MAMHOI: Factorizing Scene-Aware Human-Object Interaction through Affordances
MAMHOI factorizes scene-aware human-object interaction through affordances, reducing penetration without paired scene data.
MAMHOI factorizes scene-aware human-object interaction generation through an explicit motion-affordance interface. A scene-conditioned model predicts where and how an interaction can be executed, and an affordance-conditioned model then synthesizes the human-object motion. That split lets scene understanding and interaction dynamics be learned from complementary datasets without paired human-object-scene data. Experiments in complex indoor environments show less object-scene penetration while interaction quality is better preserved.
- A scene model predicts feasible interaction; a motion model synthesizes movement.
- Factorization avoids scarce paired human-object-scene training data.
- Indoor tests reduce object-scene penetration while preserving interaction quality.
From the original article
1 min read at arxiv.orgGenerating realistic human-object interactions (HOI) in complex 3D scenes requires two complementary capabilities: reasoning about interaction feasibility in the environment and synthesizing realistic human-object motion. However, supervision for these capabilities is rarely available jointly at scale.
By Mingyuan Lei · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.