I
nte
rna
t
io
na
l J
o
urna
l o
f
I
nfo
rm
a
t
ics a
nd
Co
m
m
un
ica
t
io
n
T
ec
hn
o
lo
g
y
(
I
J
-
I
CT
)
Vo
l.
1
4
,
No
.
2
,
A
u
g
u
s
t
20
2
5
,
p
p
.
6
1
4
~
6
2
4
I
SS
N:
2252
-
8
7
7
6
,
DOI
:
1
0
.
1
1
5
9
1
/iji
ct
.
v
1
4
i
2
.
pp
614
-
6
2
4
614
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
ict.
ia
esco
r
e.
co
m
Techniqu
es o
f
dee
p learning
neu
ra
l
network
-
ba
sed
b
uilding
feature
extractio
n
f
ro
m remo
te
sens
ing
ima
g
es: a sur
v
ey
Sh
riniv
a
s
B
.
K
ha
nd
a
re
,
M
a
no
j
B
.
Cha
nd
a
k
D
e
p
a
r
t
me
n
t
o
f
C
o
mp
u
t
e
r
S
c
i
e
n
c
e
a
n
d
En
g
i
n
e
e
r
i
n
g
,
R
a
s
h
t
r
a
sa
n
t
Tu
k
a
d
o
j
i
M
a
h
a
r
a
j
N
a
g
p
u
r
U
n
i
v
e
r
si
t
y
,
S
h
r
i
R
a
m
d
e
o
b
a
b
a
C
o
l
l
e
g
e
o
f
En
g
i
n
e
e
r
i
n
g
a
n
d
M
a
n
a
g
e
m
e
n
t
,
N
a
g
p
u
r
,
I
n
d
i
a
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Ma
r
1
1
,
2
0
2
4
R
ev
is
ed
Dec
2
4
,
2
0
2
4
Acc
ep
ted
J
an
1
9
,
2
0
2
5
Re
c
e
n
tl
y
,
d
u
e
to
e
a
rth
q
u
a
k
e
d
isa
ste
r,
m
a
n
y
p
e
o
p
le
h
a
v
e
lo
st
t
h
e
ir
li
v
e
s
a
n
d
h
o
m
e
s,
a
n
d
n
o
t
a
b
le
t
o
se
tt
le
to
n
e
w
lo
c
a
ti
o
n
s
imm
e
d
iate
ly
.
T
h
e
re
fo
re
,
a
fra
m
e
wo
rk
o
r
a
p
lan
s
h
o
u
ld
b
e
r
e
a
d
y
t
o
imm
e
d
iate
ly
re
lo
c
a
te
t
h
e
p
e
o
p
le
t
o
d
iffere
n
t
lo
c
a
ti
o
n
s
o
r
d
o
re
se
tt
le
m
e
n
t.
M
u
c
h
re
se
a
rc
h
h
a
s
b
e
e
n
d
o
n
e
in
t
h
is
field
b
u
t
stil
l
th
e
re
a
re
p
ro
b
lem
s
o
f
i
d
e
n
ti
f
y
i
n
g
c
lea
r
b
u
il
d
i
n
g
b
o
u
n
d
a
ries
,
re
c
tan
g
u
lar
h
o
u
se
s,
d
u
e
to
t
h
e
p
r
o
b
lem
o
f
d
iffere
n
t
sh
a
p
e
s
o
f
t
h
e
b
u
il
d
in
g
s.
Th
e
se
tec
h
n
iq
u
e
s
we
re
e
x
p
lo
r
e
d
fo
r
i
d
e
n
ti
fica
ti
o
n
o
f
c
lea
r
b
u
il
d
i
n
g
b
o
u
n
d
a
ries
,
re
c
tan
g
u
lar
h
o
u
se
s,
b
u
il
d
in
g
s
w
h
ich
a
re
m
o
re
h
i
g
h
li
g
h
ted
a
n
d
sm
a
ll
e
r
siz
e
b
u
il
d
i
n
g
s
fo
r
p
re
-
d
i
sa
ste
r
a
n
d
p
o
st
-
d
isa
ste
r
b
u
i
ld
i
n
g
e
x
trac
ti
o
n
sc
e
n
a
rio
s.
In
th
is
s
u
rv
e
y
o
f
b
u
il
d
in
g
e
x
trac
ti
o
n
tec
h
n
iq
u
e
s,
m
o
st
o
f
th
e
a
p
p
ro
a
c
h
is
train
i
n
g
t
h
e
n
e
two
r
k
,
se
c
o
n
d
a
p
p
ro
a
c
h
is
re
fin
i
n
g
t
h
e
train
e
d
o
u
t
p
u
t
fe
a
tu
re
s,
ru
n
n
i
n
g
th
e
trai
n
e
d
sa
m
p
les
o
n
th
e
p
re
d
e
fin
e
d
m
o
d
e
ls
o
f
n
e
u
ra
l
n
e
two
r
k
.
S
e
v
e
ra
l
issu
e
s
a
n
d
th
e
ir
a
ss
e
ss
m
e
n
t
a
re
stu
d
ied
in
t
h
e
se
tec
h
n
iq
u
e
s.
Th
e
se
a
re
b
e
n
e
ficia
l
to
th
e
v
a
rio
u
s
re
se
a
rc
h
e
rs
fo
r
d
iffere
n
t
b
u
il
d
in
g
e
x
trac
ti
o
n
s.
K
ey
w
o
r
d
s
:
Bu
ild
in
g
ex
tr
ac
tio
n
Dee
p
lear
n
in
g
Dis
aster
Featu
r
e
ex
tr
ac
tio
n
Neu
r
al
n
etwo
r
k
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
:
Sh
r
in
iv
as
B.
Kh
an
d
ar
e
Dep
ar
tm
en
t o
f
C
o
m
p
u
ter
Scie
n
ce
an
d
E
n
g
in
ee
r
in
g
,
R
ash
tr
asan
t T
u
k
ad
o
ji M
ah
ar
aj
Nag
p
u
r
Un
iv
er
s
ity
Sh
r
i Ram
d
eo
b
ab
a
C
o
lleg
e
o
f
E
n
g
in
ee
r
in
g
an
d
Ma
n
ag
em
e
n
t
Nag
p
u
r
,
Ma
h
ar
ash
tr
a,
I
n
d
ia
E
m
ail: sr
in
iv
as.k
h
an
d
ar
e@
g
m
ail.
co
m
1.
I
NT
RO
D
UCT
I
O
N
Au
to
m
ated
b
u
ild
i
n
g
f
ea
tu
r
e
e
x
tr
ac
tio
n
f
r
o
m
s
atellite
im
ag
er
y
is
cr
u
cial
f
o
r
u
r
b
an
p
lan
n
i
n
g
[
1
]
an
d
p
o
p
u
latio
n
d
e
n
s
ity
esti
m
atio
n
[
2
]
.
I
t
is
also
cr
u
cial
f
o
r
th
e
cr
ea
tio
n
,
m
ain
ten
an
ce
,
an
d
f
a
b
r
icatio
n
o
f
p
h
y
s
io
g
r
a
p
h
ical
m
a
p
s
.
Au
to
m
atic
d
etec
tio
n
is
th
er
ef
o
r
e
ess
en
tial.
E
v
en
t
h
o
u
g
h
a
u
to
m
atic
f
ea
tu
r
e
ex
tr
ac
tio
n
is
a
d
if
f
icu
lt
u
n
d
er
tak
in
g
,
in
r
e
ce
n
t
y
ea
r
s
it
h
as
b
e
co
m
e
in
c
r
ea
s
in
g
ly
im
p
o
r
tan
t.
T
h
e
d
if
f
icu
lties
ca
m
e
f
r
o
m
s
h
ad
o
ws,
n
o
is
e,
an
d
u
n
ce
r
tain
ty
in
th
e
s
atellite
d
ata
’
s
b
ac
k
g
r
o
u
n
d
p
ictu
r
es
[
3
]
.
Ma
k
i
n
g
ef
f
ec
tiv
e
an
d
ef
f
icien
t
u
s
e
o
f
t
h
e
d
ata
f
r
o
m
r
e
m
o
te
s
en
s
in
g
(
R
S)
h
as
b
ec
o
m
e
a
lab
o
r
io
u
s
task
.
I
t
is
s
im
p
le
to
o
b
tain
th
ese
ae
r
ial
d
ata
f
r
o
m
a
n
u
m
b
e
r
o
f
s
o
u
r
ce
s
[
4
]
.
T
h
is
s
u
r
p
lu
s
o
f
h
ig
h
-
r
eso
lu
t
io
n
s
atellite
d
ata
h
as
b
ec
o
m
e
a
d
ilem
m
a
f
o
r
th
e
s
cien
tific
c
o
m
m
u
n
ity
.
Neu
r
al
n
etwo
r
k
s
with
d
ee
p
lear
n
in
g
(
DL
)
h
a
v
e
g
ain
e
d
p
r
o
m
in
e
n
ce
in
s
atellite
RS
d
u
r
in
g
th
e
last
co
u
p
le
o
f
d
ec
a
d
es [
5
]
.
Natu
r
al
ca
tast
r
o
p
h
es
ca
n
h
a
v
e
an
im
p
ac
t
o
n
b
u
ild
i
n
g
s
tr
u
ctu
r
es.
T
h
is
m
ak
es
a
p
o
s
t
-
d
is
aster
au
to
m
atic
m
ea
s
u
r
em
en
t
b
ased
o
n
s
atell
ite
d
ata
cr
u
cial.
A
v
ast
am
o
u
n
t
o
f
tr
ain
in
g
d
ata
is
n
ee
d
ed
to
tr
ain
th
e
tr
ain
in
g
d
ata
s
et
f
o
r
t
h
e
n
ee
d
o
f
au
to
m
izatio
n
in
o
r
d
er
to
tr
ai
n
a
d
ee
p
n
eu
r
al
n
etwo
r
k
.
I
t
is
q
u
ite
ch
allen
g
in
g
to
g
ath
er
s
u
ch
lar
g
e
am
o
u
n
ts
o
f
tr
ain
in
g
d
ata
in
a
r
ea
l
-
tim
e
co
n
tex
t.
Sa
tellite
p
ictu
r
es
ar
e
a
ty
p
e
o
f
r
em
o
tely
s
en
s
ed
im
ag
e
th
at
m
ay
co
v
e
r
wid
e
ar
ea
s
an
d
h
av
e
h
i
g
h
r
eso
lu
tio
n
,
r
e
v
is
it
tim
e,
an
d
ca
p
a
b
i
lity
.
L
ar
g
e
m
o
u
n
d
s
o
f
d
eb
r
is
wer
e
f
o
u
n
d
b
y
co
n
ce
n
tr
atin
g
o
n
p
ar
tially
an
d
co
m
p
letely
co
llap
s
ed
b
u
ild
in
g
s
tr
u
ctu
r
es,
an
d
th
is
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J I
n
f
&
C
o
m
m
u
n
T
ec
h
n
o
l
I
SS
N:
2252
-
8
7
7
6
Tech
n
iq
u
es o
f d
ee
p
lea
r
n
in
g
n
eu
r
a
l n
etw
o
r
k
b
a
s
ed
b
u
ild
i
n
g
f
ea
tu
r
e
ex
tr
a
ctio
n
fr
o
m
…
(
S
h
r
in
iva
s
K
h
a
n
d
a
r
e
)
615
re
s
u
lted
in
s
ig
n
if
ican
t
co
n
tr
ib
u
tio
n
s
.
Per
h
ap
s
n
o
t
t
h
e
b
est
f
o
r
th
e
cr
ac
k
s
,
s
p
allin
g
,
a
n
d
f
a
ça
d
e.
Ho
wev
er
,
a
f
ew
to
p
r
esear
ch
er
s
co
n
ce
n
tr
ated
o
n
f
aç
ad
e
f
r
o
m
p
h
o
t
o
s
tak
en
af
t
er
d
is
aster
s
[
6
]
.
T
h
e
au
to
m
atic
v
ec
to
r
izatio
n
o
f
th
e
b
u
ild
in
g
’
s
s
h
ap
e
is
u
s
ed
in
u
r
b
a
n
m
an
a
g
em
en
t
an
d
g
e
o
d
atab
ase
u
p
d
atin
g
.
R
ec
en
tly
,
d
ee
p
co
n
v
o
lu
tio
n
al
n
e
u
r
al
n
etwo
r
k
(
D
C
NN
)
h
as
b
ee
n
ap
p
lied
to
ex
tr
ac
t
ar
ch
itectu
r
al
f
ea
tu
r
es;
th
e
o
u
tp
u
t
is
r
aster
o
u
tp
u
t
r
at
h
er
th
a
n
v
ec
to
r
im
a
g
e
s
.
I
t
f
r
e
q
u
en
tly
f
alls
s
h
o
r
t
o
f
t
h
e
n
ee
d
f
o
r
lo
ca
tin
g
th
e
b
u
il
d
in
g
s
.
Ma
n
y
tech
n
iq
u
es
ex
is
t
f
o
r
co
n
v
er
tin
g
r
aster
im
ag
es
to
v
ec
to
r
f
o
r
m
ats;
h
o
wev
er
,
th
e
r
esu
lt
is
o
f
ten
an
ex
ce
s
s
iv
e
n
u
m
b
er
o
f
v
ec
to
r
p
o
in
ts
an
d
asy
m
m
etr
ic
al
s
h
ap
es.
DL
alg
o
r
ith
m
s
ar
e
p
r
im
ar
ily
a
p
p
lied
to
im
ag
e
class
if
icatio
n
th
ese
d
ay
s
.
I
t
h
as
also
b
ee
n
ex
ten
s
i
v
ely
u
tili
ze
d
f
o
r
p
ictu
r
e
r
ec
o
g
n
itio
n
a
n
d
im
ag
e
s
eg
m
en
tatio
n
in
a
v
a
r
iety
o
f
a
p
p
licatio
n
s
.
T
h
ese
alg
o
r
ith
m
s
o
u
tp
er
f
o
r
m
ed
t
y
p
ical
o
r
c
o
n
v
en
tio
n
al
alg
o
r
ith
m
s
th
an
k
s
to
th
e
co
n
te
x
tu
al
f
ea
t
u
r
es
in
th
eir
n
eig
h
b
o
u
r
s
;
y
et,
th
ey
h
av
e
ce
r
tain
lo
ca
l
is
s
u
es
o
f
l
o
ca
l
f
ea
t
u
r
e
o
p
tim
is
atio
n
an
d
g
lo
b
al
f
ea
t
u
r
e
o
p
tim
is
atio
n
f
r
o
m
th
e
n
ei
g
h
b
o
u
r
h
o
o
d
o
b
jects [
7
]
.
T
h
e
task
o
f
b
u
ild
in
g
s
ep
ar
atio
n
f
r
o
m
s
atellite
d
ata
is
d
is
tin
c
t
b
ec
au
s
e
d
iv
er
s
e
b
u
ild
in
g
s
h
a
p
es
ex
is
t.
Fu
r
th
er
m
o
r
e
,
f
in
er
d
etails
o
f
elem
en
ts
lik
e
s
h
ad
o
ws,
s
p
a
ce
s
b
etwe
en
b
u
ild
in
g
s
,
an
d
f
lo
o
r
s
,
ar
e
p
r
o
v
id
e
d
b
y
h
ig
h
r
eso
lu
tio
n
im
a
g
es.
T
h
u
s
,
it
is
s
till
d
if
f
icu
lt
to
ex
tr
ac
t
b
et
ter
ac
cu
r
ac
y
f
r
o
m
h
ig
h
s
p
atial
r
eso
lu
tio
n
im
ag
es.
B
u
ild
in
g
f
o
o
tp
r
in
t
ex
tr
ac
tio
n
with
b
aselin
e
n
etwo
r
k
-
b
ased
s
em
an
tic
s
eg
m
en
tatio
n
u
tili
zin
g
DL
tec
h
n
iq
u
es
is
s
u
cc
ess
f
u
l.
T
h
er
ef
o
r
e
,
th
e
am
o
u
n
t o
f
s
p
ac
e
av
ailab
le
b
etwe
en
b
u
ild
in
g
zo
n
es is
lim
ited
[
8
]
.
I
n
m
ilit
ar
y
an
d
m
ap
im
p
lem
e
n
tatio
n
,
f
ea
t
u
r
e
e
x
tr
ac
tio
n
p
la
y
s
a
v
ital
r
o
le.
Sp
ec
tr
al
s
ig
n
atu
r
e
is
u
s
ed
to
id
en
tify
a
f
ea
t
u
r
e
f
r
o
m
h
i
g
h
r
eso
lu
tio
n
i
m
ag
e
.
T
h
e
r
e
a
r
e
s
ix
teen
m
illi
o
n
co
lo
r
s
a
v
ailab
le
in
an
y
g
i
v
en
im
ag
e.
I
t
is
d
if
f
icu
lt
t
o
id
en
tif
y
an
im
ag
e
b
ased
o
n
its
co
lo
r
s
.
Similar
co
lo
r
s
ar
e
g
r
o
u
p
e
d
to
g
eth
er
f
o
r
b
etter
an
aly
za
tio
n
o
f
th
e
im
ag
e.
Fo
r
th
is
p
u
r
p
o
s
e,
m
ac
h
i
n
e
lear
n
in
g
(
ML
)
,
DL
,
c
lu
s
ter
in
g
ap
p
r
o
a
ch
es
ar
e
b
ein
g
u
s
ed
b
y
m
a
n
y
o
f
t
h
e
r
esear
ch
e
r
s
.
Fr
o
m
th
e
s
eg
m
en
ted
im
ag
e,
s
p
e
ctr
al
in
f
o
r
m
atio
n
o
f
b
u
ild
in
g
s
,
r
o
ad
,
g
r
e
en
s
p
ac
es
f
ea
tu
r
es
ar
e
ex
tr
ac
ted
.
Sp
ec
tr
a
l
in
f
o
r
m
atio
n
al
o
n
e
is
in
s
u
f
f
ic
ien
t
as
s
o
ils
an
d
f
ea
tu
r
es
h
av
e
a
s
im
ilar
s
p
ec
tr
al
s
ig
n
atu
r
e.
R
esear
ch
er
s
h
av
e
d
ev
elo
p
ed
t
h
e
co
n
ce
p
t
o
f
m
ath
e
m
atica
l
m
o
r
p
h
o
lo
g
y
,
wh
ich
is
ap
p
lied
to
s
ep
ar
ate
f
ea
tu
r
es f
r
o
m
n
o
n
-
s
im
ilar
o
b
je
cts with
d
if
f
er
en
t sp
ec
tr
al
s
ig
n
atu
r
e
[
9
]
.
2.
T
E
CH
N
I
Q
UE
S O
F
B
UI
L
D
I
NG
F
E
AT
UR
E
E
XT
R
ACT
I
O
N
2
.
1
.
O
v
er
v
iew
T
h
er
e
ar
e
v
a
r
io
u
s
tech
n
iq
u
es
o
f
b
u
ild
in
g
f
ea
tu
r
e
ex
tr
ac
tio
n
wh
ich
ar
e
ex
p
l
o
r
ed
b
y
th
e
r
es
ea
r
ch
er
s
in
th
e
last
f
ew
y
ea
r
s
.
I
n
th
is
p
a
p
er
,
f
ew
m
ajo
r
tech
n
iq
u
es
h
av
e
b
ee
n
r
ev
iewe
d
.
T
h
e
d
etails
an
d
lim
itatio
n
s
ar
e
id
en
tifie
d
f
r
o
m
th
e
tech
n
iq
u
es.
Var
io
u
s
tech
n
iq
u
es
in
v
o
lv
e
th
e
d
i
f
f
e
r
en
tiatio
n
o
f
b
u
ild
in
g
s
b
y
im
a
g
e
s
eg
m
en
tatio
n
with
Un
et
with
d
if
f
er
en
t
en
co
d
er
s
ar
ch
itectu
r
e,
wh
ich
was
ex
p
e
r
im
en
ted
o
n
h
ig
h
-
r
eso
lu
tio
n
s
atellite
im
ag
es.
I
n
th
ese
tech
n
iq
u
es
th
er
e
ar
e
v
ar
io
u
s
p
r
o
b
lem
s
,
o
f
co
n
cr
ete
b
ase
f
lo
o
r
b
ein
g
u
n
class
if
ied
,
n
eig
h
b
o
u
r
in
g
b
u
ild
i
n
g
s
p
o
o
r
l
y
r
ec
o
g
n
is
ed
,
wr
o
n
g
ly
class
if
icatio
n
o
f
s
h
ad
o
ws
as
b
u
ild
in
g
s
in
s
o
m
e
p
ar
ts
,
b
u
ild
in
g
h
i
d
d
en
b
elo
w
v
eg
eta
tio
n
also
p
o
o
r
ly
class
if
ied
.
I
n
th
e
n
ex
t
s
ec
tio
n
all
th
e
tech
n
i
q
u
es
ar
e
id
en
tifie
d
an
d
ex
p
lain
ed
in
d
etail
f
r
o
m
p
o
in
t n
o
2
to
1
7
.
2
.
2
.
Unet
Un
et
is
th
e
b
ac
k
b
o
n
e
o
f
t
h
e
n
eu
r
al
n
etwo
r
k
.
T
h
e
lo
ca
tio
n
d
ata
o
f
th
e
u
n
d
e
r
ly
in
g
in
f
o
r
m
at
io
n
an
d
th
e
s
em
an
tic
d
ata
o
f
t
h
e
ab
s
tr
ac
t
ch
ar
ac
ter
is
tics
wer
e
o
n
ly
m
i
x
ed
in
t
h
e
o
r
ig
in
al
Un
et.
T
h
i
s
m
eth
o
d
c
o
m
b
in
es
class
f
ea
tu
r
es,
p
ix
el
f
ea
tu
r
es,
a
n
d
ab
s
tr
ac
ted
c
h
ar
ac
ter
is
tics
with
lo
ca
tio
n
in
f
o
r
m
atio
n
to
in
c
r
ea
s
e
th
e
ac
cu
r
ac
y
o
f
th
e
n
etwo
r
k
an
d
its
s
em
an
tic
r
ep
r
esen
tatio
n
[
1
0
]
.
Ov
e
r
lap
p
in
g
tile
s
tr
ateg
y
h
a
d
b
e
en
u
s
ed
to
s
eg
m
en
t
s
ea
m
less
ly
o
f
lar
g
e
ar
b
itra
r
y
b
io
m
ed
ical
im
ag
es.
T
h
e
co
n
tr
ac
tin
g
an
d
ex
p
an
d
in
g
p
ath
ar
e
u
s
ed
f
o
r
ca
p
tu
r
e
co
n
tex
t
an
d
p
r
ec
is
e
lo
ca
lizat
io
n
.
I
t
is
p
o
s
s
ib
le
to
tr
ain
t
h
is
n
etwo
r
k
with
f
ew
p
h
o
to
s
[
1
1
]
.
T
o
ex
tr
ac
t
m
u
ltis
ca
le
p
r
o
p
er
ties
o
f
th
e
tar
g
et,
th
e
m
u
lti
-
U
n
et
ap
p
r
o
ac
h
cr
ea
tes
an
ef
f
ec
tiv
e
n
etwo
r
k
r
esid
u
al
m
o
d
u
le
u
n
d
er
a
m
u
ltis
en
s
o
r
y
f
ield
(
R
MM
F).
Op
tim
izin
g
f
ea
t
u
r
e
in
f
o
r
m
atio
n
in
v
o
l
v
es
th
e
em
p
lo
y
m
en
t
o
f
an
atten
tio
n
m
ec
h
an
is
m
.
T
h
e
r
esid
u
al
m
o
d
u
le
ap
p
r
o
ac
h
lear
n
s
n
etwo
r
k
p
r
o
p
er
ties
at
v
ar
io
u
s
s
ca
les
b
y
em
p
lo
y
in
g
p
ar
allel
co
n
v
o
l
u
tio
n
al
lay
er
s
.
B
u
ild
s
l
ef
to
v
er
b
lo
ck
s
b
y
ad
d
in
g
s
h
o
r
tcu
t
co
n
n
ec
tio
n
s
b
etwe
en
s
ta
ck
ed
lay
er
s
,
f
u
s
in
g
h
ig
h
-
lev
el
s
em
an
tic
in
f
o
r
m
ati
o
n
with
lo
w
-
lev
el
g
r
a
n
u
lar
in
f
o
r
m
atio
n
.
B
y
s
u
b
s
titu
tin
g
th
e
r
esid
u
al
m
o
d
u
le
f
o
r
th
e
co
n
v
o
lu
tio
n
al
lay
er
in
th
e
U
n
et
s
tr
u
ct
u
r
e,
t
h
e
n
etw
o
r
k
s
tr
u
ct
u
r
e
is
s
tr
en
g
th
en
e
d
.
Fo
r
test
in
g
,
th
e
Gao
f
en
-
2
a
n
d
Po
ts
d
am
d
ata
s
ets we
r
e
u
tili
ze
d
[
1
2
]
.
2
.
3
.
Res2
-
Unet
T
h
e
DL
alg
o
r
ith
m
s
s
u
ch
as Un
et,
R
esUn
et,
R
e
s
Net5
0
,
R
es
Un
et
d
ilated
,
u
s
e
en
co
d
er
,
a
n
d
d
ec
o
d
er
s
in
th
e
Un
et
s
tr
u
ctu
r
e.
Hig
h
s
p
ati
al
r
eso
lu
tio
n
im
ag
e
-
b
ased
m
u
l
ti
-
s
ca
le
lear
n
in
g
at
th
e
g
r
an
u
la
r
lev
el
is
em
p
lo
y
ed
in
th
e
R
esUn
et
tech
n
o
lo
g
y
f
o
r
b
u
ild
i
n
g
d
etec
tio
n
.
I
n
d
ila
ted
R
esUn
et,
wh
ich
ex
p
lo
r
e
d
co
ar
s
e
r
eso
lu
tio
n
im
ag
es
u
s
e
d
ilatio
n
co
n
v
o
l
u
tio
n
in
s
tead
o
f
m
a
x
p
o
o
lin
g
la
y
er
.
A
b
r
o
ad
f
ield
o
f
r
ec
e
p
tio
n
o
r
wo
r
ld
v
iew
is
p
r
o
v
id
e
d
in
d
ilated
co
n
v
o
l
u
tio
n
with
o
u
t
d
r
o
p
p
in
g
th
e
r
eso
lu
tio
n
o
f
th
e
im
a
g
es.
C
o
n
tex
tu
al
in
f
o
r
m
atio
n
is
m
o
r
e
v
is
ib
le
ab
o
u
t
th
e
o
b
jects.
Sm
all
s
ize
o
b
jects
o
v
er
lap
p
e
d
with
o
th
e
r
f
ea
tu
r
es
in
co
a
r
s
e
r
eso
lu
tio
n
im
ag
es.
I
t
is
d
if
f
icu
lt
to
f
in
d
s
m
all
o
b
j
ec
ts
.
A
r
esid
u
al
u
n
it,
s
k
ip
co
n
n
ec
tio
n
s
alo
n
g
with
d
ilated
co
n
v
o
lu
tio
n
o
p
e
r
atio
n
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
7
7
6
I
n
t J I
n
f
&
C
o
m
m
u
n
T
ec
h
n
o
l
,
Vo
l.
1
4
,
No
.
2
,
A
u
g
u
s
t
20
2
5
:
614
-
624
616
ar
e
u
s
ed
in
th
e
m
o
d
el
[
1
3
]
.
I
n
th
is
tech
n
iq
u
e,
r
eso
lv
e
d
th
e
is
s
u
es
o
f
m
is
class
if
icatio
n
.
Fu
r
th
er
m
o
r
e
,
th
e
r
e
ex
is
ts
a
p
o
ten
tial
av
en
u
e
f
o
r
en
h
an
ce
m
e
n
t
co
n
ce
r
n
i
n
g
th
e
p
r
ec
is
io
n
o
f
th
e
tr
ain
in
g
an
d
v
alid
atio
n
d
atasets
.
T
h
is
m
ay
n
o
t
h
a
v
e
b
ee
n
as
co
n
ce
r
n
ed
with
in
cr
ea
s
in
g
th
e
D
AB
E
-
Net
m
eth
o
d
’
s
s
p
ee
d
a
n
d
u
s
in
g
it
to
h
an
d
le
in
cr
ea
s
in
g
ly
d
if
f
icu
lt
R
S
p
ictu
r
e
s
eg
m
e
n
tatio
n
jo
b
s
.
T
h
e
tec
h
n
iq
u
e
o
f
o
v
e
r
s
am
p
lin
g
an
d
au
g
m
en
tatio
n
o
n
th
e
s
m
all
o
b
jects
in
b
u
ild
i
n
g
d
etec
tio
n
m
ay
b
e
u
s
ed
in
th
e
R
es2
-
Un
et
f
r
am
ewo
r
k
.
B
u
ild
in
g
d
if
f
e
r
en
tiatio
n
v
ia
im
ag
e
s
eg
m
en
tatio
n
o
n
h
ig
h
-
r
eso
lu
tio
n
s
atellite
p
ictu
r
es
with
Un
et
ar
c
h
itectu
r
e
is
ca
r
r
ie
d
o
u
t
in
R
es2
-
Un
et.
T
h
is
in
v
esti
g
ates
th
e
p
o
s
s
ib
ili
ty
o
f
au
to
n
o
m
o
u
s
b
u
ild
in
g
ex
tr
ac
tio
n
in
h
ig
h
-
d
en
s
ity
r
esid
en
tial
r
eg
io
n
s
with
ar
tific
ial
in
tellig
en
ce
ass
is
tan
c
e
[
1
4
]
.
2
.
4
.
Dense
-
a
t
t
ent
i
o
n net
wo
rk
s
T
h
is
m
eth
o
d
in
v
o
lv
es
m
o
v
in
g
th
e
“
R
esam
p
ler
”
in
an
atten
tio
n
g
ate
s
o
th
at
it
is
in
lin
e
with
th
e
Sig
m
o
id
f
u
n
ctio
n
f
o
r
t
h
e
p
u
r
p
o
s
e
o
f
ex
tr
ac
tin
g
b
u
ild
in
g
s
.
T
h
ese
n
etwo
r
k
s
a
r
e
ca
p
ab
le
o
f
ef
f
icien
tly
ex
tr
ac
tin
g
b
u
ild
in
g
co
n
to
u
r
s
b
y
a
u
to
n
o
m
o
u
s
ly
lear
n
in
g
v
ar
io
u
s
k
in
d
s
o
f
b
u
ild
in
g
s
tr
u
ctu
r
es
f
r
o
m
h
ig
h
-
r
eso
lu
tio
n
RS
p
h
o
to
s
[
1
5
]
.
T
h
e
u
s
e
o
f
L
id
ar
d
ata
in
co
n
ju
n
ctio
n
with
R
S
R
GB
p
h
o
to
s
is
lim
ited
in
its
ab
ilit
y
to
e
n
h
an
ce
b
u
ild
in
g
s
h
ap
e
d
etec
tio
n
a
n
d
s
im
u
ltan
eo
u
s
ly
y
ield
b
u
ild
in
g
h
eig
h
t
esti
m
ates.
r
aise
th
e
r
eg
u
l
ar
ity
an
d
ac
cu
r
ac
y
o
f
b
u
ild
in
g
ex
tr
ac
tio
n
.
I
n
th
e
co
n
v
o
l
u
tio
n
al
n
eu
r
al
-
d
ee
p
b
el
ief
n
eu
r
al
n
etwo
r
k
s
(
C
NDBN
N)
ap
p
r
o
ac
h
,
d
ata
au
g
m
en
tatio
n
an
d
weig
h
t
r
e
g
u
lar
izatio
n
a
r
e
ap
p
lied
t
o
p
r
e
v
en
t
o
v
er
f
itti
n
g
.
T
h
e
g
en
er
ali
za
tio
n
m
o
d
el
was
r
aised
b
y
it.
T
h
er
e
is
a
s
m
aller
s
am
p
le
s
ize,
an
d
th
e
id
ea
l
s
ize
is
n
o
t
k
n
o
wn
.
Gettin
g
m
o
r
e
ac
cu
r
ate
b
u
ild
in
g
f
o
o
tp
r
i
n
ts
h
as its
lim
itatio
n
s
[
1
6
]
.
2
.
5
.
M
ultilev
el,
m
ulti
-
g
r
a
nu
la
rit
y
DL
a
nd
deep
belief
net
wo
rk
s
Usi
n
g
an
en
s
em
b
le
o
f
co
n
v
o
lu
tio
n
al
n
e
u
r
al
n
etwo
r
k
(
C
NNs
)
,
th
is
tech
n
iq
u
e
u
s
es
au
to
m
atic
b
u
ild
in
g
e
x
tr
ac
tio
n
in
h
ig
h
-
d
e
n
s
ity
r
esid
en
tial
n
eig
h
b
o
u
r
h
o
o
d
s
to
class
if
y
th
e
E
AC
d
ataset
[
1
7
]
.
R
eg
ar
d
in
g
th
e
esti
m
atin
g
is
s
u
es,
th
er
e
is
lim
ited
id
en
tific
atio
n
o
f
th
e
b
u
ild
in
g
’
s
c
o
r
n
e
r
s
with
ad
d
iti
o
n
al
co
n
f
ig
u
r
atio
n
,
wh
ich
ar
e
p
r
im
ar
ily
r
ec
tan
g
u
lar
in
s
h
ap
e
.
T
h
e
r
e
was
a
r
est
r
icted
n
u
m
b
er
o
f
h
ig
h
-
c
o
n
f
i
d
en
ce
s
am
p
les
with
m
o
r
e
u
s
ef
u
l
i
n
f
o
r
m
atio
n
ch
o
s
en
to
tak
e
p
a
r
t
in
th
e
m
o
d
el
’
s
tr
ain
in
g
an
d
i
n
cr
ea
s
e
th
e
ac
cu
r
ac
y
o
f
s
em
an
tic
s
eg
m
en
tatio
n
.
A
r
eg
u
lar
ize
d
b
u
ild
in
g
b
o
u
n
d
a
r
ies
ex
tr
ac
tio
n
f
r
o
m
R
S
d
ata
b
ased
o
n
au
g
m
e
n
t
f
ea
tu
r
e
p
y
r
am
id
n
etwo
r
k
(
AFPN)
an
d
m
o
r
p
h
o
lo
g
ical
co
n
s
tr
ain
t
(
MC
)
is
u
til
ized
in
th
e
AFPN
an
d
MC
ap
p
r
o
ac
h
.
I
n
o
r
d
e
r
to
p
r
ev
en
t
f
ea
tu
r
e
in
f
o
r
m
atio
n
lo
s
s
,
th
is
A
FP
N
i
s
co
n
s
tr
u
cted
to
g
iv
e
m
o
r
e
p
r
ec
is
e
an
d
d
e
n
s
e
g
lo
b
al
f
ea
tu
r
es
f
o
r
s
em
an
tic
s
eg
m
en
tatio
n
task
s
.
I
n
th
is
,
th
e
m
o
r
p
h
o
l
o
g
ical
asp
ec
ts
ar
e
an
aly
s
ed
,
an
d
th
e
b
u
ild
i
n
g
s
h
ap
e
is
m
an
u
ally
class
if
ied
as
lin
ea
r
o
r
cu
r
v
e
d
.
T
o
o
b
tain
th
e
f
in
er
d
ep
ictio
n
o
f
co
n
to
u
r
in
ac
co
r
d
an
ce
with
v
ar
io
u
s
b
u
ild
in
g
s
h
ap
es,
t
h
e
e
x
tr
ac
tio
n
r
esu
lts
ar
e
r
eg
u
lar
ized
.
Fin
a
lly
,
test
s
wer
e
d
o
n
e
o
n
th
e
b
e
n
ch
m
ar
k
d
ataset
to
s
ee
if
th
e
r
eg
u
lar
ize
d
b
u
ild
i
n
g
b
o
u
n
d
ar
y
e
x
tr
ac
tio
n
was a
v
ail
ab
le
[
1
8
]
.
2.
6
.
I
nte
ns
it
y
-
ba
s
ed
s
elec
t
io
n
Ma
n
y
ap
p
licatio
n
s
o
f
th
e
ML
in
R
S
im
ag
er
y
ar
e
s
u
c
ce
s
s
f
u
l
i
n
p
o
s
t
d
is
aster
s
ce
n
ar
io
s
,
f
o
r
ex
tr
ac
tin
g
b
u
ild
i
n
g
s
wh
ich
ar
e
d
am
ag
ed
.
T
r
ad
itio
n
al
m
eth
o
d
s
ca
n
n
o
t
ev
alu
ate
th
e
c
o
m
p
l
ex
ity
o
f
tr
ain
in
g
o
f
th
e
in
p
u
t
d
atasets
.
I
n
th
is
te
ch
n
iq
u
e,
n
u
m
er
ical
s
im
u
latio
n
an
d
m
o
d
elin
g
in
ten
s
ity
is
in
tr
o
d
u
ce
d
f
o
r
th
e
au
to
m
atic
ex
tr
ac
tio
n
o
f
th
e
d
etec
tin
g
d
a
m
ag
ed
b
u
ild
i
n
g
s
.
C
o
n
s
id
er
ed
l
o
w
an
d
h
ig
h
i
n
ten
s
ity
ar
ea
s
an
d
d
ep
en
d
i
n
g
u
p
o
n
i
n
ten
s
ity
o
f
t
h
e
ar
ea
,
b
u
ild
in
g
s
ar
e
ca
teg
o
r
i
ze
d
.
Mo
d
er
ate
t
o
s
ev
er
e
ar
ea
s
o
f
in
ten
s
ity
co
n
tain
s
tr
o
n
g
ly
d
am
ag
ed
b
u
ild
i
n
g
s
.
Su
p
p
o
r
t
v
ec
to
r
m
ac
h
in
es
(
SVM)
clas
s
if
ier
is
tr
ain
ed
with
th
e
u
s
e
o
f
two
p
ar
am
eter
s
o
f
r
eg
u
lar
izatio
n
u
s
in
g
th
ese
tr
ain
in
g
s
am
p
les
f
o
r
lear
n
in
g
ca
lib
r
atin
g
.
I
n
th
is
f
r
am
ewo
r
k
,
m
a
n
u
al
in
ter
p
r
etatio
n
o
f
s
elec
tio
n
o
f
tr
ain
in
g
s
am
p
les
is
a
v
o
id
ed
[
1
9
]
.
I
n
C
F
Net
:
eig
en
v
alu
e
p
r
eser
v
ed
tech
n
iq
u
e,
class
f
ea
tu
r
e
m
o
d
u
le
is
u
s
ed
i
n
d
ec
o
d
er
p
ar
t
o
f
a
tten
tio
n
n
e
two
r
k
f
o
r
s
ep
ar
atio
n
o
f
th
e
b
u
ild
in
g
f
ea
tu
r
es
an
d
th
e
b
ac
k
g
r
o
u
n
d
ar
ea
s
an
d
o
b
j
ec
tiv
e
f
ea
tu
r
e
o
u
tp
u
t
is
co
n
s
id
er
ed
f
o
r
f
in
al
o
u
t
p
u
t.
W
ith
th
e
ch
an
n
el
atten
tio
n
m
o
d
u
le
(
C
AM
)
an
d
th
e
g
lo
b
a
l
r
elatio
n
al
atten
tio
n
m
o
d
u
le
(
GAM
)
,
th
is
atten
tio
n
m
o
d
u
le
is
d
u
al
atten
tiv
e
in
n
atu
r
e.
T
h
e
lo
n
g
-
r
an
g
e
d
ep
e
n
d
en
ce
o
f
p
ix
el
s
p
ac
e
a
n
d
p
ix
el
lo
ca
tio
n
is
r
eso
lv
ed
b
y
th
e
d
u
al
-
atte
n
tio
n
m
ec
h
an
is
m
,
wh
ich
ca
n
ef
f
ec
ti
v
ely
en
a
b
le
th
e
b
ac
k
b
o
n
e
n
et
wo
r
k
to
g
ath
e
r
r
em
o
te
co
n
tex
t
u
al
in
f
o
r
m
atio
n
f
o
r
s
ce
n
e
u
n
d
er
s
tan
d
in
g
[
2
0
]
.
T
h
i
s
is
b
ec
au
s
e
s
o
m
e
ea
r
lier
r
es
ea
r
ch
b
y
ac
a
d
em
ics
p
r
o
d
u
ce
d
lo
ca
l
f
ea
tu
r
es
th
at
co
u
ld
ca
u
s
e
th
e
n
etwo
r
k
to
in
ter
p
r
et
th
e
co
n
te
n
t
in
co
r
r
e
ctly
d
u
e
to
its
lim
ited
p
er
ce
p
tu
al
f
ield
.
A
f
o
u
r
-
lay
er
co
d
in
g
lay
er
th
at
u
s
es
d
o
wn
s
am
p
lin
g
t
o
ex
tr
ac
t
f
ea
tu
r
es
m
a
k
es
u
p
th
e
e
n
tire
b
ac
k
b
o
n
e
n
et
wo
r
k
.
I
t
allo
ws
th
e
n
etwo
r
k
to
b
e
tr
ain
ed
with
o
u
t
d
if
f
icu
lty
o
win
g
to
g
r
ad
ien
t
d
is
ap
p
ea
r
an
ce
a
n
d
ex
p
lo
s
io
n
b
y
g
o
in
g
to
o
d
ee
p
,
all
th
e
wh
ile
g
u
ar
an
teei
n
g
f
ea
t
u
r
e
ex
tr
ac
tio
n
.
2.
7
.
Aut
o
enco
ders
I
n
o
r
d
er
t
o
r
ein
f
o
r
ce
th
e
f
ea
t
u
r
es,
a
s
p
atial
atten
tio
n
f
u
s
io
n
m
o
d
u
le
was
em
p
lo
y
ed
i
n
c
o
n
ju
n
ctio
n
with
an
e
n
co
d
e
r
a
n
d
a
d
ec
o
d
e
r
co
m
p
o
n
e
n
t
m
a
d
e
o
f
lig
h
tweig
h
t
Den
s
eNe
ts
.
p
r
o
p
ag
atio
n
a
n
d
th
e
in
tr
o
d
u
ctio
n
o
f
h
ig
h
er
-
lev
el
f
ea
t
u
r
e
d
ata
t
o
m
u
f
f
le
n
o
is
e
an
d
l
o
w
-
lev
el
f
ea
tu
r
es
[
5
]
.
C
h
an
g
e
v
ec
to
r
an
aly
s
is
(
C
VA)
o
f
o
p
tical
s
atellite
im
ag
e
s
was
u
s
ed
to
r
e
d
u
ce
e
r
r
o
r
s
an
d
o
m
is
s
io
n
s
b
y
c
o
m
b
in
in
g
m
o
r
e
o
p
ti
m
ized
d
ee
p
n
e
u
r
al
n
etwo
r
k
s
with
p
r
e
-
o
r
p
o
s
t
-
p
r
o
ce
s
s
in
g
m
eth
o
d
s
.
T
h
is
h
elp
ed
to
o
v
er
co
m
e
th
e
is
s
u
e
o
f
ef
f
ec
tiv
e
ex
tr
ac
tio
n
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J I
n
f
&
C
o
m
m
u
n
T
ec
h
n
o
l
I
SS
N:
2252
-
8
7
7
6
Tech
n
iq
u
es o
f d
ee
p
lea
r
n
in
g
n
eu
r
a
l n
etw
o
r
k
b
a
s
ed
b
u
ild
i
n
g
f
ea
tu
r
e
ex
tr
a
ctio
n
fr
o
m
…
(
S
h
r
in
iva
s
K
h
a
n
d
a
r
e
)
617
ac
cu
r
ac
y
a
n
d
e
f
f
icien
cy
,
wh
ic
h
was
af
f
ec
ted
b
y
c
o
m
p
lex
b
ac
k
g
r
o
u
n
d
s
,
s
p
ec
ial
b
u
ild
in
g
s
,
an
d
u
n
r
em
a
r
k
ab
le
b
u
ild
in
g
s
[
2
1
]
.
R
ath
er
th
an
h
an
d
lin
g
th
e
n
etwo
r
k
an
d
p
r
e
-
an
d
p
o
s
t
-
p
r
o
ce
s
s
in
g
tec
h
n
iq
u
es
in
d
ep
e
n
d
en
tly
,
b
o
th
a
r
e
n
o
t
tak
en
in
to
ac
co
u
n
t
at
th
e
s
am
e
tim
e
f
o
r
in
c
r
ea
s
ed
ac
cu
r
ac
y
a
n
d
e
f
f
icien
cy
[
2
2
]
.
T
o
p
in
p
o
in
t
ex
ac
t
ch
an
g
e
s
in
lan
d
-
u
s
e
an
d
lan
d
-
co
v
er
(
o
r
cr
o
p
p
in
g
p
atter
n
)
u
s
in
g
th
is
tech
n
iq
u
e.
Ap
p
r
o
p
r
iate
ca
teg
o
r
izatio
n
tech
n
iq
u
es
ar
e
n
o
t
ex
ten
d
e
d
to
d
if
f
er
en
tiate
b
etwe
en
v
ar
i
o
u
s
s
o
r
ts
o
f
ch
an
g
es.
T
h
e
ef
f
ec
tiv
en
ess
o
f
th
e
au
to
en
co
d
er
-
en
h
an
ce
d
d
is
tan
ce
m
ea
s
u
r
em
en
ts
with
in
a
d
ee
p
m
etr
ic
lear
n
in
g
f
r
am
ewo
r
k
(
s
u
ch
as
th
e
s
iam
ese
n
etwo
r
k
o
r
t
r
ip
let
n
etwo
r
k
)
m
ay
n
o
t
h
a
v
e
b
ee
n
in
v
esti
g
ated
in
t
h
is
tech
n
iq
u
e.
T
h
is
m
eth
o
d
u
s
es
a
m
u
lti
-
task
[
2
3
]
en
c
o
d
er
-
d
ec
o
d
er
n
etwo
r
k
to
s
ea
r
ch
f
o
r
v
ar
io
u
s
au
t
o
en
co
d
e
r
ar
ch
itectu
r
es
in
th
e
r
ec
o
n
s
tr
u
cti
o
n
o
f
s
p
ec
tr
al
d
ata,
s
u
ch
as c
o
n
v
o
lu
t
io
n
al
au
to
en
c
o
d
er
s
[
2
4
]
.
2
.
8
.
M
ulti
s
ca
le
f
ea
t
ure
t
r
a
ns
f
o
rm
er
B
u
ild
in
g
ex
tr
ac
tio
n
f
r
o
m
h
ig
h
s
p
atial
r
eso
lu
tio
n
,
s
atellite
im
ag
es
is
a
co
m
p
lex
task
d
u
e
to
th
e
d
if
f
er
en
ce
in
s
h
a
p
e
o
f
b
u
ild
in
g
s
an
d
co
lo
r
s
.
T
h
er
e
ar
e
c
o
m
p
lex
b
ac
k
g
r
o
u
n
d
o
b
jects
wh
ich
d
eter
th
e
b
u
ild
in
g
f
ea
tu
r
es
f
r
o
m
s
elec
tio
n
.
T
h
es
e
m
eth
o
d
s
ar
e
b
ased
o
n
D
L
tech
n
iq
u
es
with
en
c
o
d
er
a
n
d
d
ec
o
d
er
s
tr
u
ctu
r
e.
I
n
th
ese
d
etails
o
f
th
e
s
m
all
b
u
ild
in
g
s
is
n
o
t
ca
p
tu
r
ed
.
T
h
e
b
u
ild
in
g
with
b
lu
r
r
ed
b
o
u
n
d
ar
ies
is
ca
p
tu
r
ed
.
Fo
r
th
is
,
p
r
o
p
o
s
ed
f
r
e
q
u
en
c
y
-
s
p
atial
lear
n
in
g
tr
a
n
s
f
o
r
m
e
r
f
o
r
e
x
tr
ac
tin
g
f
ea
t
u
r
es
o
f
b
u
ild
in
g
s
.
I
n
th
is
tech
n
iq
u
e,
d
if
f
e
r
en
t
f
r
eq
u
en
cy
f
ea
tu
r
es
a
n
d
m
u
lti
-
s
ca
le
f
ea
tu
r
es
ar
e
ca
p
tu
r
ed
an
d
s
y
n
t
h
esized
.
I
n
th
is
,
s
p
ik
in
g
co
n
v
o
l
u
tio
n
is
u
s
ed
to
en
h
a
n
ce
f
r
e
q
u
en
c
y
.
I
t
d
is
tin
g
u
is
h
es
th
e
m
u
ltis
ca
le
f
ea
tu
r
es
f
r
o
m
th
e
b
ac
k
g
r
o
u
n
d
f
ea
tu
r
es.
Ma
s
k
ed
atten
tio
n
tr
a
n
s
f
o
r
m
er
r
ef
in
es
th
e
b
o
u
n
d
ar
y
f
ea
tu
r
es.
B
y
tr
ai
n
in
g
a
n
d
as
s
ess
in
g
p
h
o
to
s
with
v
ar
y
in
g
s
p
atial
r
eso
lu
tio
n
s
f
r
o
m
v
ar
io
u
s
r
esear
ch
a
r
ea
s
an
d
d
atasets
,
th
e
p
r
ac
tical
tr
a
n
s
f
er
ab
ilit
y
o
f
t
h
is
m
eth
o
d
was
ass
es
s
ed
in
th
is
way
.
T
h
e
b
o
u
n
d
ar
y
lo
s
s
f
u
n
ctio
n
tech
n
iq
u
e
was
e
m
p
lo
y
ed
to
o
p
tim
ize
an
d
im
p
r
o
v
e
th
e
m
u
lticlas
s
s
em
an
tic
s
eg
m
en
tatio
n
o
f
u
r
b
a
n
r
eg
i
o
n
s
b
y
ex
tr
ac
tin
g
b
u
ild
in
g
s
in
b
o
r
d
er
ar
ea
s
u
s
in
g
a
d
er
iv
ativ
e
b
o
u
n
d
ar
y
lo
s
s
f
u
n
ctio
n
.
T
h
e
s
u
g
g
ested
b
o
u
n
d
ar
y
lo
s
s
f
u
n
ctio
n
th
at
o
r
ig
in
ated
f
r
o
m
th
e
s
eg
m
en
tatio
n
n
etwo
r
k
is
c
o
m
p
u
ted
u
s
in
g
th
e
d
is
tan
ce
t
r
an
s
f
o
r
m
im
ag
e
(
DT
I
)
,
wh
ic
h
is
f
o
r
m
ed
f
r
o
m
p
r
ed
icted
an
d
g
r
o
u
n
d
-
tr
u
th
p
ic
tu
r
es
[
2
5
]
.
2
.
9
.
M
ultit
a
rg
et
do
m
a
in a
da
pta
t
io
n
W
ith
th
is
m
eth
o
d
,
R
S
ad
d
r
ess
es
v
ast
d
if
f
er
en
ce
s
in
lo
ca
tio
n
,
tim
e
o
f
y
ea
r
f
o
r
ac
q
u
is
itio
n
,
an
d
ab
u
n
d
a
n
ce
o
f
s
en
s
o
r
s
.
Du
e
to
th
e
ch
allen
g
e
o
f
o
b
tain
in
g
lab
eled
d
ata
th
at
ac
cu
r
ately
ca
p
tu
r
es
ev
er
y
ev
e
n
t,
d
ata
-
h
u
n
g
r
y
DL
m
o
d
els
ar
e
f
r
eq
u
en
tly
tr
ain
ed
u
s
in
g
lab
eled
d
ata
f
r
o
m
a
s
o
u
r
ce
d
o
m
ain
th
at
is
co
n
s
tr
ain
ed
in
th
e
af
o
r
em
en
tio
n
ed
way
s
.
T
h
i
s
s
tr
ateg
y
allo
ws
d
o
m
ain
ad
a
p
tatio
n
(
DA)
tech
n
i
q
u
es
to
m
o
d
if
y
th
ese
m
o
d
els
f
o
r
u
s
e
o
n
tar
g
et
d
o
m
ai
n
s
th
at
h
av
e
d
is
tr
ib
u
tio
n
s
th
at
d
if
f
e
r
f
r
o
m
th
e
s
o
u
r
ce
d
o
m
ain
.
I
t
is
n
ec
ess
ar
y
to
tr
ain
a
d
if
f
er
en
t
tar
g
et
class
if
ier
f
o
r
ev
er
y
tar
g
et
d
o
m
ain
b
ec
au
s
e
th
e
m
ajo
r
ity
o
f
R
S
DL
tech
n
i
q
u
es
ar
e
m
ea
n
t
f
o
r
s
in
g
le
tar
g
ets.
T
h
is
is
les
s
en
ed
b
y
m
u
ltit
ar
g
et
DA,
wh
er
e
s
ev
er
al
u
n
lab
ele
d
tar
g
et
d
o
m
ai
n
s
ar
e
lear
n
t
u
s
in
g
a
s
in
g
le
class
if
ier
.
T
h
is
m
eth
o
d
b
u
ild
s
a
m
u
ltit
ar
g
et
class
if
ier
to
ef
f
icien
tly
co
llect
c
h
ar
ac
ter
is
tics
f
r
o
m
v
ar
io
u
s
u
n
lab
eled
tar
g
et
d
o
m
ain
s
an
d
th
e
lab
eled
s
o
u
r
ce
.
T
h
is
m
eth
o
d
m
ak
es
u
s
e
o
f
co
teac
h
in
g
,
a
g
r
a
p
h
n
eu
r
al
n
etwo
r
k
-
b
ased
m
et
h
o
d
o
lo
g
y
th
at
m
ay
b
e
u
s
ed
to
u
n
lab
eled
d
ata.
I
f
th
e
n
etwo
r
k
f
in
d
s
it
ea
s
ier
to
ad
ap
t
to
th
e
clo
s
est
tar
g
et
d
o
m
ain
,
it
will
in
itially
ad
ap
t
o
n
th
e
ea
s
ier
tar
g
et
d
o
m
ain
s
u
s
in
g
th
e
s
e
q
u
en
tial
ad
ap
tatio
n
tech
n
iq
u
e
th
at
was in
tr
o
d
u
ce
d
[
2
6
]
.
2
.
1
0
.
I
m
pro
v
ed
ps
eudo
m
a
s
k
s
g
ener
a
t
io
n
T
h
is
m
eth
o
d
co
m
b
i
n
es
g
ated
co
n
v
o
l
u
tio
n
(
GC
)
an
d
ad
v
er
s
ar
ial
clim
b
in
g
(
AC
)
in
a
p
o
o
r
ly
s
u
p
er
v
is
ed
b
u
ild
in
g
e
x
tr
ac
tio
n
f
r
am
ewo
r
k
.
I
n
o
r
d
er
to
en
h
a
n
ce
p
s
eu
d
o
m
ask
g
en
er
atio
n
an
d
r
eliab
ly
ex
tr
ac
t
b
u
ild
in
g
s
f
r
o
m
h
ig
h
r
eso
lu
tio
n
(
HR
)
R
S
p
ictu
r
es,
th
e
A
C
GC
f
r
am
ewo
r
k
tr
ea
ts
im
ag
e
-
lev
el
lab
els
as
wea
k
s
u
p
er
v
is
io
n
.
T
h
er
e
a
r
e
th
r
ee
m
ain
p
ar
ts
to
th
e
en
tire
p
ip
el
in
e.
B
u
ild
in
g
class
ac
tiv
atio
n
m
ap
s
(
C
AM
s
)
ar
e
cr
ea
ted
an
d
o
p
tim
ized
with
AC
in
th
e
f
i
r
s
t
co
m
p
o
n
en
t.
Nex
t,
in
th
e
s
ec
o
n
d
co
m
p
o
n
e
n
t,
a
g
ated
co
n
v
o
lu
tio
n
m
o
d
u
le
(
GC
M)
is
u
s
ed
to
o
b
tain
th
e
class
b
o
u
n
d
ar
y
m
a
p
s
(
C
B
Ms)
o
f
th
e
b
u
ild
in
g
s
,
an
d
in
t
h
e
th
ir
d
co
m
p
o
n
e
n
t,
C
AM
s
an
d
p
air
wis
e
af
f
in
ities
f
r
o
m
th
e
C
B
Ms
ar
e
u
s
ed
to
ex
p
an
d
th
e
p
s
eu
d
o
m
ask
s
o
f
t
h
e
b
u
ild
in
g
s
.
T
h
e
g
en
er
ate
d
p
s
eu
d
o
m
ask
s
ar
e
th
e
th
r
ee
co
m
p
o
n
en
ts
o
f
th
is
tech
n
iq
u
e.
T
h
e
g
r
o
u
n
d
tr
u
th
(
GT
)
o
f
tr
ain
in
g
s
am
p
les
f
o
r
a
co
n
v
e
n
tio
n
al
s
u
p
er
v
is
ed
s
eg
m
en
tat
io
n
m
o
d
el
is
o
b
tain
e
d
f
r
o
m
th
e
p
s
eu
d
o
m
ask
s
.
T
h
e
p
r
o
g
r
a
m
was
tau
g
h
t
to
r
ec
o
g
n
ize
b
u
ild
in
g
s
u
s
in
g
th
e
s
em
an
tic
in
f
o
r
m
atio
n
f
o
u
n
d
in
HR
p
ict
u
r
es.
T
o
b
e
m
o
r
e
p
r
ec
is
e,
HR
R
S
p
h
o
to
s
wer
e
f
e
d
in
to
t
h
e
tr
ain
e
d
s
u
p
er
v
is
ed
s
eg
m
e
n
tatio
n
m
o
d
el,
wh
ich
p
r
o
d
u
ce
d
b
u
ild
in
g
m
ask
s
th
at
s
h
o
wed
th
e
ex
tr
ac
ted
b
u
ild
in
g
s
[
2
7
]
.
2
.
1
1
.
M
ultit
a
s
k
lea
rning
T
h
is
m
eth
o
d
in
v
o
lv
ed
class
if
y
in
g
d
is
aster
im
ag
es
f
r
o
m
s
o
cial
m
ed
ia
im
ag
e
s
tr
ea
m
s
an
d
u
s
in
g
a
m
u
lti
-
lev
el
f
ea
tu
r
e
en
h
a
n
ce
m
en
t
n
etwo
r
k
(
MFE
Net)
f
o
r
l
ea
r
n
in
g
f
r
o
m
im
ag
es
b
ased
o
n
C
NN
-
lik
e
C
N
N
m
o
d
u
les.
Usi
n
g
n
in
e
d
is
tin
ct
p
r
e
-
t
r
ain
ed
m
o
d
els,
class
if
icatio
n
r
es
u
lts
ar
e
p
r
o
v
id
e
d
t
h
at
d
o
n
o
t
r
eq
u
ir
e
ex
p
lo
r
atio
n
an
d
ca
n
b
e
u
s
ed
a
s
a
s
tan
d
ar
d
f
o
r
f
u
r
th
er
r
esear
ch
.
T
h
is
m
eth
o
d
,
wh
ic
h
h
as
n
o
t
b
ee
n
i
n
v
esti
g
ated
b
ef
o
r
e,
co
m
b
i
n
es
tr
an
s
f
er
lear
n
in
g
o
r
p
o
o
r
ly
s
u
p
er
v
is
ed
lea
r
n
in
g
with
th
e
ex
is
tin
g
g
l
o
b
al
d
ata
s
et
o
f
lan
d
s
li
d
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
7
7
6
I
n
t J I
n
f
&
C
o
m
m
u
n
T
ec
h
n
o
l
,
Vo
l.
1
4
,
No
.
2
,
A
u
g
u
s
t
20
2
5
:
614
-
624
618
R
S
to
cr
ea
te
L
M
n
etwo
r
k
s
w
ith
im
p
r
o
v
ed
g
en
er
aliza
tio
n
p
er
f
o
r
m
a
n
ce
.
Similar
to
th
is
,
to
o
b
tain
tim
elin
ess
an
d
ac
cu
r
ac
y
in
b
u
ild
in
g
d
a
m
ag
e,
p
er
t
u
r
b
ed
d
u
al
m
ea
n
teac
h
er
s
u
s
es
m
u
ltit
ask
-
b
ased
s
em
i
-
s
u
p
er
v
is
ed
s
em
an
tic
s
eg
m
en
tatio
n
[
2
8
]
,
f
r
am
ewo
r
k
tech
n
iq
u
e,
co
n
s
is
ten
cy
r
eg
u
lar
izatio
n
-
b
ased
s
em
i
-
s
u
p
er
v
is
ed
f
r
am
ewo
r
k
with
m
u
ltit
ask
s
em
an
tic
s
eg
m
en
tatio
n
.
T
o
co
n
f
ir
m
th
e
ef
f
icac
y
o
f
th
is
ap
p
r
o
ac
h
,
it
m
ig
h
t
n
o
t
b
e
in
v
esti
g
ated
o
n
f
u
r
t
h
er
d
atasets
with
v
ar
io
u
s
d
is
aster
s
ce
n
ar
io
s
(
s
u
ch
as
ea
r
th
q
u
ak
es
an
d
la
n
d
s
lid
es)
an
d
d
ata
s
o
u
r
ce
s
(
s
u
ch
as
ae
r
ial
an
d
U
AV
d
ata)
[
2
9
]
.
Ad
d
itio
n
ally
,
wh
en
d
ea
lin
g
with
a
h
ig
h
n
u
m
b
er
o
f
u
n
la
b
eled
s
am
p
les
to
m
ee
t
em
e
r
g
en
c
y
n
ee
d
s
in
cr
is
is
cir
cu
m
s
tan
ce
s
,
i
t
m
ig
h
t
n
o
t
i
n
cr
ea
s
e
co
m
p
u
ter
ef
f
icien
cy
.
C
o
u
l
d
o
f
f
er
a
la
n
d
s
lid
e
d
ataset
th
at
is
m
o
r
e
ac
cu
r
ate,
e
n
h
an
ce
th
e
alg
o
r
ith
m
’
s
ab
ilit
y
to
p
r
e
d
ict
lan
d
s
lid
es,
an
d
p
r
ev
en
t
th
e
d
ataset
f
r
o
m
g
r
o
w
in
g
.
W
o
r
k
e
d
with
b
o
t
h
lar
g
e
a
n
d
s
m
all
p
ictu
r
e
s
izes,
an
al
y
s
in
g
p
o
ten
tial
tr
a
d
e
-
o
f
f
s
b
etwe
en
m
em
o
r
y
co
n
s
u
m
p
tio
n
an
d
m
o
d
el
p
er
f
o
r
m
an
ce
with
v
ar
y
in
g
s
u
p
p
o
r
t im
ag
e
s
i
ze
s
[
3
0
]
.
2
.
1
2
.
M
ulti
-
s
ca
le
f
ea
t
ure
m
a
p a
g
g
re
g
a
t
io
n a
nd
s
up
er
v
is
e
d do
m
a
in a
da
pta
t
io
n
co
nv
o
l
utio
na
l
net
wo
rk
I
m
ag
e
s
eg
m
en
tatio
n
was
d
o
n
e
u
s
in
g
th
e
K
-
m
ea
n
s
clu
s
ter
in
g
alg
o
r
ith
m
,
an
d
t
h
r
ee
d
is
tin
c
t
m
eth
o
d
s
-
VGG
-
1
6
,
VGG
-
1
9
,
an
d
NAS
Net
-
wer
e
u
tili
ze
d
to
id
e
n
tify
d
am
a
g
ed
s
tr
u
ct
u
r
es
f
o
llo
win
g
an
ea
r
th
q
u
a
k
e
.
I
t
m
ay
n
o
t
h
av
e
b
ee
n
in
v
esti
g
ated
to
im
p
r
o
v
e
alg
o
r
ith
m
p
er
f
o
r
m
an
ce
,
p
ar
ticu
la
r
ly
wh
en
class
if
y
in
g
les
s
d
am
ag
ed
an
d
n
o
r
m
al
b
u
ild
i
n
g
s
to
p
r
ev
e
n
t
m
is
class
if
icatio
n
[
3
1
]
.
T
h
is
tech
n
iq
u
e,
f
o
r
th
e
m
u
lticlas
s
class
if
icatio
n
jo
b
o
f
d
em
o
lis
h
ed
s
tr
u
ctu
r
es
f
o
llo
win
g
a
n
u
n
a
n
ticip
ated
tr
ag
ed
y
,
f
o
cu
s
es
o
n
an
ef
f
icien
t
f
ea
tu
r
e
ex
tr
ac
tio
n
an
d
s
elec
tio
n
p
r
o
ce
s
s
in
co
n
ju
n
ctio
n
with
a
C
NN
m
o
d
el
[
3
2
]
.
T
h
is
m
eth
o
d
u
s
es
r
o
ad
d
am
ag
e
t
o
q
u
ick
ly
r
ed
ir
ec
t
th
e
s
h
o
r
test
r
o
u
te
b
etwe
en
th
e
h
elp
team
an
d
th
e
u
r
g
en
t
ca
r
e
f
ac
il
ity
th
at
is
n
ee
d
ed
.
Un
s
u
p
er
v
is
ed
lear
n
in
g
is
n
o
t
in
v
esti
g
ated
in
t
h
is
tech
n
iq
u
e
to
im
p
r
o
v
e
ex
tr
ac
tio
n
p
e
r
f
o
r
m
an
ce
in
u
n
lab
ele
d
d
atasets
.
Un
s
u
p
er
v
is
ed
lear
n
i
n
g
is
n
o
t
in
v
esti
g
ated
in
th
is
tech
n
iq
u
e
t
o
im
p
r
o
v
e
ex
tr
ac
ti
o
n
p
e
r
f
o
r
m
an
ce
i
n
u
n
lab
eled
d
atasets
.
T
h
is
m
eth
o
d
co
n
ce
n
tr
ated
o
n
in
c
r
ea
s
in
g
th
e
d
iv
er
s
ity
o
f
tr
ain
in
g
s
am
p
les
an
d
d
ev
elo
p
in
g
m
o
r
e
ef
f
icien
t
n
etwo
r
k
to
p
o
lo
g
ies
to
s
y
n
th
esize
tex
tu
r
al
attr
ib
u
tes
in
o
r
d
er
to
im
p
r
o
v
e
th
e
ex
tr
ac
tio
n
p
er
f
o
r
m
an
ce
o
f
s
m
all
b
u
ild
in
g
in
s
tan
ce
s
[
3
3
]
.
I
n
o
r
d
e
r
to
p
r
o
d
u
ce
s
ig
n
if
ican
tly
b
etter
r
esu
lts
,
th
e
DR
R
NN
ap
p
r
o
ac
h
is
n
o
t
i
n
v
esti
g
ated
i
n
th
is
m
eth
o
d
o
lo
g
y
f
o
r
ea
s
ily
co
n
f
u
s
ed
class
if
icatio
n
s
in
clu
d
in
g
in
d
u
s
tr
ial
lan
d
,
u
r
b
an
r
esid
en
tial
ar
ea
s
,
an
d
r
u
r
al
r
esid
en
tial
ar
ea
s
.
B
u
ild
in
g
r
e
g
io
n
s
f
r
o
m
h
ig
h
-
r
es
o
lu
tio
n
R
S
im
ag
es
(
m
u
lti
-
s
ca
le
a
n
d
m
u
lti
-
r
eso
lu
tio
n
f
ea
tu
r
es,
en
h
a
n
ce
th
e
i
n
f
o
r
m
atio
n
in
ter
ac
tio
n
b
etwe
en
d
if
f
e
r
en
t
s
ca
le
f
ea
tu
r
es,
d
ec
r
ea
s
e
th
e
d
if
f
e
r
en
ce
b
etwe
en
th
e
e
x
tr
ac
ted
r
esu
l
ts
an
d
th
e
g
r
o
u
n
d
tr
u
th
s
)
ac
cu
r
ately
u
s
in
g
C
NNs
is
p
o
s
s
ib
le
with
th
e
m
u
lti
-
s
ca
le
r
ec
u
r
r
e
n
t
co
n
d
itio
n
al
g
e
n
er
a
tiv
e
ad
v
er
s
ar
ial
n
etwo
r
k
tech
n
iq
u
e.
T
h
is
m
eth
o
d
was o
n
ly
ap
p
licab
le
to
lear
n
in
g
th
at
was u
n
s
u
p
er
v
is
ed
in
th
e
ex
tr
ac
tio
n
o
f
lab
elled
d
ataset
[
3
4
]
.
2
.
1
3
.
Ca
s
s
if
ica
t
io
n net
wo
rk
D
-
Net
a
nd
t
a
nd
em
s
t
it
ching
Dis
aster
ty
p
es
ar
e
id
en
tifie
d
i
n
th
e
ab
o
v
e
u
tili
zin
g
D
-
co
n
v
,
D
-
lin
ea
r
,
D
-
m
o
d
el,
an
d
D
-
lay
er
m
o
d
u
les.
Usi
n
g
th
is
m
eth
o
d
,
d
if
f
er
e
n
t
d
is
aster
ty
p
es
ca
n
b
e
id
en
tifie
d
u
s
in
g
Go
o
g
le
E
ar
t
h
s
atelli
te
p
h
o
to
s
o
r
o
th
e
r
o
n
lin
e
d
ata
b
ases
.
Natu
r
al
d
is
aster
s
in
th
is
ap
p
r
o
ac
h
,
k
n
o
wn
as
c
ass
if
icatio
n
n
et,
h
av
e
a
s
ig
n
if
ican
t
in
f
lu
en
c
e
o
n
p
eo
p
le
’
s
liv
es
an
d
p
r
o
p
e
r
ty
,
th
u
s
p
r
o
m
p
t
a
n
d
ef
f
icien
t
h
an
d
lin
g
o
f
d
is
aster
ca
teg
o
r
ies
is
im
p
er
ativ
e.
T
h
is
m
eth
o
d
u
s
es
th
e
D
-
c
o
n
v
,
D
-
l
in
ea
r
,
D
-
m
o
d
el,
an
d
D
-
l
ay
er
m
o
d
u
les
to
cr
ea
te
a
n
ew
d
is
aster
ca
s
s
if
icatio
n
n
etwo
r
k
,
o
r
d
is
aster
ca
s
s
if
icatio
n
n
et
(
D
-
Net
)
,
b
y
tan
d
e
m
s
titch
in
g
.
I
t
is
d
is
co
v
er
e
d
th
at
th
e
d
is
aster
ca
s
s
if
icatio
n
n
et
p
er
f
o
r
m
s
b
et
ter
th
an
C
NN
in
th
is
m
eth
o
d
o
lo
g
y
wh
e
n
th
e
s
u
g
g
ested
m
et
h
o
d
is
c
o
m
p
ar
e
d
to
“
C
NN
”
an
d
“
T
r
an
s
f
o
r
m
er
”
t
h
r
o
u
g
h
o
u
t
th
e
ex
p
er
i
m
en
t.
W
h
en
Mo
b
ileNet_
v
2
,
wh
ich
p
e
r
f
o
r
m
s
b
est
o
n
th
e
class
if
icatio
n
d
ataset,
an
d
C
C
T
n
etwo
r
k
ar
e
co
m
p
a
r
ed
with
d
is
aster
ca
s
s
if
icatio
n
n
et
o
n
th
e
f
ash
io
n
_
m
n
is
t
an
d
C
I
FAR
_
1
0
0
p
u
b
lic
d
atas
ets,
r
esp
ec
tiv
ely
,
th
e
r
esu
lts
d
em
o
n
s
tr
ate
th
at
d
i
s
aster
ca
s
s
if
icatio
n
n
et
ca
n
s
till
ac
h
iev
e
th
e
s
tate
-
of
-
th
e
-
ar
t
cl
ass
if
icatio
n
ef
f
ec
t.
I
t
was
also
f
o
u
n
d
th
at
th
is
tech
n
iq
u
e
ac
h
iev
es
th
e
ef
f
ec
t
o
f
s
tate
-
of
-
th
e
-
ar
t
(
SOTA
)
o
n
th
e
d
is
aster
d
ataset
[
3
5
]
.
2
.
1
4
.
G
ener
a
t
iv
e
a
dv
er
s
a
ria
l net
wo
rk
s
a
nd
Web
-
Net
Usi
n
g
th
e
u
ltra
-
h
ier
ar
c
h
ical
s
am
p
lin
g
(
UHS)
b
lo
ck
,
th
is
tech
n
iq
u
e
ex
tr
ac
ts
m
o
r
e
p
r
e
cisely
an
d
s
h
ar
p
ly
r
ec
o
g
n
ized
b
u
ild
i
n
g
s
f
r
o
m
th
e
p
r
ed
ictio
n
m
ap
s
,
in
clu
d
in
g
b
u
ild
in
g
ar
ea
s
th
at
ar
e
o
b
s
cu
r
ed
b
y
s
h
ad
o
ws.
Hig
h
-
r
eso
lu
tio
n
ae
r
i
al
im
ag
er
y
s
h
o
ws
s
tr
u
ctu
r
es
with
in
cr
ed
ib
ly
in
tr
icate
m
o
r
p
h
o
lo
g
ical
f
ea
tu
r
es,
in
clu
d
in
g
c
u
r
v
es,
s
tr
aig
h
t
lin
e
s
,
an
d
o
r
ien
tatio
n
s
[
3
6
]
.
T
h
e
FC
N
-
b
ased
n
etwo
r
k
s
ar
e
u
n
ab
l
e
to
d
ir
ec
tly
ex
tr
ac
t
th
e
ch
ar
ac
ter
is
tics
,
an
d
it
is
a
cr
itical
is
s
u
e
to
f
ig
u
r
e
o
u
t
h
o
w
to
in
co
r
p
o
r
ate
m
o
r
p
h
o
lo
g
ical
ch
ar
ac
ter
is
tics
in
to
C
NN
s
tr
u
ctu
r
e.
h
ig
h
lim
itatio
n
s
o
n
m
o
r
p
h
o
lo
g
ical
tr
aits
i
n
th
e
v
ar
io
u
s
ar
ch
itectu
r
al
m
o
r
p
h
o
lo
g
ies
o
f
th
e
s
tr
u
ctu
r
es in
th
e
ac
cid
en
t sit
es [
3
7
]
.
2
.
1
5
.
DCNNs
-
ba
s
ed
edg
e
de
t
ec
t
io
n
a
nd
P
CA
T
h
is
m
eth
o
d
’
s
au
to
m
atic
v
ec
to
r
izatio
n
o
f
b
u
ild
in
g
s
h
ap
es
f
r
o
m
v
er
y
h
ig
h
r
eso
l
u
tio
n
(
VHR)
RS
im
ag
er
y
is
cr
u
cial
f
o
r
a
v
ar
i
ety
o
f
ap
p
licatio
n
s
,
i
n
clu
d
in
g
g
eo
d
atab
ase
u
p
d
ates
an
d
u
r
b
an
m
an
a
g
em
en
t.
Alth
o
u
g
h
cr
ea
tin
g
e
d
g
e
d
etec
tio
n
u
s
in
g
DC
NNs
h
as
b
ee
n
ef
f
ec
tiv
e
r
ec
e
n
tly
,
th
e
r
esu
lts
ar
e
n
o
t
v
ec
to
r
ized
m
ap
s
b
u
t
r
ath
er
r
aster
im
ag
e
s
,
wh
ich
d
o
n
o
t
s
u
it
th
e
n
ee
d
s
o
f
m
a
n
y
a
p
p
licatio
n
s
.
R
aster
p
ictu
r
es
ca
n
b
e
tr
an
s
f
o
r
m
ed
in
to
v
ec
t
o
r
m
a
p
s
u
s
in
g
ce
r
tain
m
eth
o
d
s
;
th
e
s
e
v
ec
to
r
m
ap
s
f
r
eq
u
e
n
tly
in
clu
d
e
an
e
x
ce
s
s
iv
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J I
n
f
&
C
o
m
m
u
n
T
ec
h
n
o
l
I
SS
N:
2252
-
8
7
7
6
Tech
n
iq
u
es o
f d
ee
p
lea
r
n
in
g
n
eu
r
a
l n
etw
o
r
k
b
a
s
ed
b
u
ild
i
n
g
f
ea
tu
r
e
ex
tr
a
ctio
n
fr
o
m
…
(
S
h
r
in
iva
s
K
h
a
n
d
a
r
e
)
619
n
u
m
b
er
o
f
v
ec
to
r
p
o
in
ts
an
d
asy
m
m
etr
ical
s
h
ap
es.
A
p
r
i
n
cip
al
co
m
p
o
n
en
t
an
aly
s
is
(
PC
A)
-
b
ased
co
r
n
er
ex
tr
ac
tio
n
m
et
h
o
d
an
d
DC
NNs
-
b
ased
b
u
ild
in
g
ed
g
e
d
etec
tio
n
wer
e
c
o
m
b
in
e
d
to
cr
ea
te
a
b
u
il
d
in
g
f
o
r
m
v
ec
to
r
izatio
n
h
ier
ar
c
h
y
th
at
allo
wed
f
o
r
th
e
q
u
ick
ex
tr
a
ctio
n
o
f
b
u
ild
in
g
co
r
n
er
s
f
r
o
m
b
u
ild
in
g
ed
g
es.
T
ests
co
n
d
u
cted
o
n
th
e
J
ian
g
B
ei
n
ew
ar
ea
b
u
ild
in
g
s
an
d
m
ass
ac
h
u
s
etts
b
u
ild
in
g
s
d
atasets
d
em
o
n
s
tr
ated
th
at
o
u
r
s
u
g
g
ested
alg
o
r
ith
m
ex
tr
ac
ted
b
u
ild
in
g
v
ec
t
o
r
co
r
n
er
s
with
f
ewe
r
b
r
ea
k
p
o
in
ts
an
d
is
o
lated
p
o
in
ts
an
d
m
o
r
e
co
m
p
lete
an
d
r
e
g
u
lar
b
u
ild
in
g
v
ec
to
r
b
o
u
n
d
ar
ies
th
a
n
th
e
s
tate
-
of
-
th
e
-
a
r
t
co
r
n
e
r
d
etec
to
r
s
.
I
n
o
r
d
e
r
to
q
u
ick
ly
ex
tr
ac
t
b
u
ild
in
g
c
o
r
n
e
r
s
f
r
o
m
th
e
b
u
ild
in
g
ed
g
es,
th
i
s
s
tu
d
y
s
u
g
g
ested
a
b
u
ild
in
g
s
h
ap
e
v
ec
to
r
izatio
n
h
ier
ar
ch
y
th
at
in
teg
r
ated
D
C
NNs
-
b
ased
b
u
ild
in
g
ed
g
e
d
et
ec
tio
n
with
a
PC
A
-
b
ased
co
r
n
er
ex
tr
ac
tio
n
ap
p
r
o
ac
h
[
3
8
]
.
2
.
1
6
.
Act
iv
e
lea
rning
a
nd
f
e
dera
t
ed
lea
rning
Usi
n
g
f
ed
er
ate
d
lear
n
in
g
(
FL
)
,
th
is
m
et
h
o
d
elim
in
ates
th
e
n
ee
d
f
o
r
m
a
n
u
al
in
s
p
ec
tio
n
o
f
ev
er
y
tr
ain
in
g
s
am
p
le
b
y
au
t
o
m
atica
lly
s
elec
tin
g
an
d
lab
elin
g
th
e
d
ata
f
r
o
m
wh
ic
h
it
lear
n
s
.
I
t
s
ee
m
s
a
m
o
r
e
d
if
f
icu
lt
task
f
o
r
v
ar
i
o
u
s
r
ea
s
o
n
s
th
at
a
f
r
a
m
ewo
r
k
is
n
o
t
ex
ten
d
ed
to
a
n
o
n
lin
e
a
ctiv
e
lear
n
in
g
(
AL
)
b
y
in
v
esti
g
atin
g
h
o
w
u
n
lab
eled
d
ata
m
ay
b
e
u
s
ed
in
tr
ain
in
g
lo
ca
l
m
o
d
els
d
u
r
in
g
d
if
f
er
en
t
co
m
m
u
n
icatio
n
r
o
u
n
d
s
o
f
FL.
T
h
e
d
ec
o
d
er
p
o
r
tio
n
o
f
an
atten
tio
n
-
b
ased
C
NN
n
etwo
r
k
is
u
tili
ze
d
in
th
e
co
d
e
n
etwo
r
k
m
o
d
el
(
C
FNet)
,
wh
ich
is
b
ased
o
n
class
f
ea
tu
r
e
o
p
ti
m
izatio
n
an
d
th
e
class
f
ea
tu
r
e
(
C
F)
m
o
d
u
les
tech
n
iq
u
e
,
t
o
d
is
cr
im
in
ate
b
etwe
en
b
u
ild
in
g
an
d
n
o
n
-
b
u
ild
in
g
(
b
ac
k
g
r
o
u
n
d
)
ar
ea
s
.
I
t
d
is
tin
g
u
is
h
ed
b
etw
ee
n
s
ev
er
al
g
r
o
u
n
d
tex
tu
r
es th
at
wer
e
m
is
tak
en
ly
s
ee
n
as st
r
u
ctu
r
es
[
3
9
]
.
2
.
1
7
.
Do
ma
in a
da
pta
t
io
n
m
e
t
ho
ds
us
i
ng
G
NN
g
ra
ph
neu
ra
l net
wo
rk
s
Fo
r
th
e
p
u
r
p
o
s
e
o
f
class
if
y
in
g
R
S
im
ag
es,
tar
g
et
d
o
m
ain
s
w
er
e
u
s
ed
with
d
is
tr
ib
u
tio
n
s
th
a
t
d
if
f
er
e
d
f
r
o
m
th
e
s
o
u
r
ce
d
o
m
ain
.
C
r
ea
ted
a
tech
n
iq
u
e
f
o
r
id
e
n
tify
in
g
an
d
elim
in
atin
g
s
o
u
r
ce
s
am
p
les
th
at
ar
en
’
t
r
elev
an
t b
u
t
c
o
u
ld
h
elp
with
t
h
e
ad
ap
tatio
n
p
r
o
ce
s
s
.
Mix
ed
ta
r
g
et
d
o
m
ain
s
ettin
g
s
an
d
m
u
lti
s
o
u
r
ce
m
u
ltit
ar
g
et
ad
ap
tatio
n
wer
e
in
v
esti
g
ated
.
T
h
e
s
eg
m
e
n
tatio
n
o
f
u
r
b
an
b
u
ild
in
g
s
f
r
o
m
ae
r
ial
p
h
o
to
g
r
ap
h
s
an
d
en
h
a
n
ce
d
lear
n
in
g
ca
p
ac
ities
o
f
th
e
s
u
p
er
v
is
ed
lear
n
i
n
g
m
o
d
el
in
d
if
f
er
en
tiatin
g
b
u
ild
in
g
s
in
d
e
n
s
e
ar
ea
s
[
4
0
]
an
d
ex
tr
ac
tin
g
b
u
ild
in
g
s
co
v
er
ed
b
y
s
h
ad
o
ws
wer
e
ex
p
lo
r
ed
in
t
h
e
p
ar
allel
cr
o
s
s
-
lear
n
in
g
-
b
as
ed
p
ix
el
tr
a
n
s
f
er
r
ed
d
ec
o
n
v
o
lu
tio
n
al
n
etwo
r
k
(
PC
L
–
PTD
n
et)
tech
n
iq
u
e.
I
ts
ab
ilit
y
to
d
is
tin
g
u
is
h
s
p
ec
if
ic
b
o
r
d
er
s
an
d
b
u
ild
i
n
g
s
h
ap
es
is
l
im
ited
.
T
h
e
YO
L
Ov
3
o
b
ject
d
etec
tio
n
ar
ch
itectu
r
e
is
u
tili
ze
d
in
th
e
f
ea
tu
r
e
r
ewe
ig
h
tin
g
m
eth
o
d
o
l
o
g
y
to
ac
co
m
p
lis
h
f
e
w
-
s
h
o
t
lear
n
in
g
o
f
d
am
ag
e
as
s
ess
m
en
t
[
4
1
]
m
o
d
els
o
n
th
e
x
B
D
d
ataset.
W
h
en
co
m
p
ar
ed
to
th
e
b
aselin
e,
m
AP
im
p
r
o
v
e
d
with
th
e
f
ew
lab
eled
s
am
p
les.
T
h
e
class
-
s
p
ec
if
ic
r
ewe
ig
h
tin
g
v
ec
to
r
p
r
o
d
u
ce
d
b
y
th
e
r
ewe
i
g
h
tin
g
m
o
d
u
le
was
an
aly
ze
d
in
th
is
ca
s
e
u
s
in
g
t
-
SNE.
T
h
is
was
em
p
lo
y
ed
to
ass
es
s
s
im
ilar
itie
s
b
o
th
with
in
an
d
b
etwe
en
class
es
[
4
2
]
.
3.
SE
V
E
RA
L
I
S
SUE
S ID
E
N
T
I
F
I
E
D
Ma
n
y
r
esear
ch
er
s
h
av
e
f
o
u
n
d
th
at
u
s
in
g
th
e
af
o
r
em
e
n
tio
n
ed
m
eth
o
d
o
l
o
g
ies,
h
ig
h
-
r
eso
lu
tio
n
an
ten
n
a
im
ag
es
m
ay
b
e
o
b
tain
ed
wit
h
ex
ce
llen
t
ac
cu
r
ac
y
f
o
r
b
u
ild
in
g
m
ap
p
in
g
.
W
e
m
u
s
t
id
en
tify
th
ese
h
ig
h
-
r
eso
lu
tio
n
a
n
ten
n
a
p
h
o
to
s
.
R
e
g
ar
d
in
g
th
e
esti
m
atin
g
is
s
u
es,
it
is
a
ls
o
ev
i
d
en
t
th
at
id
en
tif
y
in
g
t
h
e
b
u
ild
in
g
’
s
co
r
n
er
s
r
e
q
u
ir
es
a
lo
t
o
f
la
b
o
u
r
a
n
d
is
o
f
ten
ca
lcu
lated
i
n
ac
cu
r
ately
.
T
h
o
u
g
h
th
e
o
u
tc
o
m
e
f
o
r
a
b
u
ild
in
g
ex
tr
ac
tio
n
m
ay
b
e
b
etter
,
m
o
r
e
co
n
f
ig
u
r
atio
n
is
n
ee
d
ed
in
o
r
d
er
f
o
r
th
e
m
o
d
el
to
u
n
d
er
s
ta
n
d
th
ese
s
tr
u
ctu
r
e
s
,
m
o
s
t
o
f
wh
ich
a
r
e
r
ec
ta
n
g
u
lar
in
s
h
ap
e
.
B
ased
o
n
b
u
ild
in
g
s
,
it
ap
p
ea
r
s
th
at
t
h
is
was
th
e
m
o
s
t
d
if
f
icu
lt
asp
ec
t
o
f
im
ag
e
s
eg
m
en
tatio
n
b
ec
a
u
s
e
it
was
d
if
f
icu
lt
to
esti
m
ate
th
e
co
r
n
er
s
o
f
r
ec
ta
n
g
u
la
r
b
u
ild
i
n
g
s
.
Fu
tu
r
e
r
esear
ch
in
d
icate
s
th
at
cr
ea
tin
g
co
r
n
e
r
s
in
r
ec
tan
g
u
lar
s
tr
u
ct
u
r
es is
n
ec
ess
ar
y
[
4
3
]
.
B
u
ild
in
g
f
o
o
tp
r
in
t
ex
tr
ac
tio
n
in
co
m
p
licated
m
etr
o
p
o
litan
e
n
v
ir
o
n
m
en
ts
was
n
o
t
ac
cu
r
ate
ev
en
with
th
e
u
s
e
o
f
s
tr
o
n
g
p
o
s
t
-
p
r
o
ce
s
s
in
g
tech
n
iq
u
es
an
d
th
e
cr
ea
t
io
n
o
f
m
o
r
e
r
o
b
u
s
t
d
atasets
.
A
p
o
s
t
-
p
r
o
ce
s
s
in
g
modul
e
o
r
an
au
to
m
atic
clea
n
in
g
s
o
lu
tio
n
f
o
r
th
e
tr
ain
in
g
d
ata
m
u
s
t
b
e
f
o
u
n
d
in
o
r
d
e
r
to
au
to
m
atica
lly
elim
in
ate
tiles
th
at
h
av
e
a
m
is
m
atch
b
etwe
en
t
h
e
g
r
o
u
n
d
tr
u
t
h
m
ask
an
d
th
e
R
S
p
ictu
r
es,
h
en
ce
im
p
r
o
v
in
g
th
e
p
ip
elin
e
an
d
p
r
o
d
u
ci
n
g
m
o
r
e
ac
cu
r
ate
b
u
ild
in
g
f
o
o
t
p
r
in
ts
.
DL
m
o
d
els
th
at
ar
e
m
o
r
e
r
esis
tan
t
to
n
o
is
e
m
u
s
t
b
e
cr
ea
ted
in
o
r
d
er
to
r
ed
u
ce
n
o
is
e.
No
r
m
al
d
ata
is
in
s
u
f
f
icien
t
to
im
p
r
o
v
e
th
e
d
etec
tio
n
o
f
b
u
ild
in
g
s
h
ap
es;
s
o
,
it
is
n
ec
ess
ar
y
to
d
eter
m
in
e
h
o
w
to
u
s
e
L
I
DAR
d
ata
in
co
n
ju
n
ctio
n
with
R
S
R
GB
p
h
o
to
s
.
Ad
d
itio
n
a
lly
,
esti
m
atio
n
s
o
f
co
n
s
tr
u
ctio
n
h
eig
h
ts
at
th
e
s
am
e
tim
e.
Nee
d
s
to
in
cr
ea
s
e
p
r
ec
is
io
n
an
d
p
in
p
o
in
t
t
h
e
r
eg
u
lar
izatio
n
o
f
b
u
ild
in
g
ex
tr
ac
tio
n
[
4
4
]
.
I
t
is
ch
allen
g
in
g
to
d
is
tin
g
u
is
h
b
etwe
en
th
e
m
is
class
if
icatio
n
o
f
b
u
ild
in
g
p
ix
els
in
s
p
atial
r
eso
lu
tio
n
phot
o
s
an
d
th
e
s
eg
r
e
g
atio
n
o
f
b
u
ild
in
g
b
o
u
n
d
ar
ies
in
th
e
ca
s
e
o
f
s
m
all
b
u
ild
in
g
s
th
at
ar
e
lo
ca
ted
clo
s
ely
to
g
eth
er
.
As
a
r
esu
lt,
it
is
n
e
ce
s
s
ar
y
to
d
eter
m
in
e
th
e
q
u
a
n
tity
o
f
s
am
p
le
d
ata
th
at
will
b
e
u
s
ed
to
q
u
a
n
tify
d
if
f
er
en
ce
s
.
I
n
o
r
d
er
t
o
co
n
tr
ib
u
te
to
th
e
m
o
d
el
’
s
tr
ain
in
g
an
d
en
h
an
ce
th
e
ac
c
u
r
ac
y
o
f
s
em
an
tic
s
eg
m
en
tatio
n
,
s
am
p
les
m
u
s
t
b
e
ab
le
to
d
is
tin
g
u
is
h
h
ig
h
-
c
o
n
f
id
en
ce
s
am
p
les
with
m
o
r
e
u
s
ef
u
l
in
f
o
r
m
atio
n
wh
ile
m
ain
tain
in
g
th
e
ac
cu
r
ac
y
o
f
th
e
u
n
lab
eled
d
ata.
T
h
e
d
ataset
n
ee
d
ed
to
en
h
a
n
ce
th
e
alg
o
r
ith
m
’
s
d
etec
tio
n
p
er
f
o
r
m
a
n
ce
an
d
p
r
o
v
id
e
a
m
o
r
e
ac
cu
r
ate
la
n
d
s
lid
e
d
ataset
f
o
r
n
etwo
r
k
tr
ain
in
g
is
n
o
t p
r
esen
t
[
4
5
]
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
7
7
6
I
n
t J I
n
f
&
C
o
m
m
u
n
T
ec
h
n
o
l
,
Vo
l.
1
4
,
No
.
2
,
A
u
g
u
s
t
20
2
5
:
614
-
624
620
I
t
is
n
ec
ess
ar
y
to
in
v
esti
g
ate
lar
g
e
an
d
s
m
all
p
ictu
r
e
s
izes
wh
en
an
aly
s
in
g
an
y
tr
a
d
e
-
o
f
f
s
b
etwe
en
m
o
d
el
p
er
f
o
r
m
a
n
ce
an
d
m
em
o
r
y
u
tili
za
tio
n
with
v
ar
y
in
g
s
u
p
p
o
r
t
im
ag
e
s
izes.
I
n
o
r
d
er
t
o
p
r
ev
e
n
t
is
s
u
es
lik
e
in
co
r
r
ec
tly
class
if
y
in
g
co
n
cr
et
e
f
lo
o
r
s
,
m
is
id
en
tify
in
g
n
ea
r
b
y
b
u
ild
in
g
s
,
m
is
tak
in
g
s
h
ad
o
w
s
f
o
r
b
u
ild
in
g
s
,
an
d
f
ailin
g
to
r
ec
o
g
n
ize
b
u
ild
in
g
s
co
v
er
ed
in
v
eg
etatio
n
,
it
is
n
e
ce
s
s
ar
y
to
in
c
o
r
p
o
r
ate
m
o
r
e
a
n
d
h
ig
h
er
-
r
eso
lu
tio
n
R
S
im
ag
er
y
,
in
v
esti
g
ate
p
o
s
t
-
p
r
o
ce
s
s
in
g
tech
n
iq
u
es
to
c
lar
if
y
b
u
ild
in
g
r
elatio
n
s
h
ip
s
,
an
d
i
n
cr
ea
s
e
th
e
p
r
ec
is
io
n
o
f
ex
tr
ac
tin
g
s
p
ar
s
ely
p
o
p
u
lated
b
u
ild
in
g
s
.
Fu
r
th
er
m
o
r
e,
th
er
e
e
x
is
ts
an
o
p
p
o
r
tu
n
ity
to
en
h
a
n
ce
ac
cu
r
ac
y
ab
o
u
t th
e
t
r
ain
in
g
a
n
d
v
alid
atio
n
d
atasets
[
4
6
]
.
I
t
i
s
n
e
c
e
s
s
a
r
y
t
o
s
p
e
e
d
u
p
t
h
e
D
A
B
E
-
N
e
t
a
p
p
r
o
a
c
h
a
n
d
t
a
c
k
l
e
m
o
r
e
d
i
f
f
i
c
u
l
t
R
S
i
m
a
g
e
s
e
g
m
e
n
t
a
t
i
o
n
j
o
b
s
i
n
o
r
d
e
r
t
o
p
e
r
f
o
r
m
m
u
l
t
i
c
l
a
s
s
s
e
m
a
n
t
i
c
s
e
g
m
e
n
t
a
t
i
o
n
o
f
m
e
t
r
o
p
o
l
i
t
a
n
e
n
v
i
r
o
n
m
e
n
t
s
.
I
n
o
r
d
e
r
t
o
g
u
a
r
a
n
t
e
e
p
r
e
c
i
s
e
bu
i
l
d
i
n
g
e
x
t
r
a
c
t
i
o
n
a
n
d
b
u
i
l
d
i
n
g
c
l
a
s
s
i
f
i
c
a
t
i
o
n
,
h
e
i
g
h
t
d
a
t
a
a
n
d
V
H
R
o
r
t
h
o
p
h
o
t
o
p
i
c
t
u
r
e
s
c
a
n
b
e
u
t
i
l
i
z
e
d
[
4
7
]
.
U
s
i
n
g
t
h
e
s
e
m
e
t
h
o
d
s
,
h
i
g
h
-
r
e
s
o
l
u
t
i
o
n
3
-
D
d
a
t
a
,
s
u
c
h
a
s
d
i
g
i
t
a
l
s
u
r
f
a
c
e
m
o
d
e
l
s
o
r
a
e
r
i
a
l
3
-
D
p
o
i
n
t
c
l
o
u
d
s
,
m
u
s
t
b
e
o
b
t
a
i
n
e
d
i
n
o
r
d
e
r
t
o
s
t
r
e
n
g
t
h
e
n
t
h
e
c
o
n
n
e
c
t
i
o
n
s
b
e
t
w
e
e
n
t
h
e
c
l
a
s
s
e
s
t
h
a
t
a
r
e
f
o
u
n
d
.
H
o
w
e
v
e
r
,
w
i
t
h
t
h
e
s
e
m
e
t
h
o
d
o
l
o
g
i
e
s
,
d
i
s
t
i
n
c
t
i
v
e
b
u
i
l
d
i
n
g
s
,
o
r
d
i
n
a
r
y
b
u
i
l
d
i
n
g
s
,
a
n
d
c
o
m
p
l
i
c
a
t
e
d
b
a
c
k
g
r
o
u
n
d
s
a
l
l
h
a
d
a
n
i
m
p
a
c
t
o
n
t
h
e
e
x
t
r
a
c
t
i
o
n
a
c
c
u
r
a
c
y
[
4
8
]
.
M
o
r
e
o
p
t
i
m
i
z
e
d
d
e
e
p
n
e
u
r
a
l
n
e
t
w
o
r
k
s
a
n
d
p
r
e
-
o
r
p
o
s
t
-
p
r
o
c
e
s
s
i
n
g
t
e
c
h
n
i
q
u
e
s
a
r
e
r
e
q
u
i
r
e
d
t
o
e
l
i
m
i
n
a
t
e
e
r
r
o
r
s
a
n
d
o
m
i
s
s
i
o
n
s
.
R
a
t
h
e
r
t
h
a
n
c
o
n
s
i
d
e
r
i
n
g
t
h
e
n
e
t
w
o
r
k
a
n
d
p
r
e
/
p
o
s
t
-
p
r
o
c
e
s
s
i
n
g
t
e
c
h
n
i
q
u
es
s
e
p
a
r
a
t
e
l
y
,
b
o
t
h
a
r
e
n
o
t
t
a
k
e
n
i
n
t
o
c
o
n
s
i
d
e
r
a
t
i
o
n
c
o
n
c
u
r
r
e
n
t
l
y
f
o
r
g
r
e
a
t
e
r
a
c
c
u
r
a
c
y
a
n
d
e
f
f
i
c
i
e
n
c
y
[
4
9
]
.
I
n
c
l
o
u
d
v
a
r
i
o
u
s
n
e
u
r
a
l
n
e
t
w
o
r
k
p
r
o
c
e
s
s
c
a
n
b
e
e
x
e
c
u
t
e
d
f
o
r
g
e
t
t
i
n
g
m
o
r
e
p
e
r
f
o
r
m
a
n
c
e
a
n
d
c
a
n
m
o
n
i
t
o
r
d
i
s
a
s
t
e
r
o
n
v
a
r
i
o
u
s
c
o
m
p
u
t
a
t
i
o
n
s
a
n
d
r
e
c
o
v
e
r
v
a
r
i
o
u
s
t
a
s
k
s
i
n
c
a
s
e
o
f
d
i
s
a
s
t
e
r
[
5
0
]
.
I
n
o
r
d
er
to
p
in
p
o
in
t
ex
ac
t
c
h
a
n
g
es
in
lan
d
-
u
s
e
an
d
lan
d
-
co
v
er
(
o
r
c
r
o
p
p
in
g
p
atter
n
)
,
it
is
n
ec
ess
ar
y
to
in
v
esti
g
ate
s
u
itab
le
class
if
icatio
n
an
d
alg
o
r
ith
m
s
t
h
at
d
i
f
f
er
en
tiate
b
etwe
en
v
ar
i
o
u
s
s
o
r
ts
o
f
c
h
an
g
es.
T
h
e
au
to
en
co
d
er
-
en
h
an
ce
d
d
is
tan
ce
m
ea
s
u
r
em
en
ts
p
er
f
o
r
m
ed
less
wel
l
wh
en
u
s
ed
in
a
d
ee
p
m
etr
ic
lear
n
in
g
f
r
am
ewo
r
k
(
s
u
ch
as
th
e
tr
ip
let
n
etwo
r
k
o
r
s
iam
ese
n
etwo
r
k
)
.
T
h
e
r
e
is
s
till
r
o
o
m
to
ex
p
lo
r
e
v
a
r
io
u
s
au
to
en
co
d
er
a
r
ch
itectu
r
es,
s
u
ch
as
c
o
n
v
o
lu
tio
n
al
au
to
en
c
o
d
er
s
,
f
o
r
th
e
r
ec
o
n
s
tr
u
ctio
n
o
f
s
p
ec
tr
al
d
ata.
T
h
ese
s
tr
ateg
ies
h
av
e
th
e
p
o
ten
tial
t
o
im
p
r
o
v
e
alg
o
r
ith
m
p
er
f
o
r
m
an
ce
as
m
o
r
e
d
ata
is
g
ath
e
r
ed
,
p
ar
ticu
lar
ly
in
th
e
ca
teg
o
r
izatio
n
o
f
less
d
am
ag
e
d
s
tr
u
ctu
r
es,
d
estro
y
ed
b
u
ild
i
n
g
s
,
an
d
co
r
r
ec
tly
ca
teg
o
r
ize
d
n
o
r
m
al
b
u
ild
i
n
g
s
.
T
h
ese
m
eth
o
d
s
s
h
o
u
ld
c
o
n
ce
n
tr
ate
o
n
u
s
in
g
a
C
NN
m
o
d
el
f
o
r
th
e
m
u
lticlas
s
class
if
i
ca
tio
n
p
r
o
b
lem
in
co
n
ju
n
ctio
n
with
an
ef
f
icien
t
f
ea
tu
r
e
ex
tr
ac
ti
o
n
a
n
d
s
elec
tio
n
p
r
o
ce
d
u
r
e.
I
t
is
n
ec
ess
ar
y
to
id
en
tify
v
ar
io
u
s
p
r
o
b
lem
s
,
s
u
c
h
as
r
o
a
d
d
a
m
a
g
e,
in
o
r
d
er
to
im
m
e
d
iately
r
e
d
ir
ec
t
th
e
q
u
ick
est
co
u
r
s
e
b
et
wee
n
an
em
e
r
g
en
c
y
ca
r
e
f
ac
ilit
y
an
d
a
n
aid
team
[
5
1
]
.
4.
ASSE
SS
I
NG
B
U
I
L
D
I
NG
F
E
AT
URE
E
X
T
RAC
T
I
O
N
T
E
CH
NIQU
E
S
I
n
th
e
af
o
r
em
en
tio
n
ed
m
eth
o
d
s
,
s
atellite
d
ata
ac
q
u
is
itio
n
is
u
s
ed
to
cr
ea
te
o
b
ject
-
b
ased
m
ap
s
th
at
a
r
e
eith
er
n
o
n
-
ex
is
ten
t
o
r
in
n
ee
d
o
f
u
p
d
atin
g
;
g
r
o
u
n
d
s
u
r
v
ey
in
g
o
r
R
S
s
u
r
v
e
y
in
g
is
th
e
p
r
ef
er
r
ed
p
r
im
ar
y
d
ata
ac
q
u
is
itio
n
m
eth
o
d
f
o
r
o
b
tain
in
g
n
ew
d
ata.
As
a
r
esu
lt,
th
e
p
r
im
ar
y
s
o
u
r
ce
o
f
d
ata
h
as
to
b
e
h
ig
h
-
q
u
ality
s
atellite
p
h
o
to
g
r
ap
h
y
with
a
r
e
s
o
lu
tio
n
o
f
0
.
6
m
eter
s
.
Ad
d
itio
n
al
d
ata
f
o
r
th
e
p
r
o
ject
ca
m
e
f
r
o
m
f
ield
wo
r
k
an
d
th
e
city
’
s
cu
r
r
e
n
t stre
et
g
u
id
e
[
5
2
]
.
P
re
-
p
r
o
ce
s
s
in
g
o
f
th
e
h
ig
h
-
r
e
s
o
lu
tio
n
s
atellite
d
ata
in
th
e
af
o
r
em
en
tio
n
ed
tech
n
iq
u
es:
a
t
th
is
p
o
in
t,
th
e
im
ag
e
u
n
d
er
wen
t
p
r
e
-
p
r
o
c
ess
in
g
,
an
d
all
g
eo
m
etr
ic
a
n
d
r
ad
io
m
etr
ic
co
r
r
ec
tio
n
s
wer
e
ap
p
lied
,
en
h
an
cin
g
th
e
im
ag
e
’
s
ab
ilit
y
to
b
e
s
ee
n
ex
tr
em
ely
clea
r
ly
o
n
th
e
g
r
o
u
n
d
.
Gath
e
r
in
g
d
ata:
i
t
r
ef
e
r
r
e
d
to
th
e
p
r
o
ce
s
s
o
f
d
ig
itizatio
n
,
wh
ich
co
n
v
er
ts
f
r
esh
d
ata
n
ee
d
ed
to
cr
ea
te
a
m
ap
in
to
a
d
ig
ital
f
o
r
m
at
f
o
r
ar
ch
iv
in
g
,
c
o
m
p
ar
i
n
g
with
d
ata
attr
ib
u
tes,
an
d
o
th
er
p
r
o
ce
s
s
in
g
th
at
is
r
eq
u
ir
ed
f
o
r
r
esear
ch
p
r
ep
ar
atio
n
.
T
h
e
d
a
taset,
wh
ich
is
th
e
r
esu
lt
o
f
ML
an
d
DL
tech
n
iq
u
es
u
s
ed
t
o
ex
t
r
ac
t
f
ea
tu
r
e
s
f
r
o
m
h
ig
h
-
r
eso
lu
tio
n
s
atellite
im
ag
es,
is
n
o
t
v
ec
to
r
ized
th
r
o
u
g
h
th
e
o
n
-
s
cr
ee
n
d
ig
itizin
g
p
r
o
ce
s
s
in
o
r
d
er
to
ex
tr
ac
t
c
h
ar
ac
ter
is
tics
f
r
o
m
th
e
im
ag
e
th
at
s
h
o
u
ld
b
e
in
clu
d
e
d
b
ased
o
n
th
e
in
ter
p
r
etatio
n
k
ey
an
d
au
t
o
m
atic
ex
tr
ac
tio
n
p
r
o
ce
d
u
r
e
u
s
in
g
o
b
ject
-
b
ased
ap
p
r
o
ac
h
th
at
is
lim
ited
f
o
r
id
en
tific
atio
n
o
f
f
ea
tu
r
es in
f
o
r
m
atio
n
th
em
es,
lay
er
s
cr
ea
ted
[
5
3
]
.
T
h
e
m
ain
g
o
al
o
f
th
e
r
esear
ch
p
r
o
ject
is
to
u
s
e
o
b
ject
-
b
ased
class
if
icati
o
n
to
class
if
y
all
o
f
th
e
p
ix
els
in
a
h
ig
h
-
r
eso
lu
tio
n
im
ag
e
i
n
to
g
r
o
u
p
s
o
r
th
em
es.
T
h
e
s
p
atial
o
b
ject
-
b
ased
ap
p
r
o
ac
h
is
n
o
t
u
s
ed
in
th
e
af
o
r
em
en
tio
n
ed
s
tr
ateg
ies.
I
ts
f
o
u
n
d
atio
n
is
th
e
class
if
icatio
n
o
f
p
ictu
r
e
p
ix
els
d
e
p
en
d
i
n
g
o
n
h
o
w
t
h
o
s
e
p
i
x
els
r
elate
t
o
th
e
p
ix
els
ar
o
u
n
d
th
em
.
No
t
o
n
ly
ar
e
asp
ec
ts
lik
e
p
ictu
r
e
tex
tu
r
e,
p
r
o
x
im
ity
,
f
ea
tu
r
e
s
ize,
s
h
ap
e
o
r
ien
tatio
n
,
r
e
p
etitio
n
,
an
d
c
o
n
tex
t
ig
n
o
r
ed
,
b
u
t
th
ey
ar
e
also
n
o
t
em
p
lo
y
ed
in
th
e
p
r
o
ce
s
s
o
f
v
is
u
al
in
ter
p
r
etatio
n
.
Fo
r
th
e
p
u
r
p
o
s
e
o
f
p
r
o
d
u
ci
n
g
an
in
ter
p
r
etati
o
n
k
ey
f
o
r
ev
er
y
in
f
o
r
m
ativ
e
class
,
th
e
ca
teg
o
r
ies
ar
e
n
o
t c
h
a
r
ac
ter
ized
th
r
o
u
g
h
o
u
t th
e
im
ag
e
[
5
4
]
.
Un
s
u
p
er
v
is
ed
lear
n
in
g
m
u
s
t
b
e
s
tu
d
ied
in
o
r
d
er
to
im
p
r
o
v
e
ex
tr
ac
tio
n
r
esu
lts
in
u
n
lab
eled
d
atasets
u
s
in
g
a
v
ar
iety
o
f
ap
p
r
o
ac
h
es.
B
y
in
cr
ea
s
in
g
th
e
d
i
v
er
s
ity
o
f
tr
ain
in
g
s
am
p
les
an
d
d
ev
elo
p
in
g
m
o
r
e
ef
f
icien
t
n
etwo
r
k
to
p
o
lo
g
ies
to
s
y
n
t
h
e
s
ize
tex
tu
r
al
p
r
o
p
e
r
ties
,
y
o
u
ca
n
en
h
an
c
e
th
e
ex
tr
ac
tio
n
p
er
f
o
r
m
a
n
ce
o
f
s
m
all
b
u
ild
in
g
i
n
s
tan
ce
s
.
T
ak
en
o
u
ch
i
an
d
C
h
o
h
[
5
5
]
n
ee
d
s
to
u
s
e
t
h
e
DR
R
NN
ap
p
r
o
ac
h
to
id
e
n
t
if
y
in
d
u
s
tr
ial
lan
d
,
u
r
b
an
r
esid
en
tial
ar
ea
s
,
an
d
r
u
r
al
r
esid
en
tial
r
e
g
io
n
s
th
at
a
r
e
tar
g
eted
f
o
r
d
is
aster
r
esil
ien
ce
o
r
v
u
ln
e
r
ab
ilit
y
d
u
e
to
h
ig
h
lev
els
o
f
r
eg
io
n
al
r
is
k
.
W
h
ile
cr
ea
tin
g
a
p
p
s
f
o
r
th
e
d
e
v
elo
p
m
e
n
t
o
f
m
a
p
s
u
s
ed
in
d
is
aster
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J I
n
f
&
C
o
m
m
u
n
T
ec
h
n
o
l
I
SS
N:
2252
-
8
7
7
6
Tech
n
iq
u
es o
f d
ee
p
lea
r
n
in
g
n
eu
r
a
l n
etw
o
r
k
b
a
s
ed
b
u
ild
i
n
g
f
ea
tu
r
e
ex
tr
a
ctio
n
fr
o
m
…
(
S
h
r
in
iva
s
K
h
a
n
d
a
r
e
)
621
p
r
ev
en
tio
n
,
s
o
m
e
m
a
n
u
al
o
p
er
atio
n
s
m
u
s
t
b
e
au
to
m
ate
d
(
e.
g
.
,
co
m
p
u
tatio
n
s
o
f
I
n
t.V
a
n
d
d
is
p
lay
o
f
r
o
a
d
n
etwo
r
k
co
n
d
itio
n
s
o
n
th
e
o
n
li
n
e
m
ap
)
[
5
6
]
.
I
n
ce
r
tain
m
eth
o
d
s
,
it
is
n
ec
e
s
s
ar
y
to
cr
ea
te
L
M
n
etwo
r
k
s
with
b
etter
g
en
er
aliza
tio
n
ca
p
ab
ilit
ies,
wh
ich
ca
n
b
e
d
o
n
e
b
y
f
u
s
i
n
g
t
r
an
s
f
er
lear
n
in
g
o
r
p
o
o
r
l
y
s
u
p
er
v
is
ed
lear
n
i
n
g
with
th
e
cu
r
r
en
t
g
lo
b
al
s
h
o
r
tag
e
o
f
lan
d
s
lid
e
R
S
d
ata
[
5
7
]
.
T
o
co
n
f
ir
m
th
e
ef
f
icac
y
o
f
th
is
ap
p
r
o
ac
h
,
f
u
r
th
er
d
atasets
f
ea
tu
r
in
g
v
ar
io
u
s
d
is
aster
s
itu
atio
n
s
(
s
u
ch
as
ea
r
th
q
u
a
k
e
s
an
d
ts
u
n
am
is
)
an
d
d
ata
s
o
u
r
ce
s
(
s
u
ch
as
ae
r
ial
an
d
UAV
d
ata)
m
u
s
t
b
e
f
o
u
n
d
[
5
8
]
.
I
n
cr
ea
s
ed
co
m
p
u
tatio
n
al
ef
f
icien
cy
is
r
eq
u
ir
e
d
wh
en
d
ea
lin
g
with
b
ig
v
o
lu
m
es
o
f
u
n
lab
eled
s
am
p
les
in
o
r
d
er
t
o
m
ee
t
em
e
r
g
en
c
y
r
e
q
u
ir
em
en
ts
d
u
r
in
g
ca
lam
ities
.
M
ap
o
f
ca
tast
r
o
p
h
e
i
n
ten
s
ity
m
u
s
t
b
e
lo
ca
ted
u
s
in
g
SVM.
Nee
d
s
to
im
p
r
o
v
e
th
e
p
u
r
p
o
s
ef
u
l
is
lan
d
in
g
alg
o
r
it
h
m
b
ased
o
n
ML
t
o
o
p
er
ate
m
o
r
e
q
u
ick
l
y
an
d
m
a
k
e
r
eliab
le
d
ec
is
io
n
s
.
Nee
d
s
to
lo
ca
te
d
is
aster
-
r
elate
d
s
atellite
i
m
ag
er
y
[
5
9
]
.
T
h
e
id
ea
l
s
ize
f
o
r
s
u
p
p
o
r
t
im
a
g
es
h
as
n
o
t b
ee
n
estab
lis
h
ed
i
n
th
e
m
et
h
o
d
o
lo
g
ies
u
n
d
er
s
tu
d
y
,
n
o
r
h
as
it
b
ee
n
p
o
s
s
ib
le
to
an
al
y
s
e
an
y
tr
ad
e
-
o
f
f
s
b
etwe
en
m
e
m
o
r
y
u
s
e
an
d
m
o
d
el
p
er
f
o
r
m
a
n
ce
w
ith
v
ar
y
in
g
s
u
p
p
o
r
t
im
ag
e
s
izes.
Dif
f
er
en
tiatin
g
b
etwe
en
s
u
p
p
o
r
t
p
ictu
r
e
s
ize
s
is
n
ec
ess
ar
y
.
I
n
ad
d
itio
n
to
p
r
o
v
id
in
g
v
alu
ab
le
in
s
ig
h
ts
in
to
wh
eth
er
lar
g
er
o
r
s
m
aller
s
u
p
p
o
r
t
im
ag
es
lead
to
b
etter
class
s
ep
ar
atio
n
,
class
im
b
alan
ce
in
th
e
tr
ain
in
g
s
et
q
u
er
y
im
ag
es,
a
n
d
im
p
r
o
v
ed
m
o
d
el
p
e
r
f
o
r
m
an
ce
,
t
-
SNE
an
aly
s
is
o
f
th
e
r
esu
ltin
g
r
ewe
ig
h
tin
g
v
ec
to
r
s
m
ay
also
ex
te
n
d
th
e
f
r
am
ewo
r
k
to
an
o
n
lin
e
A
L
b
y
in
v
esti
g
atin
g
h
o
w
u
n
la
b
elled
d
ata
ca
n
b
e
in
co
r
p
o
r
ated
in
tr
ain
in
g
lo
ca
l
m
o
d
els
d
u
r
in
g
d
if
f
er
en
t
co
m
m
u
n
icatio
n
r
o
u
n
d
s
o
f
FL,
wh
ich
ap
p
ea
r
s
to
b
e
a
m
o
r
e
d
if
f
icu
lt task
f
o
r
a
n
u
m
b
er
o
f
r
ea
s
o
n
s
[
6
0
]
.
T
h
er
e
ar
e
r
estrictio
n
s
o
n
th
e
b
u
ild
in
g
’
s
ex
ac
t
b
o
r
d
er
a
n
d
s
h
ap
e
s
eg
m
en
tatio
n
in
th
e
af
o
r
em
en
tio
n
ed
p
r
o
ce
d
u
r
es.
I
n
o
r
d
er
to
p
r
e
v
en
t
in
co
r
r
ec
t
r
ec
o
g
n
itio
n
m
is
tak
es
wh
en
s
o
m
e
g
r
o
u
n
d
tex
tu
r
es
r
esem
b
le
b
u
ild
in
g
s
,
a
DSM
an
d
s
o
m
e
p
o
s
tp
r
o
ce
s
s
in
g
tech
n
iq
u
es
n
ee
d
to
b
e
in
teg
r
ate
d
[
6
1
]
.
W
o
r
k
o
n
th
e
tex
t
u
r
in
g
elem
en
ts
o
f
ico
n
ic
s
tr
u
ctu
r
es
is
r
eq
u
ir
ed
to
f
in
d
a
s
o
lu
ti
o
n
to
th
is
is
s
u
e.
I
d
en
tific
ati
o
n
a
n
d
r
em
o
v
al
o
f
u
n
n
ec
ess
ar
y
s
o
u
r
ce
s
am
p
les
th
at
ca
n
aid
in
th
e
ad
a
p
tio
n
p
r
o
ce
s
s
is
n
ec
ess
ar
y
[
6
2
]
.
I
d
e
n
tific
atio
n
o
f
r
eq
u
ir
em
e
n
ts
f
o
r
m
ix
e
d
tar
g
et
d
o
m
ain
s
ettin
g
s
an
d
m
u
ltis
o
u
r
c
e,
m
u
ltit
ar
g
et
ad
a
p
tab
il
ity
.
Un
s
u
p
er
v
is
ed
lear
n
in
g
m
u
s
t
b
e
id
en
tifie
d
in
o
r
d
er
t
o
im
p
r
o
v
e
ex
tr
ac
tio
n
p
er
f
o
r
m
an
ce
in
u
n
lab
eled
d
at
asets
m
u
s
t
u
s
e
th
e
o
v
er
s
am
p
lin
g
an
d
a
u
g
m
e
n
tatio
n
tech
n
i
q
u
e
to
i
d
en
tify
s
m
all
item
s
in
b
u
ild
in
g
d
etec
tio
n
[
6
3
]
.
5.
CO
NCLU
SI
O
N
No
w
a
d
ay
s
,
in
d
is
aster
s
ce
n
ar
io
s
,
wh
ich
is
a
b
ig
h
az
a
r
d
,
wh
en
it
o
cc
u
r
s
,
it
cr
ea
tes
lo
s
s
o
f
b
u
ild
in
g
s
,
h
u
m
an
b
ein
g
s
,
a
n
im
als,
an
d
n
atu
r
al
r
eso
u
r
ce
s
.
T
o
o
v
e
r
co
m
e
th
is
p
r
o
b
lem
,
tec
h
n
iq
u
es
s
h
o
u
ld
b
e
d
e
v
elo
p
e
d
h
av
in
g
a
s
y
s
tem
atic
p
la
n
in
e
x
is
ten
ce
f
o
r
im
m
ed
ia
te
h
elp
f
o
r
id
e
n
tific
atio
n
o
f
c
o
r
r
ec
t
b
u
ild
in
g
s
tr
u
ctu
r
es
in
ca
s
e
o
f
d
is
aster
.
Mig
h
t
n
o
t
k
n
o
w,
wh
at
th
e
s
tr
u
ctu
r
e
o
f
t
h
e
b
u
ild
in
g
s
e
x
is
ted
b
ef
o
r
e
d
is
aster
,
h
o
w
m
u
c
h
a
r
e
th
e
d
am
ag
es
an
d
wh
at
ar
e
t
h
e
ca
teg
o
r
ies
o
f
r
eso
u
r
ce
s
wer
e
d
am
ag
ed
wer
e
d
am
ag
ed
.
Ho
w
ca
n
we
id
en
tify
th
em
?
No
w
a
d
ay
s
m
a
n
y
d
is
aster
s
ar
e
o
cc
u
r
r
i
n
g
,
th
er
ef
o
r
e
i
m
m
ed
iate
p
lan
is
n
o
t
r
ea
d
y
to
id
en
tify
wh
at
wer
e
th
e
b
u
ild
in
g
s
av
ailab
le
b
ef
o
r
e
d
is
aster
s
an
d
wh
at
r
em
ain
ed
af
ter
th
e
d
is
aster
h
ap
p
en
ed
.
T
h
is
is
an
ex
tr
em
ely
cr
itical
p
r
o
b
lem
.
T
o
o
v
e
r
co
m
e
th
e
ab
o
v
e
p
r
o
b
lem
a
s
y
s
tem
atic
f
r
am
ewo
r
k
s
h
o
u
l
d
b
e
d
ev
elo
p
ed
f
o
r
t
h
e
p
o
s
t
-
d
is
aster
r
esettlem
en
t
p
r
o
ce
s
s
.
I
n
th
is
s
u
r
v
e
y
o
f
b
u
ild
in
g
ex
tr
ac
tio
n
tec
h
n
iq
u
es,
a
tec
h
n
iq
u
e
n
ee
d
s
to
b
e
id
en
tifie
d
f
o
r
id
en
tif
y
in
g
s
tr
u
c
tu
r
es
ca
teg
o
r
ies
b
ef
o
r
e
d
is
aste
r
f
o
r
p
r
o
p
o
s
in
g
an
d
d
e
v
elo
p
in
g
a
f
r
am
ewo
r
k
f
o
r
p
o
s
t
-
d
is
aster
r
esettlem
en
t
u
s
i
n
g
s
atellite
im
ag
er
y
.
U
n
et
i
s
an
en
d
-
to
-
en
d
f
u
lly
co
n
v
o
lu
tio
n
al
n
etwo
r
k
;
th
er
ef
o
r
e,
it h
as n
o
t th
o
r
o
u
g
h
l
y
ex
am
in
e
d
all
th
e
d
ata
at
its
d
is
p
o
s
al,
an
d
it c
an
s
till
b
e
im
p
r
o
v
ed
.
ACK
NO
WL
E
DG
E
M
E
NT
S
T
h
e
a
u
t
h
o
r
ac
k
n
o
wled
g
es
t
h
e
Dir
ec
to
r
,
Ma
h
ar
ash
tr
a
R
S
Ap
p
licatio
n
C
en
tr
e.
th
e
d
ir
ec
to
r
p
er
m
itted
to
d
o
r
esear
ch
wo
r
k
f
o
r
ac
a
d
em
ic
p
u
r
p
o
s
es
an
d
g
u
id
an
c
e
o
f
th
e
d
ir
ec
to
r
an
d
s
cien
tis
ts
o
f
th
e
in
s
titu
tio
n
s
u
p
p
o
r
ted
f
o
r
m
o
ti
v
atio
n
to
d
o
r
esear
ch
wo
r
k
.
T
h
e
a
u
t
h
o
r
is
also
th
an
k
f
u
l
to
th
e
s
u
p
p
o
r
t
an
d
g
u
id
an
ce
d
eliv
er
ed
b
y
th
e
s
u
p
er
v
is
o
r
a
n
d
th
e
r
esear
ch
ce
n
tr
e
f
o
r
p
r
o
v
i
d
in
g
co
m
p
u
ter
s
,
s
o
f
twar
e,
in
te
r
n
et
ac
ce
s
s
,
jo
u
r
n
al
ac
ce
s
s
o
n
lin
e
an
d
p
h
y
s
ical
co
p
ies f
o
r
v
iewin
g
p
u
r
p
o
s
es.
F
UNDING
I
NF
O
R
M
A
T
I
O
N
Au
th
o
r
s
s
tate
n
o
f
u
n
d
in
g
in
v
o
lv
ed
.
AUTHO
R
CO
NT
RI
B
UT
I
O
NS ST
A
T
E
M
E
N
T
Na
m
e
o
f
Aut
ho
r
C
M
So
Va
Fo
I
R
D
O
E
Vi
Su
P
Fu
Sh
r
in
iv
as Kh
an
d
ar
e
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
M.
B
C
h
an
d
ak
✓
✓
✓
✓
✓
✓
✓
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
7
7
6
I
n
t J I
n
f
&
C
o
m
m
u
n
T
ec
h
n
o
l
,
Vo
l.
1
4
,
No
.
2
,
A
u
g
u
s
t
20
2
5
:
614
-
624
622
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
Au
th
o
r
s
s
tate
n
o
co
n
f
lict o
f
in
t
er
est.
DATA AV
AI
L
AB
I
L
I
T
Y
Data
av
ailab
ilit
y
is
n
o
t
ap
p
li
ca
b
le
to
th
is
p
ap
er
as
n
o
n
e
w
d
ata
wer
e
cr
ea
ted
o
r
an
aly
ze
d
in
th
is
s
tu
d
y
.
RE
F
E
R
E
NC
E
S
[
1
]
B
.
B
a
l
a
c
h
a
n
d
r
a
n
,
R
.
B
.
O
l
s
h
a
n
s
k
y
,
a
n
d
L
.
A
.
J
o
h
n
s
o
n
,
“
P
l
a
n
n
i
n
g
f
o
r
d
i
sas
t
e
r
-
i
n
d
u
c
e
d
r
e
l
o
c
a
t
i
o
n
o
f
c
o
m
m
u
n
i
t
i
e
s,
”
J
o
u
r
n
a
l
o
f
t
h
e
Am
e
r
i
c
a
n
Pl
a
n
n
i
n
g
Ass
o
c
i
a
t
i
o
n
,
v
o
l
.
8
8
,
n
o
.
3
,
p
p
.
2
8
8
–
3
0
4
,
Ju
l
.
2
0
2
2
,
d
o
i
:
1
0
.
1
0
8
0
/
0
1
9
4
4
3
6
3
.
2
0
2
1
.
1
9
7
8
8
5
5
.
[
2
]
H
.
Y
a
n
g
,
P
.
W
u
,
X
.
Y
a
o
,
Y
.
W
u
,
B
.
W
a
n
g
,
a
n
d
Y
.
X
u
,
“
B
u
i
l
d
i
n
g
e
x
t
r
a
c
t
i
o
n
i
n
v
e
r
y
h
i
g
h
-
r
e
so
l
u
t
i
o
n
i
m
a
g
e
r
y
b
y
d
e
n
s
e
-
a
t
t
e
n
t
i
o
n
n
e
t
w
o
r
k
s,
”
Re
m
o
t
e
S
e
n
si
n
g
,
v
o
l
.
1
0
,
n
o
.
1
1
,
p
.
1
7
6
8
,
N
o
v
.
2
0
1
8
,
d
o
i
:
1
0
.
3
3
9
0
/
r
s
1
0
1
1
1
7
6
8
.
[
3
]
Y
.
Zh
a
n
g
,
W
.
G
o
n
g
,
J
.
S
u
n
,
a
n
d
W
.
Li
,
“
W
e
b
-
N
e
t
:
a
n
o
v
e
l
n
e
st
n
e
t
w
o
r
k
s
w
i
t
h
u
l
t
r
a
-
h
i
e
r
a
r
c
h
i
c
a
l
s
a
m
p
l
i
n
g
f
o
r
b
u
i
l
d
i
n
g
e
x
t
r
a
c
t
i
o
n
f
r
o
m a
e
r
i
a
l
i
ma
g
e
r
i
e
s,
”
R
e
m
o
t
e
S
e
n
s
i
n
g
,
v
o
l
.
1
1
,
n
o
.
1
6
,
p
.
1
8
9
7
,
A
u
g
.
2
0
1
9
,
d
o
i
:
1
0
.
3
3
9
0
/
r
s1
1
1
6
1
8
9
7
.
[
4
]
G
.
Z
h
o
u
,
Y
.
W
a
n
g
,
T.
Y
u
e
,
S
.
Y
e
,
a
n
d
W
.
W
a
n
g
,
“
B
u
i
l
d
i
n
g
o
c
c
l
u
si
o
n
d
e
t
e
c
t
i
o
n
f
r
o
m
g
h
o
st
i
m
a
g
e
s
,
”
I
EE
E
T
ra
n
sa
c
t
i
o
n
s
o
n
G
e
o
sc
i
e
n
c
e
a
n
d
R
e
m
o
t
e
S
e
n
si
n
g
,
v
o
l
.
5
5
,
n
o
.
2
,
p
p
.
1
0
7
4
–
1
0
8
4
,
F
e
b
.
2
0
1
7
,
d
o
i
:
1
0
.
1
1
0
9
/
TG
R
S
.
2
0
1
6
.
2
6
1
9
1
8
4
.
[
5
]
H
.
Y
a
n
g
,
B
.
Y
u
,
J
.
L
u
o
,
a
n
d
F
.
C
h
e
n
,
“
S
e
m
a
n
t
i
c
s
e
g
m
e
n
t
a
t
i
o
n
o
f
h
i
g
h
s
p
a
t
i
a
l
r
e
so
l
u
t
i
o
n
i
ma
g
e
s
w
i
t
h
d
e
e
p
n
e
u
r
a
l
n
e
t
w
o
r
k
s,
”
G
I
S
c
i
e
n
c
e
a
n
d
Re
m
o
t
e
S
e
n
si
n
g
,
v
o
l
.
5
6
,
n
o
.
5
,
p
p
.
7
4
9
–
7
6
8
,
J
u
l
.
2
0
1
9
,
d
o
i
:
1
0
.
1
0
8
0
/
1
5
4
8
1
6
0
3
.
2
0
1
8
.
1
5
6
4
4
9
9
.
[
6
]
S
.
B
a
n
e
r
j
e
e
,
Y
.
S
.
P
a
t
e
l
,
P
.
K
u
mar,
a
n
d
M
.
B
h
u
y
a
n
,
“
T
o
w
a
r
d
s
p
o
st
-
d
i
s
a
st
e
r
d
a
m
a
g
e
a
ssess
me
n
t
u
si
n
g
d
e
e
p
t
r
a
n
sf
e
r
l
e
a
r
n
i
n
g
a
n
d
GAN
-
b
a
se
d
d
a
t
a
a
u
g
me
n
t
a
t
i
o
n
,
”
i
n
P
ro
c
e
e
d
i
n
g
s
o
f
t
h
e
2
4
t
h
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
D
i
st
r
i
b
u
t
e
d
C
o
m
p
u
t
i
n
g
a
n
d
N
e
t
w
o
rk
i
n
g
,
K
h
a
r
a
g
p
u
r
I
n
d
i
a
:
A
C
M
,
Ja
n
.
2
0
2
3
,
p
p
.
3
7
2
–
3
7
7
,
d
o
i
:
1
0
.
1
1
4
5
/
3
5
7
1
3
0
6
.
3
5
7
1
4
3
8
.
[
7
]
D
.
D
u
b
o
i
s
a
n
d
R
.
Le
p
a
g
e
,
“
F
a
st
a
n
d
e
f
f
i
c
i
e
n
t
e
v
a
l
u
a
t
i
o
n
o
f
b
u
i
l
d
i
n
g
d
a
ma
g
e
f
r
o
m
v
e
r
y
h
i
g
h
-
r
e
so
l
u
t
i
o
n
o
p
t
i
c
a
l
sat
e
l
l
i
t
e
i
m
a
g
e
s,
”
I
EEE
J
o
u
r
n
a
l
o
f
S
e
l
e
c
t
e
d
T
o
p
i
c
s
i
n
Ap
p
l
i
e
d
E
a
r
t
h
O
b
s
e
rv
a
t
i
o
n
s
a
n
d
R
e
m
o
t
e
S
e
n
si
n
g
,
v
o
l
.
7
,
n
o
.
1
0
,
p
p
.
4
1
6
7
–
4
1
7
6
,
O
c
t
.
2
0
1
4
,
d
o
i
:
1
0
.
1
1
0
9
/
JS
TA
R
S
.
2
0
1
4
.
2
3
3
6
2
3
6
.
[
8
]
S
.
D
u
,
S
.
D
u
,
B
.
L
i
u
,
X
.
Z
h
a
n
g
,
a
n
d
Z
.
Z
h
e
n
g
,
“
L
a
r
g
e
-
s
c
a
l
e
u
r
b
a
n
f
u
n
c
t
i
o
n
a
l
z
o
n
e
m
a
p
p
i
n
g
b
y
i
n
t
e
g
r
a
t
i
n
g
r
e
m
o
t
e
s
e
n
s
i
n
g
i
m
a
g
e
s
a
n
d
o
p
e
n
s
o
c
i
a
l
d
a
t
a
,
”
G
I
S
c
i
e
n
c
e
a
n
d
R
e
m
o
t
e
S
e
n
s
i
n
g
,
v
o
l
.
5
7
,
n
o
.
3
,
p
p
.
4
1
1
–
4
3
0
,
A
p
r
.
2
0
2
0
,
d
o
i
:
1
0
.
1
0
8
0
/
1
5
4
8
1
6
0
3
.
2
0
2
0
.
1
7
2
4
7
0
7
.
[
9
]
W
.
B
o
o
n
p
o
o
k
,
Y
.
Ta
n
,
K
.
T
o
r
sr
i
,
P
.
K
a
msi
n
g
,
P
.
To
r
t
e
e
k
a
,
a
n
d
A
.
N
a
r
d
k
u
l
p
a
t
,
“
P
C
L
–
P
TD
N
e
t
:
p
a
r
a
l
l
e
l
c
r
o
s
s
-
l
e
a
r
n
i
n
g
-
b
a
s
e
d
p
i
x
e
l
t
r
a
n
sf
e
r
r
e
d
d
e
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
t
w
o
r
k
f
o
r
b
u
i
l
d
i
n
g
e
x
t
r
a
c
t
i
o
n
i
n
d
e
n
s
e
b
u
i
l
d
i
n
g
a
r
e
a
s
w
i
t
h
s
h
a
d
o
w
,
”
I
E
EE
I
EEE
J
o
u
r
n
a
l
o
f
S
e
l
e
c
t
e
d
T
o
p
i
c
s
i
n
Ap
p
l
i
e
d
Ea
rt
h
O
b
s
e
rv
a
t
i
o
n
s
a
n
d
R
e
m
o
t
e
S
e
n
s
i
n
g
,
v
o
l
.
1
6
,
p
p
.
7
7
3
–
7
8
6
,
2
0
2
3
,
d
o
i
:
1
0
.
1
1
0
9
/
JS
TA
R
S
.
2
0
2
2
.
3
2
3
0
1
4
9
.
[
1
0
]
J.
L
o
n
g
,
E
.
S
h
e
l
h
a
mer,
a
n
d
T.
D
a
r
r
e
l
l
,
“
F
u
l
l
y
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
t
w
o
r
k
s
f
o
r
sema
n
t
i
c
se
g
me
n
t
a
t
i
o
n
,
”
i
n
2
0
1
5
I
E
E
E
C
o
n
f
e
re
n
c
e
o
n
C
o
m
p
u
t
e
r
Vi
s
i
o
n
a
n
d
Pa
t
t
e
rn
Re
c
o
g
n
i
t
i
o
n
(
C
VPR)
,
B
o
st
o
n
,
M
A
,
U
S
A
:
I
EEE,
Ju
n
.
2
0
1
5
,
p
p
.
3
4
3
1
–
3
4
4
0
,
d
o
i
:
1
0
.
1
1
0
9
/
C
V
P
R
.
2
0
1
5
.
7
2
9
8
9
6
5
.
[1
1]
J.
A
r
y
a
l
a
n
d
B
.
N
e
u
p
a
n
e
,
“
M
u
l
t
i
-
s
c
a
l
e
f
e
a
t
u
r
e
ma
p
a
g
g
r
e
g
a
t
i
o
n
a
n
d
s
u
p
e
r
v
i
se
d
d
o
mai
n
a
d
a
p
t
a
t
i
o
n
o
f
f
u
l
l
y
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
t
w
o
r
k
s
f
o
r
u
r
b
a
n
b
u
i
l
d
i
n
g
f
o
o
t
p
r
i
n
t
e
x
t
r
a
c
t
i
o
n
,
”
R
e
m
o
t
e
S
e
n
s
i
n
g
,
v
o
l
.
1
5
,
n
o
.
2
,
p
.
4
8
8
,
Ja
n
.
2
0
2
3
,
d
o
i
:
1
0
.
3
3
9
0
/
r
s1
5
0
2
0
4
8
8
.
[
1
2
]
S
.
R
a
n
,
J.
D
i
n
g
,
B
.
Li
u
,
X
.
G
e
,
a
n
d
G
.
M
a
,
“
M
u
l
t
i
-
U
-
N
e
t
:
r
e
si
d
u
a
l
mo
d
u
l
e
u
n
d
e
r
mu
l
t
i
se
n
s
o
r
y
f
i
e
l
d
a
n
d
a
t
t
e
n
t
i
o
n
me
c
h
a
n
i
sm
b
a
s
e
d
o
p
t
i
mi
z
e
d
U
-
N
e
t
f
o
r
V
H
R
i
ma
g
e
sem
a
n
t
i
c
s
e
g
m
e
n
t
a
t
i
o
n
,
”
S
e
n
so
rs
,
v
o
l
.
2
1
,
n
o
.
5
,
p
.
1
7
9
4
,
M
a
r
.
2
0
2
1
,
d
o
i
:
1
0
.
3
3
9
0
/
s
2
1
0
5
1
7
9
4
.
[
1
3
]
M.
D
i
x
i
t
,
K
.
C
h
a
u
r
a
s
i
a
,
a
n
d
V
.
K
.
M
i
s
h
r
a
,
“
D
i
l
a
t
e
d
-
R
e
sU
n
e
t
:
a
n
o
v
e
l
d
e
e
p
l
e
a
r
n
i
n
g
a
r
c
h
i
t
e
c
t
u
r
e
f
o
r
b
u
i
l
d
i
n
g
e
x
t
r
a
c
t
i
o
n
f
r
o
m
med
i
u
m
r
e
s
o
l
u
t
i
o
n
mu
l
t
i
-
sp
e
c
t
r
a
l
s
a
t
e
l
l
i
t
e
i
ma
g
e
r
y
,
”
Ex
p
e
rt
S
y
st
e
m
s
w
i
t
h
Ap
p
l
i
c
a
t
i
o
n
s
,
v
o
l
.
1
8
4
,
p
.
1
1
5
5
3
0
,
D
e
c
.
2
0
2
1
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
e
sw
a
.
2
0
2
1
.
1
1
5
5
3
0
.
[
1
4
]
Y
.
Z
h
a
n
g
e
t
a
l
.
,
“
O
n
o
p
t
i
m
i
z
i
n
g
mo
d
e
l
g
e
n
e
r
a
l
i
t
y
i
n
A
I
-
b
a
sed
d
i
sas
t
e
r
d
a
ma
g
e
a
sse
ssm
e
n
t
:
a
s
u
b
j
e
c
t
i
v
e
l
o
g
i
c
-
d
r
i
v
e
n
c
r
o
w
d
-
A
I
h
y
b
r
i
d
l
e
a
r
n
i
n
g
a
p
p
r
o
a
c
h
,
”
i
n
P
ro
c
e
e
d
i
n
g
s
o
f
t
h
e
T
h
i
r
t
y
-
S
e
c
o
n
d
I
n
t
e
r
n
a
t
i
o
n
a
l
J
o
i
n
t
C
o
n
f
e
re
n
c
e
o
n
Ar
t
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
,
M
a
c
a
u
,
S
A
R
C
h
i
n
a
:
I
n
t
e
r
n
a
t
i
o
n
a
l
Jo
i
n
t
C
o
n
f
e
r
e
n
c
e
s
o
n
A
r
t
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
O
r
g
a
n
i
z
a
t
i
o
n
,
A
u
g
.
2
0
2
3
,
p
p
.
6
3
1
7
–
6
3
2
5
,
d
o
i
:
1
0
.
2
4
9
6
3
/
i
j
c
a
i
.
2
0
2
3
/
7
0
1
.
[
1
5
]
M
.
Y
u
,
X
.
C
h
e
n
,
W
.
Z
h
a
n
g
,
a
n
d
Y
.
L
i
u
,
“
AGs
-
U
n
e
t
:
B
u
i
l
d
i
n
g
e
x
t
r
a
c
t
i
o
n
m
o
d
e
l
f
o
r
h
i
g
h
r
e
s
o
l
u
t
i
o
n
r
e
m
o
t
e
se
n
si
n
g
i
m
a
g
e
s
b
a
s
e
d
o
n
a
t
t
e
n
t
i
o
n
g
a
t
e
s U
n
e
t
w
o
r
k
,
”
S
e
n
so
r
s
,
v
o
l
.
2
2
,
n
o
.
8
,
p
.
2
9
3
2
,
A
p
r
.
2
0
2
2
,
d
o
i
:
1
0
.
3
3
9
0
/
s2
2
0
8
2
9
3
2
.
[
1
6
]
H
.
W
a
n
g
a
n
d
F
.
M
i
a
o
,
“
B
u
i
l
d
i
n
g
e
x
t
r
a
c
t
i
o
n
f
r
o
m rem
o
t
e
se
n
si
n
g
i
ma
g
e
s
u
s
i
n
g
d
e
e
p
r
e
s
i
d
u
a
l
U
-
N
e
t
,
”
E
u
ro
p
e
a
n
J
o
u
r
n
a
l
o
f
Re
m
o
t
e
S
e
n
si
n
g
,
v
o
l
.
5
5
,
n
o
.
1
,
p
p
.
7
1
–
8
5
,
D
e
c
.
2
0
2
2
,
d
o
i
:
1
0
.
1
0
8
0
/
2
2
7
9
7
2
5
4
.
2
0
2
1
.
2
0
1
8
9
4
4
.
[
1
7
]
S
.
T
o
u
z
a
n
i
a
n
d
J.
G
r
a
n
d
e
r
so
n
,
“
O
p
e
n
d
a
t
a
a
n
d
d
e
e
p
s
e
ma
n
t
i
c
se
g
me
n
t
a
t
i
o
n
f
o
r
a
u
t
o
m
a
t
e
d
e
x
t
r
a
c
t
i
o
n
o
f
b
u
i
l
d
i
n
g
f
o
o
t
p
r
i
n
t
s
,
”
Re
m
o
t
e
S
e
n
s
i
n
g
,
v
o
l
.
1
3
,
n
o
.
1
3
,
p
.
2
5
7
8
,
J
u
l
.
2
0
2
1
,
d
o
i
:
1
0
.
3
3
9
0
/
r
s1
3
1
3
2
5
7
8
.
[
1
8
]
M
.
J
.
B
o
s
c
o
a
n
d
G
.
W
a
n
g
,
“
M
u
l
t
i
l
e
v
e
l
a
n
d
mu
l
t
i
-
g
r
a
n
u
l
a
r
i
t
y
o
f
r
e
mo
t
e
se
n
si
n
g
i
ma
g
e
r
y
a
p
p
l
i
c
a
t
i
o
n
b
a
s
e
d
o
n
d
e
e
p
l
e
a
r
n
i
n
g
a
n
d
mac
h
i
n
e
l
e
a
r
n
i
n
g
a
l
g
o
r
i
t
h
m
,
”
i
n
2
0
2
1
T
h
e
4
t
h
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
r
e
n
c
e
o
n
M
a
c
h
i
n
e
L
e
a
r
n
i
n
g
a
n
d
M
a
c
h
i
n
e
I
n
t
e
l
l
i
g
e
n
c
e
,
H
a
n
g
z
h
o
u
C
h
i
n
a
:
A
C
M
,
S
e
p
.
2
0
2
1
,
p
p
.
1
1
1
–
1
1
9
,
d
o
i
:
1
0
.
1
1
4
5
/
3
4
9
0
7
2
5
.
3
4
9
0
7
4
2
.
[
1
9
]
L.
M
o
y
a
,
C
.
G
e
i
s,
M
.
H
a
s
h
i
m
o
t
o
,
E.
M
a
s
,
S
.
K
o
s
h
i
m
u
r
a
,
a
n
d
G
.
S
t
r
u
n
z
,
“
D
i
sast
e
r
i
n
t
e
n
si
t
y
-
b
a
se
d
s
e
l
e
c
t
i
o
n
o
f
t
r
a
i
n
i
n
g
sam
p
l
e
s f
o
r
r
e
mo
t
e
se
n
si
n
g
b
u
i
l
d
i
n
g
d
a
ma
g
e
c
l
a
ssi
f
i
c
a
t
i
o
n
,
”
I
EEE
T
r
a
n
sa
c
t
i
o
n
s
o
n
G
e
o
sc
i
e
n
c
e
a
n
d
R
e
m
o
t
e
S
e
n
s
i
n
g
,
v
o
l
.
5
9
,
n
o
.
1
0
,
p
p
.
8
2
8
8
–
8
3
0
4
,
O
c
t
.
2
0
2
1
,
d
o
i
:
1
0
.
1
1
0
9
/
TG
R
S
.
2
0
2
0
.
3
0
4
6
0
0
4
.
[
2
0
]
Y
.
R
e
n
,
Y
.
Y
u
,
a
n
d
H
.
G
u
a
n
,
“
DA
-
C
a
p
sU
N
e
t
:
a
d
u
a
l
-
a
t
t
e
n
t
i
o
n
c
a
p
s
u
l
e
U
-
N
e
t
f
o
r
ro
a
d
e
x
t
r
a
c
t
i
o
n
f
r
o
m
r
e
m
o
t
e
s
e
n
s
i
n
g
i
ma
g
e
r
y
,
”
Re
m
o
t
e
S
e
n
s
i
n
g
,
v
o
l
.
1
2
,
n
o
.
1
8
,
p
.
2
8
6
6
,
S
e
p
.
2
0
2
0
,
d
o
i
:
1
0
.
3
3
9
0
/
r
s
1
2
1
8
2
8
6
6
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J I
n
f
&
C
o
m
m
u
n
T
ec
h
n
o
l
I
SS
N:
2252
-
8
7
7
6
Tech
n
iq
u
es o
f d
ee
p
lea
r
n
in
g
n
eu
r
a
l n
etw
o
r
k
b
a
s
ed
b
u
ild
i
n
g
f
ea
tu
r
e
ex
tr
a
ctio
n
fr
o
m
…
(
S
h
r
in
iva
s
K
h
a
n
d
a
r
e
)
623
[
2
1
]
H
.
M
.
A
l
a
o
u
i
,
H
.
R
a
d
o
i
n
e
,
J.
C
h
e
n
a
l
,
H
.
H
a
j
j
i
,
a
n
d
H
.
Y
a
k
u
b
u
,
“
D
e
e
p
b
u
i
l
d
i
n
g
f
o
o
t
p
r
i
n
t
e
x
t
r
a
c
t
i
o
n
f
o
r
u
r
b
a
n
r
i
sk
a
ss
e
ssm
e
n
t
–
r
e
m
o
t
e
se
n
si
n
g
a
n
d
d
e
e
p
l
e
a
r
n
i
n
g
b
a
sed
a
p
p
r
o
a
c
h
,
”
I
n
t
e
rn
a
t
i
o
n
a
l
Ar
c
h
i
v
e
s
o
f
t
h
e
P
h
o
t
o
g
ra
m
m
e
t
ry
,
Re
m
o
t
e
S
e
n
si
n
g
a
n
d
S
p
a
t
i
a
l
I
n
f
o
rm
a
t
i
o
n
S
c
i
e
n
c
e
s
.
,
v
o
l
.
X
LV
I
I
I
-
4
/
W
3
-
2
0
2
2
,
p
p
.
8
3
–
8
6
,
D
e
c
.
2
0
2
2
,
d
o
i
:
1
0
.
5
1
9
4
/
i
s
p
r
s
-
a
r
c
h
i
v
e
s
-
X
LV
I
I
I
-
4
-
W3
-
2
0
2
2
-
83
-
2
0
2
2
.
[
2
2
]
G
.
A
n
d
r
e
s
i
n
i
,
A
.
A
p
p
i
c
e
,
D
.
I
a
i
a
,
D
.
M
a
l
e
r
b
a
,
N
.
T
a
g
g
i
o
,
a
n
d
A
.
A
i
e
l
l
o
,
“
Le
v
e
r
a
g
i
n
g
a
u
t
o
e
n
c
o
d
e
r
s
i
n
c
h
a
n
g
e
v
e
c
t
o
r
a
n
a
l
y
s
i
s
o
f
o
p
t
i
c
a
l
sa
t
e
l
l
i
t
e
i
ma
g
e
s
,
”
J
o
u
r
n
a
l
o
f
I
n
t
e
l
l
i
g
e
n
t
I
n
f
o
rm
a
t
i
o
n
S
y
st
e
m
s
,
v
o
l
.
5
8
,
n
o
.
3
,
p
p
.
4
3
3
–
4
5
2
,
Ju
n
.
2
0
2
2
,
d
o
i
:
1
0
.
1
0
0
7
/
s
1
0
8
4
4
-
021
-
0
0
6
7
0
-
9.
[
2
3
]
F
.
A
l
a
m,
T.
A
l
a
m,
M
d
.
A
.
H
a
s
a
n
,
A
.
H
a
sn
a
t
,
M
.
I
mr
a
n
,
a
n
d
F
.
O
f
l
i
,
“
M
ED
I
C
:
a
mu
l
t
i
-
t
a
s
k
l
e
a
r
n
i
n
g
d
a
t
a
set
f
o
r
d
i
sas
t
e
r
i
m
a
g
e
c
l
a
ss
i
f
i
c
a
t
i
o
n
,
”
N
e
u
r
a
l
C
o
m
p
u
t
i
n
g
a
n
d
A
p
p
l
i
c
a
t
i
o
n
s
,
v
o
l
.
3
5
,
n
o
.
3
,
p
p
.
2
6
0
9
–
2
6
3
2
,
Ja
n
.
2
0
2
3
,
d
o
i
:
1
0
.
1
0
0
7
/
s
0
0
5
2
1
-
0
2
2
-
0
7
7
1
7
-
0.
[
2
4
]
H
.
X
u
,
P
.
Z
h
u
,
X
.
L
u
o
,
T.
X
i
e
,
a
n
d
L.
Z
h
a
n
g
,
“
E
x
t
r
a
c
t
i
n
g
b
u
i
l
d
i
n
g
s
f
r
o
m
r
e
m
o
t
e
s
e
n
s
i
n
g
i
m
a
g
e
s
u
s
i
n
g
a
m
u
l
t
i
t
a
s
k
e
n
c
o
d
e
r
-
d
e
c
o
d
e
r
n
e
t
w
o
r
k
w
i
t
h
b
o
u
n
d
a
r
y
r
e
f
i
n
e
men
t
,
”
Re
m
o
t
e
S
e
n
si
n
g
,
v
o
l
.
1
4
,
n
o
.
3
,
p
.
5
6
4
,
J
a
n
.
2
0
2
2
,
d
o
i
:
1
0
.
3
3
9
0
/
r
s1
4
0
3
0
5
6
4
.
[
2
5
]
B
.
Y
u
,
F
.
C
h
e
n
,
N
.
W
a
n
g
,
L.
Y
a
n
g
,
H
.
Y
a
n
g
,
a
n
d
L.
W
a
n
g
,
“
M
S
F
Tr
a
n
s
:
a
mu
l
t
i
-
t
a
sk
f
r
e
q
u
e
n
c
y
-
sp
a
t
i
a
l
l
e
a
r
n
i
n
g
t
r
a
n
sf
o
r
mer
f
o
r
b
u
i
l
d
i
n
g
e
x
t
r
a
c
t
i
o
n
f
r
o
m
h
i
g
h
sp
a
t
i
a
l
r
e
so
l
u
t
i
o
n
r
e
m
o
t
e
s
e
n
s
i
n
g
i
m
a
g
e
s,
”
G
I
S
c
i
e
n
c
e
and
R
e
m
o
t
e
S
e
n
s
i
n
g
,
v
o
l
.
5
9
,
n
o
.
1
,
p
p
.
1
9
7
8
–
1
9
9
6
,
D
e
c
.
2
0
2
2
,
d
o
i
:
1
0
.
1
0
8
0
/
1
5
4
8
1
6
0
3
.
2
0
2
2
.
2
1
4
3
6
7
8
.
[
2
6
]
S
.
S
a
h
a
,
S
.
Zh
a
o
,
a
n
d
X
.
X
.
Z
h
u
,
“
M
u
l
t
i
t
a
r
g
e
t
d
o
ma
i
n
a
d
a
p
t
a
t
i
o
n
f
o
r
r
e
m
o
t
e
se
n
s
i
n
g
c
l
a
ss
i
f
i
c
a
t
i
o
n
u
si
n
g
g
r
a
p
h
n
e
u
r
a
l
n
e
t
w
o
r
k
,
”
I
EEE
G
e
o
sc
i
e
n
c
e
a
n
d
Re
m
o
t
e
S
e
n
si
n
g
L
e
t
t
e
rs
,
v
o
l
.
1
9
,
p
p
.
1
–
5
,
2
0
2
2
,
d
o
i
:
1
0
.
1
1
0
9
/
LG
R
S
.
2
0
2
2
.
3
1
4
9
9
5
0
.
[
2
7
]
F
.
F
a
n
g
e
t
a
l
.
,
“
I
mp
r
o
v
e
d
p
se
u
d
o
m
a
s
k
s
g
e
n
e
r
a
t
i
o
n
f
o
r
w
e
a
k
l
y
su
p
e
r
v
i
s
e
d
b
u
i
l
d
i
n
g
e
x
t
r
a
c
t
i
o
n
f
r
o
m
h
i
g
h
-
r
e
s
o
l
u
t
i
o
n
r
e
m
o
t
e
s
e
n
s
i
n
g
i
ma
g
e
r
y
,
”
I
E
EE
J
o
u
rn
a
l
o
f
S
e
l
e
c
t
e
d
T
o
p
i
c
s
i
n
Ap
p
l
i
e
d
E
a
r
t
h
O
b
ser
v
a
t
i
o
n
s
a
n
d
R
e
m
o
t
e
S
e
n
s
i
n
g
,
v
o
l
.
1
5
,
p
p
.
1
6
2
9
–
1
6
4
2
,
2
0
2
2
,
d
o
i
:
1
0
.
1
1
0
9
/
JS
TA
R
S
.
2
0
2
2
.
3
1
4
4
1
7
6
.
[
2
8
]
L.
Li
,
W
.
Z
h
a
n
g
,
X
.
Zh
a
n
g
,
M
.
Em
a
m,
a
n
d
W
.
J
i
n
g
,
“
S
e
mi
-
s
u
p
e
r
v
i
se
d
r
e
mo
t
e
se
n
si
n
g
i
ma
g
e
se
ma
n
t
i
c
se
g
men
t
a
t
i
o
n
m
e
t
h
o
d
b
a
s
e
d
o
n
d
e
e
p
l
e
a
r
n
i
n
g
,
”
El
e
c
t
r
o
n
i
c
s
,
v
o
l
.
1
2
,
n
o
.
2
,
p
.
3
4
8
,
Ja
n
.
2
0
2
3
,
d
o
i
:
1
0
.
3
3
9
0
/
e
l
e
c
t
r
o
n
i
c
s
1
2
0
2
0
3
4
8
.
[
2
9
]
L.
Lo
n
g
,
F
.
H
e
,
a
n
d
H
.
L
i
u
,
“
T
h
e
u
se
o
f
r
e
mo
t
e
se
n
s
i
n
g
sa
t
e
l
l
i
t
e
u
si
n
g
d
e
e
p
l
e
a
r
n
i
n
g
i
n
e
merg
e
n
c
y
m
o
n
i
t
o
r
i
n
g
o
f
h
i
g
h
-
l
e
v
e
l
l
a
n
d
s
l
i
d
e
s
d
i
s
a
st
e
r
i
n
J
i
n
s
h
a
R
i
v
e
r
,
”
J
o
u
r
n
a
l
S
u
p
e
r
c
o
m
p
u
t
,
v
o
l
.
7
7
,
n
o
.
8
,
p
p
.
8
7
2
8
–
8
7
4
4
,
A
u
g
.
2
0
2
1
,
d
o
i
:
1
0
.
1
0
0
7
/
s
1
1
2
2
7
-
020
-
0
3
6
0
4
-
4.
[
3
0
]
Y
.
H
e
,
J
.
W
a
n
g
,
C
.
L
i
a
o
,
X
.
Z
h
o
u
,
a
n
d
B
.
S
h
a
n
,
“
M
S
4
D
-
N
e
t
:
M
u
l
t
i
t
a
sk
-
b
a
sed
s
e
mi
-
s
u
p
e
r
v
i
se
d
s
e
ma
n
t
i
c
se
g
me
n
t
a
t
i
o
n
f
r
a
mew
o
r
k
w
i
t
h
p
e
r
t
u
r
b
e
d
d
u
a
l
me
a
n
t
e
a
c
h
e
r
s
f
o
r
b
u
i
l
d
i
n
g
d
a
ma
g
e
a
ssessm
e
n
t
f
r
o
m
h
i
g
h
-
r
e
so
l
u
t
i
o
n
r
e
mo
t
e
se
n
si
n
g
i
ma
g
e
r
y
,
”
Re
m
o
t
e
S
e
n
s
i
n
g
,
v
o
l
.
1
5
,
n
o
.
2
,
p
.
4
7
8
,
Ja
n
.
2
0
2
3
,
d
o
i
:
1
0
.
3
3
9
0
/
r
s
1
5
0
2
0
4
7
8
.
[
3
1
]
R
.
Ü
n
l
ü
a
n
d
R
.
K
i
r
i
ş,
“
D
e
t
e
c
t
i
o
n
o
f
d
a
m
a
g
e
d
b
u
i
l
d
i
n
g
s
a
f
t
e
r
a
n
e
a
r
t
h
q
u
a
k
e
w
i
t
h
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
u
r
a
l
n
e
t
w
o
r
k
s
i
n
c
o
n
j
u
n
c
t
i
o
n
w
i
t
h
i
m
a
g
e
s
e
g
m
e
n
t
a
t
i
o
n
,
”
T
h
e
V
i
s
u
a
l
C
o
m
p
u
t
e
r
,
v
o
l
.
3
8
,
n
o
.
2
,
p
p
.
6
8
5
–
6
9
4
,
F
e
b
.
2
0
2
2
,
d
o
i
:
1
0
.
1
0
0
7
/
s0
0
3
7
1
-
0
2
0
-
0
2
0
4
3
-
9.
[
3
2
]
N
.
Ty
a
g
i
a
n
d
M
.
S
a
r
a
sw
a
t
,
“
I
d
e
n
t
i
f
i
c
a
t
i
o
n
o
f
u
n
d
a
m
a
g
e
d
b
u
i
l
d
i
n
g
s a
f
t
e
r
t
h
e
e
v
e
n
t
o
f
d
i
s
a
st
e
r
u
si
n
g
d
e
e
p
l
e
a
r
n
i
n
g
,
”
i
n
Pro
c
e
e
d
i
n
g
s
o
f
t
h
e
2
0
2
2
F
o
u
r
t
e
e
n
t
h
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
C
o
n
t
e
m
p
o
ra
r
y
C
o
m
p
u
t
i
n
g
,
N
o
i
d
a
I
n
d
i
a
:
A
C
M
,
A
u
g
.
2
0
2
2
,
p
p
.
2
7
–
3
1
,
d
o
i
:
1
0
.
1
1
4
5
/
3
5
4
9
2
0
6
.
3
5
4
9
2
1
2
.
[
3
3
]
J.
L
i
,
X
.
H
u
a
n
g
,
L.
T
u
,
T
.
Zh
a
n
g
,
a
n
d
L.
W
a
n
g
,
“
A
r
e
v
i
e
w
o
f
b
u
i
l
d
i
n
g
d
e
t
e
c
t
i
o
n
f
r
o
m
v
e
r
y
h
i
g
h
r
e
s
o
l
u
t
i
o
n
o
p
t
i
c
a
l
r
e
m
o
t
e
se
n
si
n
g
i
ma
g
e
s,
”
G
I
S
c
i
e
n
c
e
a
n
d
R
e
m
o
t
e
S
e
n
s
i
n
g
,
v
o
l
.
5
9
,
n
o
.
1
,
p
p
.
1
1
9
9
–
1
2
2
5
,
D
e
c
.
2
0
2
2
,
d
o
i
:
1
0
.
1
0
8
0
/
1
5
4
8
1
6
0
3
.
2
0
2
2
.
2
1
0
1
7
2
7
.
[
3
4
]
Y
.
Ta
n
g
,
F
.
Q
i
u
,
B
.
W
a
n
g
,
D
.
W
u
,
L
.
Ji
n
g
,
a
n
d
Z
.
S
u
n
,
“
A
d
e
e
p
r
e
l
e
a
r
n
i
n
g
met
h
o
d
b
a
s
e
d
o
n
t
h
e
r
e
c
u
r
r
e
n
t
n
e
u
r
a
l
n
e
t
w
o
r
k
f
o
r
l
a
n
d
c
o
v
e
r
c
l
a
ss
i
f
i
c
a
t
i
o
n
,
”
G
I
S
c
i
e
n
c
e
a
n
d
Re
m
o
t
e
S
e
n
si
n
g
,
v
o
l
.
5
9
,
n
o
.
1
,
p
p
.
1
3
4
4
–
1
3
6
6
,
D
e
c
.
2
0
2
2
,
d
o
i
:
1
0
.
1
0
8
0
/
1
5
4
8
1
6
0
3
.
2
0
2
2
.
2
1
1
5
5
8
9
.
[
3
5
]
L.
-
L.
D
r
a
g
o
l
e
a
e
t
a
l
.
,
“
D
e
t
e
r
m
i
n
i
n
g
f
a
c
t
o
r
s
i
n
s
h
a
p
i
n
g
t
h
e
s
u
st
a
i
n
a
b
l
e
b
e
h
a
v
i
o
r
o
f
t
h
e
g
e
n
e
r
a
t
i
o
n
Z
c
o
n
s
u
me
r
,
”
Fr
o
n
t
i
e
rs
i
n
En
v
i
r
o
n
m
e
n
t
a
l
S
c
i
e
n
c
e
,
v
o
l
.
1
1
,
p
.
1
0
9
6
1
8
3
,
Ja
n
.
2
0
2
3
,
d
o
i
:
1
0
.
3
3
8
9
/
f
e
n
v
s.
2
0
2
3
.
1
0
9
6
1
8
3
.
[
3
6
]
M
.
A
.
K
u
m
a
r
a
n
d
A
.
J.
L
a
x
mi
,
“
M
a
c
h
i
n
e
l
e
a
r
n
i
n
g
b
a
s
e
d
i
n
t
e
n
t
i
o
n
a
l
i
s
l
a
n
d
i
n
g
a
l
g
o
r
i
t
h
m
f
o
r
D
E
R
s
i
n
d
i
sast
e
r
m
a
n
a
g
e
me
n
t
,
”
I
EEE
A
c
c
e
ss
,
v
o
l
.
9
,
p
p
.
8
5
3
0
0
–
8
5
3
0
9
,
2
0
2
1
,
d
o
i
:
1
0
.
1
1
0
9
/
A
C
C
ESS
.
2
0
2
1
.
3
0
8
7
9
1
4
.
[
3
7
]
R
.
Y
o
u
ssef
,
M
.
A
n
i
ss,
a
n
d
C
.
J
a
ma
l
,
“
M
a
c
h
i
n
e
l
e
a
r
n
i
n
g
a
n
d
d
e
e
p
l
e
a
r
n
i
n
g
i
n
r
e
m
o
t
e
se
n
s
i
n
g
a
n
d
u
r
b
a
n
a
p
p
l
i
c
a
t
i
o
n
:
a
s
y
st
e
ma
t
i
c
r
e
v
i
e
w
a
n
d
me
t
a
-
a
n
a
l
y
si
s
,
”
i
n
Pro
c
e
e
d
i
n
g
s
o
f
t
h
e
4
t
h
E
d
i
t
i
o
n
o
f
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
G
e
o
-
I
T
a
n
d
Wa
t
e
r
Re
s
o
u
r
c
e
s
2
0
2
0
,
G
e
o
-
I
T
a
n
d
Wa
t
e
r
Re
s
o
u
rce
s
2
0
2
0
,
A
l
-
H
o
c
e
i
m
a
M
o
r
o
c
c
o
:
A
C
M
,
M
a
r
.
2
0
2
0
,
p
p
.
1
–
5,
d
o
i
:
1
0
.
1
1
4
5
/
3
3
9
9
2
0
5
.
3
3
9
9
2
2
4
.
[
3
8
]
X
.
W
e
n
e
t
a
l
.
,
“
A
b
u
i
l
d
i
n
g
sh
a
p
e
v
e
c
t
o
r
i
z
a
t
i
o
n
h
i
e
r
a
r
c
h
y
f
r
o
m
V
H
R
r
e
mo
t
e
s
e
n
s
i
n
g
i
m
a
g
e
r
y
c
o
mb
i
n
e
d
D
C
N
N
s
-
b
a
se
d
e
d
g
e
d
e
t
e
c
t
i
o
n
a
n
d
P
C
A
-
b
a
se
d
c
o
r
n
e
r
e
x
t
r
a
c
t
i
o
n
,
”
I
EEE
J
o
u
rn
a
l
o
f
S
e
l
e
c
t
e
d
T
o
p
i
c
s
i
n
A
p
p
l
i
e
d
E
a
r
t
h
O
b
s
e
rv
a
t
i
o
n
s
a
n
d
Re
m
o
t
e
S
e
n
si
n
g
,
v
o
l
.
1
6
,
p
p
.
7
5
0
–
7
6
1
,
2
0
2
3
,
d
o
i
:
1
0
.
1
1
0
9
/
JS
TA
R
S
.
2
0
2
2
.
3
2
3
1
3
4
8
.
[
3
9
]
L.
A
h
m
e
d
,
K
.
A
h
ma
d
,
N
.
S
a
i
d
,
B
.
Q
o
l
o
ma
n
y
,
J
.
Q
a
d
i
r
,
a
n
d
A
.
A
l
-
F
u
q
a
h
a
,
“
A
c
t
i
v
e
l
e
a
r
n
i
n
g
b
a
s
e
d
f
e
d
e
r
a
t
e
d
l
e
a
r
n
i
n
g
f
o
r
w
a
st
e
a
n
d
n
a
t
u
r
a
l
d
i
s
a
st
e
r
i
ma
g
e
c
l
a
ssi
f
i
c
a
t
i
o
n
,
”
I
E
EE
Ac
c
e
ss
,
v
o
l
.
8
,
p
p
.
2
0
8
5
1
8
–
2
0
8
5
3
1
,
2
0
2
0
,
d
o
i
:
1
0
.
1
1
0
9
/
A
C
C
ESS
.
2
0
2
0
.
3
0
3
8
6
7
6
.
[
4
0
]
R
.
G
u
p
t
a
a
n
d
M
.
S
h
a
h
,
“
R
e
sc
u
e
N
e
t
:
j
o
i
n
t
b
u
i
l
d
i
n
g
se
g
m
e
n
t
a
t
i
o
n
a
n
d
d
a
m
a
g
e
a
ssessm
e
n
t
f
r
o
m
sa
t
e
l
l
i
t
e
i
ma
g
e
r
y
,
”
i
n
2
0
2
0
2
5
t
h
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
r
e
n
c
e
o
n
P
a
t
t
e
r
n
Re
c
o
g
n
i
t
i
o
n
(
I
C
P
R)
,
M
i
l
a
n
,
I
t
a
l
y
:
I
EEE,
Ja
n
.
2
0
2
1
,
p
p
.
4
4
0
5
–
4
4
1
1
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
P
R
4
8
8
0
6
.
2
0
2
1
.
9
4
1
2
2
9
5
.
[
4
1
]
K
.
R
.
N
i
a
a
n
d
G
.
M
o
r
i
,
“
B
u
i
l
d
i
n
g
d
a
mag
e
a
ssess
me
n
t
u
si
n
g
d
e
e
p
l
e
a
r
n
i
n
g
a
n
d
g
r
o
u
n
d
-
l
e
v
e
l
i
m
a
g
e
d
a
t
a
,
”
i
n
2
0
1
7
1
4
t
h
C
o
n
f
e
r
e
n
c
e
o
n
C
o
m
p
u
t
e
r
a
n
d
R
o
b
o
t
Vi
si
o
n
(
C
R
V)
,
E
d
mo
n
t
o
n
,
A
B
:
I
EEE,
M
a
y
2
0
1
7
,
p
p
.
9
5
–
1
0
2
.
d
o
i
:
1
0
.
1
1
0
9
/
C
R
V
.
2
0
1
7
.
5
4
.
[
4
2
]
J.
B
o
w
m
a
n
a
n
d
L.
Y
a
n
g
,
“
F
e
w
-
sh
o
t
l
e
a
r
n
i
n
g
f
o
r
p
o
s
t
-
d
i
sas
t
e
r
s
t
r
u
c
t
u
r
e
d
a
m
a
g
e
a
ssess
me
n
t
,
”
i
n
Pro
c
e
e
d
i
n
g
s
o
f
t
h
e
4
t
h
A
C
M
S
I
G
S
P
ATI
AL
I
n
t
e
rn
a
t
i
o
n
a
l
Wo
r
k
sh
o
p
o
n
AI
f
o
r
G
e
o
g
r
a
p
h
i
c
K
n
o
w
l
e
d
g
e
D
i
sco
v
e
ry
,
B
e
i
j
i
n
g
C
h
i
n
a
:
A
C
M
,
N
o
v
.
2
0
2
1
,
p
p
.
27
–
3
2
,
d
o
i
:
1
0
.
1
1
4
5
/
3
4
8
6
6
3
5
.
3
4
9
1
0
7
1
.
[
4
3
]
W
.
A
l
sa
b
h
a
n
a
n
d
T
.
A
l
o
t
a
i
b
y
,
“
A
u
t
o
m
a
t
i
c
b
u
i
l
d
i
n
g
e
x
t
r
a
c
t
i
o
n
o
n
sa
t
e
l
l
i
t
e
i
m
a
g
e
s
u
si
n
g
u
n
e
t
a
n
d
R
e
sN
e
t
5
0
,
”
C
o
m
p
u
t
a
t
i
o
n
a
l
I
n
t
e
l
l
i
g
e
n
c
e
a
n
d
N
e
u
r
o
sc
i
e
n
c
e
,
v
o
l
.
2
0
2
2
,
p
p
.
1
–
1
2
,
F
e
b
.
2
0
2
2
,
d
o
i
:
1
0
.
1
1
5
5
/
2
0
2
2
/
5
0
0
8
8
5
4
.
[
4
4
]
Q
.
Li
u
e
t
a
l
.
,
“
C
F
N
e
t
:
a
n
e
i
g
e
n
v
a
l
u
e
p
r
e
s
e
r
v
e
d
a
p
p
r
o
a
c
h
t
o
mu
l
t
i
sc
a
l
e
b
u
i
l
d
i
n
g
se
g
me
n
t
a
t
i
o
n
i
n
h
i
g
h
-
r
e
s
o
l
u
t
i
o
n
r
e
m
o
t
e
se
n
s
i
n
g
i
ma
g
e
s
,
”
I
E
EE
J
o
u
r
n
a
l
o
f
S
e
l
e
c
t
e
d
T
o
p
i
c
s
i
n
A
p
p
l
i
e
d
E
a
rt
h
O
b
ser
v
a
t
i
o
n
s
a
n
d
Re
m
o
t
e
S
e
n
si
n
g
,
v
o
l
.
1
6
,
p
p
.
2
4
8
1
–
2
4
9
1
,
2
0
2
3
,
d
o
i
:
1
0
.
1
1
0
9
/
JS
TA
R
S
.
2
0
2
3
.
3
2
4
4
3
3
6
.
[
4
5
]
T.
K
a
v
z
o
g
l
u
,
I
.
C
o
l
k
e
s
e
n
,
a
n
d
T.
Y
o
mr
a
l
i
o
g
l
u
,
“
O
b
j
e
c
t
-
b
a
se
d
c
l
a
ssi
f
i
c
a
t
i
o
n
w
i
t
h
r
o
t
a
t
i
o
n
f
o
r
e
s
t
e
n
sem
b
l
e
l
e
a
r
n
i
n
g
a
l
g
o
r
i
t
h
m
u
s
i
n
g
v
e
r
y
-
h
i
g
h
-
r
e
s
o
l
u
t
i
o
n
W
o
r
l
d
V
i
e
w
-
2
i
ma
g
e
,
”
Re
m
o
t
e
S
e
n
si
n
g
L
e
t
t
e
rs
,
v
o
l
.
6
,
n
o
.
1
1
,
p
p
.
8
3
4
–
8
4
3
,
N
o
v
.
2
0
1
5
,
d
o
i
:
1
0
.
1
0
8
0
/
2
1
5
0
7
0
4
X
.
2
0
1
5
.
1
0
8
4
5
5
0
.
[
4
6
]
H
.
H
o
ssei
n
p
o
u
r
,
F
.
S
a
ma
d
z
a
d
e
g
a
n
,
a
n
d
F
.
D
.
J
a
v
a
n
,
“
A
n
o
v
e
l
b
o
u
n
d
a
r
y
l
o
ss
f
u
n
c
t
i
o
n
i
n
d
e
e
p
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
t
w
o
r
k
s
t
o
i
m
p
r
o
v
e
t
h
e
b
u
i
l
d
i
n
g
s
e
x
t
r
a
c
t
i
o
n
f
r
o
m
h
i
g
h
-
r
e
so
l
u
t
i
o
n
r
e
m
o
t
e
s
e
n
si
n
g
i
ma
g
e
s
,
”
I
EEE
J
o
u
rn
a
l
o
f
S
e
l
e
c
t
e
d
T
o
p
i
c
s
i
n
A
p
p
l
i
e
d
E
a
rt
h
O
b
serv
a
t
i
o
n
s
a
n
d
R
e
m
o
t
e
S
e
n
s
i
n
g
,
v
o
l
.
1
5
,
p
p
.
4
4
3
7
–
4
4
5
4
,
2
0
2
2
,
d
o
i
:
1
0
.
1
1
0
9
/
JS
TA
R
S
.
2
0
2
2
.
3
1
7
8
4
7
0
.
[
4
7
]
Z.
W
a
n
g
,
N
.
X
u
,
B
.
W
a
n
g
,
Y
.
Li
u
,
a
n
d
S
.
Z
h
a
n
g
,
“
U
r
b
a
n
b
u
i
l
d
i
n
g
e
x
t
r
a
c
t
i
o
n
f
r
o
m
h
i
g
h
-
r
e
s
o
l
u
t
i
o
n
r
e
mo
t
e
s
e
n
s
i
n
g
i
m
a
g
e
r
y
b
a
s
e
d
o
n
m
u
l
t
i
-
sc
a
l
e
r
e
c
u
r
r
e
n
t
c
o
n
d
i
t
i
o
n
a
l
g
e
n
e
r
a
t
i
v
e
a
d
v
e
r
s
a
r
i
a
l
n
e
t
w
o
r
k
,
”
G
I
S
c
i
e
n
c
e
a
n
d
R
e
m
o
t
e
S
e
n
s
i
n
g
,
v
o
l
.
5
9
,
n
o
.
1
,
p
p
.
8
6
1
–
8
8
4
,
D
e
c
.
2
0
2
2
,
d
o
i
:
1
0
.
1
0
8
0
/
1
5
4
8
1
6
0
3
.
2
0
2
2
.
2
0
7
6
3
8
2
.
[
4
8
]
R
.
B
l
a
i
r
,
“
C
l
o
u
d
-
b
a
s
e
d
s
e
r
v
i
c
e
s
a
n
d
r
e
c
o
m
me
n
d
a
t
i
o
n
s
,
”
p
r
e
se
n
t
e
d
a
t
t
h
e
1
4
t
h
I
n
t
e
rn
a
t
i
o
n
a
l
M
u
l
t
i
-
C
o
n
f
e
re
n
c
e
o
n
C
o
m
p
l
e
x
i
t
y
,
I
n
f
o
rm
a
t
i
c
s
a
n
d
C
y
b
e
r
n
e
t
i
c
s,
Vi
r
t
u
a
l
C
o
n
f
e
re
n
c
e
,
M
a
r
.
2
0
2
3
,
p
p
.
9
8
–
1
0
4
.
d
o
i
:
1
0
.
5
4
8
0
8
/
I
M
C
I
C
2
0
2
3
.
0
1
.
9
8
.
[
4
9
]
Q
.
Z
o
u
,
G
.
Li
,
a
n
d
W
.
Y
u
,
“
C
l
o
u
d
c
o
mp
u
t
i
n
g
b
a
se
d
o
n
c
o
mp
u
t
a
t
i
o
n
a
l
c
h
a
r
a
c
t
e
r
i
s
t
i
c
s
f
o
r
d
i
s
a
st
e
r
m
o
n
i
t
o
r
i
n
g
,
”
A
p
p
l
i
e
d
S
c
i
e
n
c
e
s
,
v
o
l
.
1
0
,
n
o
.
1
9
,
p
.
6
6
7
6
,
S
e
p
.
2
0
2
0
,
d
o
i
:
1
0
.
3
3
9
0
/
a
p
p
1
0
1
9
6
6
7
6
.
Evaluation Warning : The document was created with Spire.PDF for Python.