IEEE Access,
Год журнала:
2024,
Номер
12, С. 73561 - 73580
Опубликована: Янв. 1, 2024
The
agricultural
sector
is
still
a
major
provider
of
many
countries'
economies,
but
diseases
that
continuously
infect
plants
represent
continuous
threats
to
agriculture
and
cause
massive
losses
the
country's
economy.
In
this
study,
lightweight
convolutional
neural
network
model
called
FL-ToLeD
was
proposed
for
tomato
disease
classification
based
on
soft
attention
mechanism
with
depth-wise
separable
convolution
layer.
With
size
2.5
MB
221,594
trainable
parameters,
achieved
99.5%,
99.10%,
99.04%
training,
validation
testing
accuracy
respectively,
99
%
each
precision,
recall,
f1-score,
it
also
99.90%
ROC-AUC
average
inference
time
2.06924
μs.
outperformed
H.
Ulutaş
(2023)
by
2.2%
in
terms
accuracy,
recall
f1-score.
Additionally,
performed
better
than
M.
Agarwal
(2023),
Abbas
(2021),
S.
Verma
(2020)
f1-score
8%,
2%,
6%,
respectively.
It
Arshad
4.77%,
8.92%,
35.18%
5.11%
Furthermore,
90
times
smaller
size.
All
makes
more
suitable
low-end
devices
precision
agriculture.
J — Multidisciplinary Scientific Journal,
Год журнала:
2024,
Номер
7(1), С. 48 - 71
Опубликована: Янв. 22, 2024
Chest
X-ray
imaging
plays
a
vital
and
indispensable
role
in
the
diagnosis
of
lungs,
enabling
healthcare
professionals
to
swiftly
accurately
identify
lung
abnormalities.
Deep
learning
(DL)
approaches
have
attained
popularity
recent
years
shown
promising
results
automated
medical
image
analysis,
particularly
field
chest
radiology.
This
paper
presents
novel
DL
framework
specifically
designed
for
multi-class
diseases,
including
fibrosis,
opacity,
tuberculosis,
normal,
viral
pneumonia,
COVID-19
using
images,
aiming
address
need
efficient
accessible
diagnostic
tools.
The
employs
convolutional
neural
network
(CNN)
architecture
with
custom
blocks
enhance
feature
maps
learn
discriminative
features
from
images.
proposed
is
evaluated
on
large-scale
dataset,
demonstrating
superior
performance
lung.
In
order
evaluate
effectiveness
presented
approach,
thorough
experiments
are
conducted
against
pre-existing
state-of-the-art
methods,
revealing
significant
accuracy,
sensitivity,
specificity
improvements.
findings
study
showcased
remarkable
achieving
98.88%.
metrics
precision,
recall,
F1-score,
Area
Under
Curve
(AUC)
averaged
0.9870,
0.9904,
0.9887,
0.9939
across
six-class
categorization
system.
research
contributes
provides
foundation
future
advancements
DL-based
systems
diseases.
Frontiers in Plant Science,
Год журнала:
2024,
Номер
15
Опубликована: Май 17, 2024
Tomato
is
one
of
the
most
popular
and
important
food
crops
consumed
globally.
The
quality
quantity
yield
by
tomato
plants
are
affected
impact
made
various
kinds
diseases.
Therefore,
it
essential
to
identify
these
diseases
early
so
that
possible
reduce
occurrences
effect
on
improve
overall
crop
support
farmers.
In
past,
many
research
works
have
been
carried
out
applying
machine
learning
techniques
segment
classify
leaf
images.
However,
existing
learning-based
classifiers
not
able
detect
new
types
more
accurately.
On
other
hand,
deep
with
swarm
intelligence-based
optimization
enhance
classification
accuracy,
leading
effective
accurate
detection
This
paper
proposes
a
method
for
harnessing
power
an
ensemble
model
in
sample
dataset
plants,
containing
images
pertaining
nine
different
introduces
exponential
moving
average
function
temporal
constraints
enhanced
weighted
gradient
optimizer
integrated
into
fine-tuned
Visual
Geometry
Group-16
(VGG-16)
Neural
Architecture
Search
Network
(NASNet)
mobile
training
methods
providing
improved
accuracy.
used
consists
10,000
categorized
classes
validating
additional
1,000
reserved
testing
model.
results
analyzed
thoroughly
benchmarked
performance
metrics,
thus
proving
proposed
approach
gives
better
terms
loss,
precision,
recall,
receiver
operating
characteristic
curve,
F1-score
values
98.7%,
4%,
97.9%,
98.6%,
99.97%,
respectively.
Electronics,
Год журнала:
2024,
Номер
13(11), С. 2057 - 2057
Опубликована: Май 24, 2024
The
A-star
algorithm
(A*)
is
a
traditional
and
widely
used
approach
for
route
planning
in
various
domains,
including
robotics
automobiles
smart
agriculture.
However,
notable
limitation
of
the
its
tendency
to
generate
paths
that
lack
desired
smoothness.
In
response
this
challenge,
particularly
agricultural
operations,
research
endeavours
enhance
evaluation
individual
nodes
within
search
procedure
improve
overall
smoothness
resultant
path.
So,
mitigate
inherent
choppiness
A-star-generated
agriculture,
work
adopts
novel
approach.
It
introduces
utilizing
Bezier
curves
as
postprocessing
step,
thus
refining
generated
imparting
their
This
instrumental
real-world
applications
where
continuous
safe
motion
imperative.
outcomes
simulations
conducted
part
study
affirm
efficiency
proposed
methodology.
These
results
underscore
capability
enhanced
technique
construct
smooth
pathways.
Furthermore,
they
demonstrate
performance.
are
also
well
suited
deployment
rural
conditions,
navigating
complex
terrains
with
precision
critical
necessity.
Sustainability,
Год журнала:
2023,
Номер
15(15), С. 11681 - 11681
Опубликована: Июль 28, 2023
The
growing
global
population
and
accompanying
increase
in
food
demand
has
put
pressure
on
agriculture
to
produce
higher
yields
the
face
of
numerous
challenges,
including
plant
diseases.
Tomato
is
a
widely
cultivated
essential
crop
that
particularly
susceptible
disease,
resulting
significant
economic
losses
hindrances
security.
Recently,
Artificial
Intelligence
(AI)
emerged
as
promising
tool
for
detecting
classifying
tomato
leaf
diseases
with
exceptional
accuracy
efficiency,
empowering
farmers
take
proactive
measures
prevent
damage
production
loss.
AI
algorithms
are
capable
processing
vast
amounts
data
objectively
without
human
bias,
making
them
potent
even
subtle
variations
traditional
techniques
might
miss.
This
paper
provides
comprehensive
overview
most
recent
advancements
disease
classification
using
Machine
Learning
(ML)
Deep
(DL)
techniques,
an
emphasis
how
these
approaches
can
enhance
effectiveness
classification.
Several
ML
DL
models,
convolutional
neural
networks
(CNN),
evaluated
review
highlights
various
features
used
acquisition
well
evaluation
metrics
employed
assess
performance
models.
Moreover,
this
emphasizes
address
limitations
classification,
leading
improved
more
efficient
management
ultimately
contributing
concludes
by
outlining
research
proposing
new
directions
field
AI-assisted
These
insights
will
be
value
researchers
professionals
interested
utilizing
contribute
sustainable
(SDG-3).
Frontiers in Plant Science,
Год журнала:
2023,
Номер
14
Опубликована: Окт. 16, 2023
Tomato
leaf
disease
identification
is
difficult
owing
to
the
variety
of
diseases
and
complex
causes,
for
which
method
based
on
convolutional
neural
network
effective.
While
it
challenging
capture
key
features
or
tends
lose
a
large
number
when
extracting
image
by
applying
this
method,
resulting
in
low
accuracy
identification.
Therefore,
ResNet50-DPA
model
proposed
identify
tomato
paper.
Firstly,
an
improved
ResNet50
included
model,
replaces
first
layer
convolution
basic
with
cascaded
atrous
convolution,
facilitating
obtaining
different
scales.
Secondly,
dual-path
attention
(DPA)
mechanism
search
features,
where
stochastic
pooling
employed
eliminate
influence
non-maximum
values,
two
convolutions
one
dimension
are
introduced
replace
MLP
effectively
reducing
damage
information.
In
addition,
quickly
accurately
type
disease,
DPA
module
incorporated
into
residual
obtain
enhanced
feature
map,
helps
reduce
economic
losses.
Finally,
visualization
results
Grad-CAM
presented
show
that
can
more
improve
interpretability
meeting
need
precise
diseases.
In
the
face
of
a
burgeoning
global
population
exceeding
seven
billion
and
dwindling
agricultural
land,
plants
remain
pivotal
for
sustaining
human
civilization's
food
needs.
However,
plant
health
is
threatened
by
various
diseases,
particularly
leaf
ailments
like
spots,
bacterial
infections,
black
spots.
These
afflictions,
predominantly
caused
bacteria
fungi,
jeopardize
crop
yields.
Timely
disease
detection
imperative
safeguarding
productivity.
This
study
introduces
novel
hybrid
approach
amalgamating
MobileNet,
transfer
learning-based
model,
with
SVM
(Support
Vector
Machine)
hinge
loss.
Leveraging
MobileNet's
pre-trained
capabilities,
features
are
extracted
fed
into
an
classifier
to
discern
nine
distinct
types
tomato
diseases
healthy
leaves.
Statistical
analysis
underscores
efficacy
this
surpassing
previous
benchmarks.
Notably,
it
achieves
exceptional
classification
accuracy,
precision,
recall,
AUC
values,
culminating
in
impressive
overall
accuracy
99.37%.
Journal of Phytopathology,
Год журнала:
2025,
Номер
173(1)
Опубликована: Янв. 1, 2025
ABSTRACT
Precise
detection
of
crop
disease
at
the
early
stage
is
a
crucial
task,
which
will
reduce
spreading
by
taking
preventive
measures.
The
main
goal
this
research
to
propose
hybrid
classification
system
for
detecting
utilising
Modified
Deep
Joint
(MDJ)
segmentation.
diseases
involves
five
stages.
They
are
data
acquisition,
pre‐processing,
segmentation,
feature
extraction
and
detection.
In
initial
stage,
image
diverse
crops
gathered
in
acquisition
phase.
According
work,
we
considering
Apple
corn
with
benchmark
datasets.
input
subjected
pre‐processing
median
filtering
process.
Subsequently,
pre‐processed
under
goes
segmentation
process,
where
proposed
work.
From
segmented
image,
features
like
shape,
colour,
texture‐based
Improved
Median
Binary
Pattern
(IMBP)‐based
extracted.
Finally,
extracted
given
identifying
diseases.
model
includes
Bidirectional
Long
Short‐Term
Memory
(Bi‐LSTM)
Belief
Network
(DBN)
classifiers.
outcome
both
classifiers
score,
an
improved
score
level
fusion
model,
determines
final
results.
performance
evaluated
over
existing
methods
various
metrics.
At
training
90%,
scheme
attained
accuracy
0.965,
while
conventional
achieved
less
rates.
Frontiers in Plant Science,
Год журнала:
2025,
Номер
15
Опубликована: Янв. 21, 2025
Tomatoes
are
considered
one
of
the
most
valuable
vegetables
around
world
due
to
their
usage
and
minimal
harvesting
period.
However,
effective
still
remains
a
major
issue
because
tomatoes
easily
susceptible
weather
conditions
other
types
attacks.
Thus,
numerous
research
studies
have
been
introduced
based
on
deep
learning
models
for
efficient
classification
tomato
leaf
disease.
single
architecture
does
not
provide
best
results
limited
computational
ability
complexity.
this
used
Transductive
Long
Short-Term
Memory
(T-LSTM)
with
an
attention
mechanism.
The
mechanism
in
T-LSTM
has
focus
various
parts
image
sequence.
exploits
specific
characteristics
training
instances
make
accurate
predictions.
This
can
involve
leveraging
relationships
patterns
observed
within
dataset.
is
transductive
approach
scaled
dot
product
evaluates
weights
each
step
hidden
state
patches
which
helps
classification.
data
was
gathered
from
PlantVillage
dataset
pre-processing
conducted
resizing,
color
enhancement,
augmentation.
These
outputs
were
then
processed
segmentation
stage
where
U-Net
applied.
After
segmentation,
VGG-16
feature
extraction
done
through
proposed
experimental
outcome
shows
that
classifier
achieved
accuracy
99.98%
comparably
better
than
existing
convolutional
neural
network
transfer
IBSA-NET.
Algorithms,
Год журнала:
2025,
Номер
18(2), С. 84 - 84
Опубликована: Фев. 5, 2025
Many
sciences
exploit
algorithms
in
a
large
variety
of
applications.
In
agronomy,
amounts
agricultural
data
are
handled
by
adopting
procedures
for
optimization,
clustering,
or
automatic
learning.
this
particular
field,
the
number
scientific
papers
has
significantly
increased
recent
years,
triggered
scientists
using
artificial
intelligence,
comprising
deep
learning
and
machine
methods
bots,
to
process
crop,
plant,
leaf
images.
Moreover,
many
other
examples
can
be
found,
with
different
applied
plant
diseases
phenology.
This
paper
reviews
publications
which
have
appeared
past
three
analyzing
used
classifying
agronomic
aims
crops
applied.
Starting
from
broad
selection
6060
papers,
we
subsequently
refined
search,
reducing
358
research
articles
30
comprehensive
reviews.
By
summarizing
advantages
applying
analyses,
propose
guide
farming
practitioners,
agronomists,
researchers,
policymakers
regarding
best
practices,
challenges,
visions
counteract
effects
climate
change,
promoting
transition
towards
more
sustainable,
productive,
cost-effective
encouraging
introduction
smart
technologies.