Sensors,
Год журнала:
2024,
Номер
24(22), С. 7324 - 7324
Опубликована: Ноя. 16, 2024
We
describe
a
system
for
identifying
dog
emotions
based
on
dogs'
facial
expressions
and
body
posture.
Towards
that
goal,
we
built
dataset
with
2184
images
of
ten
popular
breeds,
grouped
into
seven
similarly
sized
primal
mammalian
emotion
categories
defined
by
neuroscientist
psychobiologist
Jaak
Panksepp
as
'Exploring',
'Sadness',
'Playing',
'Rage',
'Fear',
'Affectionate'
'Lust'.
modified
the
contrastive
learning
framework
MoCo
(Momentum
Contrast
Unsupervised
Visual
Representation
Learning)
to
train
it
our
original
achieved
an
accuracy
43.2%
baseline
14%.
also
trained
this
model
second
publicly
available
resulted
in
48.46%
but
had
25%.
compared
unsupervised
approach
supervised
ResNet50
architecture.
This
model,
when
tested
labels,
74.32.
Dog
owners
are
typically
capable
of
recognizing
behavioral
cues
that
reveal
subjective
states
their
dogs,
such
as
pain.
But
automatic
recognition
the
pain
state
is
very
challenging.
This
paper
proposes
a
novel
video-based,
two-stream
deep
neural
network
approach
for
this
problem.
We
extract
and
preprocess
body
keypoints,
compute
features
from
both
keypoints
RGB
representation
over
video.
propose
an
to
deal
with
self-occlusions
missing
keypoints.
also
present
unique
video-based
dog
behavior
dataset,
collected
by
veterinary
professionals,
annotated
presence
pain,
report
good
classification
results
proposed
approach.
study
one
first
works
on
machine
learning
based
estimation
state.
Code
available
at
https://github.con/s04240051/pain_detection
International Journal of Advanced Research in Science Communication and Technology,
Год журнала:
2024,
Номер
unknown, С. 343 - 346
Опубликована: Ноя. 8, 2024
Veterinary
medicine
is
a
broad
and
developing
profession
that
covers
topics
such
as
companion
animal
health,
zoonotic
infections,
agriculture,
community
health.
The
potential
for
better
healthcare
diagnostics
has
sparked
growing
interest
in
the
application
of
computer
vision
(CV)
veterinary
science
discipline
recent
years.
This
research
investigates
extent
applications
CV
techniques,
with
focus
on
deep
learning
approaches,
medical
imaging,
thermal
video
analysis,
alignment
diagnostics,
post-surgery
pet
monitoring
clinical
settings.
Salient
Object
Deduction
(SOD),
R-CNN,
Convolutional
Attentive
Adversarial
Network
(CAAN)
are
examined
this
study
to
demonstrate
important
roles
plays
addressing
issues
enhancing
overall
Scientific Reports,
Год журнала:
2024,
Номер
14(1)
Опубликована: Ноя. 14, 2024
Affective
states
are
reflected
in
the
facial
expressions
of
all
mammals.
Facial
behaviors
linked
to
pain
have
attracted
most
attention
so
far
non-human
animals,
leading
development
numerous
instruments
for
evaluating
through
various
animal
species.
Nevertheless,
manual
expression
analysis
is
susceptible
subjectivity
and
bias,
labor-intensive
often
necessitates
specialized
expertise
training.
This
challenge
has
spurred
a
growing
body
research
into
automated
recognition,
which
been
explored
multiple
species,
including
cats.
In
our
previous
studies,
we
presented
studied
artificial
intelligence
(AI)
pipelines
recognition
cats
using
48
landmarks
grounded
cats'
musculature,
as
well
an
detector
these
landmarks.
However,
used
solely
static
information
obtained
from
hand-picked
single
images
good
quality.
study
takes
significant
step
forward
fully
detection
applications
by
presenting
end-to-end
AI
pipeline
that
requires
no
efforts
selection
suitable
or
their
landmark
annotation.
By
working
with
video
rather
than
still
images,
this
new
approach
also
optimises
temporal
dimension
visual
capture
way
not
practical
preform
manually.
The
reaches
over
70%
66%
accuracy
respectively
two
different
cat
datasets,
outperforming
landmark-based
approaches
frames
under
similar
conditions,
indicating
dynamics
matter
recognition.
We
further
define
metrics
measuring
dimensions
deficiencies
datasets
faces,
investigate
impact
on
performance
pipeline.
Sensors,
Год журнала:
2024,
Номер
24(22), С. 7324 - 7324
Опубликована: Ноя. 16, 2024
We
describe
a
system
for
identifying
dog
emotions
based
on
dogs'
facial
expressions
and
body
posture.
Towards
that
goal,
we
built
dataset
with
2184
images
of
ten
popular
breeds,
grouped
into
seven
similarly
sized
primal
mammalian
emotion
categories
defined
by
neuroscientist
psychobiologist
Jaak
Panksepp
as
'Exploring',
'Sadness',
'Playing',
'Rage',
'Fear',
'Affectionate'
'Lust'.
modified
the
contrastive
learning
framework
MoCo
(Momentum
Contrast
Unsupervised
Visual
Representation
Learning)
to
train
it
our
original
achieved
an
accuracy
43.2%
baseline
14%.
also
trained
this
model
second
publicly
available
resulted
in
48.46%
but
had
25%.
compared
unsupervised
approach
supervised
ResNet50
architecture.
This
model,
when
tested
labels,
74.32.
