I
nte
rna
t
io
na
l J
o
urna
l o
f
E
lect
rica
l a
nd
Co
m
p
ute
r
E
ng
in
ee
ring
(
I
J
E
CE
)
Vo
l.
11
,
No
.
3
,
J
u
n
e
2021
,
p
p
.
2666
~
2
6
7
3
I
SS
N:
2088
-
8708
,
DOI
: 1
0
.
1
1
5
9
1
/
i
j
ec
e
.
v
11
i
3
.
pp
2
6
6
6
-
2
6
7
3
2666
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
ec
e.
ia
esco
r
e.
co
m
Pine
w
ilt
disea
se s
prea
ding
preven
ti
o
n sy
ste
m
using
s
e
m
a
nt
ic
seg
m
en
tatio
n
Chul
hy
u
n H
w
a
ng
1
,
J
a
ea
n J
e
o
ng
2
,
H
o
ek
y
un
g
J
un
g
3
1
De
p
a
rtme
n
t
o
f
S
m
a
rt
I
T
S
o
f
t
w
a
r
e
,
Ky
u
n
g
b
o
k
Un
iv
e
rsit
y
,
Na
m
y
a
n
g
ju
,
S
o
u
t
h
Ko
re
a
2
K
-
W
a
ter,
H
w
a
se
o
n
g
si,
Gy
e
o
n
g
g
id
o
,
S
o
u
t
h
Ko
re
a
3
De
p
a
rtme
n
t
o
f
Co
m
p
u
ter E
n
g
in
e
e
rin
g
,
P
a
iCh
a
i
Un
iv
e
rsit
y
,
Da
e
jeo
n
,
S
o
u
t
h
K
o
re
a
Art
icle
I
nfo
AB
ST
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
J
u
l 3
1
,
2
0
2
0
R
ev
i
s
ed
Sep
2
2
,
2
0
2
0
A
cc
ep
ted
Oct
1
4
,
2
0
2
0
P
i
n
e
w
il
t
d
ise
a
se
is
a
d
ise
a
se
th
a
t
a
ff
e
c
ts
e
c
o
s
y
ste
m
s
b
y
ra
p
id
ly
k
i
ll
in
g
tree
s
in
a
sh
o
rt
p
e
ri
o
d
o
f
ti
m
e
d
u
e
to
th
e
c
lo
se
in
tera
c
ti
o
n
b
e
tw
e
e
n
th
re
e
f
a
c
to
rs
su
c
h
a
s
tree
s,
m
e
d
iate
s,
a
n
d
p
a
th
o
g
e
n
s.
T
h
e
re
a
r
e
n
o
1
0
0
%
m
o
rtalit
y
in
f
e
c
ti
o
u
s
f
o
re
st
p
e
sts
.
A
c
c
o
r
d
in
g
to
th
e
Ko
re
a
F
o
re
st
S
e
r
v
ice
su
rv
e
y
,
a
s
o
f
A
p
ril
2
0
1
9
,
th
e
d
a
m
a
g
e
o
f
p
in
e
re
-
n
e
m
a
to
d
e
d
ise
a
se
w
a
s
a
b
o
u
t
4
9
0
,
0
0
0
d
e
a
d
tree
s
in
1
1
7
c
it
ies
,
c
o
u
n
ti
e
s
a
n
d
w
a
rd
s
a
c
ro
ss
th
e
c
o
u
n
try
.
It'
s
a
f
a
tal
c
o
n
d
i
ti
o
n
.
In
o
rd
e
r
t
o
p
re
v
e
n
t
th
is
p
ro
b
lem
,
th
is
p
a
p
e
r
p
ro
p
o
s
e
s
a
sy
ste
m
th
a
t
d
e
tec
ts
d
e
a
d
tree
s,
e
a
rly
in
f
e
c
ti
o
n
tree
s,
a
n
d
th
e
li
k
e
,
u
si
n
g
d
e
e
p
lea
rn
in
g
-
b
a
se
d
se
m
a
n
ti
c
se
g
m
e
n
tatio
n
.
I
n
a
d
d
i
ti
o
n
,
d
ro
n
e
s
w
e
re
u
se
d
to
p
h
o
t
o
g
ra
p
h
th
e
a
re
a
o
f
th
e
f
o
re
st,
a
n
d
a
se
p
a
ra
te
p
ix
e
l
se
g
m
e
n
tatio
n
lab
e
l
c
o
u
ld
b
e
u
se
d
to
i
d
e
n
ti
f
y
th
re
e
lev
e
ls
o
f
tran
s
m
i
ss
io
n
i
n
f
o
rm
a
ti
o
n
:
S
u
s
p
icio
n
,
a
tt
e
n
ti
o
n
,
a
n
d
c
o
n
f
irma
ti
o
n
.
T
h
is
a
ll
o
w
s
th
e
u
se
r
to
g
ra
sp
in
f
o
rm
a
ti
o
n
su
c
h
a
s
a
re
a
,
lo
c
a
ti
o
n
,
a
n
d
a
larm
to
p
re
v
e
n
t
th
e
sp
re
a
d
o
f
re
-
n
e
m
a
to
d
e
d
ise
a
se
.
K
ey
w
o
r
d
s
:
Dee
p
lear
n
in
g
Dr
o
n
e
P
in
e
w
il
t d
is
ea
s
e
Se
m
a
n
tic
s
e
g
m
e
n
tatio
n
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
:
Ho
ek
y
u
n
g
J
u
n
g
Dep
ar
t
m
en
t o
f
C
o
m
p
u
ter
E
n
g
i
n
ee
r
in
g
P
aich
ai
Un
i
v
er
s
it
y
155
-
40
B
ae
j
ae
-
r
o
,
Seo
g
u
,
Dae
J
eo
n
,
So
u
th
Ko
r
ea
E
m
ail:
h
k
j
u
n
g
@
p
c
u
.
ac
.
k
r
1.
I
NT
RO
D
UCT
I
O
N
R
ec
en
t
l
y
,
w
it
h
t
h
e
d
ev
elo
p
m
e
n
t
o
f
ar
ti
f
icial
i
n
telli
g
e
n
ce
,
r
esear
ch
o
n
d
ee
p
lear
n
in
g
tec
h
n
o
lo
g
y
u
s
i
n
g
s
atellite
p
h
o
to
g
r
ap
h
s
a
n
d
ae
r
i
al
p
h
o
to
g
r
ap
h
s
h
as
b
ee
n
ac
ti
v
el
y
co
n
d
u
cted
.
I
n
p
ar
ticu
lar
,
th
e
i
n
ter
est
in
t
h
e
tech
n
o
lo
g
y
f
o
r
r
ec
o
g
n
izi
n
g
o
b
j
ec
ts
b
y
u
s
i
n
g
i
m
a
g
es
o
r
p
h
o
to
s
h
a
s
b
ee
n
v
er
y
h
i
g
h
.
Ob
j
ec
t
d
etec
tio
n
i
s
o
n
e
o
f
th
e
d
ee
p
est
tech
n
o
lo
g
ies
f
o
r
d
ee
p
lear
n
in
g
t
h
at
ca
n
id
en
ti
f
y
m
u
ltip
le
o
b
j
ec
ts
in
o
n
e
i
m
ag
e
an
d
id
en
tify
th
e
ir
lo
c
atio
n
.
I
t
is
also
ap
p
lied
as
a
tech
n
o
lo
g
y
f
o
r
d
etec
ti
n
g
p
in
e
o
b
j
ec
ts
to
s
o
lv
e
p
in
e
r
e
-
n
e
m
ato
d
e
d
is
ea
s
e.
Ho
w
e
v
er
,
th
e
e
x
is
ti
n
g
s
y
s
te
m
is
an
o
b
j
ec
t
r
ec
o
g
n
itio
n
-
b
ased
s
y
s
te
m
th
at
lear
n
s
w
h
et
h
er
a
p
in
e
tr
ee
is
i
n
f
ec
ted
b
y
lear
n
i
n
g
a
s
p
ec
i
f
ic
p
ar
t
o
f
t
h
e
i
m
ag
e,
b
u
t
it
i
s
ef
f
ec
t
iv
e
i
n
s
i
m
p
l
y
r
ec
o
g
n
izi
n
g
a
n
o
b
j
ec
t
an
d
d
eter
m
i
n
in
g
a
tar
g
et.
I
t
is
m
o
r
e
e
f
f
icie
n
t
th
a
n
th
e
e
x
i
s
ti
n
g
s
y
s
te
m
.
I
n
ad
d
itio
n
,
it
s
h
o
w
ed
h
i
g
h
ac
c
u
r
ac
y
in
r
elati
v
el
y
c
lo
s
e
p
ictu
r
es,
b
u
t lo
w
ac
cu
r
ac
y
in
h
ig
h
-
al
tit
u
d
e
ae
r
ial
an
d
s
atellite
p
h
o
to
g
r
a
p
h
s
[
1
-
7
]
.
I
n
o
r
d
er
to
s
o
lv
e
th
is
p
r
o
b
le
m
,
th
is
p
ap
er
p
r
o
p
o
s
es
a
p
in
e
r
e
n
e
m
a
to
d
e
s
p
r
ea
d
in
g
p
r
ev
e
n
ti
o
n
s
y
s
te
m
u
s
i
n
g
s
e
m
a
n
tic
s
p
litt
i
n
g
.
Se
m
an
tic
s
eg
m
e
n
tatio
n
is
a
f
ield
o
f
o
b
j
ec
t
d
etec
tio
n
th
at
r
ec
o
g
n
i
ze
s
an
d
lab
el
s
a
s
e
t
o
f
p
ix
el
s
co
n
s
tit
u
ti
n
g
d
if
f
er
en
t
c
ateg
o
r
ies.
C
o
n
v
en
tio
n
al
o
b
j
ec
t
d
etec
tio
n
d
etec
ts
o
b
j
ec
ts
b
ased
o
n
w
h
at
ar
e
in
s
id
e
a
s
p
ec
i
f
ic
b
o
u
n
d
in
g
b
o
x
,
b
u
t
s
e
m
an
tic
s
e
g
m
en
ta
ti
o
n
is
m
o
r
e
p
r
ec
is
e
t
h
an
o
t
h
er
f
o
r
m
s
o
f
o
b
j
ec
t
d
etec
tio
n
b
ec
au
s
e
i
t
u
s
es
a
s
e
g
m
e
n
tat
io
n
lab
el
ap
p
lied
to
p
ix
els
i
n
t
h
e
i
m
a
g
e.
I
n
t
h
is
p
ap
er
,
a
s
ep
ar
ate
p
ix
el
s
eg
m
e
n
tatio
n
lab
el
w
a
s
ap
p
li
ed
to
th
e
p
in
e
r
e
-
n
e
m
ato
d
e
d
ata
s
et
p
h
o
to
g
r
ap
h
ed
b
y
t
h
e
d
r
o
n
e
to
id
en
tify
t
h
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8708
P
in
e
w
ilt d
is
ea
s
e
s
p
r
ea
d
in
g
p
r
ev
en
tio
n
s
ystem
u
s
in
g
s
ema
n
tic
s
eg
men
ta
tio
n
(
C
h
u
lh
yu
n
Hw
a
n
g
)
2667
th
r
ee
s
ta
g
es
o
f
i
n
f
ec
tio
u
s
in
f
o
r
m
atio
n
s
u
c
h
as
s
u
s
p
icio
n
,
att
en
tio
n
,
a
n
d
co
n
f
ir
m
atio
n
.
T
h
is
allo
w
s
th
e
u
s
er
to
g
r
asp
in
f
o
r
m
atio
n
s
u
ch
a
s
ar
ea
,
lo
ca
tio
n
,
an
d
alar
m
to
p
r
ev
en
t th
e
s
p
r
ea
d
o
f
r
e
-
n
e
m
ato
d
e
d
is
ea
s
e.
2.
DE
T
RE
C
T
I
O
N
O
F
SU
SPE
CT
E
D
P
I
NE
WI
L
T
D
I
S
E
AS
E
2
.
1
.
Do
ub
t
det
ec
t
io
n w
it
h su
perv
is
ed
cla
s
s
if
ica
t
io
n
S.
K.
L
ee
et
a
l
.
[
8
]
ac
q
u
ir
ed
h
ig
h
r
eso
l
u
tio
n
u
n
m
a
n
n
ed
ae
r
ial
v
eh
icle
(
U
A
V)
i
m
a
g
es
i
n
r
elap
s
in
g
ar
ea
s
o
f
p
in
e
n
e
m
ato
d
es,
a
n
d
co
n
d
u
c
ted
s
u
s
p
icio
n
s
o
f
s
u
s
p
ec
ted
p
in
e
n
e
m
a
to
d
es
th
r
o
u
g
h
s
u
p
er
v
is
ed
class
i
f
icatio
n
tech
n
iq
u
e
s
s
u
c
h
as
SVM
an
d
A
NN.
B
ased
o
n
th
e
ac
q
u
ir
ed
lear
n
in
g
d
ata,
w
e
ap
p
lied
t
w
o
tech
n
iq
u
es,
ANN
a
n
d
SVM,
t
o
ca
lcu
late
9
9
.
4
3
%
an
d
8
4
.
6
0
%
o
f
u
s
er
ac
c
u
r
ac
y
a
n
d
m
an
u
f
ac
tu
r
er
ac
c
u
r
ac
y
u
s
i
n
g
A
NN,
an
d
9
9
.
2
4
% a
n
d
9
5
.
3
5
% a
cc
u
r
ac
y
u
s
i
n
g
S
VM
.
2
.
2
.
E
x
t
ra
ct
ing
s
u
s
pect
s
pa
ce
lo
c
a
t
io
ns
us
ing
R
G
B
v
a
lues
T
o
id
en
tify
t
h
e
d
a
m
ag
e
o
f
p
i
n
e
r
e
-
n
e
m
ato
d
e
d
is
ea
s
e,
th
e
e
x
p
er
im
e
n
t
s
w
er
e
co
n
d
u
cted
b
y
s
p
ec
if
y
in
g
th
e
d
a
m
a
g
e
R
GB
v
a
lu
e
s
(
R
ed
1
3
9
,
Gr
ee
n
1
1
3
,
B
lu
e
1
2
3
)
a
n
d
in
cr
ea
s
i
n
g
th
e
r
an
g
e
v
alu
e
s
co
r
r
esp
o
n
d
in
g
t
o
R
GB
v
alu
e
s
f
r
o
m
1
to
1
0
[9
-
1
2
]
.
I
n
ad
d
itio
n
,
w
h
en
a
clu
s
t
er
in
g
p
h
e
n
o
m
e
n
o
n
o
cc
u
r
s
a
m
o
n
g
th
e
ca
lc
u
lated
p
ix
els,
it
is
r
ec
o
g
n
ized
as
an
o
b
j
ec
t
as
a
c
o
lo
n
y
.
T
h
e
p
u
r
p
o
s
e
o
f
th
is
s
t
u
d
y
is
to
p
r
esen
t
an
e
x
tr
ac
tio
n
m
et
h
o
d
o
lo
g
y
t
h
at
ca
n
e
x
tr
ac
t
d
am
a
g
e
s
u
s
i
n
g
R
GB
v
al
u
es
o
f
i
m
a
g
es
ta
k
en
at
h
i
g
h
altit
u
d
es.
Ho
w
ev
er
,
d
u
e
to
th
e
ch
ar
ac
ter
is
t
ics
o
f
R
GB
v
al
u
es,
it
is
d
if
f
ic
u
lt
to
ca
lcu
late
ce
r
tain
ac
cu
r
ac
y
s
i
m
p
l
y
b
y
c
h
an
g
i
n
g
th
e
r
an
g
e
o
f
R
GB
v
al
u
es,
s
u
c
h
as
t
h
e
R
G
B
v
alu
e
s
v
ar
y
in
g
d
u
e
to
v
ar
i
o
u
s
f
ac
to
r
s
s
u
ch
a
s
w
ea
t
h
er
co
n
d
itio
n
s
,
s
h
ad
o
w
p
r
esen
ce
,
s
h
o
o
ti
n
g
h
ei
g
h
t,
a
n
d
p
ix
el
s
ize
[
1
3
-
1
4
]
.
3.
SYST
E
M
DE
SI
G
N
3
.
1
.
Da
t
a
s
et
prepro
ce
s
s
ing
T
h
e
d
ata
s
et
u
s
ed
in
t
h
is
s
tu
d
y
is
an
ae
r
ial
p
h
o
to
g
r
ap
h
w
i
th
b
o
th
in
f
ec
ted
an
d
n
o
r
m
al
tr
ee
s
.
T
h
e
p
in
e
tr
ee
in
f
ec
ted
w
it
h
t
h
e
p
i
n
e
r
e
-
n
e
m
ato
d
e
is
c
h
ar
ac
ter
ized
b
y
b
r
o
w
n
in
g
a
n
d
cl
u
s
ter
i
n
g
.
C
l
u
s
ter
in
g
i
s
a
ch
ar
ac
ter
is
tic
o
f
r
e
-
n
e
m
ato
d
e
d
is
ea
s
e,
it
s
p
r
ea
d
s
t
h
e
m
ed
ia
t
o
s
u
r
r
o
u
n
d
in
g
p
i
n
es,
s
o
it
ca
n
b
e
co
n
f
ir
m
ed
t
h
a
t
clu
s
ter
i
n
g
p
h
en
o
m
en
o
n
w
ith
s
u
r
r
o
u
n
d
i
n
g
p
in
e
s
o
cc
u
r
s
ar
o
u
n
d
th
e
ea
r
l
y
tr
ee
o
f
in
f
ec
t
io
n
.
T
h
er
ef
o
r
e,
th
e
d
ata
s
et
w
as p
r
ep
r
o
ce
s
s
ed
u
s
i
n
g
t
h
i
s
f
ea
t
u
r
e.
As
s
h
o
w
n
in
t
h
e
cir
cle
in
Fig
u
r
e
1
,
th
e
f
o
cu
s
i
n
g
ef
f
ec
t
an
d
n
o
is
e
r
ed
u
ctio
n
ar
e
p
er
f
o
r
m
ed
ar
o
u
n
d
th
e
b
r
o
w
n
ed
an
d
cl
u
s
ter
ed
g
r
o
u
p
to
i
m
p
r
o
v
e
th
e
ac
c
u
r
ac
y
o
f
t
h
e
d
ata
s
et.
I
n
ad
d
itio
n
,
t
h
e
n
o
i
s
e
g
e
n
er
ated
d
u
r
in
g
th
e
p
r
o
ce
s
s
is
r
e
m
o
v
ed
to
i
m
p
r
o
v
e
th
e
q
u
alit
y
o
f
th
e
d
ata
s
e
t.
A
s
t
h
e
p
r
ep
r
o
ce
s
s
in
g
o
f
t
h
e
i
m
a
g
e
p
r
o
ce
ed
s
,
it
ca
n
b
e
d
is
tin
g
u
is
h
ed
b
y
t
h
e
u
n
aid
ed
e
y
e
an
d
th
e
b
o
u
n
d
ar
y
w
it
h
th
e
n
o
r
m
al
tr
ee
b
ec
o
m
es
clea
r
.
T
o
ev
alu
ate
th
e
ac
cu
r
ac
y
o
f
t
h
e
d
ata
s
et
p
r
ep
r
o
ce
s
s
in
g
,
t
h
e
d
ataset
w
as
s
to
r
ed
in
a
s
ep
ar
ate
d
ir
ec
to
r
y
an
d
co
m
p
ar
ed
w
it
h
th
e
i
m
ag
e
w
i
th
o
u
t p
r
ep
r
o
ce
s
s
in
g
.
Fig
u
r
e
1
.
Data
s
et
f
o
cu
s
i
n
g
3
.
2
.
O
v
er
a
ll
s
y
s
t
e
m
de
s
ig
n
T
h
e
p
u
r
p
o
s
e
o
f
th
is
s
y
s
te
m
i
s
to
p
r
ev
en
t
th
e
s
p
r
ea
d
o
f
p
in
e
n
e
m
a
to
d
e
d
is
ea
s
e.
I
t
aim
s
to
d
etec
t
p
in
e
tr
ee
in
f
ec
ted
w
i
th
r
e
-
n
e
m
ato
d
e
d
is
ea
s
e
an
d
i
s
o
late
i
t
f
r
o
m
n
o
r
m
al
tr
e
e
b
ased
o
n
t
h
e
d
a
t
a
s
e
t
t
a
k
e
n
b
y
a
e
r
i
al
p
h
o
t
o
g
r
a
p
h
.
T
h
e
r
e
f
o
r
e
,
t
h
e
a
r
e
a
o
f
i
n
f
e
c
t
e
d
t
r
e
e
s
,
t
h
e
a
r
e
a
o
f
n
o
r
m
a
l
t
r
e
e
s
,
a
n
d
t
h
e
a
r
e
a
o
f
s
u
s
p
e
c
t
e
d
s
u
s
p
i
c
i
o
n
s
a
r
e
r
e
q
u
i
r
e
d
.
T
h
e
p
r
o
p
o
s
e
d
s
y
s
t
e
m
a
p
p
l
i
e
s
t
h
e
p
r
e
p
r
o
c
e
s
s
e
d
d
a
t
a
s
e
t
t
o
t
h
e
s
e
m
a
n
t
i
c
p
a
r
t
i
t
i
o
n
i
n
g
m
o
d
el
b
ased
o
n
Seg
Net
a
n
d
r
etu
r
n
s
ea
ch
ar
ea
.
I
t a
ls
o
v
is
u
alize
s
t
h
e
ar
ea
r
etu
r
n
ed
an
d
p
r
o
v
id
es it to
th
e
u
s
er
F
ig
u
r
e
2
.
T
h
e
s
y
s
te
m
is
a
t
w
o
-
s
tag
e
s
tr
u
ctu
r
e.
T
h
e
f
ir
s
t
s
tep
is
to
co
llect,
class
if
y
a
n
d
co
r
r
ec
t
th
e
d
ata.
T
h
e
p
r
o
ce
s
s
is
as:
I
n
t
h
e
co
llecti
n
g
p
h
a
s
e,
th
e
u
s
er
co
llects
i
m
a
g
e
d
ata
o
f
t
h
e
ar
ea
s
u
s
p
ec
ted
o
f
p
in
e
r
e
-
n
e
m
ato
d
e
d
is
ea
s
e
Fi
g
u
r
e
3
.
C
o
llected
o
b
j
ec
ts
ca
n
ef
f
ec
ti
v
el
y
co
llect
a
w
id
e
r
an
g
e
o
f
ae
r
ial
p
h
o
to
g
r
ap
h
s
a
n
d
s
atellite
i
m
a
g
es.
I
n
t
h
e
clas
s
i
f
icatio
n
p
h
ase,
y
o
u
cr
ea
te
a
tr
ain
i
n
g
s
e
t
b
y
lab
elin
g
i
m
a
g
es
s
o
t
h
at
y
o
u
ca
n
clas
s
i
f
y
ar
ea
s
b
ased
o
n
th
e
co
llected
d
ata
s
e
t
.
T
h
e
cr
iter
ia
to
class
if
y
t
h
e
ar
ea
is
a
lab
el
clas
s
i
f
ied
b
ased
o
n
t
h
e
R
GB
v
al
u
e
o
f
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
n
t J
E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
11
,
No
.
3
,
J
u
n
e
2021
:
2
6
6
6
-
2673
2668
th
e
i
m
a
g
e.
Sin
ce
t
h
e
cla
s
s
i
f
ica
tio
n
tar
g
ets
o
f
th
is
s
y
s
te
m
ar
e
in
f
ec
tio
n
tr
ee
s
,
s
u
s
p
icio
n
s
,
an
d
n
o
r
m
al
tr
ee
s
,
t
h
e
y
ar
e
d
ef
in
ed
b
ased
o
n
th
e
cla
s
s
i
f
icatio
n
tar
g
et
s
.
I
n
th
e
co
r
r
ec
ti
o
n
s
tep
,
th
e
n
o
is
e
is
r
e
m
o
v
ed
f
r
o
m
t
h
e
i
m
ag
e
a
n
d
th
e
f
o
cu
s
is
ap
p
lied
to
t
h
e
i
n
f
ec
ted
tr
ee
to
i
m
p
r
o
v
e
t
h
e
o
v
er
all
ac
cu
r
ac
y
.
T
h
e
s
ec
o
n
d
s
tep
is
to
lear
n
th
e
d
at
a
s
et
an
d
t
h
e
tr
ain
in
g
s
et,
a
n
d
to
d
er
iv
e
th
e
r
es
u
lt
s
.
I
t is d
iv
id
ed
in
to
m
o
d
eli
n
g
,
s
eg
m
e
n
tatio
n
,
an
d
o
u
tp
u
t p
h
ases
.
I
n
th
e
m
o
d
eli
n
g
s
ta
g
e,
m
o
d
el
s
b
ased
o
n
Seg
Net
ca
n
b
e
s
elec
ted
an
d
ap
p
lied
.
I
n
th
e
s
eg
m
en
tatio
n
s
tep
,
th
e
i
m
a
g
e
is
s
eg
m
e
n
ted
b
ased
o
n
th
e
lear
n
ed
r
esu
lt
s
.
T
h
e
o
u
tp
u
t
s
tep
d
er
iv
es
th
e
v
i
s
u
alize
d
i
m
ag
e
b
ased
o
n
th
e
d
iv
id
ed
i
m
a
g
e
in
t
h
e
d
i
v
id
in
g
s
tep
an
d
r
et
u
r
n
s
th
e
v
al
u
e
o
f
t
h
e
d
i
v
id
ed
ar
ea
.
I
t a
l
s
o
r
etu
r
n
s
th
e
ac
cu
r
ac
y
o
f
t
h
e
s
eg
m
e
n
ted
ar
ea
an
d
p
r
o
v
id
es it to
th
e
u
s
er
.
Fig
u
r
e
2
.
S
y
s
te
m
d
ia
g
r
a
m
Fig
u
r
e
3
.
S
y
s
te
m
f
lo
w
ch
ar
t
3
.
3
.
I
m
a
g
e
cla
s
s
if
ica
t
io
n a
re
a
def
ini
t
io
n
Dee
p
lear
n
in
g
m
o
d
el
i
s
a
m
o
d
el
th
at
m
ax
i
m
ize
s
t
h
e
ef
f
ec
t
w
h
e
n
b
u
ilt
o
n
a
lar
g
e
a
m
o
u
n
t
o
f
d
ata
.
T
h
er
ef
o
r
e,
it
is
i
m
p
o
r
tan
t
to
s
ec
u
r
e
a
lar
g
e
a
m
o
u
n
t
o
f
d
ata
in
s
e
m
a
n
tic
p
ar
titi
o
n
i
n
g
.
I
n
ad
d
itio
n
,
s
i
n
ce
s
e
m
a
n
tic
s
e
g
m
e
n
tatio
n
co
n
f
i
g
u
r
es
a
d
ata
s
et
b
y
ap
p
l
y
in
g
a
l
ab
el
ac
co
r
d
in
g
to
a
class
if
icat
i
o
n
tar
g
et,
an
i
m
a
g
e
to
b
e
class
if
ied
m
u
s
t b
e
lab
eled
b
y
ca
teg
o
r
y
.
I
m
ag
e
lab
eli
n
g
w
a
s
d
o
n
e
u
s
i
n
g
M
A
T
L
A
B
's
I
m
ag
e
L
ab
eler
.
Af
ter
lo
ad
in
g
t
h
e
i
m
a
g
e,
s
p
ec
i
f
y
a
n
ar
ea
an
d
a
lin
e
to
ass
ig
n
a
lab
el
ar
ea
to
th
e
im
a
g
e.
T
h
er
ef
o
r
e,
th
e
class
i
f
icatio
n
tar
g
ets
w
er
e
lab
eled
as
in
f
ec
tio
u
s
,
s
u
s
p
icio
u
s
a
n
d
n
o
r
m
al.
I
n
ad
d
itio
n
,
in
f
ec
ted
tr
ee
s
,
s
u
s
p
icio
u
s
tr
ee
s
,
an
d
n
o
r
m
al
tr
ee
s
w
er
e
lab
eled
an
d
lab
eled
as
o
r
an
g
e,
b
lu
e,
an
d
y
ello
w
,
r
esp
ec
tiv
el
y
.
T
h
is
d
esi
g
n
ates
th
e
lab
eled
d
ata
s
et
as
th
e
tr
ain
i
n
g
s
et
an
d
th
e
p
ar
titi
o
n
u
s
ed
f
o
r
tr
ain
in
g
.
3
.
4
.
M
o
del
s
elec
t
io
n a
nd
t
ra
ini
ng
Se
m
a
n
tic
p
ar
titi
o
n
i
n
g
is
b
a
s
ed
o
n
t
h
e
e
n
co
d
er
-
d
ec
o
d
e
r
ar
ch
itect
u
r
e.
E
ac
h
e
n
co
d
er
h
as
a
d
o
w
n
s
a
m
p
li
n
g
d
ec
o
d
er
an
d
u
p
s
a
m
p
li
n
g
t
h
e
s
tr
u
ct
u
r
e.
Am
o
n
g
th
e
m
o
d
el
s
w
i
th
t
h
e
ab
o
v
e
s
tr
u
ctu
r
e,
r
ep
r
esen
tativ
e
m
o
d
el
s
w
ith
h
i
g
h
ac
c
u
r
ac
y
ar
e
FC
N,
Un
et
a
n
d
Dee
p
L
ab
V3
,
w
h
ich
ar
e
s
i
m
ilar
to
th
e
ex
i
s
ti
n
g
Seg
Net,
b
u
t
w
it
h
h
ig
h
ac
c
u
r
a
c
y
i
n
a
s
p
ec
i
f
ic
i
m
a
g
e.
T
h
er
ef
o
r
e,
in
th
is
p
ap
er
,
w
e
d
es
i
g
n
ed
to
ev
al
u
ate
t
h
e
ac
cu
r
ac
y
u
s
i
n
g
t
h
e
test
i
m
a
g
e
s
et
tr
ain
ed
an
d
p
r
ep
ar
ed
b
a
s
ed
o
n
th
e
tr
ain
i
n
g
d
ata
s
et
p
r
ep
ar
e
d
f
o
r
all
th
e
ab
o
v
e
m
o
d
els to
co
m
p
ar
e
th
e
ac
cu
r
ac
y
i
n
a
s
p
ec
if
ic
i
m
a
g
e
[
1
5
-
1
7
]
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8708
P
in
e
w
ilt d
is
ea
s
e
s
p
r
ea
d
in
g
p
r
ev
en
tio
n
s
ystem
u
s
in
g
s
ema
n
tic
s
eg
men
ta
tio
n
(
C
h
u
lh
yu
n
Hw
a
n
g
)
2669
3
.
5
.
I
m
a
g
e
re
g
io
n seg
m
e
nta
t
io
n
I
m
ag
e
s
e
g
m
en
ta
tio
n
is
th
e
s
e
g
m
e
n
tat
io
n
o
f
te
s
t
d
ata
u
s
in
g
tr
ain
ed
w
e
ig
h
ts
a
f
ter
m
o
d
el
s
elec
tio
n
.
C
las
s
i
f
icatio
n
ar
ea
s
ar
e
d
e
f
in
ed
in
a
s
ep
ar
ate
C
SV
f
ile
to
p
r
o
v
id
e
v
i
s
u
aliza
tio
n
e
f
f
ec
t
s
to
u
s
er
s
.
Fi
g
u
r
e
4
s
h
o
w
s
t
h
e
i
m
a
g
e
in
s
id
e
th
e
C
S
V
f
ile
to
d
ef
i
n
e
th
e
R
GB
co
lo
r
.
Fig
u
r
e
4
.
T
h
e
C
SV f
ile
to
d
ef
i
n
e
th
e
R
GB
co
lo
r
I
n
f
ec
ted
,
s
u
s
p
ec
ted
i
n
f
ec
t
io
n
,
an
d
u
n
in
f
ec
ted
ar
e
d
ef
i
n
ed
to
b
e
in
f
ec
ted
,
s
u
s
p
icio
u
s
,
an
d
n
o
r
m
al,
r
esp
ec
tiv
el
y
,
an
d
m
ar
k
ed
i
n
r
e
d
,
y
e
llo
w
,
an
d
g
r
ee
n
.
T
h
e
o
th
er
Sk
y
an
d
So
d
m
ea
n
s
k
y
a
n
d
ea
r
th
,
r
esp
ec
tiv
e
l
y
.
Vo
id
m
ea
n
s
a
n
u
n
s
p
ec
if
ied
ar
ea
an
d
is
m
ar
k
ed
in
b
lack
.
3
.
6
.
F
CN
des
ig
n
T
h
e
b
ig
g
est
f
ea
t
u
r
e
o
f
F
C
N
i
s
th
e
lac
k
o
f
a
f
u
ll
y
co
n
n
ec
ted
la
y
er
.
C
h
an
g
e
t
h
e
f
u
l
l
y
co
n
n
e
cted
la
y
er
to
1
x
1
co
n
v
o
lu
tio
n
la
y
er
in
t
h
e
ex
is
ti
n
g
C
NN
s
tr
u
ct
u
r
e.
T
h
e
r
ea
s
o
n
f
o
r
d
eletin
g
t
h
e
f
u
l
l
y
c
o
n
n
ec
ted
la
y
er
is
to
o
b
tain
lo
ca
tio
n
i
n
f
o
r
m
atio
n
.
T
h
e
k
e
y
to
s
e
m
an
tic
s
e
g
m
e
n
t
atio
n
i
s
to
g
et
lo
ca
tio
n
i
n
f
o
r
m
atio
n
b
ec
au
s
e
t
h
e
lo
ca
tio
n
i
n
f
o
r
m
a
tio
n
d
i
s
ap
p
ea
r
s
th
r
o
u
g
h
t
h
e
f
u
ll
y
co
n
n
ec
ted
la
y
er
.
T
h
er
ef
o
r
e,
th
e
last
la
y
e
r
w
a
s
r
ep
lace
d
w
it
h
a
1
x
1
co
n
v
o
lu
tio
n
la
y
er
i
n
t
h
e
VGG1
6
-
b
ased
m
o
d
el.
W
e
t
h
en
u
p
s
a
m
p
led
u
s
i
n
g
tr
an
s
p
o
s
ed
co
n
v
o
lu
tio
n
a
n
d
cr
ea
ted
a
m
ap
t
h
e
s
a
m
e
s
ize
a
s
t
h
e
i
n
p
u
t.
Als
o
,
i
n
t
h
e
u
p
s
a
m
p
lin
g
p
r
o
ce
s
s
,
f
ea
t
u
r
e
m
ap
s
o
f
lo
w
er
la
y
er
s
w
er
e
ad
d
ed
to
ex
tr
ac
t
f
ea
tu
r
e
s
b
y
co
m
b
i
n
i
n
g
th
e
m
w
ith
f
ea
t
u
r
e
m
ap
s
ab
o
v
e.
Af
te
r
t
h
at,
all
th
e
f
ea
tu
r
e
s
w
er
e
r
ec
o
m
b
i
n
ed
th
r
o
u
g
h
t
h
e
s
k
ip
c
o
n
n
ec
tio
n
to
d
er
iv
e
th
e
r
es
u
lts
[
9
]
.
3
.
7
.
Dee
pL
a
bV3
de
s
ig
n
Dee
p
L
ab
V3
is
b
ased
o
n
R
es
Net
lear
n
ed
f
r
o
m
I
m
a
g
eNe
t.
R
esNet
f
ea
t
u
r
e
ex
tr
ac
to
r
s
ca
n
b
e
u
s
ed
to
ex
tr
ac
t
f
ea
tu
r
es
o
f
v
ar
y
i
n
g
s
iz
es.
At
t
h
e
co
r
e
o
f
Dee
p
L
ab
V3
is
th
e
A
tr
u
o
s
c
o
n
v
o
l
u
t
i
o
n
l
a
y
e
r
.
Y
o
u
c
a
n
d
r
i
l
l
h
o
l
e
s
in
a
t
y
p
ical
co
n
v
o
l
u
tio
n
la
y
er
to
ca
p
tu
r
e
lar
g
er
f
ea
tu
r
es
w
ith
th
e
s
a
m
e
a
m
o
u
n
t
o
f
co
m
p
u
ta
tio
n
.
T
h
er
ef
o
r
e,
in
th
is
p
ap
er
,
w
e
d
esi
g
n
ed
th
e
A
t
r
o
u
s
co
n
v
o
l
u
tio
n
la
y
er
s
i
n
p
ar
allel
to
ca
p
tu
r
e
m
o
r
e
f
ea
t
u
r
es [
1
8
-
2
2
]
.
3
.
8
.
U
-
net
des
ig
n
U
-
Net
is
a
U
-
s
h
ap
ed
n
et
w
o
r
k
w
it
h
r
ep
ea
ted
co
n
tr
ac
ti
n
g
la
y
e
r
s
an
d
u
p
s
ca
li
n
g
.
W
e
ca
n
s
ee
t
h
at
U
-
Net
d
iv
id
es
t
h
e
i
m
ag
e
i
n
to
s
e
v
er
a
l
i
m
a
g
es
w
h
ile
g
r
ad
u
al
l
y
r
ed
u
cin
g
t
h
e
s
ize
o
f
t
h
e
i
m
a
g
e
d
u
r
in
g
th
e
co
n
tr
ac
t
p
r
o
ce
s
s
.
T
h
er
ea
f
ter
,
th
e
u
p
s
a
m
p
lin
g
p
r
o
ce
s
s
is
p
er
f
o
r
m
ed
t
o
o
b
tain
th
e
d
i
v
id
ed
i
m
ag
e.
I
n
th
i
s
p
ap
er
,
U
-
Net
is
d
esig
n
ed
b
y
ad
o
p
tin
g
t
h
e
ab
o
v
e
s
tr
u
ct
u
r
e
[
2
3
-
2
5
]
.
4.
SYST
E
M
I
M
P
L
E
M
E
NT
A
T
I
O
N
T
h
is
ch
ap
ter
d
ea
ls
w
ith
th
e
i
m
p
le
m
en
ta
t
io
n
o
f
a
p
i
n
e
n
e
m
ato
d
e
d
is
ea
s
e
p
r
ev
en
tio
n
s
y
s
te
m
.
I
n
ad
d
itio
n
,
w
e
v
er
i
f
y
t
h
e
e
f
f
ec
t
iv
en
e
s
s
o
f
th
e
p
r
o
p
o
s
ed
s
y
s
t
e
m
t
h
r
o
u
g
h
co
n
s
id
er
atio
n
.
T
h
e
s
y
s
te
m
b
u
ild
i
n
g
en
v
ir
o
n
m
e
n
t
is
s
h
o
w
n
i
n
T
a
b
le
1
.
T
h
is
p
ap
er
b
u
ild
s
An
ac
o
n
d
a
4
.
5
.
1
1
en
v
ir
o
n
m
en
t
in
W
in
d
o
w
1
0
P
r
o
en
v
i
r
o
n
m
e
n
t
to
b
u
ild
d
ee
p
lear
n
in
g
e
n
v
ir
o
n
m
e
n
t.
GP
U
ac
c
eler
atio
n
is
u
s
ed
to
i
m
p
r
o
v
e
t
r
ain
in
g
s
p
ee
d
,
an
d
C
UD
A
v
er
s
io
n
is
1
0
.
0
.
A
ls
o
,
t
h
e
i
m
p
le
m
e
n
tat
io
n
o
f
th
e
s
y
s
t
e
m
u
s
es
T
en
s
er
f
lo
w
1
.
1
4
.
0
b
a
s
ed
o
n
P
y
t
h
o
n
,
a
n
d
th
e
lib
r
ar
y
d
ep
en
d
en
cie
s
ar
e
Op
en
C
V
a
n
d
n
u
m
p
y
.
T
h
e
d
ev
elo
p
m
e
n
t
to
o
l
u
s
ed
P
y
C
h
ar
m
an
d
M
A
T
L
A
B
's
I
m
ag
e
L
ab
eler
to
lab
el
th
e
tr
ai
n
in
g
s
e
t.
T
ab
le
1
.
Sy
s
te
m
d
ev
e
lo
p
m
en
t e
n
v
ir
o
n
m
e
n
t
Ty
p
e
C
o
mp
o
si
t
i
o
n
D
e
sk
t
o
p
S
p
e
c
s
C
P
U
:
I
n
t
e
l
C
o
r
e
i
7
-
4
6
7
0
@
3
.
4
0
G
H
z
,
R
A
M
:
1
6
GB
G
P
U
:
G
e
f
o
r
c
e
G
TX
1
0
6
0
t
i
4
GB
,
O
S
:
W
i
n
d
o
w
1
0
P
r
o
En
v
i
r
o
n
me
n
A
n
a
c
o
n
d
a
4
.
5
.
1
1
,
P
y
t
h
o
n
3
,
C
U
D
A
1
0
.
0
L
i
b
r
a
r
y
T
e
n
se
r
f
l
o
w
1
.
1
4
.
0
,
O
p
e
n
C
V
,
n
u
mp
y
D
e
v
e
l
o
p
me
n
t
T
o
o
l
s
P
y
C
h
a
r
m,
M
A
TL
A
B
I
mag
e
L
a
b
e
l
e
r
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
n
t J
E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
11
,
No
.
3
,
J
u
n
e
2021
:
2
6
6
6
-
2673
2670
4
.
1
.
Sy
s
t
e
m
i
m
p
le
m
ent
a
t
io
n
T
h
e
s
y
s
te
m
tr
ai
n
s
th
e
s
e
m
a
n
tic
s
e
g
m
en
tatio
n
n
et
w
o
r
k
m
o
d
el
b
ased
o
n
t
h
e
lab
eled
d
at
a
s
et
an
d
r
etu
r
n
s
i
n
f
o
r
m
atio
n
s
u
c
h
as
s
u
s
p
icio
n
,
n
o
r
m
al
tr
ee
,
an
d
in
f
ec
ted
tr
ee
o
f
th
e
test
s
et
u
s
in
g
th
e
w
eig
h
t.
P
r
o
v
id
es
th
e
u
s
er
w
it
h
a
v
is
u
alize
d
i
m
a
g
e
an
d
r
etu
r
n
s
th
e
ar
ea
o
f
th
e
s
eg
m
e
n
ted
ar
ea
.
T
h
e
in
ter
n
al
d
ir
ec
to
r
y
is
d
iv
id
ed
in
to
a
d
ata
s
et
d
ir
ec
to
r
y
an
d
a
m
o
d
el
d
ir
ec
to
r
y
.
W
h
en
y
o
u
r
u
n
tr
ain
.
p
y
an
d
s
e
lect
a
m
o
d
el,
it
tr
ain
s
b
ased
o
n
th
e
tr
ain
i
n
g
s
et
s
to
r
ed
in
th
e
'
.
.
/d
a
ta
S
et/tr
a
in
_
La
b
el'
d
ir
ec
t
o
r
y
.
tes
t.p
y
p
er
f
o
r
m
s
s
e
m
a
n
tic
p
ar
titi
o
n
in
g
o
f
th
e
d
ata
s
ets
in
th
e
'.
.
/d
a
ta
S
et/test
'
d
ir
ec
to
r
y
b
ased
o
n
th
e
tr
ain
i
n
g
w
eig
h
t
s
tr
ain
ed
in
t
h
e
ab
o
v
e
p
r
o
ce
s
s
.
T
h
e
r
esu
lt
is
s
to
r
ed
in
test
_
L
ab
el.
T
o
ap
p
ly
s
e
m
a
n
tic
s
p
lit
tin
g
,
y
o
u
n
e
ed
a
lab
ele
d
tr
ain
in
g
s
et.
I
n
t
h
is
p
ap
er
,
a
lab
ele
d
i
m
a
g
e
w
as
cr
ea
ted
b
ased
o
n
s
u
s
p
icio
u
s
,
n
o
r
m
al
a
n
d
i
n
f
ec
t
ed
tr
ee
s
,
an
d
tr
ain
i
n
g
w
ei
g
h
t
s
w
er
e
g
en
er
ated
b
y
ap
p
ly
i
n
g
s
e
m
an
tic
s
e
g
m
e
n
tat
i
o
n
n
et
w
o
r
k
m
o
d
els.
Fig
u
r
e
5
s
h
o
w
s
th
e
d
ata
s
et
s
to
r
ed
in
th
e
te
s
t
d
ir
ec
to
r
y
.
On
ce
all
t
h
e
tr
ain
in
g
w
ei
g
h
ts
h
av
e
b
ee
n
g
en
er
at
ed
,
r
u
n
te
s
t.p
y
.
Af
ter
s
elec
tin
g
th
e
m
o
d
el
an
d
i
m
p
o
r
t
in
g
t
h
e
ab
o
v
e
d
ata
s
et,
th
e
s
em
an
tic
s
eg
m
e
n
tatio
n
p
r
o
ce
ed
s
an
d
th
e
n
et
w
o
r
k
m
o
d
el
is
ev
al
u
ated
to
g
et
h
er
.
Fi
g
u
r
e
6
s
h
o
w
s
t
h
e
v
is
u
alize
d
i
m
ag
e
o
f
t
h
e
s
e
m
a
n
tic
s
eg
m
e
n
tatio
n
b
ei
n
g
co
m
p
leted
.
Gr
ee
n
is
a
n
u
n
i
n
f
ec
ted
n
o
r
m
al
tr
ee
,
an
d
r
ed
is
an
i
n
f
ec
ted
p
in
e
tr
ee
.
Yello
w
is
th
e
s
u
s
p
icio
u
s
tr
ee
an
d
b
lack
is
t
h
e
u
n
s
p
ec
if
ied
ar
ea
.
Fi
g
u
r
e
7
s
h
o
w
s
th
e
r
esu
lt
s
cr
ee
n
s
h
o
w
in
g
t
h
e
s
h
ar
e
o
f
ea
ch
clas
s
i
f
icatio
n
m
o
d
el
i
n
t
h
e
to
tal
ar
ea
.
A
cc
o
r
d
in
g
to
t
h
e
ar
ea
clas
s
if
icatio
n
,
2
7
.
3
8
%
o
f
i
n
f
ec
ted
tr
ee
s
,
3
5
.
5
2
%
o
f
s
u
s
p
icio
u
s
tr
ee
s
,
2
5
.
1
4
%
o
f
n
o
r
m
al
tr
ee
s
,
3
.
3
2
%
o
f
v
o
id
s
,
a
n
d
8
.
6
4
%
o
f
la
n
d
.
P
ix
el
clas
s
if
icatio
n
ac
cu
r
ac
y
is
ab
o
u
t 8
7
%.
Fig
u
r
e
5
.
T
est d
ata
s
et
Fig
u
r
e
6
.
Se
m
a
n
tic
s
e
g
m
en
ta
ti
o
n
r
esu
lt
s
Fig
u
r
e
7
.
Se
m
a
n
tic
s
e
g
m
en
ta
ti
o
n
ar
ea
r
esu
lts
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8708
P
in
e
w
ilt d
is
ea
s
e
s
p
r
ea
d
in
g
p
r
ev
en
tio
n
s
ystem
u
s
in
g
s
ema
n
tic
s
eg
men
ta
tio
n
(
C
h
u
lh
yu
n
Hw
a
n
g
)
2671
4
.
2
.
Dis
cu
s
s
io
n
Mo
s
t
s
u
s
p
ec
ted
p
in
e
r
e
-
n
e
m
ato
d
e
s
u
s
p
icio
n
d
etec
tio
n
s
y
s
te
m
s
u
s
ed
s
u
p
er
v
i
s
ed
class
i
f
icatio
n
tech
n
iq
u
es
to
d
ir
ec
tl
y
a
s
s
i
g
n
class
i
f
icatio
n
tar
g
ets
o
r
to
e
x
t
r
ac
t
an
d
d
etec
t
R
GB
v
al
u
es
o
f
s
u
s
p
icio
u
s
tr
ee
s
.
Ho
w
e
v
er
,
th
e
ab
o
v
e
m
et
h
o
d
is
an
o
b
j
ec
t
r
ec
o
g
n
itio
n
m
o
d
el
th
at
lear
n
s
a
s
p
ec
if
ic
p
ar
t
o
f
t
h
e
i
m
ag
e.
Ho
w
e
v
er
,
th
e
ac
c
u
r
ac
y
o
f
t
h
e
o
b
j
ec
t
r
ec
o
g
n
itio
n
m
o
d
el
i
s
o
b
s
er
v
ed
,
b
u
t
t
h
e
p
o
s
itio
n
an
d
ar
ea
o
f
th
e
o
b
j
ec
t
ca
n
n
o
t
b
e
r
etu
r
n
ed
d
u
e
to
t
h
e
c
h
ar
ac
ter
i
s
tics
o
f
t
h
e
o
b
j
ec
t
r
e
co
g
n
itio
n
m
o
d
el.
T
h
er
ef
o
r
e,
th
is
s
t
u
d
y
p
r
o
p
o
s
es
a
s
y
s
te
m
th
at
i
m
p
r
o
v
es
ac
c
u
r
ac
y
an
d
r
etu
r
n
s
th
e
p
o
s
itio
n
an
d
ar
ea
o
f
an
o
b
j
ec
t
b
y
u
s
in
g
s
e
m
a
n
t
ic
s
eg
m
e
n
tatio
n
,
a
n
o
b
j
ec
t
d
etec
tio
n
m
o
d
el,
f
o
r
s
u
s
p
icio
u
s
d
etec
tio
n
.
T
h
is
s
e
ctio
n
v
er
i
f
ies
t
h
e
ef
f
icie
n
c
y
b
y
co
m
p
ar
i
n
g
t
h
e
class
i
f
icatio
n
ac
cu
r
ac
y
o
f
th
e
ex
is
ti
n
g
s
y
s
te
m
w
it
h
th
at
o
f
th
e
p
r
o
p
o
s
ed
s
y
s
te
m
.
A
l
s
o
,
w
e
an
al
y
ze
t
h
e
b
est
m
o
d
el
b
y
co
m
p
ar
i
n
g
th
e
ac
c
u
r
ac
y
o
f
s
e
m
a
n
tic
s
e
g
m
en
tatio
n
m
o
d
el
s
ap
p
lied
to
th
e
p
r
o
p
o
s
ed
s
y
s
te
m
.
T
h
e
ex
p
er
im
e
n
tal
m
et
h
o
d
tr
ain
s
ea
c
h
m
o
d
el
u
s
i
n
g
a
tr
ai
n
in
g
s
et,
ex
tr
ac
t
s
tr
ain
in
g
w
ei
g
h
t
s
,
an
d
m
ea
s
u
r
es
p
ix
e
l
clas
s
if
icatio
n
ac
cu
r
ac
y
a
n
d
ar
ea
ac
cu
r
ac
y
u
s
in
g
a
te
s
t
s
et.
T
ab
le
2
is
a
lear
n
in
g
s
e
t
co
n
tai
n
i
n
g
th
e
en
v
ir
o
n
m
en
t
in
w
h
ich
t
h
e
ex
p
er
i
m
en
t
w
i
ll
b
e
co
n
d
u
cted
.
T
h
e
ex
p
er
im
e
n
tal
e
n
v
ir
o
n
m
en
t
u
s
ed
f
o
r
th
e
ex
p
er
i
m
e
n
t
is
s
h
o
w
n
i
n
T
ab
le
2
.
Usi
n
g
t
h
e
m
o
d
els
o
f
FC
N,
Dee
p
L
ab
V3
,
SetNet,
an
d
U
-
N
et,
B
atch
Size
is
s
e
t
to
5
0
an
d
E
p
o
ch
is
5
0
.
T
h
e
n
u
m
b
er
o
f
tr
ain
i
n
g
s
ets
to
b
e
u
s
ed
is
th
r
ee
.
T
h
e
s
e
m
a
n
tic
s
eg
m
e
n
tatio
n
i
s
p
er
f
o
r
m
ed
b
y
ap
p
l
y
i
n
g
th
e
tr
ai
n
ed
m
o
d
el
w
e
ig
h
t
s
u
s
in
g
t
h
e
ab
o
v
e
test
s
et.
Fig
u
r
e
8
i
s
a
g
r
ap
h
ica
l
r
ep
r
es
en
tatio
n
o
f
t
h
e
tr
ai
n
in
g
d
ata
f
o
r
ea
ch
m
o
d
el.
As
a
r
es
u
lt,
F
C
N,
U
-
Net,
Dee
p
L
ab
V3
an
d
Seg
Net
h
a
v
e
8
7
%,
8
8
%,
9
1
%
an
d
8
5
%
class
i
f
icatio
n
ac
c
u
r
ac
y
,
r
esp
ec
tiv
el
y
.
W
h
en
t
h
e
ex
p
er
i
m
e
n
t
w
a
s
co
n
d
u
cted
w
i
th
t
h
r
ee
test
s
ets,
t
h
e
ac
c
u
r
ac
y
is
s
h
o
w
n
i
n
T
ab
le
3
.
T
est
s
et
1
an
d
test
s
et
2
,
w
h
ic
h
ar
e
r
elativ
e
l
y
ea
s
y
to
c
lass
i
f
y
,
h
a
v
e
h
ig
h
ac
c
u
r
ac
y
with
a
v
er
ag
e
ac
c
u
r
ac
y
o
f
8
7
.
7
5
%
an
d
9
1
.
2
5
%,
b
u
t
8
1
.
7
5
%
f
o
r
test
s
et
th
at
is
d
if
f
icu
lt
to
id
en
ti
f
y
,
s
u
c
h
as
test
s
et
3
.
I
n
ad
d
itio
n
,
Dee
p
L
ab
V3
w
a
s
s
elec
ted
as
th
e
m
o
d
el
w
i
th
t
h
e
h
ig
h
es
t a
cc
u
r
a
c
y
w
it
h
an
a
v
er
ag
e
ac
c
u
r
ac
y
o
f
9
0
%.
T
ab
le
2
.
L
ea
r
n
in
g
en
v
ir
o
n
m
e
n
t
M
o
d
e
l
En
v
i
r
o
n
me
n
t
F
C
N
,
U
-
N
e
t
,
D
e
e
p
L
a
b
V
3
,
S
e
g
N
e
t
B
a
t
c
h
S
i
z
e
=
5
0
,
Ep
o
c
h
=
50
D
a
t
a
S
e
t
C
o
u
n
t
=
3
,
T
e
st
S
e
t
C
o
u
n
t
=
3
Fig
u
r
e
8
.
Mo
d
el
-
s
p
ec
if
ic
ep
o
ch
an
d
clas
s
i
f
icatio
n
ac
c
u
r
ac
y
T
ab
le
3
.
P
ix
el
class
i
f
icatio
n
ac
cu
r
ac
y
b
y
m
o
d
el
M
o
d
e
l
T
e
st
S
e
t
1
A
c
c
u
r
a
c
y
T
e
st
S
e
t
1
A
c
c
u
r
a
c
y
T
e
st
S
e
t
1
A
c
c
u
r
a
c
y
F
C
N
0
.
8
7
0
.
9
0
0
.
8
1
D
e
e
p
L
a
b
V
3
0
.
9
1
0
.
9
3
0
.
8
5
S
e
g
N
e
t
0
.
8
5
0
.
8
9
0
.
7
7
U
-
N
e
t
0
.
8
8
0
.
9
3
0
.
8
4
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
n
t J
E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
11
,
No
.
3
,
J
u
n
e
2021
:
2
6
6
6
-
2673
2672
5.
CO
NCLU
SI
O
N
P
in
e
n
e
m
ato
d
e
d
is
ea
s
e
i
s
a
f
o
r
est
p
est
ep
id
e
m
ic
w
it
h
a
1
0
0
%
m
o
r
talit
y
r
ate.
T
h
er
ef
o
r
e,
ea
r
ly
d
etec
tio
n
is
v
er
y
i
m
p
o
r
tan
t,
a
n
d
it
is
i
m
p
o
r
ta
n
t
to
h
ar
v
est
t
h
e
i
n
f
ec
ted
tr
ee
a
s
q
u
ick
l
y
as
p
o
s
s
ib
le.
T
h
e
Fo
r
est
Ser
v
ice
o
p
er
ates
a
m
o
n
ito
r
in
g
ce
n
ter
to
ta
k
e
co
n
tr
o
l
o
f
th
e
co
m
p
lain
ts
.
Ho
w
e
v
er
,
d
u
e
t
o
th
e
n
at
u
r
e
o
f
th
e
f
o
r
est
ar
ea
,
it
i
s
v
er
y
d
i
f
f
ic
u
l
t
f
o
r
h
u
m
an
s
to
s
u
r
v
e
y
al
l
ar
ea
s
.
R
ec
e
n
tl
y
,
a
s
y
s
te
m
t
h
at
co
m
b
i
n
es
m
ac
h
i
n
e
lear
n
in
g
an
d
d
r
o
n
es
to
ef
f
ec
ti
v
el
y
d
etec
t
a
w
id
e
r
an
g
e
h
a
s
b
ee
n
s
tu
d
ied
.
Du
e
to
th
e
n
at
u
r
e
o
f
p
in
e
n
e
m
ato
d
e
d
is
ea
s
e,
t
h
e
lo
ca
tio
n
a
n
d
ar
ea
d
etec
tio
n
o
f
s
u
s
p
icio
n
is
i
m
p
o
r
tan
t.
Ho
w
e
v
er
,
t
h
e
o
b
j
ec
t
r
ec
o
g
n
itio
n
m
o
d
el,
w
h
ic
h
i
s
o
n
e
o
f
th
e
w
ea
k
p
o
i
n
ts
o
f
th
e
ex
is
ti
n
g
s
y
s
te
m
,
is
ef
f
ec
tiv
e
i
n
r
ec
o
g
n
izi
n
g
o
b
j
ec
ts
b
u
t
ca
n
n
o
t
r
et
u
r
n
th
e
p
o
s
itio
n
a
n
d
ar
ea
o
f
th
e
o
b
j
ec
t.
T
o
s
o
lv
e
th
is
p
r
o
b
le
m
,
t
h
is
p
ap
er
p
r
o
p
o
s
es
a
p
r
ev
en
tio
n
s
y
s
te
m
to
p
r
ev
en
t
th
e
s
p
r
ea
d
o
f
p
in
e
n
e
m
a
to
d
e
d
is
ea
s
e
u
s
in
g
ae
r
ia
l
p
h
o
to
g
r
ap
h
s
a
n
d
s
atel
lite
i
m
ag
e
s
.
I
n
t
h
is
p
ap
er
,
s
e
m
a
n
t
ic
s
e
g
m
e
n
tat
io
n
,
a
n
o
b
j
ec
t
d
etec
ti
o
n
m
o
d
el,
is
u
s
e
d
to
class
i
f
y
i
n
f
ec
ted
tr
ee
s
,
s
u
s
p
icio
u
s
tr
ee
s
,
an
d
n
o
r
m
al
tr
e
es,
v
i
s
u
al
ize
th
e
m
,
an
d
r
etu
r
n
th
e
ar
ea
an
d
ac
cu
r
ac
y
o
f
th
e
s
ep
ar
ated
ar
ea
s
to
th
e
u
s
er
.
I
n
ad
d
itio
n
,
th
e
av
er
ag
e
class
i
f
icatio
n
ac
cu
r
ac
y
o
f
ea
c
h
m
o
d
el
w
a
s
a
b
o
u
t 8
6
.
9
%,
an
d
th
e
m
ax
i
m
u
m
clas
s
i
f
icatio
n
ac
c
u
r
ac
y
w
a
s
ab
o
u
t 9
0
%.
T
h
r
o
u
g
h
th
is
,
w
e
co
u
ld
v
er
i
f
y
t
h
at
t
h
e
o
b
j
ec
t
d
etec
tio
n
m
o
d
el
h
as
s
i
m
i
lar
class
i
f
icat
io
n
ac
c
u
r
ac
y
w
it
h
o
u
t
f
alli
n
g
class
i
f
icatio
n
ac
c
u
r
ac
y
co
m
p
a
r
ed
w
it
h
th
e
ex
is
ti
n
g
o
b
j
ec
t r
ec
o
g
n
itio
n
m
o
d
el.
T
h
er
ef
o
r
e,
it
w
as c
o
n
cl
u
d
ed
th
at
th
e
o
b
j
ec
t
r
ec
o
g
n
itio
n
m
o
d
el,
s
u
ch
a
s
th
e
r
elate
d
r
esear
ch
in
Sectio
n
2
.
3
,
is
f
ast
an
d
e
f
f
ec
tiv
e
f
o
r
s
i
m
p
le
o
b
j
ec
t
r
ec
o
g
n
itio
n
,
b
u
t
th
a
t
t
h
e
o
b
j
ec
t
d
etec
tio
n
m
o
d
el
is
s
u
itab
le
f
o
r
ae
r
o
n
au
tic
s
a
n
d
s
ate
llit
e
i
m
ag
e
s
,
w
h
ich
h
av
e
a
lo
t o
f
i
n
f
o
r
m
at
io
n
.
ACK
NO
WL
E
D
G
E
M
E
NT
S
T
h
is
w
o
r
k
w
a
s
s
u
p
p
o
r
ted
b
y
th
e
r
esear
ch
g
r
an
t o
f
P
aiC
h
ar
Un
i
v
er
s
it
y
in
2
0
2
1
.
RE
F
E
R
E
NC
E
S
[1
]
L
i
m
,
“
F
o
re
st i
n
Crisis
b
y
P
in
e
W
i
lt
Ne
m
a
to
d
e
,
”
Fo
re
st a
n
d
Cu
lt
u
re
,
v
o
l.
2
6
,
n
o
.
7
,
p
p
.
4
-
5
,
2
0
1
7
.
[2
]
S
.
Kh
a
n
a
n
d
K.
V
.
Kir
u
b
a
n
a
n
d
,
“
Co
m
p
a
rin
g
m
a
c
h
in
e
lea
rn
in
g
a
n
d
e
n
se
m
b
le
lea
rn
in
g
in
th
e
f
ield
o
f
f
o
o
tb
a
ll
,
”
In
ter
n
a
t
io
n
a
l
J
o
u
rn
a
l
o
f
E
lec
trica
l
a
n
d
C
o
mp
u
ter
En
g
in
e
e
rin
g
(
IJ
ECE
)
,
v
o
l.
9
,
n
o
.
5
,
p
p
.
4
3
2
1
-
4
3
2
5
,
2
0
1
9
.
[3
]
“
Ko
re
a
F
o
re
st
S
e
rv
ic
e
,
”
K
o
re
a
Fo
re
st
S
e
rv
ice
Ho
me
Pa
g
e
,
2
0
1
9
.
[
On
li
n
e
].
Av
a
il
a
b
le:
h
tt
p
:
//
ww
w
.
f
o
re
st.g
o
.
k
r/n
e
wk
f
s
we
b
/h
tm
l/
Htm
lP
a
g
e
.
d
o
?
p
g
=
/co
n
se
r/co
n
se
r_
0
2
0
1
0
3
.
h
tm
l&
m
n
=
KFS
_
0
2
_
0
2
_
0
1
_
0
3
.
[4
]
J.
K.
Rh
e
e
,
"
P
re
d
ictio
n
f
o
r
P
e
ri
o
d
o
n
tal
Dise
a
se
u
sin
g
G
e
n
e
Ex
p
re
ss
io
n
P
r
o
f
il
e
Da
ta
b
a
se
d
o
n
M
a
c
h
in
e
L
e
a
rn
in
g
,
"
J
o
u
rn
a
l
o
f
th
e
K
o
re
a
I
n
stit
u
te o
f
I
n
fo
rm
a
t
io
n
a
n
d
Co
mm
u
n
ica
ti
o
n
En
g
i
n
e
e
rin
g
,
v
o
l.
2
3
,
n
o
.
8
,
p
p
.
9
0
3
-
9
0
8
,
2
0
1
9
.
[5
]
“
M
a
th
W
o
rk
s,”
M
a
th
W
o
rk
s
Ho
me
p
a
g
e
,
2
0
1
9
.
[
On
l
in
e
]
.
A
v
a
il
a
b
le:
h
tt
p
s:
//
k
r.
m
a
th
w
o
rk
s.c
o
m
/s
o
lu
ti
o
n
s/
im
a
g
e
-
v
id
e
o
-
p
ro
c
e
ss
in
g
/o
b
j
ect
-
re
c
o
g
n
it
i
o
n
.
h
tm
l.
[6
]
I.
J.
Ch
o
i
,
J.
G
.
S
e
o
,
a
n
d
H.
Y.
P
a
rk
,
“
Ob
jec
t
Re
c
o
g
n
it
io
n
in
L
o
w
Re
so
lu
ti
o
n
Im
a
g
e
s
u
sin
g
a
Co
n
v
o
lu
ti
o
n
a
l
Ne
u
ra
l
Ne
tw
o
rk
a
n
d
a
n
Im
a
g
e
En
h
a
n
c
e
m
e
n
t
Ne
tw
o
rk
,
”
T
h
e
J
o
u
rn
a
l
o
f
Ko
re
a
n
I
n
stit
u
te
o
f
In
f
o
rm
a
ti
o
n
S
c
ien
ti
sts
a
n
d
En
g
i
n
e
e
r
in
g
,
v
o
l.
4
5
,
n
o
.
8
,
p
p
.
8
3
1
-
8
3
7
,
2
0
1
8
.
[7
]
G
.
S
.
M
o
o
n
a
n
d
Y.
Kim
,
“
A
Co
m
b
in
e
d
M
o
d
e
l
o
f
Ou
tl
i
n
e
F
e
a
tu
r
e
M
a
p
a
n
d
CNN
f
o
r
De
tec
ti
o
n
o
f
P
e
o
p
le
a
t
th
e
Be
a
c
h
,
”
T
h
e
J
o
u
r
n
a
l
o
f
K
o
re
a
n
I
n
stit
u
te
o
f
I
n
fo
rm
a
ti
o
n
S
c
ien
ti
sts
a
n
d
En
g
i
n
e
e
rs
,
v
o
l.
4
6
,
n
o
.
1
,
p
p
.
3
1
-
3
8
,
2
0
1
9
.
[8
]
S
.
K.
L
e
e
,
S
.
J.
P
a
rk
,
G
.
M
.
Ba
e
k
,
H.
B
.
Kim
,
a
n
d
G
.
W
.
L
e
e
,
“
D
e
tec
ti
o
n
o
f
Da
m
a
g
e
d
P
i
n
e
T
re
e
b
y
th
e
P
in
e
W
il
t
Dise
a
se
Us
in
g
U
A
V
I
m
a
g
e
,
”
T
h
e
J
o
u
rn
a
l
o
f
Ko
re
a
n
S
o
c
iety
o
f
Rem
o
te S
e
n
si
n
g
,
v
o
l.
3
5
,
n
o
.
3
,
p
p
.
3
5
9
-
3
7
3
,
2
0
1
9
.
[9
]
P
.
H.
Hu
y
n
h
,
V
.
H.
Ng
u
y
e
n
,
T
.
N.
Do
,
“
En
h
a
n
c
in
g
G
e
n
e
Ex
p
re
ss
io
n
C
las
sif
ica
t
io
n
o
f
S
u
p
p
o
rt
Ve
c
to
r
M
a
c
h
in
e
s
w
it
h
G
e
n
e
r
a
ti
v
e
A
d
v
e
rsa
rial
Ne
t
w
o
rk
s,
"
Jo
u
r
n
a
l
o
f
In
f
o
rm
a
ti
o
n
a
n
d
C
o
mm
u
n
ic
a
ti
o
n
Co
n
v
e
rg
e
n
c
e
En
g
in
e
e
rin
g
,
v
o
l.
1
7
,
n
o
.
1
,
p
p
.
14
-
2
0
,
2
0
1
9
.
[1
0
]
I.
S
.
Ch
o
,
S
.
J.
Kw
o
n
,
a
n
d
J.
Y.
Yo
o
n
,
“
Oc
c
u
rre
n
c
e
o
f
P
e
a
c
h
L
a
t
e
n
t
M
o
sa
ic
V
iro
i
d
In
d
u
c
i
n
g
P
e
a
c
h
Ca
li
c
o
Dise
a
se
o
f
P
e
a
c
h
T
re
e
ss
,
”
Ho
rticu
lt
u
re
a
b
stra
c
ts
,
p
p
.
1
7
7
-
1
7
8
,
2
0
1
7
.
[1
1
]
G
.
Y.
He
o
,
"
ROI
Ex
trac
ti
o
n
f
o
r
A
u
to
m
a
ti
c
P
lac
a
rd
Re
c
o
g
n
it
i
o
n
,
"
J
o
u
rn
a
l
o
f
t
h
e
Ko
re
a
I
n
stit
u
te
o
f
In
fo
rm
a
t
io
n
a
n
d
Co
mm
u
n
ica
ti
o
n
E
n
g
i
n
e
e
rin
g
,
v
o
l.
2
3
,
n
o
.
4
,
p
p
.
3
7
4
-
3
8
0
,
2
0
1
9
.
[1
2
]
H.
D.
L
e
e
,
“
Id
e
n
ti
f
ica
ti
o
n
o
f
Da
m
a
g
e
d
T
re
e
s
b
y
P
in
e
W
il
t
Dise
a
se
s
Us
in
g
Dro
n
e
Im
a
g
e
s
a
n
d
G
I
S
,
”
Diss
e
rtatio
n
,
K
y
u
n
g
il
Un
iv
e
rsity
,
2017.
[1
3
]
L
a
ff
e
rt
y
,
J,
M
c
Ca
ll
u
m
,
A
.
a
n
d
P
e
re
ira,
F
.
C
.
,
“
Co
n
d
it
i
o
n
a
l
ra
n
d
o
m
f
ield
s:
P
ro
b
a
b
il
ist
ic
m
o
d
e
ls
f
o
r
s
e
g
m
e
n
ti
n
g
a
n
d
la
b
e
li
n
g
se
q
u
e
n
c
e
d
a
ta,”
ICM
L
'0
1
:
Pro
c
e
e
d
in
g
s
o
f
t
h
e
Ei
g
h
tee
n
t
h
In
ter
n
a
t
io
n
a
l
C
o
n
fer
e
n
c
e
o
n
M
a
c
h
in
e
L
e
a
r
n
in
g
,
2
0
0
1
,
p
p
.
2
8
2
-
2
8
9
.
[1
4
]
H.
J.
S
h
in
a
n
d
C.
H.
Oh
,
"
M
a
c
h
in
e
L
e
a
rn
in
g
b
a
se
d
o
n
A
p
p
r
o
a
c
h
f
o
r
Clas
sif
ica
ti
o
n
o
f
A
b
n
o
rm
a
l
Da
ta
in
S
h
o
p
-
f
lo
o
r,
"
J
o
u
rn
a
l
o
f
th
e
Ko
re
a
In
stit
u
te
o
f
In
fo
rm
a
t
io
n
a
n
d
C
o
mm
u
n
ic
a
ti
o
n
En
g
in
e
e
rin
g
,
v
o
l.
2
1
,
n
o
.
1
1
,
p
p
.
2
0
3
7
-
2
0
4
2
,
2
0
1
7
.
[1
5
]
S
u
y
k
e
n
s,
J.
A
,
a
n
d
V
a
n
d
e
w
a
ll
e
,
J
.
,
“
L
e
a
st
sq
u
a
re
s
su
p
p
o
rt
v
e
c
to
r
m
a
c
h
in
e
c
las
sif
iers
,
”
Ne
u
ra
l
p
ro
c
e
ss
in
g
letter
s
,
v
o
l.
9
,
n
o
.
3
,
p
p
.
2
9
3
-
3
0
0
,
1
9
9
9
.
[1
6
]
X
.
T
.
Y
a
n
g
,
J.
S
.
L
e
e
,
a
n
d
H.
K.
Ju
n
g
,
“
F
a
u
lt
Dia
g
n
o
sis
M
a
n
a
g
e
m
e
n
t
M
o
d
e
l
u
sin
g
M
a
c
h
i
n
e
L
e
a
rn
in
g
,
”
J
o
u
r
n
a
l
o
f
In
fo
rm
a
t
io
n
a
n
d
Co
mm
u
n
ica
ti
o
n
Co
n
v
e
rg
e
n
c
e
En
g
in
e
e
rin
g
,
v
o
l
.
1
7
,
n
o
.
2
,
p
p
.
1
2
8
-
1
3
4
,
2
0
1
9
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8708
P
in
e
w
ilt d
is
ea
s
e
s
p
r
ea
d
in
g
p
r
ev
en
tio
n
s
ystem
u
s
in
g
s
ema
n
tic
s
eg
men
ta
tio
n
(
C
h
u
lh
yu
n
Hw
a
n
g
)
2673
[1
7
]
S
.
R.
S
a
lu
ti
,
“
A
su
rv
e
y
o
f
b
ig
d
a
ta
a
n
d
m
a
c
h
in
e
lea
rn
in
g
,
”
In
ter
n
a
ti
o
n
a
l
J
o
u
rn
a
l
o
f
El
e
c
trica
l
a
n
d
Co
mp
u
ter
En
g
i
n
e
e
rin
g
(
IJ
ECE
)
,
v
o
l.
1
0
,
n
o
.
1
,
p
p
.
5
7
5
-
5
8
0
,
2
0
2
0
.
[1
8
]
L
.
C.
Ch
e
n
,
G
.
P
a
p
a
n
d
re
o
u
.
G
,
I.
Ko
k
k
in
o
s,
K.
M
u
r
p
h
y
,
a
n
d
A
.
L
.
Yu
il
le,
“
De
e
p
lab
:
S
e
m
a
n
ti
c
im
a
g
e
se
g
m
e
n
tatio
n
w
it
h
d
e
e
p
c
o
n
v
o
lu
ti
o
n
a
l
n
e
ts,
a
tro
u
s
c
o
n
v
o
lu
t
io
n
,
a
n
d
f
u
ll
y
c
o
n
n
e
c
ted
CRF
S
,
”
IEE
E
tra
n
s
a
c
ti
o
n
s
o
n
p
a
tt
e
rn
a
n
a
lys
is a
n
d
ma
c
h
in
e
i
n
telli
g
e
n
c
e
,
v
o
l.
4
0
,
n
o
.
4
,
p
p
.
8
3
4
-
8
4
8
,
2
0
1
7
.
[1
9
]
J.
H.
Yi
m
a
n
d
K.
A
.
S
o
h
n
,
“
In
v
e
stig
a
ti
n
g
th
e
F
e
a
tu
re
Co
ll
e
c
ti
o
n
f
o
r
S
e
m
a
n
ti
c
S
e
g
m
e
n
tatio
n
v
ia
S
in
g
le
S
k
ip
c
o
n
n
e
c
ti
o
n
,
”
Th
e
Ko
re
a
n
I
n
stit
u
te o
f
I
n
fo
rm
a
ti
o
n
S
c
ien
ti
sts a
n
d
En
g
in
e
e
rs
,
v
o
l.
4
4
,
n
o
.
1
2
,
p
p
.
1
2
8
2
-
1
2
8
9
,
2
0
1
7
.
[2
0
]
J.
H.
Yi
m
a
n
d
K.
A
.
S
o
h
n
,
“
E
ff
icie
n
t
In
fe
re
n
c
e
o
f
I
m
a
g
e
Ob
jec
ts
u
sin
g
S
e
m
a
n
ti
c
S
e
g
m
e
n
tatio
n
,
”
J
o
u
r
n
a
l
o
f
Bro
a
d
c
a
st E
n
g
i
n
e
e
rin
g
,
v
o
l.
2
4
,
n
o
.
1
,
p
p
.
6
7
-
7
6
,
2
0
1
9.
[2
1
]
G
.
W
a
n
g
a
n
d
S
.
Y.
S
h
in
,
“
A
n
Im
p
ro
v
e
d
T
e
x
t
Clas
si
f
ic
a
ti
o
n
M
e
th
o
d
f
o
r
S
e
n
ti
m
e
n
t
Clas
si
f
ica
ti
o
n
,
”
J
o
u
r
n
a
l
o
f
In
fo
rm
a
t
io
n
a
n
d
Co
mm
u
n
ica
ti
o
n
Co
n
v
e
rg
e
n
c
e
En
g
in
e
e
rin
g
,
v
o
l
.
1
7
,
n
o
.
1
,
p
p
.
41
-
4
8
,
2
0
1
9
.
[2
2
]
B.
Ha
o
a
n
d
D.
S
.
Ka
n
g
,
“
Re
se
a
r
c
h
o
n
Im
a
g
e
S
e
m
a
n
ti
c
S
e
g
m
e
n
ta
ti
o
n
Ba
se
d
o
n
F
CN
-
V
GG
a
n
d
P
y
r
a
m
id
P
o
o
li
n
g
M
o
d
u
le,”
T
h
e
J
o
u
r
n
a
l
o
f
K
o
re
a
n
In
stit
u
te
o
f
I
n
f
o
rm
a
ti
o
n
T
e
c
h
n
o
lo
g
y
,
v
o
l.
1
6
,
n
o
.
7
,
p
p
.
1
-
8
,
2
0
1
8
.
[2
3
]
H.
K.
Ki
m
,
S
.
O.
Lee
,
a
n
d
H.
K.
Ju
n
g
,
“
Hu
m
a
n
a
c
ti
v
it
y
re
c
o
g
n
it
io
n
b
y
u
sin
g
c
o
n
v
o
lu
ti
o
n
a
l
n
e
u
ra
l
n
e
tw
o
rk
,
”
In
ter
n
a
t
io
n
a
l
J
o
u
rn
a
l
o
f
E
lec
trica
l
a
n
d
C
o
mp
u
ter
En
g
in
e
e
rin
g
(
IJ
ECE
)
,
v
o
l
.
9
,
n
o
.
6
,
p
p
.
5
2
7
0
-
5
2
7
6
,
2
0
1
9
.
[2
4
]
S
.
A
ich
,
S
.
Ch
a
k
ra
b
o
rty
,
a
n
d
H.
C.
Kim
,
“
Co
n
v
o
lu
ti
o
n
a
l
n
e
u
ra
l
n
e
tw
o
rk
-
b
a
se
d
m
o
d
e
l
f
o
r
w
e
b
-
b
a
se
d
tex
t
c
las
si
f
ica
ti
o
n
,
”
In
ter
n
a
ti
o
n
a
l
J
o
u
rn
a
l
o
f
E
lec
trica
l
a
n
d
C
o
mp
u
ter
En
g
i
n
e
e
rin
g
(
IJ
ECE
)
,
v
o
l.
9
,
n
o
.
6
,
p
p
.
5
1
8
5
-
5
1
9
1
,
2
0
1
9
.
[2
5
]
O.
Ro
n
n
e
b
e
rg
e
r,
P
.
F
isc
h
e
r,
a
n
d
T
.
Bro
x
,
“
U
-
n
e
t:
Co
n
v
o
lu
ti
o
n
a
l
n
e
tw
o
rk
s
f
o
r
b
io
m
e
d
ica
l
i
m
a
g
e
se
g
m
e
n
tatio
n
,
”
M
ICCAI
2
0
1
5
:
M
e
d
ica
l
Im
a
g
e
Co
mp
u
ti
n
g
a
n
d
C
o
mp
u
ter
-
Assiste
d
In
ter
v
e
n
ti
o
n
-
M
ICCAI
2
0
1
5
,
v
o
l
.
9
3
5
1
,
p
p
.
2
3
4
-
2
4
1
,
2
0
1
5
.
B
I
O
G
RAP
H
I
E
S
O
F
AUTH
O
RS
Chu
l
h
y
u
n
H
w
a
n
g
re
c
e
iv
e
d
th
e
M
.
S
.
d
e
g
re
e
in
1
9
9
5
a
n
d
P
h
.
D.
d
e
g
re
e
in
2
0
1
7
f
ro
m
De
p
a
rtme
n
t
o
f
Co
m
p
u
ter
En
g
in
e
e
rin
g
o
f
P
a
i
C
h
a
i
Un
iv
e
rsity
,
Ko
re
a
.
F
ro
m
1
9
9
1
to
2
0
0
0
.
He
w
o
rk
e
d
f
o
r
Ko
re
a
Na
v
y
a
s
a
Co
m
p
u
ter
Off
i
c
e
r.
S
in
c
e
2
0
1
9
,
h
e
h
a
s
w
o
rk
e
d
in
th
e
De
p
a
rtm
e
n
t
o
f
S
m
a
rt
I
T
S
o
f
t
w
a
r
e
a
t
K
y
u
n
g
Bo
k
Un
iv
e
rsit
y
,
w
h
e
re
h
e
w
o
rk
s a
s a
p
ro
f
e
ss
o
r.
His c
u
rre
n
t
re
se
a
rc
h
in
tere
sts in
c
lu
d
e
d
e
e
p
lea
rn
in
g
,
m
a
c
h
in
e
lea
rn
in
g
,
I
o
T
,
b
ig
d
a
ta an
d
a
rti
f
icia
l
in
telli
g
e
n
c
e
.
J
a
e
a
n
J
e
o
n
g
re
c
e
iv
e
d
th
e
B.
S
.
d
e
g
re
e
in
1
9
9
2
f
ro
m
th
e
De
p
a
rt
m
e
n
t
o
f
Co
m
p
u
ter
S
c
ien
c
e
o
f
Je
o
n
b
u
k
Na
ti
o
n
a
l
Un
iv
e
rsit
y
,
a
n
d
re
c
e
iv
e
d
th
e
M
.
S
d
e
g
re
e
in
2
0
2
0
f
ro
m
th
e
D
e
p
a
rt
m
e
n
t
o
f
Co
m
p
u
ter
En
g
in
e
e
rin
g
o
f
P
a
ich
a
i
Un
iv
e
rsit
y
,
Ko
re
a
.
S
in
c
e
1
9
9
3
,
h
e
h
a
s
w
o
rk
in
g
in
th
e
Ko
re
a
w
a
t
e
r
re
so
u
rc
e
s
c
o
o
p
e
ra
ti
o
n
(Kw
a
ter),
w
h
e
re
h
e
n
o
w
w
o
rk
s
a
s
a
g
e
n
e
ra
l
d
irec
to
r
o
f
Hw
a
se
o
n
g
Off
ic
e
.
His
c
u
rre
n
t
re
se
a
r
c
h
in
tere
sts
in
c
lu
d
e
b
ig
d
a
ta
a
n
a
ly
sis
p
lat
f
o
r
m
,
o
p
e
n
d
a
ta
p
latf
o
rm
,
d
a
ta
p
ro
c
e
ss
in
g
a
n
d
m
a
c
h
in
e
lea
rn
in
g
.
H
o
e
k
y
u
n
g
J
u
n
g
re
c
e
iv
e
d
th
e
M
.
S
.
d
e
g
re
e
in
1
9
8
7
a
n
d
P
h
.
D.
d
e
g
re
e
in
1
9
9
3
f
ro
m
th
e
De
p
a
rtme
n
t
o
f
Co
m
p
u
ter
En
g
in
e
e
rin
g
o
f
K
wa
n
g
w
o
o
n
Un
iv
e
rsity
,
Ko
re
a
.
F
ro
m
1
9
9
4
to
1
9
9
5
,
h
e
w
o
rk
e
d
f
o
r
ET
RI
a
s
a
r
e
se
a
r
c
h
e
r.
S
in
c
e
1
9
9
4
,
h
e
h
a
s
w
o
rk
e
d
in
th
e
De
p
a
rtm
e
n
t
o
f
Co
m
p
u
ter
En
g
in
e
e
rin
g
a
t
P
a
ich
a
i
Un
iv
e
rsity
,
w
h
e
r
e
h
e
n
o
w
w
o
rk
s
a
s
a
p
r
o
f
e
ss
o
r.
His
c
u
rre
n
t
re
se
a
r
c
h
in
tere
sts in
c
lu
d
e
m
u
lt
im
e
d
ia d
o
c
u
m
e
n
t
a
rc
h
it
e
c
tu
re
m
o
d
e
li
n
g
,
in
f
o
rm
a
ti
o
n
p
ro
c
e
ss
in
g
,
e
m
b
e
d
d
e
d
s
y
ste
m
,
m
a
c
h
in
e
lea
rn
in
g
,
b
ig
d
a
t
a
,
a
n
d
I
o
T
.
Evaluation Warning : The document was created with Spire.PDF for Python.