I
AE
S In
t
er
na
t
io
na
l J
o
urna
l o
f
Art
if
icia
l In
t
ellig
ence
(
I
J
-
AI
)
Vo
l.
1
5
,
No
.
3
,
J
u
n
e
2
0
2
6
,
p
p
.
2685
~
2
6
9
7
I
SS
N:
2
2
5
2
-
8
9
3
8
,
DOI
: 1
0
.
1
1
5
9
1
/ijai.v
15
.i
3
.
p
p
2
6
8
5
-
2
6
9
7
2685
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
a
i
.
ia
esco
r
e.
co
m
A nov
el deep
lear
ning
-
ba
sed po
llin
a
tor species cla
ss
i
ficatio
n
using
t
he m
ultim
o
da
l species
-
ima
g
e
regu
la
riza
tion
f
ra
mewo
rk
P
o
o
j
a
H
a
dim
a
ne
1
,
Ash
o
ka
K
uk
k
uv
a
da
2
1
D
e
p
a
r
t
me
n
t
o
f
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
a
n
d
E
n
g
i
n
e
e
r
i
n
g
,
B
a
p
u
j
i
I
n
st
i
t
u
t
e
o
f
E
n
g
i
n
e
e
r
i
n
g
a
n
d
T
e
c
h
n
o
l
o
g
y
(
A
f
f
i
l
i
a
t
e
d
t
o
V
i
sv
e
s
v
a
r
a
y
a
Te
c
h
n
o
l
o
g
i
c
a
l
U
n
i
v
e
r
si
t
y
,
B
e
l
a
g
a
v
i
)
,
D
a
v
a
n
g
e
r
e
,
I
n
d
i
a
2
D
e
p
a
r
t
me
n
t
o
f
I
n
f
o
r
mat
i
o
n
S
c
i
e
n
c
e
a
n
d
E
n
g
i
n
e
e
r
i
n
g
,
B
a
p
u
j
i
I
n
st
i
t
u
t
e
o
f
E
n
g
i
n
e
e
r
i
n
g
a
n
d
Te
c
h
n
o
l
o
g
y
(
A
f
f
i
l
i
a
t
e
d
t
o
V
i
sv
e
s
v
a
r
a
y
a
Te
c
h
n
o
l
o
g
i
c
a
l
U
n
i
v
e
r
si
t
y
,
B
e
l
a
g
a
v
i
)
,
D
a
v
a
n
g
e
r
e
,
I
n
d
i
a
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Dec
1
1
,
2
0
2
4
R
ev
is
ed
Ma
r
9
,
2
0
2
6
Acc
ep
ted
Ap
r
2
2
,
2
0
2
6
P
o
ll
i
n
a
to
rs,
su
c
h
a
s
b
e
e
s,
b
u
tt
e
rfli
e
s,
a
n
d
o
th
e
r
in
se
c
ts,
a
re
e
s
se
n
ti
a
l
to
m
a
in
tain
in
g
b
io
d
iv
e
rsit
y
a
n
d
e
n
su
rin
g
a
g
ricu
lt
u
ra
l
p
r
o
d
u
c
ti
v
it
y
,
with
o
v
e
r
8
0
%
o
f
fl
o
we
rin
g
p
lan
ts
a
n
d
7
5
%
o
f
g
lo
b
a
l
fo
o
d
c
ro
p
s
re
ly
i
n
g
o
n
th
e
m
fo
r
su
c
c
e
ss
fu
l
re
p
ro
d
u
c
ti
o
n
.
Ho
we
v
e
r,
p
o
ll
in
a
to
r
p
o
p
u
lati
o
n
s
a
r
e
fa
c
in
g
sig
n
ifi
c
a
n
t
d
e
c
li
n
e
s
d
u
e
t
o
e
n
v
ir
o
n
m
e
n
tal
c
h
a
n
g
e
s,
h
a
b
it
a
t
d
e
stru
c
ti
o
n
,
a
n
d
c
li
m
a
te
c
h
a
n
g
e
,
p
o
si
n
g
su
b
sta
n
t
ial
risk
s
to
e
c
o
sy
ste
m
s
a
n
d
g
l
o
b
a
l
fo
o
d
se
c
u
rit
y
.
Th
is
p
a
p
e
r
in
tr
o
d
u
c
e
s
th
e
m
u
lt
imo
d
a
l
sp
e
c
ies
-
ima
g
e
re
g
u
lariz
a
ti
on
(M
S
IR)
fra
m
e
wo
rk
fo
r
a
u
to
m
a
ti
n
g
t
h
e
c
las
sifica
ti
o
n
o
f
p
o
ll
i
n
a
to
r
sp
e
c
ies
u
sin
g
b
o
t
h
b
in
a
ry
c
las
sifica
ti
o
n
(
p
o
ll
i
n
a
to
r
v
s.
n
o
n
-
p
o
l
li
n
a
t
o
r)
a
n
d
m
u
lt
icla
ss
c
las
sifica
ti
o
n
(b
e
e
g
e
n
e
ra
).
T
h
e
fra
m
e
wo
rk
in
teg
ra
tes
m
u
lt
imo
d
a
l
d
a
ta
,
in
c
lu
d
in
g
v
isu
a
l
ima
g
e
s
o
f
p
o
ll
i
n
a
to
rs
a
n
d
sp
e
c
ies
d
e
tails
s
u
c
h
a
s
g
e
n
u
s,
fa
m
il
y
,
a
n
d
e
n
v
ir
o
n
m
e
n
tal
fa
c
to
r
s,
to
imp
r
o
v
e
a
c
c
u
ra
c
y
a
n
d
sc
a
lab
il
it
y
.
Th
e
sy
ste
m
lev
e
ra
g
e
s
th
e
m
u
lt
imo
d
a
l
c
o
n
tras
ti
v
e
lea
rn
in
g
fra
m
e
wo
rk
(
M
CLF
)
t
o
a
li
g
n
b
o
th
ima
g
e
a
n
d
sp
e
c
ies
-
d
e
t
a
il
fe
a
tu
re
s
in
t
o
a
u
n
ifi
e
d
e
m
b
e
d
d
in
g
sp
a
c
e
,
e
n
a
b
li
n
g
m
o
r
e
ffe
c
ti
v
e
c
las
sifica
ti
o
n
.
Ad
d
it
i
o
n
a
ll
y
,
t
h
e
fra
m
e
wo
rk
a
p
p
li
e
s
ima
g
e
-
sp
e
c
ies
p
ro
to
ty
p
e
re
g
u
lari
z
a
ti
o
n
(IS
P
R)
a
n
d
sp
e
c
ies
-
d
e
tail
p
ro
to
ty
p
e
re
g
u
lariz
a
ti
o
n
(S
DPR)
to
fu
rt
h
e
r
e
n
h
a
n
c
e
t
h
e
c
las
sifica
ti
o
n
a
c
c
u
ra
c
y
b
y
re
g
u
lariz
in
g
th
e
t
u
n
a
b
le
we
ig
h
ts
b
a
se
d
o
n
p
ro
t
o
t
y
p
e
a
li
g
n
m
e
n
t.
Th
e
p
ro
p
o
se
d
d
e
e
p
lea
rn
i
n
g
(DL)
m
o
d
e
l
is
e
v
a
lu
a
te
d
a
g
a
in
st
trad
it
i
o
n
a
l
m
a
c
h
in
e
lea
rn
in
g
(M
L)
m
e
th
o
d
s,
s
u
c
h
a
s
ra
n
d
o
m
fo
re
st
(RF
)
,
a
n
d
d
e
m
o
n
stra
tes
su
p
e
rio
r
p
e
rfo
rm
a
n
c
e
o
n
k
e
y
m
e
tri
c
s,
in
c
lu
d
i
n
g
a
c
c
u
ra
c
y
,
p
re
c
isi
o
n
,
re
c
a
ll
,
a
n
d
F
1
-
sc
o
re
.
K
ey
w
o
r
d
s
:
C
lim
ate
ch
an
g
e
I
m
ag
e
-
s
p
ec
ies p
r
o
t
o
ty
p
e
r
eg
u
lar
izatio
n
Mu
ltimo
d
al
co
n
tr
asti
v
e
lear
n
in
g
f
r
am
ewo
r
k
Mu
ltimo
d
al
s
p
ec
ies
-
im
ag
e
r
eg
u
lar
izatio
n
Po
llin
atio
n
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
:
Po
o
ja
Had
im
an
e
Dep
ar
tm
en
t o
f
C
o
m
p
u
ter
Scie
n
ce
an
d
E
n
g
in
ee
r
in
g
,
B
ap
u
ji
I
n
s
titu
te
o
f
E
n
g
in
ee
r
in
g
a
n
d
T
e
ch
n
o
lo
g
y
(
Af
f
iliated
to
Vis
v
esv
ar
ay
a
T
e
ch
n
o
lo
g
ical
Un
iv
er
s
ity
,
B
elag
av
i)
Dav
an
g
er
e,
I
n
d
ia
E
m
ail: p
o
o
jah
_
1
2
@
r
ed
if
f
m
ail.
co
m
1.
I
NT
RO
D
UCT
I
O
N
Po
llin
ato
r
s
p
lay
a
n
ess
en
tial
r
o
le
in
m
ai
n
tain
in
g
b
io
d
i
v
er
s
ity
an
d
en
s
u
r
i
n
g
a
g
r
icu
ltu
r
al
p
r
o
d
u
ctiv
ity
,
with
o
v
er
8
0
%
o
f
f
lo
wer
in
g
p
lan
ts
an
d
7
5
%
o
f
g
lo
b
al
f
o
o
d
cr
o
p
s
r
ely
in
g
o
n
p
o
llin
atio
n
.
Ho
wev
er
,
p
o
llin
ato
r
p
o
p
u
latio
n
s
ar
e
u
n
d
er
th
r
ea
t
d
u
e
to
en
v
ir
o
n
m
en
tal
ch
a
n
g
e
s
,
h
ab
itat
d
estru
ctio
n
,
an
d
cli
m
ate
ch
an
g
e,
wh
ich
p
o
s
e
s
ig
n
if
ican
t
r
is
k
s
to
b
o
th
ec
o
s
y
s
tem
s
an
d
f
o
o
d
s
ec
u
r
ity
.
T
r
ad
itio
n
al
m
eth
o
d
s
o
f
id
en
tif
y
in
g
an
d
class
if
y
in
g
p
o
llin
ato
r
s
p
ec
ies
ar
e
lab
o
r
-
in
ten
s
iv
e
an
d
lim
ite
d
b
y
th
e
d
ec
r
ea
s
in
g
n
u
m
b
er
o
f
tax
o
n
o
m
is
ts
.
T
h
ese
ch
allen
g
es
ca
ll
f
o
r
a
d
v
an
ce
d
s
o
lu
tio
n
s
th
at
lev
er
ag
e
a
r
tific
ial
in
tellig
en
ce
(
AI
)
an
d
d
ee
p
lear
n
in
g
(
DL
)
to
au
to
m
ate
th
e
class
if
icatio
n
p
r
o
ce
s
s
,
p
r
o
v
id
i
n
g
ac
cu
r
ate
an
d
s
ca
lab
le
s
y
s
tem
s
f
o
r
m
o
n
ito
r
in
g
p
o
llin
ato
r
s
p
ec
ies.
T
h
is
p
ap
er
in
tr
o
d
u
ce
s
a
n
o
v
el
DL
f
r
am
ewo
r
k
th
at
in
teg
r
ates
m
u
ltimo
d
al
d
ata,
in
clu
d
in
g
v
is
u
al
an
d
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
1
5
,
No
.
3
,
J
u
n
e
2
0
2
6
:
2
6
8
5
-
2
6
9
7
2686
en
v
ir
o
n
m
en
tal
f
ac
t
o
r
s
,
to
en
h
an
ce
p
o
llin
ato
r
class
if
icatio
n
an
d
v
u
ln
er
a
b
ilit
y
ass
ess
m
en
t,
ad
d
r
ess
in
g
a
cr
itical
g
ap
in
ec
o
lo
g
ical
s
tu
d
ies an
d
c
o
n
s
er
v
atio
n
e
f
f
o
r
ts
.
I
n
n
at
u
r
al
e
n
v
ir
o
n
m
en
ts
,
m
a
n
y
an
im
als
ar
e
in
ter
s
p
ec
if
ically
b
io
tic
with
p
lan
ts
,
a
ca
s
e
o
f
h
er
b
iv
o
r
es
an
d
p
o
llin
ato
r
s
.
Plan
ts
h
av
in
g
d
ef
en
s
iv
e
s
y
s
tem
s
ar
e
p
lace
d
u
n
d
er
h
ea
v
y
s
elec
tiv
e
p
r
ess
u
r
es
f
r
o
m
h
er
b
iv
o
r
es
th
at
ea
t
p
lan
t
m
ater
ial
o
n
a
lar
g
e
s
ca
le
b
y
h
er
b
iv
o
r
es.
T
h
ese
ty
p
es
o
f
d
ef
en
s
es
v
ar
y
i
n
ch
ar
ac
ter
is
tics
b
ased
o
n
th
e
ch
em
ical
s
en
s
itiv
ity
an
d
t
h
e
f
ee
d
in
g
b
e
h
av
io
r
s
o
f
h
er
b
i
v
o
r
es
o
f
v
ar
i
o
u
s
s
p
ec
ies.
Po
llin
ato
r
s
,
o
n
th
e
o
th
er
h
an
d
,
b
y
m
ea
n
s
o
f
m
u
tu
alis
tic
in
ter
ac
tio
n
s
with
p
lan
ts
,
will
s
elec
t
s
o
m
e
p
h
y
s
ical
tr
ait
a
n
d
s
ev
er
al
s
ig
n
als
o
f
in
cr
ea
s
ed
s
u
cc
ess
o
f
th
e
p
o
l
len
tr
an
s
p
o
r
t.
Selectio
n
p
r
ess
u
r
es
b
y
d
if
f
e
r
en
t
p
o
llin
at
o
r
s
ca
n
m
ain
tain
th
e
d
if
f
er
en
t
ev
o
lu
tio
n
in
p
lan
t
p
o
p
u
latio
n
s
an
d
s
p
ec
ies
s
in
ce
th
ey
p
o
s
s
ess
d
if
f
er
en
t
p
h
y
s
ical
f
ea
tu
r
es
an
d
s
en
s
o
r
y
p
r
ef
er
en
ce
s
.
T
h
e
ec
o
lo
g
ical
c
o
m
p
o
n
en
ts
ar
e
in
ter
co
n
n
ec
te
d
d
if
f
er
e
n
tly
;
th
e
s
o
il
ty
p
e
a
n
d
th
e
p
la
n
t
-
an
im
al
in
ter
ac
tio
n
s
ar
e
s
o
m
e
ex
a
m
p
les
o
f
th
is
.
I
n
a
p
r
o
ce
s
s
o
f
p
h
en
o
t
y
p
ic
p
last
icity
,
th
e
s
o
i
l
ty
p
e
h
as
a
h
u
g
e
in
f
lu
en
ce
o
n
th
e
p
lan
t
p
h
y
s
io
lo
g
y
an
d
m
o
r
p
h
o
lo
g
y
,
an
d
s
u
b
s
eq
u
en
tly
,
it
im
p
ac
ts
th
e
p
lan
t
-
p
o
llin
ato
r
an
d
p
lan
t
-
h
er
b
iv
o
r
e
in
ter
ac
tio
n
s
.
Ma
s
s
iv
e
d
ea
th
s
o
f
in
s
ec
t
s
h
a
v
e
b
ee
n
r
ep
o
r
ted
ev
en
in
th
e
s
o
-
ca
lled
p
r
is
tin
e
en
v
ir
o
n
m
en
ts
.
T
h
ese
r
ep
o
r
ts
co
n
tr
ib
u
te
to
th
e
ex
is
tin
g
is
s
u
e
o
f
b
io
d
iv
er
s
ity
o
n
th
e
g
lo
b
e.
T
h
er
e
is
m
o
u
n
tin
g
ev
id
en
ce
to
in
d
icate
th
at
i
n
s
ec
t
lo
s
s
es
ar
e
d
r
am
atic
b
o
t
h
in
ter
m
s
o
f
tax
o
n
o
m
y
an
d
g
eo
g
r
a
p
h
y
.
Simp
le
p
r
o
ce
s
s
es
in
th
e
ec
o
s
y
s
tem
,
s
u
ch
as
p
o
llin
atio
n
,
d
ec
o
m
p
o
s
itio
n
,
an
d
m
an
a
g
em
en
t
o
f
b
io
lo
g
ical
p
ests
,
h
av
e
m
an
y
s
p
ec
ies
th
at
r
ely
o
n
th
e
m
,
s
u
ch
as
h
u
m
an
b
ein
g
s
.
T
h
e
lo
s
s
o
f
th
ese
p
r
o
ce
s
s
es
is
a
s
er
io
u
s
th
r
ea
t
to
th
e
s
tab
ilit
y
o
f
th
e
ec
o
s
y
s
tem
s
an
d
b
io
d
iv
er
s
ity
[
1
]
–
[
3
]
.
I
n
s
ec
t p
o
llin
ato
r
s
ar
e
v
ital
to
th
e
welf
ar
e
o
f
th
e
ter
r
estrial
ec
o
s
y
s
tem
as
well
as
h
u
m
an
li
f
e.
B
y
a
c
o
u
n
t o
f
8
%
,
o
f
all
th
e
1
,
4
0
0
cr
o
p
s
p
ec
ies
g
r
o
wn
w
o
r
ld
wid
e
to
p
r
o
v
id
e
f
o
o
d
to
h
u
m
an
s
an
d
p
la
n
t
-
b
as
ed
in
d
u
s
tr
ial
p
r
o
d
u
cts,
an
im
al
p
o
llin
atio
n
is
r
eq
u
ir
ed
.
Po
llin
a
to
r
s
ar
e
r
eq
u
ir
ed
to
en
s
u
r
e
th
at
ec
o
s
y
s
tem
s
ar
e
p
r
eser
v
ed
,
f
o
o
d
p
r
o
d
u
ctio
n
,
an
d
n
atu
r
al
r
eso
u
r
ce
s
a
r
e
p
r
eser
v
ed
th
r
o
u
g
h
ec
o
s
y
s
tem
s
er
v
ices.
As
h
ig
h
ly
v
alu
ab
le
r
eso
u
r
ce
s
,
b
o
th
m
an
ag
ed
an
d
wild
p
o
p
u
latio
n
s
o
f
p
o
llin
ato
r
s
ar
e
f
ac
in
g
s
ev
e
r
e
p
r
ess
u
r
es
with
a
s
ig
n
if
ican
t
c
o
n
s
eq
u
e
n
ce
t
o
th
e
wo
r
ld
'
s
f
o
o
d
s
ec
u
r
ity
.
T
h
e
d
ec
lin
e
in
d
iv
er
s
ity
an
d
th
e
n
u
m
b
er
o
f
in
s
ec
ts
is
o
n
e
o
f
th
e
m
o
s
t
s
ig
n
if
ican
t
is
s
u
es
th
at
h
u
m
an
k
in
d
f
ac
es
to
d
ay
.
T
h
e
in
s
ec
ts
o
f
f
er
s
p
ec
ial
im
p
ed
im
en
ts
to
p
o
p
u
latio
n
s
u
r
v
eillan
ce
d
u
e
to
th
eir
s
m
all
s
ize
an
d
th
e
lab
o
r
i
o
u
s
n
atu
r
e
o
f
m
o
r
e
tr
ad
itio
n
al
m
ec
h
an
is
m
s
o
f
t
h
e
ir
co
llectio
n
an
d
id
en
tific
atio
n
.
I
t
is
in
cr
ea
s
in
g
ly
b
ec
o
m
in
g
d
if
f
icu
lt
to
id
e
n
tify
th
e
p
o
llin
ato
r
s
,
p
ar
ticu
lar
l
y
b
ee
s
,
tax
o
n
o
m
ically
,
d
u
e
to
th
e
d
win
d
lin
g
n
u
m
b
er
o
f
i
n
s
ec
t
tax
o
n
o
m
is
ts
.
Mo
r
eo
v
er
,
o
n
e
ca
n
n
o
t
f
in
d
m
an
y
r
ea
d
ily
ac
ce
s
s
ib
le
r
eso
u
r
ce
s
to
ed
u
ca
te
th
e
g
en
er
al
p
o
p
u
latio
n
an
d
f
u
t
u
r
e
g
en
er
atio
n
s
o
n
th
e
a
b
ilit
y
to
id
en
tify
d
if
f
e
r
en
t in
s
ec
ts
an
d
p
o
l
lin
ato
r
s
p
ec
ies.
Oth
er
f
ac
to
r
s
th
at
r
esu
lt
in
th
e
lo
s
s
o
f
p
o
llin
ato
r
s
h
av
e
b
ee
n
s
tatis
tica
lly
m
o
d
eled
o
f
o
p
p
o
r
tu
n
is
tic
d
ata
an
d
f
ield
s
tu
d
ies,
s
u
ch
as
p
esti
cid
e
ex
p
o
s
u
r
e,
la
n
d
u
s
e,
an
d
clim
ate
ch
an
g
e.
No
n
eth
eless
,
o
n
e
ca
n
n
o
t
tell
h
o
w
r
elev
an
t
ea
c
h
o
f
th
em
is
co
n
ce
r
n
i
n
g
th
e
ca
u
s
e
o
f
d
ec
r
ea
s
es
at
v
ar
y
in
g
s
izes
u
n
less
th
er
e
is
lo
n
g
-
te
r
m
ab
u
n
d
a
n
ce
d
ata.
I
n
ad
d
itio
n
,
th
e
p
r
im
ar
y
co
n
ce
r
n
o
f
th
e
r
eg
u
latio
n
o
f
th
e
wild
p
o
llin
a
to
r
s
in
th
e
Un
ited
Kin
g
d
o
m
(
UK
)
an
d
th
e
r
est
o
f
th
e
co
u
n
tr
ies
is
th
e
s
u
b
s
id
ie
s
o
f
f
er
ed
to
f
a
r
m
er
s
u
n
d
er
t
h
e
ag
r
i
-
en
v
ir
o
n
m
en
t
s
ch
em
e.
T
h
ese
s
u
b
s
id
ies
p
r
o
m
o
te
th
e
ap
p
licatio
n
o
f
ag
r
i
-
e
n
v
ir
o
n
m
en
t
m
eth
o
d
s
,
in
clu
d
in
g
f
ield
m
ar
g
in
s
th
a
t
ar
e
h
ig
h
in
f
lo
wer
s
a
n
d
h
ed
g
e
p
r
eser
v
atio
n
.
T
h
e
b
en
ef
ici
al
ef
f
ec
ts
h
av
e
b
ee
n
r
ea
lized
in
th
ese
e
f
f
o
r
ts
i
n
r
ed
u
cin
g
lo
ca
l
s
ca
les
with
r
esp
ec
t
to
t
h
e
p
o
llin
ato
r
v
ar
iety
,
ac
tiv
ity
,
an
d
p
o
llin
atio
n
s
er
v
ic
es,
b
u
t
m
o
r
e
li
k
ely
,
th
ey
ar
e
s
h
o
r
t
-
liv
ed
ter
m
i
n
i o
f
f
lo
wer
v
is
its
,
n
o
t p
er
m
a
n
en
t p
o
p
u
latio
n
i
n
cr
ea
s
e
[
4
]
.
Ag
r
icu
ltu
r
al
tech
n
o
lo
g
y
h
as
a
d
v
an
ce
d
s
ig
n
if
ican
tly
,
b
u
t
th
er
e
ar
e
s
till
n
o
s
y
s
tem
s
d
esig
n
ed
ex
p
r
ess
ly
f
o
r
th
e
au
to
m
ated
d
etec
tio
n
a
n
d
class
if
icatio
n
o
f
p
o
llin
ato
r
s
.
T
h
e
m
ajo
r
ity
o
f
cu
r
r
e
n
t a
p
p
r
o
ac
h
es tar
g
et
in
s
ec
t
s
an
d
p
ests
;
th
u
s
,
cr
ea
tiv
e
s
o
lu
tio
n
s
ar
e
n
ee
d
ed
to
s
o
lv
e
th
is
u
r
g
en
t
p
o
llin
ato
r
-
r
elate
d
is
s
u
e
.
I
d
en
tific
atio
n
an
d
ca
teg
o
r
izatio
n
o
f
p
o
llin
ato
r
s
p
ec
ies
ar
e
ess
en
tial
f
o
r
ec
o
lo
g
ical
s
tu
d
ies
an
d
ag
r
icu
ltu
r
al
u
s
es.
B
ec
au
s
e
th
ey
allo
w
m
an
y
f
lo
wer
in
g
p
la
n
ts
to
r
ep
r
o
d
u
ce
,
p
o
llin
ato
r
s
—
in
s
ec
ts
lik
e
b
ee
s
,
b
u
tter
f
lies
,
an
d
m
o
th
s
—
ar
e
cr
u
cial
to
m
ain
tain
in
g
b
i
o
d
iv
er
s
ity
a
n
d
ec
o
s
y
s
tem
s
.
Fo
r
s
ev
er
al
r
e
aso
n
s
,
th
ese
s
p
ec
ies
m
u
s
t
b
e
p
r
ec
is
ely
class
if
ied
an
d
id
en
tifie
d
[
5
]
.
C
o
n
s
er
v
atio
n
o
f
b
i
o
d
iv
er
s
ity
:
o
v
er
8
0
%
o
f
f
lo
wer
in
g
p
lan
ts
r
ely
o
n
p
o
llin
ato
r
s
f
o
r
r
ep
r
o
d
u
ctio
n
,
h
ig
h
lig
h
tin
g
th
eir
cr
itical
r
o
le
in
m
ai
n
tain
in
g
b
io
d
iv
er
s
ity
.
B
y
ac
c
u
r
ately
id
en
tify
i
n
g
an
d
ca
teg
o
r
izin
g
p
o
llin
ato
r
s
p
ec
ies,
s
cien
tis
ts
ca
n
m
o
n
ito
r
f
lu
ctu
atio
n
s
in
s
p
ec
ies
p
o
p
u
latio
n
s
a
n
d
en
h
a
n
ce
th
eir
co
m
p
r
eh
e
n
s
io
n
o
f
ec
o
lo
g
ical
d
iv
er
s
ity
.
C
o
n
s
er
v
atio
n
in
itiat
iv
es
aim
ed
at
p
r
o
tectin
g
t
h
r
e
aten
ed
s
p
ec
ies
an
d
en
s
u
r
in
g
th
e
o
v
er
all
h
ea
lth
o
f
ec
o
s
y
s
tem
s
ar
e
cr
itical.
Ag
r
icu
ltu
r
al
p
r
o
d
u
ctiv
ity
:
p
o
llin
ato
r
s
ar
e
r
esp
o
n
s
ib
le
f
o
r
th
e
s
u
cc
ess
f
u
l
p
r
o
d
u
ctio
n
o
f
ap
p
r
o
x
im
ately
7
5
%
o
f
th
e
wo
r
ld
'
s
f
o
o
d
cr
o
p
s
.
T
h
e
ess
en
tial
co
m
p
o
n
e
n
ts
f
o
r
h
u
m
an
n
u
tr
itio
n
an
d
f
o
o
d
s
ec
u
r
ity
en
c
o
m
p
ass
f
r
u
its
,
v
e
g
eta
b
les,
n
u
ts
,
an
d
s
ee
d
s
.
Acc
u
r
at
e
class
if
icatio
n
an
d
m
o
n
ito
r
in
g
o
f
p
o
llin
ato
r
s
p
ec
i
es
co
n
tr
ib
u
te
to
ass
ess
in
g
th
e
h
ea
lth
o
f
th
ese
p
o
p
u
latio
n
s
,
e
n
s
u
r
in
g
th
eir
a
b
ilit
y
to
p
r
o
v
i
d
e
ess
en
tial
s
er
v
ices.
Un
d
er
s
tan
d
in
g
th
e
e
f
f
icien
cy
o
f
v
ar
io
u
s
s
p
ec
ies
as
p
o
llin
ato
r
s
f
o
r
s
p
ec
if
ic
cr
o
p
s
en
ab
les
f
ar
m
er
s
an
d
ag
r
ic
u
ltu
r
alis
ts
to
en
h
an
ce
th
eir
p
o
ll
in
atio
n
s
tr
ateg
ies,
th
er
eb
y
in
cr
ea
s
in
g
b
o
th
cr
o
p
y
ield
s
an
d
q
u
ality
.
E
c
o
s
y
s
tem
s
er
v
ices
an
d
f
u
n
ctio
n
ality
:
p
o
l
lin
ato
r
s
p
lay
a
c
r
itical
r
o
le
b
e
y
o
n
d
a
g
r
icu
ltu
r
e
b
y
s
u
p
p
o
r
tin
g
ess
en
tial
ec
o
s
y
s
tem
s
er
v
ices,
in
clu
d
in
g
th
e
r
esto
r
atio
n
o
f
n
atu
r
al
h
ab
itats
an
d
th
e
en
h
an
ce
m
e
n
t
o
f
g
en
etic
d
iv
er
s
ity
am
o
n
g
p
lan
t
s
p
ec
ies.
Acc
u
r
ate
id
en
t
if
icatio
n
an
d
class
if
icatio
n
o
f
p
o
llin
ato
r
s
p
ec
ies
en
ab
le
r
esear
ch
er
s
to
m
o
n
ito
r
ec
o
s
y
s
tem
f
u
n
ctio
n
in
g
an
d
ass
es
s
th
e
co
n
d
itio
n
o
f
n
at
u
r
al
en
v
ir
o
n
m
en
ts
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
A
n
o
ve
l d
ee
p
lea
r
n
in
g
-
b
a
s
ed
p
o
llin
a
to
r
s
p
ec
ies cla
s
s
ifica
tio
n
u
s
in
g
th
e
mu
ltimo
d
a
l
…
(
P
o
o
ja
Ha
d
ima
n
e)
2687
T
h
e
eq
u
ilib
r
iu
m
o
f
ec
o
s
y
s
tem
s
is
co
n
tin
g
en
t
u
p
o
n
th
is
k
n
o
wled
g
e,
p
ar
ticu
la
r
ly
i
n
th
e
c
o
n
tex
t
o
f
en
v
ir
o
n
m
en
tal
ch
a
n
g
es
s
u
ch
as
p
o
llu
tio
n
,
h
ab
itat
d
eg
r
a
d
atio
n
,
a
n
d
clim
ate
ch
an
g
e.
C
lim
ate
ch
an
g
e
im
p
ac
t
ass
es
s
m
en
t:
th
e
ev
alu
atio
n
o
f
clim
ate
ch
a
n
g
e
ef
f
ec
ts
o
n
ec
o
s
y
s
tem
s
n
ec
ess
itates
ac
cu
r
ate
class
if
icatio
n
an
d
id
en
tific
atio
n
o
f
p
o
llin
ato
r
s
p
e
cies.
T
h
e
d
is
tr
ib
u
tio
n
an
d
b
e
h
av
io
r
o
f
p
o
llin
ato
r
s
ar
e
in
f
lu
e
n
ce
d
b
y
f
lu
ctu
atio
n
s
in
tem
p
er
at
u
r
e,
p
r
ec
ip
itatio
n
,
an
d
v
a
r
io
u
s
clim
atic
f
ac
to
r
s
.
Mo
n
ito
r
in
g
s
p
ec
ies
d
iv
er
s
ity
an
d
p
o
p
u
latio
n
d
y
n
am
ics
en
ab
les
r
esear
ch
er
s
to
d
etec
t
ea
r
ly
s
ig
n
s
o
f
clim
ate
-
in
d
u
ce
d
s
tr
ess
in
ec
o
s
y
s
tem
s
an
d
im
p
lem
en
t
p
r
o
tectiv
e
m
ea
s
u
r
es a
cc
o
r
d
in
g
ly
.
Po
llin
atio
n
m
an
ag
em
e
n
t
an
d
co
n
s
er
v
atio
n
s
tr
ateg
ies:
ef
f
ec
tiv
e
m
an
ag
e
m
en
t
o
f
p
o
llin
ato
r
p
o
p
u
latio
n
s
is
ess
en
tial
f
o
r
s
u
s
tain
in
g
th
eir
p
o
llin
atio
n
r
o
les
in
b
o
th
n
at
u
r
al
ec
o
s
y
s
te
m
s
an
d
ag
r
ic
u
ltu
r
al
en
v
ir
o
n
m
en
ts
.
T
h
e
id
en
tific
at
io
n
an
d
class
if
icatio
n
o
f
p
o
ll
in
ato
r
s
p
ec
ies
co
n
tr
i
b
u
te
to
th
e
f
o
r
m
u
latio
n
o
f
tar
g
eted
c
o
n
s
er
v
atio
n
s
tr
ateg
i
es.
T
h
ese
s
tr
ateg
ies
en
c
o
m
p
a
s
s
th
e
p
r
eser
v
atio
n
o
f
c
r
itical
p
o
llin
ato
r
h
ab
itats
,
h
ab
itat
r
esto
r
atio
n
ef
f
o
r
ts
,
a
n
d
th
e
cr
ea
tio
n
o
f
en
v
ir
o
n
m
en
ts
th
at
s
u
p
p
o
r
t
p
o
llin
ato
r
p
o
p
u
latio
n
s
.
T
h
is
ap
p
r
o
ac
h
en
s
u
r
es
th
e
av
ailab
il
ity
o
f
p
o
llin
ato
r
s
,
wh
ic
h
is
ess
en
tial
f
o
r
th
e
m
ain
ten
a
n
ce
o
f
n
atu
r
al
ec
o
s
y
s
tem
s
an
d
th
e
s
u
s
tain
ab
ilit
y
o
f
th
e
f
o
o
d
s
u
p
p
ly
.
C
o
n
tr
o
l
o
f
in
v
asiv
e
s
p
ec
ies:
th
e
m
an
ag
em
e
n
t
o
f
in
v
asiv
e
s
p
ec
ies
th
at
p
o
s
e
a
th
r
ea
t
to
n
ativ
e
p
o
llin
ato
r
p
o
p
u
latio
n
s
is
s
ig
n
if
ican
tly
r
elian
t
o
n
th
e
p
r
ec
is
e
id
en
tific
atio
n
o
f
p
o
llin
ato
r
s
p
ec
ies.
I
n
v
asiv
e
s
p
ec
ies
h
av
e
th
e
p
o
te
n
tial
to
alt
er
p
o
llin
atio
n
n
etwo
r
k
s
,
lead
i
n
g
to
d
is
r
u
p
ti
o
n
s
in
lo
ca
l
ec
o
s
y
s
tem
s
an
d
t
h
e
o
u
tco
m
p
etin
g
o
f
n
ativ
e
p
o
llin
ato
r
s
.
Ma
n
ag
em
en
t
p
lan
s
aim
ed
a
t
co
n
s
er
v
in
g
n
ativ
e
p
o
llin
ato
r
s
an
d
m
ain
tain
in
g
ec
o
lo
g
ical
b
alan
ce
ca
n
b
e
d
ev
el
o
p
ed
th
r
o
u
g
h
th
e
id
en
tific
atio
n
an
d
m
o
n
ito
r
in
g
o
f
th
ese
s
p
ec
ies.
Fig
u
r
e
1
s
h
o
w
s
an
ex
am
p
le
o
f
a
p
lan
t
s
p
e
cies
(
A
r
a
lia
ce
a
e
—
Oreo
p
a
n
a
x
ca
p
ita
tu
s
(
Ja
c
q
.
)
Dec
n
e.
and
P
l
a
n
ch
.
)
with
p
h
en
o
ty
p
ic
g
e
n
er
aliza
tio
n
v
is
ite
d
b
y
d
i
v
er
s
e
s
m
all
in
s
ec
ts
.
Fi
g
u
r
e
1
(
a)
s
h
o
ws
th
e
b
u
tter
f
l
y
,
Fig
u
r
e
1
(
b
)
s
h
o
ws
th
e
s
tin
g
less
b
ee
s
an
d
l
ep
id
o
p
ter
a,
Fig
u
r
e
1
(
c
)
s
h
o
ws
th
e
f
ly
,
an
d
Fig
u
r
e
1
(
d
)
s
h
o
ws
th
e
wasp
.
(
a)
(
b
)
(
c)
(
d
)
Fig
u
r
e
1
.
E
x
am
p
le
o
f
a
p
la
n
t sp
ec
ies
(
A
r
a
lia
ce
a
e
—
Oreo
p
a
n
a
x
ca
p
ita
tu
s
(
J
a
cq
.
)
Dec
n
e.
and
P
la
n
ch
.
)
with
p
h
en
o
t
y
p
ic
g
e
n
er
aliza
tio
n
v
is
i
ted
b
y
d
iv
er
s
e
s
m
all
in
s
ec
ts
f
o
r
(
a)
b
u
tter
f
ly
,
(
b
)
s
tin
g
less
b
e
es a
n
d
lep
id
o
p
ter
a,
(
c)
f
ly
,
a
n
d
(
d
)
wasp
Po
llin
ato
r
s
p
ec
ies
id
en
tific
ati
o
n
an
d
ca
teg
o
r
izatio
n
ar
e
ch
an
g
in
g
as
a
r
esu
lt
o
f
ad
v
a
n
c
es
in
AI
,
m
ac
h
in
e
lear
n
i
n
g
(
ML
)
,
an
d
DL
.
T
h
ese
tech
n
o
lo
g
ies
im
p
r
o
v
e
ac
cu
r
ac
y
an
d
m
ak
e
it
p
o
s
s
ib
le
to
ac
cu
r
ately
id
en
tify
s
p
ec
ies
t
h
at
ca
n
b
e
ch
allen
g
in
g
t
o
d
if
f
er
e
n
tiate
m
an
u
ally
b
y
a
n
aly
zin
g
v
ast d
atasets
.
R
ea
l
-
tim
e
r
esu
lts
ar
e
cr
itical
to
lar
g
e
-
s
ca
le
ec
o
lo
g
ical
s
tu
d
ies
an
d
a
g
r
icu
ltu
r
e,
an
d
AI
an
d
ML
ca
n
h
el
p
b
y
au
t
o
m
atin
g
t
h
e
id
en
tific
atio
n
p
r
o
ce
s
s
.
C
o
m
p
ilin
g
b
eh
a
v
io
r
al
,
en
v
ir
o
n
m
en
t
al,
an
d
v
is
u
al
d
ata
t
o
g
eth
er
m
ak
es
it
ea
s
ier
to
co
m
p
r
eh
e
n
d
p
o
llin
ato
r
s
p
ec
ies
an
d
h
o
w
t
h
ey
in
te
r
ac
t
with
th
eir
s
u
r
r
o
u
n
d
in
g
s
.
T
h
ese
c
o
m
p
o
n
en
ts
ar
e
ess
en
tial
f
o
r
f
o
r
ec
asti
n
g
f
u
tu
r
e
p
atter
n
s
,
ev
alu
atin
g
h
o
w
p
o
llin
ato
r
s
m
ay
b
e
af
f
ec
ted
b
y
clim
ate
ch
an
g
e,
an
d
d
ev
el
o
p
in
g
co
n
s
er
v
atio
n
p
lan
s
.
Glo
b
al
co
n
s
er
v
atio
n
ef
f
o
r
ts
m
ay
b
e
s
u
p
p
o
r
ted
b
y
AI
-
d
r
i
v
en
s
y
s
tem
s
th
at
g
u
ar
an
tee
co
n
s
is
ten
cy
,
m
in
im
ize
h
u
m
an
er
r
o
r
,
an
d
ca
n
b
e
s
ca
led
to
m
o
n
ito
r
p
o
llin
ato
r
n
u
m
b
er
s
ac
r
o
s
s
lar
g
e
ter
r
ito
r
ies
[
6
]
–
[
9
]
.
T
h
e
p
o
llin
atio
n
v
u
ln
e
r
ab
ilit
y
i
n
d
ex
(
PVI
)
is
o
n
e
im
p
o
r
tan
t
t
o
o
l
f
o
r
ass
ess
in
g
th
e
p
o
ten
tial
d
ec
lin
e
in
p
o
llin
atio
n
s
er
v
ices.
Sev
er
al
p
lan
ts
,
n
o
tab
ly
th
o
s
e
th
at
ar
e
k
e
y
to
ag
r
icu
lt
u
r
al
y
ield
a
n
d
th
e
ec
o
s
y
s
tem
,
r
ely
o
n
p
o
llin
atio
n
s
er
v
ices
f
o
r
th
eir
r
ep
r
o
d
u
ctio
n
p
r
o
ce
s
s
.
Ov
er
th
e
p
ast
f
ew
y
ea
r
s
,
DL
h
as
b
ec
o
m
e
a
m
ajo
r
tr
en
d
in
tack
lin
g
co
m
p
le
x
asp
ec
ts
o
f
cl
ass
if
icatio
n
,
esp
ec
ially
th
o
s
e
r
eq
u
ir
in
g
lar
g
e
an
d
v
ar
ied
ty
p
e
s
o
f
d
atasets
,
wh
er
e
th
is
tech
n
iq
u
e
an
d
its
v
ar
ian
t
s
h
av
e
b
ec
o
m
e
esp
ec
ially
p
o
p
u
lar
.
T
h
e
f
ac
t
th
at
DL
m
o
d
els
ar
e
ca
p
ab
le
o
f
p
r
o
v
id
i
n
g
b
o
t
h
b
in
ar
y
an
d
m
u
lti
-
class
if
icatio
n
,
in
wh
ich
d
ata
is
d
iv
id
ed
in
to
two
cl
ass
es,
r
e
s
p
ec
tiv
ely
,
an
d
o
n
e
o
f
m
u
ltip
le
class
es,
r
e
s
p
ec
tiv
ely
,
is
o
n
e
o
f
th
e
s
tr
en
g
th
s
.
DL
in
th
e
ec
o
lo
g
ical
m
o
n
ito
r
in
g
co
n
te
x
t
ca
n
b
e
u
s
ed
to
ef
f
ec
tiv
ely
a
n
d
ef
f
i
cien
tly
s
o
lv
e
p
r
o
b
lem
s
o
f
id
e
n
tify
in
g
an
d
class
if
y
in
g
s
p
ec
i
es.
T
h
e
m
o
d
els
will
b
e
ab
le
to
lea
r
n
m
o
r
e
d
etaile
d
p
atter
n
s
in
v
is
u
al
an
d
tex
t
u
al
d
ata
u
s
in
g
lay
e
r
s
o
f
ab
s
tr
a
ctio
n
,
m
ea
n
in
g
th
at
th
ey
will
b
e
m
o
r
e
ac
cu
r
ate
t
h
an
th
e
o
th
er
m
eth
o
d
s
o
f
ML
.
DL
is
esp
ec
ially
ad
ap
tab
le
,
wh
ich
is
wh
y
it
is
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
1
5
,
No
.
3
,
J
u
n
e
2
0
2
6
:
2
6
8
5
-
2
6
9
7
2688
esp
ec
ially
ap
p
r
o
p
r
iate
in
s
o
lv
i
n
g
p
r
o
b
lem
s
s
u
ch
as
p
o
llin
ato
r
s
p
ec
ies
id
en
tific
atio
n
,
wh
er
e
a
b
in
ar
y
q
u
esti
o
n
(
p
o
llin
ato
r
v
s
.
n
o
n
-
p
o
llin
ato
r
)
n
ee
d
s
to
b
e
an
s
wer
ed
,
b
u
t
a
m
u
lticlas
s
o
n
e
(
wh
at
g
en
u
s
is
it)
h
as
to
b
e
p
o
s
ed
s
im
u
ltan
eo
u
s
ly
[
1
0
]
.
Po
llin
a
to
r
s
ar
e
cr
itical
to
m
ain
tain
in
g
b
io
d
iv
er
s
ity
an
d
s
u
p
p
o
r
tin
g
g
lo
b
al
f
o
o
d
p
r
o
d
u
ctio
n
,
as
o
v
e
r
8
0
%
o
f
f
lo
wer
in
g
p
lan
ts
an
d
7
5
%
o
f
th
e
wo
r
ld
'
s
f
o
o
d
cr
o
p
s
r
ely
o
n
an
im
al
p
o
llin
atio
n
.
Desp
ite
th
eir
im
p
o
r
tan
ce
,
p
o
l
lin
ato
r
p
o
p
u
latio
n
s
,
p
ar
ticu
lar
ly
b
ee
s
,
a
r
e
d
ec
lin
i
n
g
at
alar
m
in
g
r
ates
d
u
e
to
f
ac
to
r
s
lik
e
h
ab
itat
d
estru
ctio
n
,
p
esti
cid
e
u
s
e,
an
d
clim
ate
ch
an
g
e.
T
r
a
d
itio
n
al
m
eth
o
d
s
f
o
r
id
en
tify
i
n
g
an
d
m
o
n
ito
r
in
g
p
o
llin
ato
r
s
p
ec
ies
ar
e
lab
o
r
-
in
ten
s
iv
e
,
r
ely
h
ea
v
ily
o
n
ex
p
e
r
t
tax
o
n
o
m
is
ts
,
a
n
d
lack
s
ca
lab
ilit
y
,
wh
ich
lim
its
th
eir
a
b
ilit
y
to
t
r
ac
k
p
o
llin
ato
r
p
o
p
u
latio
n
s
o
n
a
lar
g
e
s
ca
le.
T
h
is
cr
ea
tes
a
p
r
ess
in
g
n
ee
d
f
o
r
au
to
m
ated
,
ac
cu
r
ate
,
an
d
s
ca
lab
le
s
o
lu
tio
n
s
to
m
o
n
ito
r
p
o
llin
ato
r
d
iv
er
s
ity
an
d
h
ea
lth
.
Ad
v
an
ce
s
in
AI
an
d
DL
o
f
f
er
p
r
o
m
is
in
g
to
o
l
s
to
o
v
er
co
m
e
th
ese
lim
itati
o
n
s
,
b
u
t
cu
r
r
en
t
m
o
d
els
o
f
ten
f
ail
to
in
teg
r
ate
m
u
ltip
le
d
ata
ty
p
es
(
s
u
ch
as
v
is
u
al
an
d
en
v
ir
o
n
m
en
tal
d
ata)
an
d
lack
th
e
p
r
ec
is
io
n
r
e
q
u
ir
ed
f
o
r
f
in
e
-
g
r
ain
ed
s
p
ec
ies cla
s
s
if
icatio
n
.
T
h
is
r
esear
ch
is
m
o
tiv
ated
b
y
th
e
u
r
g
en
t
n
ee
d
to
d
ev
elo
p
a
r
o
b
u
s
t,
AI
-
d
r
iv
e
n
f
r
am
ewo
r
k
t
h
at
ca
n
n
o
t
o
n
ly
au
t
o
m
ate
p
o
llin
ato
r
cl
ass
if
icatio
n
b
u
t
also
im
p
r
o
v
e
th
e
ac
cu
r
ac
y
o
f
p
r
ed
ictio
n
s
b
y
in
te
g
r
atin
g
m
u
ltimo
d
al
d
ata,
ad
d
r
ess
in
g
b
o
th
b
in
ar
y
an
d
m
u
lticlas
s
class
if
icatio
n
ch
allen
g
es
in
an
ec
o
lo
g
ical
co
n
tex
t.
R
esear
ch
co
n
tr
ib
u
tio
n
in
clu
d
e
s
:
i)
Mu
ltimo
d
al
s
p
ec
ies
-
im
ag
e
r
eg
u
lar
izatio
n
(
MSI
R
)
f
r
am
ewo
r
k
:
we
in
tr
o
d
u
ce
a
n
o
v
el
DL
f
r
am
ewo
r
k
,
th
e
MSI
R
,
th
at
in
teg
r
ates
b
o
t
h
v
i
s
u
al
an
d
s
p
ec
ies
-
d
etail
d
ata
t
o
im
p
r
o
v
e
class
if
icatio
n
p
e
r
f
o
r
m
an
ce
.
T
h
is
f
r
am
ewo
r
k
en
a
b
les
th
e
s
y
s
tem
to
h
an
d
le
b
o
th
b
in
ar
y
clas
s
if
icatio
n
(
p
o
llin
ato
r
v
s
.
n
o
n
-
p
o
llin
ato
r
)
an
d
m
u
lticlas
s
clas
s
if
icatio
n
(
class
if
y
in
g
b
ee
g
en
e
r
a)
m
o
r
e
ef
f
ec
t
iv
ely
th
an
tr
a
d
itio
n
al
m
eth
o
d
s
.
ii)
Pro
to
ty
p
e
-
d
r
iv
en
r
eg
u
lar
izatio
n
:
th
e
f
r
am
ew
o
r
k
em
p
lo
y
s
im
ag
e
-
s
p
ec
ies
p
r
o
to
ty
p
e
r
eg
u
lar
i
za
tio
n
(
I
SP
R
)
an
d
s
p
ec
ies
-
d
etail
p
r
o
to
t
y
p
e
r
eg
u
lar
izatio
n
(
SDPR
)
to
en
h
a
n
ce
th
e
m
o
d
el’
s
ac
cu
r
ac
y
.
T
h
ese
tech
n
iq
u
es
alig
n
th
e
tu
n
ab
le
weig
h
ts
o
f
t
h
e
m
o
d
el
b
ased
o
n
b
o
th
im
ag
e
an
d
s
p
ec
ies
p
r
o
to
ty
p
es,
lea
d
in
g
to
b
etter
g
en
er
ali
z
atio
n
ac
r
o
s
s
d
if
f
er
en
t
s
p
ec
ies an
d
en
v
ir
o
n
m
en
tal
c
o
n
d
itio
n
s
.
iii)
Su
p
er
io
r
p
er
f
o
r
m
an
ce
m
etr
ic
s
:
th
e
p
r
o
p
o
s
ed
m
o
d
el
o
u
t
p
er
f
o
r
m
s
tr
ad
itio
n
al
ML
m
et
h
o
d
s
,
s
u
ch
as
r
an
d
o
m
f
o
r
est
(
R
F)
,
ac
h
iev
in
g
an
ac
cu
r
ac
y
o
f
9
8
.
3
4
%,
s
ig
n
if
ican
tly
s
u
r
p
ass
in
g
th
e
b
a
s
elin
e
o
f
9
2
%.
T
h
is
s
h
o
wca
s
es
th
e
p
o
te
n
tial
o
f
DL
m
o
d
els
to
r
ev
o
l
u
tio
n
i
z
e
p
o
llin
ato
r
m
o
n
ito
r
in
g
b
y
o
f
f
er
in
g
m
o
r
e
p
r
ec
is
e
an
d
s
ca
lab
le
s
o
lu
tio
n
s
.
2.
RE
L
AT
E
D
WO
RK
I
n
th
e
r
ec
en
t
p
ast,
ad
v
an
ce
m
en
ts
in
DL
an
d
AI
h
av
e
en
h
an
ce
d
t
h
e
ca
p
ac
ity
to
r
e
co
g
n
ize
an
d
ca
teg
o
r
ize
in
s
ec
ts
an
d
o
th
er
s
p
ec
ies
with
in
th
e
ec
o
lo
g
y
.
Su
ch
ad
v
a
n
ce
s
u
n
d
er
s
co
r
e
th
e
im
p
o
r
tan
ce
o
f
s
u
ch
tech
n
o
lo
g
ies
in
th
e
ag
r
ic
u
ltu
r
al
an
d
f
o
o
d
s
ec
to
r
,
an
d
t
h
e
co
n
s
er
v
atio
n
o
f
b
io
d
iv
e
r
s
ity
.
Fo
r
ex
am
p
le,
in
r
esear
ch
[
1
1
]
,
in
v
esti
g
ated
o
b
j
ec
t r
ec
o
g
n
itio
n
alg
o
r
ith
m
s
in
v
id
eo
f
o
o
tag
e
to
a
p
p
ly
AI
s
y
s
te
m
s
to
d
etec
t th
is
tle
ca
ter
p
illar
s
(
V
a
n
ess
a
ca
r
d
u
i)
.
W
ith
a
2
,
4
1
6
-
p
h
o
to
d
ataset,
th
e
YOL
Ov
5
o
b
ject
id
en
tific
ati
o
n
ar
c
h
itectu
r
e
was
tr
ain
ed
with
s
cr
atch
an
d
tr
an
s
f
er
lear
n
in
g
m
o
d
els.
T
h
e
r
esu
lt
s
ca
n
b
e
tr
ea
ted
p
o
s
itiv
ely
as
it
in
d
icate
s
th
at
DL
is
an
ef
f
icien
t
m
eth
o
d
to
d
et
ec
t
in
s
ec
ts
.
Selv
ar
aj
et
a
l.
[
1
2
]
ap
p
lied
a
d
ee
p
co
n
v
o
l
u
tio
n
al
n
eu
r
al
n
etwo
r
k
(
DC
NN)
to
id
en
tify
th
e
d
is
ea
s
es
o
n
b
an
an
as
u
s
in
g
a
d
atase
t
o
f
o
v
er
1
8
,
0
0
0
p
r
e
-
s
cr
ee
n
e
d
b
an
an
a
p
h
o
to
s
an
d
f
o
u
n
d
th
at
tr
an
s
f
er
lear
n
in
g
c
an
b
e
ef
f
ec
tiv
e
in
ag
r
icu
ltu
r
e.
T
h
e
lev
el
o
f
ac
cu
r
ac
y
ac
h
iev
ed
r
a
n
g
ed
b
etwe
en
7
0
%
an
d
9
9
%.
T
h
e
p
u
r
p
o
s
e
o
f
s
u
b
s
eq
u
en
t
r
esear
ch
[
1
3
]
was
to
p
r
ed
ict
th
e
o
cc
u
r
r
en
ce
o
f
in
s
ec
ts
b
y
m
ain
tain
in
g
a
g
az
e
at
th
e
e
n
v
ir
o
n
m
en
tal
co
n
d
itio
n
s
,
s
u
ch
as
r
elativ
e
h
u
m
id
ity
an
d
a
ir
tem
p
er
atu
r
e.
A
n
ec
o
lo
g
ical
m
o
n
ito
r
in
g
u
s
in
g
ML
in
c
o
m
b
in
ati
o
n
with
th
e
en
v
ir
o
n
m
e
n
t
d
e
m
o
n
s
tr
ated
a
7
6
.
5
%
d
etec
tio
n
ac
cu
r
ac
y
.
I
n
a
p
ap
er
a
u
th
o
r
ed
b
y
Ma
r
k
o
v
ić
et
a
l.
[
1
4
]
,
t
h
e
p
r
o
b
lem
o
f
d
r
o
n
e
-
b
ased
in
s
ec
t
an
d
p
est
d
etec
tio
n
h
as
b
ee
n
ex
p
l
o
r
ed
,
w
h
ich
f
u
r
th
er
p
r
o
v
es
th
at
AI
tech
n
o
lo
g
y
ca
n
b
e
u
s
ed
in
v
ar
io
u
s
co
n
d
it
io
n
s
wh
en
ap
p
lied
to
m
o
n
ito
r
in
g
.
W
ith
th
e
Mo
b
ileNet
ar
ch
itectu
r
e,
th
e
r
ese
ar
ch
p
a
p
er
in
[
1
5
]
h
a
d
an
a
cc
u
r
ac
y
o
f
9
7
.
5
in
d
ef
in
in
g
a
n
d
ca
teg
o
r
izin
g
p
ests
an
d
to
m
ato
d
is
ea
s
es
am
o
n
g
in
s
ec
ts
.
T
h
is
in
d
icate
s
th
at
c
o
n
v
o
lu
ti
o
n
al
n
eu
r
al
n
etwo
r
k
s
(
C
NNs)
ca
n
h
a
v
e
ap
p
licatio
n
s
in
a
g
r
icu
ltu
r
e,
c
o
m
b
in
ed
with
tr
an
s
f
er
lear
n
i
n
g
,
with
p
o
s
s
ib
le
b
en
ef
its
.
T
h
is
s
tr
ateg
y
co
m
p
l
ies
with
th
e
r
ec
en
t
m
eth
o
d
s
o
f
em
p
lo
y
in
g
s
m
all
an
d
e
f
f
ic
ien
t
m
o
d
els
to
b
e
ap
p
lied
in
ag
r
ic
u
ltu
r
e
o
n
-
tim
e
.
T
ec
h
n
o
lo
g
ical
d
ev
elo
p
m
en
ts
h
av
e
r
esu
lted
in
th
e
in
n
o
v
ati
o
n
o
f
n
ew
way
s
o
f
au
to
m
ated
d
etec
tio
n
o
f
in
s
ec
ts
.
A
class
ical
m
o
d
el
o
f
th
e
c
o
m
b
in
atio
n
o
f
DL
an
d
o
p
tim
iz
atio
n
m
eth
o
d
s
is
th
e
au
to
m
ated
in
s
ec
t
d
etec
tio
n
an
d
class
if
icatio
n
ap
p
r
o
ac
h
[
1
6
]
s
o
f
twar
e.
I
t
e
x
tr
ac
ts
f
ea
tu
r
es
th
r
o
u
g
h
Den
s
eNe
t1
2
1
an
d
th
e
Pelican
o
p
tim
izatio
n
alg
o
r
ith
m
.
T
o
d
eter
m
in
e
in
s
ec
ts
,
Ass
ir
i
et
a
l.
[
1
7
]
p
r
o
p
o
s
ed
a
n
o
p
tim
izer
,
wh
ich
is
an
ar
tific
ial
ec
o
s
y
s
tem
-
b
ased
m
o
d
el
th
at
was
f
u
s
ed
with
Mo
b
ileNet
V
2
.
T
h
is
h
ig
h
lig
h
ts
th
e
r
is
in
g
tr
en
d
o
f
h
y
b
r
id
m
o
d
els
in
co
r
p
o
r
atin
g
a
n
u
m
b
er
o
f
AI
tech
n
iq
u
es.
B
esid
es,
th
er
e
is
an
ev
id
en
t
p
r
o
o
f
u
s
in
g
th
e
ad
v
a
n
ce
d
n
e
u
r
al
n
e
two
r
k
s
in
p
est
co
n
tr
o
l,
as
it
is
d
em
o
n
s
tr
ated
in
th
e
r
ef
er
e
n
ce
[
1
8
]
wh
en
th
e
f
ar
m
lan
d
f
er
tili
ty
alg
o
r
ith
m
i
s
b
u
ilt
to
au
to
m
atica
lly
d
etec
t
th
e
r
ice
p
ests
.
T
h
e
m
o
d
els
th
at
wer
e
u
s
ed
to
ex
tr
ac
t
f
ea
tu
r
es
to
p
er
f
o
r
m
th
e
class
if
icat
io
n
p
r
o
ce
s
s
wer
e
th
e
E
lm
an
r
ec
u
r
r
e
n
t
n
eu
r
al
n
et
wo
r
k
(
E
R
NN)
an
d
n
eu
r
al
ar
ch
itectu
r
e
s
ea
r
ch
n
et
wo
r
k
-
lar
g
e
(
NASNetL
ar
g
e
)
.
L
ik
ewise
,
in
r
esear
ch
er
s
[
1
9
]
,
[
2
0
]
h
as
em
p
lo
y
e
d
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
A
n
o
ve
l d
ee
p
lea
r
n
in
g
-
b
a
s
ed
p
o
llin
a
to
r
s
p
ec
ies cla
s
s
ifica
tio
n
u
s
in
g
th
e
mu
ltimo
d
a
l
…
(
P
o
o
ja
Ha
d
ima
n
e)
2689
th
e
m
eth
o
d
o
f
in
n
o
v
ati
v
e
in
s
ec
t
d
etec
tio
n
an
d
class
if
icatio
n
(
th
at
is
,
a
co
m
b
i
n
atio
n
o
f
De
n
s
eNe
t
an
d
Gan
n
et
o
p
tim
izatio
n
alg
o
r
ith
m
)
to
s
h
o
w
th
e
f
u
r
th
er
d
e
v
elo
p
m
en
t
o
f
in
s
ec
t
-
d
etec
tin
g
m
o
d
els.
AI
h
as
g
o
n
e
a
lo
n
g
way
in
its
ef
f
o
r
t
to
ass
im
i
late
in
to
th
e
p
o
llin
atio
n
o
f
p
lan
ts
.
I
t
h
as
b
ee
n
s
u
g
g
ested
th
at
a
C
NN
an
d
AI
ca
n
b
e
ap
p
lied
to
r
eliab
ly
id
en
tif
y
f
l
o
wer
s
th
at
ca
n
b
e
p
o
llin
ated
b
y
d
r
o
n
es
o
r
r
o
b
o
ts
.
Als
o
,
Hir
ag
u
r
i
et
a
l
.
[
2
1
]
en
h
an
ce
d
t
h
e
au
to
m
atio
n
o
f
p
o
llin
ato
r
id
en
tific
atio
n
b
y
d
ev
elo
p
in
g
C
NN
m
o
d
els
o
f
b
ee
s
p
ec
ies
th
at
p
o
llin
ate
b
lu
eb
er
r
ies
an
d
co
m
b
in
in
g
th
em
with
lo
g
Me
l
-
s
p
ec
tr
o
g
r
am
r
ep
r
esen
tatio
n
s
.
Fu
r
t
h
er
m
o
r
e,
Fer
r
eir
a
et
a
l.
[
2
2
]
en
h
an
ce
d
t
h
e
au
to
m
ated
i
n
s
ec
t su
r
v
eillan
ce
,
b
o
th
i
n
ter
m
s
o
f
in
s
ec
t r
ec
o
g
n
itio
n
u
s
in
g
DL
m
o
d
els an
d
in
ter
m
s
o
f
p
r
o
v
id
in
g
a
tr
ain
in
g
d
ata
s
et
th
at
is
en
ab
led
b
y
a
s
p
ec
if
ically
d
ev
elo
p
e
d
tim
e
-
lap
s
e
ca
m
er
a
s
y
s
tem
.
T
h
is
is
o
n
e
o
f
th
e
i
d
ea
l
ex
a
m
p
les
o
f
a
DL
alg
o
r
ith
m
a
n
d
r
ea
l
-
tim
e
d
ata
g
ath
e
r
in
g
te
ch
n
iq
u
es.
L
i
k
ewise,
B
jer
g
e
et
a
l.
[
2
3
]
d
ev
elo
p
e
d
a
m
u
ltis
en
s
o
r
y
im
ag
in
g
s
y
s
tem
th
at
ca
n
b
e
ap
p
lied
to
th
e
co
n
v
en
tio
n
al
in
s
ec
t
tr
ap
s
to
en
h
a
n
ce
th
e
q
u
ality
o
f
th
e
r
eq
u
ested
im
ag
es
to
b
e
ef
f
ec
tiv
e
in
th
e
c
ateg
o
r
izatio
n
o
f
th
e
s
p
ec
ies
u
n
d
er
p
r
ec
is
e
tax
o
n
o
m
ic
class
if
icati
o
n
,
an
d
B
r
an
d
t
et
a
l.
[
2
4
]
p
r
o
p
o
s
ed
a
C
NN
-
b
ased
d
etec
t
io
n
s
y
s
tem
to
th
e
r
ea
l
-
tim
e
class
if
icatio
n
o
f
th
e
liv
e
f
r
u
it
f
ly
s
p
ec
ies,
s
h
o
win
g
th
e
ap
p
licatio
n
o
f
AI
in
p
est
m
o
n
ito
r
in
g
o
f
t
h
e
ty
p
es o
f
s
p
ec
ies
,
s
u
ch
as o
liv
e
an
d
Me
d
iter
r
a
n
ea
n
f
r
u
it f
lies
,
wh
ich
ar
e
h
ig
h
ly
d
a
n
g
er
o
u
s
to
g
lo
b
al
ag
r
icu
ltu
r
e.
DL
an
d
AI
h
av
e
ac
h
iev
e
d
a
l
o
t
in
th
e
s
p
h
er
e
o
f
i
n
s
ec
t
r
ec
o
g
n
itio
n
an
d
class
if
icatio
n
,
wh
er
e
m
o
s
t
s
tu
d
ies
h
av
e
lar
g
ely
b
ee
n
f
o
c
u
s
ed
o
n
th
e
en
v
ir
o
n
m
en
tal
o
r
v
is
u
al
im
ag
e.
W
h
en
it
co
m
es
to
m
u
ltimo
d
al
d
ata
in
teg
r
atio
n
,
m
o
r
e
r
esear
ch
is
n
ee
d
ed
to
class
if
y
th
e
s
p
ec
ie
s
o
f
p
o
llin
ato
r
s
an
d
ca
lcu
late
th
e
r
is
k
o
f
p
o
llin
atio
n
.
T
h
r
o
u
g
h
th
e
i
n
teg
r
atio
n
o
f
e
n
v
ir
o
n
m
en
tal
an
d
v
is
u
al
d
ata,
t
h
is
r
esear
ch
will f
ill its
ex
is
t
in
g
g
ap
an
d
g
en
e
r
ate
a
co
m
p
r
eh
e
n
s
iv
e
an
d
ac
c
u
r
ate
a
p
p
r
o
ac
h
to
th
e
d
ef
in
itio
n
o
f
p
o
llin
ato
r
s
p
ec
ies
an
d
th
e
ass
ess
m
en
t
o
f
th
e
PVI
.
I
t
en
h
an
ce
s
b
etter
ca
teg
o
r
izatio
n
an
d
p
r
esen
ts
in
f
o
r
m
ativ
e
d
a
ta
o
n
f
ar
m
an
d
co
n
s
er
v
atio
n
ac
tiv
ities
.
Alth
o
u
g
h
AI
an
d
DL
h
av
e
m
ad
e
a
m
a
s
s
iv
e
leap
in
th
e
in
s
ec
t
id
en
t
if
icatio
n
p
r
o
ce
s
s
,
m
o
s
t
r
esear
ch
er
s
h
av
e
s
tu
d
ied
v
is
u
al
o
r
en
v
ir
o
n
m
e
n
tal
d
ata
i
n
is
o
latio
n
,
r
estra
in
in
g
t
h
e
o
v
er
all
p
r
ec
is
io
n
an
d
u
s
ab
ilit
y
o
f
b
o
th
in
ec
o
lo
g
ical
r
esear
ch
[
2
5
]
.
T
h
e
ex
is
tin
g
m
o
d
els,
th
o
u
g
h
u
s
ef
u
l
in
id
en
ti
f
y
in
g
s
p
ec
if
ic
s
p
ec
ies
o
f
p
ests
o
r
in
d
iag
n
o
s
in
g
ag
r
icu
ltu
r
al
d
is
ea
s
es,
d
o
n
o
t
r
ef
lect
th
e
v
ar
iatio
n
o
f
t
h
e
p
o
lli
n
ato
r
s
p
ec
ies
class
if
icatio
n
an
d
m
ay
n
ee
d
to
h
av
e
a
m
o
r
e
co
m
p
r
eh
en
s
iv
e
v
iew
o
f
u
n
d
e
r
s
tan
d
in
g
t
h
e
ec
o
lo
g
ical
in
ter
ac
tio
n
s
an
d
m
u
ltimo
d
al
d
ata.
Als
o
,
th
er
e
is
a
lack
o
f
liter
atu
r
e
o
n
th
e
co
m
b
in
atio
n
o
f
e
n
v
ir
o
n
m
en
tal
a
n
d
v
is
u
al
d
ata
to
d
ete
r
m
in
e
t
h
e
s
u
s
ce
p
tib
ilit
y
o
f
p
o
llin
atio
n
p
o
p
u
latio
n
s
to
clim
ate
ch
an
g
e
an
d
h
ab
itat
lo
s
s
.
T
h
e
ex
is
ten
ce
o
f
th
is
g
ap
s
h
o
ws
th
at
ad
d
itio
n
al
an
d
m
o
r
e
ex
ten
s
iv
e
m
o
d
els
ar
e
n
ec
ess
ar
y
to
in
clu
d
e
ec
o
lo
g
i
ca
l
an
d
en
v
i
r
o
n
m
e
n
tal
in
f
o
r
m
atio
n
to
class
if
y
th
e
p
o
llin
ato
r
s
an
d
th
eir
v
u
l
n
er
a
b
ilit
y
.
T
h
e
ab
s
en
ce
o
f
a
u
to
m
ated
an
d
s
ca
lab
le
s
y
s
tem
s
to
tr
ac
k
s
p
ec
ies
o
f
p
o
llin
ato
r
s
,
wh
ich
a
r
e
im
p
o
r
ta
n
t in
b
io
d
iv
er
s
ity
an
d
in
ag
r
ic
u
ltu
r
e,
is
also
a
cr
itical
g
ap
th
a
t is cr
itical.
3.
P
RO
P
O
SE
D
M
E
T
H
O
DO
L
O
G
Y
T
h
e
p
r
o
p
o
s
ed
MSI
R
f
r
am
e
wo
r
k
f
o
r
m
u
lti
-
m
o
d
al
im
ag
e
class
if
icatio
n
ex
p
lo
r
es
f
in
e
-
tu
n
in
g
th
e
m
u
ltimo
d
al
co
n
t
r
asti
v
e
lear
n
i
n
g
f
r
am
ew
o
r
k
(
MCLF)
m
o
d
e
l
f
o
r
im
ag
e
class
if
icatio
n
b
y
in
teg
r
atin
g
b
o
th
its
v
is
u
al
an
d
tex
tu
al
en
co
d
er
s
.
I
n
itially
,
it
d
ef
in
es
a
lo
s
s
f
u
n
ctio
n
th
at
m
in
im
izes
th
e
d
i
s
cr
ep
an
cy
b
etwe
en
p
r
ed
icted
p
r
o
b
ab
ilit
ies
an
d
ac
tu
al
class
lab
el
s
u
s
in
g
co
n
tr
asti
v
e
lear
n
in
g
o
b
jectiv
es.
T
wo
p
r
im
ar
y
f
r
am
ewo
r
k
s
ar
e
ex
am
in
e
d
:
th
e
im
ag
e
-
f
ea
t
u
r
e
-
b
ased
f
r
am
ewo
r
k
,
wh
ic
h
in
p
u
ts
v
is
u
al
f
ea
tu
r
es
in
to
a
tu
n
ab
le
class
if
ier
,
an
d
th
e
lan
g
u
ag
e
-
e
n
co
d
er
-
b
as
ed
f
r
am
ewo
r
k
,
wh
ich
ad
j
u
s
ts
t
h
e
tex
t
en
co
d
e
r
with
lear
n
ab
le
to
k
en
em
b
e
d
d
in
g
s
b
u
t
f
ac
es
alig
n
m
en
t
is
s
u
es
b
etwe
en
v
is
u
al
an
d
tex
tu
al
f
ea
tu
r
es.
T
o
o
v
er
c
o
m
e
th
is
,
th
e
an
aly
s
is
in
tr
o
d
u
ce
s
p
r
o
to
ty
p
e
-
d
r
i
v
en
v
is
u
al
an
d
t
ex
t
r
eg
u
lar
izatio
n
m
eth
o
d
s
.
T
h
ese
m
eth
o
d
s
g
en
er
ate
class
p
r
o
to
ty
p
es
f
r
o
m
th
e
v
is
u
al
an
d
tex
tu
al
en
co
d
e
r
s
,
r
esp
ec
tiv
ely
,
an
d
r
eg
u
lar
ize
th
e
tu
n
ab
le
weig
h
ts
b
y
m
in
im
izin
g
th
e
Ku
llb
ac
k
-
L
eib
ler
d
iv
e
r
g
en
ce
b
etwe
en
th
ese
p
r
o
to
ty
p
es
an
d
th
e
class
if
ier
we
ig
h
ts
.
T
h
e
o
v
er
all
lo
s
s
f
u
n
ctio
n
co
m
b
in
es
th
e
cr
o
s
s
-
en
tr
o
p
y
l
o
s
s
with
th
ese
r
eg
u
lar
izatio
n
ter
m
s
,
weig
h
ted
b
y
b
alan
ce
f
ac
to
r
s
.
Fig
u
r
e
2
p
r
esen
ts
th
e
p
r
o
p
o
s
ed
wo
r
k
f
lo
w.
Fig
u
r
e
2
.
Pro
p
o
s
ed
wo
r
k
f
lo
w
T
h
is
ap
p
r
o
ac
h
aim
s
to
en
h
an
c
e
m
o
d
el
p
e
r
f
o
r
m
an
ce
b
y
en
s
u
r
in
g
b
etter
alig
n
m
en
t
a
n
d
g
e
n
er
aliza
tio
n
ac
r
o
s
s
b
o
th
m
o
d
alities
.
Fig
u
r
e
3
s
h
o
ws
th
e
p
r
o
p
o
s
ed
a
r
ch
ite
ctu
r
e.
T
h
e
MSI
R
f
r
am
ewo
r
k
i
n
teg
r
ates
v
is
u
al
an
d
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
1
5
,
No
.
3
,
J
u
n
e
2
0
2
6
:
2
6
8
5
-
2
6
9
7
2690
s
p
ec
ies
-
d
etail
in
f
o
r
m
atio
n
f
o
r
r
o
b
u
s
t
p
o
llin
ato
r
class
if
icatio
n
.
Step
1
:
p
o
llin
ato
r
im
ag
es
an
d
co
r
r
esp
o
n
d
in
g
s
p
ec
ies
-
d
etail
attr
ib
u
tes
ar
e
p
r
o
v
id
ed
as
m
u
ltimo
d
al
in
p
u
ts
.
Step
2
:
im
ag
e
f
ea
tu
r
es
ar
e
ex
tr
ac
ted
u
s
in
g
th
e
v
is
u
al
en
co
d
er
o
f
th
e
MCLF,
wh
ile
s
p
ec
ies
d
etails
ar
e
en
co
d
ed
u
s
in
g
its
tex
t
en
c
o
d
er
.
Step
3
:
a
co
n
tr
asti
v
e
lear
n
in
g
o
b
jectiv
e
alig
n
s
v
is
u
al
an
d
tex
tu
al
em
b
ed
d
in
g
s
in
a
s
h
ar
ed
f
ea
tu
r
e
s
p
ac
e.
Step
4
:
class
-
lev
el
im
ag
e
an
d
s
p
ec
ies
-
d
etail
p
r
o
to
ty
p
es
ar
e
co
n
s
tr
u
cted
to
r
ep
r
esen
t
e
ac
h
class
.
Step
5
:
I
SP
R
an
d
S
DPR
alig
n
clas
s
if
ier
weig
h
ts
with
th
eir
r
esp
ec
tiv
e
p
r
o
to
ty
p
es,
r
e
d
u
cin
g
cr
o
s
s
-
m
o
d
al
f
ea
tu
r
e
m
is
alig
n
m
e
n
t.
Ste
p
6
:
th
e
n
etwo
r
k
is
o
p
tim
ized
u
s
in
g
a
co
m
p
o
s
ite
lo
s
s
co
m
b
in
in
g
class
if
icatio
n
an
d
p
r
o
to
t
y
p
e
-
b
ased
r
eg
u
lar
iz
atio
n
.
Step
7
:
th
e
tr
ain
ed
m
o
d
el
p
er
f
o
r
m
s
b
in
a
r
y
an
d
m
u
lticlas
s
p
o
llin
ato
r
class
if
icatio
n
.
Fig
u
r
e
3
.
Pro
p
o
s
ed
ar
c
h
itectu
r
e
3
.
1
.
M
ultim
o
da
l c
o
ntr
a
s
t
iv
e
lea
rning
f
ra
m
ewo
r
k
m
o
del
T
h
e
MCLF
r
ep
r
esen
ts
a
s
ig
n
if
ican
t
b
r
ea
k
th
r
o
u
g
h
in
ML
b
y
d
em
o
n
s
tr
atin
g
h
o
w
n
at
u
r
a
l
lan
g
u
ag
e
s
u
p
er
v
is
io
n
ca
n
b
e
u
s
ed
to
tr
ain
s
tate
-
of
-
th
e
-
a
r
t
v
is
u
al
m
o
d
els.
Un
lik
e
tr
ad
itio
n
al
co
m
p
u
ter
v
is
io
n
s
y
s
tem
s
,
wh
ich
r
ely
o
n
p
r
ed
ef
i
n
ed
ca
teg
o
r
ies
o
f
o
b
jects,
MCLF
lear
n
s
f
r
o
m
a
v
ast
d
ataset
o
f
im
ag
e
-
tex
t
p
air
s
,
4
0
0
m
illi
o
n
in
to
tal,
co
llected
f
r
o
m
th
e
in
ter
n
et.
T
h
e
m
o
d
el
i
s
tr
ain
ed
to
p
r
ed
ict
wh
ic
h
tex
t
is
ass
o
ciate
d
with
wh
ich
im
ag
e
u
s
in
g
a
co
n
tr
asti
v
e
lear
n
in
g
ap
p
r
o
ac
h
,
r
ath
er
th
an
r
ely
i
n
g
o
n
lab
elled
d
atasets
lik
e
I
m
ag
eNe
t.
T
h
is
allo
ws
th
e
MCLF
m
o
d
el
to
g
en
er
alize
ac
r
o
s
s
v
ar
io
u
s
t
ask
s
an
d
p
e
r
f
o
r
m
ze
r
o
-
s
h
o
t
tr
an
s
f
er
,
wh
er
e
it
ca
n
class
if
y
im
ag
es th
at
it h
as n
ev
er
s
ee
n
b
ef
o
r
e
with
o
u
t a
d
d
itio
n
al
tr
ain
in
g
.
MCLF
ac
h
iev
es
its
v
er
s
atilit
y
b
y
jo
in
tly
t
r
ain
in
g
an
im
a
g
e
en
co
d
er
a
n
d
a
tex
t
en
c
o
d
er
.
Du
r
in
g
tr
ain
in
g
,
th
e
m
o
d
el
m
ax
im
ize
s
th
e
co
s
in
e
s
im
ilar
ity
b
etwe
en
p
air
ed
im
ag
e
an
d
tex
t
r
ep
r
esen
tatio
n
s
wh
ile
m
in
im
izin
g
s
im
ilar
ity
f
o
r
m
is
m
atch
ed
p
air
s
.
T
h
is
r
esu
lts
in
a
s
h
ar
ed
em
b
e
d
d
in
g
s
p
ac
e
f
o
r
b
o
t
h
im
ag
es
an
d
tex
t,
wh
ich
ca
n
b
e
lev
er
a
g
ed
to
p
er
f
o
r
m
class
if
icatio
n
task
s
in
a
ze
r
o
-
s
h
o
t
m
an
n
e
r
.
I
n
s
tead
o
f
tr
ain
in
g
a
class
if
ier
with
lab
elled
im
ag
es,
MCLF
ca
n
u
s
e
th
e
tex
tu
al
d
escr
ip
tio
n
s
o
f
tar
g
et
class
es
to
cr
ea
te
a
class
if
ier
on
-
th
e
-
f
ly
.
T
h
is
ab
ilit
y
to
u
s
e
n
atu
r
al
lan
g
u
ag
e
as
a
s
u
p
er
v
is
o
r
y
s
ig
n
al,
MCLF,
en
ab
les
it
to
o
u
tp
er
f
o
r
m
o
r
m
atch
f
u
lly
s
u
p
er
v
is
ed
m
o
d
els
in
m
an
y
co
m
p
u
ter
s
v
is
io
n
task
s
,
in
cl
u
d
in
g
o
b
ject
r
ec
o
g
n
itio
n
,
o
p
tical
ch
ar
ac
ter
r
ec
o
g
n
itio
n
(
OC
R
)
,
an
d
e
v
en
ac
tio
n
r
ec
o
g
n
itio
n
i
n
v
id
e
o
s
.
T
h
e
MCLF
m
o
d
el’
s
s
u
cc
ess
h
ig
h
lig
h
ts
th
e
p
o
ten
tial
o
f
s
ca
lin
g
u
p
le
ar
n
in
g
f
r
o
m
web
-
s
ca
le
n
atu
r
a
l
lan
g
u
ag
e
d
ata.
B
y
tr
ain
i
n
g
o
n
m
ass
iv
e,
d
iv
er
s
e
d
atasets
,
MCLF
d
em
o
n
s
tr
ates
im
p
r
o
v
ed
r
o
b
u
s
tn
ess
an
d
tr
a
n
s
f
er
lear
n
i
n
g
ca
p
ab
ilit
ies
co
m
p
ar
ed
to
m
o
d
els
tr
ain
ed
o
n
lim
ited
,
la
b
elled
d
atasets
.
I
ts
ze
r
o
-
s
h
o
t
p
er
f
o
r
m
an
ce
is
co
m
p
etitiv
e
ac
r
o
s
s
a
wid
e
r
an
g
e
o
f
task
s
an
d
d
o
m
ain
s
,
in
clu
d
in
g
f
in
e
-
g
r
ain
ed
im
a
g
e
class
if
icatio
n
an
d
lo
ca
lizatio
n
task
s
.
T
h
i
s
o
p
en
s
u
p
n
ew
p
o
s
s
ib
ilit
ies
f
o
r
d
ev
elo
p
in
g
m
o
r
e
g
en
er
alize
d
AI
s
y
s
tem
s
th
at
ca
n
lear
n
f
r
o
m
d
iv
er
s
e
d
ata
s
o
u
r
ce
s
with
o
u
t
th
e
n
ee
d
f
o
r
ex
ten
s
iv
e
task
-
s
p
ec
if
i
c
d
atasets
.
3
.
1
.
1
.
P
re
lim
ina
ry
a
na
ly
s
is
B
y
ass
u
m
in
g
a
co
m
m
o
n
ap
p
r
o
ac
h
,
a
p
r
e
-
t
r
ain
in
g
m
ec
h
an
is
m
th
at
lear
n
s
to
ass
o
ciate
im
ag
es
with
tex
tu
al
d
escr
ip
tio
n
s
u
s
in
g
c
o
n
tr
asti
v
e
lear
n
in
g
o
b
jectiv
es,
w
h
ich
is
f
in
e
-
tu
n
ed
f
o
r
class
if
y
in
g
im
ag
es
d
en
o
ted
by
,
b
y
co
n
s
id
er
in
g
a
tr
ain
in
g
s
et
,
=
{
(
,
)
}
=
1
,
h
er
e
is
an
im
ag
e
an
d
i
s
th
e
one
-
h
o
t
en
co
d
in
g
,
a
tu
n
ab
le
p
ar
am
eter
d
en
o
te
d
b
y
alo
n
g
th
e
MCLF
m
o
d
el,
=
{
,
}
wh
er
ein
an
d
d
ep
ict
th
e
v
is
u
al
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
A
n
o
ve
l d
ee
p
lea
r
n
in
g
-
b
a
s
ed
p
o
llin
a
to
r
s
p
ec
ies cla
s
s
ifica
tio
n
u
s
in
g
th
e
mu
ltimo
d
a
l
…
(
P
o
o
ja
Ha
d
ima
n
e)
2691
en
co
d
er
,
th
e
lear
n
in
g
o
b
jectiv
e
lead
s
to
m
in
im
izin
g
a
lo
s
s
d
en
o
ted
.
Ho
wev
er
,
(
,
,
)
d
ep
icts
th
e
p
r
o
b
a
b
ilit
y
f
o
r
th
e
s
o
f
tm
ax
cl
ass
,
d
en
o
ted
b
y
th
e
c
o
n
f
id
e
n
c
e
v
ec
to
r
,
t
h
e
f
in
e
-
t
u
n
in
g
v
ec
to
r
s
ap
p
lied
o
n
th
e
v
ar
io
u
s
o
p
tim
izatio
n
s
tr
ateg
ies f
o
r
alo
n
g
with
t
h
e
v
ar
io
u
s
ca
lcu
latio
n
tech
n
iq
u
es f
o
r
r
esp
ec
tiv
ely
in
(
1
)
.
(
,
,
)
=
−
1
∑
(
,
,
)
=
1
(
1
)
3
.
1
.
2
.
I
m
a
g
e
-
s
pecies c
la
s
s
if
ic
a
t
io
n f
ra
m
ewo
r
k
L
in
ea
r
p
r
o
b
e
n
etwo
r
k
d
en
o
ted
b
y
i
s
th
e
b
asic
m
eth
o
d
u
s
ed
f
o
r
th
e
im
ag
e
-
f
ea
tu
r
e
-
b
ased
f
i
n
e
-
tu
n
in
g
f
r
am
ewo
r
k
,
wh
ich
in
itially
e
x
t
r
ac
ts
th
e
v
is
u
al
f
ea
tu
r
es?
(
)
f
o
r
t
h
e
im
ag
e
alo
n
g
with
th
e
im
a
g
e
f
ea
tu
r
e
in
(
2
)
.
T
h
en
th
e
v
is
u
al
f
ea
tu
r
e
s
ar
e
f
ed
i
n
to
th
e
tu
n
ab
le
weig
h
ts
d
en
o
ted
b
y
,
co
n
s
id
er
ed
a
u
s
u
al
class
if
ier
to
ev
alu
ate
th
e
p
r
o
b
a
b
ilit
y
.
T
h
e
lo
s
s
f
u
n
ctio
n
is
d
en
o
ted
.
Her
e
(
.
(
)
)
d
ep
icts
th
e
ex
p
o
n
en
tial
lo
g
ic
f
o
r
th
e
−
ℎ
class
an
d
r
ep
r
esen
ts
th
e
d
o
t p
r
o
d
u
ct
i
n
(
3
)
.
−
(
,
,
)
=
−
1
∑
(
,
,
)
=
1
(
2
)
(
,
,
)
=
(
.
(
)
)
∑
(
.
(
)
)
(
3
)
3
.
1
.
3
.
Sp
ec
ies
a
t
t
ribute
enco
din
g
f
ra
m
ewo
r
k
T
h
e
p
r
o
m
p
t
lear
n
in
g
f
r
am
ew
o
r
k
ac
ts
as
th
e
b
aselin
e
m
eth
o
d
f
o
r
th
e
s
p
ec
ies
attr
ib
u
te
en
co
d
in
g
f
r
am
ewo
r
k
(
SAEF)
,
an
d
t
h
e
tu
n
ab
le
weig
h
t.
co
n
s
id
er
ed
as
le
ar
n
ab
le
to
k
e
n
em
b
e
d
d
in
g
s
f
o
r
ea
ch
class
o
f
th
e
tex
t
en
co
d
er
.
T
h
is
f
u
r
th
er
c
o
n
ca
ten
ates
with
th
e
em
b
ed
d
in
g
s
o
f
ea
ch
class
tex
t
,
th
e
tex
t
is
f
ed
in
to
th
e
tex
t
en
co
d
er
to
d
er
iv
e
th
e
te
x
t
f
ea
tu
r
es
r
ep
r
esen
ted
as
(
,
)
.
T
h
is
r
ed
u
ce
s
th
e
s
im
ilar
ity
b
et
wee
n
th
e
v
is
u
al
f
ea
tu
r
es
(
)
an
d
t
h
e
s
p
e
cies
f
ea
tu
r
es
f
o
r
g
r
o
u
n
d
tr
u
t
h
class
tex
t
in
(
4
)
.
T
h
e
lo
s
s
f
u
n
ctio
n
is
d
eter
m
in
ed
h
er
e
in
(
5
)
.
−
(
,
,
)
=
−
1
∑
(
,
,
,
,
)
=
1
(
4
)
(
,
,
,
,
)
=
(
(
(
,
)
,
(
)
)
)
∑
(
(
(
,
)
,
(
)
)
)
(
5
)
W
h
er
ein
(
.
,
.
)
r
ep
r
esen
ts
th
e
co
s
i
n
e
s
im
ilar
ity
ca
lcu
latio
n
as
∑
(
c
os
(
(
,
)
,
(
)
)
)
d
en
o
tes
th
e
ex
p
o
n
e
n
tial
p
r
ed
i
ctio
n
f
o
r
th
e
−
ℎ
class
.
T
h
e
lan
g
u
ag
e
-
en
c
o
d
er
-
b
ased
f
r
a
m
ewo
r
k
f
in
e
-
tu
n
es
th
e
tex
t
in
p
u
t
a
n
d
th
en
in
h
er
its
th
e
m
ec
h
an
is
m
.
T
h
e
p
e
r
f
o
r
m
an
ce
is
p
o
o
r
b
ec
au
s
e
o
f
th
e
m
is
alig
n
m
en
t
b
etwe
en
th
e
ch
ar
ac
te
r
is
tics
th
at
th
e
tex
t a
n
d
v
is
u
al
en
co
d
er
s
ex
tr
ac
t.
3
.
1
.
4
.
Neg
a
t
i
v
e
lo
g
-
li
k
eliho
o
d lo
s
s
Up
o
n
f
ix
i
n
g
th
e
v
is
u
al
r
ep
r
e
s
en
tatio
n
lay
er
with
th
e
tex
t
en
co
d
er
alo
n
g
with
th
e
c
o
s
in
e
class
if
ier
with
th
e
ad
ju
s
tab
le
weig
h
t
u
p
o
n
t
h
e
v
is
u
al
en
c
o
d
er
.
B
ased
o
n
th
e
v
is
u
al
e
n
co
d
er
.
T
h
e
cr
o
s
s
-
en
tr
o
p
y
lo
s
s
is
g
iv
e
n
.
T
h
e
k
ey
d
if
f
er
en
ce
b
etwe
e
n
th
e
p
r
o
p
o
s
ed
an
d
th
e
v
is
u
al
en
co
d
er
-
b
ased
f
r
am
ewo
r
k
is
th
at
th
e
b
ias
t
er
m
ass
o
ciate
d
with
it
is
ess
en
tial
f
o
r
th
e
r
e
g
u
lar
izatio
n
p
r
o
ce
s
s
,
p
r
o
p
o
s
e
p
r
o
to
ty
p
e
-
d
r
i
v
en
v
is
u
al
r
e
g
u
l
ar
izatio
n
an
d
p
r
o
to
t
y
p
e
-
d
r
iv
e
n
tex
t
r
eg
u
lar
izatio
n
t
o
u
tili
ze
th
e
v
is
u
al
an
d
tex
t
en
co
d
er
o
f
th
e
MCLF m
o
d
el.
3
.
1
.
5
.
I
m
a
g
e
-
s
pecies pro
t
o
t
y
pe
re
g
ula
riza
t
io
n
B
ased
o
n
th
e
v
is
u
al
en
co
d
er
,
th
e
im
ag
e
-
b
ased
p
r
o
to
t
y
p
e
m
o
d
el
u
s
ed
f
o
r
th
e
th
e
class
as
th
e
v
is
u
al
f
ea
tu
r
es
f
o
r
all
th
e
s
am
p
les
with
in
th
e
class
.
Ho
wev
er
,
d
en
o
tes
th
e
n
u
m
b
er
o
f
s
am
p
les
in
class
with
,
is
th
e
−
ℎ
r
o
w
o
f
.
T
h
en
,
r
e
g
u
lar
ize
th
e
tu
n
ab
le
weig
h
t.
with
th
e
v
is
u
al
p
r
o
t
o
ty
p
e.
Her
e
d
ep
icts
th
e
to
tal
n
u
m
b
er
o
f
class
es,
is
th
e
−
ℎ
r
o
w
o
f
an
d
d
en
o
tes
th
e
Ku
llb
ac
k
-
L
eib
ler
d
iv
er
g
en
ce
b
etwe
en
two
v
ec
to
r
s
b
y
(
6
)
an
d
(
7
)
.
,
=
1
∑
(
(
)
=
1
)
(
6
)
−
=
1
∑
(
,
|
|
)
=
1
(
7
)
3
.
1
.
6
.
Sp
ec
ies
-
det
a
il pro
t
o
t
y
pe
re
g
ula
riza
t
io
n
B
ased
o
n
th
e
te
x
t
e
n
co
d
e
r
,
t
h
e
t
e
x
t
-
b
ase
d
p
r
o
t
o
t
y
p
e
f
o
r
t
h
e
−
ℎ
cl
ass
as
th
e
te
x
t
f
ea
tu
r
e
o
f
th
e
cl
ass
t
ex
t
,
th
at
is
e
x
p
r
ess
e
d
i
n
(
8
)
.
H
er
e
,
is
c
o
n
s
i
d
e
r
e
d
t
h
e
te
x
t
f
o
r
th
e
−
ℎ
class
,
b
y
u
s
in
g
t
h
e
te
x
t
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
1
5
,
No
.
3
,
J
u
n
e
2
0
2
6
:
2
6
8
5
-
2
6
9
7
2692
en
co
d
er
,
a
p
r
o
m
p
t
is
d
esi
g
n
e
d
to
p
r
e
f
i
x
th
e
.
Si
m
il
ar
ly
,
t
h
e
t
e
x
t
-
b
ase
d
p
r
o
t
o
t
y
p
e
r
eg
u
l
a
r
iz
ati
o
n
is
g
i
v
en
in
(
9
)
.
W
h
e
r
e
in
is
t
h
e
cl
ass
s
ize
,
is
th
e
−
ℎ
r
o
w
o
f
a
n
d
r
e
p
r
es
en
ts
t
h
e
K
u
ll
b
ac
k
-
L
e
i
b
le
r
d
i
v
e
r
g
e
n
c
e
b
etw
ee
n
tw
o
v
ec
to
r
s
.
De
p
e
n
d
i
n
g
o
n
th
e
c
r
o
s
s
-
e
n
t
r
o
p
y
l
o
s
s
,
t
h
e
v
is
u
al
an
d
t
e
x
t
-
b
ase
d
r
e
g
u
la
r
iz
ati
o
n
,
th
e
t
o
t
al
l
o
s
s
d
e
n
o
te
d
is
g
i
v
e
n
in
(
1
0
)
.
H
o
we
v
e
r
,
h
e
r
e
1
,
2
,
a
n
d
3
d
e
p
i
ct
th
e
b
ala
n
ce
f
ac
t
o
r
s
.
,
=
(
)
(
8
)
−
=
1
∑
(
,
|
|
)
=
1
(
9
)
=
1
+
2
−
+
3
−
(
1
0
)
4.
RE
SU
L
T
S
AND
D
I
SCU
SS
I
O
N
T
h
e
d
ataset
u
s
ed
f
o
r
p
o
llin
ato
r
s
p
ec
ies
class
if
icatio
n
co
n
s
is
t
s
o
f
d
etailed
f
ield
o
b
s
er
v
atio
n
s
o
f
in
s
ec
t
v
is
its
to
f
lo
wer
in
g
p
lan
ts
,
alo
n
g
with
ass
o
ciate
d
e
n
v
ir
o
n
m
e
n
tal
an
d
la
n
d
-
u
s
e
in
f
o
r
m
atio
n
.
E
ac
h
d
ata
r
ec
o
r
d
in
clu
d
es
co
n
tex
tu
al
v
ar
iab
les
s
u
ch
as
th
e
d
ate,
latitu
d
e,
lo
n
g
itu
d
e
,
s
ite
d
escr
ip
tio
n
,
an
d
wea
th
er
,
wh
ich
p
r
o
v
id
e
a
r
ich
s
et
o
f
f
ea
tu
r
es
f
o
r
m
o
d
el
tr
ain
in
g
.
T
h
e
ac
cu
r
ac
y
,
p
r
ec
is
io
n
,
r
ec
all,
an
d
F
1
-
s
co
r
e
ar
e
th
e
k
ey
p
er
f
o
r
m
an
ce
m
etr
ics
u
s
ed
t
o
ev
al
u
ate
th
e
m
o
d
els.
Acc
u
r
ac
y
m
ea
s
u
r
es
th
e
o
v
er
all
co
r
r
ec
tn
ess
o
f
th
e
class
if
icatio
n
s
,
wh
ile
p
r
ec
is
io
n
f
o
cu
s
es
o
n
th
e
p
r
o
p
o
r
tio
n
o
f
tr
u
e
p
o
s
itiv
e
class
if
icatio
n
s
am
o
n
g
all
p
o
s
itiv
e
class
if
icatio
n
s
.
R
ec
all
ev
alu
ates
th
e
m
o
d
el'
s
ab
ili
ty
to
id
en
tify
all
r
elev
an
t
in
s
tan
ce
s
,
an
d
th
e
F1
-
s
co
r
e
p
r
o
v
id
es
a
b
alan
ce
b
etwe
en
p
r
ec
is
io
n
an
d
r
ec
all.
T
h
ese
m
e
tr
ics
co
llectiv
ely
h
elp
ass
ess
th
e
ef
f
ec
tiv
e
n
ess
o
f
th
e
m
o
d
els in
ac
cu
r
ately
class
if
y
in
g
p
o
llin
ato
r
s
p
ec
ies.
4
.
1
.
Da
t
a
s
et
det
a
ils
T
h
e
d
ata
s
et
s
h
all
co
n
tain
5
,
6
8
7
r
ec
o
r
d
s
o
f
i
n
s
ec
t
s
p
ec
im
en
s
,
th
e
s
u
b
ject
m
atter
o
f
wh
ich
will
b
e
th
e
p
o
llin
ato
r
s
p
ec
ies.
A
n
in
-
d
ep
t
h
r
ev
iew
o
f
th
e
d
ataset
in
d
icat
es
th
at
p
o
llin
atio
n
s
p
ec
ies
o
f
k
n
o
wn
s
p
ec
ies
co
u
n
t
5
,
5
7
3
,
an
d
n
o
n
-
p
o
llin
ato
r
s
co
u
n
t
1
1
4
(
wh
en
co
n
s
id
er
i
n
g
th
e
g
en
u
s
B
o
ta
n
o
p
h
ila
th
at
co
n
s
is
ts
o
f
m
o
s
tly
s
ee
d
-
f
ee
d
in
g
f
lies
)
.
T
h
e
d
at
a
co
n
tain
i
n
s
ec
ts
b
elo
n
g
in
g
to
th
r
ee
m
ajo
r
o
r
d
er
s
,
n
a
m
ely
,
Hy
m
e
n
o
p
ter
a
(
5
,
3
4
8
r
ec
o
r
d
s
)
,
Dip
ter
a
(
3
3
2
r
ec
o
r
d
s
)
,
an
d
C
o
leo
p
ter
a
(
7
r
ec
o
r
d
s
)
.
Hy
m
e
n
o
p
ter
a,
p
r
ed
o
m
in
an
tly
s
p
ec
ies
o
f
th
e
f
am
ily
Ap
id
ae
is
h
o
m
e
t
o
k
ey
p
o
llin
at
o
r
s
,
h
o
n
e
y
b
ee
s
(
A
p
is
g
en
u
s
)
an
d
b
u
m
b
le
b
e
es
(
B
o
mb
u
s
g
en
u
s
)
,
wh
ich
m
ak
e
u
p
a
lar
g
e
p
a
r
t
o
f
th
e
d
ataset.
T
h
e
o
th
er
g
en
u
s
to
n
o
te
is
Melis
s
o
d
es,
La
s
io
g
lo
s
s
u
m
,
an
d
Ha
lictu
s
,
wh
ich
ar
e
all
u
s
ef
u
l
in
t
h
e
p
o
l
lin
atio
n
o
f
cr
o
p
s
an
d
wild
f
lo
wer
s
.
T
h
e
an
aly
s
is
co
n
d
u
cte
d
at
th
e
s
p
ec
ies
lev
el
s
h
o
ws
th
at
A
p
is
mellifera
(
th
e
E
u
r
o
p
ea
n
h
o
n
ey
b
ee
)
h
as
th
e
lar
g
est
p
er
ce
n
tag
e
with
4
,
5
4
8
en
tr
ies,
in
d
icatin
g
th
at
it
is
im
p
o
r
tan
t
in
ag
r
icu
ltu
r
al
p
o
llin
atio
n
.
Als
o
,
s
p
ec
ies
s
u
ch
as
Melis
s
o
d
es
tr
in
o
d
is
,
B
o
mb
u
s
g
r
is
eo
co
llis
,
o
r
B
o
mb
u
s
tern
a
r
iu
s
ar
e
th
e
m
o
s
t
s
ig
n
if
ican
t
p
o
llin
ato
r
s
o
f
n
ativ
e
p
lan
ts
an
d
cr
o
p
s
.
T
h
e
B
o
ta
n
o
p
h
ila
g
en
u
s
,
1
1
4
r
ec
o
r
d
s
,
in
co
n
tr
ast,
co
n
s
i
s
ts
alm
o
s
t
s
o
lely
o
f
n
o
n
-
p
o
lli
n
atin
g
s
p
ec
ies
th
at
ar
e
m
o
r
e
ac
cu
r
ately
d
escr
ib
ed
as seed
f
ee
d
er
s
an
d
wh
ic
h
co
n
tr
ib
u
te
litt
le
to
p
o
llin
atio
n
.
T
h
e
im
ag
e
d
ata
o
f
o
u
r
d
ataset
is
s
h
o
wn
in
Fig
u
r
e
4
.
Fig
u
r
e
4
.
I
m
ag
e
d
ataset
4
.
2
.
Resul
t
s
4
.
2
.
1
.
B
ina
ry
cla
s
s
if
ica
t
io
n
A
b
in
ar
y
class
if
icatio
n
wo
u
ld
b
e
s
ee
n
as
a
d
iv
is
io
n
o
f
th
e
d
ata
s
et
in
to
two
d
if
f
er
e
n
t
c
ateg
o
r
ies,
in
clu
d
in
g
p
o
llin
ato
r
v
s
.
n
o
n
-
p
o
llin
ato
r
.
I
t
m
ak
es
t
h
is
p
r
o
b
le
m
ea
s
ier
to
class
if
y
an
d
is
aim
ed
at
d
eter
m
in
in
g
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
A
n
o
ve
l d
ee
p
lea
r
n
in
g
-
b
a
s
ed
p
o
llin
a
to
r
s
p
ec
ies cla
s
s
ifica
tio
n
u
s
in
g
th
e
mu
ltimo
d
a
l
…
(
P
o
o
ja
Ha
d
ima
n
e)
2693
wh
eth
er
a
n
in
s
ec
t
is
c
o
n
s
id
er
ed
to
p
lay
a
r
o
le
i
n
p
o
llin
atio
n
o
r
n
o
t.
As
a
n
illu
s
tr
atio
n
,
g
en
er
a
s
u
ch
as
A
p
is
,
B
o
mb
u
s
,
an
d
La
s
io
g
lo
s
s
u
m
,
all
co
m
m
o
n
p
o
llin
ato
r
s
,
ca
n
b
e
ter
m
ed
as
p
o
llin
ato
r
s
.
C
o
n
v
er
s
ely
,
an
y
g
en
u
s
n
o
t
ty
p
ical
o
f
p
o
llin
atio
n
(
h
ad
th
e
p
o
llin
atin
g
d
ata
av
ailab
le
in
th
e
d
ataset
wo
u
ld
b
e
ter
m
ed
a
n
o
n
-
p
o
llin
ato
r
.
T
h
is
b
in
ar
y
class
if
icatio
n
ex
er
cise
aim
s
at
esti
m
atin
g
wh
eth
er
a
p
ar
ticu
lar
in
s
ec
t,
ju
d
g
ed
b
y
its
ch
ar
ac
ter
is
tic
s
(
s
u
ch
as
in
s
ec
t
f
am
ily
,
tr
ib
e
,
an
d
s
p
ec
ies),
will
b
e
d
esc
r
ib
ed
eith
er
as
p
ar
t
o
f
th
e
p
o
llin
ato
r
o
r
as
a
non
-
p
o
llin
ato
r
.
T
h
e
m
o
d
el
is
tr
ain
ed
to
d
is
cr
im
in
ate
b
etwe
e
n
th
ese
two
class
es
o
f
f
ea
tu
r
es
g
iv
en
,
an
d
it
is
a
s
in
g
le
b
in
ar
y
la
b
el,
wh
ich
is
0
(
n
o
n
-
p
o
llin
ato
r
)
o
r
1
(
p
o
llin
at
o
r
)
.
T
h
e
ch
ar
t
is
a
co
m
p
ar
is
o
n
b
etwe
en
th
e
ac
cu
r
ac
y
o
f
two
m
o
d
els,
th
e
RF
an
d
o
u
r
p
r
o
p
o
s
ed
DL
n
etwo
r
k
,
with
p
o
llin
ato
r
s
p
e
cies
clas
s
if
icatio
n
.
DL
m
o
d
el
is
h
ig
h
ly
m
o
r
e
ef
f
ec
tiv
e
th
a
n
th
e
RF
with
an
ap
p
r
o
x
im
ate
o
f
9
8
.
3
4
%
ac
cu
r
ac
y
wh
en
co
m
p
ar
ed
to
9
2
.
T
h
e
f
in
d
in
g
s
h
o
ws
th
e
d
ee
p
er
ap
p
licab
ilit
y
o
f
th
e
DL
m
eth
o
d
t
o
th
e
co
r
r
ec
t
class
if
icatio
n
o
f
th
e
p
o
llin
ato
r
s
p
e
cies
u
n
d
er
o
u
r
m
u
ltimo
d
al
a
p
p
r
o
ac
h
.
Fig
u
r
e
5
r
ep
r
esen
ts
th
e
an
aly
s
is
o
f
ac
cu
r
ac
y
.
T
h
e
ch
a
r
t
c
o
m
p
ar
es
t
h
e
p
r
ec
i
s
io
n
o
f
th
e
RF
m
o
d
el
an
d
o
u
r
p
r
o
p
o
s
ed
DL
n
etwo
r
k
f
o
r
p
o
llin
ato
r
s
p
ec
ies
class
if
icatio
n
.
T
h
e
DL
m
o
d
el
d
em
o
n
s
tr
ates
s
ig
n
if
ica
n
tly
h
ig
h
er
p
r
ec
is
io
n
,
ar
o
u
n
d
9
5
%,
co
m
p
ar
ed
to
th
e
8
5
% a
ch
iev
e
d
b
y
RF
.
T
h
is
in
d
icate
s
th
at
th
e
DL
m
o
d
el
is
m
o
r
e
ef
f
ec
tiv
e
in
ac
cu
r
ately
i
d
en
tify
in
g
r
elev
an
t
p
o
llin
ato
r
s
p
ec
ies wh
ile
m
in
i
m
izin
g
f
alse p
o
s
itiv
es.
Fig
u
r
e
6
s
h
o
ws th
e
p
r
ec
is
io
n
c
o
m
p
ar
i
s
o
n
.
T
h
e
c
h
ar
t
c
o
m
p
ar
es
t
h
e
r
ec
a
ll
o
f
t
h
e
RF
m
o
d
el
a
n
d
o
u
r
p
r
o
p
o
s
e
d
D
L
n
etw
o
r
k
f
o
r
p
o
lli
n
a
to
r
s
p
ec
ies
class
i
f
i
ca
t
io
n
.
T
h
e
p
r
o
p
o
s
ed
D
L
n
e
tw
o
r
k
e
x
h
ib
its
a
h
ig
h
e
r
r
e
ca
ll
,
a
r
o
u
n
d
9
7
%,
c
o
m
p
ar
e
d
t
o
t
h
e
9
2
%
a
ch
ie
v
e
d
b
y
RF
.
T
h
is
i
n
d
i
ca
t
es t
h
a
t t
h
e
DL
m
o
d
el
is
m
o
r
e
ef
f
e
cti
v
e
at
id
e
n
t
if
y
i
n
g
m
o
s
t
o
f
t
h
e
r
e
le
v
a
n
t p
o
l
li
n
at
o
r
s
p
ec
ies
,
m
i
n
im
izi
n
g
f
alse
n
e
g
at
iv
es
in
t
h
e
p
r
o
c
ess
.
Fi
g
u
r
e
7
s
h
o
ws
t
h
e
r
ec
all
c
o
m
p
ar
is
o
n
.
T
h
e
ch
ar
t
co
m
p
ar
es
th
e
F1
-
s
co
r
e
o
f
th
e
RF
m
o
d
el
an
d
o
u
r
p
r
o
p
o
s
ed
DL
n
etwo
r
k
f
o
r
p
o
llin
ato
r
s
p
ec
ies
clas
s
if
icatio
n
.
T
h
e
p
r
o
p
o
s
ed
DL
n
etwo
r
k
ac
h
ie
v
e
s
a
s
ig
n
if
ican
tly
h
ig
h
er
F1
-
s
co
r
e
,
ar
o
u
n
d
9
5
%,
co
m
p
ar
ed
to
t
h
e
8
7
%
ac
h
iev
e
d
b
y
th
e
R
F
.
T
h
is
in
d
icate
s
th
at
th
e
DL
m
o
d
el
s
tr
ik
es
a
b
etter
b
alan
ce
b
etwe
en
p
r
ec
is
io
n
an
d
r
ec
all,
m
ak
i
n
g
it
m
o
r
e
e
f
f
ec
tiv
e
o
v
er
all
in
a
cc
u
r
ately
class
if
y
in
g
p
o
llin
at
o
r
s
p
ec
ies.
Fig
u
r
e
8
s
h
o
ws th
e
F1
-
s
co
r
e
co
m
p
ar
is
o
n
.
Fig
u
r
e
5
.
Acc
u
r
ac
y
c
o
m
p
ar
is
o
n
Fig
u
r
e
6
.
Pre
cisi
o
n
c
o
m
p
ar
is
o
n
Fig
u
r
e
7
.
R
ec
all
co
m
p
a
r
is
o
n
Fig
u
r
e
8
.
F1
-
s
co
r
e
co
m
p
ar
is
o
n
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
1
5
,
No
.
3
,
J
u
n
e
2
0
2
6
:
2
6
8
5
-
2
6
9
7
2694
4
.
2
.
2
.
M
ulticla
s
s
if
ica
t
io
n
I
n
a
m
u
lticlas
s
class
if
icatio
n
task
,
th
e
g
o
al
is
to
class
if
y
ea
ch
b
ee
in
t
o
o
n
e
o
f
s
ev
er
al
p
o
s
s
ib
le
ca
teg
o
r
ies,
wh
er
e
ea
ch
ca
te
g
o
r
y
r
ep
r
esen
ts
a
s
p
ec
if
ic
b
ee
g
en
u
s
.
I
n
th
is
s
ce
n
ar
io
,
th
e
m
o
d
el
ass
ig
n
s
ea
ch
in
s
ec
t to
a
p
ar
ticu
lar
g
e
n
u
s
,
s
u
ch
as
A
p
is
,
B
o
mb
u
s
,
o
r
La
s
io
g
lo
s
s
u
m
.
I
n
s
tead
o
f
g
r
o
u
p
in
g
th
e
d
ata
in
to
ju
s
t two
ca
teg
o
r
ies,
lik
e
in
b
in
ar
y
class
if
icatio
n
,
ea
ch
b
ee
g
e
n
u
s
is
tr
e
ated
as
its
o
wn
class
.
Fo
r
in
s
t
an
ce
,
class
0
co
u
ld
r
ep
r
esen
t
A
p
is
(
th
e
h
o
n
ey
b
ee
g
en
u
s
)
,
class
1
co
u
ld
r
ep
r
ese
n
t
B
o
mb
u
s
(
th
e
b
u
m
b
leb
ee
g
e
n
u
s
)
,
class
2
co
u
ld
r
ep
r
esen
t
La
s
io
g
lo
s
s
u
m
(
t
h
e
s
wea
t
b
ee
g
e
n
u
s
)
,
an
d
s
o
o
n
f
o
r
th
e
o
th
e
r
g
e
n
er
a.
T
h
e
m
o
d
el's
o
b
jectiv
e
is
to
p
r
ed
ict
th
e
g
en
u
s
o
f
ea
ch
b
ee
b
ased
o
n
its
f
ea
tu
r
es,
s
u
ch
as
f
am
ily
,
s
p
ec
ies,
o
r
tr
ib
e.
T
h
e
m
o
d
el
will
o
u
tp
u
t
o
n
e
class
,
id
en
tify
i
n
g
th
e
g
en
u
s
to
wh
ich
th
e
b
ee
b
elo
n
g
s
.
T
o
ev
alu
ate
th
e
m
o
d
el’
s
p
e
r
f
o
r
m
an
ce
,
m
etr
ics
lik
e
p
r
ec
is
io
n
,
r
ec
all,
an
d
F1
-
s
co
r
e
ar
e
ca
lcu
lated
f
o
r
ea
ch
clas
s
,
s
h
o
win
g
h
o
w
well
th
e
m
o
d
el
co
r
r
ec
tly
p
r
ed
icts
ea
ch
b
ee
g
e
n
u
s
.
Fig
u
r
e
9
s
h
o
w
s
th
e
m
u
lticlas
s
clas
s
if
icatio
n
.
T
h
e
ch
a
r
t
s
h
o
ws
th
e
ac
cu
r
ac
y
,
r
ec
all
an
d
F1
-
s
co
r
es
o
f
ea
ch
o
f
th
e
b
ee
g
e
n
er
a
i
n
th
e
f
r
a
m
e
wo
r
k
o
f
a
m
u
lticlas
s
clas
s
if
icatio
n
p
r
o
b
lem
,
th
e
o
b
jectiv
e
o
f
wh
ic
h
is
to
class
if
y
b
ee
s
p
ec
ies
co
r
r
ec
tly
i
n
to
th
eir
co
r
r
esp
o
n
d
in
g
g
en
e
r
a.
T
h
e
g
e
n
u
s
th
at
ar
e
r
ep
r
esen
ted
ar
e
A
g
a
p
o
s
temo
n
,
A
n
d
r
en
a
,
A
p
is
,
La
s
io
g
lo
s
s
u
m
,
an
d
s
o
o
n
,
with
s
co
r
es
0
to
1
,
s
h
o
win
g
th
e
a
b
ilit
y
o
f
th
e
class
if
icatio
n
m
o
d
el.
I
n
th
e
v
ast
m
ajo
r
ity
o
f
g
e
n
er
a,
in
clu
d
in
g
A
p
is
,
A
g
a
p
o
s
temo
n
,
C
era
tin
a
,
Du
fo
u
r
ea
,
Ha
lictu
s
,
Meg
a
ch
ile
,
an
d
Melis
s
o
d
es
,
th
e
m
o
d
el
h
as
v
er
y
g
o
o
d
p
er
f
o
r
m
a
n
ce
,
with
alm
o
s
t
p
er
f
ec
t
p
r
ec
is
io
n
,
r
ec
all,
an
d
F1
-
s
co
r
es.
T
h
is
im
p
lies
th
at
th
e
m
o
d
el
is
v
er
y
u
s
ef
u
l
in
th
e
d
etec
tio
n
o
f
th
ese
g
en
e
r
a,
with
f
ew
cl
ass
if
icatio
n
er
r
o
r
s
an
d
m
o
s
tly
p
r
ed
icts
m
o
s
t
o
f
th
ese
s
am
p
les.
No
n
eth
eless
,
th
e
p
er
f
o
r
m
a
n
c
e
o
f
th
e
m
o
d
el
with
r
eg
ar
d
to
a
f
ew
g
en
er
a,
s
u
ch
as
A
n
d
r
en
a
an
d
Hyla
eu
s
,
is
s
lig
h
tly
d
is
p
r
o
p
o
r
t
io
n
ate.
I
n
p
ar
ticu
lar
,
A
n
d
r
en
a
h
as lo
w
r
ec
all
an
d
F1
-
s
co
r
e
wh
en
co
m
p
ar
ed
t
o
its
p
r
ec
is
io
n
.
I
t
m
ea
n
s
th
at
alth
o
u
g
h
th
e
m
o
d
el
is
ef
f
ec
tiv
e
i
n
r
e
co
g
n
izin
g
A
n
d
r
en
a
wh
en
it
m
ak
es
a
p
r
ed
ictio
n
,
it
ca
n
n
o
t
r
ec
o
g
n
ize
all
r
ea
l
ex
a
m
p
les
o
f
th
is
g
en
u
s
,
wh
ich
i
s
wh
y
th
e
r
ec
all
d
ec
r
ea
s
es.
T
h
e
s
am
e
tr
en
d
is
o
b
s
er
v
ed
with
Hyla
eu
s
;
in
th
at
ca
s
e,
r
ec
all
is
s
o
m
ewh
at
l
o
wer
,
in
d
icatin
g
th
e
m
o
d
el
is
m
is
s
in
g
s
o
m
e
r
ea
l
Hyla
eu
s
b
ee
s
in
its
m
o
d
el.
T
h
e
co
n
f
u
s
io
n
m
atr
ix
is
d
is
p
lay
e
d
in
Fig
u
r
e
1
0
.
Fig
u
r
e
9
.
Mu
lticlas
s
class
if
icat
io
n
T
h
e
co
n
f
u
s
io
n
m
atr
i
x
in
d
icate
s
th
at
th
e
m
o
d
el
is
d
o
in
g
r
ea
ll
y
well
at
class
if
y
in
g
th
e
g
e
n
u
s
A
p
is
,
with
9
1
9
co
r
r
ec
t
an
d
n
o
m
is
class
if
icatio
n
s
.
T
h
is
m
ea
n
s
th
at
th
e
m
o
d
el
is
ac
cu
r
ate
an
d
h
as
g
o
o
d
r
ec
all.
T
h
er
e
a
r
e
also
o
th
er
g
en
e
r
a,
s
u
ch
as
Melis
s
o
d
es
an
d
La
s
io
g
lo
s
s
u
m
,
i
n
wh
ich
th
e
p
er
f
o
r
m
a
n
ce
is
h
ig
h
(
2
7
a
n
d
1
8
co
r
r
ec
t
class
if
icatio
n
s
,
r
esp
ec
tiv
ely
)
.
T
h
er
e
ar
e
m
in
o
r
m
is
class
i
f
icatio
n
s
,
e.
g
.
,
La
s
io
g
lo
s
s
u
m
is
co
n
f
u
s
ed
with
A
n
d
r
en
a
.
T
h
er
e
is
s
o
m
e
m
i
s
class
if
icatio
n
in
th
e
g
e
n
u
s
A
n
d
r
en
a
,
wh
er
e
o
n
l
y
4
/6
ca
s
es
wer
e
co
r
r
ec
tly
d
eter
m
in
ed
,
an
d
1
/6
o
f
ea
ch
c
ase
was
m
is
p
lace
d
in
th
e
g
en
u
s
La
s
io
g
lo
s
s
u
m
an
d
th
e
g
en
u
s
C
era
tin
a
.
Sm
aller
g
en
er
a
s
u
ch
as
A
g
a
p
o
s
temo
n
,
A
u
g
o
ch
l
o
r
ella
,
an
d
Du
fo
u
r
ea
h
av
e
b
ee
n
p
lace
d
in
to
p
la
ce
,
u
s
u
ally
d
esp
ite
h
av
in
g
f
ewe
r
s
am
p
les an
d
i
n
d
i
ca
te
th
e
o
v
er
all
g
o
o
d
m
u
lticlas
s
if
icatio
n
ca
p
ab
ilit
y
o
f
t
h
e
m
o
d
el.
Evaluation Warning : The document was created with Spire.PDF for Python.