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.
16
,
No
.
1
,
Feb
r
u
ar
y
20
26
,
p
p
.
383
~
394
I
SS
N:
2
0
8
8
-
8
7
0
8
,
DOI
: 1
0
.
1
1
5
9
1
/ijece.
v
16
i
1
.
p
p
3
8
3
-
3
9
4
383
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
ec
e.
ia
esco
r
e.
co
m
Im
a
g
e clas
sifica
tion usin
g
t
wo
ne
ur
a
l net
wo
rks a
nd
activ
a
tion
funct
io
ns:
a ca
se
study
on fish
spec
ies
O
pp
ir
H
uta
pea
1
,
F
o
rd
L
um
ba
n G
a
o
l
2
,
T
o
k
uro
M
a
t
s
uo
3
1
S
o
f
t
w
a
r
e
E
n
g
i
n
e
e
r
i
n
g
Te
c
h
n
o
l
o
g
y
,
F
a
c
u
l
t
y
o
f
V
o
c
a
t
i
o
n
a
l
S
t
u
d
i
e
s,
I
n
st
i
t
u
t
Te
k
n
o
l
o
g
i
D
e
l
-
S
i
t
o
l
u
a
ma,
T
o
b
a
S
a
m
o
si
r
,
I
n
d
o
n
e
si
a
2
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
D
e
p
a
r
t
me
n
t
,
B
I
N
U
S
G
r
a
d
u
a
t
e
P
r
o
g
r
a
m
-
D
o
c
t
o
r
o
f
C
o
mp
u
t
e
r
S
c
i
e
n
c
e
,
B
i
n
a
N
u
sa
n
t
a
r
a
U
n
i
v
e
r
si
t
y
,
Jak
a
r
t
a
,
I
n
d
o
n
e
si
a
3
A
d
v
a
n
c
e
d
I
n
st
i
t
u
t
e
o
f
I
n
d
u
st
r
i
a
l
Te
c
h
n
o
l
o
g
y
,
To
k
y
o
,
J
a
p
a
n
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
J
an
2
3
,
2
0
2
5
R
ev
is
ed
Sep
1
8
,
2
0
2
5
Acc
ep
ted
No
v
2
3
,
2
0
2
5
Lak
e
To
b
a
is
u
ti
li
z
e
d
f
o
r
a
q
u
a
c
u
l
tu
re
fish
i
n
g
a
s
a
c
lea
r
e
x
a
m
p
le
o
f
h
o
w
th
is
tec
h
n
o
l
o
g
y
c
a
n
b
e
a
p
p
l
ied
.
On
e
o
f
th
e
sp
e
c
ies
p
re
se
n
ts
is
th
e
re
d
d
e
v
il
fish
(
Amp
h
il
o
p
h
u
s
l
a
b
i
a
tu
s
)
,
wh
ich
is k
n
o
w
n
to
h
a
v
e
sta
rted
a
p
p
e
a
rin
g
i
n
th
e
las
t
1
0
y
e
a
rs.
Th
is
sp
e
c
ies
is
k
n
o
w
n
to
b
e
v
e
ry
a
g
g
re
ss
iv
e
a
n
d
d
a
m
a
g
e
th
e
e
c
o
sy
ste
m
.
Wh
e
n
th
e
ir
p
o
p
u
lati
o
n
s
g
o
u
n
c
h
e
c
k
e
d
,
re
d
-
d
e
v
il
s
c
a
n
c
a
u
se
a
d
e
c
li
n
e
in
l
o
c
a
l
fish
p
o
p
u
latio
n
s,
p
o
ten
t
ially
d
e
str
o
y
i
n
g
t
h
e
b
a
lan
c
e
o
f
th
e
fo
o
d
c
h
a
in
in
t
h
o
se
wa
ters
.
Th
is
stu
d
y
u
se
d
a
rti
f
icia
l
n
e
u
ra
l
n
e
two
rk
(AN
N)
a
n
d
c
o
n
v
o
l
u
ti
o
n
a
l
n
e
u
ra
l
n
e
two
r
k
(CNN
)
a
l
g
o
rit
h
m
s
to
su
c
c
e
ss
fu
ll
y
c
re
a
te
two
c
las
sifica
ti
o
n
m
o
d
e
ls
f
o
r
fis
h
sp
e
c
ies
fr
o
m
Lak
e
To
b
a
:
re
d
d
e
v
il
fis
h
(
Amp
h
il
o
p
h
u
s l
a
b
ia
t
u
s
),
m
u
ja
h
ir
f
ish
(
Or
e
o
c
h
ro
mis mo
ss
a
m
b
icu
s
),
se
p
a
t
fish
(
T
ric
h
o
g
a
ste
r
trich
o
p
ter
u
s
)
.
Th
e
p
u
rp
o
se
o
f
th
is
m
o
d
e
l
is
to
a
u
t
o
m
a
ti
c
a
ll
y
id
e
n
ti
f
y
f
ish
sp
e
c
ies
b
y
u
sin
g
ima
g
e
-
b
a
se
d
c
las
sifica
ti
o
n
t
e
c
h
n
iq
u
e
s.
Ac
c
o
rd
in
g
t
o
th
e
st
u
d
y
'
s
fin
d
i
n
g
s,
b
o
th
m
o
d
e
ls
p
e
rf
o
rm
e
d
e
x
c
e
p
t
io
n
a
ll
y
we
ll
a
n
d
h
a
d
a
h
ig
h
d
e
g
re
e
o
f
a
c
c
u
ra
c
y
.
Th
is
stu
d
y
a
d
d
re
ss
e
s
th
e
lac
k
o
f
e
ffe
c
ti
v
e
a
u
to
m
a
ted
fish
c
las
sifica
ti
o
n
s
y
ste
m
s
fo
r
e
c
o
sy
ste
m
s
li
k
e
Lak
e
To
b
a
,
In
d
o
n
e
sia
,
wh
ic
h
a
re
th
re
a
ten
e
d
b
y
i
n
v
a
si
v
e
sp
e
c
ies
su
c
h
a
s
th
e
re
d
d
e
v
il
fish
.
By
c
o
m
p
a
ri
n
g
CNN
a
n
d
AN
N
m
o
d
e
ls
with
d
iffere
n
t
a
c
ti
v
a
ti
o
n
fu
n
c
ti
o
n
s
a
n
d
o
p
ti
m
ize
rs,
we
fo
u
n
d
t
h
a
t
CNN
with
re
c
ti
fie
d
l
in
e
a
r
u
n
i
t
(
Re
LU
)
a
c
ti
v
a
ti
o
n
a
n
d
Ad
a
m
o
p
ti
m
ize
r
p
r
o
v
id
e
s
t
h
e
m
o
st
a
c
c
u
ra
te
a
n
d
sta
b
le
re
su
lt
s.
Th
e
fi
n
d
in
g
s
o
ffe
r
p
ra
c
ti
c
a
l
imp
li
c
a
ti
o
n
s
f
o
r
fish
e
ries
m
a
n
a
g
e
m
e
n
t
a
n
d
b
io
d
iv
e
rsit
y
c
o
n
se
rv
a
ti
o
n
.
K
ey
w
o
r
d
s
:
Activ
atio
n
f
u
n
ctio
n
Data
co
llectio
n
Featu
r
e
ex
tr
ac
tio
n
I
m
ag
e
d
etec
tio
n
Ma
ch
in
e
lear
n
in
g
Neu
r
al
n
etwo
r
k
Pre
p
r
o
ce
s
s
in
g
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
:
Op
p
ir
Hu
tap
ea
So
f
twar
e
E
n
g
in
ee
r
T
ec
h
n
o
lo
g
y
,
Facu
lty
o
f
Vo
ca
tio
n
al
Stu
d
i
es,
I
n
s
titu
t T
ek
n
o
lo
g
i D
el
-
Sit
o
lu
am
a
Sit
o
lu
am
a,
B
alig
e,
T
o
b
a,
No
r
th
Su
m
atr
a
2
2
3
8
1
,
I
n
d
o
n
esia
E
m
ail: o
p
p
ir
.
h
u
tap
ea
@
d
el.
ac
.
i
d
/o
p
p
ir
h
u
tap
ea
2
0
@
g
m
ail.
co
m
1.
I
NT
RO
D
UCT
I
O
N
I
n
to
d
a
y
'
s
d
ig
ital
er
a,
th
e
ap
p
licatio
n
o
f
m
ac
h
in
e
lea
r
n
in
g
(
ML
)
b
ased
tec
h
n
o
l
o
g
y
is
in
cr
ea
s
in
g
ly
wid
esp
r
ea
d
in
v
a
r
io
u
s
s
ec
to
r
s
,
r
an
g
in
g
f
r
o
m
h
ea
lth
,
tr
an
s
p
o
r
tatio
n
,
to
th
e
en
v
ir
o
n
m
e
n
t.
O
n
e
o
f
th
e
s
ec
to
r
s
th
at
is
s
tar
tin
g
to
ex
p
lo
r
e
th
e
g
r
ea
t
p
o
ten
tial
o
f
th
is
tech
n
o
l
o
g
y
is
f
is
h
er
ies,
s
p
ec
if
ically
in
ad
d
r
ess
in
g
th
e
ch
allen
g
es
r
elate
d
to
th
e
id
en
tific
atio
n
an
d
class
if
icatio
n
o
f
f
is
h
s
p
ec
ies
th
er
e
ar
e
m
an
y
s
tu
d
ies
co
n
d
u
cted
u
s
in
g
d
if
f
er
e
n
t
m
eth
o
d
s
[
1
]
,
[
2
]
.
I
n
a
d
d
itio
n
,
m
an
y
tech
n
o
l
o
g
ies
h
av
e
b
ee
n
b
u
ilt
in
th
is
co
n
tex
t
to
d
e
v
elo
p
ad
v
an
ce
d
s
y
s
tem
s
f
o
r
u
n
d
er
w
ater
f
is
h
d
etec
tio
n
an
d
r
ec
o
g
n
itio
n
.
Fis
h
ar
e
an
im
p
o
r
tan
t
p
ar
t
o
f
th
e
m
a
r
in
e
ec
o
s
y
s
tem
as
well
as
h
u
m
an
cu
ltu
r
e
an
d
in
d
u
s
tr
y
[
3
]
.
L
ak
e
T
o
b
a
is
o
n
e
o
f
t
h
e
f
r
esh
wa
ter
wate
r
s
th
at
h
as
n
atu
r
al
r
eso
u
r
ce
s
th
at
ar
e
v
er
y
im
p
o
r
tan
t
f
o
r
p
eo
p
le'
s
liv
es,
e
s
p
ec
ially
th
e
f
is
h
er
m
en
wh
o
liv
e
ar
o
u
n
d
it.
L
ak
e
T
o
b
a
as
o
n
e
o
f
th
e
lar
g
est
an
d
r
ich
est
aq
u
atic
ec
o
s
y
s
tem
s
in
I
n
d
o
n
esia,
L
ak
e
T
o
b
a
is
u
tili
ze
d
f
o
r
a
q
u
ac
u
ltu
r
e
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.
16
,
No
.
1
,
Feb
r
u
ar
y
20
26
:
3
8
3
-
394
384
f
is
h
in
g
b
ein
g
a
clea
r
ex
am
p
le
o
f
h
o
w
th
is
tech
n
o
l
o
g
y
ca
n
b
e
ap
p
lied
[
4
]
.
T
h
e
lak
e
h
as
v
e
r
y
r
ich
b
io
d
iv
e
r
s
ity
,
in
clu
d
in
g
l
o
ca
l
f
is
h
s
p
ec
ies
s
u
ch
as
th
e
d
o
m
in
an
t
r
ed
d
ev
il
[
5
]
.
I
t
co
n
tain
s
a
v
ar
iety
o
f
f
is
h
s
p
ec
ies
th
at
liv
e
an
d
b
r
ee
d
.
On
e
o
f
th
e
s
p
ec
ies
p
r
esen
ts
is
th
e
r
ed
d
ev
il
f
is
h
(
A
mp
h
ilo
p
h
u
s
la
b
ia
t
u
s
)
,
wh
ic
h
is
k
n
o
wn
to
h
a
v
e
s
tar
ted
ap
p
ea
r
in
g
in
th
e
last
1
0
y
ea
r
s
.
T
h
e
p
r
esen
ce
o
f
in
v
a
s
iv
e
s
p
ec
ies
s
u
ch
as
r
ed
d
ev
il
f
is
h
p
o
s
es
a
s
er
io
u
s
th
r
ea
t
to
th
e
b
alan
ce
o
f
th
e
aq
u
atic
ec
o
s
y
s
tem
.
T
h
is
s
p
ec
ies
is
k
n
o
wn
to
b
e
v
er
y
ag
g
r
ess
iv
e
an
d
ca
n
d
a
m
ag
e
th
e
ec
o
s
y
s
tem
.
I
ts
a
g
g
r
ess
iv
e
ch
ar
ac
ter
is
tics
n
o
t
o
n
ly
d
is
r
u
p
t
in
ter
ac
tio
n
s
b
etwe
en
f
is
h
s
p
ec
ies,
b
u
t
also
th
r
ea
ten
th
e
ex
is
ten
ce
o
f
lo
ca
l
f
is
h
es
th
at
p
lay
an
im
p
o
r
tan
t
r
o
le
in
m
ain
tain
in
g
ec
o
s
y
s
tem
b
alan
ce
.
T
h
er
ef
o
r
e,
p
r
o
p
er
m
an
a
g
em
en
t is n
ee
d
e
d
to
co
n
tr
o
l t
h
e
r
ed
d
ev
il
f
is
h
p
o
p
u
latio
n
.
On
e
o
f
th
e
m
ai
n
r
ea
s
o
n
s
f
o
r
th
e
n
ee
d
to
d
etec
t
r
ed
d
e
v
il
f
is
h
is
its
h
ig
h
ad
ap
tab
ilit
y
.
T
h
is
s
p
ec
ies
ca
n
q
u
ick
ly
ad
j
u
s
t
to
a
wid
e
r
an
g
e
o
f
aq
u
atic
en
v
ir
o
n
m
en
tal
co
n
d
itio
n
s
.
I
n
ad
d
itio
n
,
r
ed
d
ev
il
s
ar
e
v
o
r
ac
io
u
s
p
r
ed
ato
r
s
o
f
s
m
all
f
is
h
,
wh
ich
ar
e
o
f
ten
en
d
em
ic
o
r
n
at
iv
e
to
a
r
eg
io
n
[
6
]
.
W
h
en
t
h
eir
p
o
p
u
latio
n
s
g
o
u
n
ch
ec
k
e
d
,
r
ed
d
e
v
il
s
ca
n
ca
u
s
e
a
d
ec
lin
e
in
lo
ca
l
f
is
h
p
o
p
u
latio
n
s
,
p
o
ten
tially
d
estro
y
in
g
th
e
b
alan
ce
o
f
th
e
f
o
o
d
ch
ain
in
th
o
s
e
wate
r
s
.
I
n
ad
d
itio
n
to
its
ag
g
r
ess
iv
e
n
at
u
r
e
an
d
r
a
p
id
ad
ap
tatio
n
,
th
e
r
ed
d
e
v
il
also
h
as
a
v
er
y
h
i
g
h
r
e
p
r
o
d
u
ctiv
e
r
ate
[
7
]
.
T
h
is
r
ap
id
b
r
ee
d
in
g
lead
s
to
p
o
p
u
latio
n
in
cr
ea
s
es
th
at
ar
e
d
if
f
icu
lt
to
co
n
tr
o
l
in
a
s
h
o
r
t
p
er
io
d
o
f
tim
e.
I
ts
m
ain
b
r
ee
d
in
g
ar
ea
s
in
clu
d
e
a
r
o
u
n
d
L
ak
e
T
o
b
a,
wh
ich
is
o
n
e
o
f
th
e
lar
g
est
an
d
m
o
s
t
im
p
o
r
tan
t
f
r
esh
wate
r
ec
o
s
y
s
tem
s
in
I
n
d
o
n
esia.
T
h
e
a
q
u
atic
ec
o
s
y
s
tem
o
f
L
ak
e
T
o
b
a
its
elf
h
as
b
ee
n
d
is
r
u
p
ted
b
y
th
e
d
o
m
i
n
an
ce
o
f
th
is
s
p
ec
ies,
g
i
v
en
t
h
at
L
a
k
e
T
o
b
a
is
o
n
e
o
f
th
e
v
ital f
r
esh
w
ater
s
o
u
r
ce
s
f
o
r
th
e
s
u
r
r
o
u
n
d
in
g
r
eg
io
n
[
8
]
.
R
ed
d
ev
il
f
is
h
(
A
mp
h
ilo
p
h
u
s
la
b
ia
tu
s
)
is
ch
ar
ac
ter
ized
b
y
a
s
len
d
er
b
o
d
y
th
at
is
s
im
ilar
to
tilap
ia,
b
u
t
is
d
is
tin
g
u
is
h
ed
b
y
h
ar
d
,
leath
er
y
s
ca
les
an
d
p
o
in
ted
an
al
an
d
d
o
r
s
al
f
in
s
.
As
an
in
v
asiv
e
s
p
ec
ies,
it
h
as
h
ig
h
ad
ap
ta
b
ilit
y
,
f
ast
g
r
o
wth
,
an
d
ea
s
y
r
ep
r
o
d
u
cti
o
n
.
R
ed
d
ev
il
f
is
h
ten
d
to
b
e
o
m
n
iv
o
r
o
u
s
with
p
lan
k
to
n
,
esp
ec
ially
f
r
o
m
th
e
C
h
lo
r
o
p
h
y
ce
ae
class
,
b
ei
n
g
th
eir
m
a
in
f
o
o
d
,
an
d
s
h
o
w
f
lex
ib
ilit
y
in
u
tili
zin
g
f
o
o
d
r
eso
u
r
ce
s
[
9
]
.
T
h
e
b
o
d
y
s
ize
o
f
f
is
h
f
o
u
n
d
in
J
atib
ar
an
g
R
eser
v
o
ir
r
an
g
ed
f
r
o
m
8
-
1
8
.
5
c
m
,
with
an
a
v
er
ag
e
weig
h
t
o
f
4
7
.
8
g
r
am
s
T
h
e
tilap
ia
(
Oreo
ch
r
o
mis
mo
s
s
a
mb
icu
s
)
is
an
im
p
o
r
tan
t
f
r
esh
wa
ter
f
is
h
s
p
ec
ies
in
aq
u
ac
u
ltu
r
e.
T
h
is
s
p
ec
ies
h
as
a
h
ig
h
to
ler
an
ce
to
v
a
r
io
u
s
en
v
ir
o
n
m
e
n
tal
co
n
d
itio
n
s
,
f
ast
g
r
o
wth
,
an
d
is
ea
s
ily
s
p
awn
ed
,
m
ak
in
g
it
a
m
ajo
r
c
o
m
m
o
d
ity
in
v
a
r
io
u
s
r
eg
i
o
n
s
,
in
clu
d
in
g
Me
r
au
k
e
[
1
0
]
.
T
ila
p
ia
is
k
n
o
wn
as
an
in
v
asiv
e
s
p
ec
ies
th
at
is
ab
le
to
liv
e
to
g
eth
e
r
with
tilap
ia
(
Or
eo
ch
r
o
mis
n
ilo
ticu
s
)
,
o
f
te
n
ca
u
s
in
g
h
y
b
r
id
izatio
n
th
at
d
ec
r
ea
s
es
g
e
n
etic
d
iv
e
r
s
ity
.
I
n
m
o
lecu
lar
s
tu
d
ies,
tilap
ia
s
h
o
wed
lo
w
lev
els
o
f
g
e
n
etic
d
is
tan
ce
f
r
o
m
s
im
ilar
s
p
ec
ies,
w
ith
o
n
ly
o
n
e
h
ap
lo
ty
p
e
id
e
n
tifie
d
.
T
h
is
f
is
h
h
as
a
b
o
d
y
s
ize
o
f
2
0
-
4
0
cm
an
d
is
o
f
ten
f
o
u
n
d
in
ab
u
n
d
an
ce
in
n
atu
r
e,
m
ak
i
n
g
it
a
n
af
f
o
r
d
ab
le
m
ai
n
s
o
u
r
ce
o
f
p
r
o
tein
f
o
r
t
h
e
co
m
m
u
n
ity
T
h
e
s
ep
at
f
is
h
(
Tr
ich
o
g
a
s
ter
tr
ich
o
p
teru
s
)
,
a
s
p
ec
ies
o
f
th
e
f
am
ily
A
n
a
b
a
n
tid
a
e
,
h
as
m
o
r
p
h
o
lo
g
ical
c
h
ar
ac
ter
is
tics
th
at
ca
n
ad
ap
t
to
its
h
a
b
itat.
I
t o
r
ig
in
at
ed
f
r
o
m
So
u
t
h
ea
s
t
Asi
an
wate
r
s
an
d
was
in
tr
o
d
u
ce
d
to
I
n
d
o
n
esia
in
1
9
3
4
.
I
t
ca
n
liv
e
in
le
n
tic
(
s
u
ch
as
r
eser
v
o
ir
s
)
a
n
d
l
o
tic
(
s
u
ch
as
r
i
v
er
s
)
wate
r
s
[
1
1
]
.
I
t
i
s
b
o
d
y
m
o
r
p
h
o
lo
g
y
d
if
f
er
s
ac
co
r
d
in
g
t
o
h
ab
itat:
f
is
h
f
r
o
m
r
iv
er
s
h
av
e
m
o
r
e
elo
n
g
ated
b
o
d
ies
an
d
s
tr
aig
h
t
b
ac
k
s
to
ad
ap
t
to
f
ast
cu
r
r
en
ts
,
wh
ile
f
is
h
f
r
o
m
r
eser
v
o
ir
s
h
av
e
cu
r
v
e
d
b
ac
k
s
to
ad
a
p
t
to
s
till
wate
r
.
I
n
ad
d
itio
n
,
s
ep
at
f
is
h
h
av
e
th
e
ad
v
a
n
tag
e
o
f
m
o
d
if
ied
wh
ip
-
lik
e
p
ec
to
r
al
f
in
s
f
o
r
to
u
ch
,
w
h
ich
h
elp
to
s
u
r
v
iv
e
in
ex
tr
em
e
e
n
v
ir
o
n
m
en
ts
.
T
o
ad
d
r
ess
th
ese
ch
allen
g
es,
th
e
d
ev
elo
p
m
en
t
o
f
a
m
ac
h
in
e
lear
n
in
g
-
b
ased
s
y
s
tem
is
a
p
r
o
m
is
in
g
s
o
lu
tio
n
.
Usi
n
g
th
is
tech
n
o
lo
g
y
,
th
e
s
y
s
tem
ca
n
b
e
tr
ain
ed
t
o
r
ec
o
g
n
ize
th
e
u
n
i
q
u
e
tr
aits
o
f
ea
ch
s
p
ec
ies,
th
u
s
b
ein
g
ab
le
to
ac
cu
r
ately
d
is
tin
g
u
is
h
r
ed
d
ev
il
f
r
o
m
m
u
jah
ir
a
n
d
s
ep
at
.
T
h
is
s
y
s
tem
n
o
t
o
n
ly
h
elp
s
in
d
etec
tin
g
th
e
p
r
esen
ce
o
f
in
v
asiv
e
s
p
ec
ies,
b
u
t
also
s
u
p
p
o
r
ts
t
h
e
co
n
s
er
v
atio
n
o
f
lo
ca
l
f
is
h
th
at
ar
e
im
p
o
r
ta
n
t
to
th
e
ec
o
s
y
s
tem
.
T
h
r
o
u
g
h
th
e
a
p
p
li
ca
tio
n
o
f
m
ac
h
in
e
lear
n
in
g
,
f
i
s
h
s
p
ec
ies
id
en
tific
atio
n
ca
n
b
e
d
o
n
e
ef
f
icien
tly
,
b
o
th
o
n
a
r
esear
ch
a
n
d
wate
r
m
an
a
g
em
en
t
s
ca
le.
T
h
is
te
ch
n
o
lo
g
y
is
e
x
p
ec
ted
to
b
e
an
ef
f
ec
tiv
e
to
o
l
in
en
v
ir
o
n
m
en
tal
co
n
s
er
v
atio
n
ef
f
o
r
ts
,
wh
ile
s
u
p
p
o
r
tin
g
th
e
s
u
s
tain
ab
ilit
y
o
f
f
r
esh
wate
r
ec
o
s
y
s
tem
s
in
th
e
f
u
tu
r
e.
Fu
r
th
e
r
m
o
r
e
,
with
th
e
ab
ilit
y
to
q
u
ick
ly
an
d
ac
cu
r
ately
id
en
tify
r
ed
d
ev
il
s
,
th
e
s
y
s
tem
ca
n
also
h
elp
cu
ll
th
e
i
n
v
asiv
e
f
is
h
p
o
p
u
lati
o
n
in
a
tar
g
eted
an
d
s
y
s
tem
at
ic
m
an
n
e
r
[
1
2
]
,
[
1
3
]
.
B
y
u
tili
zin
g
tech
n
o
lo
g
y
t
o
tr
ac
k
an
d
c
o
n
tr
o
l
th
e
s
p
r
ea
d
o
f
r
ed
d
ev
il
s
,
we
ca
n
r
ed
u
ce
th
eir
n
eg
ativ
e
im
p
ac
t
o
n
lo
c
al
f
is
h
.
T
h
e
u
s
e
o
f
m
eth
o
d
s
s
u
ch
as
m
ass
ca
p
tu
r
e
o
r
f
o
cu
s
ed
cu
llin
g
o
f
r
ed
d
ev
il
p
o
p
u
latio
n
s
ca
n
b
e
d
o
n
e
m
o
r
e
ef
f
icien
tly
,
s
ig
n
if
ican
tly
r
ed
u
cin
g
th
e
n
u
m
b
er
o
f
in
v
asiv
e
f
is
h
,
an
d
p
r
o
v
id
in
g
s
p
ac
e
f
o
r
lo
ca
l
f
is
h
to
r
ep
r
o
d
u
ce
.
W
ith
th
is
ap
p
r
o
ac
h
,
it is
h
o
p
ed
th
at
L
ak
e
T
o
b
a
ca
n
b
e
r
esto
r
ed
,
m
ain
t
ain
ec
o
s
y
s
tem
b
alan
ce
,
an
d
s
u
p
p
o
r
t th
e
s
u
r
v
i
v
al
o
f
n
ativ
e
f
is
h
s
p
ec
ies th
at
ar
e
an
i
m
p
o
r
tan
t
r
eso
u
r
ce
f
o
r
th
e
s
u
r
r
o
u
n
d
in
g
co
m
m
u
n
ity
.
Ap
p
licatio
n
o
f
c
o
n
v
o
lu
tio
n
al
n
eu
r
al
n
etwo
r
k
(
C
NN)
alg
o
r
ith
m
f
o
r
au
to
m
atic
id
en
tif
icatio
n
o
f
v
ar
io
u
s
f
is
h
s
p
ec
ies.
C
NN
is
ef
f
ec
tiv
e
in
im
ag
e
p
r
o
ce
s
s
in
g
,
with
a
s
tr
u
ctu
r
e
co
n
s
is
tin
g
o
f
m
u
ltip
le
co
n
v
o
l
u
tio
n
a
n
d
p
o
o
lin
g
lay
er
s
th
at
ex
tr
ac
t
f
ea
t
u
r
es
f
r
o
m
f
is
h
im
ag
es.
Pre
p
r
o
ce
s
s
in
g
,
in
clu
d
in
g
im
a
g
e
r
esizin
g
an
d
n
o
r
m
aliza
tio
n
,
is
p
er
f
o
r
m
ed
p
r
io
r
t
o
m
o
d
el
tr
ai
n
in
g
.
T
h
e
d
ataset
u
s
ed
c
o
n
s
is
ted
o
f
liv
e
ca
p
tu
r
ed
f
is
h
im
ag
es
[
1
4
]
,
[
1
5
]
.
T
h
e
tr
a
in
in
g
r
esu
lts
s
h
o
wed
an
ac
cu
r
ac
y
o
f
8
5
.
1
8
%,
in
d
icatin
g
th
at
th
e
C
NN
was
ab
le
to
class
if
y
f
is
h
well,
d
esp
ite
t
h
e
ch
allen
g
e
o
f
d
is
tin
g
u
is
h
in
g
s
im
ilar
s
p
ec
ies.
T
h
is
r
esear
c
h
co
n
t
r
ib
u
tes
to
t
h
e
d
ev
elo
p
m
e
n
t
o
f
an
a
u
to
m
atic
id
en
tific
atio
n
s
y
s
tem
th
at
is
b
en
ef
icial
to
r
esear
ch
e
r
s
an
d
o
b
s
er
v
er
s
in
th
e
f
ield
o
f
aq
u
atic
b
io
lo
g
y
[
1
6
]
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8
7
0
8
I
ma
g
e
cla
s
s
ifica
tio
n
u
s
in
g
tw
o
n
eu
r
a
l n
etw
o
r
ks a
n
d
a
ctiva
ti
o
n
…
(
Op
p
ir
Hu
ta
p
ea
)
385
B
ased
o
n
th
e
f
in
d
in
g
s
an
d
a
p
p
r
o
ac
h
es
p
r
o
p
o
s
ed
b
y
two
p
r
ev
io
u
s
r
esear
ch
e
r
s
,
we
d
ec
id
ed
to
u
s
e
co
n
v
o
l
u
tio
n
al
n
eu
r
al
n
etwo
r
k
(
C
NN)
an
d
ar
tific
ial
n
eu
r
al
n
e
two
r
k
(
ANN)
al
g
o
r
ith
m
s
as
th
e
m
ain
m
eth
o
d
s
i
n
class
if
y
in
g
f
is
h
s
p
ec
ies
f
o
u
n
d
in
L
ak
e
T
o
b
a
[
1
7
]
,
[
1
8
]
.
T
h
e
C
NN
alg
o
r
ith
m
was
ch
o
s
en
b
ec
au
s
e
o
f
its
s
u
p
er
io
r
ab
ilit
y
to
r
ec
o
g
n
ize
v
is
u
al
p
atter
n
s
f
r
o
m
im
ag
es,
s
u
ch
as
s
h
ap
e,
tex
tu
r
e
,
an
d
f
i
s
h
-
s
p
ec
if
ic
f
ea
tu
r
es,
wh
ile
ANN
is
u
s
ed
to
p
r
o
ce
s
s
ad
d
itio
n
al
d
ata
o
r
o
th
e
r
n
u
m
er
ical
f
ea
tu
r
es
th
at
s
u
p
p
o
r
t
th
e
class
if
icatio
n
p
r
o
ce
s
s
.
T
h
e
co
m
b
in
atio
n
o
f
t
h
ese
two
alg
o
r
ith
m
s
is
ex
p
ec
t
ed
to
p
r
o
v
id
e
m
o
r
e
ac
cu
r
ate
a
n
d
r
eliab
le
r
esu
lts
in
id
en
tify
in
g
f
is
h
s
p
ec
ies in
th
e
r
eg
io
n
.
T
h
is
s
tu
d
y
p
r
esen
ts
a
n
o
v
el
c
o
m
b
in
atio
n
o
f
C
NN
an
d
ANN
f
o
r
m
u
lti
-
class
f
is
h
s
p
ec
ies
clas
s
if
icatio
n
u
s
in
g
im
ag
e
d
ata.
Un
lik
e
m
o
s
t
ex
is
tin
g
ap
p
r
o
ac
h
es
th
at
r
ely
o
n
a
s
in
g
le
a
r
ch
itectu
r
e,
we
ev
alu
ate
th
e
co
m
p
ar
ativ
e
p
er
f
o
r
m
a
n
ce
o
f
b
o
th
m
o
d
els
u
s
in
g
two
ac
tiv
a
tio
n
f
u
n
ctio
n
s
(
r
ec
tifie
d
lin
ea
r
u
n
it
(
R
eL
U)
an
d
T
an
h
)
,
o
p
tim
izer
s
(
Ad
am
an
d
s
to
ch
asti
c
g
r
ad
ien
t
d
escen
t
(
SGD)
)
,
an
d
m
u
ltip
le
lear
n
in
g
r
a
tes.
T
h
is
in
teg
r
ated
ap
p
r
o
ac
h
allo
ws
a
d
ee
p
er
u
n
d
er
s
tan
d
in
g
o
f
h
o
w
ac
tiv
atio
n
an
d
o
p
tim
izatio
n
ch
o
ices
af
f
ec
t
class
if
icatio
n
o
u
tco
m
es f
o
r
r
ea
l
-
wo
r
ld
aq
u
atic
d
atasets
.
2.
M
E
T
H
O
D
2
.
1
.
Resea
rc
h dia
g
ra
m
T
h
e
d
ata
co
llectio
n
p
r
o
ce
s
s
c
ar
r
ied
o
u
t
b
y
r
esear
ch
er
s
was
ca
r
r
ied
o
u
t
m
a
n
u
ally
,
b
y
tak
in
g
im
ag
e
s
am
p
les
d
ir
ec
tly
in
th
e
ar
ea
a
r
o
u
n
d
L
ak
e
T
o
b
a.
T
h
e
s
p
ec
if
i
c
lo
ca
tio
n
o
f
d
ata
co
llectio
n
was
f
o
cu
s
ed
o
n
th
e
B
alig
e
ar
ea
,
wh
ich
is
o
n
e
o
f
th
e
s
tr
ateg
ic
ar
ea
s
a
r
o
u
n
d
L
a
k
e
T
o
b
a.
T
h
e
d
ata
co
llectio
n
in
v
o
lv
ed
ca
p
tu
r
in
g
im
ag
es
o
f
f
is
h
at
v
ar
io
u
s
an
g
l
es
an
d
lig
h
tin
g
co
n
d
itio
n
s
to
en
s
u
r
e
d
ata
d
iv
er
s
ity
th
at
co
u
l
d
s
u
p
p
o
r
t
ac
c
u
r
ac
y
in
th
e
p
r
o
ce
s
s
o
f
a
n
aly
zin
g
an
d
class
if
y
in
g
f
is
h
s
p
ec
ies.
T
h
is
m
an
u
al
ap
p
r
o
ac
h
allo
ws
r
esear
ch
er
s
to
d
ir
ec
tl
y
en
s
u
r
e
th
e
q
u
ality
o
f
t
h
e
d
at
a
an
d
its
r
elev
an
ce
t
o
th
e
r
es
ea
r
ch
n
ee
d
s
.
T
h
e
co
m
p
lete
f
l
o
w
o
f
b
u
ild
in
g
t
h
e
m
ac
h
in
e
lear
n
in
g
m
o
d
el
h
as b
ee
n
s
u
m
m
ar
ized
i
n
a
r
esear
ch
d
iag
r
am
.
A
r
esear
ch
d
iag
r
am
is
a
f
o
r
m
o
f
v
is
u
aliza
tio
n
th
at
d
escr
ib
es
th
e
s
y
s
tem
atic
f
l
o
w
o
f
r
esear
ch
ca
r
r
ie
d
o
u
t
with
t
h
e
aim
o
f
m
ak
i
n
g
it e
asier
to
u
n
d
er
s
tan
d
th
e
f
lo
w
o
f
wo
r
k
.
T
h
e
d
iag
r
a
m
will
d
es
cr
ib
e
th
e
m
ain
s
tep
s
o
f
th
e
r
esear
c
h
p
r
esen
ted
in
a
s
tr
u
ctu
r
ed
m
a
n
n
er
s
tar
tin
g
f
r
o
m
p
r
o
b
lem
id
e
n
tific
atio
n
to
co
n
clu
s
io
n
.
T
h
e
u
s
e
o
f
th
is
d
iag
r
am
will
h
elp
to
e
n
s
u
r
e
th
e
r
esear
ch
r
u
n
s
an
d
c
o
n
v
ey
b
r
o
ad
er
i
n
f
o
r
m
atio
n
.
T
h
e
r
esear
ch
d
iag
r
am
ca
n
b
e
s
ee
n
in
Fig
u
r
e
1
.
Fig
u
r
e
1
.
R
esear
ch
f
lo
w
2
.
2
.
Da
t
a
c
o
llect
io
n
Af
ter
co
n
d
u
ctin
g
a
liter
atu
r
e
s
tu
d
y
,
th
e
n
e
x
t
s
tep
is
to
co
llec
t
d
ata
m
an
u
ally
.
T
h
e
d
ata
was
co
llected
m
an
u
ally
b
y
p
h
o
to
g
r
ap
h
in
g
f
r
o
m
v
ar
io
u
s
an
g
les
s
u
ch
as
th
e
s
id
e,
to
p
,
f
r
o
n
t,
a
n
d
also
th
e
b
ac
k
f
r
o
m
d
if
f
er
e
n
t
an
g
les
[
1
9
]
.
T
h
is
is
d
o
n
e
to
f
u
r
th
er
en
r
ich
th
e
v
ar
iety
o
f
d
a
ta.
Af
ter
tak
i
n
g
p
ictu
r
es
o
f
th
e
f
is
h
m
a
n
u
ally
,
th
e
o
r
ig
in
al
d
ata
was
co
llected
as
m
an
y
as
6
5
1
im
ag
es.
T
h
e
f
o
l
lo
win
g
is
s
am
p
le
d
ata
f
r
o
m
t
h
e
d
ata
th
at
will
b
e
u
s
ed
as a
d
ataset.
2
.
3
.
P
re
pro
ce
s
s
ing
da
t
a
Af
t
er
co
ll
ec
ti
n
g
th
e
im
ag
e
s
m
an
u
a
lly
,
th
e
im
ag
e
s
ar
e
en
ter
e
d
in
to
th
e
R
o
b
o
f
lo
w
f
o
r
th
e
p
r
o
ce
s
s
in
g
s
t
ag
e
.
T
h
en
t
h
e
f
i
r
s
t
th
in
g
t
o
d
o
i
s
to
lab
el
e
ac
h
i
m
ag
e
wh
er
e
th
er
e
ar
e
th
r
e
e
la
b
e
l
s
,
n
a
m
e
ly
r
ed
d
ev
il
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.
16
,
No
.
1
,
Feb
r
u
ar
y
20
26
:
3
8
3
-
394
386
(
F
ig
u
r
e
2
(
a)
)
,
m
u
jah
ir
(
F
ig
u
r
e
2
(
b
)
)
,
an
d
s
ep
a
t
(
F
ig
u
r
e
2
(
c)
)
.
Af
ter
al
l
th
e
im
ag
e
s
h
av
e
b
e
en
l
ab
e
led
,
to
f
u
r
th
e
r
in
cr
e
as
e
th
e
n
u
m
b
e
r
o
f
im
ag
e
s
,
th
e
au
g
m
en
t
at
io
n
p
r
o
ce
s
s
i
s
c
ar
r
ied
o
u
t.
T
h
e
au
g
m
en
t
at
io
n
p
r
o
ce
s
s
in
v
o
lv
e
s
r
o
ta
ti
o
n
,
f
li
p
p
in
g
,
b
r
i
g
h
tn
e
s
s
,
cr
o
p
,
an
d
zo
o
m
in
g
t
e
ch
n
iq
u
e
s
t
o
cr
ea
te
n
e
w
v
a
r
i
at
i
o
n
s
o
f
t
h
e
o
r
ig
in
a
l
d
ata
[
2
0
]
.
T
h
e
m
o
r
e
d
a
ta
u
s
ed
in
t
r
a
in
in
g
th
e
m
o
d
el,
th
e
m
o
r
e
th
e
m
o
d
el
's
ca
p
a
b
i
li
ti
e
s
wi
l
l
in
cr
ea
s
e
as
m
o
r
e
p
at
ter
n
s
o
r
i
m
ag
es
wi
ll
b
e
le
ar
n
ed
.
So
,
a
f
te
r
th
e
au
g
m
en
ta
tio
n
p
r
o
ce
s
s
,
th
e
n
u
m
b
er
o
f
i
m
ag
e
s
co
ll
ec
ted
i
s
3
,
0
6
3
i
m
ag
es
.
Pre
p
r
o
ce
s
s
in
g
s
er
v
es
to
p
r
ep
ar
e
th
e
d
atasets
r
eq
u
ir
ed
in
th
e
p
r
o
ject
in
a
s
tr
u
ct
u
r
ed
a
n
d
ef
f
i
cien
t
way
.
T
h
e
u
s
es
o
f
p
r
ep
r
o
ce
s
s
in
g
in
t
h
is
co
n
tex
t
in
clu
d
e:
Data
c
o
llectio
n
(
c
o
llectin
g
all
im
ag
e
f
il
es
f
r
o
m
th
e
d
ataset
f
o
ld
er
a
n
d
o
r
g
an
izin
g
th
em
b
y
ca
teg
o
r
y
.
T
h
is
m
ak
es
it
ea
s
y
t
o
id
en
tify
an
d
ac
ce
s
s
th
e
d
ata
r
eq
u
ir
ed
f
o
r
m
o
d
el
tr
ain
in
g
)
[
2
1
]
,
[
2
2
]
.
Data
s
et
o
r
g
an
izatio
n
(
d
iv
id
in
g
th
e
d
a
taset
in
to
th
r
ee
s
u
b
s
ets
n
am
ely
tr
ain
,
test
,
an
d
v
alid
atio
n
d
ata
wh
ich
will
ai
m
to
ac
c
u
r
ately
ev
alu
ate
th
e
p
er
f
o
r
m
an
ce
o
f
th
e
m
o
d
el)
.
T
h
is
d
iv
is
io
n
p
r
ev
e
n
ts
o
v
er
f
itti
n
g
a
n
d
en
s
u
r
es
th
e
m
o
d
el
ca
n
g
e
n
er
alize
well.
Or
g
an
ized
f
o
ld
e
r
s
tr
u
ctu
r
e
(
c
o
p
y
f
iles
in
to
s
ep
ar
ate
f
o
ld
er
s
b
ased
o
n
p
r
ed
ef
i
n
ed
s
ets.
T
h
is
ea
s
es
th
e
p
r
o
ce
s
s
o
f
d
ata
m
an
a
g
em
en
t
an
d
ac
ce
s
s
wh
en
co
n
d
u
ctin
g
tr
ain
in
g
an
d
test
in
g
)
.
(
a)
(
b
)
(
c)
Fig
u
r
e
2
.
I
m
ag
e
s
am
p
lin
g
(
a)
A
mp
h
ilo
p
h
u
s
la
b
ia
tu
s
(
r
e
d
d
e
v
il
)
,
(
b
)
Oreo
ch
r
o
mis
mo
s
s
a
m
b
icu
s
(
m
u
jah
ir
)
,
an
d
(
c)
Tr
ich
o
p
o
d
u
s
p
ec
to
r
a
lis
(
Flatfish
)
2
.
4
.
F
e
a
t
ure
ex
t
r
a
ct
io
n
At
th
is
s
tag
e,
th
e
u
n
iq
u
e
p
atte
r
n
s
o
f
ea
ch
f
is
h
s
p
ec
ies
ar
e
ex
am
in
ed
.
Featu
r
es
i
n
clu
d
e
b
o
d
y
co
lo
r
,
f
i
n
s
h
ap
e,
tex
tu
r
e
p
atter
n
s
,
an
d
o
th
er
tr
aits
.
Usi
n
g
C
NN
an
d
ANN
tech
n
iq
u
es,
all
f
ea
tu
r
es
will
b
e
ex
tr
ac
ted
to
p
r
o
v
id
e
a
m
o
r
e
ac
cu
r
ate
r
ep
r
e
s
en
tatio
n
f
o
r
u
s
e
i
n
class
if
icatio
n
[
2
3
]
,
[
2
4
]
.
Featu
r
e
e
x
tr
ac
ti
o
n
is
a
cr
u
cial
s
tag
e
in
th
e
f
is
h
class
if
icatio
n
p
r
o
ce
s
s
,
wh
er
e
th
e
d
is
tin
ctiv
e
p
h
y
s
ical
ch
ar
ac
ter
is
tics
o
f
ea
ch
s
p
ec
ies
s
u
ch
as
b
o
d
y
co
lo
r
atio
n
,
f
in
m
o
r
p
h
o
lo
g
y
,
t
ex
tu
r
e
p
atter
n
s
,
an
d
o
v
e
r
all
b
o
d
y
s
h
ap
e
ar
e
s
y
s
tem
atica
lly
an
aly
ze
d
to
ca
p
tu
r
e
m
ea
n
in
g
f
u
l
v
is
u
al
r
ep
r
esen
tati
o
n
s
.
C
NNs
p
lay
a
ce
n
tr
al
r
o
le
in
th
is
p
h
ase
d
u
e
to
th
eir
ab
ilit
y
to
au
to
m
atica
lly
lear
n
h
ier
ar
ch
ical
s
p
atial
f
ea
tu
r
es
d
ir
ec
tly
f
r
o
m
r
aw
im
ag
e
in
p
u
ts
.
T
h
r
o
u
g
h
co
n
v
o
lu
tio
n
al
f
ilter
s
,
C
NNs
ca
n
ef
f
ec
tiv
ely
d
etec
t
ed
g
es,
co
lo
r
g
r
ad
ie
n
ts
,
tex
tu
r
es,
an
d
co
m
p
lex
p
atter
n
s
th
at
d
if
f
er
e
n
tiate
o
n
e
s
p
ec
ies
f
r
o
m
an
o
th
er
,
ev
en
wh
en
v
a
r
iatio
n
s
in
lig
h
tin
g
,
o
r
ien
tatio
n
,
o
r
f
is
h
p
o
s
e
o
cc
u
r
.
T
h
is
au
to
m
at
ed
f
ea
tu
r
e
-
lear
n
in
g
ca
p
ab
ilit
y
allo
ws
C
NNs
to
g
en
er
ate
h
ig
h
ly
d
is
cr
im
in
ativ
e
a
n
d
r
o
b
u
s
t
f
ea
tu
r
e
m
ap
s
,
p
r
o
v
id
in
g
a
m
o
r
e
ac
c
u
r
ate
an
d
r
eliab
le
f
o
u
n
d
atio
n
f
o
r
th
e
class
if
icatio
n
s
tag
e.
B
y
r
ely
in
g
o
n
C
NN
-
b
ased
f
ea
t
u
r
e
e
x
t
r
ac
tio
n
,
th
e
s
y
s
tem
ca
n
ac
h
iev
e
c
o
n
s
is
ten
t p
er
f
o
r
m
an
ce
an
d
im
p
r
o
v
ed
r
ec
o
g
n
itio
n
ac
cu
r
ac
y
ac
r
o
s
s
d
iv
er
s
e
d
at
asets
[
2
4
]
.
2
.
5
.
T
ra
ini
ng
m
o
del
At
th
is
s
tag
e
th
e
m
ac
h
in
e
l
ea
r
n
in
g
m
o
d
el
is
tr
ain
e
d
u
s
i
n
g
tr
ain
in
g
d
ata
to
lear
n
p
a
tter
n
s
an
d
r
elatio
n
s
h
ip
s
b
etwe
en
in
p
u
t
a
n
d
o
u
tp
u
t
f
ea
tu
r
es.
Mo
d
elin
g
is
d
o
n
e
iter
ativ
ely
u
n
til
f
in
d
in
g
a
m
o
d
el
th
at
ca
n
wo
r
k
well
wh
er
e
it
ca
n
p
r
ed
i
ct
th
e
ty
p
e
o
f
f
is
h
,
n
am
ely
r
e
d
d
ev
il
,
m
u
jah
ir
,
a
n
d
also
s
e
p
at
.
T
h
e
r
esu
lt
is
a
m
o
d
el
th
at
ca
n
p
r
ed
ict
o
r
class
if
y
n
ew
d
ata
b
ased
o
n
lear
n
ed
p
atter
n
s
[
2
4
]
,
[
2
5
]
.
2
.
6
.
M
o
del e
v
a
lua
t
i
o
n
At
th
is
s
tag
e,
ass
ess
in
g
th
e
p
e
r
f
o
r
m
a
n
ce
o
f
th
e
tr
ain
ed
m
o
d
el
u
s
in
g
v
ali
d
atio
n
o
r
test
d
at
a.
T
h
e
aim
is
to
m
ea
s
u
r
e
h
o
w
well
th
e
m
o
d
el
wo
r
k
s
o
n
d
ata
th
at
h
a
s
n
o
t
b
ee
n
s
ee
n
b
ef
o
r
e
to
d
e
tect
is
s
u
es
s
u
ch
as
o
v
er
f
itti
n
g
o
r
u
n
d
er
f
itti
n
g
.
T
h
i
s
s
tag
e
h
elp
s
in
s
elec
tin
g
th
e
b
est r
eliab
le
m
o
d
el
f
o
r
im
p
lem
e
n
tatio
n
.
3.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
N
I
n
r
es
ea
r
ch
o
n
th
e
cl
as
s
if
ica
t
io
n
o
f
f
i
s
h
f
o
u
n
d
i
n
L
ak
e
T
o
b
a,
n
am
el
y
r
ed
d
ev
i
l
f
i
s
h
(
A
mp
h
il
o
p
h
u
s
la
b
ia
tu
s
)
,
m
u
j
ah
ir
f
i
s
h
(
O
r
e
o
ch
r
o
mi
s
mo
s
s
a
mb
i
cu
s
)
,
s
ep
at
f
i
s
h
(
T
r
ich
o
g
a
s
te
r
t
r
ich
o
p
te
r
u
s
)
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
I
ma
g
e
cla
s
s
ifica
tio
n
u
s
in
g
tw
o
n
eu
r
a
l n
etw
o
r
ks a
n
d
a
ctiva
ti
o
n
…
(
Op
p
ir
Hu
ta
p
ea
)
387
u
s
in
g
2
m
o
d
e
l
s
,
n
am
ely
co
n
v
o
lu
t
io
n
a
l
n
eu
r
al
n
e
two
r
k
an
d
ar
t
if
i
ci
al
n
eu
r
a
l
n
e
t
wo
r
k
.
W
h
er
e
al
l
d
at
a
i
s
s
t
i
ll
tak
e
n
m
an
u
a
lly
d
u
e
to
li
m
i
ted
d
a
ta
s
e
t
s
wh
i
ch
ar
e
th
en
p
r
o
c
es
s
ed
u
s
in
g
R
o
b
o
f
lo
w
s
o
th
at
th
e
am
o
u
n
t
o
f
d
at
a
u
s
ed
is
3
,
0
6
3
im
ag
e
s
.
T
o
im
p
l
em
en
t
tw
o
h
y
p
er
p
a
r
am
et
er
ac
tiv
at
io
n
f
u
n
c
tio
n
s
,
n
am
e
ly
R
e
L
U
an
d
T
an
h
[
2
6
]
,
[
2
7
]
.
Sev
e
r
a
l
ex
p
er
i
m
en
t
s
w
er
e
co
n
d
u
ct
ed
to
f
in
d
th
e
b
e
s
t
a
cc
u
r
a
cy
,
p
r
ec
i
s
io
n
,
r
e
ca
ll
an
d
F
1
-
s
co
r
e
v
a
lu
es
b
y
co
m
b
in
in
g
th
e
a
ct
iv
a
t
io
n
f
u
n
ct
io
n
a
lo
n
g
w
ith
s
ev
e
r
a
l
o
p
ti
m
iz
er
s
an
d
l
ea
r
n
in
g
r
a
te
s
.
T
h
e
ex
p
er
i
m
en
t
s
wer
e
co
n
d
u
c
ted
b
y
u
s
in
g
an
ep
o
ch
v
a
lu
e
o
f
2
5
.
T
h
en
b
y
u
s
in
g
a
le
ar
n
in
g
r
at
e
o
f
0
.
0
1
an
d
0
.
0
0
1
[
2
8
]
.
A
l
s
o
,
b
y
u
s
in
g
S
GD
o
p
t
im
iz
er
f
o
r
T
a
n
h
an
d
Ad
am
f
o
r
R
eL
U
.
I
n
th
e
ex
p
er
i
m
en
t,
th
e
ep
o
ch
v
alu
e
o
f
2
5
i
s
m
o
r
e
ef
f
e
ct
iv
e,
b
ec
au
s
e
in
th
e
tr
a
i
n
in
g
p
r
o
c
e
s
s
th
e
b
e
s
t
a
cc
u
r
ac
y
v
alu
e
ca
n
b
e
g
en
er
at
ed
a
t
ep
o
ch
s
2
1
to
3
7
b
y
u
s
in
g
th
e
E
ar
l
y
S
to
p
m
e
th
o
d
s
o
th
at
th
e
tr
ain
in
g
p
r
o
ce
s
s
ca
n
b
e
s
to
p
p
ed
a
t
a
ce
r
ta
in
ep
o
ch
wh
en
i
t
r
e
ac
h
es
th
e
lo
w
e
s
t
lo
s
s
v
a
lu
e
an
d
in
th
e
n
ex
t
5
ep
o
ch
s
i
t
d
o
e
s
n
o
t
d
ec
r
ea
s
e
th
e
lo
s
s
v
a
lu
e
.
T
h
i
s
is
d
o
n
e
to
av
o
id
o
v
er
f
i
t
tin
g
th
e
m
o
d
el
.
3
.
1
.
CNN
3
.
1
.
1
.
I
np
ut
l
a
y
er
a.
E
p
o
ch
1
5
˗
L
ea
r
n
in
g
r
ate
0
.
1
with
T
an
h
a
n
d
R
eL
U
a
ctiv
atio
n
W
h
en
th
e
lea
r
n
in
g
r
ate
is
s
et
h
ig
h
,
lik
e
0
.
1
as
s
h
o
wn
i
n
T
ab
l
e
1
,
th
e
m
o
d
el
lear
n
s
q
u
ick
ly
at
f
ir
s
t,
b
u
t
th
is
ca
n
ca
u
s
e
p
r
o
b
lem
s
in
d
ee
p
er
lay
er
s
.
T
h
e
T
an
h
ac
tiv
atio
n
f
u
n
ctio
n
m
ay
s
tr
u
g
g
l
e
with
v
an
is
h
in
g
g
r
ad
ien
ts
,
m
ak
in
g
it
h
ar
d
f
o
r
th
e
m
o
d
el
to
lear
n
ef
f
ec
tiv
el
y
.
On
th
e
o
th
er
h
an
d
,
R
eL
U
h
elp
s
b
y
k
ee
p
in
g
th
e
g
r
ad
ien
t
s
tr
o
n
g
,
wh
ich
m
a
k
es
lear
n
in
g
m
o
r
e
s
tab
le.
Ho
wev
e
r
,
a
v
er
y
h
ig
h
lear
n
in
g
r
ate
ca
n
ca
u
s
e
th
e
m
o
d
el
to
ju
m
p
o
v
er
t
h
e
b
est
s
o
lu
tio
n
,
lead
in
g
to
u
n
s
tab
le
lo
s
s
v
alu
e
s
.
I
t
i
s
im
p
o
r
tan
t
to
k
ee
p
a
n
ey
e
o
n
th
e
v
alid
atio
n
r
esu
lts
to
s
p
o
t a
n
y
s
ig
n
s
th
at
t
h
e
m
o
d
el’
s
p
er
f
o
r
m
an
ce
.
T
ab
le
1
.
T
h
e
v
alid
atio
n
r
esu
lts
with
0
.
1
(
lea
r
n
in
g
r
ate
)
A
g
g
r
e
g
a
t
e
F
u
n
c
t
i
o
n
O
p
t
i
mi
z
e
r
Le
a
r
n
i
n
g
r
a
t
e
C
l
a
s
s
V
a
l
A
c
c
P
r
e
c
i
s
i
o
n
R
e
c
a
l
l
F1
-
S
c
o
r
e
Ta
n
h
S
G
D
0
.
1
O
reo
c
h
r
o
m
i
s m
o
ss
a
m
b
i
c
u
s
0
.
9
7
1
.
0
0
1
.
0
0
1
.
0
0
Am
p
h
i
l
o
p
h
u
s
l
a
b
i
a
t
u
s
0
.
9
4
1
.
0
0
0
.
9
7
T
ri
c
h
o
p
o
d
u
s
p
e
c
t
o
r
a
l
i
s
1
.
0
0
0
.
9
4
0
.
9
7
R
e
LU
A
d
a
m
O
reo
c
h
r
o
m
i
s m
o
ss
a
m
b
i
c
u
s
0
.
3
3
0
.
0
0
0
.
0
0
0
.
0
0
Am
p
h
i
l
o
p
h
u
s
l
a
b
i
a
t
u
s
0
.
3
3
1
.
0
0
0
.
5
0
T
ri
c
h
o
p
o
d
u
s
p
e
c
t
o
r
a
l
i
s
0
.
0
0
0
.
0
0
0
.
0
0
˗
L
ea
r
n
in
g
r
ate
0
.
0
1
with
T
an
h
an
d
R
eL
U
a
ctiv
atio
n
A
lo
wer
lear
n
in
g
r
ate
(
0
.
0
1
)
e
n
co
u
r
a
g
es
m
o
r
e
s
tab
le
weig
h
t
u
p
d
ates
as
s
h
o
wn
in
T
ab
le
2
,
wh
ich
is
esp
ec
ially
u
s
ef
u
l
f
o
r
T
an
h
,
w
h
ich
s
u
f
f
er
s
f
r
o
m
d
elay
ed
co
n
v
er
g
e
n
ce
o
win
g
to
s
atu
r
atin
g
g
r
ad
ien
ts
.
R
eL
U
ad
d
s
to
th
is
o
p
tio
n
b
y
r
etain
i
n
g
g
r
a
d
ien
t
p
r
o
p
ag
atio
n
,
alth
o
u
g
h
t
h
e
m
o
d
el
m
ay
n
ee
d
lo
n
g
er
ep
o
ch
s
to
r
ea
ch
eq
u
al
ac
c
u
r
ac
y
to
th
e
h
ig
h
er
l
ea
r
n
in
g
r
ate
s
etu
p
.
T
h
is
s
etu
p
is
less
s
u
s
ce
p
tib
le
to
r
ap
i
d
p
er
f
o
r
m
an
ce
ch
a
n
g
es,
m
ak
in
g
it id
ea
l f
o
r
f
i
n
e
-
tu
n
i
n
g
.
T
ab
le
2
.
T
h
e
v
alid
atio
n
r
esu
lts
with
0
.
0
1
(
lear
n
in
g
r
ate)
A
g
g
r
e
g
a
t
e
F
u
n
c
t
i
o
n
O
p
t
i
mi
z
e
r
Le
a
r
n
i
n
g
r
a
t
e
C
l
a
s
s
V
a
l
A
c
c
P
r
e
c
i
s
i
o
n
R
e
c
a
l
l
F1
-
S
c
o
r
e
Ta
n
h
S
G
D
0
.
0
1
O
reo
c
h
r
o
m
i
s m
o
ss
a
m
b
i
c
u
s
0
.
9
8
1
.
0
0
1
.
0
0
1
.
0
0
Am
p
h
i
l
o
p
h
u
s
l
a
b
i
a
t
u
s
1
.
0
0
1
.
0
0
1
.
0
0
T
ri
c
h
o
p
o
d
u
s
p
e
c
t
o
r
a
l
i
s
1
.
0
0
1
.
0
0
1
.
0
0
R
e
LU
A
d
a
m
O
reo
c
h
r
o
m
i
s m
o
ss
a
m
b
i
c
u
s
0
.
9
3
1
.
0
0
1
.
0
0
1
.
0
0
Am
p
h
i
l
o
p
h
u
s
l
a
b
i
a
t
u
s
0
.
9
4
1
.
0
0
0
.
9
7
T
ri
c
h
o
p
o
d
u
s
p
e
c
t
o
r
a
l
i
s
1
.
0
0
0
.
9
4
0
.
9
7
3
.
1
.
2
.
H
idd
en
l
a
y
er
a.
E
p
o
ch
2
0
˗
L
ea
r
n
in
g
r
ate
0
.
1
with
T
an
h
a
n
d
R
eL
U
a
ctiv
atio
n
T
h
e
0
.
1
le
ar
n
i
n
g
r
at
e
wit
h
T
a
n
h
a
cti
v
at
io
n
s
h
o
ws
e
f
f
ec
ti
v
e
f
e
atu
r
e
tr
a
n
s
f
o
r
m
ati
o
n
i
n
th
e
h
i
d
d
e
n
l
ay
er
s
at
m
id
-
t
r
a
in
in
g
(
e
p
o
c
h
2
0
)
as
T
a
b
le
3
.
Alt
h
o
u
g
h
t
h
e
r
ap
id
l
ea
r
n
in
g
r
at
e
m
ay
r
esu
lt
i
n
o
s
c
illat
o
r
y
b
e
h
a
v
i
o
r
in
wei
g
h
t
u
p
d
at
es,
t
h
e
c
o
n
c
u
r
r
e
n
t
u
s
a
g
e
o
f
R
e
L
U
ai
d
s
i
n
m
ai
n
t
ai
n
i
n
g
g
r
a
d
ie
n
t
f
l
o
w
a
c
r
o
s
s
t
h
e
n
e
tw
o
r
k
'
s
in
t
er
m
e
d
i
at
e
r
e
p
r
es
en
tat
io
n
s
.
B
atc
h
n
o
r
m
ali
za
t
io
n
b
etw
ee
n
la
y
er
s
h
e
lp
s
t
h
is
d
esi
g
n
,
w
h
ic
h
is
es
p
ec
i
all
y
s
en
s
i
ti
v
e
t
o
ap
p
r
o
p
r
i
ate
i
n
i
tial
i
za
ti
o
n
.
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.
16
,
No
.
1
,
Feb
r
u
ar
y
20
26
:
3
8
3
-
394
388
T
ab
le
3
.
T
h
e
v
alid
atio
n
r
esu
lts
with
0
.
1
lear
n
in
g
r
ate
(
E
p
o
ch
2
0
)
A
g
g
r
e
g
a
t
e
F
u
n
c
t
i
o
n
O
p
t
i
mi
z
e
r
Le
a
r
n
i
n
g
r
a
t
e
C
l
a
s
s
V
a
l
A
c
c
P
r
e
c
i
s
i
o
n
R
e
c
a
l
l
F1
-
S
c
o
r
e
Ta
n
h
S
G
D
0
.
1
O
reo
c
h
r
o
m
i
s m
o
ss
a
m
b
i
c
u
s
0
.
9
7
1
.
0
0
1
.
0
0
1
.
0
0
Am
p
h
i
l
o
p
h
u
s
l
a
b
i
a
t
u
s
0
.
9
4
1
.
0
0
0
.
9
7
T
ri
c
h
o
p
o
d
u
s
p
e
c
t
o
r
a
l
i
s
1
.
0
0
0
.
9
4
0
.
9
7
R
e
LU
A
d
a
m
O
reo
c
h
r
o
m
i
s m
o
ss
a
m
b
i
c
u
s
0
.
5
0
0
.
0
0
0
.
0
0
0
.
0
0
Am
p
h
i
l
o
p
h
u
s
l
a
b
i
a
t
u
s
0
.
5
0
1
.
0
0
0
.
5
0
T
ri
c
h
o
p
o
d
u
s
p
e
c
t
o
r
a
l
i
s
0
.
0
0
0
.
0
0
0
.
0
0
˗
L
ea
r
n
in
g
r
ate
0
.
0
1
with
T
an
h
an
d
R
eL
U
a
ctiv
atio
n
Mo
r
e
co
n
tr
o
lled
f
ea
tu
r
e
r
ef
i
n
em
en
t
is
m
ad
e
p
o
s
s
ib
le
b
y
th
e
h
id
d
en
lay
e
r
s
'
lo
wer
lear
n
in
g
r
ate
o
f
0
.
0
1
,
T
an
h
o
f
f
er
s
s
m
o
o
th
n
o
n
lin
ea
r
tr
an
s
f
o
r
m
atio
n
s
,
wh
ile
R
eL
U
k
ee
p
s
g
r
ad
ien
t
s
f
r
o
m
co
m
p
letely
d
is
ap
p
ea
r
in
g
,
s
h
o
wn
in
T
a
b
l
e
4
.
Alth
o
u
g
h
it
co
u
ld
tak
e
m
o
r
e
ep
o
ch
s
to
attain
p
er
f
o
r
m
a
n
ce
co
m
p
ar
a
b
le
to
h
ig
h
er
lear
n
in
g
r
ate
co
n
f
ig
u
r
a
tio
n
s
,
th
is
co
m
b
in
atio
n
u
s
u
all
y
ex
h
ib
its
m
o
r
e
co
n
s
is
ten
t
lo
s
s
r
ed
u
ctio
n
d
u
r
in
g
m
id
-
tr
ain
in
g
p
h
ases
.
Fo
r
n
et
wo
r
k
s
with
in
tr
icate
h
ier
ar
ch
ical
f
ea
tu
r
e
r
ep
r
esen
tatio
n
s
,
t
h
is
co
n
f
ig
u
r
atio
n
is
f
r
eq
u
e
n
tly
p
r
e
f
er
r
ed
d
u
e
t
o
its
s
tab
ilit
y
.
T
ab
le
4
.
T
h
e
v
alid
atio
n
r
esu
lts
with
0
.
0
1
lea
r
n
in
g
r
ate
(
E
p
o
c
h
2
0
)
A
g
g
r
e
g
a
t
e
f
u
n
c
t
i
o
n
O
p
t
i
mi
z
e
r
Le
a
r
n
i
n
g
r
a
t
e
C
l
a
s
s
V
a
l
A
c
c
P
r
e
c
i
s
i
o
n
R
e
c
a
l
l
F1
-
S
c
o
r
e
Ta
n
h
S
G
D
0
.
0
1
O
reo
c
h
r
o
m
i
s m
o
ss
a
m
b
i
c
u
s
1
.
0
0
1
.
0
0
1
.
0
0
1
.
0
0
Am
p
h
i
l
o
p
h
u
s
l
a
b
i
a
t
u
s
1
.
0
0
1
.
0
0
1
.
0
0
T
ri
c
h
o
p
o
d
u
s
p
e
c
t
o
r
a
l
i
s
1
.
0
0
1
.
0
0
1
.
0
0
R
e
LU
A
d
a
m
O
reo
c
h
r
o
m
i
s m
o
ss
a
m
b
i
c
u
s
0
.
9
4
1
.
0
0
0
.
9
4
0
.
9
7
Am
p
h
i
l
o
p
h
u
s
l
a
b
i
a
t
u
s
0
.
9
4
1
.
0
0
0
.
9
7
T
ri
c
h
o
p
o
d
u
s
p
e
c
t
o
r
a
l
i
s
1
.
0
0
1
.
0
0
1
.
0
0
3
.
1
.
3
.
O
utput
l
a
y
er
a.
E
p
o
ch
2
5
˗
L
ea
r
n
in
g
r
ate
0
.
1
with
T
an
h
a
n
d
R
eL
U
a
ctiv
atio
n
T
ab
l
e
5
p
r
e
s
en
t
t
h
e
0
.
1
l
ea
r
n
i
n
g
r
a
te
w
i
th
T
an
h
a
ct
iv
a
tio
n
at
th
e
o
u
tp
u
t
lay
er
d
u
r
in
g
la
t
er
tr
a
in
in
g
s
t
ag
e
s
(
ep
o
ch
2
5
)
m
ig
h
t
s
p
e
ed
u
p
f
in
a
l
ch
an
g
e
s
to
d
e
ci
s
io
n
b
o
u
n
d
ar
i
e
s
,
wh
i
ch
i
s
es
p
ec
i
al
ly
h
elp
f
u
l
f
o
r
co
n
s
tr
ain
ed
o
u
tp
u
t
s
p
a
ce
s
.
Ho
w
ev
er
,
wh
en
ap
p
li
ed
to
f
in
a
l
l
ay
er
s
,
R
eL
U
's
u
n
b
o
u
n
d
ed
p
o
s
i
tiv
e
o
u
tp
u
t
m
ig
h
t
n
e
ed
to
b
e
ca
r
e
f
u
l
ly
s
ca
l
ed
,
wh
i
ch
c
o
u
ld
ca
u
s
e
in
s
tab
il
ity
in
th
e
n
e
two
r
k
's
p
r
e
d
ic
tio
n
s
.
T
o
a
v
o
id
o
v
er
f
i
t
tin
g
d
u
r
in
g
th
e
s
e
cr
u
ci
al
f
in
a
l
u
p
d
a
te
s
,
th
is
ar
r
an
g
e
m
en
t
n
ec
e
s
s
i
ta
te
s
ca
r
ef
u
l
m
o
n
ito
r
in
g
o
f
v
a
l
id
a
tio
n
m
etr
ic
s
.
T
ab
le
5
.
T
h
e
v
alid
atio
n
r
esu
lts
with
0
.
1
lear
n
in
g
r
ate
(
E
p
o
ch
2
5
)
A
g
g
r
e
g
a
t
e
F
u
n
c
t
i
o
n
O
p
t
i
mi
z
e
r
Le
a
r
n
i
n
g
r
a
t
e
C
l
a
s
s
V
a
l
A
c
c
P
r
e
c
i
s
i
o
n
R
e
c
a
l
l
F1
-
S
c
o
r
e
Ta
n
h
S
G
D
0
.
1
O
reo
c
h
r
o
m
i
s m
o
ss
a
m
b
i
c
u
s
0
.
9
7
1
.
0
0
1
.
0
0
1
.
0
0
Am
p
h
i
l
o
p
h
u
s
l
a
b
i
a
t
u
s
0
.
9
6
1
.
0
0
0
.
9
8
T
ri
c
h
o
p
o
d
u
s
p
e
c
t
o
r
a
l
i
s
1
.
0
0
0
.
9
6
0
.
9
8
R
e
LU
A
d
a
m
O
reo
c
h
r
o
m
i
s m
o
ss
a
m
b
i
c
u
s
0
.
5
0
0
.
0
0
0
.
0
0
0
.
0
0
Am
p
h
i
l
o
p
h
u
s
l
a
b
i
a
t
u
s
0
.
5
0
1
.
0
0
0
.
5
0
T
ri
c
h
o
p
o
d
u
s
p
e
c
t
o
r
a
l
i
s
0
.
0
0
0
.
0
0
0
.
0
0
˗
L
ea
r
n
in
g
r
ate
0
.
0
1
with
T
an
h
an
d
R
eL
U
a
ctiv
atio
n
T
an
h
p
r
o
d
u
ce
s
o
u
tp
u
ts
th
at
a
r
e
in
h
er
e
n
tly
b
o
u
n
d
ed
,
wh
ich
m
ay
b
e
u
s
ef
u
l
f
o
r
s
o
m
e
task
s
,
an
d
th
e
o
u
tp
u
t
la
y
er
'
s
0
.
0
1
lea
r
n
in
g
r
a
te
allo
ws
f
o
r
e
x
ac
t
f
in
e
-
tu
n
in
g
o
f
class
if
icatio
n
b
o
u
n
d
s
.
R
eL
U's
ad
d
itio
n
to
th
e
o
u
tp
u
t la
y
e
r
ca
n
wo
r
k
well
wh
en
co
m
b
in
e
d
with
th
is
co
n
s
er
v
ativ
e
lear
n
in
g
r
ate
a
n
d
a
p
p
r
o
p
r
iate
n
o
r
m
aliza
tio
n
m
eth
o
d
s
.
Alth
o
u
g
h
it
co
u
ld
ta
k
e
m
o
r
e
tim
e
to
tr
ain
th
e
n
etwo
r
k
to
its
m
ax
im
u
m
p
o
ten
tia
l,
th
is
co
n
f
ig
u
r
atio
n
p
r
o
d
u
ce
s
m
o
r
e
s
tab
le
f
in
al
m
o
d
els
,
as
il
lu
s
tr
ated
in
T
ab
le
6
.
I
n
th
e
g
r
ap
h
in
th
e
Fig
u
r
e
3
,
f
o
r
th
e
m
o
d
el
u
s
in
g
th
e
o
p
tim
izer
SGD
s
h
o
ws
th
at
th
e
lo
west
lo
s
s
v
alu
e
o
cc
u
r
s
at
th
e
1
0
th
e
p
o
ch
,
th
e
n
th
e
v
alu
e
o
f
th
e
b
est
ac
cu
r
ac
y
o
cc
u
r
r
e
d
at
th
e
1
0
th
e
p
o
ch
.
3
.
1
.
4
.
Co
nv
o
lutio
n
l
a
y
er
T
h
e
in
p
u
t
lay
e
r
o
f
th
e
C
NN
alg
o
r
ith
m
s
h
o
ws
th
at
th
e
C
NN
m
o
d
el
ac
ce
p
ts
im
ag
es
as
i
n
p
u
t
with
a
s
ize
o
f
2
2
4
p
i
x
els
h
ig
h
,
2
2
4
p
ix
els
wid
e,
an
d
3
co
lo
r
ch
an
n
els
(
R
GB
)
as
s
h
o
wn
in
Fig
u
r
e
4
.
T
h
ese
im
ag
es
co
m
e
f
r
o
m
a
d
ataset
th
at
is
p
r
o
ce
s
s
ed
u
s
in
g
th
e
lo
a
d
_
img
f
u
n
ctio
n
an
d
r
esized
to
f
it
th
o
s
e
d
im
en
s
io
n
s
.
T
h
is
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
I
ma
g
e
cla
s
s
ifica
tio
n
u
s
in
g
tw
o
n
eu
r
a
l n
etw
o
r
ks a
n
d
a
ctiva
ti
o
n
…
(
Op
p
ir
Hu
ta
p
ea
)
389
s
ize
is
u
s
ed
as a
s
tan
d
ar
d
to
en
ab
le
co
n
v
o
lu
tio
n
an
d
p
o
o
lin
g
o
p
er
atio
n
s
in
th
e
n
etwo
r
k
,
wh
er
e
ea
ch
im
ag
e
will
g
o
th
r
o
u
g
h
th
e
p
r
o
ce
s
s
o
f
ex
tr
ac
tin
g
f
ea
tu
r
es
s
u
ch
as
p
atter
n
s
,
tex
tu
r
es,
an
d
ed
g
es.
A
co
n
s
is
ten
t
in
p
u
t
s
ize
i
s
ess
en
tial f
o
r
th
e
m
o
d
el
t
o
o
p
ti
m
ally
lear
n
th
e
f
ea
tu
r
e
r
e
p
r
ese
n
tatio
n
at
ea
ch
lay
e
r
.
T
ab
le
6
.
T
h
e
v
alid
atio
n
r
esu
lts
with
0
.
0
1
lea
r
n
in
g
r
ate
(
E
p
o
c
h
2
5
)
A
g
g
r
e
g
a
t
e
F
u
n
c
t
i
o
n
O
p
t
i
mi
z
e
r
Le
a
r
n
i
n
g
r
a
t
e
C
l
a
s
s
V
a
l
A
c
c
P
r
e
c
i
s
i
o
n
R
e
c
a
l
l
F1
-
S
c
o
r
e
Ta
n
h
S
G
D
0
.
0
1
O
reo
c
h
r
o
m
i
s m
o
ss
a
m
b
i
c
u
s
1
.
0
0
1
.
0
0
1
.
0
0
1
.
0
0
Am
p
h
i
l
o
p
h
u
s
l
a
b
i
a
t
u
s
1
.
0
0
1
.
0
0
1
.
0
0
T
ri
c
h
o
p
o
d
u
s
p
e
c
t
o
r
a
l
i
s
1
.
0
0
1
.
0
0
1
.
0
0
R
e
LU
A
d
a
m
O
reo
c
h
r
o
m
i
s m
o
ss
a
m
b
i
c
u
s
0
.
5
0
1
.
0
0
1
.
0
0
1
.
0
0
Am
p
h
i
l
o
p
h
u
s
l
a
b
i
a
t
u
s
0
.
9
6
1
.
0
0
0
.
9
6
T
ri
c
h
o
p
o
d
u
s
p
e
c
t
o
r
a
l
i
s
1
.
0
0
0
.
9
4
0
.
9
8
Fig
u
r
e
3
.
L
o
s
s
p
lo
t v
s
ac
cu
r
ac
y
p
lo
t
Fig
u
r
e
4
.
I
m
ag
e
f
o
r
m
attin
g
3
.
2
.
ANN
3
.
2
.
1
.
E
po
ch
1
5
˗
L
ea
r
n
in
g
r
ate
0
.
1
with
T
an
h
a
n
d
R
eL
U
a
ctiv
atio
n
At
ep
o
ch
1
5
(
as
s
h
o
wn
in
T
ab
le
7
)
,
with
a
h
ig
h
lear
n
in
g
r
at
e
o
f
0
.
1
,
m
o
d
els
u
s
in
g
b
o
th
T
an
h
an
d
R
eL
U
s
tar
t to
lear
n
v
er
y
q
u
ick
ly
.
T
h
is
f
ast lea
r
n
in
g
ca
n
h
elp
r
ea
ch
th
e
b
est s
o
lu
tio
n
f
aster
,
b
u
t it
also
m
ak
es i
t
ea
s
ier
to
g
o
p
ast
th
e
b
est
p
o
in
t,
esp
ec
ially
with
T
an
h
,
wh
ich
ca
n
s
to
p
w
o
r
k
in
g
p
r
o
p
er
ly
.
R
eL
U
h
an
d
les
th
is
b
etter
b
ec
au
s
e
it
d
o
es
n
o
t
s
to
p
wo
r
k
in
g
as
ea
s
ily
,
b
u
t
it
ca
n
s
till
h
av
e
a
p
r
o
b
lem
if
th
e
ch
an
g
es
b
ec
o
m
e
to
o
b
ig
.
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.
16
,
No
.
1
,
Feb
r
u
ar
y
20
26
:
3
8
3
-
394
390
T
ab
le
7
.
T
h
e
v
alid
atio
n
r
esu
lt
with
0
.
1
lear
n
i
n
g
r
ate
A
g
g
r
e
g
a
t
e
F
u
n
c
t
i
o
n
O
p
t
i
mi
z
e
r
Le
a
r
n
i
n
g
r
a
t
e
C
l
a
s
s
V
a
l
A
c
c
P
r
e
c
i
s
i
o
n
R
e
c
a
l
l
F1
-
S
c
o
r
e
Ta
n
h
S
G
D
0
.
1
O
reo
c
h
r
o
m
i
s m
o
ss
a
m
b
i
c
u
s
0
.
9
9
1
.
0
0
1
.
0
0
1
.
0
0
Am
p
h
i
l
o
p
h
u
s
l
a
b
i
a
t
u
s
1
.
0
0
1
.
0
0
1
.
0
0
T
ri
c
h
o
p
o
d
u
s
p
e
c
t
o
r
a
l
i
s
1
.
0
0
1
.
0
0
1
.
0
0
R
e
LU
A
d
a
m
O
reo
c
h
r
o
m
i
s m
o
ss
a
m
b
i
c
u
s
0
.
3
3
0
.
0
0
0
.
0
0
0
.
0
0
Am
p
h
i
l
o
p
h
u
s
l
a
b
i
a
t
u
s
0
.
0
0
0
.
0
0
0
.
0
0
T
ri
c
h
o
p
o
d
u
s
p
e
c
t
o
r
a
l
i
s
0
.
3
3
1
.
0
0
0
.
5
0
˗
L
ea
r
n
in
g
r
ate
0
.
0
1
with
T
an
h
an
d
R
eL
U
ac
tiv
atio
n
Usi
n
g
a
lear
n
in
g
r
ate
o
f
0
.
0
1
at
ep
o
ch
1
5
h
elp
s
b
o
th
T
an
h
a
n
d
R
eL
U
co
n
v
er
g
e
m
o
r
e
s
tea
d
ily
.
T
an
h
g
iv
es
s
m
o
o
th
er
g
r
a
d
ien
t
ch
a
n
g
es
in
th
e
m
id
d
le
r
an
g
e,
w
h
ile
R
eL
U
allo
ws
q
u
ick
er
t
r
ain
in
g
b
ec
a
u
s
e
it
is
s
im
p
ler
an
d
u
s
es
less
d
ata.
B
y
th
is
p
o
in
t,
b
o
th
ac
tiv
atio
n
f
u
n
ctio
n
s
s
tar
t
to
im
p
r
o
v
e
c
o
n
s
is
ten
tly
with
les
s
ch
an
ce
o
f
b
ec
o
m
i
n
g
u
n
s
tab
le,
as sh
o
wn
in
T
ab
le
8
.
T
ab
le
8
.
T
h
e
v
alid
atio
n
r
esu
lt
with
0
.
0
1
lear
n
in
g
r
ate
A
g
g
r
e
g
a
t
e
F
u
n
c
t
i
o
n
O
p
t
i
mi
z
e
r
Le
a
r
n
i
n
g
r
a
t
e
C
l
a
s
s
V
a
l
A
c
c
P
r
e
c
i
s
i
o
n
R
e
c
a
l
l
F1
-
S
c
o
r
e
Ta
n
h
S
G
D
0
.
0
1
O
reo
c
h
r
o
m
i
s m
o
ss
a
m
b
i
c
u
s
0
.
9
8
0
.
9
7
0
.
9
7
0
.
9
7
Am
p
h
i
l
o
p
h
u
s
l
a
b
i
a
t
u
s
1
.
0
0
1
.
0
0
1
.
0
0
T
ri
c
h
o
p
o
d
u
s
p
e
c
t
o
r
a
l
i
s
0
.
9
7
0
.
9
7
0
.
9
7
R
e
LU
A
d
a
m
O
reo
c
h
r
o
m
i
s m
o
ss
a
m
b
i
c
u
s
0
.
3
3
0
.
0
0
0
.
0
0
0
.
0
0
Am
p
h
i
l
o
p
h
u
s
l
a
b
i
a
t
u
s
0
.
4
4
1
.
0
0
0
.
6
1
T
ri
c
h
o
p
o
d
u
s
p
e
c
t
o
r
a
l
i
s
1
.
0
0
0
.
7
3
0
.
8
4
3
.
2
.
2
.
E
po
ch
2
0
˗
L
ea
r
n
in
g
r
ate
0
.
1
with
T
an
h
a
n
d
R
eL
U
ac
tiv
atio
n
At
ep
o
ch
2
0
,
a
lear
n
in
g
r
ate
o
f
0
.
1
m
ig
h
t
s
till
wo
r
k
well
an
d
lead
to
q
u
ick
im
p
r
o
v
em
en
ts
,
b
u
t
th
er
e
co
u
ld
b
e
s
ig
n
s
o
f
p
r
o
b
lem
s
,
esp
ec
ially
in
n
etwo
r
k
s
th
at
u
s
e
T
an
h
.
T
h
ese
n
etwo
r
k
s
m
ig
h
t
s
tr
u
g
g
le
with
v
an
is
h
in
g
g
r
a
d
ien
ts
in
d
ee
p
e
r
lay
er
s
.
R
eL
U
ca
n
s
till
wo
r
k
,
b
u
t
if
th
e
lear
n
in
g
r
ate
is
to
o
h
i
g
h
,
it
m
ig
h
t
ca
u
s
e
th
e
lo
s
s
to
f
lu
ctu
ate
a
lo
t.
T
h
is
s
etu
p
m
ig
h
t
b
e
b
etter
f
o
r
s
im
p
ler
n
etwo
r
k
s
o
r
o
n
es
th
at
u
s
e
b
atch
n
o
r
m
aliza
tio
n
(
s
h
o
wn
in
T
ab
l
e
9
)
.
T
ab
le
9
.
T
h
e
v
alid
atio
n
r
esu
lt
with
0
.
1
lear
n
i
n
g
r
ate
(
E
p
o
c
h
2
0
)
A
g
g
r
e
g
a
t
e
F
u
n
c
t
i
o
n
O
p
t
i
mi
z
e
r
Le
a
r
n
i
n
g
r
a
t
e
C
l
a
s
s
V
a
l
A
c
c
P
r
e
c
i
s
i
o
n
R
e
c
a
l
l
F1
-
S
c
o
r
e
Ta
n
h
S
G
D
0
.
1
O
reo
c
h
r
o
m
i
s m
o
ss
a
m
b
i
c
u
s
0
.
9
9
1
.
0
0
1
.
0
0
1
.
0
0
Am
p
h
i
l
o
p
h
u
s
l
a
b
i
a
t
u
s
1
.
0
0
1
.
0
0
1
.
0
0
T
ri
c
h
o
p
o
d
u
s
p
e
c
t
o
r
a
l
i
s
0
.
9
7
0
.
9
7
0
.
9
7
R
e
LU
A
d
a
m
O
reo
c
h
r
o
m
i
s m
o
ss
a
m
b
i
c
u
s
0
.
3
3
0
.
0
0
0
.
0
0
0
.
0
0
Am
p
h
i
l
o
p
h
u
s
l
a
b
i
a
t
u
s
0
.
3
3
1
.
0
0
0
.
5
0
T
ri
c
h
o
p
o
d
u
s
p
e
c
t
o
r
a
l
i
s
0
.
7
0
0
.
4
3
0
.
2
3
˗
L
ea
r
n
in
g
r
ate
0
.
0
1
with
T
an
h
an
d
R
eL
U
a
ctiv
atio
n
W
ith
a
le
ar
n
in
g
r
a
te
o
f
0
.
0
1
,
b
o
th
ac
tiv
at
io
n
f
u
n
ct
io
n
s
s
h
o
w
s
te
ad
y
p
r
o
g
r
e
s
s
b
y
th
e
2
0
th
ep
o
ch
,
as
s
h
o
wn
in
T
ab
l
e
1
0
.
T
an
h
k
e
ep
s
p
er
f
o
r
m
in
g
we
ll
in
ta
s
k
s
wh
er
e
t
h
e
o
u
t
p
u
t
n
e
ed
s
to
s
t
ay
w
ith
in
a
ce
r
ta
in
r
an
g
e,
wh
i
le
R
eL
U
k
ee
p
s
b
e
in
g
ef
f
e
ct
iv
e
in
s
itu
at
io
n
s
wh
er
e
o
n
ly
s
o
m
e
n
eu
r
o
n
s
a
r
e
ac
t
iv
e.
T
h
e
s
l
o
w
er
l
ea
r
n
in
g
r
a
te
h
elp
s
p
r
e
v
en
t
th
e
m
o
d
e
l
f
r
o
m
g
o
in
g
o
f
f
tr
ac
k
an
d
im
p
r
o
v
es
h
o
w
w
el
l
i
t
wo
r
k
s
wi
th
n
e
w
d
ata
.
T
ab
le
1
0
.
T
h
e
v
alid
atio
n
r
esu
l
t
with
0
.
0
1
lear
n
in
g
r
ate
(
E
p
o
c
h
2
0
)
A
g
g
r
e
g
a
t
e
F
u
n
c
t
i
o
n
O
p
t
i
mi
z
e
r
Le
a
r
n
i
n
g
r
a
t
e
C
l
a
s
s
V
a
l
A
c
c
P
r
e
c
i
s
i
o
n
R
e
c
a
l
l
F1
-
S
c
o
r
e
Ta
n
h
S
G
D
0
.
0
1
O
reo
c
h
r
o
m
i
s m
o
ss
a
m
b
i
c
u
s
0
.
9
8
1
.
0
0
1
.
0
0
1
.
0
0
Am
p
h
i
l
o
p
h
u
s
l
a
b
i
a
t
u
s
1
.
0
0
1
.
0
0
1
.
0
0
T
ri
c
h
o
p
o
d
u
s
p
e
c
t
o
r
a
l
i
s
0
.
9
7
0
.
9
7
0
.
9
7
R
e
LU
A
d
a
m
O
reo
c
h
r
o
m
i
s m
o
ss
a
m
b
i
c
u
s
0
.
6
6
0
.
4
4
0
.
7
7
0
.
6
6
Am
p
h
i
l
o
p
h
u
s
l
a
b
i
a
t
u
s
0
.
7
4
1
.
0
0
0
.
8
1
T
ri
c
h
o
p
o
d
u
s
p
e
c
t
o
r
a
l
i
s
1
.
0
0
0
.
8
3
0
.
8
8
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
I
ma
g
e
cla
s
s
ifica
tio
n
u
s
in
g
tw
o
n
eu
r
a
l n
etw
o
r
ks a
n
d
a
ctiva
ti
o
n
…
(
Op
p
ir
Hu
ta
p
ea
)
391
3
.
2
.
3
.
E
po
ch
2
5
˗
L
ea
r
n
in
g
r
ate
0
.
1
with
T
an
h
a
n
d
R
eL
U
a
ctiv
atio
n
B
y
th
e
2
5
th
ep
o
c
h
(
ac
c
o
r
d
in
g
to
T
ab
le
1
1
)
,
u
s
in
g
a
lear
n
in
g
r
ate
o
f
0
.
1
ca
n
ca
u
s
e
tr
ain
in
g
t
o
b
ec
o
m
e
u
n
s
tab
le
o
r
s
to
p
im
p
r
o
v
in
g
,
esp
ec
ially
wh
en
u
s
in
g
T
an
h
b
ec
a
u
s
e
it
r
ea
cts
s
tr
o
n
g
ly
to
b
ig
c
h
an
g
es
in
g
r
a
d
ien
ts
.
R
eL
U
m
ig
h
t
s
till
wo
r
k
o
k
a
y
,
b
u
t
t
h
er
e'
s
a
ch
an
ce
th
e
m
o
d
el
co
u
ld
b
o
u
n
ce
ar
o
u
n
d
th
e
l
o
west
p
o
in
ts
with
o
u
t
s
ettlin
g
.
T
h
is
s
etu
p
is
u
s
u
ally
n
o
t
ad
v
is
ed
u
n
less
y
o
u
also
u
s
e
a
lear
n
in
g
r
ate
th
at
d
ec
r
ea
s
es
o
v
er
tim
e.
˗
L
ea
r
n
in
g
r
ate
0
.
0
1
with
T
an
h
an
d
R
eL
U
a
ctiv
atio
n
At
th
is
s
tag
e,
th
er
e
is
a
lear
n
in
g
r
ate
o
f
0
.
0
1
s
till
h
elp
s
th
e
tr
ain
in
g
g
o
s
m
o
o
t
h
ly
an
d
r
eliab
ly
f
o
r
b
o
th
ac
tiv
atio
n
f
u
n
ctio
n
s
.
T
a
n
h
g
i
v
es
o
u
tp
u
ts
th
at
ar
e
m
o
r
e
c
en
ter
ed
,
w
h
ich
ca
n
b
e
u
s
ef
u
l
in
s
o
m
e
n
etwo
r
k
d
esig
n
s
,
wh
ile
R
eL
U
k
ee
p
s
p
er
f
o
r
m
in
g
well
b
ec
au
s
e
it
i
s
s
im
p
le
an
d
ef
f
icien
t.
T
h
is
s
etu
p
wo
r
k
s
well
f
o
r
tr
ain
in
g
d
ee
p
er
th
a
n
2
5
ep
o
c
h
s
,
as illu
s
tr
ated
in
T
ab
le
1
2
.
T
ab
le
1
1
.
T
h
e
v
alid
atio
n
r
esu
l
t w
ith
0
.
1
lear
n
i
n
g
r
ate
(
E
p
o
c
h
2
5
)
A
g
g
r
e
g
a
t
e
F
u
n
c
t
i
o
n
O
p
t
i
mi
z
e
r
Le
a
r
n
i
n
g
r
a
t
e
C
l
a
s
s
V
a
l
A
c
c
P
r
e
c
i
s
i
o
n
R
e
c
a
l
l
F1
-
S
c
o
r
e
Ta
n
h
S
G
D
0
.
1
O
reo
c
h
r
o
m
i
s m
o
ss
a
m
b
i
c
u
s
0
.
9
9
1
.
0
0
1
.
0
0
1
.
0
0
Am
p
h
i
l
o
p
h
u
s
l
a
b
i
a
t
u
s
1
.
0
0
1
.
0
0
1
.
0
0
T
ri
c
h
o
p
o
d
u
s
p
e
c
t
o
r
a
l
i
s
0
.
9
7
0
.
9
7
0
.
9
7
R
e
LU
A
d
a
m
O
reo
c
h
r
o
m
i
s m
o
ss
a
m
b
i
c
u
s
0
.
3
3
0
.
0
0
0
.
0
0
0
.
0
0
Am
p
h
i
l
o
p
h
u
s
l
a
b
i
a
t
u
s
0
.
3
3
1
.
0
0
0
.
5
0
T
ri
c
h
o
p
o
d
u
s
p
e
c
t
o
r
a
l
i
s
0
.
0
0
0
.
0
0
0
.
0
0
T
ab
le
1
2
.
T
h
e
v
alid
atio
n
r
esu
l
t
with
0
.
0
1
lear
n
in
g
r
ate
(
E
p
o
c
h
2
5
)
A
g
g
r
e
g
a
t
e
F
u
n
c
t
i
o
n
O
p
t
i
mi
z
e
r
Le
a
r
n
i
n
g
r
a
t
e
C
l
a
s
s
V
a
l
A
c
c
P
r
e
c
i
s
i
o
n
R
e
c
a
l
l
F1
-
S
c
o
r
e
Ta
n
h
S
G
D
0
.
0
1
O
reo
c
h
r
o
m
i
s m
o
ss
a
m
b
i
c
u
s
0
.
9
9
1
.
0
0
1
.
0
0
1
.
0
0
Am
p
h
i
l
o
p
h
u
s
l
a
b
i
a
t
u
s
1
.
0
0
1
.
0
0
1
.
0
0
T
ri
c
h
o
p
o
d
u
s
p
e
c
t
o
r
a
l
i
s
1
.
0
0
1
.
0
0
1
.
0
0
R
e
LU
A
d
a
m
O
reo
c
h
r
o
m
i
s m
o
ss
a
m
b
i
c
u
s
0
.
9
7
1
.
0
0
1
.
0
0
1
.
0
0
Am
p
h
i
l
o
p
h
u
s
l
a
b
i
a
t
u
s
1
.
0
0
1
.
0
0
1
.
0
0
T
ri
c
h
o
p
o
d
u
s
p
e
c
t
o
r
a
l
i
s
1
.
0
0
1
.
0
0
1
.
0
0
T
h
en
Fig
u
r
e
5
s
h
o
ws
a
g
r
ap
h
o
f
ac
c
u
r
ac
y
an
d
lo
s
s
f
r
o
m
o
n
e
o
f
th
e
test
s
with
th
e
h
ig
h
e
s
t
ac
cu
r
ac
y
v
alu
e
s
h
o
wn
ab
o
v
e.
B
o
th
th
e
t
r
ain
in
g
an
d
v
alid
atio
n
l
o
s
s
es
k
ee
p
g
o
in
g
d
o
wn
as
th
e
m
o
d
el
g
o
es
th
r
o
u
g
h
m
o
r
e
tr
ain
in
g
r
o
u
n
d
s
,
wh
ic
h
m
ea
n
s
th
e
m
o
d
el
is
lear
n
in
g
well.
T
h
e
v
alid
atio
n
lo
s
s
s
tay
s
clo
s
e
to
th
e
t
r
ain
in
g
lo
s
s
with
o
u
t
a
b
ig
g
a
p
,
s
h
o
win
g
th
at
th
e
m
o
d
el
is
not
o
v
e
r
f
itti
n
g
,
m
ea
n
in
g
it
d
o
es
n
o
t
d
o
g
r
ea
t o
n
tr
ain
in
g
d
ata
b
u
t
wo
r
s
e
o
n
n
ew,
u
n
s
ee
n
d
ata.
T
h
e
f
in
al
lo
s
s
v
alu
es
a
r
e
l
o
w,
a
r
o
u
n
d
0
.
1
,
wh
ich
is
a
g
o
o
d
s
i
g
n
f
o
r
h
o
w
well
th
e
m
o
d
el
wo
r
k
s
.
B
o
th
tr
ain
in
g
a
n
d
v
alid
atio
n
ac
cu
r
ac
ies
g
o
u
p
o
v
er
tim
e,
r
ea
ch
in
g
v
er
y
h
ig
h
lev
els,
m
o
r
e
th
a
n
9
0
%.
T
h
e
v
alid
atio
n
ac
c
u
r
ac
y
f
o
llo
ws
th
e
tr
ai
n
in
g
ac
cu
r
a
cy
clo
s
ely
,
wh
ich
ag
ain
s
h
o
ws
th
e
m
o
d
el
is
n
o
t
o
v
er
f
itti
n
g
.
T
h
e
f
i
n
al
ac
cu
r
ac
y
v
alu
es a
r
e
h
ig
h
,
ar
o
u
n
d
9
5
% f
o
r
tr
ain
in
g
an
d
ju
s
t a
litt
le
lo
wer
f
o
r
v
alid
atio
n
.
Fig
u
r
e
5
.
L
o
s
s
an
d
ac
cu
r
ac
y
p
l
o
t
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.
16
,
No
.
1
,
Feb
r
u
ar
y
20
26
:
3
8
3
-
394
392
4.
CO
NCLU
SI
O
N
A
k
ey
co
n
tr
i
b
u
tio
n
o
f
th
is
wo
r
k
lies
in
th
e
s
y
s
tem
atic
ev
alu
atio
n
o
f
ac
tiv
atio
n
f
u
n
c
tio
n
s
an
d
o
p
tim
izer
s
in
b
o
th
C
NN
an
d
ANN
m
o
d
els
f
o
r
aq
u
atic
im
ag
e
class
if
icatio
n
.
T
h
is
co
m
p
ar
is
o
n
o
f
f
e
r
s
n
ew
in
s
ig
h
ts
in
to
th
e
a
r
ch
itectu
r
al
p
er
f
o
r
m
an
ce
tr
a
d
e
-
o
f
f
s
.
Fu
tu
r
e
ex
ten
s
io
n
s
in
clu
d
e
in
teg
r
atin
g
th
is
s
y
s
tem
in
t
o
r
ea
l
-
tim
e
m
o
n
ito
r
in
g
p
latf
o
r
m
s
f
o
r
in
v
asiv
e
s
p
ec
ies
m
an
ag
em
en
t
an
d
ad
ap
tin
g
th
e
f
r
am
ewo
r
k
to
class
if
y
o
th
er
s
p
ec
ies
in
d
iv
er
s
e
ec
o
s
y
s
tem
s
.
T
h
is
r
esear
ch
s
u
cc
ess
f
u
lly
d
ev
elo
p
e
d
two
class
if
icat
io
n
m
o
d
els
f
o
r
f
is
h
s
p
ec
ies
f
r
o
m
L
ak
e
T
o
b
a
,
n
am
ely
r
ed
d
ev
il
f
is
h
(
A
mp
h
ilo
p
h
u
s
la
b
ia
tu
s
)
,
m
u
jah
ir
f
is
h
(
Oreo
ch
r
o
mis
mo
s
s
a
mb
icu
s
)
,
s
ep
at
f
is
h
(
Tr
ich
o
g
a
s
ter
tr
ich
o
p
teru
s
)
u
s
i
n
g
ANN
a
n
d
C
NN
alg
o
r
ith
m
s
.
T
h
is
m
o
d
el
is
d
esig
n
ed
to
u
tili
ze
im
ag
e
-
b
a
s
ed
class
if
icatio
n
tech
n
iq
u
es
to
au
to
m
atica
lly
d
etec
t
f
is
h
s
p
ec
ies.
Fro
m
th
e
r
esear
ch
r
esu
lts
,
b
o
th
m
o
d
els
s
h
o
wed
ex
ce
llen
t
p
er
f
o
r
m
an
c
e
with
a
h
ig
h
lev
el
o
f
ac
c
u
r
ac
y
.
T
h
e
C
NN
m
o
d
el
g
av
e
th
e
b
est
r
esu
lts
o
n
th
e
te
s
t
d
ata
with
a
co
m
b
in
atio
n
o
f
R
eL
U
ac
tiv
atio
n
f
u
n
ctio
n
a
n
d
Ad
am
o
p
tim
izer
.
T
h
e
ANN
m
o
d
el,
with
a
s
im
ilar
co
n
f
ig
u
r
atio
n
,
also
s
h
o
wed
ex
ce
llen
t
p
er
f
o
r
m
a
n
ce
,
p
r
o
v
in
g
th
at
b
o
t
h
ap
p
r
o
ac
h
es
ar
e
eq
u
ally
ef
f
ec
tiv
e
in
im
ag
e
class
if
ica
tio
n
task
s
.
R
ep
o
r
ted
ly
,
th
e
C
NN
m
o
d
el
with
R
eL
U
ac
tiv
atio
n
f
u
n
ctio
n
is
m
o
r
e
s
t
ab
le
an
d
f
ast
in
th
e
tr
ain
in
g
p
r
o
ce
s
s
th
an
th
e
ANN.
T
ests
wer
e
co
n
d
u
cted
o
n
1
0
%
o
f
th
e
d
ata
f
r
o
m
th
e
to
t
al
d
ataset
(
3
,
0
6
3
im
a
g
es)
th
at
h
ad
b
ee
n
s
ep
ar
ate
d
as
test
d
ata.
T
o
ad
d
r
ess
th
e
ch
allen
g
e
o
f
lim
ited
tr
ain
in
g
d
ata
an
d
en
h
a
n
ce
m
o
d
el
r
o
b
u
s
tn
ess
,
th
e
im
p
lem
en
tatio
n
o
f
ad
v
an
ce
d
d
ata
au
g
m
en
tatio
n
tec
h
n
iq
u
es
is
s
tr
o
n
g
ly
r
ec
o
m
m
e
n
d
ed
.
B
elo
w
ar
e
th
e
d
etails
o
f
th
e
test
r
esu
lts
o
n
th
e
test
d
ata
f
o
r
ea
ch
f
is
h
s
p
ec
ies:
r
ed
d
ev
i
l
:
All
im
ag
es
wer
e
clas
s
if
ied
co
r
r
ec
tly
with
o
u
t
e
r
r
o
r
.
Mu
ja
h
ir
:
T
h
e
m
o
d
el
was
ab
le
to
r
ec
o
g
n
ize
th
is
s
p
ec
ies
with
p
er
f
ec
t
ac
cu
r
ac
y
.
Flath
e
ad
:
No
class
if
icatio
n
er
r
o
r
s
o
cc
u
r
r
ed
,
in
d
icatin
g
th
e
m
o
d
el
ca
n
d
is
tin
g
u
is
h
th
e
u
n
iq
u
e
f
ea
t
u
r
es
o
f
th
is
f
is
h
.
T
h
e
class
if
icatio
n
tech
n
iq
u
e
u
s
ed
was
im
ag
e
-
b
ased
class
if
icatio
n
with
a
d
ee
p
lea
r
n
in
g
m
o
d
el.
C
NNs
ar
e
d
esig
n
ed
to
ex
tr
ac
t
v
is
u
al
f
ea
tu
r
es
f
r
o
m
im
ag
es
u
s
in
g
co
n
v
o
l
u
tio
n
,
p
o
o
lin
g
an
d
d
e
n
s
e
lay
er
s
,
wh
ile
ANNs
p
r
o
ce
s
s
f
latten
ed
im
ag
e
d
ata
t
o
g
e
n
er
ate
p
r
e
d
ictio
n
s
.
B
o
th
tech
n
iq
u
es
en
ab
le
t
h
e
r
e
co
g
n
itio
n
o
f
co
m
p
lex
p
atter
n
s
in
f
is
h
im
ag
es,
with
f
ea
tu
r
es
s
u
ch
as
b
o
d
y
s
h
a
p
e,
tex
tu
r
e,
an
d
co
l
o
r
as
k
ey
in
d
i
ca
to
r
s
.
Ov
er
all,
th
is
r
esear
ch
m
ak
es
a
s
ig
n
if
ican
t
co
n
tr
ib
u
ti
o
n
in
s
u
p
p
o
r
tin
g
th
e
o
p
er
atio
n
al
ef
f
icien
cy
o
f
th
e
f
is
h
er
ies
s
ec
to
r
an
d
b
io
d
iv
e
r
s
ity
co
n
s
er
v
atio
n
in
L
ak
e
T
o
b
a.
T
h
e
r
esu
lts
ca
n
s
er
v
e
as a
b
asis
f
o
r
th
e
im
p
le
m
en
tatio
n
o
f
s
im
ilar
tech
n
o
lo
g
ies in
o
th
er
wate
r
a
r
ea
s
.
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
Op
p
ir
Hu
tap
ea
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
Fo
r
d
L
u
m
b
an
Gao
l
✓
✓
✓
✓
✓
✓
✓
T
o
k
u
r
o
Ma
ts
u
o
✓
✓
✓
✓
C
:
C
o
n
c
e
p
t
u
a
l
i
z
a
t
i
o
n
M
:
M
e
t
h
o
d
o
l
o
g
y
So
:
So
f
t
w
a
r
e
Va
:
Va
l
i
d
a
t
i
o
n
Fo
:
Fo
r
mal
a
n
a
l
y
s
i
s
I
:
I
n
v
e
s
t
i
g
a
t
i
o
n
R
:
R
e
so
u
r
c
e
s
D
:
D
a
t
a
C
u
r
a
t
i
o
n
O
:
W
r
i
t
i
n
g
-
O
r
i
g
i
n
a
l
D
r
a
f
t
E
:
W
r
i
t
i
n
g
-
R
e
v
i
e
w
&
E
d
i
t
i
n
g
Vi
:
Vi
su
a
l
i
z
a
t
i
o
n
Su
:
Su
p
e
r
v
i
s
i
o
n
P
:
P
r
o
j
e
c
t
a
d
mi
n
i
st
r
a
t
i
o
n
Fu
:
Fu
n
d
i
n
g
a
c
q
u
i
si
t
i
o
n
CO
NF
L
I
C
T
O
F
I
N
T
E
R
E
S
T
ST
A
T
E
M
E
NT
Au
th
o
r
s
s
tate
n
o
co
n
f
lict o
f
in
t
er
est.
DATA AV
AI
L
AB
I
L
I
T
Y
T
h
ese
d
ata
m
ay
in
clu
d
e
r
aw
d
atasets
,
p
r
o
ce
s
s
ed
r
esu
lts
,
o
r
s
u
p
p
lem
en
tar
y
m
ater
ials
th
at
wer
e
u
s
ed
to
d
r
aw
th
e
co
n
cl
u
s
io
n
s
p
r
esen
t
ed
in
th
is
r
esear
ch
.
I
n
te
r
ested
r
esear
ch
er
s
m
ay
c
o
n
tact
th
e
au
th
o
r
to
r
eq
u
est
ac
ce
s
s
,
p
r
o
v
id
ed
t
h
e
r
eq
u
est
is
f
o
r
leg
itima
te
ac
ad
em
ic
o
r
s
cien
tific
p
u
r
p
o
s
es
an
d
co
m
p
lies
with
an
y
ap
p
licab
le
eth
ical
o
r
p
r
iv
ac
y
c
o
n
s
id
er
atio
n
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.