I
nte
rna
t
io
na
l J
o
urna
l o
f
E
lect
rica
l a
nd
Co
m
pu
t
er
E
ng
ineering
(
I
J
E
CE
)
Vo
l.
1
6
,
No
.
2
,
A
p
r
il
20
2
6
,
p
p
.
964
~
972
I
SS
N:
2088
-
8
7
0
8
,
DOI
: 1
0
.
1
1
5
9
1
/ijece.
v
1
6
i
2
.
pp
9
6
4
-
9
7
2
964
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
ec
e.
ia
esco
r
e.
co
m
H
y
perpa
ra
mete
r
tuning o
f
M
o
bile
NetV2 on
forest a
nd lan
d f
i
re
sev
erity clas
sifica
tion
Ass
a
d H
ida
y
a
t
1
,
I
m
a
s
Su
k
a
e
s
ih Sita
ng
g
a
ng
1
,
L
a
ila
n Sy
a
ufina
2
1
S
c
h
o
o
l
o
f
D
a
t
a
S
c
i
e
n
c
e
,
M
a
t
h
e
ma
t
i
c
s
a
n
d
I
n
f
o
r
m
a
t
i
c
s
,
I
P
B
U
n
i
v
e
r
s
i
t
y
,
B
o
g
o
r
,
I
n
d
o
n
e
s
i
a
2
D
e
p
a
r
t
me
n
t
o
f
S
i
l
v
i
c
u
l
t
u
r
e
,
F
a
c
u
l
t
y
o
f
F
o
r
e
st
r
y
a
n
d
En
v
i
r
o
n
me
n
t
,
I
P
B
U
n
i
v
e
r
si
t
y
,
B
o
g
o
r
,
I
n
d
o
n
e
s
i
a
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Oct
1
0
,
2
0
2
4
R
ev
is
ed
Dec
4
,
2
0
2
5
Acc
ep
ted
J
an
1
5
,
2
0
2
6
F
o
re
st
a
n
d
lan
d
fires
p
o
se
sig
n
ifi
c
a
n
t
e
n
v
ir
o
n
m
e
n
tal
c
h
a
ll
e
n
g
e
s,
c
a
u
sin
g
e
c
o
n
o
m
ic
a
n
d
e
c
o
l
o
g
ica
l
d
a
m
a
g
e
d
e
p
e
n
d
i
n
g
o
n
t
h
e
ir
se
v
e
rit
y
.
Th
is
stu
d
y
p
ro
p
o
se
s
a
d
e
e
p
lea
rn
in
g
-
b
a
se
d
c
las
sifica
ti
o
n
m
o
d
e
l
to
a
ss
e
ss
fire
se
v
e
rit
y
u
sin
g
t
h
e
M
o
b
il
e
Ne
tV2
a
rc
h
it
e
c
t
u
re
.
A
d
a
tas
e
t
o
f
5
6
0
p
o
st
-
fi
re
ima
g
e
s
wa
s
c
a
teg
o
rize
d
i
n
to
fiv
e
se
v
e
rit
y
le
v
e
ls,
wit
h
d
a
tas
e
t
p
re
p
r
o
c
e
ss
in
g
in
v
o
l
v
in
g
re
siz
in
g
,
re
sc
a
li
n
g
,
a
n
d
ima
g
e
a
u
g
m
e
n
tatio
n
.
To
e
n
h
a
n
c
e
m
o
d
e
l
p
e
rfo
rm
a
n
c
e
,
K
-
m
e
a
n
s
c
lu
ste
rin
g
wa
s
a
p
p
li
e
d
fo
r
b
a
lan
c
e
d
d
a
ta
d
i
strib
u
ti
o
n
a
c
ro
s
s
c
las
se
s.
Th
e
m
o
d
e
l
wa
s
train
e
d
u
si
n
g
g
rid
se
a
rc
h
f
o
r
h
y
p
e
r
p
a
ra
m
e
ter
tu
n
i
n
g
,
with
th
e
o
p
ti
m
a
l
c
o
m
b
in
a
t
io
n
b
e
in
g
a
b
a
tch
siz
e
o
f
8
,
lea
rn
i
n
g
ra
te
o
f
0
.
0
0
0
1
,
a
n
d
d
r
o
p
o
u
t
o
f
0
.
3
.
Tr
a
in
in
g
wa
s
c
o
n
d
u
c
ted
in
5
0
e
p
o
c
h
s,
a
n
d
e
v
a
lu
a
ti
o
n
u
si
n
g
t
h
e
c
o
n
f
u
sio
n
m
a
tri
x
d
e
m
o
n
stra
ted
a
n
a
c
c
u
ra
c
y
o
f
8
5
%
,
p
re
c
isio
n
o
f
8
6
%
,
a
n
d
re
c
a
ll
o
f
8
1
%
.
Th
e
re
su
l
ts
in
d
ica
te
th
a
t
M
o
b
il
e
Ne
tV
2
e
ffe
c
ti
v
e
ly
c
las
sifies
p
o
st
-
fire
se
v
e
rit
y
le
v
e
ls,
o
ffe
ri
n
g
a
re
li
a
b
le
to
o
l
fo
r
p
o
st
-
d
isa
ste
r
a
ss
e
ss
m
e
n
t.
Th
is
st
u
d
y
h
i
g
h
li
g
h
ts
th
e
sig
n
ifi
c
a
n
c
e
o
f
d
a
tas
e
t
p
re
p
ro
c
e
ss
in
g
a
n
d
h
y
p
e
rp
a
ra
m
e
ter
tu
n
i
n
g
in
imp
ro
v
i
n
g
m
o
d
e
l
a
c
c
u
ra
c
y
.
F
u
tu
re
re
se
a
rc
h
sh
o
u
ld
e
x
p
lo
re
a
lt
e
rn
a
ti
v
e
a
rc
h
it
e
c
tu
re
s
a
n
d
e
x
p
a
n
d
th
e
d
a
tas
e
t
to
e
n
h
a
n
c
e
m
o
d
e
l
g
e
n
e
ra
li
z
a
ti
o
n
.
T
h
e
se
fin
d
i
n
g
s
c
a
n
a
id
a
u
th
o
r
it
ies
in
a
ss
e
ss
in
g
fire
im
p
a
c
t,
su
p
p
o
r
ti
n
g
m
it
i
g
a
ti
o
n
stra
teg
ies
,
a
n
d
i
m
p
ro
v
i
n
g
p
o
st
-
fire l
a
n
d
m
a
n
a
g
e
m
e
n
t.
K
ey
w
o
r
d
s
:
Fo
r
est an
d
lan
d
f
ir
es
Gr
id
s
ea
r
ch
Hy
p
er
p
ar
a
m
eter
tu
n
in
g
Mo
b
ileNetV2
Sev
er
ity
lev
el
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
:
I
m
as Su
k
ae
s
ih
Sit
an
g
g
an
g
Sch
o
o
l o
f
Data
Scien
ce
,
Ma
t
h
em
atics a
n
d
I
n
f
o
r
m
atics
,
I
PB
Un
iv
er
s
ity
1
6
6
8
0
B
o
g
o
r
,
I
n
d
o
n
esia
E
m
ail:
im
as.si
tan
g
g
an
g
@
ap
p
s
.
ip
b
.
ac
.
id
1.
I
NT
RO
D
UCT
I
O
N
Fo
r
est
an
d
lan
d
f
ir
es
ar
e
e
v
en
ts
ca
u
s
ed
b
y
h
u
m
an
o
r
n
atu
r
al
ac
tiv
ities
,
ch
ar
ac
ter
ized
b
y
w
id
esp
r
ea
d
f
ir
e
an
d
th
e
b
u
r
n
in
g
o
f
f
o
r
ests
an
d
lan
d
.
Fo
r
est f
ir
es is
o
n
e
o
f
th
e
en
v
ir
o
n
m
en
tal
is
s
u
es th
at
h
as b
o
th
ec
o
n
o
m
i
c
an
d
ec
o
lo
g
ical
d
etr
im
en
tal
i
m
p
ac
ts
[
1
]
.
Fo
r
est
f
ir
es
ar
e
c
o
m
p
lex
p
h
e
n
o
m
e
n
a
in
f
lu
en
ce
d
b
y
th
e
in
ter
ac
tio
n
o
f
n
atu
r
al
an
d
an
th
r
o
p
o
g
en
ic
f
ac
to
r
s
.
T
h
e
m
ain
d
r
iv
er
s
o
f
f
o
r
est
f
ir
es
g
en
er
ally
in
clu
d
e
c
lim
ate,
v
eg
etatio
n
,
an
d
v
ar
y
i
n
g
to
p
o
g
r
a
p
h
y
[
2
]
.
Peat
an
d
m
in
er
al
s
o
ils
ar
e
two
ch
ar
ac
ter
is
tic
ty
p
es
o
f
f
o
r
est
f
ir
e
s
ites
,
th
e
v
ar
iatio
n
b
etwe
en
th
e
two
lies
in
th
eir
o
r
g
a
n
ic
m
atter
co
n
ten
t,
with
m
in
er
al
s
o
ils
h
av
in
g
a
lo
w
co
n
ten
t,
w
h
ile
p
ea
t
s
o
ils
h
av
e
a
h
ig
h
co
n
ten
t
,
f
o
r
est
an
d
lan
d
f
ir
es
th
at
o
c
cu
r
in
p
ea
t
s
o
il
d
if
f
er
f
r
o
m
th
o
s
e
in
m
in
er
al
s
o
il
b
ec
au
s
e
f
ir
es
i
n
p
ea
t
s
o
il
ar
e
d
o
m
i
n
ated
b
y
g
r
o
u
n
d
f
i
r
es.
E
v
en
t
h
o
u
g
h
t
h
e
f
ir
e
th
at
h
a
s
b
u
r
n
e
d
tr
ee
s
an
d
g
r
ass
es
h
as
b
ee
n
ex
tin
g
u
is
h
ed
,
th
e
f
ir
e
m
ay
s
till
b
e
b
u
r
n
i
n
g
d
ee
p
b
en
ea
th
th
e
s
u
r
f
ac
e
o
f
th
e
s
o
il
[
3
]
.
T
h
e
in
cr
ea
s
e
in
th
e
n
u
m
b
er
an
d
in
ten
s
ity
o
f
f
o
r
est
f
ir
es
h
as
o
cc
u
r
r
ed
wo
r
l
d
wid
e
o
v
er
th
e
p
ast
d
ec
ad
e,
r
aisi
n
g
p
u
b
lic
co
n
ce
r
n
s
a
b
o
u
t
th
e
p
o
te
n
tial
f
u
tu
r
e
im
p
ac
ts
[
4
]
.
T
h
e
e
f
f
ec
ts
o
f
g
lo
b
al
war
m
in
g
s
u
g
g
est
th
at
th
e
n
u
m
b
er
an
d
ar
ea
o
f
f
ir
es
wo
r
ld
wid
e
a
r
e
ex
p
ec
te
d
to
co
n
tin
u
e
[
5
]
.
F
o
r
est
an
d
lan
d
f
ir
e
also
a
f
f
ec
t
s
I
n
d
o
n
esia,
wh
ich
h
as
v
ast
f
o
r
ests
,
an
d
th
e
ex
ten
t
o
f
I
n
d
o
n
esia'
s
f
o
r
ests
co
n
tin
u
es
to
d
ec
lin
e
ea
ch
y
ea
r
.
B
etwe
en
2
0
1
2
a
n
d
2
0
2
2
,
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
-
8
7
0
8
Hyp
erp
a
r
a
mete
r
tu
n
in
g
o
f Mo
b
ileN
etV
2
o
n
fo
r
est a
n
d
la
n
d
…
(
A
s
s
a
d
Hid
a
ya
t
)
965
f
o
r
est
f
ir
es
in
I
n
d
o
n
esia
h
av
e
r
ea
ch
ed
4
4
2
.
8
2
2
,
5
1
k
m
2
[
6
]
,
I
n
d
o
n
esia
is
u
n
d
er
p
r
ess
u
r
e
f
r
o
m
v
ar
i
o
u
s
p
ar
ties
to
r
ed
u
ce
f
o
r
est
f
ir
es,
an
d
ef
f
o
r
ts
h
av
e
b
ee
n
m
ad
e
to
c
o
n
tr
o
l
it
[
7
]
.
C
o
n
tr
o
l
o
f
f
o
r
est
f
ir
es
in
ar
ea
s
th
at
h
av
e
alr
ea
d
y
b
u
r
n
ed
n
ee
d
s
to
b
e
o
b
s
er
v
ed
s
o
th
at
th
e
d
ata
o
b
tain
ed
ca
n
b
e
u
s
ed
to
m
ea
s
u
r
e
th
e
s
ev
er
ity
o
f
th
e
f
ir
e
[
8
]
.
T
h
e
s
ev
er
ity
lev
el
o
f
a
f
o
r
est f
ir
e
is
d
ef
in
ed
as a
ter
m
th
at
d
escr
ib
es th
e
ex
ten
t to
wh
ich
th
e
f
ir
e
alter
s
ec
o
s
y
s
tem
co
m
p
o
n
en
ts
an
d
ca
n
b
e
u
s
ed
to
ex
p
lain
th
e
im
p
ac
t o
f
th
e
f
ir
e
[
9
]
.
Fo
r
est f
ir
es wi
ll p
r
o
d
u
ce
s
ev
er
ity
lev
els
th
at
d
ep
en
d
o
n
th
e
in
ter
ac
tio
n
b
etwe
en
th
e
ar
ea
b
u
r
n
ed
,
f
u
el
lo
ad
,
tim
in
g
,
in
ten
s
ity
,
d
u
r
atio
n
o
f
b
u
r
n
in
g
,
s
o
il
tex
tu
r
e
an
d
m
o
is
tu
r
e,
to
p
o
g
r
ap
h
y
,
v
eg
etatio
n
ty
p
e,
f
ir
e
clim
ate,
s
o
il
wate
r
s
t
atu
s
,
an
d
tim
e
s
in
ce
th
e
last
f
ir
e
[
1
0
]
.
T
h
e
s
ev
er
ity
lev
el
o
f
a
f
o
r
est
an
d
lan
d
f
ir
e
d
escr
ib
es
th
e
ec
o
s
y
s
tem
'
s
r
es
p
o
n
s
e
to
f
ir
e
an
d
is
u
s
ed
to
ex
p
lain
th
e
im
p
ac
t
o
f
f
o
r
est
a
n
d
lan
d
f
ir
e.
T
h
e
g
en
er
al
class
if
icatio
n
o
f
f
o
r
est
a
n
d
lan
d
f
ir
e
s
ev
er
ity
lev
els
is
b
ased
o
n
s
o
il
co
n
d
itio
n
s
an
d
th
eir
p
r
o
p
er
ties
in
th
e
b
u
r
n
e
d
ar
ea
.
Stu
d
y
in
g
an
d
m
e
asu
r
in
g
th
e
s
ev
er
ity
lev
el
o
f
f
o
r
est
an
d
lan
d
f
ir
es
is
im
p
o
r
tan
t
b
ec
au
s
e
it
ca
n
s
er
v
e
as
a
b
asi
s
f
o
r
in
f
o
r
m
ati
o
n
in
f
ir
e
r
ec
o
v
e
r
y
p
lan
n
in
g
,
f
o
r
est
co
n
s
er
v
atio
n
,
an
d
law
e
n
f
o
r
ce
m
en
t
[
1
1
]
.
T
h
e
s
ev
er
ity
lev
el
o
f
f
o
r
est
an
d
lan
d
f
ir
es
ca
n
b
e
m
ea
s
u
r
ed
u
s
in
g
im
a
g
e
d
ata
o
b
tain
ed
f
r
o
m
f
ield
o
b
s
er
v
atio
n
s
[
8
]
.
T
h
e
p
r
o
ce
s
s
in
g
o
f
im
ag
e
d
ata
u
s
ed
in
th
e
ca
s
e
o
f
f
o
r
est
an
d
lan
d
f
ir
es
aim
s
to
p
r
o
v
id
e
s
u
p
p
o
r
tin
g
in
f
o
r
m
atio
n
ab
o
u
t
th
e
r
esp
o
n
s
e
to
f
o
r
est
an
d
lan
d
f
ir
e
d
is
aster
s
,
en
ab
lin
g
d
ec
is
io
n
-
m
ak
er
s
to
r
esp
o
n
d
q
u
ick
ly
to
f
o
r
est
an
d
lan
d
f
ir
e
in
cid
e
n
ts
[
1
2
]
.
I
m
ag
e
d
ata
p
r
o
ce
s
s
in
g
tech
n
iq
u
es
f
o
r
p
atter
n
r
ec
o
g
n
itio
n
h
a
v
e
b
ee
n
ex
p
lo
r
ed
i
n
v
ar
io
u
s
d
o
m
ain
s
.
Fo
r
ex
am
p
le,
Ah
m
ed
et
a
l
.
[
1
3
]
p
r
o
p
o
s
ed
a
s
y
s
tem
to
im
p
r
o
v
e
f
ea
tu
r
e
ex
tr
ac
tio
n
in
im
ag
e
class
if
icatio
n
task
s
.
T
h
is
m
eth
o
d
h
ig
h
lig
h
ts
th
e
im
p
o
r
tan
ce
o
f
ef
f
ec
tiv
e
f
ea
tu
r
e
ex
tr
ac
tio
n
in
im
ag
e
-
b
ased
class
if
icatio
n
p
r
o
b
lem
s
.
I
m
ag
e
d
ata
p
r
o
ce
s
s
in
g
to
m
ea
s
u
r
e
th
e
s
ev
er
ity
lev
el
o
f
f
o
r
est
an
d
lan
d
f
ir
es
ca
n
p
r
o
v
id
e
v
alu
ab
le
in
f
o
r
m
atio
n
in
u
n
d
er
s
tan
d
i
n
g
an
d
ef
f
ec
tiv
ely
ad
d
r
ess
in
g
th
e
f
ir
es.
W
ith
th
e
co
n
tin
u
o
u
s
ad
v
a
n
ce
m
en
t
o
f
te
ch
n
o
lo
g
y
,
th
e
im
ag
e
d
ata
co
ll
ec
ted
ca
n
b
e
p
r
o
ce
s
s
ed
u
s
in
g
ce
r
tain
m
eth
o
d
s
to
p
r
o
v
id
e
in
s
ig
h
ts
in
an
aly
zin
g
f
o
r
est
an
d
lan
d
f
ir
e
s
ev
er
ity
lev
els
o
n
e
s
u
ch
m
eth
o
d
is
d
ee
p
lear
n
in
g
(
DL
)
[
1
4
]
.
DL
ca
n
b
e
c
o
n
s
id
er
ed
a
s
u
itab
le
m
eth
o
d
o
lo
g
y
f
o
r
m
o
d
el
lin
g
th
e
co
m
p
lex
i
n
ter
ac
tio
n
s
o
f
v
ar
iab
les
th
at
f
r
eq
u
e
n
tly
o
cc
u
r
i
n
E
ar
t
h
s
y
s
tem
p
r
o
b
lem
s
,
p
a
r
ticu
lar
ly
f
o
r
e
s
t
f
ir
es
[
1
5
]
.
DL
h
as
tr
an
s
f
o
r
m
ed
co
m
p
u
ter
v
is
io
n
b
y
allo
win
g
co
m
p
u
ter
s
to
lea
r
n
f
r
o
m
d
ata,
id
en
tif
y
p
atter
n
s
,
an
d
class
if
y
o
b
jects
with
r
em
ar
k
ab
le
ac
cu
r
ac
y
.
T
h
is
en
ab
les
th
e
cr
ea
tio
n
o
f
s
o
p
h
is
ticated
co
m
p
u
ter
v
is
io
n
alg
o
r
ith
m
s
an
d
ap
p
licatio
n
s
th
at
ca
n
ac
cu
r
ately
id
en
tify
o
b
jects
in
im
ag
es
[
1
6
]
.
T
h
e
d
ee
p
lear
n
i
n
g
tech
n
iq
u
e
co
m
m
o
n
l
y
u
s
ed
f
o
r
im
ag
e
d
ata
is
th
e
co
n
v
o
l
u
tio
n
al
n
e
u
r
al
n
etwo
r
k
(
C
NN)
.
C
NN
u
tili
ze
s
co
n
v
o
lu
tio
n
al
lay
er
s
th
at
ca
n
au
to
m
atica
lly
ex
tr
ac
t
f
ea
tu
r
es
f
r
o
m
im
ag
es,
allo
win
g
th
e
m
o
d
el
to
r
ec
o
g
n
ize
p
atter
n
s
an
d
o
b
jects
with
o
u
t
th
e
n
ee
d
f
o
r
co
m
p
le
x
p
r
ep
r
o
ce
s
s
in
g
[
1
7
]
.
C
NN
h
as
two
im
p
o
r
tan
t
b
lo
c
k
s
:
f
ea
tu
r
e
lear
n
in
g
an
d
class
if
icatio
n
[
1
8
]
.
C
NN
ca
n
b
e
u
s
ed
to
id
en
tify
an
d
class
if
y
ar
ea
s
af
f
ec
ted
b
y
f
ir
es;
b
y
lev
er
ag
i
n
g
im
ag
e
d
ata,
C
NN
ca
n
ev
alu
ate
th
e
im
p
ac
t
o
f
th
e
f
ir
e
o
n
th
e
ec
o
s
y
s
tem
[
1
9
]
.
C
NN
also
h
as
s
ev
er
al
ar
ch
itectu
r
es,
o
n
e
o
f
wh
ich
is
Mo
b
ileNetV2
.
T
h
e
Mo
b
ileN
etV2
ar
ch
itectu
r
e
is
d
esig
n
ed
f
o
r
co
m
p
u
tatio
n
al
e
f
f
icien
cy
t
h
r
o
u
g
h
th
e
u
s
e
o
f
Dep
th
wis
e
s
ep
ar
ab
le
co
n
v
o
l
u
tio
n
s
(
DSC
)
an
d
is
co
n
s
id
er
ed
s
tate
-
of
-
th
e
-
ar
t,
wi
th
s
ev
er
al
ad
d
itio
n
al
lay
er
s
f
o
l
lo
win
g
it
f
o
r
s
p
ec
if
ic
class
if
icatio
n
task
s
[
2
0
]
.
T
o
f
u
r
th
er
o
p
tim
ize
its
p
er
f
o
r
m
a
n
ce
,
th
is
s
tu
d
y
im
p
lem
e
n
ts
d
ata
s
et
clu
s
ter
in
g
u
s
in
g
K
-
m
ea
n
s
b
ef
o
r
e
tr
ain
in
g
a
n
d
u
s
es
p
r
in
cip
al
co
m
p
o
n
en
t
a
n
aly
s
is
(
PC
A)
to
r
ed
u
ce
th
e
f
ea
tu
r
e
d
im
e
n
s
io
n
ality
,
th
er
eb
y
r
e
d
u
cin
g
th
e
co
m
p
u
tatio
n
al
b
u
r
d
en
,
a
n
d
ap
p
lies
h
y
p
er
p
a
r
am
eter
tu
n
in
g
u
s
in
g
g
r
id
s
ea
r
ch
.
Hy
p
er
p
ar
a
m
eter
tu
n
in
g
in
Mo
b
ileNetV2
is
th
e
p
r
o
ce
s
s
o
f
ad
ju
s
tin
g
v
alu
es
th
at
ar
e
n
o
t
au
t
o
m
atica
lly
lear
n
ed
b
y
th
e
m
o
d
el,
wh
ich
p
lay
s
an
im
p
o
r
tan
t
r
o
l
e
in
m
o
d
el
p
er
f
o
r
m
an
ce
a
n
d
g
en
er
aliza
tio
n
.
Hy
p
er
p
ar
a
m
eter
s
ar
e
p
ar
am
et
er
s
d
ef
in
ed
o
u
ts
id
e
th
e
m
o
d
el
b
ein
g
tr
ain
ed
.
T
h
is
in
clu
d
es
f
ac
to
r
s
s
u
ch
as
th
e
q
u
an
tity
f
ilter
s
in
c
o
n
v
o
lu
tio
n
al
lay
er
,
n
u
m
b
er
o
f
e
p
o
ch
s
,
b
atch
s
ize,
v
ar
io
u
s
o
t
h
er
d
im
e
n
s
io
n
s
,
an
d
k
er
n
el
s
ize
[
2
1
]
.
T
h
e
aim
o
f
h
y
p
e
r
p
ar
am
eter
tu
n
in
g
is
to
id
en
tify
th
e
b
est
co
m
b
in
atio
n
o
f
h
y
p
er
p
ar
am
eter
v
alu
es,
allo
win
g
th
e
C
NN
m
o
d
el
to
e
x
ce
l
in
lear
n
in
g
th
e
p
atter
n
s
in
th
e
tr
ain
in
g
d
ata
wh
ile
p
r
ev
en
tin
g
o
v
er
f
itti
n
g
o
r
u
n
d
er
f
itti
n
g
[
2
2
]
.
A
tech
n
i
q
u
e
th
at
ca
n
b
e
u
s
ed
f
o
r
h
y
p
er
p
ar
am
eter
tu
n
i
n
g
is
g
r
id
s
ea
r
ch
.
T
h
is
tech
n
iq
u
e
in
v
o
lv
es
cr
ea
tin
g
a
g
r
id
o
f
p
ar
am
eter
s
to
b
e
test
ed
,
an
d
th
en
th
e
m
o
d
el
will
b
e
tr
ain
ed
an
d
v
alid
ated
f
o
r
ea
c
h
co
m
b
in
atio
n
.
T
h
e
r
esu
lts
o
f
ea
ch
co
m
b
in
atio
n
ar
e
m
ea
s
u
r
ed
b
ased
o
n
t
h
e
m
o
d
el'
s
p
e
r
f
o
r
m
a
n
ce
,
s
u
c
h
as
ac
cu
r
ac
y
o
r
lo
s
s
,
an
d
t
h
e
b
est
co
m
b
in
atio
n
is
s
elec
ted
to
b
e
ap
p
lied
to
th
e
d
ata
[
2
3
]
.
Sev
er
al
s
tu
d
ies
h
av
e
b
ee
n
co
n
d
u
cte
d
o
n
im
ag
e
-
b
ased
f
ir
e
s
ev
er
ity
.
On
e
o
f
th
em
is
a
s
tu
d
y
b
y
Ar
r
af
i
et
a
l.
[
2
4
]
f
o
c
u
s
ed
o
n
m
ap
p
in
g
t
h
e
s
ev
er
ity
o
f
f
o
r
es
t
an
d
lan
d
f
ir
es
u
s
in
g
t
h
e
n
o
r
m
alize
d
b
u
r
n
r
atio
(
NB
R
)
alg
o
r
ith
m
o
n
L
an
d
s
at
8
I
m
a
g
er
y
.
T
h
is
ap
p
r
o
ac
h
p
r
o
v
id
es
b
r
o
ad
in
s
ig
h
ts
b
u
t
o
f
t
en
h
as
d
if
f
icu
lty
in
ca
p
tu
r
in
g
d
etails
o
n
th
e
g
r
o
u
n
d
s
u
r
f
ac
e.
Pre
v
io
u
s
s
tu
d
ies
h
av
e
f
o
cu
s
ed
m
o
r
e
o
n
g
e
n
er
al
f
ir
e
d
etec
tio
n
o
r
s
ev
er
ity
esti
m
atio
n
u
s
in
g
s
atel
lite
im
ag
er
y
,
b
u
t
f
ew
h
av
e
d
is
cu
s
s
ed
f
ir
e
s
ev
er
ity
clas
s
if
icat
io
n
o
n
th
e
g
r
o
u
n
d
s
u
r
f
ac
e.
T
h
is
s
tu
d
y
aim
s
to
ad
d
r
ess
th
is
g
ap
b
y
d
ev
elo
p
in
g
a
Mo
b
ileNetV2
-
b
ased
m
o
d
el
to
class
if
y
p
o
s
t
-
f
ir
e
s
ev
er
ity
u
s
in
g
p
o
s
t
-
f
ir
e
ar
e
a
im
ag
er
y
.
I
n
a
d
d
i
t
i
o
n
,
t
h
i
s
c
l
as
s
i
f
i
c
a
ti
o
n
m
o
d
e
l
i
s
o
p
t
i
m
i
z
e
d
t
h
r
o
u
g
h
h
y
p
e
r
p
a
r
a
m
e
t
e
r
t
u
n
i
n
g
a
n
d
d
at
a
s
et
c
l
u
s
t
e
r
i
n
g
t
o
i
m
p
r
o
v
e
ac
c
u
r
a
c
y
a
n
d
g
e
n
e
r
a
l
i
za
t
i
o
n
i
n
p
r
e
d
i
c
t
i
n
g
f
o
r
e
s
t
a
n
d
l
a
n
d
f
i
r
e
s
e
v
e
r
it
y
c
l
as
s
e
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
1
6
,
No
.
2
,
Ap
r
il
20
2
6
:
9
6
4
-
972
966
2.
M
E
T
H
O
D
T
h
is
r
esear
ch
u
s
es
a
m
eth
o
d
o
lo
g
ical
ap
p
r
o
ac
h
th
at
f
o
cu
s
es
o
n
m
o
d
el
b
u
ild
in
g
u
s
in
g
Mo
b
ileNetV2
.
T
h
e
ap
p
lie
d
m
eth
o
d
o
lo
g
y
f
o
ll
o
ws
a
s
tr
u
ctu
r
ed
an
d
s
y
s
tem
atic
s
et
o
f
s
tag
es
to
c
r
ea
te
an
i
m
ag
e
-
b
ased
lan
d
an
d
f
o
r
est
f
ir
e
s
ev
er
ity
class
if
icat
io
n
m
o
d
el.
T
h
e
r
esear
ch
p
r
o
ce
s
s
i
s
d
iv
id
ed
in
to
s
ev
er
al
m
ain
p
h
ases
,
d
ata
co
llectio
n
,
d
ataset
id
en
tific
atio
n
,
p
r
e
-
p
r
o
ce
s
s
in
g
,
m
o
d
el
tr
a
in
in
g
,
an
d
m
o
d
el
ev
alu
atio
n
.
T
h
is
m
eth
o
d
o
lo
g
y
aim
s
to
p
r
o
d
u
ce
a
tec
h
n
o
lo
g
i
ca
l
s
o
lu
tio
n
f
o
r
f
o
r
est
an
d
la
n
d
f
ir
e
s
ev
er
ity
class
if
icatio
n
.
T
h
e
p
h
ases
ca
n
b
e
s
ee
n
in
Fig
u
r
e
1
.
Fig
u
r
e
1
.
R
esear
ch
s
tep
s
2
.
1
.
Da
t
a
c
o
llect
io
n
T
h
e
d
ataset
u
s
ed
in
th
is
r
ese
ar
ch
c
o
n
s
is
ts
o
f
im
ag
es
o
f
a
r
ea
s
af
ter
f
o
r
est
a
n
d
la
n
d
f
ir
e
h
an
d
lin
g
o
b
tain
ed
f
r
o
m
th
e
Min
is
tr
y
o
f
E
n
v
ir
o
n
m
en
t
an
d
Fo
r
es
tr
y
an
d
f
r
o
m
o
p
en
-
s
o
u
r
ce
m
ed
ia
s
u
ch
as
th
e
s
h
u
tter
s
to
ck
.
co
m
.
T
h
is
d
ataset
en
co
m
p
ass
es
a
v
ar
iety
o
f
co
n
d
itio
n
s
in
p
o
s
t
-
f
ir
e
ar
ea
s
,
to
talin
g
5
6
0
im
ag
es
f
r
o
m
v
a
r
io
u
s
f
o
r
est
f
ir
e
ev
e
n
ts
in
I
n
d
o
n
esia.
No
s
en
s
itiv
e
in
f
o
r
m
atio
n
was
co
n
tain
e
d
with
in
th
e
d
ataset.
E
x
p
er
t
v
alid
atio
n
was
c
o
n
d
u
c
ted
with
in
f
o
r
m
ed
co
n
s
en
t
f
r
o
m
p
a
r
ticip
atin
g
f
o
r
est
an
d
l
an
d
f
ir
e
s
p
ec
ialis
ts
.
E
th
ical
ap
p
r
o
v
al
was n
o
t r
eq
u
i
r
ed
,
as th
e
s
tu
d
y
d
id
n
o
t in
v
o
l
v
e
h
u
m
an
s
u
b
jects o
r
p
er
s
o
n
al
d
ata.
Alth
o
u
g
h
th
e
s
ize
o
f
th
is
d
ataset
is
r
elati
v
ely
s
m
all
f
o
r
d
ee
p
lear
n
i
n
g
ap
p
licatio
n
s
,
ex
ten
s
iv
e
d
ata
au
g
m
en
tatio
n
ca
n
o
v
er
co
m
e
th
is
lim
itatio
n
b
y
a
r
tific
ially
ex
p
an
d
in
g
th
e
d
ataset.
T
h
e
au
g
m
en
tatio
n
i
n
clu
d
es
h
o
r
izo
n
tal
f
lip
p
in
g
,
zo
o
m
in
g
,
r
o
tatio
n
,
a
n
d
co
n
tr
ast
ad
ju
s
tm
en
t
to
i
n
tr
o
d
u
ce
v
ar
iatio
n
a
n
d
im
p
r
o
v
e
g
en
er
aliza
tio
n
ab
ilit
y
.
An
o
v
er
v
iew
o
f
th
e
d
ataset
is
p
r
esen
ted
in
Fig
u
r
e
2
,
wh
er
e
Fig
u
r
e
2
(
a)
s
h
o
ws
im
ag
es
ca
teg
o
r
ized
as
lig
h
t
s
ev
er
ity
,
Fig
u
r
e
2
(
b
)
r
ep
r
esen
t
s
m
o
d
er
ate
s
ev
er
ity
,
a
n
d
Fig
u
r
e
2
(
c)
illu
s
tr
ates sev
er
e
s
ev
er
ity
.
(
a)
(
b
)
(
c)
Fig
u
r
e
2
.
Ov
e
r
v
iew
o
f
r
esear
c
h
d
atasets
(
a)
lig
h
t
,
(
b
)
m
o
d
er
a
te
,
an
d
(
c
)
s
ev
er
e
2
.
2
.
I
dentif
ica
t
io
n da
t
a
s
et
T
h
e
in
itial
d
ataset
co
llected
in
clu
d
ed
v
ar
io
u
s
im
ag
es
o
f
th
e
p
o
s
t
-
f
ir
e
m
an
ag
e
m
en
t
ar
ea
a
n
d
h
a
d
n
o
t
b
ee
n
s
eg
r
eg
ated
b
y
class
.
T
h
is
u
n
s
eg
r
eg
ated
d
ataset
was
ca
r
ef
u
lly
s
elec
ted
u
s
in
g
a
tech
n
iq
u
e
to
en
s
u
r
e
th
a
t
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
-
8
7
0
8
Hyp
erp
a
r
a
mete
r
tu
n
in
g
o
f Mo
b
ileN
etV
2
o
n
fo
r
est a
n
d
la
n
d
…
(
A
s
s
a
d
Hid
a
ya
t
)
967
th
e
m
o
d
el
co
u
ld
b
e
ef
f
ec
tiv
el
y
tr
ain
e
d
to
class
if
y
d
if
f
er
en
t
s
ev
er
ities
o
f
f
o
r
est
f
ir
es.
E
ac
h
o
f
th
ese
ca
teg
o
r
ies
o
r
class
es
d
escr
ib
es
th
e
im
p
ac
t
o
f
th
e
f
ir
e
o
n
th
e
en
v
ir
o
n
m
en
t
an
d
th
e
lev
el
o
f
d
am
ag
e
c
au
s
ed
.
T
h
e
f
lo
w
o
f
d
ataset
cr
ea
tio
n
ca
n
b
e
s
ee
n
in
Fig
u
r
e
3
.
Fig
u
r
e
3
is
a
p
r
o
ce
s
s
p
ar
t
o
f
th
e
m
eth
o
d
o
lo
g
y
in
Fig
u
r
e
1
,
n
a
m
ely
d
ataset
id
en
tific
atio
n
.
T
h
e
p
r
o
ce
s
s
s
tar
ts
b
y
in
itializin
g
th
e
d
ataset
u
s
in
g
th
e
M
o
b
ileNetV2
m
o
d
el
p
r
e
-
tr
ain
ed
with
weig
h
ts
f
r
o
m
I
m
ag
eNe
t.
T
h
e
im
ag
es
ar
e
lo
ad
e
d
f
r
o
m
a
f
o
l
d
er
,
r
esized
t
o
2
2
4
×2
2
4
,
an
d
co
n
v
er
ted
to
a
n
ar
r
ay
.
T
h
e
f
e
atu
r
es
o
f
th
e
im
ag
es
wer
e
ex
tr
ac
ted
u
s
in
g
th
e
M
o
b
ileNetV2
m
o
d
el,
w
h
ich
p
r
o
v
id
es
a
v
ec
to
r
r
ep
r
esen
tatio
n
o
f
th
e
f
ea
tu
r
es.
C
lu
s
ter
in
g
is
p
er
f
o
r
m
e
d
u
s
in
g
PC
A
to
r
ed
u
ce
t
h
e
d
im
e
n
s
io
n
ality
o
f
th
e
f
ea
t
u
r
es,
th
u
s
f
a
cilitatin
g
clu
s
ter
in
g
an
d
r
ed
u
cin
g
th
e
c
o
m
p
u
tatio
n
al
b
u
r
d
en
.
T
h
is
p
r
o
ce
s
s
is
f
o
llo
wed
b
y
th
e
K
-
m
ea
n
s
alg
o
r
it
h
m
th
at
g
r
o
u
p
s
th
e
im
ag
es
in
to
f
iv
e
clu
s
ter
s
.
Ho
wev
er
,
v
is
u
aliza
tio
n
o
f
th
e
c
lu
s
ter
r
esu
lts
s
h
o
wed
o
v
er
lap
b
etwe
en
clu
s
ter
s
,
in
d
icatin
g
th
at
th
e
s
ep
ar
atio
n
was
n
o
t
en
tire
ly
clea
r
.
T
h
e
cl
u
s
ter
r
esu
lts
wer
e
ev
alu
ated
u
s
in
g
th
r
ee
m
etr
ics:
th
e
Dav
ies
-
B
o
u
ld
in
in
d
ex
,
th
e
Sil
h
o
u
ette
s
co
r
e
,
an
d
th
e
C
ali
n
s
k
i
-
Har
ab
asz
in
d
ex
.
T
h
e
Da
v
ies
-
B
o
u
ld
in
in
d
ex
y
ield
s
a
v
alu
e
o
f
1
,
1
1
3
,
in
d
i
ca
tin
g
a
lack
o
f
s
ep
ar
atio
n
b
etwe
en
clu
s
ter
s
an
d
s
u
g
g
esti
n
g
th
at
th
e
clu
s
ter
s
o
v
er
lap
s
o
m
ewh
at.
T
h
e
Sil
h
o
u
ette
s
co
r
e
y
ield
ed
a
v
alu
e
o
f
0
,
2
8
1
,
in
d
icatin
g
th
at
th
e
clu
s
t
er
in
g
was
less
th
an
o
p
tim
al
an
d
m
a
n
y
p
o
in
ts
wer
e
b
etwe
en
clu
s
ter
s
,
in
d
icatin
g
o
v
er
lap
.
T
h
e
C
alin
s
k
i
-
Har
a
b
asz
in
d
ex
g
a
v
e
a
v
alu
e
o
f
2
6
8
,
6
0
2
,
in
d
icatin
g
th
at
th
er
e
is
f
air
ly
g
o
o
d
cl
u
s
ter
s
ep
ar
atio
n
,
alth
o
u
g
h
it
d
o
es
n
o
t
g
u
ar
an
tee
ex
ce
llen
t c
lu
s
ter
in
g
.
T
h
ese
th
r
ee
m
etr
ics
em
p
h
asi
ze
th
at
th
e
clu
s
ter
in
g
p
e
r
f
o
r
m
ed
s
till
n
ee
d
s
im
p
r
o
v
e
m
en
t
to
ac
h
iev
e
clea
r
er
s
ep
ar
atio
n
.
As
a
co
r
r
ec
tiv
e
s
tep
,
v
alid
atio
n
is
co
n
d
u
cted
b
y
f
o
r
est
an
d
lan
d
f
ir
e
ex
p
er
ts
to
en
s
u
r
e
th
e
ac
cu
r
ac
y
an
d
m
in
im
ize
n
o
is
e
d
u
e
to
o
v
er
la
p
p
in
g
class
es,
r
esu
ltin
g
in
s
o
m
e
p
h
o
to
s
b
ei
n
g
m
o
v
ed
to
m
o
r
e
ap
p
r
o
p
r
iate
class
es.
Af
ter
th
e
v
alid
atio
n
p
r
o
ce
s
s
,
th
e
d
is
tr
ib
u
tio
n
o
f
im
ag
es
in
ea
ch
cl
ass
b
ec
o
m
es
m
o
r
e
b
alan
ce
d
,
with
th
e
n
u
m
b
er
o
f
im
ag
es
r
an
g
in
g
f
r
o
m
1
0
0
to
1
2
3
d
ata
p
o
in
ts
.
T
h
e
f
in
al
s
ta
g
e
in
v
o
l
v
es
s
to
r
in
g
th
e
im
ag
es in
s
ep
ar
ate
f
o
ld
er
s
b
ased
o
n
th
e
estab
lis
h
ed
clu
s
ter
class
if
icatio
n
s
.
Fig
u
r
e
3
.
Data
s
et
id
en
tific
atio
n
f
lo
w
2
.
3
.
Da
t
a
prepro
ce
s
s
ing
T
h
e
d
ata
p
r
e
-
p
r
o
ce
s
s
in
g
s
tag
e
is
p
er
f
o
r
m
ed
t
o
en
s
u
r
e
th
at
th
e
im
ag
e
h
as
a
co
n
s
is
ten
t
s
ize
an
d
f
o
r
m
at
[
2
5
]
.
R
esize
aim
s
to
r
esize
th
e
im
ag
e
to
b
e
co
n
s
is
ten
t w
ith
th
e
in
p
u
t size
ac
ce
p
ted
b
y
th
e
Mo
b
ileNetV2
m
o
d
el
wh
ich
is
2
2
4
×2
2
4
.
R
escale
aim
s
to
n
o
r
m
alize
th
e
im
ag
e
p
ix
el
v
alu
es
to
b
e
in
th
e
s
am
e
r
an
g
e,
ch
an
g
in
g
th
e
im
ag
e
p
ix
el
v
alu
es
f
r
o
m
a
r
an
g
e
o
f
0
-
2
5
5
.
Z
o
o
m
r
a
n
g
e
is
u
s
ed
to
en
lar
g
e
o
r
r
e
d
u
ce
t
h
e
i
m
ag
e
in
th
e
d
ataset,
th
e
v
alu
e
u
s
ed
is
0
.
5
.
H
o
r
izo
n
tal
f
lip
wh
ich
aim
s
to
a
d
d
v
a
r
iety
b
y
f
lip
p
in
g
t
h
e
im
ag
e
h
o
r
izo
n
tally
wh
ich
is
ac
tiv
ated
u
s
in
g
th
e
T
r
u
e
f
u
n
c
tio
n
.
Fil
l
m
o
d
e
ai
m
s
to
h
a
n
d
l
e
em
p
ty
a
r
ea
s
th
at
ap
p
e
ar
af
t
er
tr
an
s
f
o
r
m
atio
n
s
s
u
ch
as r
o
tatio
n
o
r
cr
o
p
p
in
g
,
i
n
th
is
p
r
o
ce
s
s
u
s
in
g
n
ea
r
est to
f
ill with
th
e
n
ea
r
est p
ix
el
v
al
u
e.
2
.
4
.
Sp
lit
da
t
a
s
et
T
h
e
im
ag
es
wer
e
d
iv
id
ed
in
to
7
0
%
tr
ain
in
g
d
ata,
n
am
ely
3
9
2
im
ag
es,
an
d
3
0
%
v
alid
atio
n
d
ata,
n
am
ely
1
6
8
im
ag
es,
wh
ile
th
e
test
d
ata
is
te
s
ted
with
n
ew
d
ata,
as
m
u
ch
as
1
0
%
o
f
th
e
to
tal
d
ata
co
llected
.
T
h
e
d
ata
is
also
s
ep
ar
ated
b
ased
o
n
th
e
s
ev
e
r
ity
o
f
th
e
f
o
r
est
an
d
la
n
d
f
ir
es,
n
am
el
y
v
er
y
m
ild
,
m
ild
,
m
o
d
er
ate,
s
ev
er
e
a
n
d
v
e
r
y
s
e
v
er
e.
E
ac
h
s
u
b
s
et
h
as
a
s
p
ec
i
f
ic
p
u
r
p
o
s
e
in
tr
ain
in
g
a
n
d
ev
alu
atin
g
th
e
m
o
d
el,
wh
ich
h
elp
s
to
en
s
u
r
e
th
at
th
e
r
esu
ltin
g
m
o
d
el
p
e
r
f
o
r
m
s
well
an
d
g
en
e
r
alize
s
ef
f
ec
tiv
ely
to
p
r
ev
io
u
s
ly
u
n
s
ee
n
d
ata.
T
h
e
n
u
m
b
e
r
o
f
d
ata
s
u
b
s
ets is
p
r
esen
ted
in
T
ab
le
1
.
T
ab
le
1
.
Su
b
s
et
o
f
s
p
lit d
ata
C
l
a
s
s
Tr
a
i
n
i
n
g
d
a
t
a
V
a
l
i
d
a
t
i
o
n
d
a
t
a
V
e
r
y
m
i
l
d
7
0
I
mag
e
s
3
0
I
mag
e
s
M
i
l
d
8
4
I
mag
e
s
3
6
I
mag
e
s
M
o
d
e
r
a
t
e
7
4
I
mag
e
s
3
1
I
mag
e
s
S
e
v
e
r
e
7
9
I
mag
e
s
3
3
I
mag
e
s
V
e
r
y
S
e
v
e
r
e
8
6
I
mag
e
s
3
7
I
mag
e
s
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
1
6
,
No
.
2
,
Ap
r
il
20
2
6
:
9
6
4
-
972
968
2
.
5
.
M
o
bil
eNe
t
V2
mo
dellin
g
T
h
is
r
esear
ch
u
s
es
C
N
N
ar
ch
itectu
r
e
Mo
b
ileNetV2
to
cr
ea
te
a
s
ev
er
ity
class
if
icat
io
n
m
o
d
el
o
n
p
h
o
to
s
o
f
p
o
s
t
-
f
ir
e
a
r
ea
.
Mo
b
i
leNe
tV2
is
th
e
ar
ch
itectu
r
e
with
th
e
s
m
allest
ca
p
ac
ity
am
o
n
g
o
th
e
r
m
o
d
els.
I
t
r
an
k
s
h
ig
h
ly
as
a
lig
h
tweig
h
t
m
o
d
el,
with
a
s
m
all
m
o
d
el
s
ize
an
d
f
ast
in
f
er
en
ce
tim
e,
wh
i
ch
is
cr
u
cial
f
o
r
its
ap
p
licatio
n
i
n
m
o
b
ile
-
b
ased
s
y
s
tem
s
.
Mo
b
ileNetV2
h
as
th
e
ad
v
a
n
tag
e
o
f
r
eq
u
ir
in
g
less
s
to
r
ag
e
s
p
ac
e
a
n
d
n
ee
d
in
g
lo
we
r
co
m
p
u
tatio
n
al
p
o
wer
[
2
6
]
.
I
n
th
e
m
o
d
el
tr
ain
in
g
p
r
o
ce
s
s
,
th
e
Mo
b
ileNe
tV2
ar
ch
itectu
r
e
is
u
s
ed
as
th
e
b
ase
m
o
d
el,
with
th
e
f
in
al
lay
er
s
o
f
Mo
b
ileNetV2
f
r
o
ze
n
to
allo
w
th
e
ad
d
itio
n
o
f
n
ew
lay
er
s
tailo
r
ed
to
th
e
n
ee
d
s
o
f
f
o
r
est
an
d
lan
d
f
ir
e
im
a
g
e
class
if
icatio
n
.
Nex
t,
th
e
h
y
p
er
p
ar
am
eter
s
o
f
th
is
m
o
d
el
ar
e
tu
n
ed
u
s
in
g
th
e
g
r
id
s
ea
r
ch
tech
n
iq
u
e,
wh
e
r
e
s
ev
er
al
co
m
b
i
n
atio
n
s
o
f
h
y
p
er
p
ar
a
m
eter
s
ar
e
test
ed
to
f
in
d
th
e
m
o
s
t o
p
tim
al
co
m
b
i
n
atio
n
.
T
h
e
lis
t o
f
h
y
p
er
p
ar
am
eter
s
u
s
ed
in
th
is
s
tu
d
y
is
p
r
esen
ted
in
T
a
b
le
2
.
T
ab
le
2
.
L
is
t o
f
h
y
p
e
r
p
ar
am
et
er
s
in
th
e
m
o
d
el
H
y
p
e
r
p
a
r
a
me
t
e
r
V
a
l
u
e
B
a
t
c
h
si
z
e
[
8
,
1
6
,
3
2
]
Le
a
r
n
i
n
g
r
a
t
e
[
0
.
0
0
0
1
,
0
.
0
0
0
0
1
]
D
r
o
p
o
u
t
[
0
.
3
,
0
.
4
,
0
.
5
]
2
.
6
.
P
re
dict
io
n
a
nd
ev
a
lua
t
i
o
n m
o
del
T
h
e
class
if
icatio
n
r
esu
lts
ar
e
an
aly
ze
d
u
s
in
g
c
o
n
f
u
s
io
n
m
a
tr
ix
to
c
h
ec
k
t
h
e
ac
cu
r
ac
y
v
a
lu
e
o
f
th
e
m
o
d
elin
g
p
r
o
ce
s
s
.
C
o
n
f
u
s
io
n
m
atr
ix
f
o
r
5
class
lab
els
p
r
o
v
id
es
m
o
r
e
co
m
p
r
eh
en
s
iv
e
in
f
o
r
m
atio
n
r
eg
a
r
d
in
g
th
e
m
o
d
el'
s
p
er
f
o
r
m
an
ce
in
ca
teg
o
r
izin
g
d
ata
in
t
o
ea
ch
class
,
with
co
n
f
u
s
io
n
m
atr
ix
,
r
esu
lt
s
ca
n
b
e
ca
lcu
lated
u
s
in
g
v
ar
io
u
s
ev
al
u
atio
n
m
et
r
ics
s
u
ch
as
ac
cu
r
ac
y
p
r
esen
t
ed
in
(
1
)
,
p
r
ec
is
io
n
p
r
esen
ted
in
(
2
)
,
an
d
r
ec
all
p
r
esen
ted
in
(
3
)
[
2
7
]
.
=
+
+
+
+
×
100%
(
1
)
=
+
×
100%
(
2
)
=
+
×
100%
(
3
)
3.
RE
SU
L
T
S
T
h
e
h
y
p
er
p
a
r
am
eter
s
th
at
h
a
v
e
b
ee
n
d
eter
m
in
e
d
u
s
in
g
g
r
id
s
ea
r
ch
ar
e
p
r
o
ce
s
s
ed
in
th
e
m
o
d
el
tr
ain
in
g
u
s
in
g
5
0
e
p
o
ch
s
an
d
a
d
en
s
e
lay
er
o
f
5
1
2
.
T
h
e
m
o
s
t
o
p
tim
al
h
y
p
e
r
p
ar
am
ete
r
c
o
m
b
in
atio
n
is
th
en
u
s
ed
to
p
r
o
d
u
ce
th
e
f
in
al
m
o
d
el
th
at
h
as th
e
b
es
t p
er
f
o
r
m
an
c
e.
T
h
e
r
esu
lts
o
f
th
e
h
y
p
er
p
ar
a
m
eter
co
m
b
in
atio
n
ar
e
p
r
esen
ted
in
T
ab
le
3
.
T
ab
le
3
.
T
h
e
r
esu
lt o
f
t
h
e
co
m
b
in
atio
n
o
f
h
y
p
er
p
a
r
am
eter
s
B
a
t
c
h
S
i
z
e
Le
a
r
n
i
n
g
R
a
t
e
D
r
o
p
o
u
t
A
c
c
u
r
a
c
y
P
r
e
c
i
s
i
o
n
R
e
c
a
l
l
8
0
.
0
0
0
1
0
.
3
0
.
8
5
2
0
.
8
6
1
0
.
8
1
4
8
0
.
0
0
0
1
0
.
4
0
.
7
5
4
0
.
7
8
6
0
.
7
4
6
8
0
.
0
0
0
1
0
.
5
0
.
7
4
6
0
.
7
7
2
0
.
7
4
6
8
0
.
0
0
0
0
1
0
.
3
0
.
7
9
7
0
.
8
3
5
0
.
6
8
6
8
0
.
0
0
0
0
1
0
.
4
0
.
7
4
6
0
.
8
4
3
0
.
6
3
6
8
0
.
0
0
0
0
1
0
.
5
0
.
7
4
6
0
.
8
1
3
0
.
6
2
7
16
0
.
0
0
0
1
0
.
3
0
.
7
6
3
0
.
7
6
6
0
.
7
2
0
16
0
.
0
0
0
1
0
.
4
0
.
8
2
4
0
.
8
3
3
0
.
8
0
5
16
0
.
0
0
0
1
0
.
5
0
.
7
9
7
0
.
8
0
5
0
.
7
7
1
16
0
.
0
0
0
0
1
0
.
3
0
.
7
2
9
0
.
8
1
0
0
.
5
4
2
16
0
.
0
0
0
0
1
0
.
4
0
.
7
9
7
0
.
8
6
6
0
.
6
0
2
16
0
.
0
0
0
0
1
0
.
5
0
.
7
4
6
0
.
8
2
4
0
.
5
9
3
32
0
.
0
0
0
1
0
.
3
0
.
7
8
8
0
.
8
2
4
0
.
7
5
4
32
0
.
0
0
0
1
0
.
4
0
.
8
1
4
0
.
8
3
2
0
.
7
9
7
32
0
.
0
0
0
1
0
.
5
0
.
7
9
7
0
.
8
0
7
0
.
7
8
0
32
0
.
0
0
0
0
1
0
.
3
0
.
6
8
6
0
.
8
0
0
0
.
5
0
8
32
0
.
0
0
0
0
1
0
.
4
0
.
6
6
9
0
.
7
9
7
0
.
4
6
6
32
0
.
0
0
0
0
1
0
.
5
0
.
7
2
9
0
.
8
5
9
0
.
4
6
6
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
-
8
7
0
8
Hyp
erp
a
r
a
mete
r
tu
n
in
g
o
f Mo
b
ileN
etV
2
o
n
fo
r
est a
n
d
la
n
d
…
(
A
s
s
a
d
Hid
a
ya
t
)
969
T
h
e
m
o
s
t
o
p
tim
al
p
ar
am
ete
r
r
esu
lts
in
th
is
s
tu
d
y
ar
e
u
s
in
g
b
a
tch
_
s
iz
e
=
8
,
lea
r
n
in
g
_
r
a
te
=
0
.
0
0
0
1
,
an
d
d
r
o
p
o
u
t
=
0
.
3
.
T
h
is
p
ar
am
eter
co
m
b
in
atio
n
p
r
o
d
u
ce
s
an
ac
cu
r
ac
y
o
f
8
5
%,
with
a
p
r
ec
i
s
io
n
o
f
8
6
%,
an
d
a
r
ec
all
o
f
8
1
%.
T
h
ese
r
esu
lts
s
h
o
w
th
at
th
e
Mo
b
ileNetV2
m
o
d
el
th
at
h
as
b
ee
n
tu
n
ed
with
th
is
co
m
b
in
atio
n
o
f
h
y
p
er
p
ar
am
eter
s
p
e
r
f
o
r
m
s
q
u
i
te
well.
A
g
r
ap
h
o
f
th
e
m
o
d
el
tr
ain
in
g
p
r
o
ce
s
s
illu
s
tr
atin
g
t
h
e
d
ev
el
o
p
m
en
t
o
f
ac
cu
r
ac
y
an
d
lo
s
s
is
p
r
esen
ted
in
Fig
u
r
e
4
.
T
h
e
o
p
tim
al
m
o
d
el
tr
ain
in
g
p
r
o
ce
s
s
was e
v
alu
ated
u
s
in
g
co
n
f
u
s
io
n
m
atr
ix
to
an
aly
ze
its
p
er
f
o
r
m
an
ce
in
class
if
y
in
g
th
e
f
iv
e
ca
teg
o
r
ies
o
f
f
o
r
est
an
d
lan
d
f
ir
e
s
ev
er
ity
.
Vis
u
aliza
tio
n
o
f
th
e
co
n
f
u
s
io
n
m
atr
ix
o
f
th
e
ev
alu
atio
n
r
esu
l
ts
an
d
m
o
d
el
p
r
ed
ictio
n
s
ca
n
b
e
s
ee
n
in
Fig
u
r
e
5
.
Fig
u
r
e
4
.
Mo
d
el
tr
ain
in
g
p
r
o
c
ess
ch
ar
t
Fig
u
r
e
5
.
Vis
u
aliza
tio
n
o
f
co
n
f
u
s
io
n
m
atr
ix
o
f
m
o
d
el
ev
al
u
a
tio
n
an
d
p
r
ed
ictio
n
r
esu
lts
B
ased
o
n
th
e
r
esu
lts
o
f
t
h
e
co
n
f
u
s
io
n
m
atr
ix
v
is
u
aliza
tio
n
i
n
Fig
u
r
e
5
,
th
e
m
o
d
el
s
h
o
ws
a
f
air
ly
h
ig
h
lev
el
o
f
ac
cu
r
ac
y
in
d
is
tin
g
u
is
h
in
g
ea
ch
class
o
f
p
h
o
to
s
o
f
p
o
s
t
-
f
ir
e
h
an
d
lin
g
ar
ea
s
.
Fo
r
th
e
s
ev
er
e
class
,
th
e
m
o
d
el
co
r
r
ec
tly
i
d
en
tifie
d
1
3
o
u
t
o
f
2
2
p
h
o
to
s
,
w
h
ile
4
p
h
o
t
o
s
wer
e
m
is
class
if
ied
as
m
ild
,
1
as
v
er
y
s
ev
er
e,
1
as
v
er
y
m
ild
an
d
4
as
m
o
d
er
ate.
I
n
th
e
m
ild
class
,
th
e
m
o
d
el
co
r
r
ec
tly
p
r
e
d
icted
2
4
o
u
t
o
f
2
8
p
h
o
to
s
,
b
u
t
m
is
class
if
ied
2
p
h
o
to
s
as
s
ev
er
e,
1
as
v
er
y
m
ild
an
d
1
as
m
o
d
er
ate.
Fo
r
th
e
v
er
y
s
ev
e
r
e
class
,
th
e
m
o
d
el
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
1
6
,
No
.
2
,
Ap
r
il
20
2
6
:
9
6
4
-
972
970
co
r
r
ec
tly
class
if
ied
2
3
o
u
t
o
f
2
4
p
h
o
to
s
,
wh
ile
1
p
h
o
to
was
in
co
r
r
ec
tly
i
d
en
tifie
d
as
v
er
y
m
ild
.
I
n
th
e
v
er
y
m
ild
class
,
th
e
m
o
d
el
p
r
e
d
icted
2
2
o
u
t
o
f
2
6
p
h
o
to
s
co
r
r
e
ctly
,
b
u
t
m
is
class
if
ied
1
p
h
o
t
o
as
m
ild
,
2
as
v
er
y
m
ild
an
d
1
as
m
o
d
er
ate.
Fin
ally
,
f
o
r
th
e
m
o
d
e
r
ate
class
,
th
e
m
o
d
el
co
r
r
ec
tly
class
if
ied
1
6
o
u
t
o
f
1
8
p
h
o
to
s
,
wh
ile
1
p
h
o
to
was
m
is
id
en
tifie
d
as
s
ev
er
e,
1
as
m
ild
.
E
v
alu
atio
n
o
f
th
e
m
o
d
el'
s
p
er
f
o
r
m
an
ce
s
h
o
ws
th
at
Mo
b
ileNetV2
h
as a
f
air
ly
g
o
o
d
ab
ilit
y
to
class
if
y
im
ag
es in
t
o
th
e
f
iv
e
s
ev
er
ity
class
es.
T
h
e
tr
ain
ed
m
o
d
el
was
th
en
s
av
ed
in
.
h
5
f
o
r
m
at
an
d
co
n
v
er
ted
in
to
T
e
n
s
o
r
Flo
w
L
i
te
(
T
FLite)
f
o
r
m
at
s
o
th
at
it
c
an
b
e
im
p
le
m
en
ted
in
to
a
s
y
s
tem
in
th
e
f
u
tu
r
e
.
4.
DIS
CU
SS
I
O
N
T
h
is
s
tu
d
y
f
o
cu
s
es
o
n
d
ev
el
o
p
in
g
a
f
o
r
est
an
d
la
n
d
f
ir
e
s
ev
er
ity
class
if
icatio
n
m
o
d
el
u
s
in
g
C
NN
ar
ch
itectu
r
e
Mo
b
ileNetV2
,
a
n
d
h
as
s
h
o
wn
s
ig
n
if
ican
t
im
p
r
o
v
em
en
t
co
m
p
ar
e
d
to
p
r
ev
io
u
s
r
esear
ch
th
at
also
u
s
es
th
e
s
am
e
ar
ch
itectu
r
e.
Pre
v
io
u
s
r
esear
ch
co
n
d
u
cted
b
y
Hid
ay
at
et
a
l
.
[
8
]
,
wh
er
e
th
e
Mo
b
ileNetV2
m
o
d
el
was
u
s
ed
to
class
if
y
th
e
s
ev
er
ity
o
f
f
o
r
est
f
ir
es
with
a
to
tal
o
f
9
0
d
ata
a
n
d
two
im
ag
e
s
izes,
n
am
ely
2
2
4
×2
2
4
an
d
1
1
2
×1
1
2
,
r
esu
lted
in
th
e
b
est
ac
cu
r
ac
y
r
ea
ch
in
g
7
7
.
7
%.
Me
an
wh
ile,
in
th
is
s
tu
d
y
,
th
e
m
o
d
el
tr
ain
ed
with
h
y
p
er
p
ar
am
eter
o
p
tim
izatio
n
,
n
am
ely
b
a
tch
_
s
iz
e
=
8
,
le
a
r
n
in
g
_
r
a
te
=
0
.
0
0
0
1
,
a
n
d
d
r
o
p
o
u
t
=
0
.
3
r
esu
lted
in
h
ig
h
er
ac
c
u
r
ac
y
,
wh
ich
is
8
5
%,
with
p
r
ec
is
io
n
8
6
% a
n
d
r
ec
all
8
1
%.
T
h
e
m
ain
d
if
f
e
r
en
ce
b
etwe
en
th
ese
two
s
tu
d
ies
lies
in
th
e
d
ata
p
r
o
ce
s
s
in
g
a
p
p
r
o
ac
h
a
n
d
t
h
e
u
s
e
o
f
h
y
p
er
p
ar
am
eter
tu
n
in
g
tech
n
i
q
u
es.
T
h
e
p
r
e
v
io
u
s
s
tu
d
y
u
s
ed
lim
ited
d
ata,
wh
ic
h
r
esu
lted
in
th
e
m
o
d
el
o
n
ly
b
ein
g
ab
le
to
ca
p
t
u
r
e
a
lim
ited
v
ar
iatio
n
in
s
ev
er
ity
.
As
a
r
esu
lt,
alth
o
u
g
h
Mo
b
ileNetV2
p
r
o
v
id
es
ad
e
q
u
ate
p
er
f
o
r
m
an
ce
,
th
er
e
is
s
till
r
o
o
m
f
o
r
im
p
r
o
v
e
m
en
t,
esp
ec
ially
in
ter
m
s
o
f
i
n
cr
ea
s
in
g
th
e
v
a
r
iety
o
f
f
ield
d
ata
to
m
ak
e
th
e
m
o
d
el
m
o
r
e
r
o
b
u
s
t.
I
n
co
n
t
r
ast,
th
is
s
tu
d
y
u
tili
ze
s
a
m
o
r
e
d
i
v
er
s
e
d
ataset,
co
v
er
i
n
g
a
wid
e
r
a
n
g
e
o
f
f
o
r
est
an
d
lan
d
f
ir
e
ev
en
ts
in
I
n
d
o
n
esia,
wh
ich
g
iv
es
th
e
m
o
d
el
ac
ce
s
s
to
a
g
r
ea
ter
v
ar
iety
o
f
s
ev
er
ity
lev
els.
T
h
e
ap
p
licatio
n
o
f
h
y
p
er
p
ar
a
m
eter
tu
n
in
g
in
th
is
s
tu
d
y
s
ig
n
if
ican
tly
im
p
r
o
v
ed
th
e
m
o
d
e
l's
p
er
f
o
r
m
an
ce
.
T
h
e
ap
p
lied
g
r
i
d
s
ea
r
ch
h
elp
s
to
f
i
n
d
th
e
m
o
s
t
o
p
tim
al
co
m
b
in
at
io
n
o
f
p
ar
am
eter
s
,
w
h
ile
in
p
r
ev
io
u
s
s
tu
d
ies,
th
e
h
y
p
er
p
ar
am
eter
tu
n
in
g
p
r
o
ce
s
s
h
as
n
o
t
b
ee
n
o
p
tim
ized
.
T
h
e
m
o
d
el
in
th
is
s
tu
d
y
is
m
o
r
e
a
cc
u
r
ate,
s
tab
le,
an
d
ca
n
g
en
e
r
alize
to
n
ew
d
ata.
T
o
f
u
r
th
er
e
v
alu
ate
th
e
co
n
tr
i
b
u
tio
n
o
f
d
if
f
er
en
t
m
o
d
el
co
m
p
o
n
en
ts
,
a
n
ab
latio
n
s
tu
d
y
f
o
cu
s
ed
o
n
d
r
o
p
o
u
t,
le
ar
n
in
g
r
ate,
an
d
u
s
in
g
a
p
r
e
-
t
r
ain
ed
Mo
b
ileNetV2
m
o
d
el.
T
h
is
s
tu
d
y
aim
s
to
an
aly
ze
th
e
im
p
ac
t
o
f
ea
ch
e
lem
en
t
o
n
th
e
o
v
er
all
class
if
icatio
n
p
er
f
o
r
m
an
ce
.
Dr
o
p
o
u
t
is
a
r
eg
u
lar
izatio
n
tech
n
iq
u
e
t
h
at
r
e
d
u
ce
s
o
v
er
f
it
tin
g
b
y
r
an
d
o
m
ly
d
ea
ctiv
atin
g
n
eu
r
o
n
s
d
u
r
in
g
tr
ain
i
n
g
.
Var
io
u
s
d
r
o
p
o
u
t
r
ates
(
0
.
1
,
0
.
3
,
an
d
0
.
5
)
wer
e
test
e
d
,
an
d
th
e
r
esu
lts
s
h
o
wed
th
a
t
a
d
r
o
p
o
u
t
r
ate
o
f
0
.
3
p
r
o
v
i
d
ed
th
e
b
est
b
alan
ce
b
etwe
en
g
en
e
r
aliza
tio
n
an
d
p
er
f
o
r
m
a
n
ce
.
T
h
e
lear
n
i
n
g
r
ate
d
eter
m
in
es
th
e
s
tep
s
ize
wh
en
u
p
d
atin
g
m
o
d
el
p
ar
am
eter
s
d
u
r
in
g
tr
ain
in
g
.
E
x
p
er
im
en
ts
with
th
r
ee
d
if
f
er
en
t
lear
n
in
g
r
ates
(
0
.
0
1
,
0
.
0
0
1
,
a
n
d
0
.
0
0
0
1
)
in
d
icate
d
th
at
0
.
0
0
0
1
was
th
e
o
p
tim
al
ch
o
ice.
A
h
ig
h
er
le
ar
n
in
g
r
ate
(
0
.
0
1
)
ca
u
s
ed
u
n
s
tab
le
co
n
v
er
g
en
ce
,
lead
in
g
to
s
u
b
o
p
tim
al
r
esu
lts
,
wh
ile
a
lo
wer
lear
n
in
g
r
ate
(
0
.
0
0
1
)
p
r
o
lo
n
g
ed
th
e
tr
ain
in
g
p
r
o
ce
s
s
with
o
u
t
s
ig
n
if
ican
t p
er
f
o
r
m
an
ce
g
ain
s
.
T
h
is
s
tu
d
y
wen
t
th
r
o
u
g
h
a
p
r
o
ce
s
s
o
f
clu
s
ter
in
g
th
e
d
ataset
b
ef
o
r
e
s
tar
tin
g
th
e
m
o
d
el
tr
ain
in
g
,
an
ap
p
r
o
ac
h
th
at
was
n
o
t
tak
en
in
p
r
e
v
io
u
s
s
tu
d
ies.
T
h
e
clu
s
ter
in
g
p
r
o
ce
s
s
,
v
alid
ated
b
y
Fo
r
est
an
d
L
an
d
Fir
e
E
x
p
er
ts
,
en
s
u
r
e
d
th
at
th
e
im
a
g
es
u
s
ed
in
t
h
e
m
o
d
el
tr
ain
in
g
wer
e
m
o
r
e
e
v
en
ly
d
is
tr
ib
u
te
d
in
ea
c
h
class
.
T
h
e
r
esu
lts
o
b
tain
ed
in
th
is
s
tu
d
y
s
h
o
w
th
at
a
m
o
r
e
co
m
p
r
eh
en
s
iv
e
ap
p
r
o
ac
h
to
d
ata
co
llectio
n
,
h
y
p
e
r
p
ar
a
m
eter
o
p
tim
izatio
n
,
an
d
d
ataset
clu
s
ter
in
g
h
as
s
ig
n
if
ican
tly
im
p
r
o
v
e
d
m
o
d
el
p
er
f
o
r
m
a
n
c
e.
Dev
elo
p
in
g
th
is
class
if
icatio
n
m
o
d
el
in
ev
itab
l
y
in
tr
o
d
u
ce
s
p
o
ten
tial
b
iases
in
th
e
d
ataset,
esp
ec
ially
r
eg
ar
d
in
g
im
a
g
e
q
u
ality
an
d
g
eo
g
r
ap
h
ic
co
v
e
r
ag
e.
Var
iatio
n
s
in
im
a
g
e
q
u
ality
,
s
u
c
h
as
d
if
f
e
r
en
ce
s
in
r
eso
lu
tio
n
,
l
ig
h
tin
g
c
o
n
d
itio
n
s
,
an
d
s
en
s
o
r
ty
p
es,
ca
n
ca
u
s
e
in
ter
f
er
en
ce
in
t
h
e
tr
ain
in
g
p
r
o
ce
s
s
,
lead
in
g
to
in
co
n
s
is
ten
t
m
o
d
el
p
er
f
o
r
m
a
n
ce
.
T
o
ad
d
r
ess
th
ese
b
iases
,
ap
p
ly
in
g
ad
v
a
n
ce
d
d
ata
au
g
m
e
n
tatio
n
tech
n
iq
u
es
ca
n
h
elp
m
itig
ate
th
e
im
p
ac
t
o
f
im
ag
e
d
iv
er
s
ity
lim
itatio
n
s
b
y
ar
tific
ially
in
cr
ea
s
in
g
th
e
d
iv
er
s
ity
o
f
tr
ain
in
g
s
am
p
les.
Fu
tu
r
e
wo
r
k
s
h
o
u
ld
also
ex
p
an
d
th
e
d
ataset
to
in
c
lu
d
e
im
ag
es
f
r
o
m
d
if
f
er
en
t
s
o
u
r
ce
s
an
d
g
eo
g
r
ap
h
ic
lo
ca
tio
n
s
s
o
th
at
th
e
m
o
d
el
will b
e
m
o
r
e
r
o
b
u
s
t.
5.
CO
NCLU
SI
O
N
T
h
is
s
tu
d
y
s
u
cc
ess
f
u
lly
cr
e
ated
a
f
o
r
est
an
d
lan
d
f
ir
e
s
ev
er
ity
class
if
icatio
n
m
o
d
el
u
s
in
g
Mo
b
ileNetV2
ar
ch
itectu
r
e.
B
y
ap
p
ly
in
g
h
y
p
e
r
p
ar
am
eter
tu
n
in
g
tech
n
iq
u
es
th
r
o
u
g
h
g
r
id
s
ea
r
ch
,
an
o
p
tim
al
p
ar
am
eter
co
m
b
in
atio
n
o
f
b
a
tch
_
s
iz
e
=
8
,
lea
r
n
in
g
_
r
a
te
=
0
.
0
0
0
1
,
an
d
d
r
o
p
o
u
t
=
0
.
3
was
o
b
tain
ed
.
T
h
e
r
esu
ltin
g
m
o
d
el
s
h
o
ws
g
o
o
d
p
er
f
o
r
m
a
n
ce
with
ac
cu
r
ac
y
r
ea
ch
in
g
8
5
%,
p
r
ec
is
io
n
o
f
8
6
%,
an
d
r
ec
all
o
f
8
1
%.
Hy
p
er
p
ar
a
m
eter
tu
n
in
g
u
s
in
g
g
r
id
s
ea
r
ch
,
as
well
as
a
d
at
aset
clu
s
ter
in
g
ap
p
r
o
ac
h
u
s
in
g
K
-
m
ea
n
s
p
r
io
r
to
m
o
d
el
tr
ai
n
in
g
p
r
o
v
ed
ef
f
ec
ti
v
e
in
im
p
r
o
v
in
g
th
e
m
o
d
el'
s
p
er
f
o
r
m
a
n
ce
in
class
if
y
in
g
t
h
e
f
iv
e
s
ev
er
ity
class
es
o
f
f
o
r
est
an
d
la
n
d
f
i
r
es:
v
er
y
m
ild
,
m
ild
,
m
o
d
e
r
ate,
s
ev
er
e,
an
d
v
er
y
s
ev
er
e.
E
v
alu
atio
n
u
s
in
g
c
o
n
f
u
s
io
n
m
atr
ix
s
h
o
ws th
at
th
e
m
o
d
el
h
as a
g
o
o
d
a
b
ilit
y
to
d
is
tin
g
u
is
h
th
e
f
iv
e
s
ev
er
ity
class
es.
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
-
8
7
0
8
Hyp
erp
a
r
a
mete
r
tu
n
in
g
o
f Mo
b
ileN
etV
2
o
n
fo
r
est a
n
d
la
n
d
…
(
A
s
s
a
d
Hid
a
ya
t
)
971
A
s
u
g
g
esti
o
n
f
o
r
f
u
t
u
r
e
r
esear
ch
is
to
ex
p
an
d
th
e
d
ataset
with
m
o
r
e
im
ag
es
f
r
o
m
d
if
f
er
e
n
t
lo
ca
tio
n
s
an
d
d
if
f
er
en
t
f
o
r
est
f
ir
e
co
n
d
itio
n
s
.
T
h
is
aim
s
to
im
p
r
o
v
e
th
e
m
o
d
el'
s
ab
ilit
y
to
g
en
er
alize
n
ew
d
ata
an
d
h
an
d
le
m
o
r
e
c
o
m
p
lex
v
ar
iatio
n
s
.
E
x
p
lo
r
atio
n
o
f
o
t
h
er
m
o
d
e
l
ar
ch
itectu
r
es
s
u
ch
as
E
f
f
icie
n
tNet
ca
n
b
e
d
o
n
e
to
co
m
p
ar
e
m
o
d
el
p
er
f
o
r
m
an
ce
o
n
t
h
e
s
am
e
class
if
icatio
n
task
.
T
h
e
a
d
d
itio
n
o
f
m
o
r
e
v
a
r
ied
d
ata
au
g
m
en
tatio
n
m
eth
o
d
s
ca
n
also
h
elp
th
e
m
o
d
el
im
p
r
o
v
e.
ACK
NO
WL
E
DG
E
M
E
NT
S
T
h
e
au
th
o
r
s
wo
u
l
d
lik
e
to
th
a
n
k
th
o
s
e
wh
o
p
r
o
v
id
e
d
f
ield
d
ata
f
r
o
m
th
e
J
am
b
i
r
eg
io
n
,
wh
ich
g
r
ea
tly
s
u
p
p
o
r
ted
th
e
im
p
lem
en
tatio
n
o
f
th
is
r
esear
ch
.
T
h
e
d
ata
p
r
o
v
id
ed
was
in
v
al
u
ab
le
i
n
th
e
d
ev
elo
p
m
e
n
t
o
f
t
h
e
f
o
r
est an
d
lan
d
f
ir
e
s
ev
e
r
ity
cl
ass
if
icatio
n
m
o
d
el
u
s
ed
in
th
is
s
tu
d
y
.
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
.
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
a
v
aila
b
i
lit
y
is
n
o
t
a
p
p
li
ca
b
le
t
o
t
h
is
p
ap
er
as
n
o
n
e
w
d
a
t
a
w
er
e
cr
ea
te
d
o
r
an
al
y
z
e
d
i
n
t
h
is
s
t
u
d
y
.
RE
F
E
R
E
NC
E
S
[
1
]
Z.
A
k
h
y
a
r
,
M
.
K
i
p
t
i
a
h
,
D
.
A
.
R
u
c
h
l
i
y
a
d
i
,
a
n
d
M
.
El
m
y
,
“
L
e
g
a
l
a
w
a
r
e
n
e
ss
o
f
c
i
t
i
z
e
n
s
i
n
p
r
e
v
e
n
t
i
n
g
f
o
r
e
st
a
n
d
l
a
n
d
f
i
r
e
s
(
K
a
r
h
u
t
l
a
)
i
n
S
o
u
t
h
K
a
l
i
ma
n
t
a
n
P
r
o
v
i
n
c
e
,
”
J
o
u
r
n
a
l
o
f
Ec
o
h
u
m
a
n
i
sm
,
v
o
l
.
3
,
n
o
.
4
,
p
p
.
1
5
1
–
1
6
1
,
Ju
l
.
2
0
2
4
,
d
o
i
:
1
0
.
6
2
7
5
4
/
j
o
e
.
v
3
i
4
.
3
5
6
6
.
[
2
]
W
.
M
a
,
Z
.
F
e
n
g
,
Z
.
C
h
e
n
g
,
S
.
C
h
e
n
,
a
n
d
F
.
W
a
n
g
,
“
I
d
e
n
t
i
f
y
i
n
g
f
o
r
e
s
t
f
i
r
e
d
r
i
v
i
n
g
f
a
c
t
o
r
s
a
n
d
r
e
l
a
t
e
d
i
m
p
a
c
t
s
i
n
C
h
i
n
a
u
s
i
n
g
r
a
n
d
o
m
f
o
r
e
s
t
a
l
g
o
r
i
t
h
m,
”
F
o
res
t
s
,
v
o
l
.
1
1
,
n
o
.
5
,
p
.
5
0
7
,
M
a
y
2
0
2
0
,
d
o
i
:
1
0
.
3
3
9
0
/
F
1
1
0
5
0
5
0
7
.
[
3
]
L.
S
y
a
u
f
i
n
a
a
n
d
A
.
A
.
H
a
mz
a
h
,
“
C
h
a
n
g
e
s
o
f
t
r
e
e
s
p
e
c
i
e
s
d
i
v
e
r
s
i
t
y
i
n
p
e
a
t
l
a
n
d
i
mp
a
c
t
e
d
b
y
m
o
d
e
r
a
t
e
f
i
r
e
s
e
v
e
r
i
t
y
a
t
T
e
l
u
k
M
e
r
a
n
t
i
,
P
e
l
a
l
a
w
a
n
,
R
i
a
u
p
r
o
v
i
n
c
e
,
I
n
d
o
n
e
s
i
a
,
”
Bi
o
d
i
v
e
rsi
t
a
s
,
v
o
l
.
2
2
,
n
o
.
5
,
p
p
.
2
8
9
9
–
2
9
0
8
,
2
0
2
1
,
d
o
i
:
1
0
.
1
3
0
5
7
/
b
i
o
d
i
v
/
d
2
2
0
5
5
5
.
[
4
]
A
.
Ta
r
i
q
e
t
a
l
.
,
“
F
o
r
e
s
t
f
i
r
e
mo
n
i
t
o
r
i
n
g
u
si
n
g
sp
a
t
i
a
l
-
st
a
t
i
s
t
i
c
a
l
a
n
d
G
e
o
-
sp
a
t
i
a
l
a
n
a
l
y
si
s
o
f
f
a
c
t
o
r
s
d
e
t
e
r
mi
n
i
n
g
f
o
r
e
s
t
f
i
r
e
i
n
M
a
r
g
a
l
l
a
H
i
l
l
s
,
I
sl
a
ma
b
a
d
,
P
a
k
i
s
t
a
n
,
”
G
e
o
m
a
t
i
c
s
,
N
a
t
u
r
a
l
H
a
z
a
r
d
s
a
n
d
Ri
s
k
,
v
o
l
.
1
2
,
n
o
.
1
,
p
p
.
1
2
1
2
–
1
2
3
3
,
2
0
2
1
,
d
o
i
:
1
0
.
1
0
8
0
/
1
9
4
7
5
7
0
5
.
2
0
2
1
.
1
9
2
0
4
7
7
.
[
5
]
A
.
L.
R
.
W
e
s
t
e
r
l
i
n
g
,
“
I
n
c
r
e
a
si
n
g
w
e
st
e
r
n
U
S
f
o
r
e
st
w
i
l
d
f
i
r
e
a
c
t
i
v
i
t
y
:
S
e
n
si
t
i
v
i
t
y
t
o
c
h
a
n
g
e
s
i
n
t
h
e
t
i
mi
n
g
o
f
s
p
r
i
n
g
,
”
Ph
i
l
o
s
o
p
h
i
c
a
l
T
ra
n
s
a
c
t
i
o
n
s
o
f
t
h
e
Ro
y
a
l
S
o
c
i
e
t
y
B:
Bi
o
l
o
g
i
c
a
l
S
c
i
e
n
c
e
s
,
v
o
l
.
3
7
1
,
n
o
.
1
6
9
6
,
J
u
n
.
2
0
1
6
,
d
o
i
:
1
0
.
1
0
9
8
/
r
st
b
.
2
0
1
5
.
0
1
7
8
.
[
6
]
A
.
N
.
I
k
h
sa
n
,
D
.
S
.
H
a
d
mo
k
o
,
a
n
d
P
.
W
i
d
a
y
a
n
i
,
“
S
p
a
t
i
a
l
mo
d
e
l
i
n
g
o
f
f
o
r
e
st
a
n
d
l
a
n
d
f
i
r
e
s
u
s
c
e
p
t
i
b
i
l
i
t
y
u
s
i
n
g
t
h
e
i
n
f
o
r
ma
t
i
o
n
v
a
l
u
e
m
e
t
h
o
d
i
n
K
o
t
a
w
a
r
i
n
g
i
n
B
a
r
a
t
R
e
g
e
n
c
y
,
I
n
d
o
n
e
s
i
a
,
”
Fi
re
,
v
o
l
.
6
,
n
o
.
4
,
p
.
1
7
0
,
A
p
r
.
2
0
2
3
,
d
o
i
:
1
0
.
3
3
9
0
/
f
i
r
e
6
0
4
0
1
7
0
.
[
7
]
K
.
B
u
d
i
n
i
n
g
si
h
e
t
a
l
.
,
“
F
o
r
e
st
ma
n
a
g
e
me
n
t
u
n
i
t
s’
p
e
r
f
o
r
ma
n
c
e
i
n
f
o
r
e
st
f
i
r
e
ma
n
a
g
e
me
n
t
i
m
p
l
e
m
e
n
t
a
t
i
o
n
i
n
C
e
n
t
r
a
l
K
a
l
i
ma
n
t
a
n
a
n
d
S
o
u
t
h
S
u
mat
r
a
,
”
F
o
res
t
s
,
v
o
l
.
1
3
,
n
o
.
6
,
p
.
8
9
4
,
J
u
n
.
2
0
2
2
,
d
o
i
:
1
0
.
3
3
9
0
/
f
1
3
0
6
0
8
9
4
.
[
8
]
A
.
H
i
d
a
y
a
t
,
I
.
S
.
S
i
t
a
n
g
g
a
n
g
,
a
n
d
L.
S
y
a
u
f
i
n
a
,
“
C
l
a
ssi
f
i
c
a
t
i
o
n
o
f
f
o
r
e
st
a
n
d
l
a
n
d
f
i
r
e
se
v
e
r
i
t
y
l
e
v
e
l
s
u
si
n
g
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
,
”
BI
O
W
e
b
o
f
C
o
n
f
e
re
n
c
e
s
,
v
o
l
.
1
2
3
,
p
.
1
0
3
0
,
A
u
g
.
2
0
2
4
,
d
o
i
:
1
0
.
1
0
5
1
/
b
i
o
c
o
n
f
/
2
0
2
4
1
2
3
0
1
0
3
0
.
[
9
]
K
.
H
.
J
o
d
h
a
n
i
e
t
a
l
.
,
“
A
ssessm
e
n
t
o
f
f
o
r
e
st
f
i
r
e
se
v
e
r
i
t
y
a
n
d
l
a
n
d
s
u
r
f
a
c
e
t
e
mp
e
r
a
t
u
r
e
u
s
i
n
g
G
o
o
g
l
e
E
a
r
t
h
En
g
i
n
e
:
a
c
a
se
s
t
u
d
y
o
f
G
u
j
a
r
a
t
S
t
a
t
e
,
I
n
d
i
a
,
”
F
i
re
Ec
o
l
o
g
y
,
v
o
l
.
2
0
,
n
o
.
1
,
p
.
2
3
,
M
a
r
.
2
0
2
4
,
d
o
i
:
1
0
.
1
1
8
6
/
s4
2
4
0
8
-
024
-
0
0
2
5
4
-
2.
[
1
0
]
K
.
K
.
M
c
La
u
c
h
l
a
n
e
t
a
l
.
,
“
F
i
r
e
a
s
a
f
u
n
d
a
me
n
t
a
l
e
c
o
l
o
g
i
c
a
l
p
r
o
c
e
ss
:
R
e
sea
r
c
h
a
d
v
a
n
c
e
s
a
n
d
f
r
o
n
t
i
e
r
s,”
J
o
u
r
n
a
l
o
f
E
c
o
l
o
g
y
,
v
o
l
.
1
0
8
,
n
o
.
5
,
p
p
.
2
0
4
7
–
2
0
6
9
,
S
e
p
.
2
0
2
0
,
d
o
i
:
1
0
.
1
1
1
1
/
1
3
6
5
-
2
7
4
5
.
1
3
4
0
3
.
[
1
1
]
L.
S
y
a
u
f
i
n
a
,
I
.
S
.
S
i
t
a
n
g
g
a
n
g
,
a
n
d
F
.
S
.
A
f
i
n
a
,
“
C
h
a
l
l
e
n
g
e
s
o
n
f
i
r
e
se
v
e
r
i
t
y
a
ssess
me
n
t
i
n
I
n
d
o
n
e
s
i
a
:
A
v
e
g
e
t
a
t
i
o
n
d
i
v
e
r
si
t
y
c
h
a
n
g
e
s
p
e
r
s
p
e
c
t
i
v
e
,
”
i
n
I
O
P
C
o
n
f
e
re
n
c
e
S
e
r
i
e
s:
E
a
rt
h
a
n
d
E
n
v
i
r
o
n
m
e
n
t
a
l
S
c
i
e
n
c
e
,
N
o
v
.
2
0
2
1
,
v
o
l
.
8
8
6
,
n
o
.
1
,
d
o
i
:
1
0
.
1
0
8
8
/
1
7
5
5
-
1
3
1
5
/
8
8
6
/
1
/
0
1
2
1
0
7
.
[
1
2
]
M
.
P
a
r
k
,
D
.
Q
.
Tr
a
n
,
D
.
Ju
n
g
,
a
n
d
S
.
P
a
r
k
,
“
W
i
l
d
f
i
r
e
-
d
e
t
e
c
t
i
o
n
me
t
h
o
d
u
s
i
n
g
D
e
n
seN
e
t
a
n
d
C
y
c
l
e
G
A
N
d
a
t
a
a
u
g
me
n
t
a
t
i
o
n
-
b
a
se
d
r
e
mo
t
e
c
a
m
e
r
a
i
ma
g
e
r
y
,
”
Re
m
o
t
e
S
e
n
si
n
g
,
v
o
l
.
1
2
,
n
o
.
2
2
,
p
p
.
1
–
1
6
,
N
o
v
.
2
0
2
0
,
d
o
i
:
1
0
.
3
3
9
0
/
r
s
1
2
2
2
3
7
1
5
.
[
1
3
]
A
.
K
.
A
h
m
e
d
,
M
.
Q
.
T
a
h
a
,
a
n
d
A
.
S
.
M
u
s
t
a
f
a
,
“
O
n
-
r
o
a
d
a
u
t
o
mo
b
i
l
e
l
i
c
e
n
se
p
l
a
t
e
r
e
c
o
g
n
i
t
i
o
n
u
s
i
n
g
c
o
-
o
c
c
u
r
r
e
n
c
e
mat
r
i
x
,
”
J
o
u
rn
a
l
o
f
Ad
v
a
n
c
e
d
R
e
se
a
rc
h
i
n
D
y
n
a
m
i
c
a
l
&
C
o
n
t
ro
l
S
y
s
t
e
m
s
,
v
o
l
.
1
0
,
n
o
.
7
,
2
0
1
8
.
[
1
4
]
R
.
G
h
a
l
i
a
n
d
M
.
A
.
A
k
h
l
o
u
f
i
,
“
D
e
e
p
l
e
a
r
n
i
n
g
a
p
p
r
o
a
c
h
e
s
f
o
r
w
i
l
d
l
a
n
d
f
i
r
e
s
u
s
i
n
g
sa
t
e
l
l
i
t
e
r
e
m
o
t
e
s
e
n
s
i
n
g
d
a
t
a
:
d
e
t
e
c
t
i
o
n
,
map
p
i
n
g
,
a
n
d
p
r
e
d
i
c
t
i
o
n
,
”
F
i
re
,
v
o
l
.
6
,
n
o
.
5
,
p
.
1
9
2
,
M
a
y
2
0
2
3
,
d
o
i
:
1
0
.
3
3
9
0
/
f
i
r
e
6
0
5
0
1
9
2
.
[
1
5
]
S
.
K
o
n
d
y
l
a
t
o
s
e
t
a
l
.
,
“
W
i
l
d
f
i
r
e
d
a
n
g
e
r
p
r
e
d
i
c
t
i
o
n
a
n
d
u
n
d
e
r
st
a
n
d
i
n
g
w
i
t
h
d
e
e
p
l
e
a
r
n
i
n
g
,
”
G
e
o
p
h
y
s
i
c
a
l
Re
s
e
a
r
c
h
L
e
t
t
e
rs
,
v
o
l
.
4
9
,
n
o
.
1
7
,
S
e
p
.
2
0
2
2
,
d
o
i
:
1
0
.
1
0
2
9
/
2
0
2
2
G
L0
9
9
3
6
8
.
[
1
6
]
M
.
Y
a
n
d
o
u
z
i
e
t
a
l
.
,
“
F
o
r
e
s
t
f
i
r
e
s
d
e
t
e
c
t
i
o
n
u
s
i
n
g
d
e
e
p
t
r
a
n
sf
e
r
l
e
a
r
n
i
n
g
,
”
I
n
t
e
r
n
a
t
i
o
n
a
l
J
o
u
rn
a
l
o
f
A
d
v
a
n
c
e
d
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
a
n
d
Ap
p
l
i
c
a
t
i
o
n
s
,
v
o
l
.
1
3
,
n
o
.
8
,
p
p
.
2
6
8
–
2
7
5
,
2
0
2
2
,
d
o
i
:
1
0
.
1
4
5
6
9
/
I
JA
C
S
A
.
2
0
2
2
.
0
1
3
0
8
3
2
.
[
1
7
]
B
.
Ö
z
e
l
,
M
.
S
.
A
l
a
m
,
a
n
d
M
.
U
.
K
h
a
n
,
“
R
e
v
i
e
w
o
f
m
o
d
e
r
n
f
o
r
e
s
t
f
i
r
e
d
e
t
e
c
t
i
o
n
t
e
c
h
n
i
q
u
e
s:
i
n
n
o
v
a
t
i
o
n
s
i
n
i
m
a
g
e
p
r
o
c
e
ssi
n
g
a
n
d
d
e
e
p
l
e
a
r
n
i
n
g
,
”
I
n
f
o
rm
a
t
i
o
n
(
S
w
i
t
zer
l
a
n
d
)
,
v
o
l
.
1
5
,
n
o
.
9
,
p
.
5
3
8
,
S
e
p
.
2
0
2
4
,
d
o
i
:
1
0
.
3
3
9
0
/
i
n
f
o
1
5
0
9
0
5
3
8
.
[
1
8
]
R
.
P
.
S
a
d
e
w
a
,
B
.
I
r
a
w
a
n
,
a
n
d
C
.
S
e
t
i
a
n
i
n
g
s
i
h
,
“
F
i
r
e
d
e
t
e
c
t
i
o
n
u
s
i
n
g
i
m
a
g
e
p
r
o
c
e
ss
i
n
g
t
e
c
h
n
i
q
u
e
s
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
2
0
1
9
2
n
d
I
n
t
e
r
n
a
t
i
o
n
a
l
S
e
m
i
n
a
r o
n
Re
s
e
a
r
c
h
o
f
I
n
f
o
rm
a
t
i
o
n
T
e
c
h
n
o
l
o
g
y
a
n
d
I
n
t
e
l
l
i
g
e
n
t
S
y
st
e
m
s,
I
S
RI
T
I
2
0
1
9
,
D
e
c
.
2
0
1
9
,
p
p
.
2
9
0
–
2
9
5
,
d
o
i
:
1
0
.
1
1
0
9
/
I
S
R
I
TI
4
8
6
4
6
.
2
0
1
9
.
9
0
3
4
6
4
2
.
[
1
9
]
A
.
S
a
n
t
o
p
a
o
l
o
,
S
.
S
.
S
a
i
f
,
A
.
P
i
e
t
r
a
b
i
s
sa,
a
n
d
A
.
G
i
u
se
p
p
i
,
“
F
o
r
e
s
t
f
i
r
e
r
i
sk
p
r
e
d
i
c
t
i
o
n
f
r
o
m sa
t
e
l
l
i
t
e
d
a
t
a
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
2
0
2
1
2
9
t
h
M
e
d
i
t
e
rr
a
n
e
a
n
C
o
n
f
e
r
e
n
c
e
o
n
C
o
n
t
r
o
l
a
n
d
Au
t
o
m
a
t
i
o
n
(
ME
D
)
,
J
u
n
.
2
0
2
1
,
p
p
.
3
6
0
–
3
6
7
,
d
o
i
:
1
0
.
1
1
0
9
/
M
ED
5
1
4
4
0
.
2
0
2
1
.
9
4
8
0
2
2
6
.
[
2
0
]
K
.
D
o
n
g
,
C
.
Zh
o
u
,
Y
.
R
u
a
n
,
a
n
d
Y
.
Li
,
“
M
o
b
i
l
e
N
e
t
V
2
m
o
d
e
l
f
o
r
i
ma
g
e
c
l
a
ssi
f
i
c
a
t
i
o
n
,
”
i
n
Pr
o
c
e
e
d
i
n
g
s
-
2
0
2
0
2
n
d
I
n
t
e
r
n
a
t
i
o
n
a
l
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
1
6
,
No
.
2
,
Ap
r
il
20
2
6
:
9
6
4
-
972
972
C
o
n
f
e
re
n
c
e
o
n
I
n
f
o
rm
a
t
i
o
n
T
e
c
h
n
o
l
o
g
y
a
n
d
C
o
m
p
u
t
e
r
A
p
p
l
i
c
a
t
i
o
n
,
I
T
C
A
2
0
2
0
,
D
e
c
.
2
0
2
0
,
p
p
.
4
7
6
–
4
8
0
,
d
o
i
:
1
0
.
1
1
0
9
/
I
TC
A
5
2
1
1
3
.
2
0
2
0
.
0
0
1
0
6
.
[
2
1
]
E.
Tu
b
a
,
N
.
B
a
č
a
n
i
n
,
I
.
S
t
r
u
m
b
e
r
g
e
r
,
a
n
d
M
.
Tu
b
a
,
“
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
h
y
p
e
r
p
a
r
a
m
e
t
e
r
s
t
u
n
i
n
g
,
”
i
n
Ar
t
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
:
T
h
e
o
r
y
a
n
d
Ap
p
l
i
c
a
t
i
o
n
s
,
2
0
2
1
,
p
p
.
6
5
–
8
4
.
[
2
2
]
M
.
W
o
j
c
i
u
k
,
Z.
S
w
i
d
e
r
s
k
a
-
C
h
a
d
a
j
,
K
.
S
i
w
e
k
,
a
n
d
A
.
G
e
r
t
y
c
h
,
“
T
h
e
r
o
l
e
o
f
h
y
p
e
r
p
a
r
a
me
t
e
r
o
p
t
i
mi
z
a
t
i
o
n
i
n
f
i
n
e
-
t
u
n
i
n
g
o
f
C
N
N
mo
d
e
l
s
,
”
S
S
R
N
E
l
e
c
t
r
o
n
i
c
J
o
u
r
n
a
l
,
2
0
2
2
,
d
o
i
:
1
0
.
2
1
3
9
/
ssr
n
.
4
0
8
7
6
4
2
.
[
2
3
]
G
.
Zh
a
n
g
,
M
.
W
a
n
g
,
a
n
d
K
.
Li
u
,
“
F
o
r
e
st
f
i
r
e
s
u
sc
e
p
t
i
b
i
l
i
t
y
m
o
d
e
l
i
n
g
u
si
n
g
a
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
f
o
r
Y
u
n
n
a
n
P
r
o
v
i
n
c
e
o
f
C
h
i
n
a
,
”
I
n
t
e
rn
a
t
i
o
n
a
l
J
o
u
r
n
a
l
o
f
D
i
sa
s
t
e
r
R
i
sk
S
c
i
e
n
c
e
,
v
o
l
.
1
0
,
n
o
.
3
,
p
p
.
3
8
6
–
4
0
3
,
S
e
p
.
2
0
1
9
,
d
o
i
:
1
0
.
1
0
0
7
/
s
1
3
7
5
3
-
0
1
9
-
0
0
2
3
3
-
1.
[
2
4
]
M
.
A
r
r
a
f
i
,
L
.
S
o
ma
n
t
r
i
,
a
n
d
R
.
R
i
d
w
a
n
a
,
“
M
a
p
p
i
n
g
t
h
e
se
v
e
r
i
t
y
l
e
v
e
l
s
o
f
f
o
r
e
s
t
a
n
d
l
a
n
d
f
i
r
e
s
u
si
n
g
t
h
e
n
o
r
m
a
l
i
z
e
d
b
u
r
n
r
a
t
i
o
(
N
B
R
)
a
l
g
o
r
i
t
h
m
o
n
L
a
n
d
s
a
t
8
i
m
a
g
e
r
y
i
n
M
u
a
r
o
Jam
b
i
R
e
g
e
n
c
y
(
i
n
B
a
h
a
s
a
)
,
”
J
u
rn
a
l
G
e
o
s
a
i
n
s
d
a
n
R
e
m
o
t
e
S
e
n
s
i
n
g
,
v
o
l
.
3
,
n
o
.
1
,
p
p
.
1
0
–
1
9
,
M
a
y
2
0
2
2
,
d
o
i
:
1
0
.
2
3
9
6
0
/
j
g
r
s.
2
0
2
2
.
v
3
i
1
.
6
8
.
[
2
5
]
K
.
M
a
h
a
r
a
n
a
,
S
.
M
o
n
d
a
l
,
a
n
d
B
.
N
e
m
a
d
e
,
“
A
r
e
v
i
e
w
:
D
a
t
a
p
r
e
-
p
r
o
c
e
ss
i
n
g
a
n
d
d
a
t
a
a
u
g
me
n
t
a
t
i
o
n
t
e
c
h
n
i
q
u
e
s,”
G
l
o
b
a
l
T
r
a
n
si
t
i
o
n
s
Pro
c
e
e
d
i
n
g
s
,
v
o
l
.
3
,
n
o
.
1
,
p
p
.
9
1
–
9
9
,
Ju
n
.
2
0
2
2
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
g
l
t
p
.
2
0
2
2
.
0
4
.
0
2
0
.
[
2
6
]
J.
-
H
.
R
o
h
,
S
.
-
H
.
M
i
n
,
a
n
d
M
.
-
S
.
K
o
n
g
,
“
A
n
a
l
y
si
s
o
f
f
i
r
e
p
r
e
d
i
c
t
i
o
n
p
e
r
f
o
r
man
c
e
o
f
i
ma
g
e
c
l
a
ssi
f
i
c
a
t
i
o
n
mo
d
e
l
s
b
a
s
e
d
o
n
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
,
”
F
i
re
S
c
i
e
n
c
e
a
n
d
En
g
i
n
e
e
r
i
n
g
,
v
o
l
.
3
6
,
n
o
.
6
,
p
p
.
7
0
–
7
7
,
D
e
c
.
2
0
2
2
,
d
o
i
:
1
0
.
7
7
3
1
/
k
i
f
se
.
9
e
9
0
6
e
7
a
.
[
2
7
]
C
.
I
w
e
n
d
i
,
S
.
K
h
a
n
,
J.
H
.
A
n
a
j
e
m
b
a
,
M
.
M
i
t
t
a
l
,
M
.
A
l
e
n
e
z
i
,
a
n
d
M
.
A
l
a
z
a
b
,
“
T
h
e
u
se
o
f
e
n
sem
b
l
e
mo
d
e
l
s
f
o
r
mu
l
t
i
p
l
e
c
l
a
ss
a
n
d
b
i
n
a
r
y
c
l
a
ss
c
l
a
ssi
f
i
c
a
t
i
o
n
f
o
r
i
m
p
r
o
v
i
n
g
i
n
t
r
u
si
o
n
d
e
t
e
c
t
i
o
n
sy
s
t
e
ms
,
”
S
e
n
s
o
rs
,
v
o
l
.
2
0
,
n
o
.
9
,
p
.
2
5
5
9
,
A
p
r
.
2
0
2
0
,
d
o
i
:
1
0
.
3
3
9
0
/
s
2
0
0
9
2
5
5
9
.
B
I
O
G
RAP
H
I
E
S O
F
AUTH
O
RS
As
sa
d
H
id
a
y
a
t
is
a
m
a
ste
r
stu
d
e
n
t
o
f
t
h
e
Co
m
p
u
ter
S
c
ie
n
c
e
S
tu
d
y
P
r
o
g
ra
m
F
M
IP
A
IP
B
U
n
iv
e
rsit
y
wh
o
is
c
u
rre
n
tl
y
c
o
n
d
u
c
ti
n
g
th
e
sis
re
se
a
rc
h
in
t
h
e
field
o
f
c
o
m
p
u
ter
sc
ien
c
e
a
n
d
fo
re
stry
.
He
p
re
v
i
o
u
sly
wo
rk
e
d
a
s
a
p
ra
c
ti
ti
o
n
e
r
i
n
t
h
e
field
o
f
tec
h
n
o
lo
g
y
a
n
d
in
fo
rm
a
ti
o
n
i
n
o
n
e
o
f
th
e
G
o
v
e
rn
m
e
n
t
Ag
e
n
c
ies
.
He
h
a
s
writt
e
n
o
r
c
o
-
a
u
t
h
o
re
d
se
v
e
ra
l
n
a
ti
o
n
a
ll
y
a
n
d
in
tern
a
ti
o
n
a
ll
y
in
d
e
x
e
d
p
u
b
li
c
a
ti
o
n
s.
His
re
se
a
rc
h
in
tere
sts
i
n
c
lu
d
e
a
rti
ficia
l
in
telli
g
e
n
c
e
a
n
d
d
a
ta m
in
i
n
g
.
He
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
h
i
d
a
y
a
t
a
ss
a
d
@g
m
a
il
.
c
o
m
.
Im
a
s
S
u
k
a
e
s
ih
S
it
a
n
g
g
a
n
g
re
c
e
i
v
e
d
h
e
r
P
h
D
i
n
c
o
m
p
u
ter
sc
ie
n
c
e
fr
o
m
t
h
e
F
a
c
u
l
t
y
o
f
C
o
m
p
u
ter
S
c
ie
n
c
e
a
n
d
I
n
f
o
rm
a
t
i
o
n
T
e
c
h
n
o
l
o
g
y
,
U
n
i
v
e
r
sit
i
P
u
t
ra
M
a
la
y
s
ia,
i
n
2
0
1
3
.
S
h
e
is
a
lec
t
u
re
r
i
n
S
c
h
o
o
l
o
f
D
a
ta
S
c
ie
n
c
e
,
M
a
th
e
m
a
t
ics
a
n
d
In
f
o
rm
a
t
ics
a
t
I
P
B
U
n
i
v
e
rs
it
y
,
In
d
o
n
e
s
ia
.
He
r
m
a
i
n
re
se
a
rc
h
i
n
te
re
st
s
i
n
c
l
u
d
e
s
p
a
ti
a
l
d
a
ta
m
i
n
i
n
g
,
m
a
c
h
i
n
e
lea
r
n
i
n
g
,
a
n
d
d
a
ta
wa
re
h
o
u
si
n
g
.
S
h
e
c
a
n
b
e
c
o
n
tac
te
d
a
t
e
m
a
il
:
ima
s.
si
ta
n
g
g
a
n
g
@a
p
p
s
.
i
p
b
.
a
c
.
i
d
.
La
il
a
n
S
y
a
u
fin
a
is
a
p
ro
fe
ss
o
r
in
fo
re
st
p
ro
tec
ti
o
n
wit
h
fo
re
st
fire
a
s
h
e
r
m
a
jo
r
a
t
De
p
a
rtme
n
t
o
f
S
il
v
ic
u
lt
u
re
F
a
c
u
lt
y
o
f
F
o
re
str
y
a
n
d
E
n
v
ir
o
n
m
e
n
t
IP
B.
He
r
u
n
d
e
rg
ra
d
u
a
te
fro
m
Bo
g
o
r
Ag
ricu
l
tu
ra
l
Un
iv
e
rsity
(I
P
B),
M
a
ste
r
d
e
g
re
e
fro
m
G
e
o
rg
Au
g
u
st
U
n
iv
e
rsit
y
G
e
rm
a
n
y
a
n
d
P
h
.
D.
fr
o
m
Un
i
v
e
rsiti
P
u
tra
M
a
lay
sia
.
He
r
field
o
f
i
n
tere
st
in
c
lu
d
in
g
fo
re
st
f
ire
in
th
e
a
sp
e
c
ts
o
f
fire
se
v
e
rit
y
a
ss
e
ss
m
e
n
t,
fire
m
a
n
a
g
e
m
e
n
t,
p
e
a
tl
a
n
d
fire,
fire
-
b
io
d
i
v
e
rsity
,
f
ire
-
c
li
m
a
te,
a
n
d
fire
-
e
m
issio
n
.
At
t
h
e
n
a
ti
o
n
a
l
lev
e
l,
s
h
e
h
a
s
b
e
e
n
a
re
so
u
rc
e
p
e
rso
n
o
f
M
o
EF
fo
r
fo
re
st
fire
a
n
d
p
e
a
tl
a
n
d
re
late
d
p
o
li
c
ies
si
n
c
e
2
0
0
2
.
S
h
e
h
a
s
e
x
p
e
rien
c
e
d
wit
h
in
tern
a
t
io
n
a
l
c
o
ll
a
b
o
ra
ti
o
n
,
a
s
re
so
u
rc
e
p
e
rso
n
a
n
d
e
v
a
lu
a
t
o
r
fo
r
JICA
fo
re
st
f
ire
p
ro
jec
t
a
n
d
h
a
s
b
e
e
n
a
n
a
ti
o
n
a
l
e
x
p
e
rt
f
o
r
AS
EAN
P
e
a
tl
a
n
d
F
o
re
st
P
r
o
jec
t
(
2
0
1
0
-
2
0
1
4
),
a
n
d
a
s
a
re
se
a
rc
h
c
o
ll
a
b
o
ra
t
o
r
o
n
p
e
a
tl
a
n
d
fire
re
se
a
rc
h
u
n
d
e
r
NEW
TON
F
UN
D
sc
h
e
m
e
with
Lee
d
s
Un
iv
e
rsity
,
UK
(2
0
1
8
-
2
0
1
9
),
NA
P
C
(Ne
two
rk
e
d
ASE
AN
P
e
a
t
S
wa
m
p
F
o
re
st
Co
m
m
u
n
it
ies
-
2
0
1
9
-
2
0
2
1
),
a
n
d
Ne
t
-
P
e
a
t:
Ne
two
rk
e
d
ASE
AN
P
e
a
tl
a
n
d
Co
m
m
u
n
it
ies
fo
r
Tran
sb
o
u
n
d
a
ry
Ha
z
e
Ale
rt
(2
0
2
2
)
wit
h
UP
M
,
UTB
a
n
d
NIC
T.
S
h
e
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
lailan
s@
a
p
p
s.i
p
b
.
a
c
.
i
d
.
Evaluation Warning : The document was created with Spire.PDF for Python.