.png){kind=icon}
Return
Skeleton-based lightweight action recognition framework in complex scenes
.png){kind=icon}
PRE
AI
DOI：10.1016/j.image.2026.117714.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
• A lightweight skeleton-based framework targets complex-scene action recognition.
• Dynamic adaptive graph convolution learns topology from joint and bone motion cues.
• KP-GCN models ordered latent dynamics with class-conditional transition operators.
• A training-only FR-Module refines hidden features to separate similar actions.
• The framework achieves competitive accuracy on six benchmarks with 2.78M parameters.
Keywords:
Action recognition
Complex scenes
Lightweight
Skeleton-based
Journal
S
IF:
2.7
Papers:
127
Citations:
0
Organization
Cited Papers
No cited papers available