MVGD from Toyota Research Institute: Zero Shot 3D Scene Reconstruction

MVGD from Toyota Research Institute: Zero Shot 3D Scene Reconstruction

Toyota Research Institute Researchers have unveiled Multi-View Geometric Diffusion (MVGD), a groundbreaking diffusion-based architecture that directly synthesizes high-fidelity novel RGB and depth maps from sparse, posed images, bypassing the need for explicit 3D representations like NeRF or 3D Gaussian splats. This innovation promises to redefine the frontier of 3D synthesis by offering a streamlined, robust, and scalable solution for generatin…