New method enables mask-free 3D object reconstruction for physics simulation

Researchers have developed a novel mask-free method for reconstructing complete 3D objects from sparse and occluded real-world views. This technique utilizes 3D Gaussian Splatting and a SAM2-trained segmentation field to achieve instance partitioning, followed by a joint Score Distillation Sampling process. This process integrates reference-view supervision with novel-view synthesis guided by diffusion priors to ensure texture fidelity and 3D consistency, enabling physically plausible interactions via Material Point Method simulation. AI

IMPACT Enables more realistic dynamic interactions in simulations by improving 3D object reconstruction from limited visual data.

RANK_REASON The item is an academic paper detailing a new method for 3D reconstruction. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CV →

• 3D Gaussian Splatting
• alphaXiv
• arXivLabs
• CatalyzeX Code Finder for Papers
• CORE Recommender
• DagsHub
• Gotit.pub
• Hugging Face
• Influence Flower
• Material Point Method
• SAM2
• ScienceCast
• Score Distillation Sampling

• paper
• other

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