PoseRieFlow framework generates detailed 3D human poses from text

Researchers have developed PoseRieFlow, a novel framework for generating full-body 3D human poses from text descriptions. Unlike previous methods that often overlook hand articulation or model rotations in Euclidean space, PoseRieFlow operates directly on the complex manifold of SMPL-H joint rotations. This approach utilizes a Diffusion Transformer to process body parts efficiently, dedicating specific tokens to hands for improved detail and kinematic accuracy. The framework boasts a compact size with only 21.6 million parameters and achieves state-of-the-art results on the PoseScript benchmark for whole-body kinematic error, outperforming existing methods that generate hand poses. AI

IMPACT This new framework could advance applications in animation, virtual reality, and motion capture by enabling more accurate and detailed text-to-pose generation.

RANK_REASON The cluster contains a research paper detailing a new generative model for 3D human poses. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CV →

• arXiv
• Diffusion Transformer
• Hugging Face
• PoseRieFlow
• PoseScript
• rotation group SO(3)
• SMPL-H

• paper
• model release

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →