Researchers have developed GroundControl, a novel diffusion model that leverages satellite imagery to improve novel-view synthesis from sparse street-level images. This approach addresses the tendency of existing multi-view diffusion models to hallucinate unrealistic spatial layouts when generating views outside the input range. By incorporating cross-attention conditioning from both satellite and street-level views, GroundControl generates seamless and realistic novel views without requiring explicit 3D scene reconstruction. Experiments show its superiority over baseline methods, particularly in challenging extrapolation scenarios. AI
IMPACT Enhances generative model capabilities for 3D scene reconstruction and novel-view synthesis, particularly for outdoor environments.
RANK_REASON The item describes a new research paper and model for novel-view synthesis. [lever_c_demoted from research: ic=1 ai=1.0]
Read on Hugging Face Daily Papers →
- 3D Scene Reconstruction Using Omnidirectional Vision and LiDAR: A Hybrid Approach
- GroundControl
- Hugging Face
- Multi-view diffusion maps
- Niantic
- satellite imagery
- street-level input images
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
