Seed2GS method extracts objects from 3D scenes without training or cameras

Researchers have developed Seed2GS, a novel method for extracting objects from 3D Gaussian Splatting scenes without requiring camera information or scene-specific training. This approach separates object identity from 3D coverage, using QD-SAM3 to select a reliable reference mask and then employing seed lift and visibility-adaptive virtual orbits to expose the object from new viewpoints. Seed2GS achieves high accuracy on benchmarks like LERF-MASK and 3D-OVS, demonstrating significantly faster processing times compared to existing methods. AI

IMPACT This method could streamline 3D editing workflows by enabling faster and more accessible object extraction from existing 3D scenes.

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

Read on arXiv cs.CV →

• 3D Gaussian Splatting
• 3D-OVS
• arXiv
• LERF-MASK
• QD-SAM3
• Seed2GS

• paper
• other

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