A lightweight skeleton-based model that learns from velocity and acceleration cues via distillation outperforms larger pose-based baselines on a new synthetic-to-real classroom incident benchmark.
Exploring au- tomated recognition of instructional activity and discourse from multimodal classroom data
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.CV 1years
2026 1verdicts
CONDITIONAL 1roles
background 1polarities
background 1representative citing papers
citing papers explorer
-
Robust and Efficient Motion Reasoning for Privacy-Aware Classroom Incident Recognition
A lightweight skeleton-based model that learns from velocity and acceleration cues via distillation outperforms larger pose-based baselines on a new synthetic-to-real classroom incident benchmark.