A pose-conditioned cVAE plus automated interaction data augmentation and a capsule-based collision-fix module synthesizes diverse, contact-preserving, non-penetrating 3D human-human poses better than prior generative baselines.
Title resolution pending
1 Pith paper cite this work, alongside 86 external citations. Polarity classification is still indexing.
1
Pith paper citing it
86
external citations · OpenAlex
fields
cs.CV 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
GNOCHI: Generative Neural mOdel for Close Human-Human Interactions
A pose-conditioned cVAE plus automated interaction data augmentation and a capsule-based collision-fix module synthesizes diverse, contact-preserving, non-penetrating 3D human-human poses better than prior generative baselines.