Description
Generating multiple views from one image, diffusion models often produce views that don't agree geometrically. GLD from KAIST repurposes geometric foundation models for multi-view diffusion, keeping views geometrically consistent.
Leveraging the geometry model's 3D understanding, the views are better suited to downstream reconstruction.
Consistency:Aligned views.
Multi-view:From one image.
Reconstruction-ready:Easier 3D.
Leveraging the geometry model's 3D understanding, the views are better suited to downstream reconstruction.
Features
Consistency:Aligned views.
Multi-view:From one image.
Reconstruction-ready:Easier 3D.
