New FAH models offer faster, more accurate 3D hand estimation

Researchers have developed FAH, a new family of hand estimation models that improve accuracy and speed by directly predicting 2D and 3D joints and surface points. These models are fine-tuned from SAM 3D Body's hand branch and can fit to MANO or MHR-Hand models, optionally estimating anatomical parts and contact. Experiments demonstrate significant improvements in alignment, geometry, contact estimation, and world-space tracking compared to existing methods. The default ViT-L model achieves 65 fps on an RTX 4070 Ti SUPER, with a smaller ViT-S model reaching 125 fps. AI

IMPACT Improves accuracy and speed for 3D hand tracking, potentially benefiting AR/VR and robotics applications.

RANK_REASON New research paper detailing a novel model for hand estimation. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CV →

• arXiv
• GeForce RTX 4070 Ti Super
• MANO
• MHR-Hand
• SAM 3D Body
• Vision Transformer Large

• paper
• model release

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