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.
15
,
No
.
6
,
Decem
b
er
20
25
,
p
p
.
5
3
2
7
~
5
3
3
5
I
SS
N:
2088
-
8
7
0
8
,
DOI
: 1
0
.
1
1
5
9
1
/ijece.
v
15
i
6
.
pp
5
3
2
7
-
5
3
3
5
5327
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
ec
e.
ia
esco
r
e.
co
m
Enha
ncing
sem
a
ntic
seg
men
tatio
n
with
a
boun
da
ry
-
sens
itive
lo
ss
f
unction
:
a
n
o
v
el
a
ppro
a
ch
G
a
nes
h R.
P
a
da
lk
a
r
,
M
a
dh
uri B.
K
ha
m
bet
e
D
e
p
a
r
t
me
n
t
o
f
El
e
c
t
r
o
n
i
c
s a
n
d
T
e
l
e
c
o
mm
u
n
i
c
a
t
i
o
n
s
,
M
K
S
S
S
’
s
C
u
mm
i
n
s
C
o
l
l
e
g
e
o
f
E
n
g
i
n
e
e
r
i
n
g
f
o
r
W
o
m
e
n
,
P
u
n
e
,
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
Sep
1
2
,
2
0
2
4
R
ev
is
ed
J
u
l 1
9
,
2
0
2
5
Acc
ep
ted
Sep
1
4
,
2
0
2
5
S
e
m
a
n
ti
c
se
g
m
e
n
tatio
n
is
c
r
u
c
ial
ste
p
i
n
a
u
t
o
n
o
m
o
u
s
d
r
iv
i
n
g
,
m
e
d
ica
l
ima
g
in
g
,
a
n
d
sc
e
n
e
u
n
d
e
rsta
n
d
in
g
.
Tra
d
it
i
o
n
a
l
a
p
p
r
o
a
c
h
e
s
l
e
v
e
ra
g
in
g
m
a
n
u
a
ll
y
e
x
trac
ted
p
i
x
e
l
p
r
o
p
e
rti
e
s
a
n
d
p
ro
b
a
b
il
isti
c
m
o
d
e
ls,
h
a
v
e
a
c
h
iev
e
d
re
a
so
n
a
b
le
p
e
rf
o
rm
a
n
c
e
b
u
t
su
ffe
r
fro
m
li
m
it
e
d
g
e
n
e
ra
li
z
a
ti
o
n
a
n
d
th
e
n
e
e
d
fo
r
e
x
p
e
rt
-
d
ri
v
e
n
fe
a
tu
re
se
lec
ti
o
n
.
Th
e
r
ise
o
f
d
e
e
p
lea
rn
i
n
g
a
rc
h
it
e
c
tu
re
s
h
a
s
sig
n
ifi
c
a
n
tl
y
im
p
ro
v
e
d
se
g
m
e
n
tatio
n
a
c
c
u
ra
c
y
b
y
e
n
a
b
li
n
g
a
u
to
m
a
ti
c
fe
a
tu
re
e
x
trac
ti
o
n
a
n
d
c
a
p
tu
ri
n
g
in
tri
c
a
te
o
b
jec
t
d
e
tails.
Ho
we
v
e
r,
th
e
se
me
th
o
d
s
sti
ll
fa
c
e
c
h
a
ll
e
n
g
e
s,
in
c
l
u
d
i
n
g
th
e
n
e
e
d
f
o
r
lar
g
e
d
a
tas
e
ts,
e
x
ten
siv
e
h
y
p
e
r
p
a
ra
m
e
ter
tu
n
in
g
,
a
n
d
c
a
r
e
fu
l
lo
ss
f
u
n
c
ti
o
n
se
lec
ti
o
n
.
T
h
is
p
a
p
e
r
p
ro
p
o
se
s
a
n
o
v
e
l
b
o
u
n
d
a
r
y
-
se
n
si
ti
v
e
lo
ss
f
u
n
c
ti
o
n
,
wh
ich
c
o
m
b
in
e
s
re
g
io
n
lo
ss
a
n
d
b
o
u
n
d
a
ry
lo
ss
,
t
o
e
n
h
a
n
c
e
b
o
th
re
g
i
o
n
c
o
n
siste
n
c
y
a
n
d
e
d
g
e
d
e
li
n
e
a
ti
o
n
i
n
se
g
m
e
n
tatio
n
tas
k
s.
Im
p
lem
e
n
ted
wi
th
i
n
a
m
o
d
if
i
e
d
S
e
g
Ne
t
fra
m
e
wo
rk
,
th
e
a
p
p
ro
a
c
h
p
ro
p
o
se
d
in
th
e
p
a
p
e
r
is
e
v
a
l
u
a
ted
with
t
h
e
se
m
a
n
ti
c
b
o
u
n
d
a
r
y
d
a
tas
e
t
(
S
B
D
)
d
a
tas
e
t
u
si
n
g
sta
n
d
a
r
d
se
g
m
e
n
tatio
n
m
e
tri
c
s.
Ex
p
e
r
ime
n
tal
re
su
lt
s
i
n
d
ica
te
imp
r
o
v
e
d
se
g
m
e
n
tati
o
n
a
c
c
u
ra
c
y
,
su
b
sta
n
ti
a
ti
n
g
to
p
r
o
p
o
se
d
m
e
th
o
d
.
K
ey
w
o
r
d
s
:
B
o
u
n
d
ar
y
r
ef
in
em
e
n
t n
etwo
r
k
s
C
o
n
d
itio
n
al
r
an
d
o
m
f
ield
s
C
o
n
v
o
lu
tio
n
al
n
eu
r
al
n
etwo
r
k
s
Ma
r
k
o
v
r
an
d
o
m
f
ield
s
Py
r
am
id
s
ce
n
e
p
ar
s
in
g
n
etwo
r
k
R
eg
io
n
s
with
co
n
v
o
lu
ti
o
n
al
n
eu
r
al
n
etwo
r
k
s
Sem
an
tic
b
o
u
n
d
ar
y
d
etec
tio
n
d
ataset
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
:
Gan
esh
R
.
Pad
alk
ar
D
e
p
a
r
t
m
e
n
t
o
f
E
l
ec
t
r
o
n
i
c
s
a
n
d
T
e
l
e
c
o
m
m
u
n
i
c
a
ti
o
n
s
,
MK
SS
S
’
s
C
u
m
m
i
n
s
C
o
l
l
e
g
e
o
f
E
n
g
i
n
e
er
i
n
g
f
o
r
W
o
m
e
n
C
u
m
m
in
s
C
o
lleg
e
R
d
,
Kar
v
en
ag
ar
,
Pu
n
e,
Ma
h
ar
ash
tr
a
4
1
1
0
5
2
,
I
n
d
ia
E
m
ail:
g
an
esh
.
p
ad
al
k
ar
@
cu
m
m
in
s
co
lleg
e.
in
1.
I
NT
RO
D
UCT
I
O
N
Sem
an
tic
s
eg
m
en
tatio
n
is
a
cr
u
cial
s
tep
o
f
g
iv
in
g
a
lab
el
to
im
ag
e
p
ix
els
f
r
o
m
p
r
e
d
ef
in
ed
ca
teg
o
r
ies
[
1
]
.
I
t
h
as
n
u
m
er
o
u
s
ap
p
licatio
n
s
[
2
]
–
[
7
]
.
T
r
ad
itio
n
al
m
eth
o
d
s
r
elied
o
n
ex
tr
ac
ted
f
ea
tu
r
es,
co
m
b
in
ed
with
p
r
o
b
a
b
ilis
tic
m
o
d
els
co
n
d
itio
n
al
r
an
d
o
m
f
ield
s
(
C
R
Fs
)
[
8
]
a
n
d
Ma
r
k
o
v
r
an
d
o
m
f
ield
s
(
MRF
s
)
[
9
]
.
Ho
wev
er
,
s
u
ch
ap
p
r
o
ac
h
es
s
tr
u
g
g
le
with
g
e
n
er
aliza
tio
n
,
p
ar
ticu
lar
ly
wh
en
d
ea
lin
g
with
v
ar
iatio
n
s
in
lig
h
tin
g
,
o
b
ject
ap
p
ea
r
an
ce
,
an
d
o
r
ie
n
tatio
n
ac
r
o
s
s
d
atasets
[
1
0
]
.
Ad
d
i
tio
n
ally
,
h
a
n
d
-
cr
a
f
ted
f
ea
tu
r
es
ar
e
f
ix
ed
,
tim
e
-
co
n
s
u
m
in
g
to
s
elec
t,
an
d
in
ef
f
ec
tiv
e
f
o
r
n
o
v
el
o
r
u
n
s
ee
n
d
ata
[
1
1
]
.
W
h
ile
p
r
o
b
ab
ilis
tic
m
o
d
els
h
elp
m
o
d
el
s
p
atial
d
ep
en
d
en
cies,
th
ey
f
ac
e
d
if
f
icu
lties
h
an
d
lin
g
n
o
n
-
lin
ea
r
,
h
ig
h
-
d
im
en
s
io
n
al
r
elatio
n
s
h
ip
s
,
m
ak
in
g
th
em
in
ef
f
icien
t
f
o
r
lar
g
e
-
s
ca
le,
h
ig
h
-
r
eso
lu
tio
n
im
a
g
es
[
1
2
]
.
Du
e
to
th
eir
lo
ca
l
f
o
cu
s
,
tr
ad
itio
n
a
l
m
eth
o
d
s
o
f
ten
f
ail
to
ca
p
tu
r
e
th
e
b
r
o
ad
er
im
a
g
e
co
n
tex
t,
lead
in
g
to
s
u
b
o
p
ti
m
al
s
eg
m
en
tatio
n
,
esp
ec
ially
f
o
r
f
in
e
d
etails
o
r
co
m
p
lex
o
b
ject
b
o
u
n
d
a
r
ies
[
1
3
]
.
T
h
e
em
er
g
e
n
ce
o
f
d
ee
p
ar
c
h
itectu
r
es
h
as
s
h
if
ted
p
er
f
o
r
m
an
ce
f
o
r
o
b
tain
in
g
s
em
an
tic
s
eg
m
en
tatio
n
to
h
ig
h
e
r
lev
el
b
y
m
ak
in
g
p
o
s
s
ib
le
to
in
f
er
b
o
t
h
lo
w
-
a
n
d
h
i
g
h
-
lev
el
f
ea
tu
r
es
a
u
to
m
atica
lly
[
1
]
.
Dee
p
ar
ch
itectu
r
es
av
o
id
h
u
m
a
n
ly
d
ec
id
in
g
an
d
ex
tr
ac
tin
g
th
e
f
ea
tu
r
es,
en
h
an
cin
g
s
ca
lab
ilit
y
ac
r
o
s
s
d
iv
er
s
e
d
o
m
ain
s
[
1
4
]
.
T
h
e
ad
v
a
n
ce
m
en
t
m
o
d
els
[
1
]
,
[
3
]
,
[
5
]
,
[
6
]
,
[
1
5
]
,
[
1
6
]
h
as
led
to
im
p
r
o
v
e
d
s
eg
m
en
tatio
n
p
r
ec
is
io
n
an
d
ef
f
icien
c
y
.
Ho
wev
er
,
tr
ain
in
g
th
ese
m
o
d
els
p
r
esen
ts
ch
allen
g
es
s
u
ch
as
th
e
n
ee
d
f
o
r
lar
g
e
d
atasets
,
ex
ten
s
iv
e
h
y
p
er
p
ar
am
eter
tu
n
in
g
,
a
n
d
h
ig
h
co
m
p
u
tatio
n
al
r
eso
u
r
ce
s
.
R
ese
ar
ch
er
s
co
n
tin
u
e
to
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.
15
,
No
.
6
,
Decem
b
e
r
20
25
:
5
3
2
7
-
5
3
3
5
5328
ex
p
lo
r
e
way
s
to
ad
d
r
ess
th
ese
ch
allen
g
es,
in
clu
d
in
g
m
eth
o
d
s
th
at
ac
h
iev
e
ac
cu
r
ate
s
eg
m
en
tatio
n
with
lim
ited
d
ata
[
1
7
]
an
d
s
tr
ateg
ies f
o
r
o
p
tim
izin
g
tr
ain
in
g
ef
f
icien
cy
[
1
8
]
.
Dev
elo
p
m
en
t
o
f
s
em
an
tic
s
eg
m
en
tatio
n
m
o
d
el
is
p
o
s
s
ib
le
b
y
tr
ain
in
g
a
m
u
lti
-
lay
er
n
etwo
r
k
s
[
1
9
]
.
T
h
e
m
o
d
el
lear
n
s
h
ier
ar
c
h
ical
f
ea
tu
r
es,
with
th
e
p
r
im
a
r
y
o
b
jectiv
e
o
f
o
p
tim
izin
g
p
ar
am
eter
s
—
s
u
ch
as
weig
h
ts
an
d
b
iases
—
to
g
en
er
alize
well
o
n
u
n
s
ee
n
d
ata
[
2
0
]
.
B
ac
k
p
r
o
p
a
g
atio
n
co
m
p
u
tes
g
r
ad
ien
ts
b
ased
o
n
th
e
lo
s
s
f
u
n
ctio
n
,
g
u
id
in
g
p
ar
am
eter
ad
ju
s
tm
en
ts
to
m
in
im
i
ze
p
r
ed
ictio
n
e
r
r
o
r
s
[
1
0
]
,
[
2
1
]
.
T
h
e
ef
f
ec
tiv
e
n
ess
o
f
m
o
d
el
tr
ain
in
g
d
e
p
en
d
s
o
n
m
u
ltip
le
f
ac
to
r
s
,
in
cl
u
d
in
g
n
et
wo
r
k
ar
c
h
itectu
r
e,
d
ata
q
u
alit
y
,
an
d
o
p
tim
izatio
n
s
tr
ateg
ies
[
2
0
]
.
L
o
s
s
f
u
n
ctio
n
s
,
in
p
a
r
ticu
lar
,
a
r
e
cr
itical,
a
s
th
ey
d
i
r
ec
tly
in
f
lu
en
ce
p
ar
a
m
eter
u
p
d
ates
an
d
m
o
d
el
p
er
f
o
r
m
a
n
ce
[
2
2
]
.
Sem
an
tic
s
eg
m
en
tatio
n
r
eq
u
ir
es
n
o
t
o
n
ly
ac
cu
r
ate
r
eg
io
n
cl
ass
if
icatio
n
b
u
t
also
p
r
ec
is
e
d
elin
ea
tio
n
o
f
o
b
ject
b
o
u
n
d
ar
ies.
C
lear
ly
d
ef
in
ed
b
o
u
n
d
ar
ies
ar
e
es
s
en
tial
f
o
r
d
is
tin
g
u
is
h
in
g
ad
jace
n
t
r
eg
io
n
s
an
d
im
p
r
o
v
in
g
o
b
ject
d
etec
tio
n
ac
cu
r
ac
y
[
2
3
]
,
[
2
4
]
.
Ad
d
itio
n
ally
,
b
o
u
n
d
ar
ies
o
f
ten
ex
is
t
b
etwe
en
s
im
ilar
r
eg
io
n
s
,
m
ak
in
g
t
h
em
d
i
f
f
icu
lt
to
s
eg
m
en
t
ac
cu
r
ately
.
Ad
d
r
ess
in
g
th
ese
ch
allen
g
es
is
ess
en
tial
f
o
r
im
p
r
o
v
i
n
g
d
ee
p
lear
n
in
g
m
o
d
els’
ab
ilit
y
to
ca
p
tu
r
e
b
o
th
r
eg
io
n
al
an
d
b
o
u
n
d
a
r
y
in
f
o
r
m
atio
n
e
f
f
ec
tiv
ely
.
T
h
e
r
esear
ch
o
n
s
em
an
tic
b
o
u
n
d
ar
y
s
eg
m
en
tatio
n
in
v
iew
r
ef
in
in
g
o
b
ject
b
o
u
n
d
ar
ies
an
d
s
eg
m
en
tatio
n
in
co
m
p
le
x
s
ce
n
ar
io
s
led
to
in
tr
o
d
u
ce
a
b
o
u
n
d
ar
y
-
awa
r
e
d
ee
p
lear
n
i
n
g
m
o
d
el
[
2
5
]
th
at
im
p
r
o
v
e
d
s
eg
m
en
tatio
n
ac
c
u
r
ac
y
,
esp
ec
ially
at
o
b
ject
b
o
u
n
d
ar
ies.
A
lig
h
tweig
h
t
n
etwo
r
k
[
2
6
]
th
at
lev
er
ag
es
b
o
u
n
d
ar
y
-
awa
r
e
lear
n
in
g
to
b
o
o
s
t
p
er
f
o
r
m
a
n
ce
o
n
d
atasets
lik
e
C
ity
s
ca
p
es
an
d
ADE
2
0
K.
A
s
em
i
-
s
u
p
er
v
is
ed
ap
p
r
o
ac
h
th
at
a
d
ap
ts
to
s
tr
u
ct
u
r
ed
o
u
tp
u
t
s
p
ac
es,
im
p
r
o
v
in
g
b
o
u
n
d
a
r
y
d
elin
ea
tio
n
with
l
im
ited
lab
eled
d
ata
[
2
7
]
.
T
h
e
r
esear
ch
er
s
f
o
cu
s
ed
o
n
a
m
u
lti
-
s
ca
le
f
u
s
io
n
n
etwo
r
k
[
2
8
]
a
n
d
s
em
a
n
tic
h
ier
ar
c
h
y
-
awa
r
e
m
o
d
el
[
2
9
]
,
wh
ich
en
h
a
n
ce
s
b
o
u
n
d
ar
y
p
r
e
cisi
o
n
b
y
u
tili
zin
g
h
ier
ar
ch
ical
r
elatio
n
s
h
ip
s
b
etwe
en
o
b
ject
class
es.
T
h
is
r
esear
ch
aim
s
to
en
h
an
ce
s
em
an
tic
s
eg
m
en
tatio
n
b
y
in
tr
o
d
u
cin
g
a
b
o
u
n
d
ar
y
-
s
en
s
itiv
e
lo
s
s
f
u
n
ctio
n
,
wh
ich
in
teg
r
ates
r
eg
io
n
lo
s
s
an
d
b
o
u
n
d
ar
y
lo
s
s
.
T
r
ad
itio
n
al
tr
ain
in
g
m
eth
o
d
s
p
r
ed
o
m
in
an
tly
em
p
h
asize
r
eg
i
o
n
l
o
s
s
,
wh
ich
,
wh
ile
e
f
f
ec
tiv
e
f
o
r
o
v
er
all
s
eg
m
en
tatio
n
,
m
ay
o
v
er
lo
o
k
p
r
ec
is
e
b
o
u
n
d
ar
y
alig
n
m
en
t.
I
n
co
r
p
o
r
atin
g
b
o
u
n
d
ar
y
l
o
s
s
in
to
th
e
tr
ain
in
g
p
r
o
ce
s
s
en
ab
les
m
o
d
els
to
r
ef
i
n
e
th
eir
p
r
e
d
ictio
n
s
,
en
s
u
r
in
g
m
o
r
e
ac
cu
r
ate
s
eg
m
en
tatio
n
o
f
o
b
ject
ed
g
es.
T
h
is
b
o
u
n
d
ar
y
-
awa
r
e
a
p
p
r
o
ac
h
en
h
an
ce
s
p
er
f
o
r
m
an
ce
in
co
m
p
lex
s
ce
n
es
wh
er
e
b
o
u
n
d
ar
y
d
etails
ar
e
cr
itical
f
o
r
ac
cu
r
ate
s
eg
m
en
tatio
n
[
3
0
]
–
[
3
2
]
.
T
h
is
r
esear
ch
wo
r
k
p
r
o
p
o
s
es
a
n
o
v
el
b
o
u
n
d
ar
y
-
s
en
s
itiv
e
lo
s
s
f
u
n
ctio
n
d
esig
n
ed
to
im
p
r
o
v
e
b
o
th
s
em
an
tic
r
eg
io
n
a
n
d
b
o
u
n
d
ar
y
s
eg
m
e
n
tatio
n
.
B
y
jo
in
tly
o
p
tim
izin
g
f
o
r
r
eg
io
n
class
if
icatio
n
an
d
b
o
u
n
d
ar
y
alig
n
m
en
t,
o
u
r
ap
p
r
o
ac
h
en
h
an
ce
s
s
eg
m
en
tatio
n
ac
cu
r
ac
y
an
d
d
elin
ea
tio
n
p
r
ec
is
io
n
.
T
h
e
m
o
d
el
ev
alu
atio
n
i
s
p
er
f
o
r
m
ed
u
s
in
g
p
u
b
lically
a
v
ailab
le
im
a
g
e
d
ataset,
d
em
o
n
s
tr
atin
g
th
at
o
u
r
b
o
u
n
d
ar
y
-
awa
r
e
lo
s
s
f
u
n
ctio
n
s
ig
n
if
ica
n
tly
im
p
r
o
v
es seg
m
en
tatio
n
p
er
f
o
r
m
an
ce
co
m
p
ar
ed
to
co
n
v
en
tio
n
al
lo
s
s
f
u
n
ctio
n
s
[
3
3
]
,
[
3
4
]
.
T
h
e
p
ap
er
is
s
tr
u
ctu
r
ed
as:
Sectio
n
2
p
r
o
v
id
es
an
ex
te
n
s
iv
e
liter
atu
r
e.
Sectio
n
3
o
u
tlin
es
th
e
p
r
o
p
o
s
ed
m
eth
o
d
.
Sectio
n
4
is
d
ed
icate
d
f
o
r
th
e
r
esu
lts
an
d
d
is
cu
s
s
io
n
.
Sectio
n
5
c
o
n
clu
d
es
with
s
u
m
m
ar
y
o
f
co
n
tr
ib
u
tio
n
s
an
d
f
u
tu
r
e
r
esear
ch
d
ir
ec
tio
n
s
.
2.
CO
M
P
RE
H
E
NS
I
V
E
T
H
E
O
RO
T
I
CA
L
B
ASI
S
Sem
an
tic
s
eg
m
en
tatio
n
ass
ig
n
s
ea
ch
p
ix
el
with
a
lab
el
lean
e
d
f
o
r
m
t
h
e
g
r
o
u
n
d
t
r
u
th
d
ata,
aim
in
g
f
o
r
ac
cu
r
ate
s
eg
m
en
tatio
n
.
R
esear
ch
in
th
is
ar
ea
d
ir
ec
ted
to
e
n
h
an
ce
s
eg
m
en
tatio
n
p
er
f
o
r
m
an
ce
.
Pix
el
-
b
ased
s
eg
m
en
tatio
n
tech
n
iq
u
es
in
clu
d
e
th
r
esh
o
ld
in
g
-
b
ased
m
eth
o
d
s
,
r
eg
io
n
g
r
o
win
g
,
r
eg
io
n
s
p
litt
in
g
an
d
m
er
g
in
g
,
wate
r
s
h
ed
tr
an
s
f
o
r
m
,
g
r
a
p
h
c
u
ts
,
an
d
m
ea
n
-
s
h
if
t
[
3
5
]
–
[
3
7
]
.
Ho
wev
er
,
r
esu
lts
f
r
o
m
p
ix
e
l
-
b
ased
an
d
b
lo
ck
-
b
ased
m
eth
o
d
s
m
ay
n
o
t
alwa
y
s
b
e
s
u
f
f
icien
t
f
o
r
co
m
p
le
x
co
m
p
u
te
r
v
is
io
n
ap
p
licatio
n
s
.
T
o
ad
d
r
ess
th
is
,
s
em
an
tic
s
eg
m
en
tatio
n
ca
n
b
e
ac
h
iev
ed
th
r
o
u
g
h
b
o
th
s
u
p
er
v
is
ed
an
d
s
em
i
-
s
u
p
er
v
is
ed
m
eth
o
d
s
.
T
h
e
alg
o
r
ith
m
s
ca
n
b
e
tr
ai
n
ed
o
n
ex
tr
ac
ted
f
ea
tu
r
es
a
n
d
c
o
r
r
esp
o
n
d
in
g
lab
els
to
p
er
f
o
r
m
ef
f
ec
tiv
e
s
em
an
tic
s
eg
m
en
tatio
n
[
3
8
]
,
[
3
9
]
.
T
h
e
tr
an
s
itio
n
f
r
o
m
tr
ad
itio
n
a
l
m
eth
o
d
s
to
d
ee
p
ar
ch
itectu
r
e
-
b
ased
s
em
an
tic
s
eg
m
en
tatio
n
is
d
r
iv
en
b
y
th
e
s
u
p
er
io
r
ac
cu
r
ac
y
an
d
ef
f
icien
cy
th
at
t
h
ey
ca
n
ac
h
ie
v
e
in
s
im
p
lify
in
g
th
e
co
n
ten
ts
in
th
e
im
a
g
e.
Dee
p
n
etwo
r
k
s
h
av
e
r
ev
o
lu
tio
n
ize
d
s
em
an
tic
s
eg
m
en
tatio
n
b
y
en
a
b
lin
g
ac
cu
r
ate
p
ix
el
-
lev
el
class
if
icatio
n
in
im
ag
es
[
4
0
]
.
Key
ar
c
h
itectu
r
es
in
t
h
is
f
ield
in
clu
d
e
f
u
lly
co
n
v
o
lu
ti
o
n
al
n
etwo
r
k
s
(
FC
Ns),
an
d
v
ar
ian
ts
lik
e
FC
N
-
8
,
U
-
Net
an
d
Dee
p
L
ab
[
6
]
.
Seg
Net
[
3
1
]
,
[
3
2
]
an
o
th
er
ef
f
ec
tiv
e
ar
ch
itectu
r
e,
in
co
r
p
o
r
ates
B
a
y
esian
in
f
er
en
ce
f
o
r
h
an
d
lin
g
u
n
ce
r
tain
ty
,
d
em
o
n
s
t
r
atin
g
h
ig
h
ac
cu
r
ac
y
o
n
th
e
C
a
m
Vid
d
ataset
[
7
]
.
Oth
er
in
f
lu
e
n
tial m
o
d
els in
th
e
f
ield
in
clu
d
e
PS
PNet,
Ma
s
k
R
-
C
NN,
HR
Net,
an
d
E
f
f
ic
ien
tNet,
ea
ch
co
n
tr
ib
u
tin
g
t
o
ad
v
a
n
ce
m
en
ts
in
s
em
an
tic
s
eg
m
en
tatio
n
th
r
o
u
g
h
d
ee
p
lear
n
in
g
tec
h
n
iq
u
es
[
1
5
]
.
I
n
d
ee
p
ar
c
h
itectu
r
es,
f
ea
tu
r
e
ex
tr
ac
tio
n
in
v
o
lv
es
p
r
o
ce
s
s
in
g
in
p
u
t
im
ag
es
th
r
o
u
g
h
m
u
ltip
le
lay
er
s
to
ex
tr
ac
t
h
ig
h
-
le
v
el
f
ea
tu
r
es.
T
h
ese
f
ea
tu
r
es
p
r
o
p
ag
ate
d
th
r
o
u
g
h
th
e
n
etwo
r
k
to
g
en
e
r
ate
th
e
f
in
al
s
eg
m
en
tatio
n
m
ap
.
Stan
d
ar
d
f
ea
tu
r
e
p
r
o
p
a
g
atio
n
m
eth
o
d
s
m
ay
l
o
s
e
cr
itical
d
etails,
p
ar
ticu
lar
ly
ar
o
u
n
d
o
b
ject
b
o
u
n
d
ar
ies,
lead
in
g
to
in
ac
cu
r
ate
s
eg
m
en
t
atio
n
r
esu
lts
.
So
m
e
ap
p
r
o
ac
h
e
s
in
v
o
lv
e
esti
m
atin
g
b
o
u
n
d
ar
i
es
as
a
s
ep
ar
ate
task
b
u
t
in
tr
o
d
u
ctio
n
o
f
b
o
u
n
d
ar
y
-
awa
r
e
f
ea
tu
r
es,
en
h
a
n
cin
g
t
h
e
p
r
eser
v
atio
n
o
f
e
d
g
es
an
d
im
p
r
o
v
in
g
o
v
e
r
all
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
E
n
h
a
n
cin
g
s
ema
n
tic
s
eg
men
t
a
tio
n
w
ith
a
b
o
u
n
d
a
r
y
-
s
en
s
itive
…
(
Ga
n
esh
R
.
P
a
d
a
lka
r
)
5329
s
eg
m
en
tatio
n
q
u
ality
[
1
]
,
[
1
5
]
.
T
h
e
au
th
o
r
s
d
em
o
n
s
tr
ate
im
p
r
o
v
ed
p
er
f
o
r
m
an
ce
o
n
s
ev
er
al
b
en
ch
m
ar
k
d
atasets
.
B
is
ch
k
e
et
a
l.
[
4
1
]
p
r
o
p
o
s
es
an
ap
p
r
o
ac
h
,
wh
er
e
th
e
n
etwo
r
k
s
im
u
ltan
eo
u
s
ly
lear
n
s
to
p
er
f
o
r
m
s
eg
m
en
tin
g
th
e
b
u
ild
in
g
f
o
o
t
p
r
in
ts
an
d
p
r
ed
ictin
g
th
e
b
o
u
n
d
ar
y
ed
g
es.
I
n
co
r
p
o
r
atin
g
b
o
u
n
d
a
r
y
in
f
o
r
m
atio
n
im
p
r
o
v
es
ass
is
t
m
o
d
el
to
m
a
r
k
clo
s
ely
p
ac
k
ed
b
u
ild
i
n
g
s
an
d
th
eir
b
o
u
n
d
ar
ies,
r
esu
ltin
g
in
b
etter
o
v
er
all
p
er
f
o
r
m
an
ce
.
T
h
e
ap
p
r
o
ac
h
was
test
ed
o
n
lar
g
e
-
s
ca
le
d
atasets
,
s
h
o
win
g
s
ig
n
if
ican
t
im
p
r
o
v
e
m
en
ts
o
v
er
tr
ad
itio
n
al
m
eth
o
d
s
[
4
1
]
.
T
h
e
co
n
v
o
lu
tio
n
al
lay
er
s
ar
e
in
ev
itab
le
p
ar
t
o
f
co
n
v
o
lu
tio
n
a
l
n
eu
r
al
n
etwo
r
k
s
(
C
NNs)
[
4
2
]
.
T
h
o
u
g
h
th
e
lay
er
s
ex
tr
ac
t
th
e
in
f
o
r
m
a
tio
n
o
n
ev
er
y
s
tag
e
o
f
th
e
n
e
two
r
k
,
th
ese
ca
n
b
l
u
r
o
r
lo
s
e
f
in
e
d
etails
at
th
e
ed
g
es
b
etwe
en
d
if
f
er
e
n
t
r
e
g
io
n
s
o
r
o
b
jects,
wh
ic
h
ca
n
l
ea
d
to
i
n
ac
cu
r
ac
ies
in
s
eg
m
en
tin
g
o
b
jects
with
co
m
p
lex
b
o
u
n
d
ar
ies
[
4
2
]
.
E
d
g
e
-
awa
r
e
co
n
v
o
lu
tio
n
s
ar
e
d
e
s
ig
n
ed
to
ad
d
r
ess
th
is
lim
itat
io
n
b
y
s
p
ec
if
ically
in
co
r
p
o
r
atin
g
in
f
o
r
m
atio
n
ab
o
u
t e
d
g
es o
r
b
o
u
n
d
ar
ies d
u
r
in
g
th
e
co
n
v
o
lu
tio
n
p
r
o
ce
s
s
[
4
3
]
.
T
h
ey
p
r
eser
v
e
ed
g
e
d
etails.
B
o
u
n
d
ar
y
r
ef
in
em
en
t
n
etwo
r
k
s
(
B
R
Ns)
ar
e
ad
v
an
ce
d
n
eu
r
al
n
etwo
r
k
ar
c
h
itectu
r
es
d
esig
n
ed
to
im
p
r
o
v
e
th
e
o
b
ject
b
o
u
n
d
ar
y
d
elin
ea
tio
n
[
4
4
]
.
Su
ch
n
etwo
r
k
s
s
p
ec
if
ically
f
o
cu
s
o
n
r
ef
in
in
g
th
e
ed
g
es
an
d
b
o
u
n
d
ar
ies o
f
s
eg
m
e
n
ted
o
b
je
cts to
p
r
o
d
u
ce
clea
r
er
a
n
d
m
o
r
e
p
r
ec
is
e
s
eg
m
en
tatio
n
r
esu
lts
[
4
4
]
.
Pix
el
class
if
ica
tio
n
ac
cu
r
ac
y
d
ep
en
d
s
o
n
ca
p
tu
r
in
g
f
in
e
d
etails
at
o
b
ject
ed
g
es
[
4
5
]
.
T
h
is
lim
itatio
n
b
ec
o
m
es
p
ar
ticu
lar
ly
p
r
o
b
lem
atic
in
im
ag
es
with
co
m
p
lex
s
ce
n
es,
wh
er
e
o
b
jects
h
av
e
in
tr
icate
b
o
u
n
d
ar
ies,
o
r
wh
en
d
if
f
er
en
t
class
es
ar
e
clo
s
ely
p
ac
k
ed
to
g
eth
e
r
.
B
o
u
n
d
a
r
y
-
b
ased
l
o
s
s
f
u
n
ctio
n
s
s
p
ec
if
ically
ad
d
r
ess
th
is
is
s
u
e
b
y
em
p
h
asizin
g
th
e
im
p
o
r
tan
ce
o
f
c
o
r
r
ec
t
b
o
u
n
d
ar
y
p
r
ed
ictio
n
[
4
5
]
.
T
h
ese
lo
s
s
f
u
n
ctio
n
s
ar
e
d
esig
n
ed
to
p
en
alize
er
r
o
r
s
n
ea
r
o
b
ject
ed
g
es
m
o
r
e
h
ea
v
ily
,
th
e
r
eb
y
en
co
u
r
ag
i
n
g
th
e
n
etwo
r
k
to
f
o
cu
s
o
n
lear
n
in
g
p
r
ec
is
e
b
o
u
n
d
ar
y
r
e
p
r
esen
tatio
n
s
[
3
]
.
T
h
is
ap
p
r
o
ac
h
lead
s
to
s
h
ar
p
er
an
d
m
o
r
e
ac
cu
r
ate
s
eg
m
en
tatio
n
o
u
tp
u
ts
,
wh
ich
ar
e
cr
u
cial
in
ap
p
licatio
n
s
lik
e
m
e
d
ical
im
ag
i
n
g
,
wh
er
e
p
r
ec
is
e
b
o
u
n
d
ar
y
id
e
n
tific
atio
n
is
ess
en
tial
f
o
r
task
s
lik
e
tu
m
o
r
d
etec
tio
n
o
r
o
r
g
an
s
eg
m
en
tatio
n
[
3
]
.
T
h
e
m
eth
o
d
is
test
ed
o
n
u
r
b
an
s
ce
n
e
d
atasets
an
d
s
h
o
ws
s
u
p
er
io
r
p
e
r
f
o
r
m
an
ce
c
o
m
p
a
r
ed
to
tr
a
d
itio
n
al
m
eth
o
d
s
[
4
6
]
.
T
h
e
au
th
o
r
s
d
em
o
n
s
tr
at
e
u
s
ef
u
ln
ess
o
f
th
e
tech
n
iq
u
e
b
y
m
ak
i
n
g
its
im
p
l
em
en
tatio
n
i
n
m
e
d
ical
ap
p
licatio
n
s
wh
er
e
p
r
ec
is
e
b
o
u
n
d
ar
y
d
etec
tio
n
is
v
ital
[
3
]
.
3.
P
RO
P
O
SE
D
M
E
T
H
O
D
I
n
Sem
an
tic
s
eg
m
e
n
tatio
n
,
b
o
u
n
d
ar
y
p
i
x
el
s
eg
m
en
tatio
n
is
a
cr
itical
task
.
I
t
d
eter
m
in
es
t
h
e
p
r
ec
is
e
s
ep
ar
atio
n
b
etwe
en
o
b
jects
an
d
b
ac
k
g
r
o
u
n
d
o
r
n
eig
h
b
o
r
i
n
g
o
b
jects.
B
o
u
n
d
a
r
y
p
ix
els
ar
e
d
if
f
icu
lt
to
class
if
y
ac
cu
r
ately
d
u
e
to
th
e
s
im
ilar
ity
o
f
p
ix
el
p
r
o
p
er
ties
in
ad
jace
n
t
r
eg
io
n
s
,
wh
ich
m
ak
e
s
d
if
f
icu
lt
to
lear
n
d
is
tin
ctio
n
b
etwe
en
o
b
ject
ed
g
es
an
d
r
eg
io
n
s
.
T
o
o
v
er
co
m
e
th
is
p
r
o
b
lem
,
b
o
u
n
d
ar
y
p
ix
el
s
eg
m
en
tatio
n
is
ad
d
r
ess
ed
s
ep
ar
ately
b
y
[
4
7
]
,
[
4
8
]
.
Dee
p
a
r
ch
itectu
r
e
m
o
d
els
h
av
e
b
ee
n
d
ev
el
o
p
ed
to
class
i
f
y
b
o
u
n
d
a
r
y
p
i
x
els
[
4
9
]
–
[
5
1
]
.
T
h
is
is
ch
allen
g
in
g
p
r
o
b
lem
as
th
e
n
u
m
b
er
o
f
b
o
u
n
d
a
r
y
p
ix
els
ar
e
v
er
y
less
co
m
p
ar
ed
to
o
b
ject
r
eg
io
n
p
ix
els.
I
n
p
r
o
p
o
s
ed
m
e
th
o
d
,
b
o
u
n
d
ar
y
s
en
s
itiv
e
lo
s
s
f
u
n
ctio
n
is
in
clu
d
ed
.
T
h
is
lo
s
s
f
u
n
ctio
n
en
h
an
ce
s
m
o
d
el'
s
ab
ilit
y
to
d
is
t
in
g
u
is
h
ad
jace
n
t
o
b
jects,
im
p
r
o
v
i
n
g
s
eg
m
en
tatio
n
p
e
r
f
o
r
m
an
ce
,
esp
ec
ially
in
co
m
p
lex
s
ce
n
ar
io
s
.
3
.
1
.
T
ra
ini
ng
o
f
deep
net
wo
rk
wit
h lo
s
s
f
un
ct
io
n
Fo
r
s
em
an
tic
s
eg
m
en
tatio
n
w
h
en
d
ee
p
n
etwo
r
k
is
u
s
ed
,
t
h
e
y
ar
e
tr
ain
ed
u
s
in
g
tr
ain
in
g
d
a
taset.
Dee
p
n
etwo
r
k
ex
tr
ac
t
f
ea
tu
r
es
an
d
l
ea
r
n
s
p
atter
n
s
d
u
r
in
g
th
e
tr
ain
in
g
.
B
ased
o
n
th
is
lear
n
in
g
class
o
f
ea
ch
p
ix
el
is
p
r
ed
icted
.
T
h
e
er
r
o
r
b
etwe
en
p
r
ed
icted
class
lab
el
an
d
ac
tu
al
class
lab
el
is
ex
p
ec
te
d
as
m
in
im
al
as
p
o
s
s
ib
le.
T
o
co
m
p
u
te
er
r
o
r
b
etwe
en
p
r
e
d
icted
lab
els
a
n
d
ac
tu
al
la
b
el
l
o
s
s
f
u
n
ctio
n
s
ar
e
u
s
ed
.
T
h
e
cr
o
s
s
-
en
tr
o
p
y
lo
s
s
is
ca
lcu
lated
u
s
in
g
(
1
)
.
ℒ
=
−
1
∑
∑
,
l
og
(
,
)
=
1
=
1
(
1
)
w
h
er
e,
an
d
ar
e
s
am
p
les an
d
c
lass
es
in
d
ataset
r
esp
ec
tiv
ely
;
,
is
ac
tu
al
lab
el;
is
th
e
p
r
o
b
ab
ilit
y
g
iv
en
.
T
h
e
So
f
tMa
x
p
r
o
v
i
d
es
p
r
ed
ict
ed
p
r
o
b
ab
ilit
ies
f
o
r
ea
c
h
class
.
T
h
e
f
u
n
ctio
n
(
,
)
wh
er
e
is
th
e
lo
s
s
f
u
n
ctio
n
co
m
p
u
tes
th
e
lo
s
s
ass
o
ciate
d
with
p
r
ed
icted
an
d
ac
tu
al
lab
els.
T
h
e
g
ar
d
ien
t
d
escen
t
alg
o
r
ith
m
o
p
tim
izes
th
e
n
etwo
r
k
p
ar
am
eter
s
to
g
et
m
i
n
im
u
m
lo
s
s
.
Ou
r
m
eth
o
d
p
r
o
p
o
s
es
n
o
v
el
l
o
s
s
f
u
n
ctio
n
wh
ich
in
clu
d
es b
o
u
n
d
ar
y
lo
s
s
alo
n
g
with
r
eg
io
n
lo
s
s
.
Deta
ils
o
f
p
r
o
p
o
s
ed
m
eth
o
d
s
ar
e
g
iv
en
.
3.
2.
B
lo
ck
dia
g
ra
m o
f
pro
po
s
ed
m
et
ho
d
T
h
e
f
lo
w
o
f
th
e
in
p
u
t
an
d
u
p
d
atin
g
o
f
n
etwo
r
k
p
ar
am
eter
s
b
y
ca
lcu
latin
g
th
e
lo
s
s
at
d
if
f
er
en
t
s
tag
es
is
s
h
o
wn
in
Fig
u
r
e
1
.
T
h
e
p
r
o
p
o
s
ed
m
eth
o
d
u
s
es
Seg
N
et
n
etwo
r
k
as
a
b
ac
k
b
o
n
e
.
Pre
tr
ain
ed
Seg
Net
ar
ch
itectu
r
e
is
u
s
ed
to
im
p
le
m
en
t
p
r
o
p
o
s
ed
m
eth
o
d
.
T
h
is
ar
ch
itectu
r
e
is
tr
ain
ed
u
s
in
g
s
em
an
tic
b
o
u
n
d
a
r
y
d
ataset
(
SB
D)
wh
ich
co
n
tain
s
r
eg
io
n
g
r
o
u
n
d
tr
u
th
(R
-
GT
)
an
d
b
o
u
n
d
ar
y
g
r
o
u
n
d
tr
u
th
(
B
-
GT
)
as
g
iv
en
Fig
u
r
e
2
.
T
h
e
d
ataset
o
f
1
,
0
2
6
im
ag
es
h
as
1
5
class
es.
I
n
Fig
u
r
e
1
,
I
m
a
g
e
I
is
p
r
o
v
id
e
d
t
o
n
etwo
r
k
,
R
-
GT
is
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.
15
,
No
.
6
,
Decem
b
e
r
20
25
:
5
3
2
7
-
5
3
3
5
5330
r
eg
io
n
g
r
o
u
n
d
tr
u
th
o
f
I
,
B
-
GT
is
b
o
u
n
d
a
r
y
g
r
o
u
n
d
tr
u
th
o
f
th
e
o
f
I
,
R
O
is
th
e
r
eg
io
n
s
eg
m
en
tatio
n
(
R
O)
o
f
th
e
I
,
B
-
Ma
p
is
b
o
u
n
d
ar
y
m
ap
o
f
th
e
r
eg
io
n
s
eg
m
en
tatio
n
.
Fig
u
r
e
1
.
T
h
e
b
lo
c
k
d
iag
r
am
I
m
ag
e
R
-
GT
B
-
GT
Fig
u
r
e
2
.
Sam
p
le
im
a
g
es o
f
R
-
GT
an
d
B
-
GT
f
r
o
m
th
e
SB
D
d
ataset
3
.
3
.
Net
wo
rk
t
r
a
ini
ng
T
h
e
p
r
etr
ain
e
d
n
etwo
r
k
is
tr
a
in
ed
u
s
in
g
SB
D
d
ataset
[
1
0
]
wh
ich
co
n
tain
lab
elle
d
g
r
o
u
n
d
tr
u
th
f
o
r
r
eg
io
n
(
R
-
GT
)
as
well
as
b
o
u
n
d
ar
y
(
B
-
GT
)
.
L
o
s
s
b
etwe
en
p
r
ed
icted
r
eg
io
n
p
ix
el
lab
els
an
d
R
-
GT
lab
els
(
1
)
is
ca
lcu
lated
u
s
in
g
(
2
)
.
Fo
r
p
r
ed
icted
im
ag
e
m
ap
a
n
d
R
-
G
T
h
av
in
g
h
eig
h
t
(
)
an
d
wid
t
h
(
)
an
d
o
b
ject
class
es
C
:
1
=
−
1
×
∑
∑
∑
−
(
,
,
)
l
og
(
(
,
,
)
)
=
1
=
1
=
1
(
2
)
h
er
e,
(
,
)
is
p
ix
el
co
o
r
d
i
n
ates a
n
d
c
ar
e
o
b
ject
class
es.
E
d
g
e
m
ap
o
f
p
r
ed
icted
lab
elle
d
im
ag
e
is
o
b
tain
ed
u
s
in
g
C
an
n
y
ed
g
e
d
etec
to
r
[
5
2
]
,
[
5
3
]
as
s
h
o
wn
in
Fig
u
r
e
3
.
T
h
e
lo
s
s
b
etwe
en
th
e
p
r
ed
icted
s
em
an
tic
b
o
u
n
d
ar
y
m
ap
[
5
4
]
,
[
5
5
]
(
r
ef
er
r
e
d
to
as
B
-
Ma
p
)
an
d
th
e
g
r
o
u
n
d
tr
u
th
b
o
u
n
d
ar
y
m
ap
(
B
-
GT
)
is
ca
lcu
lated
u
s
in
g
(
3
)
.
T
h
is
lo
s
s
,
d
en
o
ted
as
2
,
m
ea
s
u
r
es
h
o
w
ac
cu
r
ately
th
e
m
o
d
el
p
r
ed
icts
o
b
ject
b
o
u
n
d
a
r
ies.
T
h
e
p
r
ed
i
ctio
n
an
d
g
r
o
u
n
d
t
r
u
th
m
a
p
s
b
o
th
h
a
v
e
a
h
ei
g
h
t
(
)
,
wid
th
(
)
,
an
d
r
ep
r
esen
t
o
b
ject
class
es.
T
h
e
lo
s
s
2
is
s
h
o
wn
in
(
3
)
:
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
E
n
h
a
n
cin
g
s
ema
n
tic
s
eg
men
t
a
tio
n
w
ith
a
b
o
u
n
d
a
r
y
-
s
en
s
itive
…
(
Ga
n
esh
R
.
P
a
d
a
lka
r
)
5331
2
=
−
1
×
∑
∑
∑
−
(
,
,
)
l
og
(
−
(
,
,
)
)
=
1
=
1
=
1
(
3
)
T
h
is
eq
u
atio
n
en
s
u
r
es
th
at
t
h
e
m
o
d
el
is
p
en
alize
d
m
o
r
e
wh
en
it
f
ails
to
ac
cu
r
ately
p
r
ed
ict
b
o
u
n
d
a
r
ies
o
f
o
b
jects
in
th
e
im
ag
e.
T
o
g
u
id
e
th
e
m
o
d
el
in
lear
n
in
g
b
o
th
r
eg
io
n
s
eg
m
en
tatio
n
an
d
b
o
u
n
d
ar
y
p
r
ed
ictio
n
,
a
to
tal
lo
s
s
(
)
is
u
s
ed
.
T
h
is
is
co
m
b
in
atio
n
o
f
1
(
r
eg
io
n
lo
s
s
)
an
d
2
(
b
o
u
n
d
ar
y
lo
s
s
)
as sh
o
wn
in
(
4
)
:
=
1+
2
(
4
)
3
.
4
.
T
ra
ini
ng
s
et
u
p a
nd
hy
perpa
ra
m
et
er
c
o
nfig
ura
t
io
n
T
h
e
SB
D
[
5
5
]
is
u
s
ed
f
o
r
tr
ain
in
g
an
d
ev
alu
atin
g
th
e
Seg
Net
d
ee
p
ar
ch
itectu
r
e
[
7
]
.
T
h
is
d
ataset
,
d
er
iv
ed
f
r
o
m
th
e
PASC
AL
VOC
d
atase
t
[
5
6
]
,
in
clu
d
es
d
etailed
b
o
u
n
d
a
r
y
an
n
o
tatio
n
s
f
o
r
2
1
class
es,
in
clu
d
in
g
“
b
ac
k
g
r
o
u
n
d
”
class
.
Fo
r
tr
ain
in
g
t
h
e
Seg
Net
[
7
]
1
,
0
2
6
im
a
g
es,
f
o
r
v
alid
atio
n
2
9
0
im
ag
es,
an
d
f
o
r
test
in
g
1
3
6
im
a
g
es
ar
e
u
tili
ze
d
f
r
o
m
t
h
e
d
ataset.
T
h
e
p
r
o
p
o
s
ed
m
eth
o
d
f
o
r
s
em
an
tic
s
eg
m
e
n
tatio
n
was
tr
ain
ed
o
n
an
HP
Z
4
T
o
wer
W
o
r
k
s
tatio
n
with
an
I
n
tel
Xeo
n
W
-
2
2
4
5
p
r
o
ce
s
s
o
r
,
3
2
GB
DDR4
R
AM
,
an
NVI
DI
A
R
T
X
A5
0
0
0
GPU,
an
d
a
1
T
B
HDD,
u
s
in
g
MA
T
L
AB
R
2
0
1
4
a.
Fo
r
h
y
p
er
p
a
r
am
eter
tu
n
in
g
,
th
e
Seg
Net
[
7
]
m
o
d
el
is
co
n
f
ig
u
r
ed
with
s
tar
tin
g
lear
n
i
n
g
r
ate
o
f
0
.
0
1
,
p
iec
ewise
lear
n
in
g
r
ate
s
ch
ed
u
le
with
a
d
r
o
p
p
er
io
d
o
f
1
0
ep
o
ch
s
an
d
a
d
r
o
p
f
ac
to
r
o
f
0
.
1
,
b
atch
s
ize
o
f
1
(
s
to
ch
asti
c
g
r
ad
ien
t
d
escen
t)
,
a
to
tal
o
f
5
0
ep
o
ch
s
,
m
o
m
en
tu
m
o
f
0
.
9
,
a
n
d
v
alid
atio
n
f
r
e
q
u
en
c
y
o
f
1
0
e
p
o
ch
s
.
3
.
5
.
E
v
a
lua
t
io
n
m
et
rics
T
h
is
wo
r
k
ev
alu
ates
th
e
co
n
tr
ib
u
tio
n
o
f
p
r
o
p
o
s
ed
lo
s
s
f
u
n
ctio
n
f
o
r
b
o
th
s
em
an
t
ic
r
eg
io
n
s
eg
m
en
tatio
n
an
d
b
o
u
n
d
ar
y
d
etec
tio
n
task
s
.
T
o
en
s
u
r
e
a
co
m
p
r
eh
e
n
s
iv
e
ass
ess
m
e
n
t,
s
ev
er
al
wid
ely
r
ec
o
g
n
ized
m
etr
ics
ar
e
em
p
lo
y
ed
,
in
clu
d
in
g
g
lo
b
al
ac
cu
r
ac
y
[
5
6
]
,
m
ea
n
ac
cu
r
ac
y
[
5
6
]
,
an
d
(
m
ea
n
I
o
U)
[
5
7
]
.
Ad
d
itio
n
ally
,
weig
h
ted
in
ter
s
ec
tio
n
o
v
er
u
n
io
n
(
w
eig
h
te
d
I
o
U)
[
5
8
]
is
in
c
o
r
p
o
r
ated
to
ass
ig
n
h
ig
h
er
im
p
o
r
tan
ce
to
u
n
d
er
r
e
p
r
esen
te
d
class
es,
en
s
u
r
in
g
f
air
ev
alu
a
tio
n
ac
r
o
s
s
d
if
f
er
en
t
o
b
ject
ca
teg
o
r
ies.
T
h
e
m
ea
n
b
o
u
n
d
ar
y
F
-
Sco
r
e
(
m
ea
n
B
FS
co
r
e)
[
5
9
]
is
p
ar
ticu
lar
ly
cr
u
cial
f
o
r
ass
ess
in
g
b
o
u
n
d
ar
y
ac
cu
r
ac
y
,
as
it
q
u
an
tifie
s
h
o
w
well
th
e
p
r
ed
icted
s
eg
m
en
tatio
n
alig
n
s
wit
h
o
b
ject
b
o
u
n
d
ar
ies,
ca
p
tu
r
in
g
f
in
e
d
etails
o
f
ten
o
v
er
lo
o
k
ed
b
y
s
tan
d
ar
d
r
e
g
io
n
-
b
ased
m
etr
ics.
T
h
ese
m
etr
ics
co
llectiv
ely
p
r
o
v
id
e
a
h
o
lis
tic
ev
alu
atio
n
.
4.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
N
T
h
e
p
er
f
o
r
m
an
ce
o
f
in
teg
r
ate
d
lo
s
s
f
u
n
ctio
n
with
r
eg
io
n
lo
s
s
an
d
th
e
p
r
o
p
o
s
ed
b
o
u
n
d
ar
y
-
s
en
s
itiv
e
lo
s
s
f
u
n
ctio
n
is
s
h
o
wn
in
T
ab
l
e
1
.
T
h
e
tab
le
h
ig
h
lig
h
ts
ac
cu
r
ac
y
,
I
o
U,
a
n
d
Me
an
B
FS
co
r
e,
f
o
r
b
o
t
h
m
eth
o
d
s
.
T
ab
le
2
p
r
o
v
id
es
an
o
v
er
all
s
u
m
m
ar
y
o
f
h
o
w
th
e
s
eg
m
en
ta
tio
n
m
o
d
el
p
er
f
o
r
m
s
wh
en
u
s
in
g
o
n
ly
t
h
e
r
eg
io
n
lo
s
s
co
m
p
ar
ed
to
t
h
e
p
r
o
p
o
s
ed
b
o
u
n
d
ar
y
-
s
en
s
itiv
e
lo
s
s
.
Fro
m
th
e
r
esu
lts
it
is
ev
id
en
t
th
at
with
a
s
in
g
le
d
ee
p
n
etwo
r
k
alo
n
g
with
a
b
o
u
n
d
ar
y
-
s
en
s
itiv
e
lo
s
s
,
im
p
r
o
v
es
s
em
an
tic
s
eg
m
en
tatio
n
o
v
er
a
ll.
B
y
f
o
cu
s
in
g
o
n
o
b
ject
b
o
u
n
d
ar
ies,
it
m
ak
es
th
e
s
eg
m
en
tatio
n
m
o
r
e
ac
cu
r
ate
.
T
h
is
ap
p
r
o
ac
h
also
p
er
f
o
r
m
s
b
etter
th
an
s
ev
er
al
o
th
er
m
eth
o
d
s
th
at
r
ely
o
n
m
u
ltip
le
n
etwo
r
k
s
,
m
ak
in
g
it b
o
t
h
ef
f
ec
tiv
e
a
n
d
ef
f
icien
t.
T
ab
le
1
.
C
o
m
p
a
r
is
o
n
o
f
s
em
an
tic
s
eg
m
en
tatio
n
u
s
in
g
o
n
ly
r
eg
io
n
lo
s
s
an
d
p
r
o
p
o
s
ed
b
o
u
n
d
ar
y
-
s
en
s
itiv
e
lo
s
s
S
r
.
N
o
O
b
j
e
c
t
c
l
a
ss
A
c
c
u
r
a
c
y
I
o
U
M
e
a
n
B
F
s
c
o
r
e
O
n
l
y
r
e
g
i
o
n
l
o
ss
P
r
o
p
o
se
d
b
o
u
n
d
a
r
y
-
sen
s
i
t
i
v
e
L
o
ss
O
n
l
y
r
e
g
i
o
n
l
o
ss
P
r
o
p
o
se
d
b
o
u
n
d
a
r
y
-
sen
s
i
t
i
v
e
l
o
ss
O
n
l
y
r
e
g
i
o
n
l
o
ss
P
r
o
p
o
se
d
b
o
u
n
d
a
r
y
-
sen
s
i
t
i
v
e
l
o
ss
1.
B
a
c
k
g
r
o
u
n
d
0
.
9
7
7
0
.
9
9
6
0
.
6
0
8
0
.
6
2
1
0
.
4
7
2
0
.
4
7
9
2.
A
e
r
o
p
l
a
n
e
0
.
1
8
9
0
.
6
6
7
0
.
1
5
9
0
.
5
4
8
0
.
2
1
0
0
.
4
7
1
3.
B
i
c
y
c
l
e
0
.
2
3
9
0
.
2
7
9
0
.
2
2
3
0
.
2
6
7
0
.
0
3
4
0
.
0
3
9
4.
B
i
r
d
0
.
1
6
3
0
.
4
2
8
0
.
0
4
5
0
.
1
6
4
0
.
3
5
2
0
.
5
1
9
5.
B
u
s
0
.
1
4
7
0
.
1
8
4
0
.
1
3
1
0
.
1
3
6
0
.
1
5
5
0
.
1
4
6
6.
Car
0
.
5
1
6
0
.
1
6
0
0
.
4
0
1
0
.
1
3
9
0
.
1
7
6
0
.
1
1
6
7.
Cat
0
.
0
0
6
0
.
2
4
6
0
.
0
0
5
0
.
2
2
8
0
.
0
4
7
0
.
1
3
2
8.
C
h
a
i
r
0
.
3
1
4
0
.
1
4
3
0
.
1
9
3
0
.
0
8
4
0
.
2
7
7
0
.
2
7
6
9.
C
o
w
0
.
0
3
6
0
.
0
0
2
0
.
0
0
8
0
.
0
0
1
0
.
1
8
4
N
a
N
1
0
.
D
o
g
0
.
0
5
2
0
.
0
4
0
0
.
0
4
9
0
.
0
4
0
0
.
0
0
1
0
.
0
4
6
1
1
.
H
o
r
se
0
.
0
1
1
0
.
0
2
6
0
.
0
1
0
0
.
0
2
5
0
.
0
7
6
0
.
1
0
0
1
2
.
M
o
t
o
r
b
i
k
e
0
.
0
2
9
0
.
1
1
2
0
.
0
2
2
0
.
0
9
7
0
.
0
8
7
0
.
0
9
6
1
3
.
P
e
r
so
n
0
.
2
5
9
0
.
1
4
9
0
.
2
2
2
0
.
1
4
5
0
.
2
6
4
0
.
1
3
2
1
4
.
S
o
f
a
0
.
1
0
5
0
.
0
6
4
0
.
1
0
1
0
.
0
6
2
0
.
0
8
0
0
.
0
4
3
1
5
.
T
V
mo
n
i
t
o
r
0
.
0
8
2
0
.
0
5
7
0
.
0
7
9
0
.
0
5
5
0
.
0
4
4
0
.
0
0
7
1
6
.
A
v
e
r
a
g
e
v
a
l
u
e
s
0
.
2
0
8
0
.
2
3
7
0
.
1
5
0
0
.
1
7
4
0
.
1
6
4
0
.
1
8
6
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.
15
,
No
.
6
,
Decem
b
e
r
20
25
:
5
3
2
7
-
5
3
3
5
5332
T
ab
le
2
.
C
o
m
p
a
r
is
o
n
o
f
s
em
an
tic
s
eg
m
en
tatio
n
with
r
e
g
io
n
lo
s
s
an
d
p
r
o
p
o
s
ed
b
o
u
n
d
ar
y
-
s
en
s
itiv
e
lo
s
s
S
r
.
N
o
M
e
t
r
i
c
O
n
l
y
r
e
g
i
o
n
l
o
ss
P
r
o
p
o
se
d
b
o
u
n
d
a
r
y
-
s
e
n
s
i
t
i
v
e
l
o
ss
1.
G
l
o
b
a
l
a
c
c
u
r
a
c
y
0
.
5
9
7
9
0
.
6
0
9
0
2.
M
e
a
n
a
c
c
u
r
a
c
y
0
.
1
5
2
4
0
.
1
7
4
5
3.
M
e
a
n
I
o
U
0
.
1
1
0
7
0
.
1
2
8
6
4.
W
e
i
g
h
t
e
d
I
o
U
0
.
3
8
1
7
0
.
3
9
0
8
5.
M
e
a
n
BF
s
c
o
r
e
0
.
2
9
0
7
0
.
3
3
6
2
5.
CO
NCLU
SI
O
N
T
h
e
ex
p
er
im
en
t
d
em
o
n
s
tr
ates
th
e
s
ig
n
if
ica
n
t
im
p
ac
t
o
f
in
c
o
r
p
o
r
ati
n
g
b
o
u
n
d
ar
y
lo
s
s
(
2
)
al
o
n
g
s
id
e
r
eg
io
n
lo
s
s
(
1
)
in
s
em
an
tic
s
eg
m
en
tatio
n
.
W
h
en
th
e
m
o
d
el
is
tr
ain
ed
s
o
lely
with
r
eg
io
n
lo
s
s
,
it
p
er
f
o
r
m
s
well
in
s
eg
m
en
tin
g
lar
g
e
an
d
f
r
eq
u
en
t
o
b
jects
b
u
t
s
tr
u
g
g
les
with
s
m
al
l
o
b
jects,
lead
in
g
to
p
o
o
r
b
o
u
n
d
a
r
y
d
elin
ea
tio
n
an
d
lo
wer
I
o
U
an
d
b
o
u
n
d
ar
y
ac
cu
r
ac
y
.
I
n
co
n
t
r
ast,
wh
en
b
o
th
r
eg
io
n
an
d
b
o
u
n
d
a
r
y
lo
s
s
es
ar
e
in
clu
d
ed
,
th
e
m
o
d
el
ac
h
iev
es
b
etter
s
eg
m
en
tatio
n
p
e
r
f
o
r
m
a
n
ce
,
p
a
r
ticu
lar
ly
in
p
r
eser
v
in
g
o
b
ject
b
o
u
n
d
ar
ies.
T
h
e
im
p
r
o
v
e
m
en
ts
ar
e
m
o
s
t
ev
id
en
t
f
o
r
o
b
jects
with
well
-
d
ef
in
ed
ed
g
es,
as
r
ef
lecte
d
i
n
h
ig
h
er
I
o
U
an
d
b
o
u
n
d
ar
y
F
-
s
co
r
es.
Ho
wev
er
,
th
e
s
eg
m
en
tatio
n
o
f
s
m
aller
o
b
jects
r
em
ain
s
ch
allen
g
in
g
,
lik
ely
d
u
e
to
class
im
b
alan
ce
an
d
th
e
co
m
p
lex
ity
o
f
d
etec
tin
g
f
in
e
d
etails.
T
h
e
r
esu
lts
alig
n
with
r
ec
en
t
r
esear
ch
f
in
d
in
g
s
,
r
ein
f
o
r
cin
g
th
e
id
ea
th
at
b
o
u
n
d
a
r
y
-
awa
r
e
m
o
d
els
en
h
an
ce
s
eg
m
en
tatio
n
ac
c
u
r
ac
y
.
T
h
is
s
tu
d
y
d
em
o
n
s
tr
at
es
th
at
b
alan
cin
g
r
eg
io
n
an
d
b
o
u
n
d
ar
y
lo
s
s
es
co
n
tr
ib
u
tes
to
m
o
r
e
d
etailed
a
n
d
p
r
ec
is
e
s
eg
m
e
n
tatio
n
,
m
a
k
in
g
it
p
ar
ticu
lar
l
y
u
s
ef
u
l
i
n
ad
v
an
ce
d
ap
p
licatio
n
s
,
wh
er
e
b
o
u
n
d
ar
y
p
r
ec
is
io
n
is
cr
itical.
T
h
e
p
r
o
p
o
s
ed
ap
p
r
o
a
ch
in
tr
o
d
u
ce
s
a
n
o
v
el
lo
s
s
f
u
n
ctio
n
th
at
im
p
r
o
v
es
s
eg
m
en
tatio
n
ac
cu
r
ac
y
.
Fu
tu
r
e
r
esear
ch
s
h
o
u
ld
f
o
cu
s
o
n
ad
d
r
ess
in
g
th
e
lim
itatio
n
s
o
b
s
er
v
ed
in
s
m
all
o
b
ject
s
eg
m
en
tatio
n
b
y
e
x
p
lo
r
i
n
g
ad
ap
tiv
e
weig
h
tin
g
s
tr
ateg
ies
to
m
itig
ate
class
im
b
alan
ce
.
E
x
p
an
d
in
g
th
e
m
o
d
el
to
h
an
d
le
m
u
lti
-
class
s
eg
m
en
tati
o
n
in
m
o
r
e
co
m
p
lex
s
ce
n
es is
an
o
th
er
a
v
en
u
e
f
o
r
im
p
r
o
v
em
en
t.
ACK
NO
WL
E
DG
M
E
N
T
S
T
h
e
au
th
o
r
s
wo
u
l
d
lik
e
t
o
ac
k
n
o
wled
g
e
MK
SS
S
'
s
C
u
m
m
i
n
s
C
o
lleg
e
o
f
E
n
g
in
ee
r
in
g
f
o
r
W
o
m
en
,
Pu
n
e,
f
o
r
p
r
o
v
id
i
n
g
lab
o
r
ato
r
y
f
ac
ilit
ies
eq
u
ip
p
e
d
with
GP
U
r
eso
u
r
ce
s
an
d
f
o
r
s
u
p
p
o
r
tin
g
th
is
r
esear
ch
wo
r
k
th
r
o
u
g
h
in
s
titu
tio
n
al
af
f
iliatio
n
.
F
UNDING
I
NF
O
R
M
A
T
I
O
N
T
h
e
au
th
o
r
s
s
tate
th
at
n
o
f
u
n
d
i
n
g
was in
v
o
lv
e
d
in
s
u
p
p
o
r
tin
g
th
is
r
esear
ch
wo
r
k
.
AUTHO
R
CO
NT
RI
B
UT
I
O
NS ST
A
T
E
M
E
N
T
T
h
is
jo
u
r
n
al
u
s
es
th
e
C
o
n
tr
ib
u
to
r
R
o
les
T
ax
o
n
o
m
y
(
C
R
ed
iT)
to
r
ec
o
g
n
ize
in
d
iv
id
u
al
au
th
o
r
co
n
tr
ib
u
tio
n
s
,
r
ed
u
ce
au
th
o
r
s
h
ip
d
is
p
u
tes,
an
d
f
ac
ilit
ate
co
llab
o
r
atio
n
.
Na
m
e
o
f
Aut
ho
r
C
M
So
Va
Fo
I
R
D
O
E
Vi
Su
P
Fu
Gan
esh
R
.
Pad
alk
ar
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
Ma
d
h
u
r
i B.
Kh
am
b
ete
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
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.
I
NF
O
RM
E
D
CO
NS
E
N
T
W
e
h
av
e
o
b
tain
ed
in
f
o
r
m
ed
c
o
n
s
en
t f
r
o
m
all
in
d
iv
id
u
als in
c
lu
d
ed
in
t
h
is
s
tu
d
y
.
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
E
n
h
a
n
cin
g
s
ema
n
tic
s
eg
men
t
a
tio
n
w
ith
a
b
o
u
n
d
a
r
y
-
s
en
s
itive
…
(
Ga
n
esh
R
.
P
a
d
a
lka
r
)
5333
E
T
H
I
CAL AP
P
RO
V
AL
T
h
e
r
esear
ch
r
elate
d
to
an
i
m
a
l
u
s
e
h
as
b
ee
n
co
m
p
lied
with
all
th
e
r
elev
an
t
n
atio
n
al
r
eg
u
l
atio
n
s
an
d
in
s
titu
tio
n
al
p
o
licies f
o
r
th
e
ca
r
e
an
d
u
s
e
o
f
an
im
als.
DATA AV
AI
L
AB
I
L
I
T
Y
T
h
e
d
ata
s
u
p
p
o
r
tin
g
th
e
f
i
n
d
in
g
s
o
f
th
is
s
tu
d
y
a
r
e
b
ased
o
n
t
h
e
Sem
an
tic
B
o
u
n
d
a
r
ies
Data
s
et
(
SB
D)
,
wh
ich
was
ac
ce
s
s
ed
b
y
th
e
a
u
th
o
r
s
wh
en
t
h
e
o
r
i
g
in
al
r
e
p
o
s
ito
r
y
was
ac
tiv
e.
C
u
r
r
en
tly
,
th
e
d
ataset
is
av
ailab
le
th
r
o
u
g
h
th
e
f
o
llo
win
g
s
o
u
r
ce
s
:
-
Pap
er
s
with
co
d
e
SB
D
p
ag
e
:
h
ttp
s
:
//p
a
p
ers
w
ith
co
d
e.
co
m/d
a
ta
s
et/s
b
d
-
Py
T
o
r
ch
T
o
r
ch
Vis
io
n
L
ib
r
ar
y
:
h
ttp
s
:
//d
o
cs.p
yto
r
ch
.
o
r
g
/vis
io
n
/ma
in
/g
en
era
ted
/to
r
ch
visi
o
n
.
d
a
ta
s
ets.S
B
Da
ta
s
et.
h
tml
RE
F
E
R
E
NC
E
S
[
1
]
J.
Lo
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
s
e
g
me
n
t
a
t
i
o
n
,
”
i
n
Pr
o
c
e
e
d
i
n
g
s
o
f
t
h
e
I
EE
E
c
o
n
f
e
re
n
c
e
o
n
c
o
m
p
u
t
e
r v
i
si
o
n
a
n
d
p
a
t
t
e
r
n
r
e
c
o
g
n
i
t
i
o
n
,
2
0
1
7
,
p
p
.
3
4
3
1
–
3
4
4
0
.
[
2
]
I
.
P
a
p
a
d
e
a
s,
L.
Tso
c
h
a
t
z
i
d
i
s
,
A
.
A
ma
n
a
t
i
a
d
i
s,
a
n
d
I
.
P
r
a
t
i
k
a
k
i
s,
“
R
e
a
l
-
t
i
me
sema
n
t
i
c
i
ma
g
e
se
g
me
n
t
a
t
i
o
n
w
i
t
h
d
e
e
p
l
e
a
r
n
i
n
g
f
o
r
a
u
t
o
n
o
mo
u
s
d
r
i
v
i
n
g
:
a
s
u
r
v
e
y
,
”
A
p
p
l
i
e
d
S
c
i
e
n
c
e
s
,
v
o
l
.
1
1
,
n
o
.
1
9
,
p
.
8
8
0
2
,
S
e
p
.
2
0
2
1
,
d
o
i
:
1
0
.
3
3
9
0
/
a
p
p
1
1
1
9
8
8
0
2
.
[
3
]
O
.
R
o
n
n
e
b
e
r
g
e
r
,
P
.
F
i
sc
h
e
r
,
a
n
d
T.
B
r
o
x
,
“
U
-
n
e
t
:
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
t
w
o
r
k
s
f
o
r
b
i
o
me
d
i
c
a
l
i
ma
g
e
se
g
m
e
n
t
a
t
i
o
n
,
”
i
n
I
n
t
e
rn
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
M
e
d
i
c
a
l
i
m
a
g
e
c
o
m
p
u
t
i
n
g
a
n
d
c
o
m
p
u
t
e
r
-
a
ss
i
st
e
d
i
n
t
e
rv
e
n
t
i
o
n
,
2
0
1
5
,
p
p
.
2
3
4
–
2
4
1
.
[
4
]
N
.
S
.
D
’
S
o
u
z
a
e
t
a
l
.
,
“
D
e
e
p
sr
-
D
D
L:
D
e
e
p
st
r
u
c
t
u
r
a
l
l
y
r
e
g
u
l
a
r
i
z
e
d
d
y
n
a
mi
c
d
i
c
t
i
o
n
a
r
y
l
e
a
r
n
i
n
g
t
o
i
n
t
e
g
r
a
t
e
m
u
l
t
i
m
o
d
a
l
a
n
d
d
y
n
a
mi
c
f
u
n
c
t
i
o
n
a
l
c
o
n
n
e
c
t
o
m
i
c
s
d
a
t
a
f
o
r
m
u
l
t
i
d
i
me
n
si
o
n
a
l
c
l
i
n
i
c
a
l
c
h
a
r
a
c
t
e
r
i
z
a
t
i
o
n
s,”
N
e
u
r
o
I
m
a
g
e
,
v
o
l
.
2
4
1
,
p
.
1
1
8
3
8
8
,
N
o
v
.
2
0
2
1
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
n
e
u
r
o
i
m
a
g
e
.
2
0
2
1
.
1
1
8
3
8
8
.
[
5
]
L.
-
C
.
C
h
e
n
,
G
.
P
a
p
a
n
d
r
e
o
u
,
I
.
K
o
k
k
i
n
o
s
,
K
.
M
u
r
p
h
y
,
a
n
d
A
.
L.
Y
u
i
l
l
e
,
“
S
e
ma
n
t
i
c
i
ma
g
e
se
g
me
n
t
a
t
i
o
n
w
i
t
h
d
e
e
p
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
t
s a
n
d
f
u
l
l
y
c
o
n
n
e
c
t
e
d
C
R
F
s,”
a
r
Xi
v
p
r
e
p
ri
n
t
a
r
Xi
v
:
1
4
1
2
.
7
0
6
2
,
p
p
.
7
7
8
–
7
9
0
,
2
0
1
4
.
[
6
]
L.
-
C
.
C
h
e
n
,
G
.
P
a
p
a
n
d
r
e
o
u
,
I
.
K
o
k
k
i
n
o
s,
K
.
M
u
r
p
h
y
,
a
n
d
A
.
L.
Y
u
i
l
l
e
,
“
D
e
e
p
La
b
:
sem
a
n
t
i
c
i
ma
g
e
se
g
m
e
n
t
a
t
i
o
n
w
i
t
h
d
e
e
p
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
t
s,
A
t
r
o
u
s
c
o
n
v
o
l
u
t
i
o
n
,
a
n
d
f
u
l
l
y
c
o
n
n
e
c
t
e
d
C
R
F
s,
”
I
E
EE
T
r
a
n
s
a
c
t
i
o
n
s
o
n
Pa
t
t
e
rn
A
n
a
l
y
s
i
s
a
n
d
M
a
c
h
i
n
e
I
n
t
e
l
l
i
g
e
n
c
e
,
v
o
l
.
4
0
,
n
o
.
4
,
p
p
.
8
3
4
–
8
4
8
,
A
p
r
.
2
0
1
8
,
d
o
i
:
1
0
.
1
1
0
9
/
TPA
M
I
.
2
0
1
7
.
2
6
9
9
1
8
4
.
[
7
]
V
.
B
a
d
r
i
n
a
r
a
y
a
n
a
n
,
A
.
K
e
n
d
a
l
l
,
a
n
d
R
.
C
i
p
o
l
l
a
,
“
S
e
g
N
e
t
:
a
d
e
e
p
c
o
n
v
o
l
u
t
i
o
n
a
l
e
n
c
o
d
e
r
-
d
e
c
o
d
e
r
a
r
c
h
i
t
e
c
t
u
r
e
f
o
r
i
ma
g
e
seg
m
e
n
t
a
t
i
o
n
,
”
I
E
EE
T
r
a
n
sa
c
t
i
o
n
s o
n
Pa
t
t
e
rn
An
a
l
y
si
s
a
n
d
M
a
c
h
i
n
e
I
n
t
e
l
l
i
g
e
n
c
e
,
v
o
l
.
3
9
,
n
o
.
1
2
,
p
p
.
2
4
8
1
–
2
4
9
5
,
D
e
c
.
2
0
1
7
,
d
o
i
:
1
0
.
1
1
0
9
/
TPA
M
I
.
2
0
1
6
.
2
6
4
4
6
1
5
.
[
8
]
A
.
Q
u
a
t
t
o
n
i
,
S
.
W
a
n
g
,
L.
-
P
.
M
o
r
e
n
c
y
,
M
.
C
o
l
l
i
n
s,
a
n
d
T
.
D
a
r
r
e
l
l
,
“
H
i
d
d
e
n
c
o
n
d
i
t
i
o
n
a
l
r
a
n
d
o
m
f
i
e
l
d
s,
”
I
EEE
T
ra
n
sa
c
t
i
o
n
s
o
n
Pa
t
t
e
r
n
An
a
l
y
si
s
a
n
d
Ma
c
h
i
n
e
I
n
t
e
l
l
i
g
e
n
c
e
,
v
o
l
.
2
9
,
n
o
.
1
0
,
p
p
.
1
8
4
8
–
1
8
5
2
,
O
c
t
.
2
0
0
7
,
d
o
i
:
1
0
.
1
1
0
9
/
TPA
M
I
.
2
0
0
7
.
1
1
2
4
.
[
9
]
Y
.
Y
.
B
o
y
k
o
v
a
n
d
M
.
-
P
.
J
o
l
l
y
,
“
I
n
t
e
r
a
c
t
i
v
e
g
r
a
p
h
c
u
t
s
f
o
r
o
p
t
i
m
a
l
b
o
u
n
d
a
r
y
&
r
e
g
i
o
n
se
g
me
n
t
a
t
i
o
n
o
f
o
b
j
e
c
t
s
i
n
N
D
i
ma
g
e
s
,
”
i
n
Pro
c
e
e
d
i
n
g
s
Ei
g
h
t
h
I
EEE
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
m
p
u
t
e
r
Vi
si
o
n
.
I
C
C
V
2
0
0
1
,
v
o
l
.
1
,
p
p
.
1
0
5
–
1
1
2
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
C
V
.
2
0
0
1
.
9
3
7
5
0
5
.
[
1
0
]
A
.
K
r
i
z
h
e
v
s
k
y
,
I
.
S
u
t
s
k
e
v
e
r
,
a
n
d
G
.
E.
H
i
n
t
o
n
,
“
I
mag
e
N
e
t
c
l
a
s
si
f
i
c
a
t
i
o
n
w
i
t
h
d
e
e
p
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
,
”
A
d
v
a
n
c
e
s
i
n
n
e
u
ra
l
i
n
f
o
rm
a
t
i
o
n
p
r
o
c
e
ss
i
n
g
s
y
st
e
m
s
,
v
o
l
.
2
5
,
2
0
1
2
.
[
1
1
]
H
.
R
.
R
o
t
h
,
L
.
L
u
,
C
.
L
.
F
a
r
a
g
o
,
a
n
d
E.
S
.
L
.
H
.
G
.
(
H
)
.
L
i
,
“
D
e
e
p
l
e
a
r
n
i
n
g
f
o
r
me
d
i
c
a
l
i
m
a
g
e
a
n
a
l
y
s
i
s:
a
c
o
m
p
r
e
h
e
n
si
v
e
r
e
v
i
e
w
,
”
Me
d
i
c
a
l
I
m
a
g
e
A
n
a
l
y
s
i
s
,
v
o
l
.
5
4
,
p
p
.
2
1
5
–
2
3
4
,
2
0
1
9
.
[
1
2
]
C
.
M
.
B
i
sh
o
p
a
n
d
N
.
M
.
N
a
sr
a
b
a
d
i
,
P
a
t
t
e
r
n
r
e
c
o
g
n
i
t
i
o
n
a
n
d
m
a
c
h
i
n
e
l
e
a
r
n
i
n
g
,
v
o
l
.
4
,
n
o
.
4
.
S
p
r
i
n
g
e
r
,
2
0
0
6
.
[
1
3
]
R
.
G
i
r
s
h
i
c
k
,
J.
D
o
n
a
h
u
e
,
T.
D
a
r
r
e
l
l
,
a
n
d
J.
M
a
l
i
k
,
“
R
i
c
h
f
e
a
t
u
r
e
h
i
e
r
a
r
c
h
i
e
s
f
o
r
a
c
c
u
r
a
t
e
o
b
j
e
c
t
d
e
t
e
c
t
i
o
n
a
n
d
sem
a
n
t
i
c
seg
m
e
n
t
a
t
i
o
n
,
”
i
n
Pr
o
c
e
e
d
i
n
g
s
o
f
t
h
e
I
EEE
c
o
n
f
e
re
n
c
e
o
n
c
o
m
p
u
t
e
r v
i
si
o
n
a
n
d
p
a
t
t
e
r
n
r
e
c
o
g
n
i
t
i
o
n
,
2
0
1
4
,
p
p
.
5
8
0
–
5
8
7
.
[
1
4
]
K
.
H
e
,
X
.
Z
h
a
n
g
,
S
.
R
e
n
,
a
n
d
J
.
S
u
n
,
“
D
e
e
p
r
e
si
d
u
a
l
l
e
a
r
n
i
n
g
f
o
r
i
ma
g
e
r
e
c
o
g
n
i
t
i
o
n
,
”
i
n
2
0
1
6
I
EEE
C
o
n
f
e
r
e
n
c
e
o
n
C
o
m
p
u
t
e
r
Vi
si
o
n
a
n
d
Pa
t
t
e
rn
Re
c
o
g
n
i
t
i
o
n
(
C
V
P
R)
,
Ju
n
.
2
0
1
6
,
p
p
.
7
7
0
–
778
,
d
o
i
:
1
0
.
1
1
0
9
/
C
V
P
R
.
2
0
1
6
.
9
0
.
[
1
5
]
H
.
Z
h
a
o
,
J
.
S
h
i
,
X
.
Q
i
,
X
.
W
a
n
g
,
a
n
d
J
.
J
i
a
,
“
P
y
r
a
mi
d
sc
e
n
e
p
a
r
si
n
g
n
e
t
w
o
r
k
,
”
i
n
Pro
c
e
e
d
i
n
g
s
o
f
t
h
e
I
E
E
E
c
o
n
f
e
re
n
c
e
o
n
c
o
m
p
u
t
e
r
v
i
s
i
o
n
a
n
d
p
a
t
t
e
rn
re
c
o
g
n
i
t
i
o
n
,
2
0
1
7
,
p
p
.
2
8
8
1
–
2
8
9
0
.
[
1
6
]
K
.
H
e
,
G
.
G
k
i
o
x
a
r
i
,
P
.
D
o
l
l
á
r
,
a
n
d
R
.
G
i
r
sh
i
c
k
,
“
M
a
sk
R
-
C
N
N
,
”
i
n
P
ro
c
e
e
d
i
n
g
s
o
f
t
h
e
I
EEE
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
m
p
u
t
e
r
v
i
s
i
o
n
,
2
0
1
7
,
p
p
.
2
9
6
1
–
2
9
6
9
.
[
1
7
]
B
.
X
i
e
,
S
.
Li
,
M
.
Li
,
C
.
H
.
L
i
u
,
G
.
H
u
a
n
g
,
a
n
d
G
.
W
a
n
g
,
“
S
e
P
i
C
o
:
sem
a
n
t
i
c
-
g
u
i
d
e
d
p
i
x
e
l
c
o
n
t
r
a
s
t
f
o
r
d
o
ma
i
n
a
d
a
p
t
i
v
e
sem
a
n
t
i
c
seg
m
e
n
t
a
t
i
o
n
,
”
I
EEE
T
ra
n
s
a
c
t
i
o
n
s
o
n
P
a
t
t
e
r
n
A
n
a
l
y
s
i
s
a
n
d
M
a
c
h
i
n
e
I
n
t
e
l
l
i
g
e
n
c
e
,
v
o
l
.
4
5
,
n
o
.
7
,
p
p
.
9
0
0
4
–
9
0
2
1
,
Ju
l
.
2
0
2
3
,
d
o
i
:
1
0
.
1
1
0
9
/
TPA
M
I
.
2
0
2
3
.
3
2
3
7
7
4
0
.
[
1
8
]
X
.
C
a
o
,
H
.
Zo
u
,
J.
L
i
,
X
.
Y
i
n
g
,
a
n
d
S
.
H
e
,
“
O
B
B
I
n
st
:
r
e
m
o
t
e
s
e
n
s
i
n
g
i
n
st
a
n
c
e
se
g
me
n
t
a
t
i
o
n
w
i
t
h
o
r
i
e
n
t
e
d
b
o
u
n
d
i
n
g
b
o
x
su
p
e
r
v
i
si
o
n
,
”
I
n
t
e
r
n
a
t
i
o
n
a
l
J
o
u
r
n
a
l
o
f
Ap
p
l
i
e
d
E
a
rt
h
O
b
ser
v
a
t
i
o
n
a
n
d
G
e
o
i
n
f
o
rm
a
t
i
o
n
,
v
o
l
.
1
2
8
,
p
.
1
0
3
7
1
7
,
A
p
r
.
2
0
2
4
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
j
a
g
.
2
0
2
4
.
1
0
3
7
1
7
.
[
1
9
]
L.
B
o
t
t
o
u
,
“
S
t
o
c
h
a
s
t
i
c
g
r
a
d
i
e
n
t
d
e
s
c
e
n
t
t
r
i
c
k
s,
”
L
e
c
t
u
r
e
N
o
t
e
s
i
n
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
(
i
n
c
l
u
d
i
n
g
su
b
se
ri
e
s
L
e
c
t
u
re
N
o
t
e
s
i
n
Art
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
a
n
d
L
e
c
t
u
r
e
N
o
t
e
s
i
n
B
i
o
i
n
f
o
rm
a
t
i
c
s)
,
v
o
l
.
7
7
0
0
L
EC
TU
,
p
p
.
4
2
1
–
4
3
6
,
2
0
1
2
,
d
o
i
:
1
0
.
1
0
0
7
/
9
7
8
-
3
-
6
4
2
-
3
5
2
8
9
-
8
_
2
5
.
[
2
0
]
Y
.
Le
C
u
n
,
Y
.
B
e
n
g
i
o
,
a
n
d
G
.
H
i
n
t
o
n
,
“
D
e
e
p
l
e
a
r
n
i
n
g
,
”
N
a
t
u
re
,
v
o
l
.
5
2
1
,
n
o
.
7
5
5
3
,
p
p
.
4
3
6
–
4
4
4
,
M
a
y
2
0
1
5
,
d
o
i
:
1
0
.
1
0
3
8
/
n
a
t
u
r
e
1
4
5
3
9
.
[
2
1
]
G
.
H
i
n
t
o
n
e
t
a
l
.
,
“
D
e
e
p
n
e
u
r
a
l
n
e
t
w
o
r
k
s
f
o
r
a
c
o
u
s
t
i
c
m
o
d
e
l
i
n
g
i
n
s
p
e
e
c
h
r
e
c
o
g
n
i
t
i
o
n
:
t
h
e
sh
a
r
e
d
v
i
e
w
s
o
f
f
o
u
r
r
e
sea
r
c
h
g
r
o
u
p
s
,
”
I
EEE
S
i
g
n
a
l
Pr
o
c
e
ssi
n
g
M
a
g
a
z
i
n
e
,
v
o
l
.
2
9
,
n
o
.
6
,
p
p
.
8
2
–
9
7
,
N
o
v
.
2
0
1
2
,
d
o
i
:
1
0
.
1
1
0
9
/
M
S
P
.
2
0
1
2
.
2
2
0
5
5
9
7
.
[
2
2
]
D
.
P
.
K
i
n
g
m
a
a
n
d
J.
B
a
,
“
A
d
a
m
:
a
m
e
t
h
o
d
f
o
r
st
o
c
h
a
s
t
i
c
o
p
t
i
mi
z
a
t
i
o
n
,
”
3
rd
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
L
e
a
r
n
i
n
g
Re
p
r
e
se
n
t
a
t
i
o
n
s
(
I
C
L
R)
,
2
0
1
5
.
[
2
3
]
A
.
A
r
n
a
b
,
S
.
J
a
y
a
s
u
ma
n
a
,
S
.
Zh
e
n
g
,
a
n
d
P
.
H
.
S
.
T
o
r
r
,
“
H
i
g
h
e
r
o
r
d
e
r
c
o
n
d
i
t
i
o
n
a
l
r
a
n
d
o
m
f
i
e
l
d
s
i
n
d
e
e
p
n
e
u
r
a
l
n
e
t
w
o
r
k
s,
”
i
n
Eu
r
o
p
e
a
n
c
o
n
f
e
re
n
c
e
o
n
c
o
m
p
u
t
e
r
v
i
s
i
o
n
,
2
0
1
6
,
p
p
.
5
2
4
–
5
4
0
.
[
2
4
]
Y
.
Y
u
a
n
,
X
.
C
h
e
n
,
a
n
d
J.
W
a
n
g
,
“
O
b
j
e
c
t
-
c
o
n
t
e
x
t
u
a
l
r
e
p
r
e
s
e
n
t
a
t
i
o
n
s
f
o
r
sema
n
t
i
c
se
g
m
e
n
t
a
t
i
o
n
,
”
i
n
Eu
r
o
p
e
a
n
c
o
n
f
e
re
n
c
e
o
n
c
o
m
p
u
t
e
r
v
i
s
i
o
n
,
2
0
2
0
,
p
p
.
1
7
3
–
1
9
0
.
[
2
5
]
H
.
W
u
,
Y
.
L
i
,
L
.
C
h
e
n
,
X
.
L
i
u
,
a
n
d
P
.
L
i
,
“
D
e
e
p
b
o
u
n
d
a
r
y
-
a
w
a
r
e
s
e
m
a
n
t
i
c
i
m
a
g
e
se
g
me
n
t
a
t
i
o
n
,
”
C
o
m
p
u
t
e
r
An
i
m
a
t
i
o
n
a
n
d
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.
15
,
No
.
6
,
Decem
b
e
r
20
25
:
5
3
2
7
-
5
3
3
5
5334
Vi
rt
u
a
l
W
o
rl
d
s
,
v
o
l
.
3
2
,
n
o
.
3
–
4
,
2
0
2
1
,
d
o
i
:
1
0
.
1
0
0
2
/
c
a
v
.
2
0
2
3
.
[
2
6
]
Q
.
L
i
u
,
H
.
S
u
,
M
.
El
-
K
h
a
m
y
,
a
n
d
K
.
-
B
.
S
o
n
g
,
“
D
e
e
p
G
B
A
S
S
:
d
e
e
p
g
u
i
d
e
d
b
o
u
n
d
a
r
y
-
a
w
a
r
e
se
ma
n
t
i
c
se
g
me
n
t
a
t
i
o
n
,
”
i
n
I
C
A
S
S
P
2
0
2
2
-
2
0
2
2
I
EEE
I
n
t
e
rn
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
Ac
o
u
st
i
c
s,
S
p
e
e
c
h
a
n
d
S
i
g
n
a
l
Pr
o
c
e
ssi
n
g
(
I
C
A
S
S
P)
,
M
a
y
2
0
2
2
,
p
p
.
1
5
2
5
–
1
5
2
9
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
A
S
S
P
4
3
9
2
2
.
2
0
2
2
.
9
7
4
7
8
9
2
.
[
2
7
]
W
.
H
u
a
n
g
a
n
d
F
.
Z
h
a
n
g
,
“
S
e
mi
-
s
u
p
e
r
v
i
s
e
d
s
e
ma
n
t
i
c
s
e
g
me
n
t
a
t
i
o
n
w
i
t
h
s
t
r
u
c
t
u
r
e
d
o
u
t
p
u
t
s
p
a
c
e
a
d
a
p
t
i
o
n
,
”
i
n
I
C
A
S
S
P
2
0
2
3
-
2
0
2
3
I
EEE
I
n
t
e
rn
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
Ac
o
u
s
t
i
c
s,
S
p
e
e
c
h
a
n
d
S
i
g
n
a
l
Pr
o
c
e
s
si
n
g
(
I
C
A
S
S
P)
,
Ju
n
.
2
0
2
3
,
p
p
.
1
–
5
,
doi
:
1
0
.
1
1
0
9
/
I
C
A
S
S
P
4
9
3
5
7
.
2
0
2
3
.
1
0
0
9
5
6
2
8
.
[
2
8
]
T.
Ji
a
n
g
,
Y
.
J
i
n
,
T.
Li
a
n
g
,
X
.
W
a
n
g
,
a
n
d
Y
.
Li
,
“
B
o
u
n
d
a
r
y
c
o
r
r
e
c
t
e
d
m
u
l
t
i
-
sc
a
l
e
f
u
s
i
o
n
n
e
t
w
o
r
k
f
o
r
r
e
a
l
-
t
i
me
s
e
ma
n
t
i
c
seg
m
e
n
t
a
t
i
o
n
,
”
Pr
o
c
e
e
d
i
n
g
s
-
I
n
t
e
rn
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
I
m
a
g
e
Pro
c
e
ssi
n
g
,
I
C
I
P
,
p
p
.
1
8
8
6
–
1
8
9
0
,
2
0
2
2
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
I
P
4
6
5
7
6
.
2
0
2
2
.
9
8
9
7
9
0
7
.
[
2
9
]
M
.
Y
a
n
g
,
K
.
Y
u
,
C
.
Z
h
a
n
g
,
Z
.
L
i
,
a
n
d
K
.
Y
a
n
g
,
“
S
e
ma
n
t
i
c
h
i
e
r
a
r
c
h
y
-
a
w
a
r
e
se
g
me
n
t
a
t
i
o
n
,
”
I
EEE
T
r
a
n
s
a
c
t
i
o
n
s
o
n
P
a
t
t
e
r
n
An
a
l
y
si
s
a
n
d
M
a
c
h
i
n
e
I
n
t
e
l
l
i
g
e
n
c
e
,
v
o
l
.
4
6
,
n
o
.
1
0
,
p
p
.
1
1
1
1
9
–
1
1
1
3
5
,
A
p
r
.
2
0
2
4
,
d
o
i
:
1
0
.
1
1
0
9
/
TPA
M
I
.
2
0
2
3
.
3
3
3
2
4
3
5
.
[
3
0
]
H
.
K
e
r
v
a
d
e
c
,
J.
B
o
u
c
h
t
i
b
a
,
C
.
D
e
s
r
o
si
e
r
s
,
E
.
G
r
a
n
g
e
r
,
J
.
D
o
l
z
,
a
n
d
I
.
B
e
n
A
y
e
d
,
“
B
o
u
n
d
a
r
y
l
o
ss
f
o
r
h
i
g
h
l
y
u
n
b
a
l
a
n
c
e
d
seg
m
e
n
t
a
t
i
o
n
,
”
i
n
I
n
t
e
r
n
a
t
i
o
n
a
l
c
o
n
f
e
ren
c
e
o
n
m
e
d
i
c
a
l
i
m
a
g
i
n
g
w
i
t
h
d
e
e
p
l
e
a
r
n
i
n
g
,
2
0
1
9
,
p
p
.
2
8
5
–
2
9
6
.
[
3
1
]
L.
-
C
.
C
h
e
n
,
Y
.
Zh
u
,
G
.
P
a
p
a
n
d
r
e
o
u
,
F
.
S
c
h
r
o
f
f
,
a
n
d
H
.
A
d
a
m
,
“
E
n
c
o
d
e
r
-
d
e
c
o
d
e
r
w
i
t
h
A
t
r
o
u
s
se
p
a
r
a
b
l
e
c
o
n
v
o
l
u
t
i
o
n
f
o
r
sem
a
n
t
i
c
i
ma
g
e
s
e
g
me
n
t
a
t
i
o
n
,
”
i
n
Pro
c
e
e
d
i
n
g
s
o
f
t
h
e
Eu
r
o
p
e
a
n
c
o
n
f
e
r
e
n
c
e
o
n
c
o
m
p
u
t
e
r
v
i
s
i
o
n
(
E
C
C
V)
,
2
0
1
8
,
p
p
.
8
0
1
–
8
1
8
.
[
3
2
]
F
.
Y
u
a
n
d
V
.
K
o
l
t
u
n
,
“
M
u
l
t
i
-
sca
l
e
c
o
n
t
e
x
t
a
g
g
r
e
g
a
t
i
o
n
b
y
d
i
l
a
t
e
d
c
o
n
v
o
l
u
t
i
o
n
s
,
”
a
rX
i
v
p
r
e
p
ri
n
t
a
rX
i
v
:
1
5
1
1
.
0
7
1
2
2
,
2
0
1
5
.
[
3
3
]
T.
Ta
k
i
k
a
w
a
,
D
.
A
c
u
n
a
,
V
.
J
a
m
p
a
n
i
,
a
n
d
S
.
F
i
d
l
e
r
,
“
G
a
t
e
d
-
S
C
N
N
:
g
a
t
e
d
s
h
a
p
e
C
N
N
s fo
r
sem
a
n
t
i
c
se
g
me
n
t
a
t
i
o
n
,
”
i
n
Pro
c
e
e
d
i
n
g
s
o
f
t
h
e
I
E
EE/
C
V
F
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
c
o
m
p
u
t
e
r
v
i
s
i
o
n
,
2
0
1
9
,
p
p
.
5
2
2
9
–
5
2
3
8
.
[
3
4
]
Z.
C
h
e
n
,
A
.
V
O
i
k
o
n
o
mo
u
,
a
n
d
P
.
F
i
e
g
u
t
h
,
“
S
e
g
me
n
t
a
t
i
o
n
-
a
w
a
r
e
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
t
w
o
r
k
s
u
si
n
g
l
o
c
a
l
a
t
t
e
n
t
i
o
n
m
a
s
k
s,”
i
n
Pro
c
e
e
d
i
n
g
s
o
f
t
h
e
I
E
EE
I
n
t
e
rn
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
C
o
m
p
u
t
e
r V
i
si
o
n
,
2
0
1
7
,
v
o
l
.
2
9
,
p
p
.
5
0
3
8
–
5
0
4
7
.
[
3
5
]
D
.
C
o
ma
n
i
c
i
u
a
n
d
P
.
M
e
e
r
,
“
M
e
a
n
sh
i
f
t
:
a
r
o
b
u
s
t
a
p
p
r
o
a
c
h
t
o
w
a
r
d
f
e
a
t
u
r
e
s
p
a
c
e
a
n
a
l
y
si
s
,
”
I
EEE
T
r
a
n
sa
c
t
i
o
n
s
o
n
P
a
t
t
e
r
n
An
a
l
y
si
s
a
n
d
M
a
c
h
i
n
e
I
n
t
e
l
l
i
g
e
n
c
e
,
v
o
l
.
2
4
,
n
o
.
5
,
p
p
.
6
0
3
–
6
1
9
,
M
a
y
2
0
0
2
,
d
o
i
:
1
0
.
1
1
0
9
/
3
4
.
1
0
0
0
2
3
6
.
[
3
6
]
Y
.
B
o
y
k
o
v
,
O
.
V
e
k
s
l
e
r
,
a
n
d
R
.
Z
a
b
i
h
,
“
F
a
s
t
a
p
p
r
o
x
i
ma
t
e
e
n
e
r
g
y
mi
n
i
mi
z
a
t
i
o
n
v
i
a
g
r
a
p
h
c
u
t
s,
”
I
E
EE
T
r
a
n
sa
c
t
i
o
n
s
o
n
P
a
t
t
e
r
n
An
a
l
y
si
s
a
n
d
M
a
c
h
i
n
e
I
n
t
e
l
l
i
g
e
n
c
e
,
v
o
l
.
2
3
,
n
o
.
1
1
,
p
p
.
1
2
2
2
–
1
2
3
9
,
2
0
0
1
,
d
o
i
:
1
0
.
1
1
0
9
/
3
4
.
9
6
9
1
1
4
.
[
3
7
]
J.
M
c
Q
u
e
e
n
,
“
S
o
me
me
t
h
o
d
s
f
o
r
c
l
a
ssi
f
i
c
a
t
i
o
n
a
n
d
a
n
a
l
y
s
i
s
o
f
m
u
l
t
i
v
a
r
i
a
t
e
o
b
ser
v
a
t
i
o
n
s,”
i
n
F
i
f
t
h
B
e
rk
e
l
e
y
S
y
m
p
o
s
i
u
m
o
n
Ma
t
h
e
m
a
t
i
c
a
l
S
t
a
t
i
st
i
c
s
a
n
d
Pr
o
b
a
b
i
l
i
t
y
,
1
9
6
7
,
p
p
.
2
8
1
–
2
9
7
.
[
3
8
]
L.
B
r
e
i
m
a
n
,
“
R
a
n
d
o
m fo
r
e
st
s,
”
Ma
c
h
i
n
e
L
e
a
r
n
i
n
g
,
v
o
l
.
4
5
,
n
o
.
1
,
p
p
.
5
–
3
2
,
2
0
0
1
.
[
3
9
]
T.
C
o
v
e
r
a
n
d
P
.
H
a
r
t
,
“
N
e
a
r
e
st
n
e
i
g
h
b
o
r
p
a
t
t
e
r
n
c
l
a
ss
i
f
i
c
a
t
i
o
n
,
”
I
EE
E
T
ra
n
s
a
c
t
i
o
n
s
o
n
I
n
f
o
rm
a
t
i
o
n
T
h
e
o
ry
,
v
o
l
.
1
3
,
n
o
.
1
,
p
p
.
2
1
–
2
7
,
Ja
n
.
1
9
6
7
,
d
o
i
:
1
0
.
1
1
0
9
/
TI
T.
1
9
6
7
.
1
0
5
3
9
6
4
.
[
4
0
]
A
.
G
a
r
c
i
a
-
G
a
r
c
i
a
,
S
.
O
r
t
s
-
Es
c
o
l
a
n
o
,
S
.
O
p
r
e
a
,
V
.
V
i
l
l
e
n
a
-
M
a
r
t
i
n
e
z
,
a
n
d
J.
G
.
R
o
d
r
i
g
u
e
z
,
“
A
r
e
v
i
e
w
o
n
d
e
e
p
l
e
a
r
n
i
n
g
t
e
c
h
n
i
q
u
e
s
a
p
p
l
i
e
d
t
o
sem
a
n
t
i
c
se
g
m
e
n
t
a
t
i
o
n
,
”
I
E
EE
T
r
a
n
s
a
c
t
i
o
n
s
o
n
C
y
b
e
rn
e
t
i
c
s
,
v
o
l
.
5
0
,
n
o
.
4
,
p
p
.
1
2
0
5
–
1
2
2
0
,
2
0
2
0
.
[
4
1
]
B
.
B
i
sc
h
k
e
,
P
.
H
e
l
b
e
r
,
J.
F
o
l
z
,
D
.
B
o
r
t
h
,
a
n
d
A
.
D
e
n
g
e
l
,
“
M
u
l
t
i
-
t
a
s
k
l
e
a
r
n
i
n
g
f
o
r
s
e
g
m
e
n
t
a
t
i
o
n
o
f
b
u
i
l
d
i
n
g
f
o
o
t
p
r
i
n
t
s
w
i
t
h
d
e
e
p
n
e
u
r
a
l
n
e
t
w
o
r
k
s
,
”
i
n
2
0
1
9
I
EEE
I
n
t
e
rn
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
I
m
a
g
e
Pro
c
e
ssi
n
g
(
I
C
I
P)
,
S
e
p
.
2
0
1
9
,
p
p
.
1
4
8
0
–
1
4
8
4
,
doi
:
1
0
.
1
1
0
9
/
I
C
I
P
.
2
0
1
9
.
8
8
0
3
0
5
0
.
[
4
2
]
K
.
S
i
mo
n
y
a
n
a
n
d
A
.
Z
i
sserm
a
n
,
“
V
e
r
y
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
f
o
r
l
a
r
g
e
-
sc
a
l
e
i
m
a
g
e
r
e
c
o
g
n
i
t
i
o
n
,
”
i
n
3
rd
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
L
e
a
rn
i
n
g
Re
p
res
e
n
t
a
t
i
o
n
s
,
I
C
L
R
2
0
1
5
-
C
o
n
f
e
re
n
c
e
T
r
a
c
k
P
ro
c
e
e
d
i
n
g
s
,
2
0
1
5
,
p
p
.
1
–
1
4
.
[
4
3
]
L.
Y
a
n
,
J.
H
u
a
n
g
,
H
.
X
i
e
,
P
.
W
e
i
,
a
n
d
Z
.
G
a
o
,
“
Ef
f
i
c
i
e
n
t
d
e
p
t
h
f
u
s
i
o
n
t
r
a
n
sf
o
r
mer
f
o
r
a
e
r
i
a
l
i
ma
g
e
sem
a
n
t
i
c
se
g
me
n
t
a
t
i
o
n
,
”
Re
m
o
t
e
S
e
n
s
i
n
g
,
v
o
l
.
1
4
,
n
o
.
5
,
p
.
1
2
9
4
,
M
a
r
.
2
0
2
2
,
d
o
i
:
1
0
.
3
3
9
0
/
r
s1
4
0
5
1
2
9
4
.
[
4
4
]
X
.
Z
h
a
n
g
,
Y
.
W
e
i
,
J
.
F
e
n
g
,
Y
.
Y
a
n
g
,
a
n
d
T
.
S
.
H
u
a
n
g
,
“
A
d
v
e
r
sari
a
l
c
o
mp
l
e
m
e
n
t
a
r
y
l
e
a
r
n
i
n
g
f
o
r
w
e
a
k
l
y
s
u
p
e
r
v
i
se
d
o
b
j
e
c
t
l
o
c
a
l
i
z
a
t
i
o
n
,
”
i
n
I
E
EE
C
o
n
f
e
re
n
c
e
o
n
C
o
m
p
u
t
e
r
V
i
si
o
n
a
n
d
P
a
t
t
e
r
n
R
e
c
o
g
n
i
t
i
o
n
(
C
V
PR)
,
J
u
n
.
2
0
1
8
,
p
p
.
1
3
3
5
–
1
3
4
4
,
d
o
i
:
1
0
.
1
1
0
9
/
C
V
P
R
.
2
0
1
8
.
0
0
1
4
5
.
[
4
5
]
T.
C
h
e
n
,
L
.
L
i
n
,
X
.
W
u
,
N
.
X
i
a
o
,
a
n
d
X
.
L
u
o
,
“
L
e
a
r
n
i
n
g
t
o
s
e
g
m
e
n
t
o
b
j
e
c
t
c
a
n
d
i
d
a
t
e
s
v
i
a
r
e
c
u
r
s
i
v
e
n
e
u
r
a
l
n
e
t
w
o
r
k
s,”
I
EEE
T
ra
n
s
a
c
t
i
o
n
s
o
n
I
m
a
g
e
Pro
c
e
ss
i
n
g
,
v
o
l
.
2
7
,
n
o
.
1
2
,
p
p
.
5
8
2
7
–
5
8
3
9
,
D
e
c
.
2
0
1
8
,
d
o
i
:
1
0
.
1
1
0
9
/
TI
P
.
2
0
1
8
.
2
8
5
9
0
2
5
.
[
4
6
]
M
.
C
o
r
d
t
s
e
t
a
l
.
,
“
T
h
e
c
i
t
y
s
c
a
p
e
s d
a
t
a
set
f
o
r
se
ma
n
t
i
c
u
r
b
a
n
sc
e
n
e
u
n
d
e
r
s
t
a
n
d
i
n
g
,
”
i
n
2
0
1
6
I
E
EE
C
o
n
f
e
re
n
c
e
o
n
C
o
m
p
u
t
e
r
V
i
si
o
n
a
n
d
Pa
t
t
e
rn
Re
c
o
g
n
i
t
i
o
n
(
C
V
PR)
,
Ju
n
.
2
0
1
6
,
p
p
.
3
2
1
3
–
3
2
2
3
,
d
o
i
:
1
0
.
1
1
0
9
/
C
V
P
R
.
2
0
1
6
.
3
5
0
.
[
4
7
]
T.
L
u
,
D
.
M
i
n
g
,
X
.
Li
n
,
Z.
H
o
n
g
,
X
.
B
a
i
,
a
n
d
J
.
F
a
n
g
,
“
D
e
t
e
c
t
i
n
g
b
u
i
l
d
i
n
g
e
d
g
e
s
f
r
o
m
h
i
g
h
s
p
a
t
i
a
l
r
e
s
o
l
u
t
i
o
n
r
e
mo
t
e
se
n
si
n
g
i
ma
g
e
r
y
u
si
n
g
r
i
c
h
e
r
c
o
n
v
o
l
u
t
i
o
n
f
e
a
t
u
r
e
s
n
e
t
w
o
r
k
,
”
R
e
m
o
t
e
S
e
n
s
i
n
g
,
v
o
l
.
1
0
,
n
o
.
9
,
p
.
1
4
9
6
,
S
e
p
.
2
0
1
8
,
d
o
i
:
1
0
.
3
3
9
0
/
r
s
1
0
0
9
1
4
9
6
.
[
4
8
]
Y
.
Lu
,
R
.
J
i
a
o
,
S
.
D
a
i
,
a
n
d
H
.
Z
h
a
n
g
,
“
B
o
u
n
d
a
r
y
-
a
w
a
r
e
f
e
a
t
u
r
e
p
r
o
p
a
g
a
t
i
o
n
f
o
r
sc
e
n
e
se
g
me
n
t
a
t
i
o
n
,
”
i
n
Pr
o
c
e
e
d
i
n
g
s
o
f
t
h
e
I
EEE/
C
VF
C
o
n
f
e
r
e
n
c
e
o
n
C
o
m
p
u
t
e
r
Vi
si
o
n
a
n
d
Pa
t
t
e
rn
Re
c
o
g
n
i
t
i
o
n
(
C
V
P
R)
,
O
c
t
.
2
0
2
2
,
p
p
.
1
2
4
0
–
1
2
4
9
.
[
4
9
]
F
.
Y
u
,
V
.
K
o
l
t
u
n
,
a
n
d
T.
F
u
n
k
h
o
u
s
e
r
,
“
D
i
l
a
t
e
d
r
e
si
d
u
a
l
n
e
t
w
o
r
k
s,
”
i
n
2
0
1
7
I
E
EE
C
o
n
f
e
re
n
c
e
o
n
C
o
m
p
u
t
e
r
V
i
si
o
n
a
n
d
Pa
t
t
e
r
n
Re
c
o
g
n
i
t
i
o
n
(
C
VPR)
,
J
u
l
.
2
0
1
7
,
p
p
.
6
3
6
–
6
4
4
,
d
o
i
:
1
0
.
1
1
0
9
/
C
V
P
R
.
2
0
1
7
.
7
5
.
[
5
0
]
Y
.
W
a
n
g
,
Z.
J
i
n
,
S
.
Z
h
a
o
,
a
n
d
X
.
C
h
e
n
,
“
3
5
9
B
o
u
n
d
a
r
y
-
a
w
a
r
e
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
sk
i
n
l
e
si
o
n
seg
m
e
n
t
a
t
i
o
n
i
n
c
l
i
n
i
c
a
l
i
ma
g
e
s
,
”
J
o
u
r
n
a
l
o
f
I
n
v
e
s
t
i
g
a
t
i
v
e
D
e
rm
a
t
o
l
o
g
y
,
v
o
l
.
1
4
1
,
n
o
.
5
,
p
.
S
6
3
,
M
a
y
2
0
2
1
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
j
i
d
.
2
0
2
1
.
0
2
.
3
8
1
.
[
5
1
]
X
.
H
u
a
n
g
,
Y
.
Li
,
a
n
d
C
.
C
.
L
o
y
,
“
Le
a
r
n
i
n
g
p
i
x
e
l
-
l
e
v
e
l
sema
n
t
i
c
a
f
f
i
n
i
t
y
w
i
t
h
i
ma
g
e
-
l
e
v
e
l
su
p
e
r
v
i
si
o
n
f
o
r
w
e
a
k
l
y
su
p
e
r
v
i
s
e
d
sema
n
t
i
c
s
e
g
me
n
t
a
t
i
o
n
,
”
I
EE
E
T
r
a
n
s
a
c
t
i
o
n
s
o
n
P
a
t
t
e
r
n
A
n
a
l
y
si
s
a
n
d
M
a
c
h
i
n
e
I
n
t
e
l
l
i
g
e
n
c
e
,
v
o
l
.
4
4
,
n
o
.
1
0
,
p
p
.
6
8
3
4
–
6
8
4
2
,
2
0
2
2
.
[
5
2
]
J.
C
a
n
n
y
,
“
A
c
o
m
p
u
t
a
t
i
o
n
a
l
a
p
p
r
o
a
c
h
t
o
e
d
g
e
d
e
t
e
c
t
i
o
n
,
”
I
EEE
T
ra
n
s
a
c
t
i
o
n
s
o
n
Pa
t
t
e
rn
A
n
a
l
y
s
i
s
a
n
d
M
a
c
h
i
n
e
I
n
t
e
l
l
i
g
e
n
c
e
,
v
o
l
.
P
A
M
I
-
8
,
n
o
.
6
,
p
p
.
6
7
9
–
6
9
8
,
N
o
v
.
1
9
8
6
,
d
o
i
:
1
0
.
1
1
0
9
/
TPA
M
I
.
1
9
8
6
.
4
7
6
7
8
5
1
.
[
5
3
]
R
.
G
o
n
z
a
l
e
z
a
n
d
R
.
W
o
o
d
s
,
D
i
g
i
t
a
l
i
m
a
g
e
p
ro
c
e
ss
i
n
g
,
3
r
d
e
d
.
N
e
w
J
e
r
se
y
:
P
r
e
n
t
i
c
e
-
H
a
l
l
,
2
0
0
8
.
[
5
4
]
T.
P
a
v
l
i
d
i
s
,
“
A
t
h
i
n
n
i
n
g
a
l
g
o
r
i
t
h
m
f
o
r
d
i
s
c
r
e
t
e
b
i
n
a
r
y
i
m
a
g
e
s,”
C
o
m
p
u
t
e
r
G
ra
p
h
i
c
s
a
n
d
I
m
a
g
e
Pr
o
c
e
ssi
n
g
,
v
o
l
.
1
3
,
n
o
.
2
,
p
p
.
1
4
2
–
1
5
7
,
J
a
n
.
1
9
8
0
,
d
o
i
:
1
0
.
1
0
1
6
/
S
0
1
4
6
-
6
6
4
X
(
8
0
)
8
0
0
3
7
-
2.
[
5
5
]
A
.
B
o
z
o
r
g
p
o
u
r
,
Y
.
S
a
d
e
g
h
e
i
h
,
A
.
K
a
z
e
r
o
u
n
i
,
R
.
A
z
a
d
,
a
n
d
D
.
M
e
r
h
o
f
,
“
D
e
r
mo
S
e
g
D
i
f
f
:
A
b
o
u
n
d
a
r
y
-
a
w
a
r
e
se
g
me
n
t
a
t
i
o
n
d
i
f
f
u
s
i
o
n
mo
d
e
l
f
o
r
s
k
i
n
l
e
si
o
n
d
e
l
i
n
e
a
t
i
o
n
,
”
2
0
2
3
,
p
p
.
1
4
6
–
158
,
d
o
i
:
1
0
.
1
0
0
7
/
9
7
8
-
3
-
0
3
1
-
4
6
0
0
5
-
0
_
1
3
.
[
5
6
]
M
.
Ev
e
r
i
n
g
h
a
m
e
t
a
l
.
,
“
T
h
e
P
A
S
C
A
L
v
i
su
a
l
o
b
j
e
c
t
c
l
a
sses
(
V
O
C
)
c
h
a
l
l
e
n
g
e
,
”
I
n
t
e
r
n
a
t
i
o
n
a
l
J
o
u
r
n
a
l
o
f
C
o
m
p
u
t
e
r
V
i
si
o
n
m
a
n
u
scr
i
p
t
,
v
o
l
.
8
8
,
n
o
.
2
,
p
p
.
3
0
3
–
3
3
8
,
2
0
1
2
.
[
5
7
]
T.
-
Y
.
Li
n
,
P
.
D
o
l
l
a
r
,
R
.
G
i
r
s
h
i
c
k
,
K
.
H
e
,
B
.
H
a
r
i
h
a
r
a
n
,
a
n
d
S
.
B
e
l
o
n
g
i
e
,
“
F
e
a
t
u
r
e
p
y
r
a
m
i
d
n
e
t
w
o
r
k
s
f
o
r
o
b
j
e
c
t
d
e
t
e
c
t
i
o
n
,
”
i
n
Pro
c
e
e
d
i
n
g
s
o
f
t
h
e
I
E
EE
C
o
n
f
e
r
e
n
c
e
o
n
C
o
m
p
u
t
e
r
Vi
s
i
o
n
a
n
d
P
a
t
t
e
r
n
R
e
c
o
g
n
i
t
i
o
n
(
C
VP
R)
,
J
u
l
.
2
0
1
7
,
p
p
.
2
1
1
7
–
2
1
2
5
.
[
5
8
]
S
.
S
.
M
.
S
a
l
e
h
i
,
D
.
Er
d
o
g
mu
s
,
a
n
d
A
.
G
h
o
l
i
p
o
u
r
,
“
T
v
e
r
s
k
y
l
o
ss
f
u
n
c
t
i
o
n
f
o
r
i
ma
g
e
s
e
g
m
e
n
t
a
t
i
o
n
u
si
n
g
3
D
f
u
l
l
y
c
o
n
v
o
l
u
t
i
o
n
a
l
d
e
e
p
n
e
t
w
o
r
k
s,
”
2
0
1
7
,
p
p
.
3
7
9
–
3
8
7
,
d
o
i
:
1
0
.
1
0
0
7
/
9
7
8
-
3
-
3
1
9
-
6
7
3
8
9
-
9
_
4
4
.
[
5
9
]
S
.
X
i
e
a
n
d
Z.
T
u
,
“
H
o
l
i
s
t
i
c
a
l
l
y
-
n
e
s
t
e
d
e
d
g
e
d
e
t
e
c
t
i
o
n
,
”
i
n
2
0
1
5
I
EEE
I
n
t
e
rn
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
C
o
m
p
u
t
e
r
Vi
s
i
o
n
(
I
C
C
V)
,
D
e
c
.
2
0
1
5
,
p
p
.
1
3
9
5
–
1
4
0
3
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
C
V
.
2
0
1
5
.
1
6
4
.
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
E
n
h
a
n
cin
g
s
ema
n
tic
s
eg
men
t
a
tio
n
w
ith
a
b
o
u
n
d
a
r
y
-
s
en
s
itive
…
(
Ga
n
esh
R
.
P
a
d
a
lka
r
)
5335
B
I
O
G
RAP
H
I
E
S O
F
AUTH
O
RS
G
a
n
e
sh
R.
Pa
d
a
lk
a
r
re
c
e
iv
e
d
h
is
M
.
E.
d
e
g
re
e
fr
o
m
S
h
iv
a
ji
Un
i
v
e
rsity
,
Ko
lh
a
p
u
r,
M
a
h
a
ra
sh
tra,
In
d
ia
in
2
0
0
9
.
He
is
wo
rk
in
g
a
s
a
ss
istan
t
p
ro
fe
ss
o
r
a
t
M
KSS
S
S
’
Cu
m
m
in
s
Co
ll
e
g
e
o
f
E
n
g
i
n
e
e
rin
g
f
o
r
W
o
m
e
n
,
a
n
a
u
to
n
o
m
o
u
s
in
stit
u
te
a
ffil
iate
d
wit
h
S
a
v
it
rib
a
i
P
h
u
le
P
u
n
e
Un
iv
e
rsity
,
P
u
n
e
,
M
a
h
a
ra
sh
tra,
I
n
d
ia.
He
is
a
lso
a
re
se
a
rc
h
c
a
n
d
id
a
te
a
t
S
a
v
it
rib
a
i
P
h
u
le
P
u
n
e
U
n
iv
e
rsit
y
,
P
u
n
e
.
G
a
n
e
sh
R.
P
a
d
a
l
k
a
r
is
th
e
a
u
t
h
o
r
o
f
6
tec
h
n
ica
l
p
u
b
li
c
a
ti
o
n
s.
His
re
se
a
rc
h
in
tere
sts
in
c
lu
d
e
ima
g
e
p
ro
c
e
ss
in
g
a
n
d
d
e
e
p
lea
rn
in
g
.
He
is
a
li
fe
m
e
m
b
e
r
o
f
th
e
In
d
ian
S
o
c
iet
y
fo
r
Tec
h
n
ica
l
Ed
u
c
a
ti
o
n
.
He
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
g
a
n
e
sh
.
p
a
d
a
lk
a
r@c
u
m
m
in
sc
o
ll
e
g
e
.
in
.
Ma
d
h
u
r
i
B.
K
h
a
m
b
e
te
re
c
e
i
v
e
d
h
e
r
P
h
D
d
e
g
re
e
in
2
0
0
9
i
n
th
e
a
re
a
o
f
ima
g
e
p
ro
c
e
ss
in
g
fro
m
th
e
S
a
v
it
ri
b
a
i
P
h
u
le
P
u
n
e
Un
iv
e
rsity
,
P
u
n
e
,
M
a
h
a
ra
sh
tra,
In
d
ia.
S
h
e
is
se
rv
in
g
a
s
a
d
irec
to
r
wit
h
t
h
e
M
KSS
S
’s
C
u
m
m
in
s
Co
ll
e
g
e
o
f
En
g
in
e
e
rin
g
f
o
r
W
o
m
e
n
,
Ka
rv
e
n
a
g
a
r,
P
u
n
e
.
S
h
e
a
lso
is
a
P
h
D
g
u
i
d
e
i
n
t
h
e
sa
m
e
re
se
a
rc
h
c
e
n
ter
.
Dr
M
a
d
h
u
ri
B.
Kh
a
m
b
e
te
is
t
h
e
a
u
t
h
o
r
o
f
o
v
e
r
2
4
tec
h
n
ica
l
p
u
b
li
c
a
ti
o
n
s.
He
r
re
se
a
rc
h
in
tere
sts
i
n
c
lu
d
e
ima
g
e
p
ro
c
e
ss
in
g
,
c
o
m
p
u
ter
v
isio
n
a
n
d
p
a
t
tern
re
c
o
g
n
it
i
o
n
.
S
h
e
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
p
rin
c
i
p
a
l@c
u
m
m
in
sc
o
ll
e
g
e
.
i
n
.
Evaluation Warning : The document was created with Spire.PDF for Python.