DiffusionGS: Baking Gaussian Splatting into Diffusion Denoiser for Fast and Scalable Single-stage Image-to-3D Generation and Reconstruction
Yuanhao Cai1, He Zhang2, Kai Zhang2,
Yixun Liang3, Mengwei Ren2, Fujun Luan2, Qing Liu2,
Soo Ye Kim2,
Jianming Zhang2, Zhifei Zhang2, Yuqian Zhou2,
Yulun Zhang4, Xiaokang Yang4,
Zhe Lin2, Alan Yuille1
1 Johns Hopkins University 2 Adobe Research 3 HKUST 4 Shanghai Jiao Tong University
A 3DGS-based diffusion generates objects and reconstructs scenes from a single view in 6 seconds.
