Researchers have introduced RoGe, a novel framework for novel view synthesis that integrates reconstruction and generation into a single end-to-end process. Unlike previous methods that rely on intermediate rendered images or explicit 3D representations, RoGe uses a feed-forward reconstruction model to create an implicit scene representation. This representation's geometric features are then directly injected as conditioning into a video diffusion model, enabling the generation of temporally coherent videos from sparse input views and a camera trajectory. Experiments on the DL3DV dataset show RoGe outperforming existing reconstruction-based, generation-based, and hybrid baselines. AI
IMPACT This research could lead to more efficient and accurate methods for generating 3D scenes and videos from limited input data.
RANK_REASON This is a research paper detailing a new method for novel view synthesis. [lever_c_demoted from research: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
