Researchers have developed Genie Sim PanoWorld, a novel pipeline for generating infinite indoor 3D worlds from a single panoramic image. This system bypasses the need for per-scene optimization or multi-view capture, addressing limitations in existing methods regarding metric trajectory control and handling large disocclusions. The pipeline uses a two-stage approach, first generating a trajectory-controllable panoramic video via a latent video diffusion model and then reconstructing this video into a high-fidelity 3D Gaussian scene. This output is directly usable for embodied AI applications, demonstrating superior performance in both video generation and downstream 3D reconstruction compared to geometry-conditioned baselines. AI
IMPACT Enables creation of large-scale, navigable 3D environments for embodied AI training and simulation.
RANK_REASON The cluster describes a new research paper detailing a novel pipeline for 3D world generation. [lever_c_demoted from research: ic=1 ai=1.0]
- arXiv
- Genie Sim PanoWorld
- Hugging Face
- SE(3)-Transformers: 3D Roto-Translation Equivariant Attention Networks
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
