Researchers have developed JointHOI, a novel single-stage diffusion framework designed to generate realistic 3D hand-object interactions from text prompts. This method jointly generates hand and object motion alongside dynamic contact maps, enabling the model to learn the coupling between contact and motion during training. By using contact-guided sampling during inference, JointHOI enforces consistency between the generated contact maps and the implied geometry of the motion, leading to improved temporal stability and reduced artifacts like penetration and floating. Experiments on the GRAB and ARCTIC datasets show that JointHOI outperforms previous methods in both text adherence and physical plausibility. AI
IMPACT This research could advance robotics and immersive applications by enabling more physically plausible and text-driven generation of human-object interactions.
RANK_REASON The cluster contains a research paper detailing a new method for generating 3D hand-object interactions. [lever_c_demoted from research: ic=1 ai=1.0]
Read on Hugging Face Daily Papers →
AI-generated summary · Google Gemini · from 2 sources. How we write summaries →
