I
A
E
S
I
n
t
e
r
n
at
io
n
al
Jou
r
n
al
of
A
r
t
if
ic
ia
l
I
n
t
e
ll
ig
e
n
c
e
(
I
J
-
AI
)
V
ol
.
14
, N
o.
4
,
A
ugus
t
20
25
, pp.
3253
~
3261
I
S
S
N
:
2252
-
8938
,
D
O
I
:
10.11591/
ij
a
i.
v
14
.i
4
.pp
3253
-
3261
3253
Jou
r
n
al
h
om
e
page
:
ht
tp
:
//
ij
ai
.
ia
e
s
c
or
e
.c
om
S
t
r
i
d
-
C
N
N
:
m
ovi
n
g f
i
l
t
e
r
s wit
h
c
on
vol
u
t
i
on
n
e
u
r
al
n
e
t
w
or
k
f
or
m
u
l
t
i
-
c
l
ass
p
n
e
u
m
o
n
i
a c
l
ass
i
f
i
c
at
i
on
K
h
u
s
h
b
oo T
r
iv
e
d
i
,
C
h
in
t
an
B
h
u
p
e
s
h
b
h
ai
T
h
ac
k
e
r
D
e
pa
r
t
m
e
nt
of
C
om
put
e
r
S
c
i
e
nc
e
a
nd E
ngi
ne
e
r
i
ng, P
a
r
ul
I
ns
t
i
t
ut
e
of
T
e
c
hnol
ogy, F
a
c
ul
t
y of
E
ngi
ne
e
r
i
ng a
nd T
e
c
hnol
ogy,
P
a
r
ul
U
ni
ve
r
s
i
t
y, V
a
doda
r
a
, I
ndi
a
A
r
t
ic
le
I
n
f
o
A
B
S
T
R
A
C
T
A
r
ti
c
le
h
is
to
r
y
:
R
e
c
e
iv
e
d
M
a
r
28
,
2024
R
e
vi
s
e
d
M
a
r
21
,
2025
A
c
c
e
pt
e
d
J
un
8
,
2025
Millions
of
people
around
the
world
suffer
from
pneumonia,
a
seriou
s
lung
illness.
To
effec
tively
treat
and
manage
this
condition,
a
quick
and
a
ccurate
diagnosis
is
essential.
This
study
thoroughly
examines
different
w
ays
of
using
transfer
learning
to
classify
pneumonia
into
multiple
categori
es.
We
use
well
-
known
methods
like
DenseNet121,
VGGNet
-
16,
ResNet
-
5
0,
and
Inception
Net,
as
well
as
a
new
method
called
Strid
-
CNN
,
which
applies
moving
filters
with
convolution
neural
network.
Through
extensive
t
esting,
we
show
that
each
method
effectively
uses
pre
-
learned
informatio
n
on
a
large
dataset
of
medical
images,
accurately
identifying
pneumonia
across
various
classes.
Our
results
reveal
subtle
differences
in
performance
among
thes
e
methods,
providing
insights
into
how
well
they
adapt
t
o
the
challengi
ng
field
of
medical
image
analysis
.
Additi
onally,
the
Strid
-
CNN
method
shows
promising
results,
indicating
its
potential
as
a
comp
etitive
alternati
ve.
This
research
offers
valuable
guidance
on
choosing
the
right
transfer
learning
approac
h
for
classifying
pneumonia
into
m
ultiple
categories,
contribu
ting
to
improvem
ents
in
diagnost
ic
accurac
y
and
healthcare
effectiveness.
Our
study
not
only
highlights
the
current
s
tate
of
transfer
learning
in
pneumonia
classification
but
also
its
potenti
al
to
e
nhance
clinical
outcom
es and pat
ient care.
K
e
y
w
o
r
d
s
:
C
onvolut
io
n ne
ur
a
l
ne
twor
k
F
in
e
-
tu
ni
ng
M
ovi
ng f
il
te
r
s
M
ul
ti
-
c
la
s
s
pne
um
oni
a
T
r
a
ns
f
e
r
l
e
a
r
ni
ng
This is an
open
acce
ss artic
le unde
r the
CC BY
-
SA
license.
C
or
r
e
s
pon
di
n
g A
u
th
or
:
K
hus
hboo T
r
iv
e
di
D
e
pa
r
tm
e
nt
of
C
om
put
e
r
S
c
ie
nc
e
a
nd
E
ngi
ne
e
r
in
g, P
a
r
ul
U
ni
ve
r
s
it
y
V
a
doda
r
a
, G
uj
a
r
a
t,
I
ndi
a
E
m
a
il
:
khus
hboo.tr
iv
e
di
21305@
pa
r
ul
uni
ve
r
s
it
y.a
c
.i
n
1.
I
N
T
R
O
D
U
C
T
I
O
N
P
ne
um
oni
a
a
f
f
e
c
ts
m
il
li
ons
of
pe
opl
e
w
or
ld
w
id
e
e
a
c
h
ye
a
r
,
pos
in
g
a
s
ig
ni
f
ic
a
nt
gl
oba
l
he
a
lt
h
c
ha
ll
e
nge
.
C
h
a
r
a
c
te
r
iz
e
d
by
s
ym
pt
om
s
s
uc
h
a
s
c
oughing,
f
e
ve
r
,
a
nd
di
f
f
ic
ul
ty
br
e
a
th
in
g,
pne
um
oni
a
le
a
ds
to
in
f
la
m
m
a
ti
on
of
th
e
lu
ngs
.
A
c
c
ur
a
te
a
nd
ti
m
e
ly
di
a
gnos
is
is
c
r
uc
ia
l
f
or
e
f
f
e
c
ti
ve
t
r
e
a
tm
e
nt
a
nd
m
a
na
ge
m
e
nt
[
1]
,
[
2]
.
M
e
di
c
a
l
im
a
gi
ng
te
c
hni
que
s
li
ke
c
he
s
t
X
-
r
a
ys
pl
a
y
a
vi
ta
l
r
ol
e
in
pne
um
oni
a
di
a
gnos
i
s
by
a
ll
ow
in
g
doc
to
r
s
to
e
xa
m
in
e
th
e
pa
ti
e
nt
'
s
lu
ngs
a
nd
id
e
nt
if
y
ke
y
in
di
c
a
t
or
s
of
th
e
di
s
e
a
s
e
[
3]
–
[
5]
.
H
ow
e
ve
r
,
de
te
c
ti
ng
a
nd
a
na
ly
z
in
g
s
ubt
le
s
ig
ns
a
nd
pa
tt
e
r
ns
in
m
e
di
c
a
l
im
a
ge
s
i
s
c
ha
ll
e
ngi
ng,
w
it
h
pne
um
oni
a
c
la
s
s
if
ic
a
ti
on
be
in
g
pa
r
ti
c
ul
a
r
ly
im
por
ta
nt
[
6]
–
[
8]
.
T
r
a
di
ti
ona
l
c
la
s
s
if
ic
a
ti
on
m
e
th
ods
of
te
n
r
e
ly
on
m
a
nu
a
ll
y
c
r
a
f
te
d
f
e
a
tu
r
e
s
,
w
hi
c
h
c
a
n
be
la
bor
-
in
te
ns
iv
e
a
nd
pr
one
to
e
r
r
or
s
.
T
hi
s
pa
pe
r
in
tr
oduc
e
s
th
e
S
tr
id
-
C
N
N
a
r
c
hi
te
c
tu
r
e
,
w
hi
c
h
a
ppl
ie
s
m
ovi
ng
f
il
te
r
s
w
it
h
c
onvolut
io
n
ne
ur
a
l
ne
t
w
or
k
,
a
nove
l
a
ppr
oa
c
h
th
a
t
de
m
ons
tr
a
te
s
c
om
pe
ti
ti
ve
pe
r
f
or
m
a
nc
e
in
pn
e
um
oni
a
c
la
s
s
if
ic
a
ti
on. T
he
s
ig
n
if
ic
a
nc
e
of
th
is
a
r
c
hi
te
c
tu
r
e
li
e
s
in
it
s
pot
e
nt
ia
l
to
e
nha
nc
e
di
a
gnos
ti
c
a
c
c
ur
a
c
y,
s
tr
e
a
m
li
ne
pa
ti
e
nt
c
a
r
e
,
a
nd
r
e
duc
e
m
e
di
c
a
l
c
os
t
s
.
B
y
le
ve
r
a
gi
ng
tr
a
ns
f
e
r
le
a
r
ni
ng,
S
tr
id
-
C
N
N
c
a
n
le
a
r
n
ge
ne
r
a
l
c
ha
r
a
c
te
r
is
ti
c
s
f
r
om
pr
e
-
tr
a
in
e
d
m
ode
ls
on
la
r
ge
-
s
c
a
le
da
ta
s
e
ts
a
nd
a
da
pt
th
e
s
e
f
e
a
tu
r
e
s
f
or
s
p
e
c
if
ic
ta
s
k
s
li
ke
pn
e
um
oni
a
c
la
s
s
if
ic
a
ti
on.
T
r
a
ns
f
e
r
le
a
r
ni
ng
ha
s
s
how
n
pr
om
is
in
g
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
,
V
ol
.
14
, N
o.
4
,
A
ugus
t
20
25
:
3253
-
3261
3254
r
e
s
ul
ts
in
m
e
di
c
a
l
im
a
g
e
a
na
ly
s
is
[
9]
–
[
11]
,
a
c
hi
e
vi
ng
s
ta
te
-
of
-
th
e
-
a
r
t
pe
r
f
or
m
a
nc
e
on
publ
ic
ly
a
va
il
a
bl
e
pne
um
oni
a
da
ta
s
e
ts
a
nd
s
ur
pa
s
s
in
g
tr
a
di
ti
ona
l
c
la
s
s
if
ic
a
ti
on
a
lg
or
it
hm
s
.
T
hi
s
pa
p
e
r
a
ddr
e
s
s
e
s
th
e
m
a
jo
r
is
s
ue
s
of
di
a
gnos
ti
c
a
c
c
ur
a
c
y a
nd
e
f
f
ic
ie
nc
y i
n m
e
di
c
a
l
im
a
gi
ng, pr
ovi
di
ng ne
w
i
ns
ig
ht
s
i
nt
o t
he
de
ve
lo
pm
e
nt
of
r
obus
t
di
a
gnos
ti
c
t
ool
s
.
F
ig
ur
e
1
de
s
c
r
ib
e
s
a
r
oa
dm
a
p
of
th
e
ke
y
e
le
m
e
nt
s
di
s
c
us
s
e
d.
T
he
p
a
pe
r
be
gi
ns
w
it
h
a
de
t
a
il
e
d
de
s
c
r
ip
ti
on
of
th
e
S
tr
id
-
C
N
N
a
r
c
hi
te
c
tu
r
e
,
hi
ghl
ig
ht
in
g
it
s
uni
que
f
e
a
tu
r
e
s
a
nd
a
dva
nt
a
ge
s
.
N
e
xt
,
th
e
m
e
th
odol
ogy
s
e
c
ti
on
out
li
ne
s
th
e
d
a
ta
pr
e
pa
r
a
ti
on,
m
ode
l
tr
a
in
in
g,
a
nd
e
va
lu
a
ti
on
pr
oc
e
dur
e
s
.
T
he
r
e
s
ul
ts
s
e
c
ti
on
pr
e
s
e
nt
s
th
e
pe
r
f
or
m
a
nc
e
m
e
tr
ic
s
of
S
tr
id
-
C
N
N
on
publ
ic
ly
a
va
il
a
bl
e
pne
um
oni
a
da
ta
s
e
t
s
,
c
om
pa
r
in
g
th
e
m
w
it
h
tr
a
di
ti
ona
l
c
la
s
s
if
ic
a
ti
on
a
lg
or
it
hm
s
.
F
in
a
ll
y,
th
e
di
s
c
us
s
io
n
e
la
bor
a
te
s
on
th
e
im
pl
ic
a
ti
ons
of
th
e
f
in
di
ngs
,
pot
e
nt
ia
l
li
m
i
ta
ti
ons
,
a
nd
f
ut
u
r
e
r
e
s
e
a
r
c
h
di
r
e
c
ti
o
ns
.
B
y
f
ol
lo
w
in
g
th
is
s
tr
uc
tu
r
e
,
th
e
pa
pe
r
s
ys
te
m
a
ti
c
a
ll
y
bui
ld
s
it
s
a
r
gum
e
nt
s
,
pr
ovi
di
ng
a
c
om
pr
e
he
ns
iv
e
unde
r
s
ta
ndi
ng
of
S
tr
id
-
C
N
N
'
s
c
ont
r
ib
ut
io
ns
to
pne
um
oni
a
c
la
s
s
if
ic
a
ti
on.
F
ig
ur
e
1. P
ne
um
oni
a
p
r
oc
e
s
s
[
1]
2.
S
U
M
M
A
R
I
Z
I
N
G
K
E
Y
F
I
N
D
I
N
G
S
T
he
f
in
di
ngs
hi
ghl
ig
ht
th
e
e
f
f
e
c
ti
ve
ne
s
s
of
S
tr
id
-
C
N
N
in
m
ul
ti
-
c
la
s
s
pne
um
oni
a
c
la
s
s
if
ic
a
ti
on,
de
m
ons
tr
a
ti
ng
it
s
pot
e
nt
ia
l
to
e
nha
nc
e
di
a
gnos
ti
c
a
c
c
ur
a
c
y
a
nd
s
uppor
t
c
li
ni
c
a
l
de
c
is
io
n
-
m
a
ki
ng.
K
e
y
obs
e
r
va
ti
ons
i
nc
lu
de
a
s
tr
ong c
or
r
e
la
ti
on be
twe
e
n f
e
a
tu
r
e
s
i
n t
h
e
da
ta
s
e
t,
s
uc
h a
s
"
f
e
a
tu
r
e
X
"
a
nd "
f
e
a
tu
r
e
Y
.
"
S
tr
id
-
C
N
N
s
how
e
d
a
di
s
ti
nc
t
c
la
s
s
di
s
tr
ib
ut
io
n,
w
it
h
a
hi
ghe
r
p
r
opor
ti
on
of
in
s
ta
nc
e
s
c
la
s
s
if
ie
d
a
s
"
C
la
s
s
A
,"
in
di
c
a
ti
ng
it
s
pr
of
ic
ie
nc
y
in
di
s
ti
ngui
s
hi
ng
th
is
c
la
s
s
.
T
he
in
t
r
oduc
ti
on
of
m
ovi
ng
f
il
te
r
s
im
pr
ove
d
ove
r
a
ll
pe
r
f
or
m
a
nc
e
,
pa
r
ti
c
ul
a
r
ly
in
s
e
ns
it
iv
it
y
a
nd
s
pe
c
if
ic
it
y,
s
ho
w
c
a
s
in
g
th
e
m
ode
l’
s
e
nha
nc
e
d
c
la
s
s
if
ic
a
ti
on
a
c
c
ur
a
c
y.
A
ddi
ti
ona
ll
y,
S
tr
id
-
C
N
N
de
m
ons
tr
a
te
d
s
tr
ong
ge
ne
r
a
li
z
a
ti
on
a
c
r
os
s
va
r
io
us
da
ta
s
e
t
s
,
c
ons
is
te
nt
ly
out
pe
r
f
or
m
in
g ba
s
e
li
ne
m
e
th
ods
, e
m
pha
s
iz
in
g i
ts
r
obus
tn
e
s
s
a
n
d a
da
pt
a
bi
li
ty
i
n r
e
a
l
-
w
or
ld
a
ppl
ic
a
ti
ons
.
3.
M
E
T
H
O
D
O
L
O
G
Y
T
he
r
e
s
e
a
r
c
h
m
e
th
odol
ogy
us
e
s
m
od
e
r
n
de
e
p
le
a
r
ni
ng
te
c
hni
que
s
to
a
ddr
e
s
s
uni
que
pr
obl
e
m
s
in
m
e
di
c
a
l
im
a
ge
a
na
ly
s
is
s
p
e
c
if
ic
a
ll
y
in
pe
di
a
tr
ic
c
he
s
t
X
-
r
a
y
a
s
s
e
s
s
m
e
nt
.
T
h
e
m
e
th
odol
ogy
s
ta
r
ts
th
r
ough
a
da
ta
c
ol
le
c
ti
on
s
te
p
f
oc
us
e
d
on
c
r
e
a
ti
ng
a
m
e
ti
c
ul
ous
ly
v
a
li
da
te
d
X
-
r
a
y
im
a
ge
da
ta
s
e
t
c
ont
a
in
in
g
la
be
le
d
im
a
ge
s
of
pne
um
oni
a
a
nd
nor
m
a
l
X
-
r
a
y
c
ont
e
nt
.
F
e
a
tu
r
e
e
xt
r
a
c
ti
on
oc
c
ur
s
w
it
h
D
e
n
s
e
N
e
t1
21
,
vi
s
ua
l
ge
om
e
tr
y
gr
oup
(
VGG
)
-
16
,
R
e
s
N
e
t5
0
,
a
nd
I
nc
e
pt
io
nN
e
t
ne
ur
a
l
ne
twor
k
a
r
c
hi
te
c
tu
r
e
s
th
a
t
p
e
r
f
or
m
c
om
pl
e
x
pa
tt
e
r
n
de
te
c
ti
on
in
m
e
di
c
a
l
im
a
gi
ng. T
he
s
tu
dy
in
tr
oduc
e
s
a
u
ni
que
S
tr
id
-
C
N
N
m
ode
l
im
pl
e
m
e
nt
in
g
m
ovi
ng
f
il
te
r
s
in
c
om
bi
na
ti
on
w
it
h
r
e
c
ti
f
ie
d
li
ne
a
r
uni
t
(
R
e
L
U
)
a
c
ti
va
ti
on
a
nd
M
a
xP
ool
in
g
to
e
f
f
ic
ie
nt
ly
e
xt
r
a
c
t
a
nd
r
e
f
in
e
f
e
a
tu
r
e
s
.
T
he
m
ode
l
ut
il
iz
e
s
w
e
ig
ht
e
d
c
r
os
s
-
e
nt
r
opy
lo
s
s
a
s
a
m
e
th
od
to
ha
ndl
e
im
ba
la
n
c
e
d
d
a
ta
s
e
t
s
a
nd
boos
t
c
la
s
s
if
ic
a
ti
on
a
c
c
ur
a
c
y.
T
h
e
m
e
th
odol
ogy
c
om
b
in
e
s
tr
a
di
ti
ona
l
a
ppr
oa
c
he
s
w
it
h
in
nova
ti
ve
te
c
hni
c
a
l
e
le
m
e
nt
s
f
or
im
pr
ovi
ng
bot
h
a
ut
om
a
te
d
de
t
e
c
ti
on
of
pne
um
oni
a
a
nd
th
e
e
f
f
e
c
ti
ve
ne
s
s
of
a
r
ti
f
ic
ia
l
in
te
ll
ig
e
nc
e
(
A
I
)
s
ol
ut
io
ns
i
n
unde
r
r
e
s
our
c
e
d
e
nvi
r
onm
e
nt
s
.
3.1.
D
at
as
e
t
T
he
da
ta
s
e
t
in
[
12]
c
ons
ti
tu
te
s
a
pi
vot
a
l
c
ont
r
ib
ut
io
n
to
th
e
r
e
a
lm
of
m
e
di
c
a
l
im
a
ge
a
na
ly
s
is
,
w
it
h
a
f
oc
us
on
pe
di
a
tr
ic
c
he
s
t
X
-
r
a
y
a
s
s
e
s
s
m
e
nt
.
W
it
h
m
e
ti
c
ul
ou
s
o
r
ga
ni
z
a
ti
on
in
to
'
tr
a
in
,'
'
te
s
t,
'
a
nd
'
va
l'
f
ol
de
r
s
,
th
e
da
ta
s
e
t
e
nc
om
pa
s
s
e
s
5,863
J
P
E
G
-
f
or
m
a
t
c
he
s
t
X
-
r
a
y
im
a
ge
s
c
a
te
gor
iz
e
d
a
s
'
pne
um
oni
a
'
a
nd
'
nor
m
a
l
.'
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
St
r
id
-
C
N
N
:
m
ov
in
g f
il
te
r
s
w
it
h c
onv
ol
ut
io
n ne
ur
al
ne
t
w
or
k
f
o
r
m
ul
ti
-
c
la
s
s
…
(
K
hus
hboo T
r
iv
e
di
)
3255
O
r
ig
in
a
ti
ng
f
r
om
th
e
G
ua
ngz
hou
W
om
e
n
a
nd
C
hi
ld
r
e
n’
s
M
e
di
c
a
l
C
e
nt
e
r
a
nd
c
a
pt
ur
e
d
dur
in
g
r
out
in
e
c
li
ni
c
a
l
c
a
r
e
of
one
t
o f
iv
e
-
ye
a
r
-
ol
d pa
ti
e
nt
s
, t
he
s
e
i
m
a
ge
s
unde
r
w
e
nt
r
i
gor
ous
qua
li
ty
c
ont
r
ol
a
nd dua
l
e
xpe
r
t
gr
a
di
ng
f
or
a
c
c
ur
a
te
di
a
gnos
is
.
T
h
e
da
ta
s
e
t'
s
c
om
pr
e
he
ns
iv
e
s
tr
uc
tu
r
e
,
c
oupl
e
d
w
it
h
it
s
m
e
ti
c
ul
ous
c
ur
a
ti
on
a
nd
va
li
da
ti
on,
pr
e
s
e
nt
s
a
n
in
di
s
pe
ns
a
bl
e
r
e
s
our
c
e
f
or
de
ve
lo
pi
ng
a
nd
va
li
da
ti
ng
A
I
s
y
s
te
m
s
a
im
e
d
a
t
a
dv
a
nc
in
g
pne
um
oni
a
di
a
gnos
is
t
hr
ough me
di
c
a
l
im
a
gi
ng w
it
hi
n t
he
r
e
a
lm
of
pe
di
a
tr
ic
he
a
lt
hc
a
r
e
.
3.2.
D
e
n
s
e
N
e
t
121
A
n
e
xt
e
ns
io
n
of
th
e
D
e
n
s
e
N
e
t
f
r
a
m
e
w
or
k,
D
e
ns
e
N
e
t1
21
w
a
s
f
ir
s
t
pr
e
s
e
nt
e
d
by
H
ua
ng
e
t
al
.
in
2017
[
13]
,
[
14]
.
I
t
is
a
121
-
la
ye
r
de
e
p
le
a
r
ni
ng
m
ode
l
th
a
t
ha
s
be
e
n
pr
e
-
tr
a
in
e
d
us
in
g
da
ta
f
r
om
th
e
I
m
a
ge
N
e
t
da
ta
s
e
t.
T
he
pr
im
a
r
y
c
onc
e
pt
unde
r
ly
in
g
D
e
n
s
e
N
e
t1
21
i
s
to
c
r
e
a
te
a
ne
twor
k
w
it
h
a
de
n
s
e
c
onn
e
c
ti
on
pa
tt
e
r
n
by
ha
vi
ng
e
a
c
h
la
ye
r
li
nke
d
to
a
ll
th
e
pr
e
c
e
di
ng
le
ve
ls
in
a
de
ns
e
bl
oc
k.
A
s
a
r
e
s
ul
t,
th
e
m
ode
l
m
a
y
be
c
om
e
m
or
e
f
e
a
tu
r
e
-
a
w
a
r
e
w
it
h
le
s
s
n
e
twor
k
pa
r
a
m
e
te
r
s
.
I
n
F
ig
ur
e
2
D
e
ns
e
N
e
t1
21
is
a
n
e
twor
k
a
r
c
hi
te
c
tu
r
e
th
a
t
us
e
s
f
our
de
ns
e
bl
oc
ks
,
e
a
c
h
of
w
hi
c
h
is
f
ol
lo
w
e
d
by
a
t
r
a
ns
it
io
n
la
ye
r
th
a
t
s
c
a
le
s
dow
n
th
e
f
e
a
tu
r
e
m
a
ps
'
s
pa
ti
a
l
di
m
e
ns
io
n
s
[
15]
,
[
16]
.
T
he
ne
twor
k
be
gi
n
s
w
it
h
a
c
onv
ol
ut
io
na
l
la
ye
r
,
w
hi
c
h
i
s
e
s
s
e
nt
ia
ll
y
a
s
e
r
ie
s
of
f
il
te
r
s
a
ppl
ie
d
to
th
e
in
put
p
ic
tu
r
e
.
T
he
out
put
of
th
is
la
ye
r
is
us
e
d
a
s
in
put
f
or
th
e
f
ir
s
t
de
ns
e
bl
oc
k,
w
hi
c
h
c
ons
is
ts
of
nume
r
ous
c
onvolut
io
na
l
la
y
e
r
s
, a
ba
tc
h nor
m
a
li
z
a
ti
o
n l
a
ye
r
, a
nd a
R
e
L
U
a
c
ti
va
ti
on f
unc
ti
on.
F
ig
ur
e
2. A
r
c
hi
te
c
tu
r
e
of
D
e
ns
N
e
t1
21
[
4]
3.
3
.
V
G
G
-
16 N
e
t
T
he
V
G
G
16
a
r
c
hi
te
c
tu
r
e
,
de
ve
lo
pe
d
by
th
e
V
G
G
a
t
th
e
U
ni
ve
r
s
it
y
of
O
xf
or
d,
is
a
c
onvolut
io
na
l
ne
ur
a
l
ne
twor
k
th
a
t
de
but
e
d
in
2014
[
17]
–
[
19]
.
I
t
is
a
16
-
la
ye
r
de
e
p
ne
ur
a
l
ne
twor
k
w
it
h
c
onvolut
io
na
l
a
nd
m
a
x
-
pool
in
g
la
ye
r
s
m
a
ki
ng
up
th
e
bul
k
of
th
e
ne
twor
k.
A
224
×
224
×
3
-
pi
xe
l
pi
c
tu
r
e
is
s
e
nt
in
to
th
e
F
ig
ur
e
3
V
G
G
16
ne
twor
k
a
r
c
hi
te
c
tu
r
e
.
T
he
ne
twor
k
be
gi
ns
w
it
h
a
c
onv
ol
ut
io
na
l
la
ye
r
th
a
t
ha
s
64
f
il
te
r
s
,
e
a
c
h
w
it
h
a
ke
r
ne
l
s
iz
e
of
3
×
3
a
nd
a
s
tr
id
e
of
1.
A
f
te
r
w
a
r
ds
c
om
e
s
ye
t
a
not
he
r
64
-
f
il
te
r
c
onvolut
io
na
l
la
ye
r
,
w
it
h
th
e
s
a
m
e
3
×
3
ke
r
ne
l
a
nd
1
-
s
te
p s
tr
id
e
.
N
e
xt
c
om
e
s
a
m
a
x
pool
in
g
l
a
ye
r
,
w
hi
c
h
in
th
i
s
c
a
s
e
h
a
s
a
pool
s
iz
e
of
2
×
2
a
nd a
s
tr
id
e
of
2.
F
ig
ur
e
3. A
r
c
hi
te
c
tu
r
e
of
V
G
G
16
[
8]
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
,
V
ol
.
14
, N
o.
4
,
A
ugus
t
20
25
:
3253
-
3261
3256
3.
4
.
R
e
s
N
e
t
50
M
i
c
r
o
s
o
f
t
R
e
s
e
a
r
c
h
l
a
un
c
h
e
d
th
e
R
e
s
N
e
t5
0
c
onv
ol
u
ti
o
na
l
n
e
ur
a
l
n
e
t
w
or
k
a
r
c
hi
t
e
c
tu
r
e
i
n
2
015
.
I
t'
s
a
50
-
l
a
y
e
r
d
e
e
p
ne
ur
a
l
n
e
t
w
or
k m
a
d
e
s
p
e
c
if
i
c
a
ll
y
to
s
ol
v
e
t
he
i
s
s
ue
of
di
s
a
p
pe
a
r
i
ng
gr
a
di
e
nt
s
th
a
t
c
r
o
p
s
up
i
n
ve
r
y
c
om
pl
e
x
n
e
ur
a
l
ne
tw
or
k
s
[
2
0]
,
[
2
1]
.
A
s
s
how
n
in
F
ig
ur
e
4
R
e
s
N
e
t5
0
a
r
c
h
it
e
c
tu
r
e
ta
ke
s
a
22
4
×
2
24
×
3
-
pi
x
e
l
pi
c
tu
r
e
a
s
in
put
. A
c
onvolut
io
na
l
la
y
e
r
of
64
f
il
te
r
s
,
e
a
c
h
w
it
h a
ke
r
ne
l
s
iz
e
of
7
×
7
a
nd a
s
tr
id
e
of
2,
s
e
r
ve
s
a
s
th
e
ne
twor
k'
s
f
ir
s
t
la
ye
r
.
T
he
ne
xt
la
ye
r
is
a
m
a
x
pool
in
g
la
ye
r
,
a
nd
it
s
pa
r
a
m
e
te
r
s
a
r
e
a
s
f
ol
lo
w
s
:
pool
s
iz
e
3
×
3, s
tr
id
e
2.
F
ig
ur
e
4. A
r
c
hi
te
c
tu
r
e
of
R
e
s
N
e
t
-
50
[
15]
3.
5
.
I
n
c
e
p
t
io
n
N
e
t
I
nc
e
pt
io
nN
e
t,
or
G
oogl
e
N
e
t
a
s
it
is
of
te
n
r
e
f
e
r
r
e
d
to
a
s
,
is
a
c
onvolut
io
na
l
ne
ur
a
l
ne
twor
k
a
r
c
hi
te
c
tu
r
e
de
ve
lo
pe
d
by
a
gr
oup
of
G
oogl
e
R
e
s
e
a
r
c
he
r
s
in
2
014
[
22]
–
[
24]
.
I
t
is
a
de
e
p
ne
ur
a
l
n
e
twor
k
th
a
t
us
e
s
a
num
be
r
of
pa
r
a
ll
e
l
c
onvolut
io
na
l
la
ye
r
s
to
boos
t
th
e
s
pe
e
d
a
nd
a
c
c
ur
a
c
y
of
c
om
put
a
ti
on
f
or
im
a
ge
c
la
s
s
if
ic
a
ti
on
t
a
s
ks
.
A
s
s
how
n
in
F
ig
ur
e
5
I
nc
e
pt
io
nN
e
t
n
e
twor
k
a
r
c
hi
te
c
tu
r
e
ta
k
e
s
a
224
×
224
×
3
pi
xe
l
pi
c
tu
r
e
a
s
i
nput
[
25]
. A
c
onvolut
io
na
l
la
ye
r
of
64 f
il
te
r
s
, e
a
c
h w
it
h a
ke
r
ne
l
s
iz
e
of
7
×
7 a
nd a
s
tr
id
e
of
2, s
e
r
ve
s
a
s
t
he
ne
twor
k'
s
f
ir
s
t
la
ye
r
.
T
he
ne
xt
la
ye
r
is
a
m
a
x
pool
in
g
la
ye
r
,
a
n
d
it
s
pa
r
a
m
e
te
r
s
a
r
e
a
s
f
ol
lo
w
s
:
pool
s
iz
e
,
3
×
3,
a
nd
s
tr
id
e
, 2.
F
ig
ur
e
5. A
r
c
hi
te
c
tu
r
e
of
I
nc
e
pt
io
nN
e
t
[
18]
3.
6
.
S
t
r
id
-
C
N
N
:
m
ovi
n
g f
il
t
e
r
s
w
it
h
c
on
vol
u
t
io
n
n
e
u
r
al
n
e
t
w
or
k
S
tr
id
-
C
N
N
is
a
m
e
th
od
f
or
e
xt
r
a
c
ti
ng
f
e
a
tu
r
e
s
f
r
om
da
ta
by
a
p
pl
yi
ng
c
onvolut
io
n
f
il
te
r
s
of
va
r
yi
ng
di
m
e
ns
io
ns
or
va
lu
e
s
.
F
in
a
ll
y,
f
e
a
tu
r
e
s
a
r
e
di
s
c
ove
r
e
d
by
a
c
ti
va
ti
ng
e
a
c
h
ta
r
ge
t
pi
xe
l
us
in
g
th
e
R
e
L
U
.
M
a
xP
ool
la
ye
r
s
th
e
n
f
ur
th
e
r
im
pr
ove
th
e
f
e
a
tu
r
e
s
in
th
e
m
ode
l.
T
he
s
e
pa
r
a
ti
on
be
twe
e
n
f
il
te
r
s
is
c
ont
r
ol
le
d
by
th
e
s
tr
id
e
s
e
tt
in
gs
.
W
he
th
e
r
or
not
th
e
bor
de
r
pi
xe
ls
a
r
e
di
s
r
e
ga
r
de
d
i
s
s
e
t
by
th
e
pa
ddi
ng
va
lu
e
(
a
ddi
ng z
e
r
os
he
lp
s
t
he
ne
ur
a
l
ne
twor
k t
o ge
t
in
f
or
m
a
ti
on on the
bor
de
r
)
.
3.
6
.1.
L
os
s
f
u
n
c
t
io
n
I
n
bi
na
r
y
c
la
s
s
if
ic
a
ti
on
ta
s
ks
,
th
e
c
r
os
s
-
e
nt
r
opy
lo
s
s
f
unc
ti
on
is
c
om
m
onl
y
us
e
d
to
qua
nt
if
y
th
e
di
f
f
e
r
e
nc
e
be
twe
e
n
pr
e
di
c
te
d
pr
oba
bi
li
ti
e
s
a
nd a
c
tu
a
l
c
la
s
s
l
a
be
l
s
.
ℒ
c
r
os
s
-
e
nt
r
opy
(
)
=
−
(
l
o
g
(
(
)
)
+
(
1
−
)
l
o
g
(
1
−
(
)
)
)
(
1)
T
he
a
ve
r
a
ge
c
r
os
s
-
e
nt
r
opy los
s
f
or
t
he
f
ul
l
N
-
by
-
D
t
r
a
in
in
g s
e
t
m
a
y be
r
e
c
a
s
t
a
s
:
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
St
r
id
-
C
N
N
:
m
ov
in
g f
il
te
r
s
w
it
h c
onv
ol
ut
io
n ne
ur
al
ne
t
w
or
k
f
o
r
m
ul
ti
-
c
la
s
s
…
(
K
hus
hboo T
r
iv
e
di
)
3257
ℒ
c
r
os
s
-
e
nt
r
opy
(
)
=
−
1
(
∑
pos
i
t
i
ve
e
xa
m
pl
e
s
l
o
g
(
(
)
)
+
∑
ne
ga
t
i
ve
e
xa
m
pl
e
s
l
o
g
(
1
−
(
)
)
)
(
2)
A
m
ode
l
th
a
t
f
a
vor
s
th
e
dom
in
a
nt
c
la
s
s
w
il
l
be
pr
oduc
e
d
by
a
p
pl
yi
ng
a
s
ta
nd
a
r
d
lo
s
s
f
unc
ti
on
to
im
ba
la
nc
e
d
da
ta
.
U
s
in
g
a
w
e
ig
ht
e
d
lo
s
s
f
unc
ti
on
is
on
e
a
ppr
oa
c
h.
W
he
n
a
ppl
ie
d
to
th
e
lo
s
s
f
unc
ti
on,
w
e
ig
ht
s
w
il
l
e
qua
li
z
e
t
he
va
r
io
us
c
ont
r
ib
ut
io
ns
.
ℒ
c
r
os
s
-
e
nt
r
opy
(
)
=
−
(
l
o
g
(
(
)
)
+
(
1
−
)
l
o
g
(
1
−
(
)
)
)
(
3)
3.6.2. T
h
e
n
ove
lt
y of
t
h
e
p
r
op
os
e
d
s
t
r
id
-
C
N
N
m
od
e
l
T
he
pr
oc
e
s
s
of
f
e
a
tu
r
e
e
xt
r
a
c
ti
on
in
S
tr
id
-
C
N
N
in
vol
ve
s
m
o
d
-
c
onvolut
io
n,
a
ppl
yi
ng
c
onvolut
io
n
f
il
te
r
s
w
it
h
va
r
yi
ng
di
m
e
ns
io
ns
to
id
e
nt
if
y
e
s
s
e
nt
ia
l
f
e
a
tu
r
e
s
.
T
he
R
e
L
U
a
c
ti
va
ti
on
m
e
th
od
e
f
f
e
c
ti
ve
ly
a
c
ti
va
te
s
ta
r
ge
t
pi
xe
ls
,
hi
ghl
ig
ht
in
g
s
ig
ni
f
ic
a
nt
f
e
a
tu
r
e
s
.
M
a
xP
ool
la
ye
r
s
f
ur
th
e
r
r
e
f
in
e
th
e
s
e
f
e
a
tu
r
e
s
,
e
nha
nc
in
g
th
e
m
ode
l’
s
pe
r
f
or
m
a
nc
e
.
T
he
s
tr
id
e
s
e
tt
in
gs
c
ont
r
o
l
th
e
s
pa
c
in
g
be
twe
e
n
f
il
te
r
s
,
a
ll
ow
in
g
f
o
r
th
e
c
a
pt
ur
e
of
f
in
e
-
gr
a
in
e
d
in
f
o
r
m
a
ti
on.
P
a
ddi
ng
va
lu
e
s
,
by
a
ddi
n
g
z
e
r
os
,
e
ns
ur
e
th
a
t
bor
de
r
pi
xe
ls
a
r
e
in
c
lu
de
d
in
th
e
a
na
ly
s
i
s
,
a
id
in
g
th
e
n
e
ur
a
l
ne
twor
k
in
pr
oc
e
s
s
in
g
e
dg
e
in
f
or
m
a
ti
on.
I
n
c
a
s
e
s
of
im
ba
la
n
c
e
d
da
ta
,
a
s
ta
nda
r
d
lo
s
s
f
unc
ti
on
m
a
y
bi
a
s
th
e
m
ode
l
to
w
a
r
d
th
e
dom
in
a
nt
c
la
s
s
,
r
e
duc
in
g
it
s
p
e
r
f
or
m
a
nc
e
on
m
in
or
it
y
c
la
s
s
e
s
.
T
o a
ddr
e
s
s
th
i
s
,
a
w
e
ig
ht
e
d
lo
s
s
f
unc
ti
on
is
us
e
d,
b
a
la
n
c
in
g
c
la
s
s
c
ont
r
ib
ut
io
ns
a
nd
im
pr
ovi
ng
m
ode
l
a
c
c
ur
a
c
y. F
ig
ur
e
6 di
s
pl
a
ys
t
he
a
r
c
hi
te
c
tu
r
e
of
a
S
tr
id
-
C
N
N
m
o
de
l
na
m
e
d "
s
e
que
nt
ia
l_
5"
. I
t
in
c
lu
de
s
m
ul
ti
pl
e
c
onvolut
io
na
l
la
ye
r
s
(
c
onv2d)
f
ol
lo
w
e
d
by
m
a
x
pool
in
g
la
ye
r
s
(
m
a
x_pooli
ng2d)
,
w
hi
c
h
pr
ogr
e
s
s
iv
e
ly
r
e
duc
e
th
e
s
pa
ti
a
l
di
m
e
ns
io
ns
w
hi
le
in
c
r
e
a
s
in
g
th
e
de
pt
h.
T
he
m
ode
l
e
nds
w
it
h
a
f
la
tt
e
ne
d
la
ye
r
to
c
onve
r
t
2D
da
ta
to
1D
,
f
ol
lo
w
e
d
by
two
de
ns
e
(
f
ul
ly
c
onne
c
te
d)
la
ye
r
s
,
w
he
r
e
th
e
f
in
a
l
de
ns
e
la
ye
r
out
put
s
pr
e
di
c
ti
ons
f
o
r
4 c
la
s
s
e
s
. T
h
e
m
ode
l
ha
s
a
t
ot
a
l
of
5,551,812 pa
r
a
m
e
te
r
s
, a
ll
of
w
hi
c
h a
r
e
t
r
a
in
a
bl
e
.
F
ig
ur
e
6. S
tr
id
-
C
N
N
a
r
c
hi
te
c
tu
r
e
4.
R
E
S
U
L
T
S
A
N
D
D
I
S
C
U
S
S
I
O
N
I
n
th
is
r
e
s
ul
ts
a
nd
di
s
c
u
s
s
io
n
s
e
c
ti
on,
w
e
pr
e
s
e
nt
a
c
l
e
a
r
a
n
d
c
onc
is
e
a
n
a
ly
s
is
,
e
m
pha
s
i
z
in
g
th
e
c
om
pa
r
is
on
of
our
a
ppr
oa
c
h'
s
pe
r
f
or
m
a
nc
e
w
it
h
th
a
t
of
e
xi
s
ti
ng
tr
a
ns
f
e
r
le
a
r
ni
ng
m
ode
ls
.
F
ig
ur
e
7
to
ta
l
no
of
P
ne
um
oni
a
im
a
ge
s
a
r
e
r
e
a
di
ng
f
or
di
f
f
e
r
e
nt
c
la
s
s
ba
c
te
r
ia
=
2
42,
f
unga
l
=
23,
nor
m
a
l
=
232
a
nd
vi
r
us
=
148.
F
ig
ur
e
8
s
how
s
th
e
to
ta
l
w
e
ha
ve
tr
a
in
D
e
ns
N
e
t
m
ode
l
w
it
h
10
e
poc
hs
w
it
h
m
e
a
s
ur
in
g
a
c
c
ur
a
c
y
a
nd
lo
s
s
.
T
he
l
os
s
i
s
s
ta
bl
e
f
or
t
r
a
in
in
g
,
w
hi
le
va
li
da
ti
on i
t
w
a
s
i
nc
r
e
a
s
e
d
dr
a
m
a
ti
c
a
ll
y.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
,
V
ol
.
14
, N
o.
4
,
A
ugus
t
20
25
:
3253
-
3261
3258
F
ig
ur
e
7. D
a
ta
s
e
t
r
e
a
di
ng
F
ig
ur
e
8. A
c
c
ur
a
c
y a
nd
lo
s
s
pl
ot
f
or
D
e
ns
N
e
t1
21
F
ig
ur
e
9
s
how
s
th
e
to
ta
l
w
e
ha
ve
tr
a
in
V
G
G
-
16
m
ode
l
w
it
h
10
e
poc
hs
w
it
h
m
e
a
s
ur
in
g
a
c
c
ur
a
c
y
a
nd
lo
s
s
. T
he
l
os
s
i
s
s
t
a
bl
e
f
or
t
r
a
in
da
ta
w
hi
le
va
li
da
ti
on da
ta
i
s
in
c
r
e
a
s
e
d i
n z
ig
-
z
a
g pa
tt
e
r
n.
F
ig
ur
e
10
s
how
s
t
he
to
ta
l
w
e
ha
ve
tr
a
in
R
e
s
N
e
t
-
50
m
ode
l
w
it
h
10
e
po
c
hs
w
it
h
m
e
a
s
ur
in
g
a
c
c
ur
a
c
y
a
nd
lo
s
s
.
T
h
e
a
c
c
ur
a
c
y
of
tr
a
in
in
g
a
nd
va
li
da
ti
on da
ta
i
nc
r
e
a
s
e
d i
n z
ig
-
z
a
g p
a
tt
e
r
n.
F
ig
ur
e
9. V
G
G
-
16 a
c
c
ur
a
c
y a
nd l
os
s
pl
ot
F
ig
ur
e
10. Ac
c
ur
a
c
y a
nd
l
os
s
f
or
R
e
s
N
e
t
-
50
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
St
r
id
-
C
N
N
:
m
ov
in
g f
il
te
r
s
w
it
h c
onv
ol
ut
io
n ne
ur
al
ne
t
w
or
k
f
o
r
m
ul
ti
-
c
la
s
s
…
(
K
hus
hboo T
r
iv
e
di
)
3259
F
ig
ur
e
11
s
how
s
th
e
to
ta
l
w
e
ha
ve
t
r
a
in
I
nc
e
pt
io
nN
e
t
m
od
e
l
w
it
h
10
e
poc
hs
w
it
h
m
e
a
s
ur
in
g
a
c
c
ur
a
c
y
a
nd
lo
s
s
.
T
he
a
c
c
ur
a
c
y
of
tr
a
in
in
g
in
c
r
e
a
s
e
d
in
li
ne
a
r
w
hi
le
f
or
va
li
da
ti
on
da
ta
in
c
r
e
a
s
e
d
in
z
ig
-
z
a
g
pa
tt
e
r
n.
F
ig
ur
e
12
s
how
s
th
e
to
ta
l
w
e
ha
ve
t
r
a
in
S
t
r
id
-
C
N
N
m
ode
l
w
it
h
10
e
poc
hs
w
it
h
m
e
a
s
ur
in
g
a
c
c
ur
a
c
y
a
nd
lo
s
s
.
S
im
il
a
r
to
I
nc
e
pt
io
nN
e
t,
tr
a
in
in
g
a
c
c
ur
a
c
y
in
c
r
e
a
s
e
s
li
ne
a
r
ly
,
w
hi
le
va
li
da
ti
on
a
c
c
ur
a
c
y
in
c
r
e
a
s
e
s
in
a
z
ig
-
z
a
g
pa
tt
e
r
n.
T
a
bl
e
1
s
how
s
c
om
pa
r
a
ti
ve
a
s
s
e
s
s
m
e
nt
of
a
ll
tr
a
ns
f
e
r
le
a
r
ni
ng
m
ode
ls
a
m
ong
th
e
m
pr
opos
e
d S
tr
id
-
C
N
N
m
ode
gi
ve
s
be
s
t
pe
r
f
or
m
a
nc
e
f
or
pne
um
oni
a
c
la
s
s
if
ic
a
ti
on.
F
ig
ur
e
11. T
he
I
nc
e
pt
io
nN
e
t
a
c
c
ur
a
c
y
a
nd l
os
s
pl
ot
F
ig
ur
e
12.
S
tr
id
-
C
N
N
a
c
c
ur
a
c
y a
nd l
os
s
pl
ot
T
a
bl
e
1. A
s
s
e
s
s
m
e
nt
of
l
e
a
r
ni
ng t
r
a
ns
f
e
r
s
tr
a
te
gi
e
s
M
ode
l
P
r
e
c
i
s
i
on
(%)
R
e
c
a
l
l
(%)
F1
-
s
c
or
e
(%)
A
c
c
ur
a
c
y
(%)
N
o. of
pa
r
a
m
e
t
e
r
s
N
o. of
ope
r
a
t
i
ons
D
e
ns
ne
t
121
90
86
87
88
7,037,504
2.78 bi
l
l
i
on F
L
O
P
s
VGG
-
16
90
86
88
88
14,714,688
15 bi
l
l
i
on F
L
O
P
s
R
e
s
N
e
t
-
50
53
46
43
64
23,587,712
3.86 bi
l
l
i
on F
L
O
P
s
I
nc
e
pt
i
onN
e
t
83
79
81
79
22,928,673
6.7 m
i
l
l
i
on F
L
O
P
s
S
t
r
i
d
-
C
N
N
94
94
94
93
55,51,812
2.8 m
i
l
l
i
on F
L
O
P
s
5.
C
O
N
C
L
U
S
I
O
N
I
n
c
onc
lu
s
io
n,
th
is
r
e
s
e
a
r
c
h
de
m
ons
tr
a
te
s
th
a
t
u
s
in
g
X
-
r
a
y
im
a
ge
s
of
th
e
lu
ngs
,
w
e
c
a
n
e
f
f
e
c
ti
ve
ly
di
s
ti
ngui
s
h
be
twe
e
n
h
e
a
lt
hy
a
nd
pne
um
oni
a
-
a
f
f
e
c
te
d
c
ondi
ti
ons
us
in
g
tr
a
ns
f
e
r
le
a
r
ni
ng.
T
h
e
pr
e
-
tr
a
in
e
d
m
ode
ls
, i
nc
lu
di
ng D
e
ns
e
N
e
t1
21, VG
G
-
16, R
e
s
N
e
t
-
50, I
nc
e
pt
io
n N
e
t,
a
nd S
tr
id
-
C
N
N
, a
c
hi
e
ve
d hi
gh l
e
ve
ls
of
a
c
c
ur
a
c
y,
pr
e
c
is
io
n,
r
e
c
a
ll
,
a
nd
F
1
-
s
c
or
e
in
c
la
s
s
if
yi
ng
pne
um
o
ni
a
,
unde
r
s
c
or
in
g
th
e
e
f
f
e
c
ti
ve
ne
s
s
of
tr
a
ns
f
e
r
le
a
r
ni
ng
in
th
is
c
ont
e
xt
.
N
ot
a
bl
y,
S
tr
id
-
C
N
N
de
m
ons
tr
a
te
d
a
n
a
c
c
ur
a
c
y
of
93%
in
di
s
ti
ngui
s
hi
ng
be
twe
e
n
th
e
two c
la
s
s
e
s
. A
ddi
ti
ona
ll
y, i
nc
or
por
a
ti
ng dive
r
s
e
t
r
a
in
in
g da
ta
a
nd us
in
g da
ta
a
ugm
e
nt
a
ti
on t
e
c
hni
que
s
s
uc
h a
s
r
ot
a
ti
on a
nd hor
iz
ont
a
l
f
li
ppi
ng f
u
r
th
e
r
e
nha
nc
e
d t
he
m
ode
ls
'
pe
r
f
or
m
a
nc
e
. T
he
s
e
f
in
di
ngs
pr
ovi
de
c
onc
lu
s
iv
e
e
vi
de
nc
e
s
uppor
ti
ng
th
e
pot
e
nt
ia
l
of
tr
a
ns
f
e
r
le
a
r
ni
ng
in
a
c
c
ur
a
te
ly
c
a
te
gor
iz
in
g
pne
um
oni
a
,
w
hi
c
h
c
oul
d
a
s
s
is
t
doc
to
r
s
in
m
a
ki
ng
c
or
r
e
c
t
di
a
gnos
e
s
.
T
o
im
pr
ove
th
e
e
f
f
ic
ie
nc
y
of
c
la
s
s
if
ic
a
ti
on
m
ode
l
s
a
nd
br
oa
de
n
th
e
s
c
ope
of
f
ut
ur
e
in
ve
s
ti
ga
ti
ons
,
it
is
r
e
c
om
m
e
nde
d
to
e
xpl
or
e
a
ddi
ti
ona
l
pr
e
-
t
r
a
in
e
d
m
ode
ls
a
nd
e
m
pl
oy
va
r
ie
d da
ta
a
ugm
e
nt
a
ti
on s
tr
a
te
gi
e
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
,
V
ol
.
14
, N
o.
4
,
A
ugus
t
20
25
:
3253
-
3261
3260
F
U
N
D
I
N
G
I
N
F
O
R
M
A
T
I
O
N
A
ut
hor
s
s
ta
te
no f
undi
ng i
nvol
ve
d.
A
U
T
H
O
R
C
O
N
T
R
I
B
U
T
I
O
N
S
S
T
A
T
E
M
E
N
T
T
hi
s
jo
ur
na
l
us
e
s
th
e
C
ont
r
ib
ut
or
R
ol
e
s
T
a
xonomy
(
C
R
e
di
T
)
to
r
e
c
ogni
z
e
in
di
vi
dua
l
a
ut
hor
c
ont
r
ib
ut
io
ns
, r
e
duc
e
a
ut
hor
s
hi
p di
s
put
e
s
,
a
nd f
a
c
il
it
a
te
c
ol
la
bo
r
a
ti
on.
N
am
e
o
f
A
u
t
h
or
C
M
So
Va
Fo
I
R
D
O
E
Vi
Su
P
Fu
K
hus
hboo T
r
iv
e
di
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
C
hi
nt
a
n
B
hupe
s
hbh
a
i
T
ha
c
ke
r
✓
✓
✓
✓
✓
✓
✓
C
:
C
onc
e
pt
ua
l
i
z
a
t
i
on
M
:
M
e
t
hodol
ogy
So
:
So
f
t
w
a
r
e
Va
:
Va
l
i
da
t
i
on
Fo
:
Fo
r
m
a
l
a
na
l
ys
i
s
I
:
I
nve
s
t
i
ga
t
i
on
R
:
R
e
s
our
c
e
s
D
:
D
a
t
a
C
ur
a
t
i
on
O
:
W
r
i
t
i
ng
-
O
r
i
gi
na
l
D
r
a
f
t
E
:
W
r
i
t
i
ng
-
R
e
vi
e
w
&
E
di
t
i
ng
Vi
:
Vi
s
ua
l
i
z
a
t
i
on
Su
:
Su
pe
r
vi
s
i
on
P
:
P
r
oj
e
c
t
a
dm
i
ni
s
t
r
a
t
i
on
Fu
:
Fu
ndi
ng a
c
qui
s
i
t
i
on
C
O
N
F
L
I
C
T
O
F
I
N
T
E
R
E
S
T
S
T
A
T
E
M
E
N
T
A
ut
hor
s
s
ta
te
no c
onf
li
c
t
of
i
nt
e
r
e
s
t.
D
A
T
A
A
V
A
I
L
A
B
I
L
I
T
Y
D
a
ta
a
va
il
a
bi
li
ty
is
not
a
ppl
ic
a
bl
e
to
th
is
pa
pe
r
a
s
no
ne
w
da
ta
w
e
r
e
c
r
e
a
te
d
or
a
na
ly
z
e
d
in
th
is
s
tu
dy.
R
E
F
E
R
E
N
C
E
S
[
1]
C
.
J
.
G
hi
a
a
nd
G
.
S
.
R
a
m
bha
d,
“
S
ys
t
e
m
a
t
i
c
r
e
vi
e
w
a
nd
m
e
t
a
-
a
na
l
ys
i
s
of
c
om
or
bi
di
t
i
e
s
a
nd
a
s
s
oc
i
a
t
e
d
r
i
s
k
f
a
c
t
or
s
i
n
I
ndi
a
n
pa
t
i
e
nt
s
of
c
om
m
uni
t
y
-
a
c
qui
r
e
d pne
um
oni
a
,”
SA
G
E
O
pe
n M
e
di
c
i
ne
, vol
. 10, 2022, doi
:
10.1177/
20503121221095485.
[
2]
V
.
R
a
j
a
gur
u,
T
.
H
.
K
i
m
,
J
.
S
h
i
n,
S
.
G
.
L
e
e
,
a
nd
W
.
H
a
n,
“
A
bi
l
i
t
y
of
t
he
L
A
C
E
i
nde
x
t
o
pr
e
di
c
t
30
-
da
y
r
e
a
dm
i
s
s
i
ons
i
n
pa
t
i
e
nt
s
w
i
t
h a
c
ut
e
m
yoc
a
r
di
a
l
i
nf
a
r
c
t
i
on,”
J
our
nal
of
P
e
r
s
onal
i
z
e
d M
e
di
c
i
n
e
, vol
. 12, no. 7, 2022, doi
:
10.3390/
j
pm
12071085.
[
3]
M
.
O
.
L
e
w
i
s
,
P
.
T
.
T
r
a
n,
Y
.
H
ua
ng,
R
.
A
.
D
e
s
a
i
,
Y
.
S
he
n,
a
nd
J
.
D
.
B
r
ow
n,
“
D
i
s
e
a
s
e
s
e
v
e
r
i
t
y
a
nd
r
i
s
k
f
a
c
t
or
s
of
30
-
da
y
ho
s
pi
t
a
l
r
e
a
dm
i
s
s
i
on
i
n
pe
di
a
t
r
i
c
hos
pi
t
a
l
i
z
a
t
i
ons
f
or
pne
um
oni
a
,”
J
ou
r
nal
of
C
l
i
ni
c
al
M
e
di
c
i
ne
,
vol
.
11,
no.
5,
2022
,
doi
:
10.3390/
j
c
m
11051185.
[
4]
A
.
K
.
M
onda
l
,
A
.
B
ha
t
t
a
c
ha
r
j
e
e
,
P
.
S
i
ngl
a
,
a
nd
A
.
P
.
P
r
a
t
hos
h,
“
xV
i
T
C
O
S
:
e
xpl
a
i
na
bl
e
vi
s
i
on
t
r
a
ns
f
or
m
e
r
ba
s
e
d
c
ovi
d
-
19
s
c
r
e
e
ni
ng
us
i
ng
r
a
di
ogr
a
phy,”
I
E
E
E
J
our
nal
of
T
r
ans
l
at
i
onal
E
ngi
ne
e
r
i
ng
i
n
H
e
al
t
h
and
M
e
di
c
i
ne
,
vo
l
.
10,
pp.
1
–
10,
2022,
doi
:
10.1109/
J
T
E
H
M
.2021.3134096.
[
5]
C
.
I
e
r
a
c
i
t
a
no
e
t
al
.
,
“
A
f
uz
z
y
-
e
nha
nc
e
d
de
e
p
l
e
a
r
ni
ng
a
ppr
oa
c
h f
or
e
a
r
l
y
de
t
e
c
t
i
on
of
C
ovi
d
-
19
pne
um
oni
a
f
r
om
po
r
t
a
bl
e
c
he
s
t
X
-
r
a
y i
m
a
ge
s
,”
N
e
u
r
oc
om
put
i
ng
, vol
. 481, pp. 202
–
215, 2022, doi
:
10.1016/
j
.ne
uc
om
.2022.01.055.
[
6]
H
.
R
e
n
e
t
al
.
,
“
I
nt
e
r
pr
e
t
a
bl
e
pne
um
oni
a
de
t
e
c
t
i
on
by
c
om
bi
ni
ng
de
e
p
l
e
a
r
n
i
n
g
a
nd
e
xpl
a
i
na
bl
e
m
ode
l
s
w
i
t
h
m
ul
t
i
s
our
c
e
da
t
a
,”
I
E
E
E
A
c
c
e
s
s
, vol
. 9, pp. 95872
–
95883, 2021, doi
:
10.1109/
A
C
C
E
S
S
.2021.3090215.
[
7]
L
.
K
ong
a
nd
J
.
C
he
ng,
“
B
a
s
e
d
on
i
m
pr
ove
d
de
e
p
c
onvol
ut
i
ona
l
ne
ur
a
l
ne
t
w
or
k
m
ode
l
pne
um
oni
a
i
m
a
ge
c
l
a
s
s
i
f
i
c
a
t
i
on,”
P
L
oS
O
N
E
, vol
. 16, no. 11, 2021, doi
:
10.1371/
j
our
na
l
.pone
.0258804.
[
8]
M
.
R
os
t
a
m
i
a
nd
M
.
O
us
s
a
l
a
h,
“
A
nove
l
e
xpl
a
i
na
bl
e
c
ovi
d
-
19
di
a
gnos
i
s
m
e
t
hod
by
i
nt
e
gr
a
t
i
on
of
f
e
a
t
ur
e
s
e
l
e
c
t
i
on
w
i
t
h
r
a
ndom
f
or
e
s
t
,”
I
nf
or
m
at
i
c
s
i
n M
e
di
c
i
ne
U
nl
oc
k
e
d
, vol
. 30, 2022, doi
:
10.1016/
j
.i
m
u.2022.100941.
[
9]
M
.
T
oğa
ç
a
r
,
N
.
M
uz
oğl
u,
B
.
E
r
ge
n,
B
.
S
.
B
.
Y
a
r
m
a
n,
a
nd
A
.
M
.
H
a
l
e
f
oğ
l
u,
“
D
e
t
e
c
t
i
on
of
c
ovi
d
-
19
f
i
ndi
ngs
by
t
he
l
oc
a
l
i
nt
e
r
pr
e
t
a
bl
e
m
ode
l
-
a
gnos
t
i
c
e
xpl
a
na
t
i
on
m
e
t
hod
of
t
ype
ba
s
e
d
a
c
t
i
va
t
i
ons
e
xt
r
a
c
t
e
d
f
r
om
C
N
N
s
,”
B
i
om
e
di
c
al
Si
gna
l
P
r
oc
e
s
s
i
n
g
and C
ont
r
ol
, vol
. 71, 2022, doi
:
10.1016/
j
.bs
pc
.2021.103128.
[
10]
D
.
M
a
ha
pa
t
r
a
,
Z
.
G
e
,
a
nd
M
.
R
e
ye
s
,
“
S
e
l
f
-
s
upe
r
vi
s
e
d
ge
ne
r
a
l
i
z
e
d
z
e
r
o
s
hot
l
e
a
r
ni
ng
f
or
m
e
di
c
a
l
i
m
a
ge
c
l
a
s
s
i
f
i
c
a
t
i
on
us
i
ng
nov
e
l
i
nt
e
r
pr
e
t
a
bl
e
s
a
l
i
e
nc
y
m
a
p
s
,”
I
E
E
E
T
r
ans
ac
t
i
ons
on
M
e
di
c
al
I
m
agi
ng
,
vol
.
41,
no.
9,
pp.
2443
–
2456,
2022,
doi
:
10.1109/
T
M
I
.2022.3163232.
[
11]
M
.
L
i
e
t
al
.
,
“
E
xpl
a
i
na
bl
e
c
ovi
d
-
19
i
n
f
e
c
t
i
ons
i
de
nt
i
f
i
c
a
t
i
on
a
nd
de
l
i
ne
a
t
i
on
u
s
i
ng
c
a
l
i
br
a
t
e
d
ps
e
udo
l
a
be
l
s
,”
I
E
E
E
T
r
ans
ac
t
i
ons
on E
m
e
r
gi
ng T
opi
c
s
i
n C
om
put
at
i
onal
I
nt
e
l
l
i
ge
nc
e
, vol
. 7, no. 1, pp. 26
–
35, 2023, doi
:
10.1109/
T
E
T
C
I
.2022.3189054.
[
12]
D
.
K
e
r
m
a
ny,
K
.
Z
ha
ng,
M
.
G
ol
dba
um
,
a
nd
ot
he
r
s
,
“
L
a
be
l
e
d
opt
i
c
a
l
c
ohe
r
e
nc
e
t
om
ogr
a
phy
(
O
C
T
)
a
nd
c
he
s
t
x
-
r
a
y
i
m
a
ge
s
f
or
c
l
a
s
s
i
f
i
c
a
t
i
on,”
M
e
nde
l
e
y
D
at
a,
V
2,
2018, doi
:
10.17632/
r
s
c
bj
br
9s
j
.2.
[
13]
A
.
I
.
A
vi
l
e
s
-
R
i
ve
r
o,
P
.
S
e
l
l
a
r
s
,
C
.
B
.
S
c
hönl
i
e
b,
a
nd
N
.
P
a
pa
da
ki
s
,
“
G
r
a
phX
C
O
V
I
D
:
E
xpl
a
i
na
bl
e
de
e
p
gr
a
ph
di
f
f
us
i
on
ps
e
udo
-
l
a
be
l
l
i
ng f
or
i
de
nt
i
f
yi
ng c
ovi
d
-
19 on c
he
s
t
x
-
r
a
ys
,”
P
at
t
e
r
n R
e
c
ogni
t
i
on
, vol
. 122, 2022, doi
:
10.1016/
j
.pa
t
c
og.2021.108274.
[
14]
A
.
M
a
l
hot
r
a
e
t
al
.
,
“
M
ul
t
i
-
t
a
s
k
dr
i
ve
n
e
xpl
a
i
na
bl
e
di
a
gnos
i
s
of
C
O
V
I
D
-
19
us
i
ng
c
he
s
t
X
-
r
a
y
i
m
a
ge
s
,
”
P
at
t
e
r
n
R
e
c
ogni
t
i
on
,
vol
. 122, 2022, doi
:
10.1016/
j
.pa
t
c
og.2021.108243.
[
15]
G
.
L
i
a
ng
a
nd
L
.
Z
he
ng,
“
A
t
r
a
ns
f
e
r
l
e
a
r
ni
ng
m
e
t
hod
w
i
t
h
de
e
p
r
e
s
i
dua
l
ne
t
w
or
k
f
or
pe
di
a
t
r
i
c
pne
um
oni
a
di
a
gnos
i
s
,”
C
om
put
e
r
M
e
t
hods
and P
r
ogr
am
s
i
n B
i
om
e
di
c
i
ne
, vol
. 187, 2020, doi
:
10.1016/
j
.c
m
pb.20
19.06.023.
[
16]
W
H
O
,
“
W
H
O
D
i
r
e
c
t
or
-
G
e
ne
r
a
l
’
s
ope
ni
ng
r
e
m
a
r
ks
a
t
t
he
m
e
di
a
b
r
i
e
f
i
ng
on
c
ovi
d
-
19
-
11
M
a
r
c
h
2020,”
W
or
l
d
H
e
al
t
h
O
r
gani
z
at
i
on
,
2020.
[
O
nl
i
ne
]
.
A
va
i
l
a
bl
e
:
ht
t
ps
:
/
/
w
w
w
.w
ho.i
nt
/
di
r
e
c
t
or
-
ge
ne
r
a
l
/
s
pe
e
c
he
s
/
de
t
a
i
l
/
w
ho
-
di
r
e
c
t
or
-
ge
ne
r
a
l
-
s
-
ope
ni
ng
-
r
e
m
a
r
ks
-
at
-
t
he
-
m
e
di
a
-
br
i
e
f
i
ng
-
on
-
c
ovi
d
-
19
-
-
-
11
-
m
a
r
c
h
-
2020
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
St
r
id
-
C
N
N
:
m
ov
in
g f
il
te
r
s
w
it
h c
onv
ol
ut
io
n ne
ur
al
ne
t
w
or
k
f
o
r
m
ul
ti
-
c
la
s
s
…
(
K
hus
hboo T
r
iv
e
di
)
3261
[
17]
M
.
M
.
H
e
l
l
ou
e
t
al
.
,
“
N
u
c
l
e
i
c
a
c
i
d
a
m
pl
i
f
i
c
a
t
i
on
t
e
s
t
s
on
r
e
s
pi
r
a
t
or
y
s
a
m
pl
e
s
f
or
t
he
di
a
gnos
i
s
of
c
or
ona
vi
r
us
i
nf
e
c
t
i
ons
:
a
s
ys
t
e
m
a
t
i
c
r
e
vi
e
w
a
nd
m
e
t
a
-
a
na
l
ys
i
s
,”
C
l
i
ni
c
al
M
i
c
r
obi
ol
ogy
and
I
nf
e
c
t
i
on
,
vol
.
27,
no.
3,
pp.
341
–
351,
M
a
r
.
2021,
doi
:
10.1016/
j
.c
m
i
.2020.11.002.
[
18]
F
.
C
ol
a
vi
t
a
e
t
al
.
,
“
C
ovi
d
-
19
r
a
pi
d
a
nt
i
ge
n
t
e
s
t
a
s
s
c
r
e
e
ni
ng
s
t
r
a
t
e
gy
a
t
poi
nt
s
of
e
nt
r
y:
E
xpe
r
i
e
nc
e
i
n
l
a
z
i
o
r
e
gi
on,
c
e
nt
r
a
l
i
t
a
l
y,
a
ugus
t
–
oc
t
obe
r
2020,”
B
i
om
ol
e
c
ul
e
s
, vol
. 11, no. 3, pp. 1
–
8, 2021, doi
:
10.339
0/
bi
om
11030425.
[
19]
J
. C
. H
s
u, F
. H
. W
u,
H
. H
.
L
i
n, D
. J
. L
e
e
, Y
. F
.
C
he
n, a
nd
C
. S
. L
i
n,
“
A
I
m
ode
l
s
f
or
pr
e
di
c
t
i
ng r
e
a
dm
i
s
s
i
on of
pne
um
oni
a
pa
t
i
e
nt
s
w
i
t
hi
n 30 da
ys
a
f
t
e
r
di
s
c
ha
r
ge
,”
E
l
e
c
t
r
oni
c
s
, vol
. 11, no. 5, 2022, doi
:
10.3390/
e
l
e
c
t
r
oni
c
s
11050673.
[
20]
A
.
F
ur
t
a
do,
L
.
A
ndr
a
de
,
D
.
F
r
i
a
s
,
T
.
M
a
i
a
,
R
.
B
a
da
r
ó,
a
nd
E
.
G
.
S
pe
r
a
n
di
o
N
a
s
c
i
m
e
nt
o,
“
D
e
e
p
l
e
a
r
ni
ng
a
ppl
i
e
d
t
o
c
he
s
t
r
a
di
ogr
a
ph
c
l
a
s
s
i
f
i
c
a
t
i
on
—
a
c
ovi
d
-
19
pne
um
oni
a
e
xpe
r
i
e
nc
e
,”
A
ppl
i
e
d
Sc
i
e
nc
e
s
,
vol
.
12,
no.
8,
2022,
doi
:
10.3390/
a
pp12083712.
[
21]
Z
. S
a
l
a
huddi
n, H
. C
. W
oodr
uf
f
, A
. C
ha
t
t
e
r
j
e
e
, a
nd P
.
L
a
m
bi
n, “
T
r
a
ns
pa
r
e
nc
y o
f
de
e
p ne
ur
a
l
ne
t
w
or
ks
f
or
m
e
di
c
a
l
i
m
a
ge
a
n
a
l
ys
i
s
:
A
r
e
vi
e
w
of
i
nt
e
r
pr
e
t
a
bi
l
i
t
y
m
e
t
hods
,”
C
om
put
e
r
s
i
n
B
i
ol
ogy
and
M
e
di
c
i
ne
,
vol
.
140,
2022,
doi
:
10.1016/
j
.c
om
pbi
om
e
d.2021.105111.
[
22]
A
.
R
a
nj
a
n,
C
.
K
um
a
r
,
R
.
K
.
G
upt
a
,
a
nd
R
.
M
i
s
r
a
,
“
T
r
a
ns
f
e
r
l
e
a
r
ni
ng
ba
s
e
d
a
p
pr
oa
c
h
f
or
pne
um
oni
a
de
t
e
c
t
i
on
us
i
ng
c
us
t
om
i
z
e
d
V
G
G
16
de
e
p
l
e
a
r
ni
ng
m
ode
l
,”
i
n
I
nt
e
r
ne
t
of
T
hi
ngs
and
C
onne
c
t
e
d
T
e
c
hnol
ogi
e
s
,
2022,
pp.
17
–
28
,
doi
:
10.1007/
978
-
3
-
030
-
94507
-
7_2.
[
23]
M
.
M
.
R
a
hm
a
n,
S
.
N
oor
uddi
n,
K
.
M
.
A
.
H
a
s
a
n,
a
nd
N
.
K
.
D
e
y,
“
H
O
G
+
C
N
N
N
e
t
:
di
a
gnos
i
ng
c
ovi
d
-
19
a
nd
pne
um
oni
a
by
de
e
p
ne
ur
a
l
ne
t
w
or
k f
r
om
c
he
s
t
x
-
r
a
y i
m
a
ge
s
,”
SN
C
om
put
e
r
S
c
i
e
nc
e
, vol
. 2, no. 5, 2
021, doi
:
10.1007/
s
42979
-
021
-
00762
-
x.
[
24]
T
.
H
.
M
a
nde
e
l
,
S
.
M
.
A
w
a
d,
a
nd
S
.
N
a
j
i
,
“
P
ne
um
oni
a
b
i
na
r
y
c
l
a
s
s
i
f
i
c
a
t
i
on
us
i
ng
m
ul
t
i
-
s
c
a
l
e
f
e
a
t
ur
e
c
l
a
s
s
i
f
i
c
a
t
i
on
ne
t
w
or
k
on
c
he
s
t
x
-
r
a
y
i
m
a
ge
s
,”
I
A
E
S
I
nt
e
r
nat
i
onal
J
our
nal
of
A
r
t
i
f
i
c
i
al
I
nt
e
l
l
i
ge
nc
e
,
vol
.
11,
no.
4,
pp.
1469
–
1477,
2022,
doi
:
10.11591/
i
j
a
i
.v11.i
4.pp1469
-
1477.
[
25]
Y
.
B
r
i
m
a
,
M
.
A
t
e
m
ke
ng,
S
.
T
.
D
j
i
oka
p,
J
.
E
bi
e
l
e
,
a
nd
F
.
T
c
ha
kount
é
,
“
T
r
a
ns
f
e
r
l
e
a
r
ni
ng
f
or
t
he
de
t
e
c
t
i
on
a
nd
di
a
gnos
i
s
of
t
yp
e
s
of
pne
um
oni
a
i
nc
l
udi
ng
pne
um
oni
a
i
ndu
c
e
d
by
C
O
V
I
D
-
19
f
r
om
c
he
s
t
X
-
r
a
y
i
m
a
ge
s
,”
D
i
agnos
t
i
c
s
,
vol
.
11,
no.
8,
2021
,
doi
:
10.3390/
di
a
gnos
t
i
c
s
11081480.
B
I
O
G
R
A
P
H
I
E
S
O
F
A
U
T
H
O
R
S
Khushboo
Trivedi
holds
a
master’s
degree
in
computer
science
a
nd
engineering
from
Parul
University
and
is
currently
pursuing
a
Ph.D
.
in
the
doma
in
of
deep
learning
and
computer
vision
f
rom Parul
Universit
y, Vadodara, Guj
arat.
She has
12
+
years of
experience in
academia.
Her
research
interest
s
are
in
machine
learning,
deep
lear
ning,
AI
,
and
computer
visio
n.
Currently
,
s
he
serves
as
an
assistan
t
professor
in
Department
of
Computer
Science
and
Engineering
in
Parul
Institute
of
Technology,
Parul
University.
She
can
be
contacted
at
email:
khushboo.trivedi21305@
paruluniversity.ac.in.
Dr.
Chintan
Bhupeshbhai
Thacker
received
the
Ph.D.
degree
in
the
domain
of
AI
and
computer
vis
ion
from
Gujarat
Technical
University
in
the
year
2022.
He
had
served
as
head
of
the
Department
of
Computer
Science
and
Engineering
at
HJ
D
Institute
of
Technical
Education
and
Research,
Kera,
India.
He
has
1+
years
of
experience
i
n
industry
and
12+
years
of
experience
in
academia.
Currently,
he
serves
as
an
assistant
prof
essor
in
Department
of
Computer
Science
and
Enginee
ring
in
Parul
Institute
of
Enginee
ring
Technol
ogy,
Parul
University,
Vadodara,
Gujarat.
In
addition,
he
has
also
guided
severa
l
doctorate
students
and
has
been
active
in
conducting
several
workshops
in
the
domain
o
f
computer
vision
.
Hi
s
research
interests
are
in
machine
learning,
AI
,
deep
learning,
and
computer
vi
sion.
He
can
be
contacted
at email
: chint
an.thacker19
435@
paruluni
versity.
ac.in.
Evaluation Warning : The document was created with Spire.PDF for Python.