3D Shape Generation and Completion through Point-Voxel Diffusion
Authors:
Preprint
Abstract
We propose a novel approach for probabilistic generative modeling of 3D shapes. Unlike most existing models that learn to deterministically translate a latent vector to a shape, our model, Point-Voxel Diffusion (PVD), is a unified, probabilistic formulation for unconditional shape generation and conditional, multi-modal shape completion. PVD marries denoising diffusion models with the hybrid, point-voxel representation of 3D shapes. It can be viewed as a series of denoising steps, reversing the diffusion process from observed point cloud data to Gaussian noise, and is trained by optimizing a variational lower bound to the (conditional) likelihood function. Experiments demonstrate that PVD is capable of synthesizing high-fidelity shapes, completing partial point clouds, and generating multiple completion results from single-view depth scans of real objects.Keywords
All Related Versions
This publication has 21 references indexed in Scilit:
- Learning Gradient Fields for Shape Generation2020
- Discrete Point Flow Networks for Efficient Point Cloud Generation2020
- Multimodal Shape Completion via Conditional Generative Adversarial Networks2020
- Adversarial autoencoders for compact representations of 3D point cloudsComputer Vision and Image Understanding, 2020
- PartNet: A Large-Scale Benchmark for Fine-Grained and Hierarchical Part-Level 3D Object Understanding2019
- Multiresolution Tree Networks for 3D Point Cloud Processing2018
- PointGrid: A Deep Network for 3D Shape Understanding2018
- Learning Descriptor Networks for 3D Shape Synthesis and Analysis2018
- FoldingNet: Point Cloud Auto-Encoder via Deep Grid Deformation2018
- A Point Set Generation Network for 3D Object Reconstruction from a Single Image2017