PeO-HOI combines word-embedding prototype features, a propensity-reweighted cross-entropy loss, and a spatio-temporal transformer to improve human-object interaction detection in livestreaming, reporting 1-3 mAP point gains over the Ni et al. baseline.
Video-based human- object interaction detection from tubelet tokens ,
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
2025 1verdicts
CONDITIONAL 1roles
background 1polarities
unclear 1representative citing papers
citing papers explorer
-
Prototype Embedding Optimization for Human-Object Interaction Detection in Livestreaming
PeO-HOI combines word-embedding prototype features, a propensity-reweighted cross-entropy loss, and a spatio-temporal transformer to improve human-object interaction detection in livestreaming, reporting 1-3 mAP point gains over the Ni et al. baseline.