ArtCrafter generates articulated 3D objects from images using Diffusion Transformer

Researchers have developed ArtCrafter, a novel feed-forward generative model capable of creating articulated 3D objects from just two images. This model uses a Diffusion Transformer architecture with a unique slot layout to jointly generate meshes for both rest and open articulation states in a single pass. ArtCrafter ensures geometric consistency and prevents part penetration through specialized attention mechanisms and a repulsion loss, analytically deriving URDF joint parameters for interpretable results. Experiments show ArtCrafter surpasses existing methods in geometry quality and articulation precision. AI

IMPACT This model could advance the creation of 3D assets for robotics and simulation by enabling more precise and consistent generation from simple image inputs.

RANK_REASON The cluster describes a novel generative model presented in an arXiv paper. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CV →

• ArtCrafter
• arXiv
• Diffusion Transformer
• Hugging Face
• Slot Attention
• Triplet Attention
• URDF

• paper
• model release

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