SceneExpander enables text-guided 3D scene expansion with improved consistency

Researchers have developed SceneExpander, a novel method for text-guided 3D scene expansion through free-form view insertion. This approach addresses the challenge of maintaining multi-view consistency when generative models introduce misaligned views or hallucinated content. SceneExpander utilizes test-time adaptation with two distillation signals: anchor distillation to stabilize the scene using geometric cues and inserted-view self-distillation to accommodate misaligned views. Experiments on ETH scenes and other datasets show improved expansion quality and reconstruction accuracy. AI

IMPACT This research could enhance content creation tools for 3D environments and simulations by allowing more intuitive scene expansion.

RANK_REASON The cluster contains an academic paper detailing a new method for 3D scene expansion. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CV →

• arXiv
• ETH scenes
• Hugging Face
• Zijian He

• paper
• product

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