USAD combines diffusion-based data augmentation, multi-branch spatiotemporal attention, and adaptive loss weighting, reporting 98.84% on WISDM, 94.07% on PAMAP2, and 84.60% on OPPORTUNITY, though the abstract lists different values.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2025 1verdicts
REJECT 1representative citing papers
citing papers explorer
-
USAD: End-to-End Human Activity Recognition via Diffusion Model with Spatiotemporal Attention
USAD combines diffusion-based data augmentation, multi-branch spatiotemporal attention, and adaptive loss weighting, reporting 98.84% on WISDM, 94.07% on PAMAP2, and 84.60% on OPPORTUNITY, though the abstract lists different values.