Description
3D generative models lack large training datasets and struggle from scratch. DiffSplat repurposes pretrained image diffusion models into 3D Gaussian splat generators, leveraging their vast 2D knowledge.
It supports text-to-3D and image-to-3D, fast and scalable; published at ICLR 2025.
Reused diffusion:2D knowledge.
Text to 3D:Describe it.
Image to 3D:One image.
Fast:Scalable.
It supports text-to-3D and image-to-3D, fast and scalable; published at ICLR 2025.
Features
Reused diffusion:2D knowledge.
Text to 3D:Describe it.
Image to 3D:One image.
Fast:Scalable.
