Few-shot multi-scale railway obstacle detection via lightweight linear transformer and precise feature reweighting
Measurement,
Год журнала:
2025,
Номер
unknown, С. 117584 - 117584
Опубликована: Апрель 1, 2025
Язык: Английский
GCapNet-FSD: A heterogeneous graph capsule network for few-shot object detection
Neural Networks,
Год журнала:
2025,
Номер
189, С. 107570 - 107570
Опубликована: Май 16, 2025
Язык: Английский
Decoupled DETR for Few-Shot Object Detection
Lecture notes in computer science,
Год журнала:
2024,
Номер
unknown, С. 158 - 174
Опубликована: Дек. 6, 2024
Язык: Английский
Multi-level similarity transfer and adaptive fusion data augmentation for few-shot object detection
Journal of Visual Communication and Image Representation,
Год журнала:
2024,
Номер
105, С. 104340 - 104340
Опубликована: Ноя. 12, 2024
Язык: Английский
Enhanced enchondroma detection from x‐ray images using deep learning: A step towards accurate and cost‐effective diagnosis
Journal of Orthopaedic Research®,
Год журнала:
2024,
Номер
42(12), С. 2826 - 2834
Опубликована: Июль 15, 2024
This
study
investigates
the
automated
detection
of
enchondromas,
benign
cartilage
tumors,
from
x-ray
images
using
deep
learning
techniques.
Enchondromas
pose
diagnostic
challenges
due
to
their
potential
for
malignant
transformation
and
overlapping
radiographic
features
with
other
conditions.
Leveraging
a
data
set
comprising
1645
1173
patients,
deep-learning
model
implemented
Detectron2
achieved
an
accuracy
0.9899
in
detecting
enchondromas.
The
employed
rigorous
validation
processes
compared
its
findings
existing
literature,
highlighting
superior
performance
approach.
Results
indicate
machine
improving
reducing
healthcare
costs
associated
advanced
imaging
modalities.
underscores
significance
early
accurate
enchondromas
effective
patient
management
suggests
avenues
further
research
musculoskeletal
tumor
detection.
Язык: Английский
Optimized Design of Instrument Recognition Based on CNN Model
Applied Mathematics and Nonlinear Sciences,
Год журнала:
2024,
Номер
9(1)
Опубликована: Янв. 1, 2024
Abstract
Intelligent
recognition
of
instrument
features
plays
an
important
role
in
automation
management
and
overhaul
also
facilitates
the
realization
accurate
reading
key
parameters
complex
environments.
The
dial
intelligent
system
proposed
this
paper
consists
geometry
correction,
pointer
segmentation,
modules.
Combining
idea
GhostNet
model
to
improve
structure
backbone
network
Mask
RCNN
model,
attention
mechanism
is
introduced
into
U-Net
minimum
outer
rectangle
method
used
for
recognition.
Under
different
viewpoint
rotation
angles,
errors
paper’s
are
relatively
stable,
they
less
than
1%.
region
segmentation
precision,
recall,
accuracy
99.39%,
99.05%,
98.38%,
respectively.
average
error
results
only
-0.04°C,
which
satisfactory
Язык: Английский
Robust internal representations for domain generalization
AI Magazine,
Год журнала:
2023,
Номер
44(4), С. 467 - 481
Опубликована: Окт. 26, 2023
Abstract
This
paper,
which
is
part
of
the
New
Faculty
Highlights
Invited
Speaker
Program
AAAI'23,
serves
as
a
comprehensive
survey
my
research
in
transfer
learning
by
utilizing
embedding
spaces.
The
work
reviewed
this
paper
specifically
revolves
around
inherent
challenges
associated
with
continual
and
limited
availability
labeled
data.
By
providing
an
overview
past
ongoing
contributions,
aims
to
present
holistic
understanding
research,
paving
way
for
future
explorations
advancements
field.
My
delves
into
various
settings
learning,
including,
few‐shot
zero‐shot
domain
adaptation,
distributed
learning.
I
hope
provides
forward‐looking
perspective
researchers
who
would
like
focus
on
similar
directions.
Язык: Английский