New diffusion model estimates multi-person 3D body pose from egocentric views

Researchers have developed a novel diffusion-based approach for estimating the 3D body poses of multiple interacting individuals from their egocentric camera views. This method integrates data from egocentric cameras and IMU sensors, utilizing VIO SLAM for camera tracking. The system fuses pose estimates derived from head motion with sparse, intermittent, and variable-reliability exocentric observations of other people. Trained on both motion-capture and multi-person video data, the model learns priors for body motion and observation reliability, outperforming vision-only and motion-only baselines. AI

RANK_REASON The cluster contains a research paper detailing a new method for 3D body pose estimation. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CV →

• 3D Body Pose Estimation Using an Adaptive Person Model for Articulated ICP
• arXiv
• Charless C. Fowlkes
• diffusion-based approach
• egocentric views
• Motion capture data fitting system
• multi-person video
• VIO SLAM

• paper
• other

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