f
r
ui
t
in
th
e
im
a
ge
.
T
hi
s
a
nnot
a
ti
on
s
ta
ge
la
be
ls
th
e
m
a
ngos
te
e
n
f
r
ui
t
in
to
two
c
a
t
e
gor
ie
s
,
n
a
m
e
ly
e
xpor
t
-
w
or
th
y
a
nd
unw
or
th
y.
T
he
r
e
s
ul
t
s
of
th
e
m
a
ngo
s
te
e
n
a
nnot
a
ti
on
w
e
r
e
obt
a
in
e
d
w
it
h
a
to
ta
l
of
4,014 ma
ngos
te
e
n f
r
ui
ts
, w
it
h a
di
s
tr
ib
ut
io
n of
2,717 f
r
ui
ts
f
r
o
m
pr
im
a
r
y da
ta
a
nd
1,297 f
r
om
s
e
c
onda
r
y da
ta
.
T
a
bl
e
2
s
how
s
th
e
di
s
tr
ib
ut
io
n
of
f
e
a
s
ib
le
a
nd
unf
i
t
la
be
ls
a
f
t
e
r
c
a
r
r
yi
ng
out
th
e
a
nnot
a
ti
on
pr
oc
e
s
s
on
th
e
a
nnot
a
te
d m
a
ngos
te
e
n f
r
ui
t
im
a
ge
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
R
e
al
-
ti
m
e
obj
e
c
t
de
te
c
ti
on t
o c
la
s
s
if
y
e
x
por
t
qual
it
y
of
m
angos
t
e
e
n …
(
D
ia
n Sa’
adi
ll
ah M
ay
la
w
at
i)
121
T
a
bl
e
2. D
is
tr
ib
ut
io
n of
m
a
ngos
te
e
n a
nnot
a
ti
on
D
a
t
a
s
our
c
e
s
N
um
be
r
of
e
xpor
t
w
or
t
hy
N
um
be
r
of
e
xpor
t
unw
or
t
hy
T
ot
a
l
a
nnot
a
t
i
on
P
r
i
m
a
r
y
1,602
1,115
2,717
S
e
c
onda
r
y
856
441
1,297
T
he
n,
th
e
da
ta
s
pl
it
ti
ng
a
nd
a
ugm
e
nt
a
ti
on
pr
oc
e
s
s
is
c
a
r
r
ie
d
out
s
e
que
nt
ia
ll
y.
A
f
te
r
th
e
pr
oc
e
s
s
of
di
vi
di
ng
th
e
tr
a
in
in
g,
va
li
da
ti
on,
a
nd
te
s
ti
ng
da
ta
w
it
h
a
r
a
ti
o
va
r
ia
ti
on
of
80:
10:
10,
70:
20:
10,
a
nd
60:
20:
20,
th
e
ne
xt
s
te
p
is
th
e
da
t
a
a
ugm
e
nt
a
ti
on
pr
oc
e
s
s
,
w
it
h
th
e
a
ug
m
e
nt
a
ti
on
r
e
s
ul
ts
,
a
s
s
how
n
in
T
a
bl
e
3.
T
hi
s
a
ugm
e
nt
a
ti
on
pr
oc
e
s
s
a
im
s
to
e
nr
ic
h
th
e
d
a
ta
s
e
t
b
e
c
a
u
s
e
th
e
de
e
p
le
a
r
ni
ng
m
e
th
od
on
Y
O
L
O
v8
is
m
or
e
opt
im
a
l
f
or
la
r
ge
da
ta
[
35
]
–
[
39]
.
S
om
e
of
th
e
a
ugm
e
nt
a
ti
on
m
e
c
ha
ni
s
m
s
us
e
d
in
c
lu
de
r
ot
a
ti
ng,
c
r
oppi
ng
(0
-
15%
)
, a
nd bounding box f
li
p.
T
a
bl
e
3. D
is
tr
ib
ut
io
n of
s
pl
it
ti
ng a
nd a
ugm
e
nt
a
ti
on da
ta
S
pl
i
t
t
i
ng r
a
t
i
o
A
c
t
i
vi
t
y
P
r
i
m
a
r
y
(
a
ugm
e
nt
a
t
i
on)
da
t
a
S
e
c
onda
r
y
(
a
ugm
e
nt
a
t
i
on)
da
t
a
T
ot
a
l
(
a
ugm
e
nt
a
t
i
on)
da
t
a
80:
10:
10
T
r
a
i
ni
ng
1,379 (
4,137)
196 (
588)
1,575 (
4,725)
V
a
l
i
da
t
i
on
172
24
196
T
e
s
t
i
ng
171
24
195
70:
20:
10
T
r
a
i
ni
ng
1,209 (
4,137)
172 (
513)
1,380 (
4,650)
V
a
l
i
da
t
i
on
342
48
391
T
e
s
t
i
ng
171
24
195
60:
20:
20
T
r
a
i
ni
ng
1,038 (
3,546)
148 (
444)
1,186 (
3,990)
V
a
l
i
da
t
i
on
342
48
390
T
e
s
t
i
ng
342
48
390
3.2. M
od
e
li
n
g an
d
e
val
u
at
io
n
r
e
s
u
lt
T
he
m
ode
l
-
bui
ld
in
g
e
xpe
r
im
e
nt
in
th
is
s
tu
dy
c
om
bi
ne
s
f
iv
e
Y
O
L
O
v8
va
r
ia
nt
s
,
na
m
e
ly
na
no,
s
m
a
ll
,
m
e
di
um
, l
a
r
ge
, a
nd e
xt
r
a
la
r
ge
, w
it
h va
r
ia
ti
ons
i
n s
pl
it
ti
ng da
ta
tr
a
in
in
g, va
li
da
ti
on, a
nd t
e
s
ti
ng. T
he
r
e
s
ul
ts
of
th
e
e
xpe
r
im
e
nt
w
e
r
e
th
e
n
e
va
lu
a
t
e
d
us
in
g
a
c
onf
us
io
n
m
a
tr
ix
th
a
t
s
how
s
th
e
v
a
lu
e
s
of
a
c
c
ur
a
c
y,
pr
e
c
is
io
n,
a
nd
r
e
c
a
ll
us
in
g
(
1)
to
(
3)
[
40]
.
T
he
e
xpe
r
im
e
nt
w
a
s
c
a
r
r
ie
d
out
us
in
g
a
hype
r
pa
r
a
m
e
te
r
c
onf
ig
ur
a
ti
on
in
th
e
f
or
m
o
f
a
m
a
xi
m
u
m
e
poc
h
of
100,
a
ba
tc
h
s
iz
e
of
16,
a
nd
a
pa
ti
e
nc
e
of
20.
T
he
us
e
of
a
uni
f
or
m
hype
r
pa
r
a
m
e
te
r
c
onf
ig
ur
a
ti
on
e
ns
ur
e
s
th
a
t
th
e
di
f
f
e
r
e
nc
e
in
p
e
r
f
or
m
a
nc
e
be
twe
e
n
s
c
e
n
a
r
io
s
c
om
e
s
e
nt
ir
e
ly
f
r
om
th
e
da
ta
r
a
ti
o
a
nd
m
ode
l
a
r
c
hi
te
c
tu
r
e
va
r
ia
nt
s
,
w
it
hout
be
in
g
in
f
lu
e
nc
e
d
by
di
f
f
e
r
e
nc
e
s
in
tr
a
in
in
g
pa
r
a
m
e
te
r
s
.
T
h
e
r
e
s
ul
ts
of
th
is
s
ta
ge
a
r
e
in
th
e
f
or
m
of
pr
e
c
is
i
on,
r
e
c
a
ll
,
a
nd
a
c
c
ur
a
c
y.
T
he
s
e
va
lu
e
s
w
il
l
be
th
e
be
nc
hm
a
r
k
f
or
th
e
s
uc
c
e
s
s
of
th
e
m
ode
l
f
or
th
is
s
tu
dy,
a
nd
th
e
a
c
c
ur
a
c
y
r
e
s
ul
ts
a
r
e
gi
ve
n
a
th
r
e
s
hol
d
c
onf
id
e
nc
e
va
lu
e
of
0.5 a
nd 0.7.
=
+
(
1)
=
(
+
)
(
2)
=
(
+
)
(
3)
T
he
s
e
th
r
e
e
e
qua
ti
ons
ha
ve
a
c
r
uc
ia
l
r
ol
e
in
e
va
lu
a
ti
ng
th
e
pe
r
f
or
m
a
nc
e
of
th
e
obj
e
c
t
de
te
c
ti
on
m
ode
l
us
e
d
in
th
e
s
tu
dy
[
41]
,
[
42]
.
A
c
c
ur
a
c
y
m
e
a
s
ur
e
s
th
e
ove
r
a
ll
c
or
r
e
c
tn
e
s
s
of
th
e
m
ode
l’
s
pr
e
di
c
ti
ons
on
th
e
te
s
t
da
ta
s
e
t.
P
r
e
c
is
io
n
e
va
lu
a
te
s
th
e
r
e
li
a
bi
li
ty
of
pos
it
iv
e
pr
e
di
c
ti
ons
,
a
nd
r
e
c
a
ll
m
e
a
s
ur
e
s
th
e
m
ode
l’
s
a
bi
li
ty
to
f
in
d
a
ll
r
e
le
va
nt
pos
it
iv
e
c
a
s
e
s
.
B
e
s
id
e
s
,
th
is
m
ode
l
is
a
ls
o
e
va
lu
a
te
d
us
in
g
m
A
P
50
a
nd
m
A
P
50
-
95.
T
he
m
e
tr
ic
s
m
A
P
50
a
nd
m
A
P
50
-
95
a
r
e
c
r
it
ic
a
l
f
or
a
s
s
e
s
s
i
ng
th
e
pe
r
f
or
m
a
nc
e
of
obj
e
c
t
id
e
nt
if
ic
a
ti
on
a
lg
or
it
hm
s
,
pr
ovi
di
ng
in
f
or
m
a
ti
on
a
bout
th
e
ir
a
c
c
ur
a
c
y
a
nd
r
obus
tn
e
s
s
.
m
A
P
50
c
a
lc
ul
a
te
s
th
e
m
A
P
a
t
a
n
in
te
r
s
e
c
ti
on
ove
r
uni
on
(
I
oU
)
c
r
it
e
r
io
n
of
0.50,
w
he
r
e
a
s
m
A
P
50
-
95
a
ve
r
a
ge
s
th
e
a
ve
r
a
ge
pr
e
c
i
s
io
n
a
c
r
o
s
s
s
e
ve
r
a
l
I
oU
th
r
e
s
hol
ds
r
a
ngi
ng
f
r
om
0.50
to
0.95.
T
hi
s
dua
l
t
e
c
hni
que
e
na
bl
e
s
a
f
ul
l
a
s
s
e
s
s
m
e
nt
of
m
ode
l
pe
r
f
or
m
a
nc
e
[
43]
–
[
45
]
.
F
ig
ur
e
4
s
how
s
a
n
e
xa
m
pl
e
of
t
he
Y
O
L
O
v8
tr
a
in
in
g
p
r
oc
e
s
s
to
r
e
c
ogni
z
e
m
a
ngos
te
e
ns
th
a
t
a
r
e
s
ui
ta
bl
e
a
nd
un
s
ui
ta
bl
e
f
or
e
xpor
t.
W
he
r
e
a
va
lu
e
of
1
in
di
c
a
te
s
a
m
a
ngos
t
e
e
n
th
a
t
i
s
s
ui
ta
bl
e
f
or
e
xpor
t,
w
hi
le
a
va
lu
e
of
0
in
di
c
a
te
s
a
m
a
ngos
te
e
n
th
a
t
is
uns
ui
ta
bl
e
f
or
e
xpor
t.
F
ur
th
e
r
m
or
e
,
F
ig
ur
e
5
s
how
s
a
n
e
xa
m
pl
e
of
th
e
va
li
da
ti
on
la
be
ls
a
nd
pr
e
di
c
ti
on
pr
oc
e
s
s
f
or
m
a
ngos
te
e
ns
,
“
la
y
a
k
”
m
e
a
n
s
“
w
or
th
y”
, w
hi
le
“
ti
dak
l
ay
ak
”
m
e
a
ns
“
unw
or
th
y”
. L
a
s
t,
t
he
m
o
de
l
te
s
ti
ng s
ta
ge
i
s
i
ll
us
tr
a
te
d i
n F
ig
ur
e
6.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
, V
ol
.
15
, N
o.
1
,
F
e
br
ua
r
y
20
26
:
116
-
128
122
F
ig
ur
e
4. T
he
e
xa
m
pl
e
of
t
he
t
r
a
in
in
g pr
oc
e
s
s
w
it
h pr
im
a
r
y a
nd
s
e
c
onda
r
y m
a
ngos
t
e
e
n da
ta
s
e
ts
F
ig
ur
e
5. T
he
e
xa
m
pl
e
of
t
he
va
li
da
ti
on pr
oc
e
s
s
w
it
h pr
im
a
r
y a
nd s
e
c
onda
r
y m
a
ngos
t
e
e
n da
ta
s
e
ts
F
ig
ur
e
6. T
he
e
xa
m
pl
e
of
t
e
s
ti
ng pr
oc
e
s
s
w
it
h pr
im
a
r
y a
nd s
e
c
o
nda
r
y m
a
ngos
te
e
n da
ta
s
e
ts
B
a
s
e
d
on
th
e
r
e
s
ul
ts
of
e
xpe
r
im
e
nt
s
w
it
h
15
Y
O
L
O
v8
va
r
ia
nt
s
c
e
na
r
io
s
as
in
T
a
bl
e
4,
it
s
how
s
th
a
t
th
e
na
no va
r
ia
nt
ha
s
t
he
be
s
t
a
c
c
ur
a
c
y r
e
s
ul
ts
i
n de
te
c
ti
ng t
he
e
xpor
t
e
li
gi
bi
li
ty
o
f
m
a
ngos
te
e
n f
r
ui
t.
H
ow
e
ve
r
,
hi
gh
a
c
c
ur
a
c
y
is
not
th
e
onl
y
d
e
te
r
m
in
in
g
f
a
c
to
r
in
th
e
qua
li
ty
of
th
e
Y
O
L
O
v8
va
r
ia
nt
,
b
e
c
a
us
e
ot
he
r
f
a
c
to
r
s
a
f
f
e
c
t
th
e
qua
li
ty
of
th
e
va
r
ia
nt
,
in
c
lu
di
ng
da
ta
s
e
ts
a
nd
ot
he
r
hype
r
pa
r
a
m
e
te
r
s
.
T
he
r
e
s
ul
t
s
of
e
a
c
h
m
ode
l
th
a
t
ha
s
be
e
n
bui
lt
gr
e
a
tl
y
a
f
f
e
c
t
th
e
qua
li
ty
of
th
e
m
ode
l,
da
ta
s
e
t,
a
nd
s
e
ve
r
a
l
ot
he
r
pa
r
a
m
e
te
r
s
.
T
he
e
xpe
r
im
e
nt
a
l
r
e
s
ul
ts
s
how
e
vi
de
n
c
e
th
a
t
onl
y
th
e
two
be
s
t
s
c
e
na
r
io
s
e
xc
e
e
d
th
e
e
xp
e
c
te
d
a
c
c
ur
a
c
y
li
m
it
of
70%
,
a
lm
os
t
a
ll
th
e
80:
10:
10
a
nd
70
:2
0:
10
s
pl
it
ti
ng
s
c
e
na
r
io
s
w
it
h
th
e
Y
O
L
O
v8
a
c
hi
e
vi
ng
a
c
c
ur
a
c
y
va
lu
e
s
m
or
e
th
a
n
70%
,
e
xc
e
pt
Y
O
L
O
v8
na
no
w
it
h
s
pl
it
ti
ng
da
ta
80
:1
0:
10.
T
he
hi
ghe
s
t
va
lu
e
s
a
r
e
hi
ghl
ig
ht
e
d
in
bl
ue
c
ol
or
i
n t
he
T
a
bl
e
4.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
R
e
al
-
ti
m
e
obj
e
c
t
de
te
c
ti
on t
o c
la
s
s
if
y
e
x
por
t
qual
it
y
of
m
angos
t
e
e
n …
(
D
ia
n Sa’
adi
ll
ah M
ay
la
w
at
i)
123
T
a
bl
e
4. E
xpe
r
im
e
nt
r
e
s
ul
t
S
pl
i
t
i
ng
Y
O
L
O
v8
va
r
i
a
nt
s
S
t
oppe
d
e
poc
h
P
r
e
c
i
s
i
on
R
e
c
a
l
l
A
c
c
ur
a
c
y (
%
)
m
A
P
50
m
A
P
50
-
95
T
r
a
i
ni
ng
T
e
s
t
i
ng
C
onf
. 0.5
C
onf
. 0.7
08:
10:
10
N
a
no
75
0.817
0.668
83
40
69
0.767
0.507
S
m
a
l
l
85
0.832
0.697
80
50
71
0.777
0.521
M
e
di
um
73
0.823
0.689
81
47
70
0.776
0.525
L
a
r
ge
75
0.871
0.655
81
45
70
0.782
0.534
E
xt
r
a
l
a
r
ge
81
0.827
0.694
80
50
72
0.782
0.529
70:
20:
10
N
a
no
97
0.907
0.792
71
69
81
0.848
0.603
S
m
a
l
l
84
0.898
0.77
68
60
78
0.842
0.601
M
e
di
um
85
0.905
0.813
72
70
82
0.832
0.602
L
a
r
ge
98
0.911
0.75
65
58
76
0.856
0.616
E
xt
r
a
l
a
r
ge
100
0.909
0.79
67
65
80
0.848
0.603
60:
20:
20
N
a
no
85
0.576
0.682
51
33
45
0.647
0.321
S
m
a
l
l
78
0.598
0.67
48
31
45
0.663
0.324
M
e
di
um
89
0.598
0.663
51
35
48
0.652
0.322
L
a
r
ge
46
0.511
0.75
50
22
45
0.637
0.321
E
xt
r
a
l
a
r
ge
83
0.598
0.66
49
50
51
0.657
0.322
T
he
e
xpe
r
im
e
nt
a
l
r
e
s
ul
t
s
in
T
a
bl
e
4
s
how
th
a
t
th
e
Y
O
L
O
v8
m
e
di
um
va
r
ia
nt
m
ode
l
w
it
h
70:
20:
10
da
ta
s
pl
it
ti
ng
ha
s
th
e
be
s
t
pe
r
f
or
m
a
nc
e
w
it
h
th
e
hi
ghe
s
t
r
e
c
a
ll
va
lu
e
,
w
hi
c
h
is
0.813.
T
he
hi
ghe
s
t
a
c
c
ur
a
c
y
va
lu
e
is
73%
f
or
a
c
onf
id
e
nt
s
c
or
e
of
0.5
a
nd
82%
f
or
a
c
onf
id
e
nt
s
c
or
e
of
0.7.
T
he
hi
ghe
s
t
pr
e
c
is
io
n,
m
A
P
50,
m
A
P
50
-
95,
is
r
e
a
c
he
d
by
th
e
Y
O
L
O
v8
la
r
ge
va
r
ia
nt
w
it
h
70:
20:
10
da
ta
s
pl
it
ti
ng
in
s
e
que
nc
e
0.911,
0.856,
a
nd
0.616
c
om
pa
r
e
d
to
ot
he
r
m
ode
ls
.
H
ow
e
ve
r
,
th
e
Y
O
L
O
v8
na
no
va
r
ia
nt
ha
s
a
f
a
ir
ly
hi
gh
e
poc
h
s
to
ppi
ng
ti
m
e
,
w
hi
c
h
s
to
ps
a
t
e
poc
h
80
w
he
n
th
e
m
ode
l
ha
s
c
onve
r
ge
d
(
s
ta
bl
e
)
.
T
hi
s
Y
O
L
O
v8
na
no
va
r
ia
nt
a
ls
o
ha
s
th
e
be
s
t
a
c
c
ur
a
c
y
va
lu
e
f
or
a
c
onf
id
e
nc
e
va
lu
e
of
0.5
f
or
a
ll
da
ta
s
pl
it
ti
ng
va
r
ia
ti
ons
.
D
a
ta
s
pl
it
ti
ng
80:
10:
10 ha
s
t
he
be
s
t
tr
a
in
in
g pr
oc
e
s
s
a
c
c
ur
a
c
y va
lu
e
, w
hi
c
h i
s
83%
i
n t
he
na
no
va
r
ia
nt
.
T
hi
s
s
tu
dy
f
ound
a
li
gnm
e
nt
w
it
h
p
r
e
vi
ous
s
tu
di
e
s
w
he
r
e
th
e
Y
O
L
O
v8
na
no
va
r
ia
nt
ha
d
th
e
hi
ghe
s
t
a
c
c
ur
a
c
y
in
r
e
c
ogni
z
in
g
obj
e
c
ts
[
46]
.
T
h
e
Y
O
L
O
v8
m
e
di
u
m
va
r
ia
nt
ha
d
th
e
hi
ghe
s
t
pr
e
c
is
io
n
va
lu
e
in
de
te
c
ti
ng
obj
e
c
ts
[
47]
.
H
ig
he
r
Y
O
L
O
v8
va
r
ia
nt
s
c
e
r
ta
in
l
y
r
e
qui
r
e
e
xpe
ns
iv
e
c
om
put
in
g
r
e
s
our
c
e
s
.
C
onve
r
s
e
ly
,
s
m
a
ll
e
r
m
ode
l
s
li
ke
Y
O
L
O
v8
na
no
pr
ovi
de
a
pr
a
c
ti
c
a
l
s
ol
ut
io
n
f
or
r
e
a
l
-
ti
m
e
ta
s
ks
w
it
h
li
m
it
e
d
r
e
s
our
c
e
s
[
47]
,
[
48]
.
W
hi
le
la
r
ge
r
da
ta
s
e
t
s
ge
ne
r
a
ll
y
in
c
r
e
a
s
e
i
m
a
ge
pr
oc
e
s
s
in
g
qua
li
ty
,
a
tt
e
nt
io
n
s
houl
d
a
l
s
o
be
pa
id
to
da
ta
r
e
pr
e
s
e
nt
a
ti
ve
ne
s
s
a
nd
qua
li
ty
.
B
a
la
nc
in
g
s
iz
e
a
nd
qua
li
ty
a
r
e
c
r
it
ic
a
l
f
or
a
c
hi
e
vi
ng
opt
im
a
l
m
ode
l
pe
r
f
or
m
a
nc
e
.
I
nc
r
e
a
s
in
g
th
e
num
be
r
of
tr
a
in
in
g
da
ta
im
pr
ove
s
c
la
s
s
if
ie
r
pe
r
f
or
m
a
nc
e
,
a
s
s
e
e
n
in
la
nd
c
ove
r
m
a
ppi
ng e
xpe
r
im
e
nt
s
w
he
r
e
a
c
c
ur
a
c
y pe
a
k
e
d a
t
a
s
p
e
c
if
ie
d t
r
a
in
in
g/
te
s
t
r
a
ti
o
[
49]
.
L
a
r
ge
r
da
ta
s
e
ts
a
r
e
e
xt
r
e
m
e
ly
be
ne
f
ic
ia
l
in
ne
ur
a
l
ne
twor
ks
,
s
uc
h
a
s
de
e
p
le
a
r
ni
ng,
w
hi
c
h
c
ons
id
e
r
e
a
c
h
s
a
m
pl
e
dur
in
g
tr
a
in
in
g.
T
he
y
he
lp
de
te
r
m
in
e
c
la
s
s
bor
de
r
s
m
or
e
a
c
c
ur
a
te
ly
.
H
ow
e
ve
r
,
w
hi
le
hi
gh
-
qua
li
ty
tr
a
in
in
g
da
ta
is
r
e
qui
r
e
d
f
o
r
be
s
t
m
ode
l
pe
r
f
o
r
m
a
nc
e
,
th
e
tr
a
de
-
of
f
s
be
twe
e
n
da
ta
qua
li
ty
a
nd
qua
nt
it
y
m
us
t
a
ls
o
be
c
on
s
id
e
r
e
d.
H
ig
h
w
it
hi
n
-
im
a
ge
di
ve
r
s
it
y,
w
h
ic
h
in
c
lu
de
s
a
va
r
ie
ty
of
obj
e
c
ts
,
a
nd
lo
w
c
om
pr
e
s
s
io
n
a
r
ti
f
a
c
ts
in
da
ta
s
e
ts
,
im
pr
ove
s
th
e
pe
r
f
or
m
a
nc
e
of
i
m
a
ge
s
upe
r
-
r
e
s
ol
ut
io
n
m
ode
ls
[
50]
.
E
f
f
e
c
ti
ve
pr
e
pr
oc
e
s
s
in
g
pr
oc
e
dur
e
s
,
in
c
lu
di
ng
r
e
s
iz
in
g,
nor
m
a
li
z
a
ti
on,
d
a
ta
a
ugm
e
nt
a
ti
on,
a
nd
c
r
oppi
ng,
a
r
e
c
r
it
ic
a
l
f
or
im
pr
ovi
ng da
ta
qua
li
ty
[
51]
.
B
e
s
id
e
s
m
a
ngos
te
e
n,
th
e
pr
o
pos
e
d
Y
O
L
O
v
8
-
ba
s
e
d
f
r
a
m
e
w
or
k
c
a
n
be
ge
ne
r
a
l
iz
e
d
to
ot
he
r
hi
gh
-
va
lu
e
tr
opi
c
a
l
e
xpor
t
c
om
m
odi
ti
e
s
s
uc
h
a
s
dur
ia
n,
a
voc
a
do,
a
nd
dr
a
gon
f
r
ui
t,
w
hi
c
h
a
ls
o
r
e
qui
r
e
pr
e
c
is
e
gr
a
di
ng
ba
s
e
d
on
s
ur
f
a
c
e
te
xt
ur
e
,
c
ol
or
uni
f
or
m
it
y,
a
nd
phys
ic
a
l
de
f
e
c
ts
.
A
c
tu
a
ll
y,
s
e
ve
r
a
l
s
tu
di
e
s
ha
v
e
a
ls
o
c
onduc
te
d
f
r
ui
t
pos
t
-
ha
r
ve
s
t
qua
li
ty
c
ont
r
ol
us
in
g
Y
O
L
O
,
s
uc
h
a
s
ki
w
if
r
ui
t
[
52
]
a
nd
a
ppl
e
[
53]
.
T
he
a
da
pt
a
bi
li
ty
of
th
e
m
ode
l
a
r
c
hi
te
c
tu
r
e
a
nd
da
ta
pi
p
e
li
ne
e
na
bl
e
s
tr
a
ns
f
e
r
le
a
r
ni
ng,
w
he
r
e
pr
e
tr
a
in
e
d
w
e
ig
ht
s
f
r
om
m
a
ngos
te
e
n
da
ta
s
e
ts
c
a
n
be
f
in
e
-
tu
ne
d
f
or
ne
w
f
r
ui
t
ty
pe
s
w
it
h
m
in
im
a
l
a
ddi
ti
ona
l
da
ta
c
ol
le
c
ti
on.
T
hi
s
s
c
a
la
bi
li
ty
ope
ns
oppor
tu
ni
ti
e
s
f
or
de
ve
lo
pi
ng
a
n
in
te
gr
a
t
e
d,
a
r
ti
f
ic
ia
l
in
te
ll
ig
e
nc
e
-
dr
iv
e
n
pos
th
a
r
ve
s
t
in
s
pe
c
ti
on
s
y
s
te
m
th
a
t
s
uppor
ts
I
ndone
s
ia
’
s
br
oa
de
r
a
gr
ic
ul
tu
r
a
l
e
xpor
t
s
e
c
to
r
,
e
nha
nc
in
g
c
om
pe
ti
ti
ve
ne
s
s
a
nd c
ons
is
te
n
c
y i
n gl
oba
l
m
a
r
ke
ts
.
3.3. De
p
lo
ym
e
n
t
T
he
de
pl
oym
e
nt
pr
oc
e
s
s
is
bui
lt
s
im
pl
y
us
in
g
f
r
a
m
e
w
or
ks
s
u
c
h
a
s
P
yT
or
c
h
w
it
h
G
P
U
s
uppor
t
a
nd
T
e
ns
or
F
lo
w
t
o r
un t
he
Y
O
L
O
v8 mode
l.
T
he
i
nf
e
r
e
nc
e
e
ngi
n
e
u
s
e
d i
s
O
pe
n
C
V
t
o pr
oc
e
s
s
i
m
a
ge
s
i
n r
e
a
l
-
ti
m
e
.
A
de
qua
te
R
A
M
is
e
s
s
e
nt
ia
l.
T
h
e
r
e
f
or
e
,
pr
e
vi
ous
r
e
s
e
a
r
c
h
r
e
c
o
m
m
e
nds
us
in
g
de
vi
c
e
s
th
a
t
ty
pi
c
a
ll
y
r
e
qui
r
e
a
t
le
a
s
t
4
G
B
to
ha
ndl
e
th
e
m
ode
l'
s
ope
r
a
ti
ons
e
f
f
ic
ie
nt
ly
[
54]
.
I
n
a
ddi
ti
on,
F
la
s
k
is
u
s
e
d
to
bui
ld
a
n
in
f
e
r
e
nc
e
a
ppl
ic
a
ti
on pr
ogr
a
m
m
in
g i
nt
e
r
f
a
c
e
(
A
P
I
)
if
t
he
m
ode
l
is
a
c
c
e
s
s
e
d vi
a
a
ne
twor
k s
e
r
vi
c
e
. T
he
be
s
t
m
ode
l
f
r
om
th
e
e
xpe
r
im
e
nt
is
s
e
le
c
te
d
in
th
e
de
pl
oym
e
nt
pr
oc
e
s
s
.
F
ig
ur
e
7
s
how
s
a
n
e
xa
m
pl
e
of
th
e
be
s
t
m
ode
l
de
pl
oym
e
nt
in
de
te
c
ti
ng
m
a
ngos
t
e
e
ns
th
a
t
a
r
e
w
or
th
y
a
nd
unw
or
th
y
f
or
e
xpor
t,
m
a
ki
ng
it
e
a
s
y
a
nd
e
f
f
ic
ie
nt
in
t
he
s
or
ti
ng pr
oc
e
s
s
.
T
hi
s
m
ode
l
c
a
n d
e
te
c
t
85%
c
or
r
e
c
tl
y of
m
a
ngos
te
e
n i
n r
e
a
l
-
ti
m
e
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
, V
ol
.
15
, N
o.
1
,
F
e
br
ua
r
y
20
26
:
116
-
128
124
F
ig
ur
e
7. T
he
e
xa
m
pl
e
of
r
e
a
l
-
ti
m
e
m
a
ngos
te
e
n de
t
e
c
ti
on w
it
h s
e
c
onda
r
y a
nd pr
im
a
r
y vi
de
o
4.
C
O
N
C
L
U
S
I
O
N
T
hi
s
w
or
k
s
uc
c
e
s
s
f
ul
ly
c
r
e
a
te
d
a
r
e
a
l
-
ti
m
e
m
a
ngos
te
e
n
s
or
ti
ng
m
ode
l
th
a
t
us
e
s
th
e
Y
O
L
O
v8
a
lg
or
it
hm
to
c
a
te
gor
iz
e
e
xpor
t
-
qua
li
ty
f
r
ui
t
ba
s
e
d
on
vi
s
ua
l
c
ha
r
a
c
te
r
is
ti
c
s
s
uc
h
a
s
s
ki
n
f
la
w
s
a
nd
s
t
a
lk
c
ondi
ti
on.
T
he
e
xpe
r
im
e
nt
a
l
r
e
s
ul
ts
s
ho
w
th
a
t
th
e
Y
O
L
O
v8
m
e
di
um
va
r
ia
ti
on
pe
r
f
or
m
e
d
be
s
t
ove
r
a
ll
,
w
it
h
th
e
hi
ghe
s
t
a
c
c
ur
a
c
y
82%
,
but
th
e
Y
O
L
O
v8
na
no
va
r
ia
ti
on
pr
ovi
de
d
a
r
e
a
li
s
ti
c
s
ol
ut
io
n
f
or
r
e
a
l
-
ti
m
e
ta
s
ks
,
m
a
in
ta
in
in
g
s
ta
bl
e
pe
r
f
or
m
a
nc
e
a
t
a
lo
w
e
r
c
om
put
in
g
c
os
t,
w
he
r
e
it
a
c
hi
e
ve
d
th
e
hi
ghe
s
t
tr
a
in
in
g
a
c
c
ur
a
c
y
(
83%
)
.
T
he
pr
opos
e
d
m
ode
l
of
f
e
r
s
a
de
pe
nda
bl
e
a
nd
e
f
f
e
c
ti
ve
te
c
hni
que
f
or
gr
a
di
ng
m
a
ngos
te
e
n
qua
li
ty
,
im
pr
ovi
ng
c
om
pl
ia
nc
e
w
it
h
s
tr
in
ge
nt
e
xpor
t
r
e
gul
a
ti
ons
,
a
nd
in
c
r
e
a
s
in
g
I
ndone
s
ia
'
s
a
gr
ic
ul
tu
r
a
l
c
om
pe
ti
ti
ve
ne
s
s
in
in
te
r
na
ti
ona
l
m
a
r
ke
ts
.
T
he
r
e
f
or
e
,
th
is
s
tu
d
y
c
onf
ir
m
s
pr
io
r
f
in
di
ngs
on
th
e
e
f
f
ic
ie
nc
y
of
s
m
a
ll
e
r
m
ode
ls
li
ke
Y
O
L
O
v8
na
no
f
or
r
e
s
our
c
e
-
li
m
it
e
d
e
nvi
r
onm
e
nt
s
a
nd
th
e
im
por
ta
nc
e
of
la
r
ge
r
m
ode
l
s
li
ke
Y
O
L
O
v8
m
e
di
um
a
nd
la
r
ge
f
or
hi
gh
-
pr
e
c
is
io
n
ta
s
ks
.
I
t
a
ls
o
e
m
pha
s
iz
e
s
th
a
t
da
ta
s
e
t
s
iz
e
a
nd
qua
li
ty
a
r
e
c
r
it
ic
a
l
f
a
c
to
r
s
in
opt
im
iz
in
g
de
e
p
le
a
r
ni
ng
m
ode
l
pe
r
f
or
m
a
nc
e
.
I
nc
r
e
a
s
in
g
da
ta
di
ve
r
s
it
y,
im
pr
ovi
ng
pr
e
pr
oc
e
s
s
in
g
s
te
p
s
,
a
nd
ba
la
nc
in
g
da
ta
qua
li
ty
w
it
h
qua
nt
it
y
a
r
e
e
s
s
e
nt
ia
l
to
e
nha
nc
in
g
m
ode
l
a
c
c
ur
a
c
y
a
nd
ge
ne
r
a
li
z
a
bi
li
ty
.
F
ut
ur
e
r
e
s
e
a
r
c
h
c
oul
d
in
c
r
e
a
s
e
th
e
da
ta
s
e
t'
s
di
ve
r
s
it
y
to
in
c
r
e
a
s
e
th
e
m
ode
l'
s
ge
ne
r
a
li
z
a
bi
li
ty
a
c
r
os
s
di
f
f
e
r
e
nt
e
nvi
r
onm
e
nt
a
l
c
ir
c
um
s
ta
nc
e
s
a
nd
m
a
ngos
te
e
n
ki
nds
.
F
ur
th
e
r
m
or
e
,
ne
w
opt
im
iz
a
ti
on
a
ppr
oa
c
he
s
li
ke
hype
r
pa
r
a
m
e
te
r
tu
ni
ng
a
nd
qua
nt
iz
a
ti
on
m
a
y
im
pr
ove
th
e
a
c
c
ur
a
c
y
a
nd
de
pl
oym
e
n
t
e
f
f
ic
ie
nc
y
of
hi
ghe
r
Y
O
L
O
v8
va
r
ia
ti
ons
.
I
nt
e
gr
a
ti
ng
th
e
m
od
e
l
w
it
h
a
ut
om
a
te
d
s
or
ti
ng
ha
r
dw
a
r
e
,
s
uc
h
a
s
c
onve
yor
be
lt
s
or
r
obot
ic
a
r
m
s
,
a
nd
te
s
ti
ng
it
s
pe
r
f
or
m
a
nc
e
in
r
e
a
l
-
w
or
ld
s
c
e
na
r
io
s
w
il
l
he
lp
to
pr
ove
it
s
s
c
a
la
bi
li
ty
a
nd
f
e
a
s
ib
il
it
y.
E
xt
e
ndi
ng
th
is
m
e
th
od
to
ot
he
r
e
xpor
t
c
om
m
odi
ti
e
s
m
a
y
w
id
e
n
a
r
ti
f
ic
ia
l
in
te
ll
ig
e
nc
e
'
s
i
m
pa
c
t
in
i
m
pr
ovi
ng a
gr
ic
ul
tu
r
a
l
s
ys
te
m
s
.
A
C
K
N
O
WL
E
D
G
E
M
E
N
T
S
T
he
a
ut
hor
s
th
a
nk
to
D
ir
e
c
to
r
a
te
of
I
s
la
m
i
c
R
e
li
gi
ou
s
E
duc
a
ti
o
n,
th
e
D
ir
e
c
to
r
a
te
G
e
ne
r
a
l
of
I
s
l
a
m
ic
E
duc
a
ti
on,
M
in
is
tr
y
of
R
e
li
gi
ous
A
f
f
a
ir
s
of
th
e
R
e
publ
ic
of
I
ndone
s
ia
.
I
n
m
os
t
c
a
s
e
s
,
s
pons
or
a
nd
f
in
a
nc
ia
l
s
uppor
t
a
c
knowle
dgm
e
nt
s
.
F
U
N
D
I
N
G
I
N
F
O
R
M
A
T
I
O
N
T
hi
s
r
e
s
e
a
r
c
h
w
a
s
f
unde
d
by
D
ir
e
c
to
r
a
te
of
I
s
la
m
ic
R
e
li
gi
ous
E
duc
a
ti
on,
th
e
D
ir
e
c
to
r
a
te
G
e
ne
r
a
l
of
I
s
la
m
ic
E
duc
a
ti
on,
M
in
i
s
tr
y
of
R
e
li
gi
ous
A
f
f
a
ir
s
of
th
e
R
e
p
ubl
ic
of
I
ndone
s
ia
,
b
a
s
e
d
on
D
e
c
is
io
n
of
th
e
C
om
m
it
m
e
nt
-
M
a
ki
ng
O
f
f
ic
ia
l
of
th
e
D
ir
e
c
to
r
a
te
of
H
ig
he
r
I
s
la
m
ic
R
e
li
gi
ous
E
duc
a
ti
on,
D
ir
e
c
to
r
a
te
G
e
ne
r
a
l
of
I
s
la
m
ic
E
duc
a
ti
on,
M
in
is
tr
y
of
R
e
li
gi
ous
A
f
f
a
ir
s
N
um
be
r
5157
of
2024:
C
onc
e
r
ni
ng
R
e
c
ip
ie
nt
s
of
R
e
s
e
a
r
c
h
A
s
s
i
s
ta
nc
e
b
a
s
e
d on N
a
ti
ona
l
S
tr
a
te
gi
c
P
ol
ic
y f
or
t
he
2024 F
is
c
a
l
Y
e
a
r
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
R
e
al
-
ti
m
e
obj
e
c
t
de
te
c
ti
on t
o c
la
s
s
if
y
e
x
por
t
qual
it
y
of
m
angos
t
e
e
n …
(
D
ia
n Sa’
adi
ll
ah M
ay
la
w
at
i)
125
A
U
T
H
O
R
C
O
N
T
R
I
B
U
T
I
O
N
S
S
T
A
T
E
M
E
N
T
T
hi
s
jo
ur
na
l
us
e
s
th
e
C
ont
r
ib
ut
or
R
ol
e
s
T
a
xonomy
(
C
R
e
di
T
)
to
r
e
c
ogni
z
e
in
di
vi
dua
l
a
ut
hor
c
ont
r
ib
ut
io
ns
, r
e
duc
e
a
ut
hor
s
hi
p di
s
put
e
s
,
a
nd f
a
c
il
it
a
te
c
ol
la
bo
r
a
ti
on
N
am
e
o
f
A
u
t
h
or
C
M
So
Va
Fo
I
R
D
O
E
Vi
Su
P
Fu
D
i
a
n
S
a
’
a
d
i
l
l
a
h
M
a
y
l
a
w
a
t
i
✓
✓
✓
✓
✓
✓
✓
✓
✓
M
i’
r
a
j
F
ua
di
✓
✓
✓
✓
✓
✓
K
ur
ni
a
w
a
n Y
uni
a
r
to
✓
✓
✓
✓
✓
Y
uhe
ndr
a
A
P
✓
✓
✓
✓
R
iz
ky R
a
hm
a
t
N
ugr
a
ha
✓
✓
✓
✓
✓
✓
✓
✓
A
kba
r
H
id
a
ya
tu
ll
a
h
H
a
r
a
ha
p
✓
✓
✓
✓
✓
✓
A
gung W
a
ha
na
✓
✓
✓
✓
C
:
C
onc
e
pt
ua
l
i
z
a
t
i
on
M
:
M
e
t
hodol
ogy
So
:
So
f
t
w
a
r
e
Va
:
Va
l
i
da
t
i
on
Fo
:
Fo
r
m
a
l
a
na
l
ys
i
s
I
:
I
nve
s
t
i
ga
t
i
on
R
:
R
e
s
our
c
e
s
D
:
D
a
t
a
C
ur
a
t
i
on
O
:
W
r
i
t
i
ng
-
O
r
i
gi
na
l
D
r
a
f
t
E
:
W
r
i
t
i
ng
-
R
e
vi
e
w
&
E
di
t
i
ng
Vi
:
Vi
s
ua
l
i
z
a
t
i
on
Su
:
Su
pe
r
vi
s
i
on
P
:
P
r
oj
e
c
t
a
dm
i
ni
s
t
r
a
t
i
on
Fu
:
Fu
ndi
ng a
c
qui
s
i
t
i
on
C
O
N
F
L
I
C
T
O
F
I
N
T
E
R
E
S
T
S
T
A
T
E
M
E
N
T
T
he
a
ut
hor
s
s
t
a
te
no c
onf
li
c
t
of
i
nt
e
r
e
s
t.
I
N
F
O
R
M
E
D
C
O
N
S
E
N
T
W
e
ha
ve
obt
a
in
e
d i
nf
or
m
e
d c
ons
e
nt
f
r
om
a
ll
i
ndi
vi
dua
ls
i
nc
lu
de
d i
n t
hi
s
s
tu
dy.
E
T
H
I
C
A
L
A
P
P
R
O
V
A
L
T
he
a
ut
hor
s
s
t
a
te
t
ha
t
th
e
r
e
s
e
a
r
c
h i
s
r
e
la
te
d t
o r
e
gul
a
ti
ons
a
nd i
ns
ti
tu
ti
ona
l
pol
ic
ie
s
.
D
A
T
A
A
V
A
I
L
A
B
I
L
I
T
Y
T
he
a
va
il
a
bi
li
ty
of
t
he
pr
im
a
r
y a
nd s
e
c
onda
r
y da
t
a
, i
nc
lu
di
ng:
‒
P
r
im
a
r
y da
ta
w
a
s
c
ol
le
c
te
d f
r
om
m
a
ngos
te
e
n pl
a
nt
a
ti
on
s
a
nd s
o
r
ti
ng l
oc
a
ti
ons
i
n t
he
M
a
ta
r
a
m
.
‒
S
e
c
onda
r
y da
ta
w
a
s
c
ol
le
c
t
e
d f
r
om
t
he
i
nt
e
r
ne
t
w
it
h e
xi
s
ti
ng s
e
a
r
c
h e
ngi
ne
s
, s
uc
h
a
s
G
oogl
e
a
nd K
a
ggl
e
a
t
ht
tp
s
:/
/ww
w
.ka
ggl
e
.c
om
/d
a
ta
s
e
ts
/m
a
r
ia
m
e
r
e
s
/d
e
e
p
-
le
a
r
ni
ng
-
kl
a
s
if
ik
a
s
i
-
je
ni
s
-
bua
h
-
m
a
nggi
s
.
R
E
F
E
R
E
N
C
E
S
[
1]
D
.
R
.
F
a
u
z
i
a
na
,
M
a
r
i
m
i
n,
H
.
K
.
S
uw
a
r
s
i
na
h,
a
nd
E
.
A
.
P
r
a
s
e
t
i
o,
“
W
ha
t
f
a
c
t
or
s
i
m
pa
c
t
t
h
e
a
dopt
i
on
of
pos
t
ha
r
ve
s
t
l
os
s
-
r
e
duc
t
i
on
t
e
c
hnol
ogi
e
s
i
n
m
a
ngos
t
e
e
n
s
uppl
y
c
ha
i
n?
,”
J
our
nal
of
O
pe
n
I
nnov
at
i
on:
T
e
c
hnol
ogy
,
M
ar
k
e
t
,
and
C
om
pl
e
x
i
t
y
,
vol
.
9,
no.
3,
S
e
p. 2023, doi
:
10.1016/
j
.j
oi
t
m
c
.2023.100102.
[
2]
A
.
L
.
S
a
ye
kt
i
e
t
al
.
,
“
T
he
pol
i
c
y
i
m
pa
c
t
s
a
nd
i
m
pl
i
c
a
t
i
ons
f
r
om
t
he
i
ns
i
ght
s
of
m
a
ngos
t
e
e
n
e
xpor
t
c
ha
i
n
s
,”
I
O
P
C
onf
e
r
e
nc
e
Se
r
i
e
s
:
E
ar
t
h and E
nv
i
r
onm
e
nt
al
Sc
i
e
nc
e
, vol
. 1153, no. 1, M
a
y 2023, doi
:
10.1
088/
1755
-
1315/
1153/
1/
012014.
[
3]
I
.
B
a
r
oh,
“
C
om
pe
t
i
t
i
ve
ne
s
s
of
I
ndone
s
i
a
n
m
a
ngo
s
t
e
e
n
i
n
t
he
i
nt
e
r
na
t
i
ona
l
m
a
r
ke
t
,”
I
nt
e
r
nat
i
onal
J
our
nal
of
Soc
i
al
Sc
i
e
nc
e
a
nd
H
um
an R
e
s
e
ar
c
h
, vol
. 5, no. 1, J
a
n. 2022, doi
:
10.47191/
i
j
s
s
hr
/
v5
-
i1
-
42.
[
4]
S
.
R
i
ya
di
,
A
.
M
.
A
.
R
a
t
i
w
i
,
C
.
D
a
m
a
r
j
a
t
i
,
T
.
K
.
H
a
r
i
a
di
,
I
.
P
r
a
ba
s
a
r
i
,
a
nd
N
.
A
.
U
t
a
m
a
,
“
C
l
a
s
s
i
f
i
c
a
t
i
on
of
m
a
ngos
t
e
e
n
s
ur
f
a
c
e
qua
l
i
t
y
us
i
ng
pr
i
nc
i
pa
l
c
om
pon
e
nt
a
na
l
y
s
i
s
,”
E
m
e
r
gi
ng
I
nf
or
m
at
i
on
Sc
i
e
nc
e
and
T
e
c
hnol
ogy
,
vol
.
1,
no.
1,
2020,
doi
:
10.18196/
e
i
s
t
.115.
[
5]
A
.
T
ha
m
m
a
s
t
i
t
kul
a
nd
T
.
K
l
a
yj
um
l
a
ng,
“
M
a
ngos
t
e
e
n
qua
l
i
t
y
gr
a
di
ng
f
or
e
xpor
t
m
a
r
ke
t
s
us
i
ng
di
gi
t
a
l
i
m
a
ge
pr
oc
e
s
s
i
ng
t
e
c
hni
que
s
,”
I
nt
e
r
nat
i
onal
J
ou
r
nal
on
A
dv
anc
e
d
Sc
i
e
nc
e
,
E
ngi
ne
e
r
i
ng
and
I
nf
or
m
at
i
on
T
e
c
hnol
ogy
,
vol
.
11,
no.
6,
pp. 2452
–
2458, D
e
c
. 2021, doi
:
10.18517/
i
j
a
s
e
i
t
.11.6.14003.
[
6]
S
.
R
i
ya
di
e
t
al
.
,
“
C
l
a
s
s
i
f
i
c
a
t
i
on
of
m
a
ngos
t
e
e
n
s
ur
f
a
c
e
qua
l
i
t
y
ba
s
e
d
on
i
m
a
ge
pr
oc
e
s
s
i
ng
us
i
ng
s
uppor
t
ve
c
t
or
m
a
c
hi
ne
,”
I
nt
e
r
nat
i
onal
J
our
nal
of
I
nt
e
gr
at
e
d E
ngi
ne
e
r
i
ng
, vol
. 13, no. 5, J
ul
. 2021, doi
:
10.30880/
i
j
i
e
.2021.13.05.029.
[
7]
K
.
J
a
r
oe
ns
ut
a
s
i
ne
e
,
M
.
J
a
r
oe
ns
ut
a
s
i
ne
e
,
P
.
B
oons
a
nong,
a
nd
E
.
S
pa
r
r
ow
,
“
M
a
ngos
t
e
e
n
yi
e
l
d
a
nd
f
r
ui
t
qua
l
i
t
y
unde
r
r
e
gul
a
t
e
d
i
r
r
i
ga
t
i
on w
i
t
h s
e
ns
or
s
a
nd
I
oT
,”
J
our
nal
of
H
um
an,
E
ar
t
h, and F
ut
ur
e
,
vol
. 5,
no. 2, pp. 205
–
215, J
un. 2024, doi
:
10.28991/
H
E
F
-
2024
-
05
-
02
-
05.
[
8]
D
.
K
e
ys
e
r
s
,
T
.
D
e
s
e
l
a
e
r
s
,
C
.
G
ol
l
a
n,
a
nd
H
.
N
e
y,
“
D
e
f
or
m
a
t
i
on
m
ode
l
s
f
or
i
m
a
ge
r
e
c
ogni
t
i
on,”
I
E
E
E
T
r
ans
ac
t
i
ons
on
P
at
t
e
r
n
A
nal
y
s
i
s
and M
ac
hi
ne
I
nt
e
l
l
i
ge
nc
e
, vol
. 29, no. 8, pp. 1422
–
1435, A
ug. 2007, doi
:
10.1109/
T
P
A
M
I
.2007.1153.
[
9]
Z
.
L
uo,
“
T
he
a
ppl
i
c
a
t
i
on
a
nd
de
v
e
l
opm
e
nt
of
i
m
a
ge
r
e
c
ogni
t
i
on
t
e
c
hnol
og
y
i
n
t
he
f
i
e
l
d
of
c
om
put
e
r
vi
s
i
on,”
H
i
ghl
i
ght
s
i
n
Sc
i
e
nc
e
, E
ngi
ne
e
r
i
ng and T
e
c
hnol
ogy
, vol
. 85, pp. 909
–
914,
M
a
r
. 2024, doi
:
10.54097/
g63vgv77.
[
10]
B
.
Z
hu,
“
R
e
t
a
i
l
c
om
m
odi
t
y
i
m
a
ge
r
e
c
ogni
t
i
on
ba
s
e
d
on
W
S
-
D
A
N
,”
i
n
2022
I
nt
e
r
nat
i
onal
C
onf
e
r
e
nc
e
on
M
ac
hi
ne
L
e
ar
ni
ng
and
I
nt
e
l
l
i
ge
nt
Sy
s
t
e
m
s
E
ngi
ne
e
r
i
ng (
M
L
I
SE
)
, A
ug. 2022, pp. 310
–
316
,
doi
:
10.110
9/
M
L
I
S
E
57402.2022.00068.
Evaluation Warning : The document was created with Spire.PDF for Python.