CVPR2023
Diffusion-SDF: Text-to-Shape via Voxelized Diffusion
Muheng Li, Yueqi Duan, Jie Zhou, Jiwen Lu
Abstract
Text-to-Shape Generation Text-Guided Shape Completion Text-Guided Shape Manipulation A chair with straight legs. A chair with curving legs. Incomplete Shape A straight chair. A sofa. A ball-shaped revolving chair. A sofa with curving legs. A two-layered triangular table .
An artistic chair. Figure 1 . We propose a text-to-shape synthesis approach named Diffusion-SDF. Our method is capable of performing various text-to-shape tasks including text-to-shape generation (directly generating 3D shapes from texts), text-guided shape completion (generating the missing part of an incomplete shape under the guidance of texts), and text-guided shape manipulation (modifying a given shape according to texts).