CVPR2024

SVDTree: Semantic Voxel Diffusion for Single Image Tree Reconstruction

Yuan Li, Zhihao Liu, Bedrich Benes, Xiaopeng Zhang, Jianwei Guo

Abstract

Figure 1 . SVDTree for single image tree reconstruction. Given a masked image, we use a diffusion model to automatically infer a semantic voxel structure of the tree, which guides a hybrid geometry reconstruction algorithm to produce a 3D tree with high visual fidelity.