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
.
3
,
J
u
n
e
20
25
,
p
p
.
3118
~
3
1
2
6
I
SS
N:
2088
-
8
7
0
8
,
DOI
: 1
0
.
1
1
5
9
1
/ijece.
v
15
i
3
.
pp
3
1
1
8
-
3
1
2
6
3118
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
ec
e.
ia
esco
r
e.
co
m
SIGAN
:
A
g
enera
tive a
dv
ersa
ria
l n
etwork archi
tec
tu
re f
o
r
sketch
to pho
to sy
nthesis
B
ud
da
nn
a
g
a
ri
L
a
t
ha
,
At
hiy
o
o
r
K
a
nn
a
n Ve
lm
urug
a
n
D
e
p
a
r
t
me
n
t
o
f
C
o
mp
u
t
e
r
S
c
i
e
n
c
e
a
n
d
En
g
i
n
e
e
r
i
n
g
,
K
o
n
e
r
u
La
k
s
h
ma
i
a
h
E
d
u
c
a
t
i
o
n
F
o
u
n
d
a
t
i
o
n
,
G
u
n
t
u
r
,
I
n
d
i
a
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
J
u
l 1
3
,
2
0
2
4
R
ev
is
ed
Dec
2
2
,
2
0
2
4
Acc
ep
ted
Ma
r
4
,
2
0
2
5
Of
late
,
with
th
e
rise
o
f
a
rti
ficia
l
in
telli
g
e
n
c
e
(AI)
a
n
d
d
e
e
p
lea
r
n
in
g
(DL)
m
o
d
e
ls,
ima
g
e
tran
sla
ti
o
n
h
a
s
b
e
c
o
m
e
a
v
e
ry
imp
o
rtan
t
p
h
e
n
o
m
e
n
a
wh
ich
c
o
u
ld
p
r
o
d
u
c
e
re
a
li
stic
p
h
o
to
g
ra
p
h
ic
re
su
lt
s
.
S
y
n
t
h
e
siz
in
g
n
e
w
i
m
a
g
e
s
is
wid
e
ly
u
se
d
in
d
iffere
n
t
a
p
p
li
c
a
ti
o
n
s
in
c
lu
d
i
n
g
th
e
o
n
e
s
u
se
d
b
y
in
v
e
stig
a
t
io
n
a
g
e
n
c
ies
.
Im
a
g
e
g
e
n
e
ra
ti
o
n
fr
o
m
h
a
n
d
-
d
ra
wn
sk
e
tch
to
re
a
li
stic
p
h
o
to
s
a
n
d
v
ice
v
e
rsa
is
re
q
u
ired
in
d
iffere
n
t
c
o
m
p
u
ter
v
isi
o
n
a
p
p
li
c
a
ti
o
n
s.
G
e
n
e
r
a
ti
v
e
a
d
v
e
rsa
rial
n
e
two
r
k
(
G
A
N)
a
rc
h
it
e
c
tu
re
is
e
x
ten
siv
e
l
y
e
m
p
lo
y
e
d
fo
r
g
e
n
e
ra
ti
n
g
ima
g
e
s.
Ho
we
v
e
r,
th
e
re
is
n
e
e
d
fo
r
in
v
e
stig
a
ti
n
g
f
u
rth
e
r
o
n
imp
ro
v
isi
n
g
G
AN
a
r
c
h
it
e
c
tu
re
a
n
d
t
h
e
u
n
d
e
rly
i
n
g
lo
ss
fu
n
c
ti
o
n
s
to
wa
rd
s
lev
e
ra
g
in
g
p
e
rfo
rm
a
n
c
e
.
I
n
t
h
is
p
a
p
e
r,
we
p
u
t
f
o
rt
h
a
G
AN
a
rc
h
it
e
c
tu
re
k
n
o
w
n
a
s
sk
e
tc
h
-
ima
g
e
G
AN
(S
IG
AN
)
fo
r
sy
n
t
h
e
siz
in
g
re
a
li
stic
p
h
o
to
s
fro
m
h
a
n
d
-
d
ra
wn
s
k
e
tch
e
s.
B
o
th
g
e
n
e
ra
to
r
(G
)
a
n
d
d
isc
rimi
n
a
to
r
(D)
c
o
m
p
o
n
e
n
ts
a
re
d
e
sig
n
e
d
b
a
se
d
o
n
DL
m
o
d
e
ls
fo
ll
o
win
g
a
n
o
n
-
c
o
o
p
e
ra
ti
v
e
g
a
m
e
th
e
o
ry
t
o
wa
rd
s
imp
r
o
v
i
n
g
ima
g
e
g
e
n
e
ra
ti
o
n
p
e
rfo
rm
a
n
c
e
.
S
IG
AN
e
x
p
lo
it
s
imp
r
o
v
ise
d
i
m
a
g
e
re
p
re
se
n
tatio
n
a
n
d
lea
rn
i
n
g
o
f
d
a
ta
d
i
strib
u
ti
o
n
.
Th
e
a
lg
o
rit
h
m
we
h
a
v
e
p
ro
p
o
se
d
is
k
n
o
wn
a
s
lea
rn
in
g
-
b
a
se
d
sk
e
t
c
h
-
ima
g
e
g
e
n
e
ra
ti
o
n
(L
b
S
IG
).
T
h
is
a
l
g
o
rit
h
m
e
x
p
l
o
it
s
S
IG
AN
a
rc
h
it
e
c
tu
re
fo
r
e
fficie
n
tl
y
g
e
n
e
ra
ti
n
g
re
a
li
stic
p
h
o
to
fr
o
m
g
iv
e
n
h
a
n
d
-
d
ra
wn
sk
e
tc
h
.
S
IG
AN
is
a
ss
e
ss
e
d
u
sin
g
a
b
e
n
c
h
m
a
rk
d
a
tas
e
t
c
a
ll
e
d
CUH
K
fa
c
e
sk
e
tch
d
a
tab
a
se
(CUFS
).
F
r
o
m
th
e
e
m
p
iri
c
a
l
stu
d
y
,
it
is
o
b
se
rv
e
d
th
a
t
th
e
p
ro
p
o
se
d
S
IG
AN
a
rc
h
it
e
c
tu
re
with
u
n
d
e
rl
y
in
g
d
e
e
p
lea
rn
in
g
m
o
d
e
ls
c
o
u
ld
o
u
tp
e
rfo
r
m
e
x
isti
n
g
G
A
N m
o
d
e
ls i
n
term
s o
f
F
ré
c
h
e
t
in
c
e
p
ti
o
n
d
istan
c
e
(F
ID) wit
h
3
8
.
2
3
4
6
%
.
K
ey
w
o
r
d
s
:
Ar
tific
ial
in
tellig
en
ce
Dee
p
lear
n
in
g
Gen
er
ativ
e
ad
v
e
r
s
ar
ial
n
etwo
r
k
I
m
ag
e
g
e
n
er
atio
n
Sk
etch
-
p
h
o
t
o
s
y
n
th
esis
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
:
B
u
d
d
an
n
ag
a
r
i L
ath
a
Dep
ar
tm
en
t o
f
C
o
m
p
u
ter
Scie
n
ce
an
d
E
n
g
in
ee
r
in
g
,
K
o
n
er
u
L
ak
s
h
m
aiah
E
d
u
ca
tio
n
Fo
u
n
d
atio
n
Gr
ee
n
Field
s
,
Vad
d
eswar
am
,
Gu
n
tu
r
,
An
d
h
r
a
Pra
d
esh
,
I
n
d
i
a
E
m
ail:
lath
a.
r
ed
d
y
5
8
0
8
@
g
m
a
il.c
o
m
1.
I
NT
RO
D
UCT
I
O
N
T
h
e
f
ield
o
f
co
m
p
u
ter
v
is
io
n
an
d
p
atter
n
r
ec
o
g
n
itio
n
h
a
s
in
cr
ea
s
in
g
ly
f
o
cu
s
ed
o
n
f
a
ce
s
k
etch
s
y
n
th
esis
in
r
ec
en
t
y
ea
r
s
.
T
h
is
is
d
u
e
to
its
s
ig
n
if
ican
t
u
s
ef
u
ln
ess
in
d
ig
ital
en
ter
t
ain
m
en
t
an
d
law
en
f
o
r
ce
m
e
n
t.
W
h
en
it
co
m
es
to
cr
im
in
al
p
r
o
s
ec
u
tio
n
s
,
th
e
q
u
an
tity
an
d
q
u
ality
o
f
s
u
r
v
ei
llan
ce
ca
m
er
as
ca
n
leav
e
m
u
ch
in
f
o
r
m
atio
n
ab
o
u
t
th
e
s
u
s
p
ec
ts
u
n
r
eliab
le.
I
n
s
u
ch
ca
s
es,
ar
tis
ts
cr
ea
te
s
k
etch
es
b
ased
o
n
ey
ewitn
ess
es
m
em
o
r
ies
ar
e
ty
p
ically
u
s
ed
as
a
s
tan
d
-
in
f
o
r
th
e
d
ef
en
d
a
n
ts
'
id
en
titi
e
s
.
B
y
o
b
tain
in
g
th
e
f
ac
e
d
atab
ases
f
r
o
m
law
e
n
f
o
r
ce
m
en
t
ag
e
n
cies
o
r
co
m
b
in
in
g
th
e
s
k
etch
es
with
s
ec
u
r
ity
ca
m
e
r
a
v
id
e
o
,
th
e
p
o
lice
ca
n
u
s
e
th
e
s
k
etch
es
to
r
ed
u
ce
th
e
n
u
m
b
er
o
f
p
o
te
n
tial
s
u
s
p
ec
ts
o
n
th
eir
lis
t
[
1
]
.
Ad
d
itio
n
ally
,
th
e
f
ac
ia
l
s
k
etch
es
ar
e
u
tili
ze
d
in
th
e
cr
e
atio
n
o
f
an
im
atio
n
s
a
n
d
as
s
o
c
ial
m
ed
ia
av
atar
s
.
Ma
tch
i
n
g
f
a
ce
s
k
etch
p
ictu
r
es
to
p
h
o
to
im
ag
es
in
f
ac
e
s
k
etc
h
to
p
h
o
to
r
ec
o
g
n
itio
n
p
o
s
es
a
g
r
ea
ter
ch
allen
g
e
c
o
m
p
ar
e
d
t
o
h
o
m
o
g
e
n
o
u
s
f
ac
e
r
ec
o
g
n
itio
n
b
ec
a
u
s
e
o
f
th
e
s
ig
n
if
ican
t
m
o
d
ality
d
is
p
ar
ity
b
e
twee
n
d
ig
ital
p
h
o
to
s
an
d
f
ac
e
d
r
awin
g
s
.
Va
r
io
u
s
m
eth
o
d
o
l
o
g
ies
h
av
e
b
ee
n
em
p
lo
y
ed
to
tack
le
th
e
is
s
u
e
o
f
m
o
d
ality
d
is
p
ar
ity
in
f
ac
e
s
k
etch
r
ec
o
g
n
itio
n
,
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
S
I
GA
N
:
a
g
e
n
era
tive
a
d
ve
r
s
a
r
ia
l n
etw
o
r
k
a
r
ch
itectu
r
e
fo
r
…
(
B
u
d
d
a
n
n
a
g
a
r
i La
th
a
)
3119
in
clu
d
in
g
tech
n
iq
u
es
in
clu
d
in
g
p
ictu
r
e
t
o
s
k
etch
s
y
n
th
es
is
,
co
m
m
o
n
s
u
b
s
p
ac
e
p
r
o
jectio
n
,
a
n
d
m
o
d
ality
-
in
v
ar
ian
t
f
ea
tu
r
e
ex
tr
ac
tio
n
.
T
h
e
s
y
s
tem
u
tili
ze
s
p
h
o
to
-
s
k
etc
h
im
ag
e
p
air
s
as
tr
ain
in
g
d
ata
to
tr
an
s
f
o
r
m
f
ac
ial
r
ep
r
esen
tatio
n
s
f
r
o
m
p
h
o
to
g
r
ap
h
s
to
s
k
etch
es,
ef
f
ec
tiv
ely
m
in
im
izin
g
th
e
d
if
f
er
en
ce
s
b
etwe
en
th
ese
two
f
ac
ial
m
o
d
alities
.
C
o
n
v
en
tio
n
al
h
o
m
o
g
en
o
u
s
f
ac
e
r
ec
o
g
n
it
io
n
tech
n
iq
u
es
m
ay
b
e
u
s
ed
to
r
ec
o
g
n
ize
f
ac
es
s
k
etch
ed
af
ter
th
e
y
h
av
e
b
ee
n
cr
ea
ted
f
r
o
m
d
ig
ital p
ictu
r
es
[
2
]
.
T
h
er
e
ar
e
m
an
y
ap
p
r
o
ac
h
es
f
o
u
n
d
in
th
e
liter
atu
r
e
to
s
o
lv
e
th
e
p
r
o
b
lem
s
o
f
im
ag
e
tr
an
s
latio
n
b
ased
o
n
g
en
er
ativ
e
ad
v
er
tis
ed
n
etwo
r
k
(
GAN)
ar
ch
itectu
r
e
[
3
]
,
[
4
]
.
Yan
et
a
l.
[
5
]
in
tr
o
d
u
ce
d
th
e
id
en
tific
atio
n
-
s
en
s
itiv
e
g
en
er
ativ
e
ad
v
er
s
ar
i
al
n
etwo
r
k
(
I
SGAN)
as
a
s
o
lu
tio
n
to
th
e
p
r
o
b
lem
o
f
p
r
eser
v
in
g
id
e
n
tific
atio
n
in
f
o
r
m
atio
n
in
f
ac
e
p
h
o
to
-
s
k
etch
s
y
n
th
esis
.
Su
b
s
eq
u
en
t
r
esear
ch
en
d
ea
v
o
r
s
to
au
g
m
en
t
ef
f
icac
y
an
d
b
r
o
ad
e
n
th
e
m
eth
o
d
o
l
o
g
y
f
o
r
cr
ea
tin
g
ca
r
icatu
r
es.
B
i
et
a
l.
[
6
]
ex
am
in
ed
th
e
ap
p
licatio
n
o
f
co
n
d
itio
n
al
g
en
er
ativ
e
ad
v
er
s
ar
ial
n
etwo
r
k
s
(
cGA
N)
f
o
r
tr
a
n
s
latin
g
f
ac
es
in
to
s
k
e
tch
es.
W
an
et
a
l.
[
7
]
d
is
cu
s
s
ed
th
e
r
eten
tio
n
o
f
f
ac
ial
in
f
o
r
m
atio
n
i
n
f
ac
e
s
k
e
tch
s
y
n
th
esis
an
d
s
u
g
g
ests
a
GAN
-
b
ased
m
eth
o
d
.
W
an
et
a
l.
[
8
]
p
r
esen
ted
a
s
tr
u
ctu
r
e
f
o
r
s
im
u
ltan
e
o
u
s
f
a
ce
s
k
etch
g
en
er
atio
n
an
d
r
ec
o
g
n
itio
n
u
s
in
g
a
r
esid
u
al
d
e
n
s
e
U
-
Net
g
en
er
ato
r
,
wh
ich
is
b
ased
o
n
GANs.
A
m
u
lti
-
task
d
is
cr
im
in
ato
r
e
x
tr
ac
ts
d
is
cr
im
in
ativ
e
ch
a
r
ac
t
er
is
tics
an
d
d
ir
ec
ts
s
y
n
th
esis
.
Z
h
en
g
et
a
l.
[
9
]
th
e
p
r
o
p
o
s
ed
en
co
d
er
g
u
id
ed
g
en
er
ativ
e
ad
v
e
r
s
ar
ial
n
etwo
r
k
(
E
GGAN
)
m
o
d
el
u
tili
ze
s
a
cy
cle
-
co
n
s
is
ten
t
G
AN
ar
ch
itectu
r
e,
em
p
lo
y
i
n
g
two
g
en
er
ato
r
s
an
d
two
d
is
c
r
im
in
ato
r
s
,
f
o
r
f
ac
e
p
h
o
to
-
s
k
etc
h
s
y
n
th
esis
.
T
h
e
liter
atu
r
e
in
d
icate
s
th
at
b
ec
au
s
e
o
f
th
eir
lear
n
in
g
-
b
ased
m
eth
o
d
o
l
o
g
ies,
th
e
cu
r
r
en
t
GAN
m
o
d
els
u
s
ed
f
o
r
im
ag
e
tr
an
s
latio
n
co
u
ld
p
er
f
o
r
m
b
etter
.
Ho
wev
er
,
t
h
e
GAN
m
o
d
els
n
ee
d
to
b
e
im
p
r
o
v
e
d
f
u
r
th
er
to
war
d
s
g
en
er
atin
g
m
o
r
e
r
ea
lis
tic
im
ag
es f
r
o
m
g
iv
e
n
h
a
n
d
-
d
r
awn
s
k
etch
es
[
1
0
]
–
[
1
3
]
.
An
ex
ten
d
ed
U
-
Net,
two
d
is
cr
im
in
ato
r
s
,
an
d
an
id
en
tity
co
n
s
tr
ain
t a
r
e
th
e
co
m
p
o
n
en
ts
o
f
t
h
e
id
en
tity
m
ain
tain
ed
ad
v
er
s
ar
ial
m
o
d
el
(
I
PAM)
,
wh
ich
tack
les
th
e
p
r
o
b
lem
o
f
f
ac
e
s
k
etch
-
p
h
o
to
s
y
n
th
esis
[
1
4
]
–
[
1
8
]
.
B
etter
f
ac
ial
r
ec
o
g
n
itio
n
r
esu
lts
ar
e
s
h
o
wn
.
Haja
r
o
lasv
a
d
i
in
clu
d
e
d
a
n
ap
p
licatio
n
[
1
9
]
s
u
c
h
as
f
ac
ial
ex
p
r
ess
io
n
,
v
o
ice,
an
d
cr
o
s
s
-
m
o
d
al
s
y
n
th
esis
ar
e
co
v
er
ed
i
n
g
r
ea
t
d
etail
b
y
g
en
e
r
ativ
e
m
o
d
els,
s
p
ec
if
ically
GANs,
in
th
e
f
ield
o
f
h
u
m
an
em
o
tio
n
s
y
n
th
esis
.
Kh
an
et
a
l.
[
2
0
]
h
i
g
h
lig
h
ts
im
p
r
o
v
ed
f
in
d
in
g
s
an
d
p
r
esen
ts
a
f
u
lly
tr
ain
ed
GAN
f
o
r
te
x
t
-
to
-
f
ac
e
s
y
n
th
esis
.
T
h
e
tech
n
iq
u
e
th
at
h
as
b
ee
n
s
u
g
g
ested
in
teg
r
ates
m
an
y
d
atasets
to
co
n
d
u
ct
th
o
r
o
u
g
h
ass
ess
m
e
n
ts
an
d
y
ield
s
en
co
u
r
ag
in
g
r
e
s
u
lts
.
Den
s
er
f
ac
e
-
r
elate
d
in
f
o
r
m
atio
n
is
th
e
g
o
a
l
o
f
f
u
t
u
r
e
in
v
esti
g
atio
n
.
L
i
et
a
l.
[
2
1
]
p
r
o
v
i
d
ed
a
f
ac
e
s
k
et
ch
s
y
n
th
esis
tech
n
iq
u
e
ca
lled
r
eg
u
lar
ize
d
b
r
o
ad
lear
n
in
g
s
y
s
tem
(
R
B
L
S)
th
at
u
s
es
an
in
cr
e
m
en
tal
lear
n
i
n
g
m
eth
o
d
o
lo
g
y
to
p
r
eser
v
e
r
ic
h
f
ea
tu
r
es.
Desp
ite
s
ev
er
al
lim
its
in
co
m
p
licated
cir
cu
m
s
tan
ce
s
,
ex
p
e
r
im
en
ts
s
h
o
w
its
u
s
ef
u
ln
ess
an
d
ef
f
icien
cy
.
Su
b
s
eq
u
en
t
wo
r
k
will
f
o
cu
s
o
n
im
p
r
o
v
in
g
s
p
atial
co
r
r
esp
o
n
d
en
ce
,
h
a
n
d
lin
g
p
r
o
b
lem
s
,
a
n
d
in
v
esti
g
atin
g
n
ew
d
atasets
with
in
tr
icate
s
ce
n
er
ies
an
d
d
ep
th
.
Z
h
an
g
[
2
2
]
ad
d
r
ess
ed
th
e
s
h
o
r
tco
m
in
g
s
o
f
p
r
e
v
io
u
s
s
y
s
tem
s
b
y
in
tr
o
d
u
cin
g
a
ca
s
ca
d
ed
f
ac
e
s
k
etch
g
en
er
ati
o
n
m
eth
o
d
th
at
is
r
esis
tan
t
to
d
if
f
er
en
t
lig
h
tin
g
co
n
d
itio
n
s
.
T
h
e
o
u
tco
m
es o
f
th
e
ex
p
er
im
en
ts
i
n
d
icate
a
n
o
ta
b
le
en
h
a
n
ce
m
en
t a
n
d
p
o
s
s
ib
ilit
y
f
o
r
u
s
ef
u
l
o
p
t
ical
s
y
s
tem
s
.
T
h
e
ap
p
r
o
ac
h
e
n
s
u
r
es
ae
s
th
etica
lly
p
leasin
g
v
is
u
als
b
y
co
m
b
in
in
g
tex
t
u
al
an
d
v
is
u
al
asp
ec
ts
.
KO
et
al
s
u
g
g
ested
Su
p
er
s
tar
GAN,
an
en
h
an
ce
d
Star
GAN
v
ar
ian
t
f
o
r
ex
p
a
n
s
iv
e
d
o
m
ai
n
s
th
at
u
s
es
C
o
n
tr
o
lGAN
to
o
v
er
co
m
e
d
r
aw
b
ac
k
s
.
I
m
p
r
o
v
ed
p
er
f
o
r
m
a
n
ce
o
n
a
v
ar
iety
o
f
d
atasets
with
r
ed
u
ce
d
F
r
éc
h
e
t
in
ce
p
tio
n
d
is
tan
ce
(
FID
)
an
d
lear
n
ed
p
e
r
ce
p
tu
al
im
ag
e
p
atch
s
im
ilar
ity
(
L
PIPS)
.
Fro
m
th
e
liter
atu
r
e,
it
is
o
b
s
er
v
ed
th
at
th
e
ex
is
tin
g
GAN
m
o
d
els
u
s
ed
f
o
r
im
ag
e
tr
an
s
latio
n
co
u
ld
i
m
p
r
o
v
e
p
e
r
f
o
r
m
an
ce
d
u
e
to
th
eir
lear
n
in
g
-
b
ased
ap
p
r
o
ac
h
es.
Ho
wev
er
,
th
e
G
AN
m
o
d
els
n
ee
d
to
b
e
im
p
r
o
v
ed
f
u
r
th
er
t
o
war
d
s
g
en
er
at
in
g
m
o
r
e
r
ea
lis
tic
im
ag
es f
r
o
m
g
iv
en
h
an
d
-
d
r
aw
n
s
k
etch
es
[
2
3
]
–
[
2
9
]
.
2.
M
E
T
H
O
D
T
h
is
s
ec
tio
n
p
r
esen
ts
th
e
p
r
o
p
o
s
ed
m
eth
o
d
o
lo
g
y
wh
ich
in
clu
d
es
o
u
r
SIG
AN
ar
ch
itectu
r
e,
d
etails
ab
o
u
t
g
en
er
at
o
r
a
n
d
d
is
cr
im
in
ato
r
,
p
r
o
p
o
s
ed
alg
o
r
ith
m
an
d
ev
alu
atio
n
m
eth
o
d
o
lo
g
y
.
A
g
e
n
er
ativ
e
a
d
v
er
s
ar
ial
n
etwo
r
k
is
m
ad
e
u
p
o
f
two
p
a
r
ts
,
th
e
g
en
er
ato
r
an
d
th
e
d
is
cr
im
in
ato
r
,
th
at
ar
e
tr
ai
n
ed
to
g
et
h
er
v
ia
ad
v
er
s
ar
ial
tr
ain
in
g
.
T
h
e
d
is
cr
im
in
ato
r
is
u
s
ed
to
im
p
r
o
v
e
its
ab
ilit
y
t
o
d
is
ce
r
n
b
etwe
en
ac
tu
al
an
d
f
ak
e
d
ata,
an
d
th
e
g
en
er
ato
r
tak
es r
an
d
o
m
n
o
is
e
as in
p
u
t a
n
d
g
en
er
ates sy
n
th
eti
c
d
ata
clo
s
e
to
th
e
r
ea
l
d
ata.
2
.
1
.
P
r
o
blem
d
ef
ini
t
io
n
Pro
v
id
ed
a
h
an
d
-
d
r
awn
s
k
etch
,
d
ev
elo
p
in
g
a
n
o
v
el
GAN
ar
c
h
itectu
r
e
wh
ich
ex
p
lo
its
DL
m
o
d
els
f
o
r
b
u
ild
in
g
g
en
e
r
ato
r
a
n
d
d
is
cr
im
in
ato
r
with
a
n
o
n
-
co
o
p
er
ativ
e
g
am
e
b
etwe
en
th
e
m
f
o
r
a
u
to
m
atic
g
en
er
atio
n
o
f
r
ea
lis
tic
p
h
o
to
f
r
o
m
th
e
g
iv
e
n
s
k
etch
is
th
e
ch
allen
g
in
g
p
r
o
b
lem
co
n
s
id
er
ed
.
Gen
er
ativ
e
ad
v
er
s
ar
ial
n
etwo
r
k
s
(
GANs)
h
av
e
em
er
g
ed
as
a
p
o
wer
f
u
l
to
o
l
r
a
p
id
ly
ad
v
an
cin
g
th
e
s
tate
-
of
-
th
e
-
a
r
t
in
n
u
m
e
r
o
u
s
d
o
m
ain
s
.
T
h
is
p
ap
er
c
o
n
d
u
cts
a
c
o
m
p
r
e
h
en
s
iv
e
r
ev
iew
to
a
n
aly
s
es
th
e
ap
p
licatio
n
s
o
f
GANs
in
th
e
co
n
s
tr
u
ctio
n
in
d
u
s
tr
y
o
v
er
th
e
y
ea
r
s
,
a
n
d
th
e
r
ev
ie
w
aim
s
to
en
r
ich
t
h
e
b
o
d
y
o
f
k
n
o
wled
g
e
o
n
th
is
em
er
g
i
n
g
d
ee
p
lear
n
in
g
(
DL
)
alg
o
r
ith
m
in
t
h
e
co
n
s
tr
u
cti
o
n
s
ec
to
r
.
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
.
3
,
J
u
n
e
20
25
:
3
1
1
8
-
3
1
2
6
3120
2
.
2
.
P
r
o
po
s
ed
G
AN
a
rc
hite
ct
ure
R
ec
en
t
r
esear
ch
in
s
em
an
tic
in
p
ain
tin
g
r
eg
ar
d
s
,
in
p
ain
tin
g
as
a
task
o
f
lim
ited
p
ictu
r
e
g
en
er
atio
n
is
s
u
e
[
3
0
]
it
is
co
n
s
id
er
ed
e
s
s
en
tial
th
at
th
e
g
e
n
er
ated
c
o
n
ten
t
m
ai
n
tain
s
s
em
an
tic
c
o
h
er
en
ce
with
in
th
e
o
b
s
er
v
ed
co
n
tex
t
an
d
s
ea
m
less
ly
in
teg
r
ates
with
th
e
s
u
r
r
o
u
n
d
in
g
p
i
x
els.
Similar
ly
,
we
f
r
am
e
th
e
p
r
o
b
lem
o
f
p
ictu
r
e
p
r
o
d
u
ctio
n
as
o
n
e
o
f
im
ag
e
co
m
p
leten
ess
,
with
s
k
etch
ac
tin
g
as
a
wea
k
co
n
te
x
tu
al
co
n
s
tr
ain
t.
T
h
e
GAN
ar
ch
itectu
r
e
s
u
g
g
ested
in
[
3
1
]
s
er
v
es
as
th
e
f
o
u
n
d
atio
n
f
o
r
o
u
r
d
ee
p
m
o
d
el,
with
th
e
f
o
llo
win
g
tech
n
o
lo
g
ical
m
o
d
if
icatio
n
s
.
2
.
2
.
1
.
I
m
a
g
e
re
presenta
t
io
n
W
e
s
u
g
g
est
m
o
d
elin
g
s
k
etch
an
d
p
ictu
r
e
in
a
co
m
b
in
e
d
in
p
u
t
s
p
ac
e,
as
o
p
p
o
s
ed
to
th
e
co
n
v
en
tio
n
al
m
eth
o
d
s
o
f
s
ep
ar
atin
g
th
em
.
Sp
ec
if
ically
,
we
s
p
atially
co
m
b
in
e
g
en
u
i
n
e
p
h
o
to
g
r
ap
h
s
(
B
)
with
th
ei
r
co
r
r
esp
o
n
d
in
g
s
k
etch
s
ty
les
(
A)
to
f
o
r
m
jo
in
t
s
k
etch
-
im
a
g
e
p
air
s
(
AB
)
b
ased
o
n
a
c
o
r
p
u
s
o
f
s
am
p
les.
T
h
is
jo
in
t
im
ag
e
in
h
er
en
tly
ca
p
tu
r
es
th
e
co
n
tex
tu
al
r
elatio
n
s
h
ip
b
etwe
en
th
e
s
k
etch
an
d
th
e
p
h
o
to
co
m
p
o
n
e
n
ts
,
aid
in
g
in
u
n
d
e
r
s
tan
d
in
g
o
f
th
e
ir
co
m
b
in
e
d
d
is
tr
ib
u
tio
n
th
r
o
u
g
h
GAN.
W
e
co
m
m
en
ce
t
h
e
tr
ain
in
g
p
r
o
ce
s
s
o
f
a
GAN
m
o
d
el
u
s
in
g
th
ese
co
m
b
in
ed
im
ag
es,
allo
win
g
it
to
lev
er
ag
e
th
e
c
o
n
tex
tu
al
in
f
o
r
m
ati
o
n
p
r
o
v
id
e
d
b
y
th
e
s
k
etch
co
m
p
o
n
en
t
to
au
to
m
at
ically
p
r
ed
ict
an
d
r
ec
o
n
s
tr
u
ct
th
e
m
is
s
in
g
p
ar
ts
o
f
th
e
im
ag
e.
I
n
co
n
tr
ast
to
ea
r
lier
wo
r
k
[
3
2
]
)
wh
er
e
z
was
s
o
lely
an
im
ag
e
em
b
ed
d
in
g
,
th
e
g
en
er
at
o
r
p
r
o
d
u
ce
s
a
co
n
s
o
lid
ated
r
ep
r
esen
tatio
n
b
y
m
ap
p
i
n
g
th
e
m
er
g
ed
s
k
etch
an
d
im
ag
e
i
n
to
a
n
o
n
-
lin
ea
r
jo
in
t
s
p
ac
e
c
alled
z.
I
n
s
tead
o
f
d
ir
ec
tly
lim
itin
g
th
e
g
en
er
ated
im
ag
e
with
th
e
co
m
p
lete
z,
w
e
ca
n
in
d
ir
ec
tly
im
p
o
s
e
r
estrictio
n
s
o
n
it
b
y
o
n
ly
u
s
in
g
th
e
s
k
etch
o
f
th
e
jo
in
t
e
m
b
ed
d
in
g
z
o
f
th
e
i
n
p
u
t.
T
h
is
allo
ws
u
s
to
m
ain
tain
f
aith
f
u
ln
ess
wh
ile
allo
win
g
f
o
r
a
ce
r
tain
lev
el
o
f
f
lex
ib
ilit
y
in
th
e
v
is
u
al
p
r
esen
tatio
n
o
f
th
e
r
esu
ltin
g
im
ag
e
.
2
.
2
.
2
.
O
bje
ct
iv
e
f
un
ct
io
n
Ou
r
p
u
r
p
o
s
e
is
to
d
is
co
v
er
a
g
en
er
ated
jo
in
t
im
ag
e,
(
^
)
,
th
at
clo
s
ely
r
esem
b
les
th
e
in
p
u
t
s
k
e
tch
,
to
ac
co
m
p
lis
h
th
e
m
o
s
t
ac
cu
r
ate
m
ap
p
in
g
b
etwe
en
th
e
d
is
to
r
ted
an
d
r
ec
o
v
er
ed
j
o
in
t
im
a
g
es.
Ou
r
g
o
al
is
to
co
n
s
tr
u
ct
th
e
lo
s
s
f
u
n
ctio
n
to
i
n
co
r
p
o
r
ate
two
lo
s
s
es,
u
tili
zin
g
th
e
r
an
d
o
m
ly
s
elec
ted
in
p
u
t
∼
_
.
T
o
ass
ess
th
e
co
n
tex
t
u
al
r
esem
b
lan
ce
b
etwe
en
th
e
u
n
af
f
ec
te
d
s
ec
tio
n
s
th
at
is
th
e
in
p
u
t
s
k
etc
h
alo
n
g
with
th
e
r
ec
o
n
s
tr
u
cted
d
r
awin
g
we
u
tili
ze
a
co
n
tex
tu
al
l
o
s
s
[
2
0
]
,
w
h
ich
is
“
ex
p
r
ess
ed
as in
(
1
)
.
_
(
)
=
_
(
⨀
,
⨀
(
)
)
(
1
)
wh
er
e
M
r
ep
r
esen
ts
th
e
Had
am
ar
d
p
r
o
d
u
ctio
n
an
d
is
th
e
b
i
n
ar
y
m
ask
o
f
th
e
d
am
ag
ed
j
o
in
t
p
ictu
r
e.
Un
lik
e
[
2
0
]
,
we
em
p
lo
y
th
e
KL
-
d
iv
er
g
en
ce
to
m
ea
s
u
r
e
th
e
s
im
ilar
it
y
b
etwe
en
th
e
d
is
tr
ib
u
tio
n
o
f
two
d
r
awin
g
s
.
T
h
is
ap
p
r
o
ac
h
en
h
an
ce
s
th
e
alig
n
m
en
t o
f
s
k
etch
es,
tak
in
g
in
to
ac
co
u
n
t th
at
a
s
k
etch
is
a
b
in
ar
y
im
ag
e”
in
s
tead
o
f
a
n
atu
r
al
im
ag
e
.
I
n
an
id
ea
l
s
c
en
ar
io
,
ea
c
h
p
i
x
el
in
th
e
s
k
et
ch
ar
ea
s
o
f
b
o
th
y
an
d
(
)
wo
u
ld
h
av
e
b
ee
n
in
d
is
tin
g
u
is
h
ab
le,
r
esu
ltin
g
in
_
(
)
=
0
.
C
o
n
s
eq
u
e
n
tly
,
we
im
p
o
s
e
a
p
en
alty
o
n
(
)
f
o
r
its
f
ailu
r
e
to
g
e
n
er
ate
a
d
r
awin
g
t
h
at
m
o
s
t
clo
s
ely
m
atch
es
th
e
o
b
s
er
v
ed
in
p
u
t
s
k
etch
y
.
W
e
u
s
ed
th
e
ad
v
e
r
s
ar
ial
lo
s
s
o
f
th
e
G
n
etwo
r
k
;
th
e
p
er
ce
p
tu
al
lo
s
s
p
r
eser
v
es
th
e
s
em
an
tic
in
f
o
r
m
atio
n
o
f
th
e
a
n
ticip
ated
“im
ag
e
as
in
(
2
)
.
_
(
)
=
(
1
−
(
(
)
)
)
(
2
)
T
h
e
d
esire
d
f
u
n
ctio
n
f
o
r
z^
i
s
f
o
r
m
u
lated
b
y
c
o
m
b
in
i
n
g
t
h
e
two
lo
s
s
es
th
r
o
u
g
h
a
weig
h
ted
s
u
m
m
atio
n
,
as
elu
cid
ated
in
(
3
)
.
^
=
(
)
┬
(
_
(
)
+
_
(
)
)
(
3
)
wh
er
e
λ
is
a
h
y
p
er
p
ar
am
eter
”
th
at
u
s
es
th
e
in
p
u
t
to
r
estrict
t
h
e
o
u
tp
u
t
p
ictu
r
e
.
A
m
o
d
est
λ
will
en
s
u
r
e
th
at
th
e
in
p
u
t a
n
d
o
u
t
p
u
t lo
o
k
th
e
s
am
e.
2
.
2
.
3
.
O
ur
s
k
et
ch
-
ima
g
e
G
A
N
(
SI
G
AN)
T
h
e
t
r
a
i
n
i
n
g
a
n
d
c
o
m
p
l
e
ti
o
n
s
ta
g
e
s
m
a
k
e
u
p
o
u
r
GA
N
.
E
x
ce
p
t
f
o
r
u
s
i
n
g
j
o
i
n
t
p
i
c
t
u
r
es
f
o
r
o
u
r
t
r
a
i
n
i
n
g
s
a
m
p
l
es
,
t
h
e
p
h
as
e
o
f
a
t
r
a
i
n
i
n
g
i
s
i
d
e
n
ti
c
a
l
t
o
c
l
ass
i
c
t
r
a
i
n
i
n
g
o
f
a
G
A
N
.
A
f
t
e
r
c
o
m
p
le
t
i
n
g
th
e
t
r
a
i
n
i
n
g
p
r
o
c
e
s
s
,
w
e
s
e
le
c
t
a
“
g
e
n
e
r
at
i
v
e
n
e
t
wo
r
k
G
t
h
a
t
e
f
f
e
c
t
i
v
e
l
y
r
e
p
r
o
d
u
c
e
s
t
h
e
c
o
m
b
i
n
e
d
d
i
s
t
r
i
b
u
t
i
o
n
o
f
i
m
a
g
e
d
a
t
a
b
y
c
o
n
v
e
r
t
i
n
g
s
a
m
p
l
es
f
r
o
m
t
h
e
n
o
i
s
e
d
i
s
t
r
i
b
u
t
i
o
n
_
i
n
t
o
t
h
e
d
a
ta
d
i
s
t
r
i
b
u
ti
o
n
_
.
T
o
a
c
h
i
e
v
e
o
u
r
d
e
s
i
r
e
d
o
u
t
c
o
m
e
s
,
w
e
m
u
s
t
p
r
o
v
i
d
e
e
it
h
e
r
t
h
e
i
n
p
u
t
i
m
a
g
e
o
f
t
h
e
d
am
a
g
e
d
j
o
i
n
t
o
r
t
h
e
m
a
s
k
e
d
-
out
i
m
a
g
e
c
o
m
p
o
n
e
n
t
.
T
h
i
s
r
e
p
r
es
e
n
t
at
i
o
n
wi
l
l
e
n
a
b
l
e
u
s
t
o
c
h
o
o
s
e
t
h
e
i
m
a
g
e
o
n
t
h
e
m
a
n
i
f
o
l
d
o
f
G
t
h
at
is
c
l
o
s
es
t
in
t
h
e
l
at
e
n
t
”
s
p
a
c
e
.
W
e
d
e
te
r
m
i
n
e
t
h
e
^
v
e
c
t
o
r
i
n
(
3
)
t
h
a
t
m
i
n
i
m
i
z
es
o
u
r
o
b
j
e
c
t
i
v
e
f
u
n
c
t
i
o
n
r
a
t
h
e
r
t
h
a
n
m
a
x
i
m
i
zi
n
g
(
)
.
T
h
i
s
s
h
o
w
s
t
h
a
t
t
h
e
d
i
s
t
o
r
t
e
d
i
n
p
u
t
is
b
e
i
n
g
p
r
o
je
c
t
e
d
o
n
t
o
t
h
e
g
e
n
e
r
a
t
o
r
'
s
z
s
p
a
c
e
b
y
r
e
p
e
t
it
i
v
e
b
a
c
k
p
r
o
p
a
g
a
t
i
o
n
.
I
n
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
S
I
GA
N
:
a
g
e
n
era
tive
a
d
ve
r
s
a
r
ia
l n
etw
o
r
k
a
r
ch
itectu
r
e
fo
r
…
(
B
u
d
d
a
n
n
a
g
a
r
i La
th
a
)
3121
p
a
r
t
i
c
u
l
a
r
,
t
h
e
i
n
p
u
t
c
o
m
p
r
is
e
s
a
c
o
m
b
i
n
e
d
i
m
a
g
e
f
e
a
t
u
r
i
n
g
s
o
l
e
l
y
t
h
e
le
f
t
-
s
i
d
e
d
r
a
w
i
n
g
,
w
h
i
l
e
t
h
e
r
i
g
h
t
-
s
i
d
e
i
m
a
g
e
r
e
m
a
i
n
s
c
o
n
c
e
a
l
e
d
,
a
n
d
“
a
v
e
c
t
o
r
z
s
ta
r
t
e
d
wi
t
h
e
v
e
n
l
y
d
i
s
t
r
i
b
u
te
d
r
a
n
d
o
m
n
o
i
s
e
.
T
o
m
o
d
if
y
th
e
r
an
d
o
m
ly
c
h
o
s
en
in
p
u
t
z
o
f
n
etwo
r
k
G,
we
ap
p
ly
t
h
e
lo
s
s
f
u
n
ctio
n
d
escr
i
b
ed
in
(
3
)
.
At
p
r
esen
t,
ju
s
t
th
e
in
p
u
t
v
ec
to
r
z
is
m
o
d
if
ied
b
y
u
s
in
g
g
r
ad
ien
t
d
escen
t,
wh
er
ea
s
th
e
r
elativ
e
weig
h
ts
o
f
n
etwo
r
k
s
G
an
d
D
s
tay
u
n
c
h
a
n
g
ed
.
Fig
u
r
e
1
illu
s
tr
ates
th
e
p
r
o
ce
s
s
o
f
tr
a
v
er
s
in
g
th
e
late
n
t
s
p
ac
e
d
u
r
in
g
b
ac
k
-
p
r
o
p
a
g
atio
n
,
s
h
o
wca
s
in
g
f
o
u
r
iter
atio
n
s
”
.
Kee
p
in
m
in
d
th
at
[
3
3
]
u
s
es
an
an
alo
g
o
u
s
o
p
tim
iz
atio
n
tech
n
iq
u
e
o
f
g
r
ad
ien
t
d
escen
t
f
o
r
in
v
e
r
s
e
m
ap
p
in
g
.
“
Fo
llo
win
g
th
e
b
ac
k
-
p
r
o
p
a
g
atio
n
p
r
o
ce
s
s
,
th
e
co
r
r
u
p
ted
i
n
p
u
t
(
y
)
is
m
ap
p
ed
to
th
e
cl
o
s
est
v
ec
to
r
^
in
th
e
laten
t
s
p
ac
e.
T
h
is
v
ec
to
r
is
th
en
f
ed
in
to
th
e
G
n
etwo
r
k
to
g
e
n
er
at
e
(
)
.
T
h
e
im
ag
e
is
cr
ea
ted
b
y
u
tili
zin
g
(
^
)
to
co
m
p
lete
th
e
ab
s
en
t
v
alu
es
o
f
y
,
wh
ich
r
ep
r
esen
t
s
th
e
im
ag
e
s
ec
tio
n
:
_
=
⨀
+
(
1
−
)
⨀
(
^
)
(
4
)
W
e
u
tili
ze
a
n
o
is
e
v
ec
to
r
”
th
at
is
u
n
if
o
r
m
ly
s
am
p
led
as
i
n
p
u
t.
T
h
e
in
f
lu
en
ce
o
f
th
e
i
n
itializatio
n
o
n
t
h
e
r
esu
ltan
t
im
ag
e
is
a
clea
r
co
n
ce
r
n
.
I
f
th
e
in
itial
s
k
etch
s
ec
tio
n
o
f
(
)
p
er
ce
p
tu
ally
d
ev
iate
s
s
ig
n
if
ican
tly
f
r
o
m
th
e
in
p
u
t
s
k
etch
,
g
r
ad
ien
t
d
escen
t
will
f
ac
e
ch
allen
g
es
in
m
ap
p
in
g
th
e
d
a
m
ag
ed
p
ict
u
r
e
to
th
e
clo
s
est
z
in
th
e
laten
t
s
p
ac
e.
Fai
lu
r
e
s
am
p
les
will
ar
i
s
e
f
r
o
m
th
is
,
ev
en
if
we
estab
lis
h
a
m
in
u
s
cu
le
in
(
3
)
.
W
e
im
p
r
o
v
e
th
e
in
itializatio
n
as
f
o
llo
ws
to
s
o
lv
e
th
is
is
s
u
e:
W
e
u
s
e
a
f
o
r
war
d
p
ass
to
s
am
p
le
N
n
o
is
e
v
ec
to
r
s
u
n
if
o
r
m
ly
at
r
an
d
o
m
an
d
ex
tr
ac
t
th
e
in
itialized
d
r
awin
g
s
f
o
r
ea
ch
o
n
e
.
Nex
t,
we
d
eter
m
in
e
th
e
p
air
wis
e
KL
-
d
iv
er
g
en
ce
b
etwe
en
th
ese
N
in
itialized
d
r
awin
g
s
an
d
th
e
in
p
u
t
s
k
etch
.
Ou
t
o
f
all
th
e
N
s
am
p
les,
th
e
in
itial
s
k
etch
will
b
e
th
e
o
n
e
with
th
e
least
“
KL
-
d
iv
er
g
en
ce
,
s
ig
n
if
y
in
g
th
e
b
est
in
itializatio
n
.
I
n
Fig
u
r
e
1
,
th
e
en
tire
n
etwo
r
k
is
v
is
ib
le.
Gen
er
ato
r
G
r
ec
eiv
es
a
1
0
0
-
D
r
an
d
o
m
n
o
is
e
v
ec
to
r
wh
ich
is
u
n
if
o
r
m
l
y
s
am
p
led
f
r
o
m
−1
to
1
.
T
h
i
s
is
d
o
n
e
in
c
o
m
p
lia
n
ce
with
[
8
]
.
Af
ter
war
d
,
th
e
in
p
u
t
is
r
esh
a
p
ed
to
4
×8
×
5
1
2
u
s
in
g
an
8
1
9
2
×
2
lin
ea
r
lay
er
.
Fig
u
r
e
1
.
Ar
c
h
itectu
r
al
o
v
e
r
v
i
ew
o
f
SIG
AN
W
e
em
p
lo
y
f
i
v
e
u
p
-
co
n
v
o
lu
ti
o
n
al
lay
er
s
with
a
s
tr
id
e
o
f
t
wo
an
d
a
k
er
n
el
s
ize
o
f
f
iv
e.
T
o
s
p
ee
d
u
p
tr
ain
in
g
an
d
s
tab
ilize
lear
n
in
g
,
af
ter
ev
er
y
u
p
-
c
o
n
v
o
lu
tio
n
al
lay
er
,
ex
ce
p
t
f
o
r
th
e
f
in
al
o
n
e,
“
we
ad
d
a
b
atch
n
o
r
m
aliza
tio
n
lay
er
.
Ad
d
itio
n
ally
,
all
lay
e
r
s
u
tili
ze
th
e
leak
y
r
ec
tifie
d
lin
ea
r
u
n
it
(
L
R
eL
U)
ac
tiv
atio
n
.
T
an
h
is
ap
p
lied
at
th
e
o
u
tp
u
t
lay
e
r
a
t
th
e
en
d
.
A
h
i
g
h
er
r
eso
lu
tio
n
p
ictu
r
e
m
ea
s
u
r
i
n
g
6
4
×
1
2
8
is
p
r
o
d
u
ce
d
v
ia
a
n
o
n
lin
ea
r
weig
h
te
d
u
p
s
am
p
lin
g
o
f
th
e
laten
t
s
p
ac
e
”
th
r
o
u
g
h
a
s
eq
u
e
n
ce
o
f
u
p
-
co
n
v
o
lu
tio
n
s
an
d
n
o
n
-
lin
ea
r
ities
.
A
p
ictu
r
e
with
d
im
en
s
io
n
s
o
f
6
4
b
y
1
2
8
b
y
3
is
u
s
ed
as
th
e
d
is
cr
im
in
ato
r
'
s
in
p
u
t.
I
t
is
f
o
llo
wed
b
y
f
o
u
r
co
n
v
o
l
u
tio
n
al
lay
er
s
,
ea
c
h
o
f
wh
ich
h
as twice
as m
an
y
ch
an
n
els as th
e
lay
er
b
ef
o
r
e
it
an
d
h
alf
th
e
f
ea
tu
r
e
m
ap
'
s
d
im
en
s
io
n
.
T
o
g
en
e
r
ate
a
4
×
8
×5
1
2
o
u
tp
u
t,
we
s
p
ec
if
i
ca
lly
ad
d
4
c
o
n
v
o
lu
tio
n
al
la
y
e
r
s
with
k
er
n
el
s
ize
5
an
d
s
tr
id
e
2
.
Af
ter
r
esh
a
p
in
g
th
e
o
u
tp
u
t
to
o
n
e
d
im
e
n
s
io
n
with
a
f
u
lly
co
n
n
ec
te
d
lay
e
r
,
we
co
m
p
u
te
lo
s
s
u
s
in
g
a
So
f
tMa
x
lay
e
r
.
Gen
u
in
e
f
r
ee
h
an
d
d
r
awin
g
s
co
m
e
in
a
wid
e
r
an
g
e
o
f
s
ty
les
an
d
ca
n
d
if
f
er
s
ig
n
if
ica
n
tly
f
r
o
m
s
y
n
th
esis
d
r
awin
g
s
th
at
ar
e
m
ec
h
an
ically
cr
ea
te
d
f
r
o
m
p
ict
u
r
es.
W
e
en
h
an
ce
o
u
r
tr
ain
i
n
g
d
ata
b
y
em
p
lo
y
in
g
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
.
3
,
J
u
n
e
20
25
:
3
1
1
8
-
3
1
2
6
3122
s
ev
er
al
k
in
d
s
o
f
d
r
awin
g
s
as
t
h
e
tr
ain
in
g
s
et
to
p
r
e
v
en
t
o
v
er
f
itti
n
g
to
a
ce
r
tain
s
ty
le
o
f
s
k
e
tch
im
ag
e
p
air
in
g
s
an
d
to
in
cr
ea
s
e
th
e
g
en
er
ality
o
f
th
e
n
etwo
r
k
.
T
o
m
ak
e
d
iv
er
s
e
ty
p
es
o
f
d
r
awin
g
s
,
we
s
p
ec
i
f
ically
em
p
lo
y
th
e
FDo
G
f
ilter
p
r
o
v
id
ed
in
[
1
0
]
,
th
e
Ph
o
to
co
p
y
ef
f
ec
t
[
1
]
in
P
h
o
to
s
h
o
p
,
an
d
th
e
XDo
G
ed
g
e
d
etec
to
r
p
r
o
p
o
s
ed
in
[
1
1
]
.
W
e
also
u
s
e
[
1
2
]
to
s
im
p
lify
th
e
e
d
g
e
p
ictu
r
es
s
o
th
at
th
e
y
m
o
r
e
clo
s
ely
r
esem
b
le
h
an
d
-
d
r
awn
d
r
awin
g
s
.
T
o
tr
ain
d
is
tin
ct
s
ty
le
m
o
d
els,
we
d
iv
id
e
th
e
d
ata
i
n
ea
ch
s
ty
le
i
n
to
tr
ain
i
n
g
a
n
d
t
esti
n
g
s
ets.
Firstl
y
,
we
r
etr
iev
e
th
e
p
r
e
-
t
r
ain
ed
X
Do
G
s
ty
le
m
o
d
el,
r
ath
er
th
an
tr
ain
in
g
all
s
ty
le
m
o
d
els
f
r
o
m
s
tar
t.
T
h
e
n
etwo
r
k
s
ar
e
th
en
r
ef
in
e
d
u
s
in
g
d
r
awin
g
s
in
d
if
f
er
en
t sty
les,
s
u
ch
as FDo
G,
th
e
s
im
p
lific
atio
n
,
an
d
t
h
e
co
p
ier
s
ty
le.
T
h
e
r
atio
n
ale
is
th
at
XDo
G,
in
o
u
r
o
p
in
io
n
,
is
m
o
r
e
d
etailed
an
d
m
o
r
e
ak
in
to
th
e
o
r
i
g
in
al
p
h
o
to
g
r
a
p
h
ic
im
ag
e.
T
h
is
en
s
u
r
es
th
at
th
e
n
etwo
r
k
is
tr
ain
ed
o
n
h
ig
h
-
q
u
ality
lo
c
al
m
in
im
a
b
ef
o
r
e
in
co
r
p
o
r
atin
g
ad
d
itio
n
al
s
k
etch
s
ty
les.
Du
r
in
g
th
e
e
x
p
er
im
e
n
t,
we
d
em
o
n
s
tr
ate
h
o
w
th
e
au
g
m
en
tin
g
s
ty
les
allo
w
f
o
r
s
o
m
e
d
eg
r
ee
o
f
ap
p
ea
r
an
ce
f
lex
ib
ilit
y
wh
ile
al
s
o
im
p
r
o
v
in
g
th
e
g
en
er
aliza
tio
n
o
f
th
e
s
k
etch
-
im
ag
e
r
elatio
n
s
h
ip
.
2
.
3
.
I
m
ple
m
ent
a
t
io
n
d
et
a
ils
U
s
i
n
g
SG
A
N
,
w
e
p
r
e
t
r
a
i
n
t
h
e
n
e
t
w
o
r
k
f
o
r
e
v
e
r
y
c
a
t
e
g
o
r
y
.
B
o
t
h
t
h
e
g
e
n
e
r
a
t
o
r
a
n
d
d
i
s
c
r
i
m
i
n
a
t
o
r
n
e
t
w
o
r
k
s
e
m
p
l
o
y
t
h
e
A
d
a
m
o
p
t
i
m
i
z
e
r
[
1
3
]
w
it
h
a
b
e
t
a
v
a
l
u
e
o
f
0
.
5
a
n
d
a
l
e
a
r
n
i
n
g
r
a
t
e
o
f
0
.
0
0
0
2
.
T
h
e
t
r
a
i
n
i
n
g
d
u
r
a
t
i
o
n
v
a
r
i
e
s
b
as
e
d
o
n
t
h
e
d
a
t
a
s
et'
s
m
a
g
n
it
u
d
e
,
s
p
a
n
n
i
n
g
f
r
o
m
6
t
o
4
8
h
o
u
r
s
.
T
h
i
s
i
s
a
cc
o
m
p
l
i
s
h
e
d
b
y
u
t
i
li
z
i
n
g
a
b
a
t
c
h
s
i
z
e
o
f
6
4
a
n
d
r
u
n
n
i
n
g
t
h
e
t
r
a
i
n
i
n
g
p
r
o
c
e
s
s
f
o
r
2
0
0
e
p
o
c
h
s
.
O
n
c
e
w
e
h
a
v
e
a
c
q
u
i
r
e
d
a
h
i
g
h
l
y
s
k
i
l
l
e
d
v
e
r
s
i
o
n
o
f
t
h
e
X
D
o
G
s
t
y
l
e
,
we
u
t
i
l
iz
e
“
t
h
e
s
a
m
e
n
e
t
w
o
r
k
s
tr
u
c
t
u
r
e
t
o
t
r
a
i
n
o
t
h
e
r
d
r
a
w
i
n
g
s
t
y
l
es
s
e
q
u
e
n
t
i
al
l
y
.
T
h
i
s
i
s
a
c
h
i
e
v
e
d
b
y
e
m
p
l
o
y
i
n
g
a
l
o
w
e
r
le
a
r
n
i
n
g
r
a
t
e
(
e
.
g
.
,
1
e^
(
-
5
)
)
t
o
p
r
o
d
u
c
e
m
o
d
e
ls
f
o
r
t
h
e
s
e
d
i
f
f
e
r
e
n
t
s
t
y
l
es
.
T
h
r
o
u
g
h
o
u
t
t
h
e
c
o
m
p
l
e
t
i
o
n
p
r
o
c
e
s
s
,
t
h
e
i
n
p
u
t
z
is
m
o
d
i
f
i
e
d
th
r
o
u
g
h
t
h
e
i
n
c
o
r
p
o
r
a
t
i
o
n
o
f
a
co
n
t
e
x
t
u
a
l
l
o
s
s
a
n
d
a
p
e
r
c
e
p
t
u
a
l
l
o
s
s
.
T
h
e
c
o
n
t
e
x
t
u
a
l
l
o
s
s
a
n
d
p
e
r
c
e
p
t
u
a
l
l
o
s
s
a
r
e
ass
i
g
n
e
d
a
λ
v
a
l
u
e
o
f
0
.
0
1
a
n
d
a
m
o
m
e
n
t
u
m
v
a
l
u
e
o
f
0
.
9
,
r
e
s
p
e
c
t
i
v
e
l
y
”
.
B
a
c
k
-
p
r
o
p
a
g
a
t
i
o
n
u
t
i
l
i
ze
s
s
t
o
c
h
as
t
i
c
cl
ip
p
i
n
g
.
W
e
o
p
t
i
m
i
z
e
a
t
te
s
t
-
t
im
e
t
o
m
a
x
i
m
i
z
e
t
h
e
s
k
e
t
c
h
p
a
r
t
i
n
t
h
e
p
r
o
d
u
c
e
d
p
i
ct
u
r
e
t
h
a
t m
o
s
t c
l
o
s
e
l
y
r
e
s
e
m
b
les
t
h
e
o
r
i
g
i
n
a
l d
r
a
w
i
n
g
b
y
s
et
t
i
n
g
a
r
e
l
at
i
v
e
l
y
s
m
al
l
λ
,
w
h
i
c
h
m
a
k
e
s
c
o
n
t
e
x
t
u
a
l
l
o
s
s
m
o
r
e
s
i
g
n
i
f
i
c
a
n
t
’
’
.
D
u
r
i
n
g
b
a
c
k
-
p
r
o
p
a
g
a
t
i
o
n
,
t
h
e
d
i
s
c
r
i
m
i
n
a
to
r
a
n
d
g
e
n
e
r
a
t
o
r
a
r
e
f
i
x
e
d
.
T
h
i
s
u
p
d
a
t
e
m
a
y
b
e
co
m
p
l
e
t
e
d
i
n
5
0
0
i
t
e
r
at
i
o
n
s
b
a
s
e
d
o
n
t
h
e
e
x
p
e
r
i
m
e
n
ta
l
f
i
n
d
i
n
g
s
(
t
h
e
l
o
s
s
(
3
)
c
o
n
v
e
r
g
e
s
q
u
i
c
k
l
y
w
i
t
h
o
u
r
r
ev
i
s
e
d
i
n
i
t
i
al
i
z
at
i
o
n
,
u
s
u
al
l
y
r
ea
c
h
i
n
g
s
t
a
b
i
l
it
y
w
it
h
i
n
1
0
0
it
e
r
at
i
o
n
s
i
n
l
es
s
t
h
a
n
1
s
e
c
o
n
d
)
.
F
o
r
e
a
c
h
o
f
t
h
e
t
h
r
e
e
ca
t
e
g
o
r
i
e
s
,
w
e
e
m
p
l
o
y
t
h
e
s
a
m
e
n
e
t
w
o
r
k
d
e
s
i
g
n
.
2
.
4
.
P
r
o
po
s
ed
a
lg
o
rit
hm
Ou
r
p
r
o
p
o
s
itio
n
e
n
tails
an
alg
o
r
ith
m
r
ec
o
g
n
ized
as
l
ea
r
n
in
g
-
b
ased
s
k
etch
-
im
a
g
e
g
en
er
atio
n
(
L
b
SIG
)
.
T
h
is
alg
o
r
ith
m
ex
p
lo
its
SIG
AN
ar
ch
itectu
r
e
f
o
r
ef
f
icien
tly
g
en
e
r
atin
g
r
ea
lis
tic
p
h
o
to
f
r
o
m
g
iv
en
h
an
d
-
d
r
awn
s
k
etch
.
As
p
r
esen
ted
in
alg
o
r
ith
m
1
,
it
h
as
a
lear
n
in
g
-
b
ased
ap
p
r
o
ac
h
f
o
r
g
en
e
r
atin
g
im
ag
es
f
r
o
m
s
k
etch
es,
u
s
in
g
th
e
C
UHK
f
ac
e
s
k
etch
d
atab
ase
(
C
UFS)
d
ataset
as
in
p
u
t.
T
h
e
o
u
tp
u
t
o
f
th
e
alg
o
r
ith
m
in
clu
d
es
th
e
g
en
er
ated
im
ag
e
r
esu
lts
(
R
)
an
d
p
er
f
o
r
m
an
ce
s
tati
s
tics
(
P).
T
h
e
p
r
o
ce
s
s
b
eg
in
s
with
d
ata
au
g
m
en
tatio
n
o
f
th
e
C
UFS
d
ataset
(
D)
,
wh
ich
is
th
en
s
p
lit
in
to
two
s
u
b
s
ets
k
n
o
wn
as
tr
ain
in
g
s
et
(
T
1
)
a
n
d
test
s
et
(
T
2
)
.
T
h
e
co
r
e
o
f
th
e
alg
o
r
ith
m
is
th
e
c
r
ea
tio
n
an
d
t
r
ain
in
g
o
f
th
e
SIG
AN
m
o
d
el.
T
h
e
ar
c
h
itectu
r
e
o
f
SIG
AN
is
co
n
f
ig
u
r
ed
as
s
h
o
wn
in
Fig
u
r
e
1
,
an
d
th
e
m
o
d
e
l
is
th
en
co
m
p
iled
.
T
h
e
tr
ain
i
n
g
p
r
o
ce
s
s
in
v
o
lv
es
u
s
in
g
th
e
f
ir
s
t
s
u
b
s
et
(
T
1
)
t
o
tr
ain
th
e
SIG
AN
m
o
d
el
(
m
)
,
wh
ich
is
th
en
p
er
s
is
ted
f
o
r
f
u
tu
r
e
u
s
e.
O
n
ce
th
e
m
o
d
el
is
tr
ain
ed
,
it
is
lo
ad
e
d
,
an
d
th
e
s
ec
o
n
d
s
u
b
s
et
(
T
2
)
is
u
s
ed
to
g
e
n
er
ate
im
ag
es
with
th
e
tr
ain
ed
m
o
d
el
(
m
'
)
.
T
h
e
p
er
f
o
r
m
an
ce
o
f
th
e
g
en
er
ated
im
ag
es
is
e
v
alu
ated
ag
ain
s
t
th
e
g
r
o
u
n
d
tr
u
th
,
an
d
b
o
th
t
h
e
r
esu
lts
(
R
)
an
d
th
e
p
er
f
o
r
m
an
ce
s
tatis
tics
(
P)
ar
e
d
is
p
lay
ed
.
I
n
s
u
m
m
a
r
y
,
th
e
alg
o
r
ith
m
f
o
llo
ws
a
s
tr
u
ctu
r
ed
a
p
p
r
o
ac
h
to
tr
ain
a
n
e
u
r
al
n
etwo
r
k
o
n
au
g
m
en
ted
d
ata,
g
e
n
er
ate
im
ag
es
f
r
o
m
s
k
etch
es,
an
d
ev
alu
ate
t
h
e
g
en
er
ate
d
im
ag
es
b
ased
o
n
th
eir
f
i
d
elity
to
th
e
g
r
o
u
n
d
tr
u
th
.
T
h
e
k
ey
co
m
p
o
n
e
n
ts
in
clu
d
e
d
ata
p
r
e
p
ar
atio
n
,
m
o
d
el
c
o
n
f
ig
u
r
atio
n
an
d
tr
ain
in
g
,
im
ag
e
g
en
er
atio
n
,
an
d
p
er
f
o
r
m
an
ce
e
v
alu
atio
n
.
Alg
o
r
ith
m
1
.
L
ea
r
n
in
g
b
ased
s
k
etch
-
im
ag
e
g
en
er
atio
n
Input: CUFS dataset D
Output: Image
generation results R, performance statistics P
1.
Begin
2.
D' = Data Augmentation(D)
3.
(T1, T2) = Split Data(D')
Building and Training SIGAN
4.
Configure SIGAN architecture (as shown in Figure 1)
5.
Compile SIGAN model m
6.
m’ = Train SIGAN(T1)
7.
Persist model m'
Image Generation
8.
Load m'
9.
R=Image Generation (T2, m')
10.
P=Evaluation (R, ground truth)
11.
Display R
12.
Display P13.
End
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
S
I
GA
N
:
a
g
e
n
era
tive
a
d
ve
r
s
a
r
ia
l n
etw
o
r
k
a
r
ch
itectu
r
e
fo
r
…
(
B
u
d
d
a
n
n
a
g
a
r
i La
th
a
)
3123
2
.
5
.
Da
t
a
s
et
d
et
a
ils
C
UFS
d
ata
s
et
[
1
4
]
is
u
s
ed
f
o
r
th
e
em
p
ir
ical
s
tu
d
y
in
th
is
p
ap
er
.
T
h
is
d
ataset
is
wid
el
y
u
s
ed
f
o
r
im
ag
e
tr
an
s
latio
n
an
d
co
m
p
u
t
er
v
is
io
n
ap
p
licatio
n
s
.
T
h
is
d
ataset
it
h
as
6
0
6
f
ac
es
an
d
th
er
e
is
a
co
r
r
esp
o
n
d
in
g
s
k
etch
f
o
r
ea
ch
f
ac
e.
An
o
t
h
er
d
ataset
n
am
ed
C
UFSF
[
1
5
]
–
[
3
4
]
is
also
u
s
ed
in
th
e
ex
p
er
i
m
en
tal
s
tu
d
y
o
f
t
h
is
p
ap
er
.
I
t h
as 1
,
1
9
4
s
am
p
les.
2
.
6
.
E
v
a
lua
t
i
o
n
m
et
ho
do
lo
g
y
T
h
e
r
ea
lis
m
an
d
d
iv
er
s
ity
o
f
s
y
n
th
etic
p
h
o
to
g
r
ap
h
s
an
d
d
r
a
win
g
s
ar
e
ass
e
s
s
ed
in
th
i
s
s
tu
d
y
u
s
in
g
th
e
(
FID
)
Fré
ch
et
in
ce
p
tio
n
d
is
tan
ce
.
Giv
en
its
g
r
ea
t
d
eg
r
ee
o
f
ag
r
ee
m
en
t
with
h
u
m
a
n
v
is
io
n
,
FID
h
as
f
o
u
n
d
wid
esp
r
ea
d
u
s
e
in
p
ictu
r
e
g
en
er
atin
g
ap
p
licatio
n
s
.
R
ea
l
an
d
s
y
n
th
etic
d
ata
d
is
tr
ib
u
tio
n
s
a
r
e
clo
s
er
wh
en
th
e
FID
v
alu
e
is
lo
wer
.
3.
E
XP
E
R
I
M
E
N
T
A
L
RE
SUL
T
S
I
n
o
u
r
ex
p
er
im
en
ts
,
s
h
o
wn
in
Fig
u
r
es
2
to
4
an
d
T
ab
le
s
1
a
n
d
2
.
we
m
ak
e
u
s
e
o
f
all
th
e
tes
t
s
am
p
les
an
d
ca
lcu
late
th
e
FID
u
s
in
g
t
h
e
2
0
4
8
-
d
im
en
s
io
n
al
f
e
atu
r
es
ex
tr
ac
ted
f
r
o
m
th
e
I
n
ce
p
tio
n
-
v
3
n
etwo
r
k
.
T
h
is
n
etwo
r
k
h
as
b
ee
n
s
u
b
jecte
d
to
p
r
e
-
tr
ain
in
g
o
n
I
m
a
g
eNe
t.
T
o
im
p
ar
tially
ev
alu
ate
th
e
q
u
ality
o
f
th
e
s
y
n
th
esized
p
ictu
r
e,
“
we
u
tili
z
e
th
e
f
ea
tu
r
e
s
im
ilar
ity
in
d
ex
m
etr
ic
(
FS
I
M)
f
o
r
c
o
n
tr
asti
n
g
th
e
s
y
n
th
etic
im
ag
e
with
th
e
co
r
r
esp
o
n
d
i
n
g
g
r
o
u
n
d
-
tr
u
th
im
ag
e
.
I
n
ter
esti
n
g
ly
,
wh
ile
FS
I
M
h
as
g
ain
ed
p
o
p
u
l
ar
ity
in
th
e
r
ea
lm
o
f
f
ac
e
p
h
o
to
-
s
k
etch
s
y
n
th
esis
an
d
h
as
”
p
r
o
v
en
e
f
f
ec
tiv
e
in
ass
ess
in
g
th
e
q
u
ality
o
f
r
ea
l
im
ag
es,
it
f
alls
s
h
o
r
t
in
ter
m
s
o
f
alig
n
in
g
with
h
u
m
a
n
p
er
ce
p
tio
n
w
h
en
it
c
o
m
es
to
s
y
n
th
esized
f
ac
e
p
h
o
to
s
an
d
d
r
awin
g
s
.
B
y
u
tili
zin
g
th
e
ar
tific
ially
g
en
er
ated
“
p
h
o
to
s
/s
k
etch
es
as
t
h
e
im
ag
es
in
th
e
g
aller
y
an
d
th
e
au
th
en
ti
c
p
h
o
to
s
/s
k
etch
es
as
th
e
p
r
o
b
e
im
ag
e,
we
u
ltima
tely
c
o
n
d
u
ct
a
s
tatis
tical
ev
alu
atio
n
o
f
th
e
ac
cu
r
ac
y
o
f
f
ac
e
r
ec
o
g
n
itio
n
.
Nu
ll
-
s
p
ac
e
lin
ea
r
d
is
cr
im
in
an
t
an
aly
s
is
(
NL
DA)
is
em
p
lo
y
ed
t
o
co
n
d
u
ct
t
h
e
f
ac
e
r
ec
o
g
n
itio
n
test
s
.
B
ef
o
r
e
r
ep
o
r
tin
g
th
e
av
e
r
ag
e
ac
cu
r
ac
y
,
we
c
o
n
d
u
ct
ev
er
y
s
in
g
le
f
ac
e
r
ec
o
g
n
itio
n
e
x
p
er
im
en
t
2
0
tim
es,
r
an
d
o
m
l
y
p
ar
titi
o
n
i
n
g
th
e
d
ata
”
d
u
r
in
g
ea
ch
iter
atio
n
.
Fig
u
r
e
2
.
C
o
m
p
u
te
L
2
-
n
o
r
m
a
n
d
SS
I
M
T
ab
le
1
.
Per
f
o
r
m
an
ce
co
m
p
a
r
is
o
n
with
C
UFS d
ataset
S
k
e
t
c
h
-
i
ma
g
e
g
e
n
e
r
a
t
i
o
n
me
t
h
o
d
P
e
r
f
o
r
ma
n
c
e
(
%)
F
I
D
F
S
I
M
N
LD
A
BP
-
GAN
8
6
.
1
8
6
1
6
9
.
1
6
9
1
9
3
.
1
9
3
1
C
o
n
d
i
t
i
o
n
a
l
G
A
N
4
3
.
2
4
3
2
7
1
.
1
7
1
1
9
5
.
5
9
5
5
S
I
G
A
N
(
P
r
o
p
o
s
e
d
)
3
8
.
2
3
4
6
7
1
.
2
5
3
1
9
5
.
6
8
5
4
Tab
le 2
.
P
e
rfo
rm
a
n
c
e
c
o
m
p
a
riso
n
with
CUF
S
F
d
a
tas
e
t
S
k
e
t
c
h
-
i
ma
g
e
g
e
n
e
r
a
t
i
o
n
me
t
h
o
d
P
e
r
f
o
r
ma
n
c
e
(
%)
F
I
D
F
S
I
M
N
LD
A
BP
-
GAN
4
2
.
9
4
2
9
6
8
.
2
6
8
2
6
7
.
5
6
7
5
C
o
n
d
i
t
i
o
n
a
l
G
A
N
2
9
.
2
2
9
2
7
2
.
8
7
2
8
8
0
.
9
8
0
9
S
I
G
A
N
(
P
r
o
p
o
s
e
d
)
2
0
.
3
1
4
2
7
2
.
8
9
6
5
7
8
.
0
1
4
3
9
0
.
6
8
9
2
.
3
8
9
3
.
4
5
9
3
.
6
2
89
8
9
.
5
90
9
0
.
5
91
9
1
.
5
92
9
2
.
5
93
9
3
.
5
94
A
x
i
s
T
i
t
l
e
M
odel
s
L2
-
n
o
rm
G
e
n
e
r
a
t
e
d
_
Pi
x
e
l_
1
Ge
n
e
r
a
t
e
d
_
P
i
x
e
l_8
Ge
n
e
r
a
t
e
d
_
P
i
x
e
l_5
G
e
n
e
r
a
t
e
d
_
Pi
x
e
l_
2
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
.
3
,
J
u
n
e
20
25
:
3
1
1
8
-
3
1
2
6
3124
Fig
u
r
e
3
.
Per
f
o
r
m
an
c
e
co
m
p
ar
is
o
n
with
C
UFS d
ataset
Fig
u
r
e
4
.
Per
f
o
r
m
an
c
e
co
m
p
ar
is
o
n
with
C
UFS
F
d
ataset
4.
DIS
CU
SS
I
O
N
W
i
t
h
t
h
e
e
m
e
r
g
e
n
c
e
o
f
A
I
an
d
D
L
m
o
d
e
l
s
,
t
h
e
r
e
h
a
s
b
ee
n
a
n
i
n
c
r
e
a
s
e
d
s
u
c
c
es
s
r
at
e
i
n
s
o
l
v
i
n
g
p
r
o
b
l
e
m
s
i
n
m
a
n
y
r
e
a
l
-
w
o
r
l
d
d
o
m
a
i
n
s
.
C
o
m
p
u
t
e
r
v
i
s
i
o
n
ap
p
l
i
c
a
t
i
o
n
s
a
r
e
w
i
d
el
y
u
s
e
d
to
s
i
m
p
li
f
y
c
o
m
p
l
e
x
a
p
p
l
i
c
a
t
i
o
n
s
.
I
m
a
g
e
t
r
a
n
s
l
a
ti
o
n
i
s
o
n
e
o
f
t
h
e
i
m
p
o
r
t
a
n
t
a
p
p
l
ic
at
i
o
n
s
d
es
i
r
e
d
b
y
m
a
n
y
d
o
m
a
i
n
s
.
T
h
e
g
e
n
e
r
a
ti
o
n
o
f
a
s
k
e
t
c
h
i
m
a
g
e
f
r
o
m
a
g
i
v
e
n
r
e
a
l
i
m
a
g
e
a
n
d
a
l
s
o
g
e
n
e
r
a
ti
o
n
o
f
a
r
e
a
l
i
m
a
g
e
f
r
o
m
t
h
e
g
i
v
e
n
s
k
e
t
c
h
i
m
a
g
e
a
r
e
t
w
o
i
m
p
o
r
t
a
n
t
a
c
t
i
v
it
i
es
o
f
i
m
a
g
e
s
y
n
t
h
e
s
is
.
T
r
a
d
i
t
i
o
n
a
l
a
p
p
r
o
ac
h
e
s
f
o
r
i
m
a
g
e
s
y
n
t
h
e
s
is
c
o
u
l
d
n
o
t
p
r
o
v
i
d
e
t
h
e
r
e
q
u
i
r
e
d
p
e
r
f
o
r
m
a
n
c
e
.
I
n
o
t
h
e
r
w
o
r
d
s
,
i
m
a
g
e
p
r
o
c
ess
i
n
g
a
n
d
h
e
u
r
i
s
ti
c
a
p
p
r
o
a
c
h
es
s
h
o
w
e
d
l
i
m
i
t
at
i
o
n
s
i
n
p
r
o
d
u
c
i
n
g
a
c
c
u
r
a
t
e
o
u
t
c
o
m
e
s
.
O
f
l
a
t
e
,
l
e
a
r
n
i
n
g
b
as
e
d
ap
p
r
o
a
c
h
e
s
c
a
m
e
i
n
t
o
e
x
is
t
e
n
c
e
t
o
h
a
v
e
m
o
r
e
c
o
m
p
r
e
h
e
n
s
i
v
e
a
p
p
r
o
a
c
h
i
n
l
e
ar
n
i
n
g
f
r
o
m
i
n
p
u
t
i
m
a
g
es
a
n
d
to
g
e
n
e
r
a
t
e
d
es
i
r
e
d
o
u
t
c
o
m
es
wi
t
h
b
e
t
te
r
a
c
c
u
r
a
c
y
.
P
a
r
t
i
c
u
l
a
r
l
y
,
G
AN
a
r
c
h
i
te
c
t
u
r
e
s
w
e
r
e
d
e
v
el
o
p
e
d
f
o
r
d
a
t
a
a
u
g
m
e
n
t
a
t
i
o
n
a
n
d
s
y
n
t
h
e
s
i
s
o
f
n
e
w
i
m
a
g
e
s
i
n
t
h
e
c
o
m
p
u
t
e
r
v
i
s
i
o
n
d
o
m
a
i
n
.
T
h
i
s
p
a
p
e
r
f
o
c
u
s
e
s
o
n
g
e
n
e
r
at
i
o
n
o
f
r
e
a
l
i
m
a
g
e
f
r
o
m
t
h
e
g
i
v
e
n
h
a
n
d
-
d
r
a
w
n
s
k
e
t
c
h
.
T
h
e
G
A
N
a
r
c
h
i
te
c
t
u
r
e
p
r
o
v
i
d
e
d
i
n
[
8
]
s
e
r
v
e
s
a
s
b
a
s
i
s
f
o
r
o
u
r
w
o
r
k
i
n
p
r
o
p
o
s
i
n
g
a
n
o
v
e
l
G
A
N
a
r
c
h
i
t
e
c
t
u
r
e
k
n
o
w
n
a
s
S
I
G
A
N
.
S
I
GA
N
is
f
o
u
n
d
t
o
h
a
v
e
b
e
t
t
e
r
e
f
f
e
c
ti
v
e
n
e
s
s
c
o
m
p
a
r
e
d
t
o
s
t
at
e
-
of
-
t
h
e
-
a
r
t
a
r
c
h
it
e
c
t
u
r
e
s
.
5.
CO
NCLU
SI
O
N
AND
F
U
T
U
RE
WO
RK
A
GAN
ar
ch
itectu
r
e
k
n
o
wn
a
s
s
k
etch
-
im
ag
e
GAN
(
SIG
AN
)
was
p
r
o
p
o
s
ed
f
o
r
s
y
n
th
esizin
g
r
ea
lis
tic
p
h
o
to
s
f
r
o
m
h
an
d
-
d
r
aw
n
s
k
etch
es.
B
o
th
g
en
er
ato
r
(
G)
a
n
d
d
is
cr
im
in
ato
r
(
D)
co
m
p
o
n
e
n
ts
ar
e
d
esig
n
ed
b
ased
o
n
DL
m
o
d
els
f
o
llo
win
g
a
n
o
n
-
co
o
p
er
ativ
e
g
a
m
e
th
eo
r
y
to
war
d
s
im
p
r
o
v
i
n
g
im
ag
e
g
en
er
atio
n
p
er
f
o
r
m
an
ce
.
SIG
AN
ex
p
lo
its
im
p
r
o
v
is
ed
im
ag
e
r
ep
r
esen
tatio
n
an
d
lear
n
in
g
o
f
d
ata
d
is
tr
ib
u
t
io
n
.
Ou
r
p
r
o
p
o
s
al
en
co
m
p
ass
es
an
alg
o
r
ith
m
tit
led
lear
n
in
g
-
b
ased
s
k
etch
-
im
a
g
e
g
en
e
r
atio
n
(
L
b
SIG
)
.
T
h
is
alg
o
r
ith
m
e
x
p
lo
its
SIG
AN
ar
ch
itectu
r
e
f
o
r
ef
f
ici
en
tly
g
en
er
atin
g
r
ea
lis
tic
p
h
o
to
f
r
o
m
g
i
v
en
h
an
d
-
d
r
awn
s
k
e
tch
.
T
h
e
co
n
tex
t
u
al
in
f
o
r
m
atio
n
,
o
r
th
e
r
elatio
n
s
h
i
p
b
etwe
en
th
e
s
k
etch
an
d
p
ictu
r
e
co
m
p
o
n
e
n
ts
,
is
au
to
m
atica
lly
ca
p
tu
r
ed
b
y
th
e
jo
in
t
im
ag
e
in
o
u
r
s
y
s
tem
,
a
n
d
th
is
is
u
s
ef
u
l
f
o
r
lear
n
in
g
th
eir
co
llectiv
e
d
is
tr
ib
u
tio
n
u
tili
zin
g
GAN.
I
n
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
S
I
GA
N
:
a
g
e
n
era
tive
a
d
ve
r
s
a
r
ia
l n
etw
o
r
k
a
r
ch
itectu
r
e
fo
r
…
(
B
u
d
d
a
n
n
a
g
a
r
i La
th
a
)
3125
p
ar
ticu
lar
,
we
u
s
e
jo
i
n
t
p
ictu
r
es
to
tr
ain
a
GAN
m
o
d
el,
w
h
ich
th
en
u
s
es
th
e
co
n
tex
t
o
f
t
h
e
m
atch
in
g
s
k
etch
co
m
p
o
n
en
t
to
au
to
m
atica
lly
p
r
ed
ict
th
e
d
a
m
ag
ed
im
ag
e
p
ar
t.
SIG
AN
is
ass
e
s
s
ed
u
s
in
g
a
b
en
ch
m
ar
k
d
ataset
ca
lled
C
UHK
f
ac
e
s
k
etch
d
atab
ase
(
C
UFS).
Fro
m
th
e
em
p
ir
ical
s
tu
d
y
,
it
is
o
b
s
er
v
e
d
th
at
th
e
p
r
o
p
o
s
ed
SIG
AN
ar
ch
itectu
r
e
with
u
n
d
er
ly
in
g
d
ee
p
lear
n
in
g
m
o
d
els
co
u
ld
o
u
t
p
er
f
o
r
m
ex
is
tin
g
GA
N
m
o
d
els
in
ter
m
s
o
f
Fré
ch
et
in
ce
p
tio
n
d
is
tan
ce
(
FID
)
with
3
8
.
2
3
4
6
%.
I
n
f
u
tu
r
e,
we
in
ten
d
to
im
p
r
o
v
e
o
v
er
SIG
AN
ar
ch
itectu
r
e
with
an
en
co
d
er
f
o
r
f
ac
e
ap
p
ea
r
an
ce
an
d
an
en
co
d
er
f
o
r
f
ac
e
lab
els
to
war
d
s
b
etter
p
er
f
o
r
m
an
ce
.
An
o
th
er
d
ir
ec
tio
n
f
o
r
f
u
tu
r
e
wo
r
k
is
to
s
tack
GAN
m
o
d
els to
war
d
s
im
p
r
o
v
i
n
g
p
e
r
f
o
r
m
an
ce
f
u
r
th
er
.
F
UNDING
I
NF
O
R
M
A
T
I
O
N
Au
th
o
r
s
s
tate
n
o
f
u
n
d
in
g
in
v
o
lv
ed
.
AUTHO
R
CO
NT
RI
B
UT
I
O
NS ST
A
T
E
M
E
N
T
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
B
u
d
d
an
n
ag
a
r
i L
ath
a
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
Ath
iy
o
o
r
Kan
n
an
Velm
u
r
u
g
an
✓
✓
✓
✓
✓
✓
✓
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
No
co
n
f
lict o
f
in
ter
est.
DATA AV
AI
L
AB
I
L
I
T
Y
D
a
t
a
a
v
a
il
a
b
i
li
t
y
is
n
o
t
a
p
p
l
i
ca
b
l
e
t
o
t
h
is
p
a
p
e
r
a
s
n
o
n
e
w
d
at
a
w
e
r
e
c
r
e
a
t
e
d
o
r
a
n
al
y
z
e
d
i
n
t
h
is
s
t
u
d
y
.
RE
F
E
R
E
NC
E
S
[
1
]
M
.
S
i
g
a
l
a
,
A
.
B
e
e
r
,
L
.
H
o
d
g
so
n
,
a
n
d
A
.
O
’
C
o
n
n
o
r
,
B
i
g
d
a
t
a
f
o
r
m
e
a
su
ri
n
g
t
h
e
i
m
p
a
c
t
o
f
t
o
u
r
i
sm
e
c
o
n
o
m
i
c
d
e
v
e
l
o
p
m
e
n
t
p
ro
g
r
a
m
m
e
s:
A
p
r
o
c
e
ss
a
n
d
q
u
a
l
i
t
y
c
ri
t
e
r
i
a
f
r
a
m
e
w
o
rk
f
o
r
u
si
n
g
b
i
g
d
a
t
a
.
2
0
1
9
.
d
o
i
:
1
0
.
1
0
0
7
/
9
7
8
-
981
-
13
-
6
3
3
9
-
9
_
4
.
[
2
]
G
.
N
g
u
y
e
n
e
t
a
l
.
,
“
M
a
c
h
i
n
e
l
e
a
r
n
i
n
g
a
n
d
d
e
e
p
l
e
a
r
n
i
n
g
f
r
a
mew
o
r
k
s a
n
d
l
i
b
r
a
r
i
e
s
f
o
r
l
a
r
g
e
-
sc
a
l
e
d
a
t
a
mi
n
i
n
g
:
A
s
u
r
v
e
y
,
”
Ar
t
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
Re
v
i
e
w
,
v
o
l
.
5
2
,
n
o
.
1
,
p
p
.
7
7
–
1
2
4
,
2
0
1
9
,
d
o
i
:
1
0
.
1
0
0
7
/
s
1
0
4
6
2
-
0
1
8
-
0
9
6
7
9
-
z.
[
3
]
M
.
S
.
S
a
n
n
i
d
h
a
n
,
G
.
A
n
a
n
t
h
P
r
a
b
h
u
,
D
.
E
.
R
o
b
b
i
n
s
,
a
n
d
C
.
S
h
a
s
k
y
,
“
E
v
a
l
u
a
t
i
n
g
t
h
e
p
e
r
f
o
r
m
a
n
c
e
o
f
f
a
c
e
s
k
e
t
c
h
g
e
n
e
r
a
t
i
o
n
u
s
i
n
g
g
e
n
e
r
a
t
i
v
e
a
d
v
e
r
s
a
r
i
a
l
n
e
t
w
o
r
k
s
,
”
P
a
t
t
e
r
n
R
e
c
o
g
n
i
t
i
o
n
L
e
t
t
e
r
s
,
v
o
l
.
1
2
8
,
p
p
.
4
5
2
–
4
5
8
,
D
e
c
.
2
0
1
9
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
p
a
t
r
e
c
.
2
0
1
9
.
1
0
.
0
1
0
.
[
4
]
W
.
C
h
a
o
,
L.
C
h
a
n
g
,
X
.
W
a
n
g
,
J.
C
h
e
n
g
,
X
.
D
e
n
g
,
a
n
d
F
.
D
u
a
n
,
“
H
i
g
h
-
F
i
d
e
l
i
t
y
F
a
c
e
S
k
e
t
c
h
-
To
-
P
h
o
t
o
S
y
n
t
h
e
si
s
U
si
n
g
G
e
n
e
r
a
t
i
v
e
A
d
v
e
r
sari
a
l
N
e
t
w
o
r
k
,
”
i
n
2
0
1
9
I
EE
E
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
I
m
a
g
e
Pr
o
c
e
ssi
n
g
(
I
C
I
P)
,
S
e
p
.
2
0
1
9
,
p
p
.
4
6
9
9
–
4
7
0
3
.
d
o
i
:
1
0
.
1
1
0
9
/
I
C
I
P
.
2
0
1
9
.
8
8
0
3
5
4
9
.
[
5
]
L.
Y
a
n
,
W
.
Z
h
e
n
g
,
C
.
G
o
u
,
a
n
d
F
.
-
Y
.
W
a
n
g
,
“
I
sG
A
N
:
i
d
e
n
t
i
t
y
-
se
n
si
t
i
v
e
g
e
n
e
r
a
t
i
v
e
a
d
v
e
r
sari
a
l
n
e
t
w
o
r
k
f
o
r
f
a
c
e
p
h
o
t
o
-
s
k
e
t
c
h
sy
n
t
h
e
si
s,”
P
a
t
t
e
r
n
R
e
c
o
g
n
i
t
i
o
n
,
v
o
l
.
1
1
9
,
p
.
1
0
8
0
7
7
,
N
o
v
.
2
0
2
1
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
p
a
t
c
o
g
.
2
0
2
1
.
1
0
8
0
7
7
.
[
6
]
H
.
B
i
,
N
.
L
i
,
H
.
G
u
a
n
,
D
.
L
u
,
a
n
d
L
.
Y
a
n
g
,
“
A
m
u
l
t
i
-
s
c
a
l
e
c
o
n
d
i
t
i
o
n
a
l
g
e
n
e
r
a
t
i
v
e
a
d
v
e
r
s
a
r
i
a
l
n
e
t
w
o
r
k
f
o
r
f
a
c
e
s
k
e
t
c
h
s
y
n
t
h
e
s
i
s
,
”
i
n
2
0
1
9
I
E
E
E
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
I
m
a
g
e
P
r
o
c
e
s
s
i
n
g
(
I
C
I
P
)
,
S
e
p
.
2
0
1
9
,
p
p
.
3
8
7
6
–
3
8
8
0
.
d
o
i
:
1
0
.
1
1
0
9
/
I
C
I
P
.
2
0
1
9
.
8
8
0
3
6
2
9
.
[
7
]
W
.
W
a
n
,
Y
.
Y
a
n
g
,
a
n
d
H
.
J.
Le
e
,
“
G
e
n
e
r
a
t
i
v
e
a
d
v
e
r
sar
i
a
l
l
e
a
r
n
i
n
g
f
o
r
d
e
t
a
i
l
-
p
r
e
ser
v
i
n
g
f
a
c
e
sk
e
t
c
h
sy
n
t
h
e
si
s,
”
N
e
u
ro
c
o
m
p
u
t
i
n
g
,
v
o
l
.
4
3
8
,
p
p
.
1
0
7
–
1
2
1
,
M
a
y
2
0
2
1
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
n
e
u
c
o
m
.
2
0
2
1
.
0
1
.
0
5
0
.
[
8
]
W
.
W
a
n
a
n
d
H
.
J.
L
e
e
,
“
G
e
n
e
r
a
t
i
v
e
a
d
v
e
r
sar
i
a
l
mu
l
t
i
-
t
a
s
k
l
e
a
r
n
i
n
g
f
o
r
f
a
c
e
sk
e
t
c
h
s
y
n
t
h
e
s
i
s
a
n
d
r
e
c
o
g
n
i
t
i
o
n
,
”
i
n
Pr
o
c
e
e
d
i
n
g
s
-
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
I
m
a
g
e
P
ro
c
e
ssi
n
g
,
I
C
I
P
,
2
0
1
9
,
v
o
l
.
2
0
1
9
-
S
e
p
t
e
,
p
p
.
4
0
6
5
–
4
0
6
9
.
d
o
i
:
1
0
.
1
1
0
9
/
I
C
I
P
.
2
0
1
9
.
8
8
0
3
6
1
7
.
[
9
]
J.
Z
h
e
n
g
,
W
.
S
o
n
g
,
Y
.
W
u
,
R
.
X
u
,
a
n
d
F
.
Li
u
,
“
F
e
a
t
u
r
e
e
n
c
o
d
e
r
g
u
i
d
e
d
g
e
n
e
r
a
t
i
v
e
a
d
v
e
r
sari
a
l
n
e
t
w
o
r
k
f
o
r
f
a
c
e
p
h
o
t
o
-
s
k
e
t
c
h
sy
n
t
h
e
si
s,”
I
EEE
Ac
c
e
ss
,
v
o
l
.
7
,
p
p
.
1
5
4
9
7
1
–
1
5
4
9
8
5
,
2
0
1
9
,
d
o
i
:
1
0
.
1
1
0
9
/
A
C
C
ESS
.
2
0
1
9
.
2
9
4
9
0
7
0
.
[
1
0
]
K
.
D
u
,
H
.
Z
h
o
u
,
L.
C
a
o
,
Y
.
G
u
o
,
a
n
d
T.
W
a
n
g
,
“
M
H
G
A
N
:
mu
l
t
i
-
h
i
e
r
a
r
c
h
i
e
s
g
e
n
e
r
a
t
i
v
e
a
d
v
e
r
sari
a
l
n
e
t
w
o
r
k
f
o
r
h
i
g
h
-
q
u
a
l
i
t
y
f
a
c
e
sk
e
t
c
h
s
y
n
t
h
e
si
s,”
I
EEE
Ac
c
e
ss
,
v
o
l
.
8
,
p
p
.
2
1
2
9
9
5
–
2
1
3
0
1
1
,
2
0
2
0
,
d
o
i
:
1
0
.
1
1
0
9
/
A
C
C
ESS
.
2
0
2
0
.
3
0
4
1
2
8
4
.
[
1
1
]
Z.
Li
,
C
.
D
e
n
g
,
E.
Y
a
n
g
,
a
n
d
D
.
Ta
o
,
“
S
t
a
g
e
d
s
k
e
t
c
h
-
to
-
i
m
a
g
e
sy
n
t
h
e
si
s
v
i
a
sem
i
-
su
p
e
r
v
i
s
e
d
g
e
n
e
r
a
t
i
v
e
a
d
v
e
r
sari
a
l
n
e
t
w
o
r
k
s,
”
I
EEE
T
r
a
n
s
a
c
t
i
o
n
s
o
n
M
u
l
t
i
m
e
d
i
a
,
v
o
l
.
2
3
,
p
p
.
2
6
9
4
–
2
7
0
5
,
2
0
2
1
,
d
o
i
:
1
0
.
1
1
0
9
/
T
M
M
.
2
0
2
0
.
3
0
1
5
0
1
5
.
[
1
2
]
D
.
W
a
d
h
w
a
,
U
.
M
a
h
a
r
a
n
a
,
D
.
S
h
a
h
,
V
.
Y
a
d
a
v
,
a
n
d
P
.
P
a
n
d
e
y
,
“
H
u
ma
n
sk
e
t
c
h
r
e
c
o
g
n
i
t
i
o
n
u
si
n
g
g
e
n
e
r
a
t
i
v
e
a
d
v
e
r
sari
a
l
n
e
t
w
o
r
k
s
a
n
d
o
n
e
-
sh
o
t
l
e
a
r
n
i
n
g
,
”
i
n
2
0
1
9
T
w
e
l
f
t
h
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
C
o
n
t
e
m
p
o
ra
r
y
C
o
m
p
u
t
i
n
g
(
I
C
3
)
,
A
u
g
.
2
0
1
9
,
p
p
.
1
–
6
.
d
o
i
:
1
0
.
1
1
0
9
/
I
C
3
.
2
0
1
9
.
8
8
4
4
8
8
5
.
[
1
3
]
L.
W
a
n
g
,
W
.
C
h
e
n
,
W
.
Y
a
n
g
,
F
.
B
i
,
a
n
d
F
.
R
.
Y
u
,
“
A
s
t
a
t
e
-
of
-
t
h
e
-
a
r
t
r
e
v
i
e
w
o
n
i
m
a
g
e
s
y
n
t
h
e
s
i
s
w
i
t
h
g
e
n
e
r
a
t
i
v
e
a
d
v
e
r
sari
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.
15
,
No
.
3
,
J
u
n
e
20
25
:
3
1
1
8
-
3
1
2
6
3126
n
e
t
w
o
r
k
s,”
I
EEE
Ac
c
e
ss
,
v
o
l
.
8
,
p
p
.
6
3
5
1
4
–
6
3
5
3
7
,
2
0
2
0
,
d
o
i
:
1
0
.
1
1
0
9
/
A
C
C
ESS
.
2
0
2
0
.
2
9
8
2
2
2
4
.
[
1
4
]
J.
Z
h
a
o
,
X
.
X
i
e
,
L.
W
a
n
g
,
M
.
C
a
o
,
a
n
d
M
.
Z
h
a
n
g
,
“
G
e
n
e
r
a
t
i
n
g
p
h
o
t
o
g
r
a
p
h
i
c
f
a
c
e
s
f
r
o
m
t
h
e
sk
e
t
c
h
g
u
i
d
e
d
b
y
a
t
t
r
i
b
u
t
e
u
si
n
g
G
A
N
,
”
I
EEE
Ac
c
e
ss
,
v
o
l
.
7
,
p
p
.
2
3
8
4
4
–
2
3
8
5
1
,
2
0
1
9
,
d
o
i
:
1
0
.
1
1
0
9
/
A
C
C
ES
S
.
2
0
1
9
.
2
8
9
9
4
6
6
.
[
1
5
]
J.
Z
h
e
n
g
,
Y
.
W
u
,
W
.
S
o
n
g
,
R
.
X
u
,
a
n
d
F
.
Li
u
,
“
M
u
l
t
i
-
sca
l
e
f
e
a
t
u
r
e
c
h
a
n
n
e
l
a
t
t
e
n
t
i
o
n
g
e
n
e
r
a
t
i
v
e
a
d
v
e
r
sar
i
a
l
n
e
t
w
o
r
k
f
o
r
f
a
c
e
sk
e
t
c
h
s
y
n
t
h
e
si
s,”
I
EEE
Ac
c
e
ss
,
v
o
l
.
8
,
p
p
.
1
4
6
7
5
4
–
1
4
6
7
6
9
,
2
0
2
0
,
d
o
i
:
1
0
.
1
1
0
9
/
A
C
C
ESS
.
2
0
2
0
.
3
0
1
5
3
1
2
.
[
1
6
]
Y
.
F
a
n
g
,
W
.
D
e
n
g
,
J.
D
u
,
a
n
d
J.
H
u
,
“
I
d
e
n
t
i
t
y
-
a
w
a
r
e
C
y
c
l
e
G
A
N
f
o
r
f
a
c
e
p
h
o
t
o
-
s
k
e
t
c
h
sy
n
t
h
e
si
s
a
n
d
r
e
c
o
g
n
i
t
i
o
n
,
”
P
a
t
t
e
rn
Re
c
o
g
n
i
t
i
o
n
,
v
o
l
.
1
0
2
,
p
.
1
0
7
2
4
9
,
Ju
n
.
2
0
2
0
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
p
a
t
c
o
g
.
2
0
2
0
.
1
0
7
2
4
9
.
[
1
7
]
M
.
Zh
u
,
J.
L
i
,
N
.
W
a
n
g
,
a
n
d
X
.
G
a
o
,
“
A
d
e
e
p
c
o
l
l
a
b
o
r
a
t
i
v
e
f
r
a
m
e
w
o
r
k
f
o
r
f
a
c
e
p
h
o
t
o
-
sk
e
t
c
h
s
y
n
t
h
e
si
s
,
”
I
EEE
T
ra
n
s
a
c
t
i
o
n
s
o
n
N
e
u
ra
l
N
e
t
w
o
rks
a
n
d
L
e
a
r
n
i
n
g
S
y
st
e
m
s
,
v
o
l
.
3
0
,
n
o
.
1
0
,
p
p
.
3
0
9
6
–
3
1
0
8
,
O
c
t
.
2
0
1
9
,
d
o
i
:
1
0
.
1
1
0
9
/
TN
N
LS.2
0
1
8
.
2
8
9
0
0
1
8
.
[
1
8
]
Y
.
Li
n
,
S
.
Li
n
g
,
K
.
F
u
,
a
n
d
P
.
C
h
e
n
g
,
“
A
n
i
d
e
n
t
i
t
y
-
p
r
e
ser
v
e
d
m
o
d
e
l
f
o
r
f
a
c
e
sk
e
t
c
h
-
p
h
o
t
o
sy
n
t
h
e
si
s,
”
I
E
EE
S
i
g
n
a
l
Pro
c
e
ss
i
n
g
L
e
t
t
e
rs
,
v
o
l
.
2
7
,
p
p
.
1
0
9
5
–
1
0
9
9
,
2
0
2
0
,
d
o
i
:
1
0
.
1
1
0
9
/
LSP
.
2
0
2
0
.
3
0
0
5
0
3
9
.
[
1
9
]
N
.
H
a
j
a
r
o
l
a
s
v
a
d
i
,
M
.
A
.
R
a
mi
r
e
z
,
W
.
B
e
c
c
a
r
o
,
a
n
d
H
.
D
e
m
i
r
e
l
,
“
G
e
n
e
r
a
t
i
v
e
a
d
v
e
r
sari
a
l
n
e
t
w
o
r
k
s
i
n
h
u
m
a
n
e
m
o
t
i
o
n
sy
n
t
h
e
si
s
:
a
r
e
v
i
e
w
,
”
I
EEE
Ac
c
e
ss
,
v
o
l
.
8
,
p
p
.
2
1
8
4
9
9
–
2
1
8
5
2
9
,
2
0
2
0
,
d
o
i
:
1
0
.
1
1
0
9
/
A
C
C
ESS
.
2
0
2
0
.
3
0
4
2
3
2
8
.
[
2
0
]
M
.
Z.
K
h
a
n
e
t
a
l
.
,
“
A
r
e
a
l
i
st
i
c
i
m
a
g
e
g
e
n
e
r
a
t
i
o
n
o
f
f
a
c
e
f
r
o
m
t
e
x
t
d
e
s
c
r
i
p
t
i
o
n
u
si
n
g
t
h
e
f
u
l
l
y
t
r
a
i
n
e
d
g
e
n
e
r
a
t
i
v
e
a
d
v
e
r
sari
a
l
n
e
t
w
o
r
k
s,”
I
EEE
Ac
c
e
ss
,
v
o
l
.
9
,
p
p
.
1
2
5
0
–
1
2
6
0
,
2
0
2
1
,
d
o
i
:
1
0
.
1
1
0
9
/
A
C
C
E
S
S
.
2
0
2
0
.
3
0
1
5
6
5
6
.
[
2
1
]
P
.
Li
,
B
.
S
h
e
n
g
,
a
n
d
C
.
L.
P
.
C
h
e
n
,
“
F
a
c
e
s
k
e
t
c
h
s
y
n
t
h
e
si
s
u
s
i
n
g
r
e
g
u
l
a
r
i
z
e
d
b
r
o
a
d
l
e
a
r
n
i
n
g
s
y
st
e
m,
”
I
EE
E
T
ra
n
s
a
c
t
i
o
n
s
o
n
N
e
u
ra
l
N
e
t
w
o
rks
a
n
d
L
e
a
r
n
i
n
g
S
y
st
e
m
s
,
v
o
l
.
3
3
,
n
o
.
1
0
,
p
p
.
5
3
4
6
–
5
3
6
0
,
O
c
t
.
2
0
2
2
,
d
o
i
:
1
0
.
1
1
0
9
/
TN
N
LS.2
0
2
1
.
3
0
7
0
4
6
3
.
[
2
2
]
M
.
Z
h
a
n
g
,
Y
.
L
i
,
N
.
W
a
n
g
,
Y
.
C
h
i
,
a
n
d
X
.
G
a
o
,
“
C
a
s
c
a
d
e
d
f
a
c
e
s
k
e
t
c
h
sy
n
t
h
e
s
i
s u
n
d
e
r
v
a
r
i
o
u
s i
l
l
u
mi
n
a
t
i
o
n
s,”
I
E
EE
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
9
,
p
p
.
1
5
0
7
–
1
5
2
1
,
2
0
2
0
,
d
o
i
:
1
0
.
1
1
0
9
/
TI
P
.
2
0
1
9
.
2
9
4
2
5
1
4
.
[
2
3
]
Q
.
Li
u
,
H
.
Z
h
a
o
,
Y
.
W
a
n
g
,
F
.
Z
h
a
n
g
,
a
n
d
M
.
R
a
mas
a
my
,
“
S
k
e
t
c
h
t
o
p
o
r
t
r
a
i
t
g
e
n
e
r
a
t
i
o
n
w
i
t
h
g
e
n
e
r
a
t
i
v
e
a
d
v
e
r
sar
i
a
l
n
e
t
w
o
r
k
s
a
n
d
e
d
g
e
c
o
n
s
t
r
a
i
n
t
,
”
C
o
m
p
u
t
e
rs
a
n
d
El
e
c
t
ri
c
a
l
E
n
g
i
n
e
e
r
i
n
g
,
v
o
l
.
9
5
,
2
0
2
1
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
c
o
m
p
e
l
e
c
e
n
g
.
2
0
2
1
.
1
0
7
3
3
8
.
[
2
4
]
X
.
Li
,
F
.
G
a
o
,
a
n
d
F
.
H
u
a
n
g
,
“
H
i
g
h
-
q
u
a
l
i
t
y
f
a
c
e
sk
e
t
c
h
s
y
n
t
h
e
s
i
s
v
i
a
g
e
o
met
r
i
c
n
o
r
ma
l
i
z
a
t
i
o
n
a
n
d
r
e
g
u
l
a
r
i
z
a
t
i
o
n
,
”
i
n
2
0
2
1
I
EEE
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
Mu
l
t
i
m
e
d
i
a
a
n
d
Ex
p
o
(
I
C
M
E)
,
J
u
l
.
2
0
2
1
,
p
p
.
1
–
6
.
d
o
i
:
1
0
.
1
1
0
9
/
I
C
M
E
5
1
2
0
7
.
2
0
2
1
.
9
4
2
8
3
4
8
.
[
2
5
]
J.
G
u
o
a
n
d
Y
.
Li
u
,
“
F
a
c
i
a
l
p
a
r
t
s
sw
a
p
p
i
n
g
w
i
t
h
g
e
n
e
r
a
t
i
v
e
a
d
v
e
r
sari
a
l
n
e
t
w
o
r
k
s,
”
J
o
u
r
n
a
l
o
f
V
i
su
a
l
C
o
m
m
u
n
i
c
a
t
i
o
n
a
n
d
I
m
a
g
e
Re
p
r
e
se
n
t
a
t
i
o
n
,
v
o
l
.
7
8
,
p
.
1
0
3
1
5
2
,
Ju
l
.
2
0
2
1
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
j
v
c
i
r
.
2
0
2
1
.
1
0
3
1
5
2
.
[
2
6
]
H
.
B
i
,
Z
.
L
i
u
,
L.
Y
a
n
g
,
K
.
W
a
n
g
,
a
n
d
N
.
Li
,
“
F
a
c
e
s
k
e
t
c
h
s
y
n
t
h
e
s
i
s:
a
s
u
r
v
e
y
,
”
M
u
l
t
i
m
e
d
i
a
T
o
o
l
s
a
n
d
A
p
p
l
i
c
a
t
i
o
n
s
,
v
o
l
.
8
0
,
n
o
.
1
2
,
p
p
.
1
8
0
0
7
–
1
8
0
2
6
,
M
a
y
2
0
2
1
,
d
o
i
:
1
0
.
1
0
0
7
/
s
1
1
0
4
2
-
0
2
0
-
1
0
3
0
1
-
0.
[
2
7
]
S
.
Ji
a
n
g
,
Z.
Ta
o
,
a
n
d
Y
.
F
u
,
“
G
e
o
m
e
t
r
i
c
a
l
l
y
e
d
i
t
a
b
l
e
f
a
c
e
i
ma
g
e
t
r
a
n
sl
a
t
i
o
n
w
i
t
h
a
d
v
e
r
sari
a
l
n
e
t
w
o
r
k
s,”
I
E
EE
T
ra
n
s
a
c
t
i
o
n
s
o
n
I
m
a
g
e
Pr
o
c
e
ssi
n
g
,
v
o
l
.
3
0
,
p
p
.
2
7
7
1
–
2
7
8
3
,
2
0
2
1
,
d
o
i
:
1
0
.
1
1
0
9
/
TI
P
.
2
0
2
1
.
3
0
5
2
0
8
4
.
[
2
8
]
K
.
K
.
B
a
b
u
a
n
d
S
.
R
.
D
u
b
e
y
,
“
C
S
G
A
N
:
c
y
c
l
i
c
-
s
y
n
t
h
e
s
i
z
e
d
g
e
n
e
r
a
t
i
v
e
a
d
v
e
r
sar
i
a
l
n
e
t
w
o
r
k
s
f
o
r
i
ma
g
e
-
to
-
i
ma
g
e
t
r
a
n
sf
o
r
mat
i
o
n
,
”
Ex
p
e
rt
S
y
st
e
m
s w
i
t
h
Ap
p
l
i
c
a
t
i
o
n
s
,
v
o
l
.
1
6
9
,
p
.
1
1
4
4
3
1
,
M
a
y
2
0
2
1
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
e
sw
a
.
2
0
2
0
.
1
1
4
4
3
1
.
[
2
9
]
W
.
X
i
a
,
Y
.
Y
a
n
g
,
a
n
d
J.
-
H
.
X
u
e
,
“
C
a
l
i
-
s
k
e
t
c
h
:
s
t
r
o
k
e
c
a
l
i
b
r
a
t
i
o
n
a
n
d
c
o
mp
l
e
t
i
o
n
f
o
r
h
i
g
h
-
q
u
a
l
i
t
y
f
a
c
e
i
m
a
g
e
g
e
n
e
r
a
t
i
o
n
f
r
o
m
h
u
m
a
n
-
l
i
k
e
sk
e
t
c
h
e
s,
”
N
e
u
ro
c
o
m
p
u
t
i
n
g
,
v
o
l
.
4
6
0
,
p
p
.
2
5
6
–
2
6
5
,
O
c
t
.
2
0
2
1
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
n
e
u
c
o
m.
2
0
2
1
.
0
7
.
0
2
9
.
[
3
0
]
N
.
A
.
M
a
sh
u
d
i
,
N
.
A
h
ma
d
,
a
n
d
N
.
M
o
h
d
N
o
o
r
,
“
L
i
W
G
A
N
:
a
l
i
g
h
t
me
t
h
o
d
t
o
i
m
p
r
o
v
e
t
h
e
p
e
r
f
o
r
m
a
n
c
e
o
f
g
e
n
e
r
a
t
i
v
e
a
d
v
e
r
sar
i
a
l
n
e
t
w
o
r
k
,
”
I
EEE
Ac
c
e
ss
,
v
o
l
.
1
0
,
p
p
.
9
3
1
5
5
–
9
3
1
6
7
,
2
0
2
2
,
d
o
i
:
1
0
.
1
1
0
9
/
A
C
C
ESS
.
2
0
2
2
.
3
2
0
3
0
6
5
.
[
3
1
]
U
.
O
sah
o
r
a
n
d
N
.
M
.
N
a
sr
a
b
a
d
i
,
“
T
e
x
t
-
g
u
i
d
e
d
sk
e
t
c
h
-
to
-
p
h
o
t
o
i
m
a
g
e
s
y
n
t
h
e
s
i
s,”
I
EEE
Ac
c
e
ss
,
v
o
l
.
1
0
,
p
p
.
9
8
2
7
8
–
9
8
2
8
9
,
2
0
2
2
,
d
o
i
:
1
0
.
1
1
0
9
/
A
C
C
ESS
.
2
0
2
2
.
3
2
0
6
7
7
1
.
[
3
2
]
K
.
K
o
,
T.
Y
e
o
m,
a
n
d
M
.
Le
e
,
“
S
u
p
e
r
st
a
r
G
A
N
:
g
e
n
e
r
a
t
i
v
e
a
d
v
e
r
sar
i
a
l
n
e
t
w
o
r
k
s
f
o
r
i
m
a
g
e
-
to
-
i
m
a
g
e
t
r
a
n
sl
a
t
i
o
n
i
n
l
a
r
g
e
-
sc
a
l
e
d
o
m
a
i
n
s,
”
N
e
u
ra
l
N
e
t
w
o
rks
,
v
o
l
.
1
6
2
,
p
p
.
3
3
0
–
3
3
9
,
M
a
y
2
0
2
3
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
n
e
u
n
e
t
.
2
0
2
3
.
0
2
.
0
4
2
.
[
3
3
]
X
.
W
a
n
g
a
n
d
X
.
Ta
n
g
,
“
F
a
c
e
p
h
o
t
o
-
sk
e
t
c
h
s
y
n
t
h
e
s
i
s
a
n
d
r
e
c
o
g
n
i
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
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
1
,
n
o
.
1
1
,
p
p
.
1
9
5
5
–
1
9
6
7
,
2
0
0
9
,
d
o
i
:
1
0
.
1
1
0
9
/
TPA
M
I
.
2
0
0
8
.
2
2
2
.
[
3
4
]
W
.
Zh
a
n
g
,
X
.
W
a
n
g
,
a
n
d
X
.
Ta
n
g
,
“
C
o
u
p
l
e
d
i
n
f
o
r
ma
t
i
o
n
-
t
h
e
o
r
e
t
i
c
e
n
c
o
d
i
n
g
f
o
r
f
a
c
e
p
h
o
t
o
-
s
k
e
t
c
h
r
e
c
o
g
n
i
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
m
p
u
t
e
r
S
o
c
i
e
t
y
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
R
e
c
o
g
n
i
t
i
o
n
,
2
0
1
1
,
p
p
.
5
1
3
–
5
2
0
.
d
o
i
:
1
0
.
1
1
0
9
/
C
V
P
R
.
2
0
1
1
.
5
9
9
5
3
2
4
.
B
I
O
G
RAP
H
I
E
S O
F
AUTH
O
RS
Bu
d
d
a
n
n
a
g
a
r
i
La
th
a
is
p
u
r
su
in
g
h
e
r
P
h
.
D
.
i
n
a
rti
ficia
l
i
n
tel
li
g
e
n
c
e
a
t
Ko
n
e
ru
Lak
sh
m
a
i
a
h
E
d
u
c
a
ti
o
n
F
o
u
n
d
a
ti
o
n
,
Vijay
a
wa
d
a
,
A
n
d
h
ra
P
ra
d
e
s
h
.
S
h
e
h
o
l
d
s
a
n
M
.
Tec
h
i
n
Co
m
p
u
ter
S
c
ie
n
c
e
En
g
i
n
e
e
rin
g
f
ro
m
JN
TUH
(2
0
1
5
)
a
n
d
a
B.
Te
c
h
in
C
o
m
p
u
ter
S
c
ien
c
e
&
En
g
i
n
e
e
rin
g
fr
o
m
S
ri
d
e
v
i
W
o
m
e
n
'
s
En
g
in
e
e
rin
g
Co
ll
e
g
e
,
JN
TU
H
(2
0
0
8
).
S
h
e
is
p
re
se
n
tl
y
wo
rk
i
n
g
a
s
a
n
a
ss
istan
t
p
r
o
fe
ss
o
r
in
th
e
De
p
a
rtme
n
t
o
f
C
o
m
p
u
ter
S
c
ien
c
e
a
n
d
En
g
in
e
e
rin
g
a
t
Ra
ji
v
G
a
n
d
h
i
Un
iv
e
rsit
y
o
f
Kn
o
wle
d
g
e
Tec
h
n
o
l
o
g
ies
-
Ba
sa
r,
Te
lan
g
a
n
a
,
a
n
d
h
a
s
1
0
y
e
a
rs
o
f
tea
c
h
in
g
e
x
p
e
rien
c
e
.
He
r
re
se
a
rc
h
in
tere
sts
i
n
c
lu
d
e
d
e
e
p
lea
rn
i
n
g
a
n
d
m
a
c
h
in
e
lea
rn
i
n
g
.
S
h
e
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
lath
a
.
re
d
d
y
5
8
0
8
@
g
m
a
il
.
c
o
m
.
Athi
y
o
o
r
K
a
n
n
a
n
Ve
l
m
u
r
u
g
a
n
g
ra
d
u
a
ted
i
n
c
o
m
p
u
ter
sc
ien
c
e
a
n
d
e
n
g
in
e
e
rin
g
fro
m
Aru
n
a
i
E
n
g
in
e
e
rin
g
Co
l
leg
e
,
a
ffil
iate
d
wit
h
t
h
e
Un
iv
e
rsit
y
o
f
M
a
d
ra
s,
Ch
e
n
n
a
i.
H
e
o
b
tai
n
e
d
h
is
M
.
Tec
h
.
in
c
o
m
p
u
t
e
r
sc
ien
c
e
a
n
d
e
n
g
i
n
e
e
rin
g
fr
o
m
Dr.
M
G
R
Ed
u
c
a
ti
o
n
a
l
a
n
d
Re
se
a
rc
h
In
stit
u
te
(De
e
m
e
d
Un
i
v
e
rsity
),
C
h
e
n
n
a
i,
a
n
d
e
a
rn
e
d
h
is
P
h
.
D.
in
Co
m
p
u
ter
S
c
ie
n
c
e
a
n
d
En
g
in
e
e
rin
g
fro
m
S
t
.
P
e
ter’s
In
stit
u
te
o
f
Hi
g
h
e
r
Ed
u
c
a
ti
o
n
a
n
d
Re
se
a
rc
h
(De
e
m
e
d
Un
iv
e
rsity
)
,
Ch
e
n
n
a
i.
He
is
c
u
rr
e
n
tl
y
w
o
rk
i
n
g
a
s
a
p
ro
f
e
ss
o
r
in
t
h
e
De
p
a
rtme
n
t
o
f
C
o
m
p
u
ter
S
c
ien
c
e
a
n
d
E
n
g
i
n
e
e
rin
g
a
t
K
L
De
e
m
e
d
to
b
e
Un
i
v
e
rsity
,
Va
d
d
e
s
wa
ra
m
,
Vijay
a
wa
d
a
,
An
d
h
ra
P
ra
d
e
sh
.
He
h
a
s
2
3
y
e
a
rs
o
f
a
c
a
d
e
m
ic
e
x
p
e
rien
c
e
a
n
d
h
a
s
p
u
b
li
sh
e
d
n
u
m
e
ro
u
s
re
se
a
rc
h
a
rti
c
les
in
r
e
p
u
ted
j
o
u
r
n
a
ls,
b
o
o
k
c
h
a
p
ter
s,
a
n
d
c
o
n
fe
re
n
c
e
p
r
o
c
e
e
d
in
g
s
.
His
a
re
a
s
o
f
in
tere
st
in
c
l
u
d
e
wire
les
s
se
n
so
r
n
e
two
r
k
s,
in
ter
n
e
t
o
f
th
in
g
s,
a
rti
ficia
l
in
telli
g
e
n
c
e
,
a
n
d
m
a
c
h
in
e
lea
rn
in
g
.
H
e
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
a
k
v
e
lm
u
ru
g
a
n
7
4
@g
m
a
il
.
c
o
m
.
Evaluation Warning : The document was created with Spire.PDF for Python.